Research paper · November 2026
Receipt Required
Retail media reports its own results. In the public record at 3 November 2026, 36 of the 37 retailer and marketplace services that document a default report attributed sales with no counterfactual, one of 22 documented lift offers discloses random assignment, and every MRC accreditation that maps to an eligible service covers delivery metrics, not outcomes.
Evgeny Popov · No Fluff Advisory
Research cutoff 3 November 2026 · Only information available by that date is used
Short on time? Read the plain-language version: the same findings in fewer, shorter words.
Abstract
Abstract
What retail and commerce media report as a result in 2026: the number a buyer gets by default, who computes and accredits it, whether the buyer can check it, how off-site loops close, what agent placements report and how big the market is. Research current to 3 November 2026, when the status-sensitive facts were read, with the United States as the main market.
Of the 37 retailer and marketplace services that say what their standard report delivers, 36 report attributed sales with no counterfactual . Of 22 lift offers beside the default, 21 do not say how groups were assigned . The seller computes the number on its own definitions, and at Amazon its count is also the bill . Of six major services, only Amazon lets the buyer query the records, and only aggregates leave . Every Media Rating Council (MRC) accreditation that maps to an eligible service covers delivery metrics; one outcome-type entry, named only "Epsilon", is tied to no service on current evidence, so H2 is unresolved . Three of 13 off-site services document that the ad is matched to the purchase outside the retailer's own surface, so H4b is contradicted . Inside AI agents, the only generally available sponsored formats come with the seller's own report . Each top row on the Receipt Quotient rests on one assurance instrument .
Treat a retail media sales figure as the seller's attribution until the seller shows otherwise. Ask what the report counts and over what window, how lift groups were assigned, who corrects the number, what you can query, what an accreditation covers, how off-site exposure met the purchase and what an agent placement reports: chapter 13 has seven questions. The paper audits public documentation; it does not test whose numbers are right.
The author works at Samba TV and co-leads the Signals & Measurement working group of AgenticAdvertising.org, which maintains AdCP. His past employers include Verve Group, Hearts & Science (an Omnicom agency; Omnicom owns Flywheel and Perpetua) and Lotame. Rows tied to those interests carry a mark wherever a marked finding is used; marked rows are excluded in a sensitivity run, and a mark never alters a score. No vendor funded, reviewed or saw the paper. Every service is scored under the same published rules. Chapter 14 has the full statement.
The question
A retailer can match an ad to its own purchase records, and by default it computes the result, usually attributed sales and a return on ad spend (ROAS), attributed sales divided by ad spend . This paper asks what a buyer can verify about that number: who produced it, what kind of number it is, whether the buyer can re-run it and whether anyone independent has attested it. The author's published positions, among them that every MRC accreditation in retail media covers delivery metrics , were registered as rival hypotheses, not evidence.
How it was done
15 hypotheses and questions, a product register and scoring rules with fixed anchors were frozen on 3 November 2026, before any claim was extracted or service scored. AI research agents ran 15 workstreams, 1 vendor-universe sweep and 8 deep profiles, and a separate AI verifier, told to refute, checked each workstream and profile file: of 1123 checks, 460 needed a correction and 29 found the item wrong. The ledger holds 1108 claims from 1058 source records; 737 were checked by a verifier, and findings no verifier reached are labelled unresolved. One scoring agent per slice scored 91 service rows on 12 dimensions in 5 cohorts, and only cells a separate cell verifier confirmed or qualified count. Coding rulings made after extraction are dated and logged; two of them moved H4b, which ends contradicted , and the corrections ledger holds 609 entries. This is AI-assisted research with a published process, not an audit. A separate AI review spot-checked selected claims against primary sources, tested the scoring rules and inspected the print output, and every point it raised was checked at its source before anything changed; no independent party audited the evidence or the scores, and no journal has reviewed the paper (chapter 14).
What it found
- No public number sizes US retail media on its own. The Interactive Advertising Bureau (IAB) and PwC's US$63.4 billion for 2025 counts commerce media on a survey definition with no placement split, and the same values were labelled "Retail media network advertising revenue" a year earlier . Company lines such as Amazon's global US$68,635 million cannot be added or ranked , and with no series split by placement, H4a is unresolved . Chapter 2.
- Ownership keeps moving. From 2023 to 3 November 2026, eleven changes of ownership completed, among them Circana's purchase of NCSolutions and Publicis Groupe's of LiveRamp on 2 November 2026 , and three more had no dated close, among them DoorDash's agreement of 28 October 2026 to acquire Instacart . Chapter 2.
- Attributed sales are the default. Of 49 retailer and marketplace services, 37 document a default and 36 of those report attribution with no counterfactual, as do all 20 networks whose representative service documents one, so H1a is supported on every reading tried . Chapter 3.
- Lift is an offer, and its groups are rarely shown to be random. Of 22 controlled-lift offers beside the defaults, 21 leave assignment unknown, and the one that discloses random assignment, Instacart's Sponsored Products holdout, does so on pages from 2022 and 2023; H1b carries no verdict . Chapter 3.
- The seller holds the pen. Wherever outcome reporting is documented, the operator that sold the media produces the number on its own definitions, with no documented separation of selling from measuring , and Amazon's agreement says "Fees are based solely on our measurements" . Only 7 of 69 retailer and commerce media rows could be scored on correction governance, none above 3 of 5 . Chapter 4.
- One network lets the buyer query; none hands over the events. Amazon Sponsored Products advertisers can query Amazon's impression and click records in Amazon Marketing Cloud and take out aggregates, so H3 is contradicted at its first step, though none of the six tested documents event-level exposure data for the buyer or an independent measurer . The conversions there are Amazon's own 14-day attributions, so the analysis supplies no counterfactual . Chapter 5.
- Most off-site loops do not say how they close. Of 13 retailer and marketplace services that tie outcome measurement to an off-site channel, 3 (23.1%) document that the join of ad to purchase runs outside the operator's logged-in surface, so H4b is contradicted; every stricter reading agrees, and only the loosest, at 8, would leave it unresolved . None of the three names its match keys , and Kroger's and Roundel's clean-room statements name no purchase outcome . Chapter 6.
- MRC accreditation stops at delivery metrics where it can be mapped. The seven eligible services with completed MRC accreditation hold it for delivery and invalid-traffic metrics . MRC's one outcome-type entry, "Correlated Outcomes", belongs to a service named only "Epsilon", which this review ties to a service outside the register only by inference from a 2021 release, so under its registered decision order H2 is unresolved; on that inference it would be supported . Kroger's outcome metrics are only in pre-audit , and the Alliance for Audited Media (AAM) certified DoorDash's incrementality measurement, which H2 does not test . Chapter 7.
- The retail media studies are the networks' own. No refereed study from 2024 to 2026 of a brand's incremental sales from a US retail media network with a published design was found , and in Albertsons Media Collective's study with Ovative, a campaign's highest and lowest iROAS (incremental return on ad spend) under different matching choices differed by 6.5 times on average . Of thirteen outcome figures shown to buyers, one states random assignment and none prints an interval . Chapter 8.
- The software underneath documents the same. Criteo Commerce Yield is the only enablement platform that states default attribution windows , and Moloco the only one that documents a randomised holdout for clients' campaigns, in an undated blog . No optimiser documents an assignment or counterfactual method for incrementality , and no outcome measurement provider documents random assignment for its standard design . Chapter 9.
- Privacy law offers opt-outs; the federal record is about prices. The US rules read govern purchase records in advertising mainly through opt-outs, and through consent, or in Maryland strict necessity, once a purchase reveals health , while the Federal Trade Commission's (FTC) orders of 28 October 2026 ask nine companies, Instacart, Target and Walmart among them, about personalised pricing . The searches logged found no order, settlement, court finding or breach notice about a network's use of shopper data for retail media . Chapter 10.
- Inside AI agents, the seller reports. Of eleven sponsored formats on US agent surfaces, the two stated generally available, Amazon's Sponsored Products and Sponsored Brands prompts, come with Amazon's own report of sales and ROAS, so H6a is supported . None of the eleven documents explicit general availability of third-party attribution, causal measurement or ad-quality verification, and most leave those functions undocumented rather than shown absent, so H6b to H6d are unresolved . Chapter 11.
- On the Receipt Quotient, the top rests on one cell each. Of 91 scored rows, 39 carry a band and 2 reach Most documented, DoorDash Ads' CPG (consumer packaged goods) Sponsored Products and Sponsored Brands on AAM's certification and Moloco Commerce Media on an IAB Europe certificate ; with 1s in their unscored cells, both would be Well documented, and without its assurance cell either would be a thin record. Of the 48 thin records, 44 have no counted cell on assurance, governance or privacy. Chapter 12.
What it means
On this paper's reading, retail media sells a number its seller computes, defines and corrects, and at several services bills on, with lift sold beside it, assignment rarely disclosed, and accreditation that stops at delivery. Chapter 13 sets out the questions a buyer can ask before the money moves.
Chapter 1
What is being sold
Retail and commerce media sell advertising that the seller can match to its own record of purchases, and by default the seller also computes the result: in documentation read by 3 November 2026, 36 of the 37 retailer and marketplace services that say what their standard report delivers report attribution with no counterfactual. This paper audits that number's receipt, meaning what a buyer can verify about it, one purchasable service at a time, because method, data access and accreditation differ between services of the same network. The register frozen on 3 November 2026 and amended once that day holds 132 records, 107 of them service rows in five cohorts.
The Interactive Advertising Bureau (IAB) and Media Rating Council (MRC) retail media guidelines of January 2024 call the common practice "event-level attribution (also sometimes referred to as closed-loop attribution)" . Of 49 retailer and marketplace service rows across 31 networks, 37 document a default outcome method, and 36 of those are attribution with no counterfactual . Of 22 controlled-lift offers documented beside those defaults, 21 leave the assignment of test and control groups unknown . On MRC's list read on 3 November 2026, the Amazon, Criteo, Instacart and Walmart Connect entries carry only delivery and invalid-traffic metrics, and the one outcome-type metric belongs to a service named only "Epsilon" . IAB/PwC's US$63.4 billion for 2025, the one US market series read in full, counts commerce media on a survey definition, partly from PwC's estimates, with no on-site, off-site or in-store split . TikTok Shop's help pages document attributed revenue with 7-day click and 1-day view windows, so its exclusion here is a choice of scope .
The receipt
A shop receipt lets a buyer check what was sold, by whom and at what price against what is in the bag. A retail media buyer is handed an outcome number instead, usually attributed sales and a return on ad spend (ROAS) . This paper uses "the receipt" for what a buyer can verify about that number: who produced it, what kind of number it is, what its documentation discloses, whether the buyer or its agent can re-run it, and whether anyone independent has attested it. It does not test whether any number is right.
The seller holds the pen. Amazon Ads computes and reports attributed sales, ROAS and its modelled long-term sales on its own metric definitions . In every commerce media service whose outcome reporting is documented, the operator that sold the media produces the default number, and named third parties appear only in optional studies or assessments the operator engaged or publishes . The seller's count can also be the invoice: Amazon's Advertising Agreement says "Fees are based solely on our measurements" , and at eBay the attributed sale is itself the billing event . IAB Europe's commerce media standards make whoever provides the metrics to the brand responsible for following them, and name IAB Europe's certification as the only way to guarantee it .
The industry's own texts treat attribution as the norm and lift as a separate exercise. The IAB/MRC guidelines add that the practice often involves extrapolation models . For incrementality they recommend randomised controlled trials when feasible and concede that holdouts are not always available or appropriate , and they recommend that retail media measurement eventually be accredited . On MRC's lists as screened on 3 November 2026, every accreditation that maps to an eligible service covers delivery metrics, but the one outcome-type entry, named only "Epsilon", is tied to a service outside the register only by inference, so H2 is unresolved (chapter 7) . IAB/PwC explain slower growth in 2025 partly by "brands challenging retailers to prove the value of incremental advertising spend amidst measurement challenges" .
The author's own standards page states that every MRC accreditation in retail media covers delivery metrics . That page and the author's essays on retail media measurement are treated as rival hypotheses, never as evidence (chapter 14).
What is sold, and where
A retail media network, in this paper, sells advertising as a retailer, or as a marketplace on which the advertised goods are sold, on inventory it controls or against its own shopper data. Its placements fall into three classes that are never pooled: on-site (its own sites and apps: sponsored products and search, display, video), off-site (third-party media bought with its data or audiences, including the open web, social platforms and connected TV, or CTV) and in-store (screens, audio, shelf and checkout media in its stores).
The same inventory can carry more than one label. IAB's DOOH (digital out-of-home) Measurement Guide of July 2025 files retail stores under "Retail and Venue-Based DOOH" . On their current pages the in-store sellers call the same inventory store ads or in-store media, but Walmart's release of 28 January 2021 described its in-store screens as making Walmart Connect one of the largest in-store and digital out-of-home ad platforms . The paper keeps every valid label, records which convention a source uses and never adds overlapping market totals.
Commerce media sells against transaction, loyalty, payment or travel data to brands whose goods are not sold on the operator's surface. The boundary is drawn per service and advertiser, so one operator can sit on both sides. Uber Eats sponsored listings sell restaurant ads to merchants on Uber Eats, while Uber Journey Ads sell in-ride display and video to brands . Walmart Connect sells its demand-side platform (DSP), Offsite Display and store placements to brands in automotive, financial services, insurance, quick-service restaurants and travel; its pages describe no outcome measurement for them, and its help centre names two approved attribution partners without stating parameters or terms . Inference: the "closed loop" Walmart documents ties exposure to purchases at Walmart, so it yields no sales outcome for a brand that does not sell there . A service sold to both kinds of advertiser without a documented split stays in cohort A, as DICK'S Media does .
The terms
- Attributed sales: an exposure or click matched to a later purchase in the operator's own transaction data within a stated window, with no counterfactual. Incremental lift: the difference against a counterfactual that was not exposed. The IAB and IAB Europe incrementality guidelines measure against "what would have occurred in the absence of marketing activity" and class platform-reported incrementality as a proxy without an independent control .
- ROAS is attributed sales over ad spend, as the operator defines both; incremental ROAS (iROAS) is incremental sales over ad spend. Neither implies profit, and an attributed figure is never called a return. New-to-brand (NTB) counts attributed purchases by shoppers new to the brand within the operator's lookback.
- "Closed loop" is the sellers' term for the attribution match, kept in quotes beside the coded fields; it proves neither causality nor assurance. Amazon's Alexa for Shopping page of 11 June 2026 promises "closed-loop measurement" with no holdout, control or lift wording .
- Assignment and counterfactual are separate fields. A holdout, matched group or test-and-control design is not randomised unless it discloses how units were allocated; ghost ads, matched markets and synthetic controls name a counterfactual, not an allocation.
- The reporting party is coded per claim: the operator or its enablement platform, a named third party computing on the operator's data on the operator's terms, or a party computing on data it controls. A third party is not thereby independent; funding, data provenance and analytical control decide.
A service's default is what its own documentation says the standard report delivers, either an attributed metric or a lift design. A "closed loop" claim that names no reported metric counts as a term only, and H1a's primary reading treats it as undocumented (chapter 3) . An offer is a lift study sold beside the default, usually on conditions. Roundel's incrementality "is available for eligible campaigns" . DoorDash's Sales Lift Measurement, based on ghost ads, is for select advertisers through its managed service . Kroger Precision Marketing's PrecisionView 360 is a paid, opt-in study for campaigns that may qualify, with at least US$400,000 across three or more channels . An offer never changes how a default is coded.
A revenue line is read by its perimeter, the business it covers: service, network, broader parent proxy, company-wide unallocated or third-party estimate. Amazon's US$68,635 million advertising services line for 2025 is global, includes advertising sold to publishers and authors, and is context only . Walmart's "nearly $6.4 billion, including VIZIO" for fiscal 2026 comes from earnings releases, is booked in net sales or as a reduction of cost of sales, and is a broader parent proxy on every Walmart Connect row . Uber Advertising's US$1.5 billion is a run rate for the first quarter of 2025 across food delivery and rides, and a run rate is never read as revenue . Expedia Group's US$758 million for 2025 is a network-level line with no endemic split . Series from different publishers are never spliced, and undisclosed service revenue is never prorated from a broader line (chapter 2).
Audience counts come in units that do not compare. Kroger's 10-K says it serves about 63 million households a year and that over 95% of customer transactions are tied to a loyalty card, a company-level share with no period or definition, while Kroger Precision Marketing cites 62+ million verified households . Albertsons Media Collective advertises 175M premium shoppers, and Albertsons' 10-K reports 51.2 million loyalty members . None of these is the share of an advertiser's sales that can be matched to an ad exposure, and none of the 20 commerce media rows discloses that share .
Five cohorts
Cohorts follow the buying task: who buys the service, and for what.
| Cohort | What is bought, and by whom | Rows |
|---|---|---|
| A · Retailer and marketplace media networks | Advertising for goods or services purchasable on the operator's own store, marketplace or booking surface; bought by brands and agencies | 49, across 31 networks |
| B · Commerce media networks | Non-endemic advertising sold against transaction, loyalty, payment or travel data; bought by brands and agencies | 20 |
| C · Enablement platforms | Ad serving, sponsored-product engines, ad platforms, portals and measurement modules that retailers license | 20 |
| D · Cross-network optimisation platforms | Software that plans, buys, bids and reports across several networks for brands and agencies | 11 |
| E · Outcome measurement providers | Measurement of retail or commerce media outcomes sold to brands, agencies or networks | 7 |
| Listed | A register record with no scored row: outside the cohorts' tasks, not shown to be purchasable, or excluded by scope | 22 |
| Historical | Retired, sold or no longer shown to operate | 3 |
Every one of cohort A's 31 networks has an on-site service . Amazon is a retailer and sits here with four rows, and Walmart Connect with five. Cohort B holds banks, card-linked offer networks, payment and reward apps, an airline, a hotel group, rideshare operators and the non-endemic offers of Walmart and Expedia Group, among them Chase Media Solutions, Cardlytics, MARRIOTT MEDIA and Kinective Media by United Airlines .
Cohort C is scored on the functions a platform documents as supported and configurable, never on what one client deployed. The same company can appear in three places: Criteo's retailer-side Commerce Yield and its supply-side Commerce Grid are cohort C rows, and its buyer-side Commerce Max is cohort D . Carrot Ads, which Instacart licenses to other retailers, is cohort C while Instacart's own ads are cohort A . Cohort D is not scored on identity basis, and its access dimension is read as the provenance of the network outcome feeds it reports. Cohort E is scored on its own measurement service, with independence coded claim by claim.
Some measurement services fit no scoring ladder. Store-level matched-market tests, marketing-mix and other aggregate models, and geographic experiments have no attribution window and no household panel, so NIQ Matched Market Testing, Ipsos MMA, Kantar LIFT ROI, Measured, Haus, INCRMNTAL, WorkMagic and LiftLab are kept as descriptive profiles and listed, not scored . Costco is listed because its own domain shows only an undated ad-specs page, and Affirm because its 10-K names merchant placements with no price, revenue line or outside advertisers .
Google, Meta and TikTok Shop advertising are an editorial exclusion: the paper covers operators whose transaction data arises from their own commerce and leaves out platforms whose primary business is media. Google's Commerce Media Suite documents named US retailers and product-level reporting inside Google's own buying tools, and no service that retailers license from Google . TikTok Shop's help pages document last-touch attribution of Shop revenue with 7-day click and 1-day view windows . The three historical records are NCSolutions, which Circana agreed to buy on 26 August 2024, Quotient Technology's platform and Shopkick .
One row per purchasable service
A row is a named service a buyer can purchase for a defined task and placement scope. Placements, report tiers, access tiers and configurations a buyer cannot buy separately are child records of that service. A second row exists only for a separately purchasable task whose default method, analytical access or assurance scope differs. Roundel is one row, because its pages name products without product-level defaults, access or assurance . Walmart's VIZIO placements are a child record of Walmart Connect Offsite Display .
Where the documentation differs, the rows split. DoorDash's restaurant ads are a separate row from its consumer goods Sponsored Products and Sponsored Brands, because the Alliance for Audited Media's certification and the managed sales-lift product cover only the consumer goods formats . eBay's four promoted-ad types are four rows because they credit and bill sales differently; the general strategy, for one, credits a sale after any buyer's click within 30 days and charges its fee on that sale, while priority campaigns charge per click . Ibotta's pay-per-sale offers and its fixed-price ad products are two rows with different revenue lines in its 10-K .
Renames are aliases on the same row. On 29 September 2026 Amazon renamed its unified platform Amazon Ads Agent, opened Full-Funnel Campaigns to all US advertisers and announced DVA+ from late October; both are campaign types with no separate default, access or assurance, so neither is a row . Amazon's page says Sponsored Display is now known as display ads .
Evidence, parameters and accreditation never transfer between rows. MRC accredited Walmart Connect's first-party measurement of Sponsored Search for clicks, click-through rate, impressions and invalid-traffic metrics on 26 September 2025 . Onsite Display is listed by MRC as "Audit in process as of 11/2025", and no Walmart DSP, Offsite Display, Store Ads or non-endemic entry appears on MRC's lists . Each Walmart row is read on its own record, and no network receives a pooled score. H1a also counts each network once, through one service chosen before extraction by a fixed priority, so splitting a network into rows cannot tilt the result .
Ownership events change rows only when they complete. DoorDash agreed on 28 October 2026 to acquire Instacart, with a close expected at the end of 2027 or in the first half of 2028, and until then the two operate as separate companies . Their rows, Carrot Ads included, stay apart .
The universe and the freeze
The register comes from one bounded sweep, workstream U, run on 3 November 2026 by six AI research agents and a browser. It opened 317 sources and screened every company and service named in the brief's seed list, an external reviewer's leads or its own discovery for status, ownership, cohort, row granularity, US availability and scale. It is a reasoned sample, not a census. Its table holds 146 records, among them separate scale and ownership records for services that also have their own line, and it coded 125 as live, 2 as beta, 1 as retired and 18 as unconfirmed.
Several unconfirmed records rest on pages the sweep could not read: Kroger Precision Marketing's two hosts refused the fetch tool, AD Retail Media's site served an empty shell, and Instacart's US product pages redirected to a European page . Later reads settled several. Kroger's onsite service is live, its self-serve platform open to all advertisers , and Albertsons announced off-site placements at its network's 2021 launch and has sold streaming and CTV under Collective TV since 13 June 2024 .
Vendor universe
A reasoned sample, not a census: how it was assembled and what was left out is stated in the method chapter.
Machine-readable copy: data/vendor-universe.csv.
| Company / product | Legal entity / owner | HQ | Scope group | Formats | Buying routes | Main geographies | Status at cutoff | Src |
|---|---|---|---|---|---|---|---|---|
| Amazon Sponsored Products | Amazon.com, Inc. (10-K registrant, RTL-U-S007) | Retailer and marketplace media networks (cohort A) | live | |||||
| Amazon Sponsored Brands | Amazon.com, Inc. (RTL-U-S007) | Retailer and marketplace media networks (cohort A) | live | |||||
| Amazon display ads | Amazon.com, Inc. (RTL-U-S007) | Retailer and marketplace media networks (cohort A) | live | |||||
| Amazon DSP | Amazon.com, Inc. (RTL-U-S007) | Retailer and marketplace media networks (cohort A) | live | |||||
| Walmart Connect Sponsored Search | Walmart Inc. (RTL-U-S017) | Retailer and marketplace media networks (cohort A) | live | |||||
| Walmart Connect Onsite Display | Walmart Inc. (RTL-U-S017) | Retailer and marketplace media networks (cohort A) | live | |||||
| Walmart DSP | Walmart Inc. (RTL-U-S017) | Retailer and marketplace media networks (cohort A) | live | |||||
| Walmart Connect Offsite Display | Walmart Inc. (RTL-U-S017) | Retailer and marketplace media networks (cohort A) | live | |||||
| Walmart Connect Store Ads | Walmart Inc. (RTL-U-S017) | Retailer and marketplace media networks (cohort A) | live | |||||
| Walmart Connect non-endemic advertising | Walmart Inc. (RTL-U-S017) | Commerce media networks (cohort B) | live | |||||
| Sam's Club Connect | Walmart Inc. (RTL-U-S019, RTL-U-S020) | Retailer and marketplace media networks (cohort A) | live | |||||
| Instacart Sponsored Products | Maplebear Inc.; DoorDash, Inc. agreement to acquire dated 28 October 2026, pending (RTL-U-S027) | Retailer and marketplace media networks (cohort A) | unconfirmed | |||||
| Instacart display and video ads | Maplebear Inc.; DoorDash agreement pending (RTL-U-S027) | Retailer and marketplace media networks (cohort A) | unconfirmed | |||||
| Roundel (Target) | Target Corporation (fact sheet on corporate.target.com, RTL-U-S029) | Retailer and marketplace media networks (cohort A) | live | |||||
| Kroger Precision Marketing | The Kroger Co. (seed; unverified here) | Retailer and marketplace media networks (cohort A) | unconfirmed | |||||
| Albertsons Media Collective onsite | Albertsons Companies, Inc. (RTL-U-S032) | Retailer and marketplace media networks (cohort A) | live | |||||
| Albertsons Media Collective offsite | Albertsons Companies, Inc. (RTL-U-S032) | Retailer and marketplace media networks (cohort A) | unconfirmed | |||||
| Albertsons Media Collective in-store | Albertsons Companies, Inc. (RTL-U-S033) | Retailer and marketplace media networks (cohort A) | live | |||||
| AD Retail Media | Koninklijke Ahold Delhaize N.V. (seed; unverified here) | Retailer and marketplace media networks (cohort A) | unconfirmed | |||||
| DoorDash Ads CPG Sponsored Products and Sponsored Brands | DoorDash, Inc.; agreement to acquire Maplebear Inc. dated 28 October 2026, pending (RTL-U-S027) | Retailer and marketplace media networks (cohort A) | live | |||||
| DoorDash Ads restaurant Sponsored Listings | DoorDash, Inc. (RTL-U-S027) | Retailer and marketplace media networks (cohort A) | live | |||||
| Uber Eats sponsored listings | Uber Technologies, Inc. (RTL-U-S042) | Retailer and marketplace media networks (cohort A) | live | |||||
| Uber Journey Ads | Uber Technologies, Inc. (RTL-U-S043) | Commerce media networks (cohort B) | live | |||||
| Chewy Ads | Chewy, Inc. (operator per its pages; ultimate control not settled by U2: no filing opened) | Retailer and marketplace media networks (cohort A) | live | |||||
| Best Buy Ads Sponsored Products | Best Buy Co., Inc. (seed; corporate newsroom RTL-U-S048 is the issuer's; no filing opened by U2) | Retailer and marketplace media networks (cohort A) | live | |||||
| Best Buy Ads display, video, search and CTV media | Best Buy Co., Inc. (seed; RTL-U-S048 is the issuer's newsroom; no filing opened by U2) | Retailer and marketplace media networks (cohort A) | live | |||||
| Orange Apron Media | The Home Depot, Inc. (issuer of RTL-U-S050) | Retailer and marketplace media networks (cohort A) | live | |||||
| Lowe's Media Network | Lowe's Companies, Inc. | Retailer and marketplace media networks (cohort A) | live | |||||
| CVS Media Exchange | CVS Health Corporation (per the copyright line; no filing opened by U2) | Retailer and marketplace media networks (cohort A) | live | |||||
| Walgreens Advertising Group | Walgreens Boots Alliance, Inc., since 28 August 2025 'a wholly owned subsidiary of Parent' (Blazing Star Parent, LLC; 'Parent and Merger Sub are affiliates of investment funds managed by Sycamore Partners Management, L.P.') (RTL-U-S061) | Retailer and marketplace media networks (cohort A) | live | |||||
| Macy's Media Network | Macy's, Inc. (seed; the August 2026 10-Q in the leads was not opened by U2) | Retailer and marketplace media networks (cohort A) | live | |||||
| UB Media | Ulta Beauty, Inc. | Retailer and marketplace media networks (cohort A) | live | |||||
| Kohl's Media Network | Kohl's Corporation (seed; unverified by U2) | Retailer and marketplace media networks (cohort A) | unconfirmed | |||||
| Costco retail media | Costco Wholesale Corporation | Outside the scored cohorts | unconfirmed | |||||
| DG Media Network | Dollar General Corporation | Retailer and marketplace media networks (cohort A) | live | |||||
| Wayfair advertising | Wayfair Inc. (seed; 'Wayfair Inc.' does not appear on the page; no filing opened by U2) | Retailer and marketplace media networks (cohort A) | live | |||||
| eBay Promoted Listings General | eBay Inc. | Retailer and marketplace media networks (cohort A) | live | |||||
| eBay Promoted Listings Priority | eBay Inc. | Retailer and marketplace media networks (cohort A) | live | |||||
| eBay Promoted Stores | eBay Inc. | Retailer and marketplace media networks (cohort A) | live | |||||
| eBay Promoted Offsite | eBay Inc. | Retailer and marketplace media networks (cohort A) | live | |||||
| Etsy Ads | Etsy, Inc. | Retailer and marketplace media networks (cohort A) | live | |||||
| Etsy Offsite Ads | Etsy, Inc. | Retailer and marketplace media networks (cohort A) | live | |||||
| Sephora media | LVMH (seed; the newsroom media kit in the leads was not opened by U2) | Outside the scored cohorts | unconfirmed | |||||
| Gopuff Ads | GoBrands, Inc. (private; controlling ownership not settled by U2) | Retailer and marketplace media networks (cohort A) | live | |||||
| Hy-Vee RedMedia | Hy-Vee, Inc. (employee-owned, RTL-U-S081) | Retailer and marketplace media networks (cohort A) | live | |||||
| Expedia Group Advertising sponsored listings (TravelAds and Flights Sponsored Listings) | Expedia Group, Inc. (issuer of RTL-U-S083) | Retailer and marketplace media networks (cohort A) | live | |||||
| Vrbo Sponsored Listings | Expedia Group, Inc. | Retailer and marketplace media networks (cohort A) | live | |||||
| Gulp Media Network | unverified: no parent company is named in the opened 7-Eleven release (RTL-U-S094) or on the site (RTL-U-S084); the seed's Seven & i Holdings Co., Ltd. was not confirmed by U2 | Retailer and marketplace media networks (cohort A) | live | |||||
| DICK'S Media | DICK'S Sporting Goods, Inc. (seed; no filing opened by U2) | Retailer and marketplace media networks (cohort A) | live | |||||
| Petco RMN | Petco Health and Wellness Company, Inc. (controlling ownership not settled by U2; no filing opened) | Retailer and marketplace media networks (cohort A) | live | |||||
| Staples Media Network | Staples, Inc.; the seed's Sycamore Partners control was not verified by U2 (no filing or release opened) | Retailer and marketplace media networks (cohort A) | live | |||||
| TikTok Shop Ads | ByteDance Ltd. (seed; not verified by U2) | Outside the scored cohorts | live | |||||
| Leap Media Group (Giant Eagle) | Giant Eagle, Inc. (private; control not settled by U2) | Retailer and marketplace media networks (cohort A) | unconfirmed | |||||
| Lyft Ads in-app advertising | Lyft, Inc. (issuer; its 10-K was not opened by U3) | Commerce media networks (cohort B) | live | |||||
| Chase Media Solutions (Chase Offers) | JPMorgan Chase & Co. (RTL-U-S105: "JPMorgan Chase & Co. (NYSE: JPM)"; RTL-U-S106 copyright "© 2026 JPMorgan Chase & Co.") | Commerce media networks (cohort B) | live | |||||
| PayPal Ads | PayPal Holdings, Inc. (issuer; not named on the opened pages; 10-K not opened by U3) | Commerce media networks (cohort B) | live | |||||
| Cardlytics card-linked offers | Cardlytics, Inc. (public company; CIK 1666071) | Commerce media networks (cohort B) | live | |||||
| Fetch Special Offers | Private (RTL-U-S117: "Fetch is a privately owned company founded in Madison, WI."); ultimate controlling owner unverified | Commerce media networks (cohort B) | live | |||||
| Ibotta Performance Network offers | Ibotta, Inc. (public company; 10-K RTL-U-S121) | Commerce media networks (cohort B) | live | |||||
| Ibotta D2C ad products | Ibotta, Inc. | Commerce media networks (cohort B) | live | |||||
| Ibotta white-label retailer enablement | Ibotta, Inc. | Outside the scored cohorts | live | |||||
| MARRIOTT MEDIA | Marriott International, Inc. | Commerce media networks (cohort B) | live | |||||
| Kinective Media by United Airlines | United Airlines Holdings, Inc. (RTL-U-S124) | Commerce media networks (cohort B) | live | |||||
| Expedia Group Advertising for non-endemic brands | Expedia Group, Inc. (RTL-U-S128 "Expedia Group, Inc. (NASDAQ: EXPE)"; RTL-U-S131) | Commerce media networks (cohort B) | live | |||||
| American Express Amex Offers and Amex Ads | American Express Company (issuer; not confirmed on an opened page; 10-K not opened by U3) | Commerce media networks (cohort B) | unconfirmed | |||||
| Capital One Ad Solutions | Capital One Financial Corporation (issuer; not confirmed on an opened page; 10-K not opened by U3) | Commerce media networks (cohort B) | unconfirmed | |||||
| Klarna Ads Manager | Klarna Group plc (SEC registrant, CIK 2003292; 20-F for fiscal 2025 filed 2026-02-26; RTL-U-S142, RTL-U-S144) | Commerce media networks (cohort B) | live | |||||
| Klarna Search & Compare sponsored placements | Klarna Group plc (RTL-U-S142, RTL-U-S144) | Commerce media networks (cohort B) | live | |||||
| Affirm marketplace placements | Affirm Holdings, Inc. (RTL-U-S137) | Outside the scored cohorts | unconfirmed | |||||
| Zillow Media Manager | Zillow Group, Inc. (issuer; not confirmed on an opened page; 10-K not opened by U3) | Commerce media networks (cohort B) | unconfirmed | |||||
| Rokt Ads | Rokt (private; ownership unverified) | Commerce media networks (cohort B) | live | |||||
| Upside merchant promotions | Upside Services, Inc (private; ownership unverified) | Commerce media networks (cohort B) | live | |||||
| Criteo Commerce Yield | Criteo S.A. (listed; per the brief's filing record) | Enablement platforms (cohort C) | live | |||||
| Criteo Commerce Grid | Criteo S.A. | Enablement platforms (cohort C) | live | |||||
| Epsilon Retail Media | Publicis Groupe S.A. (per the brief's record of Publicis releases of 15 July and 14 October 2021; not re-opened here) | Enablement platforms (cohort C) | live | |||||
| Amazon Retail Ad Service | Amazon.com, Inc. | Enablement platforms (cohort C) | beta | |||||
| Amazon Ads sponsored products across retailers (beta) | Amazon.com, Inc. | Outside the scored cohorts | beta | |||||
| Koddi Ads | Koddi Inc. (no ownership, investor or parent statement on S10) | Enablement platforms (cohort C) | live | |||||
| Topsort core platform (T-Platform) | Topsort Technologies, Inc. (private; no ownership statement on S12) | Enablement platforms (cohort C) | live | |||||
| Pentaleap | Operator-described as 'effectively an employee-owned company' (S15); cap table, control perimeter and spin-off completion date not stated on any opened page | Enablement platforms (cohort C) | live | |||||
| Moloco Commerce Media | Moloco, Inc. (private; no ownership statement on S16) | Enablement platforms (cohort C) | live | |||||
| Vantage retail media platform | Vantage (private; no ownership statement on S17) | Enablement platforms (cohort C) | live | |||||
| Kevel commerce media platform (Retail Media Cloud) | Kevel (private; no ownership statement) | Enablement platforms (cohort C) | live | |||||
| Mirakl Ads | Mirakl SAS (private; no ownership statement on S23) | Enablement platforms (cohort C) | live | |||||
| Carrot Ads | Maplebear Inc. (DoorDash acquisition agreement of 28 October 2026 pending; per the brief's 8-K record, not re-opened) | Enablement platforms (cohort C) | live | |||||
| NIQ Activate Retail Media Intelligence | NIQ Global Intelligence plc (per the seed; the IPO pricing release was not re-opened) | Enablement platforms (cohort C) | live | |||||
| Rippl | PAR Technology Corporation, by inference from the Bridg asset sale (S29, S30), if Rippl is among the Bridg assets; neither release names Rippl, so the inference is marked | Enablement platforms (cohort C) | live | |||||
| ADvendio retail media solution (AdOne, AdPortal, AdGateway, AdFinance) | ADvendio (private; unverified) | Enablement platforms (cohort C) | live | |||||
| Ibotta white-label 'rewards as a service' platform (publisher platform of the Ibotta Performance Network) | Ibotta, Inc. (listed) | Enablement platforms (cohort C) | live | |||||
| Microsoft Retail Media (PromoteIQ) | Microsoft Corporation | Outside the scored cohorts | unconfirmed | |||||
| Zitcha | Zitcha Pty Ltd (private; unverified) | Enablement platforms (cohort C) | live | |||||
| Osmos Retail Media OS | Osmos, Inc. (private; no ownership statement) | Enablement platforms (cohort C) | live | |||||
| Streaem | WPP plc via VML (unverified; the redirect is not an ownership statement) | Enablement platforms (cohort C) | unconfirmed | |||||
| LiveRamp Clean Room for Retailers | LiveRamp Holdings, Inc. (per seed) | Outside the scored cohorts | live | |||||
| The Trade Desk commerce and travel activation | The Trade Desk, Inc. | Outside the scored cohorts | live | |||||
| Yahoo DSP commerce | Yahoo Inc. (per seed; not verified here) | Outside the scored cohorts | live | |||||
| Google Commerce Media Suite | Alphabet Inc. | Outside the scored cohorts | live | |||||
| STRATACACHE in-store digital display technology | STRATACACHE (private; unverified) | Enablement platforms (cohort C) | live | |||||
| Swiftly retail media | Swiftly Systems, Inc. (private; unverified) | Enablement platforms (cohort C) | live | |||||
| Grocery TV | Grocery TV (private; unverified) | Outside the scored cohorts | live | |||||
| Skai (retail media platform) | Private; investor logos Sequoia, Tenaya Capital, Bain Capital Ventures, Arts Alliance, Qumra Capital (RTL-U-S194); controlling owner unverified | Cross-network optimisation platforms (cohort D) | live | |||||
| Pacvue | unverified; the 27 March 2023 Assembly-to-Pacvue rebrand is a lead not opened | Cross-network optimisation platforms (cohort D) | unconfirmed | |||||
| Criteo Commerce Max | Criteo S.A. (listed; not re-verified from a filing in this sweep) | Cross-network optimisation platforms (cohort D) | live | |||||
| Flywheel Commerce Cloud — Retail Media (Search, Display & Video, Measurement & Reporting) | Omnicom Group Inc. — 'today announced it has closed the previously announced acquisition of Flywheel Digital', 2 January 2024 (RTL-U-S200) | Cross-network optimisation platforms (cohort D) | live | |||||
| Perpetua | Omnicom Group Inc. (via Flywheel; Omnicom named in the privacy contact address, RTL-U-S202; Flywheel acquisition closed 2 January 2024, RTL-U-S200) | Cross-network optimisation platforms (cohort D) | live | |||||
| Quartile | unverified (no ownership or investors on RTL-U-S203) | Cross-network optimisation platforms (cohort D) | live | |||||
| Teikametrics ARI Platform | unverified (no ownership on RTL-U-S204) | Cross-network optimisation platforms (cohort D) | live | |||||
| CommerceIQ Retail Media Management | unverified (no ownership on RTL-U-S205) | Cross-network optimisation platforms (cohort D) | live | |||||
| Stackline Ad Manager | unverified (no ownership on RTL-U-S206) | Cross-network optimisation platforms (cohort D) | live | |||||
| Intentwise Optimize | Private; founders Sreenath Reddy and Raghu Kashyap ('Founded in 2016', RTL-U-S208); no investors or acquirer on the pages; controlling owner unverified | Cross-network optimisation platforms (cohort D) | live | |||||
| Xnurta | unverified (no ownership on the pages) | Cross-network optimisation platforms (cohort D) | live | |||||
| Circana Media Lift (household-level A/B testing) | Private; the IRI/NPD merger and the Circana rename dates in the seed were not re-opened here; controlling owner unverified | Outcome measurement providers (cohort E) | live | |||||
| Numerator sales lift studies | Operating business Numerator, 'a private company with private equity backing from Bain Capital'; 'not owned or operated by Kantar' (RTL-U-S214). The brief records the reporting-group level (The Kantar Group) and ultimate control (Bain Capital Europe Fund V, SCSp) from the group's investor page, which this sweep did not open; 'The Kantar Group' does not appear on RTL-U-S214. | Outcome measurement providers (cohort E) | live | |||||
| NIQ Matched Market Testing | NIQ Global Intelligence plc (seed; listed July 2025 per the brief; not re-verified from a filing in this sweep) | Outcome measurement providers (cohort E) | live | |||||
| Catalina HUB360 (measurement) | Infillion — acquisition 'today announced' 23 February 2026 by both parties (RTL-U-S218, RTL-U-S219); closing NOT stated on either release; prior owner not stated | Outcome measurement providers (cohort E) | live | |||||
| Ipsos MMA Marketing Mix Modeling / Unified Marketing Measurement (retail) | Ipsos: 'In 2018 Ipsos completed the acquisition of MMA from Hunting Ridge, LLC, forming Ipsos MMA.' (RTL-U-S221) | Outcome measurement providers (cohort E) | live | |||||
| Nielsen Sales Lift | unverified from the pages opened (Nielsen is private; the brief leaves the controlling owner to confirm) | Outcome measurement providers (cohort E) | live | |||||
| Kantar LIFT ROI | The Kantar Group (Bain Capital) per the seed; not re-verified here | Outcome measurement providers (cohort E) | live | |||||
| Attain In-Flight Measurement (retail) | unverified (no ownership on RTL-U-S226/RTL-U-S227/RTL-U-S228) | Outcome measurement providers (cohort E) | live | |||||
| Measured Incrementality Testing and calibrated MMM (omnichannel retail) | unverified (private per seed; no ownership on RTL-U-S229) | Outcome measurement providers (cohort E) | live | |||||
| Haus Experiments (GeoLift, Fixed Geo Tests, Time Tests) | unverified (private per seed) | Outcome measurement providers (cohort E) | live | |||||
| INCRMNTAL Explorer (continuous measurement) | unverified (private per seed) | Outcome measurement providers (cohort E) | live | |||||
| WorkMagic Geo Incrementality Testing (Lift Test) | unverified (private per seed) | Outcome measurement providers (cohort E) | live | |||||
| Affinity Solutions Consumer Purchase Lift | unverified (no ownership on RTL-U-S235/RTL-U-S236) | Outside the scored cohorts | live | |||||
| Incremental — Measurement & Reporting | unverified | Outcome measurement providers (cohort E) | live | |||||
| LiftLab Agile MMM and Incrementality Testing Suite (retail media) | unverified | Outcome measurement providers (cohort E) | live | |||||
| NCSolutions (Nielsen Catalina Solutions) | Circana: agreement 26 August 2024 (RTL-U-S241); completion per Circana's page title 'Circana Officially Acquires NCSolutions' dated 2 June 2025 in a search snippet (body unread, RTL-U-S242); domain ncsolutions.com 301-redirects to circana.com on 2026-11-03 (RTL-U-S238) | Outside the scored cohorts | retired | |||||
| Amazon Sponsored Products (scale record; applies equally to the seed rows Amazon Sponsored Brands, Amazon display ads and Amazon DSP) | Amazon.com, Inc. (registrant of the 10-K, RTL-U-S253, RTL-U-S007) | Retailer and marketplace media networks (cohort A) | live | |||||
| Walmart Connect Sponsored Search (scale record; applies equally to Walmart Connect Onsite Display, Walmart DSP, Walmart Connect Offsite Display, Walmart Connect Store Ads, Walmart Connect VIZIO placements and Walmart Connect non-endemic advertising) | Walmart Inc. (RTL-U-S017) | Retailer and marketplace media networks (cohort A) | live | |||||
| Instacart Sponsored Products (scale record; applies equally to Instacart display and video ads and, as parent context, to Carrot Ads) | Maplebear Inc. (RTL-U-S259) | Retailer and marketplace media networks (cohort A) | live | |||||
| Criteo Commerce Yield (scale context record; the cohort C unit is the count of named US retailer networks, which U6 did not screen) | Criteo S.A. (RTL-U-S260) | Enablement platforms (cohort C) | live | |||||
| Cardlytics platform (scale record) | Cardlytics, Inc. (RTL-U-S263) | Commerce media networks (cohort B) | live | |||||
| Shopkick | unverified (no readable operator or docket page) | Outside the scored cohorts | unconfirmed | |||||
| Quotient Technology platform | Neptune Retail Solutions since 5 September 2023 (RTL-U-S270); Neptune's own owner unverified (no ownership statement on RTL-U-S269 or RTL-U-S270) | Outside the scored cohorts | unconfirmed | |||||
| Catalina HUB360 (ownership record only) | Infillion (acquisition announced as made on 23 February 2026; no closing date stated) | Outcome measurement providers (cohort E) | live | |||||
| Numerator sales lift (ownership record only) | Reporting group: Kantar Global Holdings S.a r.l. ('reporting entity for Numerator | Kantar', RTL-U-S273); ultimate controlling party: Bain Capital Europe Fund V, SCSp (RTL-U-S274 p.15); division: Numerator, one of 'two separate divisions' of The Kantar Group (RTL-U-S272) | Outcome measurement providers (cohort E) | live | |||||
| Kroger Precision Marketing onsite search and display | The Kroger Co. | Retailer and marketplace media networks (cohort A) | live | |||||
| Kroger Precision Marketing offsite programmatic | The Kroger Co. | Retailer and marketplace media networks (cohort A) | live | |||||
| Staples Media Network | Staples, Inc. (owner beyond the issuer not shown on the page) | live | ||||||
| Pacvue | Pacvue (private; Assembly took the Pacvue name 27 March 2023 per the seed; controlling owner not on the page) | live | ||||||
| Kohl's Media Network | Kohl's Corporation | live | ||||||
| Amex Offers | American Express Company | Commerce media networks (cohort B) | live | |||||
| MARRIOTT MEDIA | Marriott International, Inc. | live | ||||||
| Kinective Media by United Airlines | United Airlines Holdings, Inc. | live | ||||||
| Zillow Elevate Media Manager | Zillow Group, Inc. | Commerce media networks (cohort B) | live | |||||
| Citi Commerce Media | Citigroup Inc. | Commerce media networks (cohort B) | live |
The register was frozen on 3 November 2026 with 131 records, before any service was scored; the hypotheses and scoring rules carry the same date. A later product enters only by a versioned amendment that states its effect on every denominator, and the frozen copy is kept for a sensitivity run. One amendment was made that day, before any score was seen: version 1.1 added NCS Sales Effect, taking cohort E from six rows to seven and changing no hypothesis population, because a trigger written into the frozen NCSolutions record was met. Circana's page lists NCS Sales Effect as a separately branded study service that measures incremental sales lift and ROAS, with its own contact; it states no price, contract or product code, so separate purchase is implied by the offer page, not documented in terms . Coding decisions taken after extraction began, none of which changed a decision threshold, are logged in chapter 14.
How each row is read
Each scored row is read on twelve dimensions in three groups. Depth asks what the number is: the default outcome method and controlled-lift availability, attribution definition disclosure, buyer-controlled outcome analysis, identity and data-basis disclosure, and validation. Reach asks where it works: placement measurement coverage, activation and integration routes, scale and US availability. Standing asks whether a buyer can trust and trade it: independent assurance, change and correction governance, and privacy and consent posture. Each cell runs from 0 to 5 against anchors fixed before scoring. Missing evidence is n/e, never zero, and only cells a verifier confirmed or qualified count. A row needs four counted cells to be admitted: 91 of the 107 rows were, 16 were not.
The Receipt Quotient (RQ) summarises a row's counted cells within its cohort and places it in a band against the cohort's field. It measures documented measurement capability, not measurement quality, effectiveness or business quality. Rows are never compared across cohorts. Chapter 12 sets out the quotient, and chapter 14 the method, the verification and the limits. The research was AI-assisted: agents researched each workstream and a separate agent verified each file.
The 15 hypotheses and questions registered on 3 November 2026 are read in chapter 2 (Q3, Q4), chapter 3 (H1a, H1b, Q1), chapter 4 (Q2), chapter 5 (H3), chapter 6 (H4a, H4b), chapter 7 (H2), chapter 8 (Q5) and chapter 11 (H6a to H6d). H2, H3 and H4b test the public record itself, so for them an absence in the screened documentation is the evidence; for the others an absence leaves the question unresolved. Every statement about a service describes its public documentation as read by 3 November 2026, and silence is coded unknown, never as the absence of a control.
Chapter 2
How big, by whose count, and who owns it
No public number measures US retail media on its own. The one US market series read, IAB/PwC's, puts commerce media advertising revenue at US$63.4 billion for 2025 on a survey definition that carried a "retail media network" label a year earlier, and the company lines beside it each cover a different perimeter, most of them global, mixed with other revenue, net or annualised, so none can be added to or ranked against another. From 1 January 2023 to 3 November 2026, eleven ownership deals closed and three more were agreed or announced with no dated close.
IAB/PwC reports US commerce media advertising revenue of US$63.4 billion for 2025, partly from PwC's estimates and with no placement split . The previous report labelled the same values "Retail media network advertising revenue", and neither explains the change . Amazon's US$68,635 million "Advertising services" line for 2025 is global and covers advertising sold to publishers and authors . Walmart's "nearly $6.4 billion, including VIZIO" comes from earnings releases and is booked partly as a reduction of cost of sales . Kroger and Albertsons name media revenue only inside mixed "Other" lines, and DoorDash's 10-K books its advertising inside Marketplaces revenue with no amount . No series read splits a US total by placement on one definition, so H4a is unresolved . DoorDash's purchase of Instacart and Citi's of Kard were agreed, with no close documented by 3 November 2026 .
Q4, on the denominators, and Q3, on ownership, are coding questions with no verdict. A forecast stays a forecast, a run rate is never revenue, and overlapping totals are never added, spliced or ranked (chapter 1).
One US series, two labels
IAB (the Interactive Advertising Bureau) and PwC's revenue report for 2025, published in April 2026, carries one line for this market: US commerce media advertising revenue of US$37.5 billion in 2022, US$43.7 billion in 2023, US$53.7 billion in 2024 and US$63.4 billion in 2025 . PwC compiles it from a survey, public disclosures and its own estimates for non-participants, without saying how much is estimated, and the report's 43 pages carry no on-site, off-site or in-store split . It explains slower growth in 2025, 18.0% after 23.0%, partly by "brands challenging retailers to prove the value of incremental advertising spend amidst measurement challenges" .
What the line counts depends on the page. The survey definition respondents report against centres on the retailer: "Commerce Media Network ads that allow a retailer to promote/ sponsor their own white-label products and/or the products they sell from their vendors directly on their own eCommerce site or on other sites" . The narrative says commerce media now spans on-site, off-site and in-store environments and entrants beyond retail . No sentence reconciles the two, and neither says whether banks, travel companies, store screens or connected TV are counted .
The label changed and the numbers did not. The report for 2024, published in April 2025, charted the same 2022-2024 values as "Retail media network advertising revenue", in a section already headed "Commerce media"; the 2025 report renamed the survey term from "Retail Media Network ads (Retail Media / Shopper)" to "Commerce Media Network ads", left the definition otherwise unchanged, and neither report explains why . The record fits a wider scope, as the narrative implies, or a renamed instrument, as the unrevised values suggest; without the publisher's word, the paper treats the series as one publisher's survey count and never as retail plus bank, travel and delivery commerce media .
There is no 2026 figure. The report's one forward statement, that commerce media spend will grow 30% faster than the overall advertising market, comes from IAB's own outlook survey and has no base that would turn it into a value .
The series that could not be read
Six other publishers were screened, and none yielded a US figure with its definition: eMarketer's series page returned an access error, WPP Media's and MAGNA's pages returned 404, dentsu's report sits behind a form, an Advertiser Perceptions page was not opened and no Coresight Research report was found; no gated form was submitted .
The two free eMarketer articles read concern other perimeters, financial media networks and Amazon's ad business, with forecasts on undisclosed definitions . eMarketer's US retail media figures, seen only in search listings and in two vintages for 2026, stay leads, as does a snippet suggesting that it counts in-store media in both retail media and out-of-home .
H4a needed a US series splitting on-site, off-site and in-store spending on one definition from 2023 to 2026; none screened discloses one, so H4a has no eligible series and is unresolved by its own rule (chapter 6) .
Where store screens are counted is a convention. IAB's DOOH (digital out-of-home) Measurement Guide of July 2025 files retail stores under "Retail and Venue-Based DOOH" . The sellers' current pages call the same inventory store ads or in-store media with no DOOH label, but Walmart's release of 28 January 2021 described its store screens as making Walmart Connect one of the largest in-store and digital out-of-home ad platforms, so one screen can sit inside a retail media total and a DOOH total at once .
What the company lines contain
The figure draws 15 company lines from 11 companies, each in the perimeter group the scoring rules assign. None measures a US retail media network's revenue.
Amazon's audited "Advertising services" line, US$68,635 million in 2025, is defined as sales of advertising services "to sellers, vendors, publishers, authors, and others, through programs such as sponsored ads, display, and video advertising", worldwide and with no split by product, placement or country . The nine months to September 2026 brought US$56,689 million, up 19.8%, with the third quarter alone up 10.9% . It bounds Amazon's retail media revenue from above without measuring it, and the paper uses it as context, never as the scale of the four Amazon rows . That this global line exceeds IAB/PwC's US total for 2025 shows how far the perimeters differ .
Walmart has no advertising revenue line; its earnings releases give a global advertising business of about US$3.4 billion in fiscal 2024, US$4.4 billion in fiscal 2025 and "nearly $6.4 billion, including VIZIO" in fiscal 2026, the year to 31 January 2026, each recorded in net sales or as a reduction to cost of sales depending on the arrangement . A full-text search of Walmart's filings with the Securities and Exchange Commission finds the phrase "global advertising business" in releases, annual reports to shareholders and proxies but in no 10-K or 10-Q; whether those reports state the amount was not read . Walmart Connect in the US appears only as growth, most recently 43% excluding VIZIO in the quarter to 31 July 2026, when international advertising, led by Flipkart, grew 20% . The same global figure is the scale record for every Walmart Connect row in the register, so it counts as one piece of evidence .
Instacart's audited "Advertising and other" revenue was US$871 million, US$958 million and US$1,065 million for 2023 to 2025: sponsored products, display, coupons and other online advertising, with third-party arrangements booked gross or net by contract, plus software subscriptions the note calls immaterial . The note does not name Carrot Ads, the platform Instacart licenses to other retailers, so its presence in the line is an inference, and the line is company-wide, not shown to be US-only . The earnings release of 29 October 2026 gives US$331 million for the third quarter, 3.1% of gross transaction value, and suspended guidance and withdrew long-term guidance because of the DoorDash agreement; the quarter's 10-Q had not been filed by 3 November .
Criteo sells technology to retailers and buying tools to brands, and its audited Retail Media segment reported revenue of US$263.9 million for 2025 and Contribution ex-TAC, revenue less traffic acquisition costs, of US$259.7 million, two different quantities . The segment spans Commerce Yield and Commerce Max and, by Criteo's account, which the profile's verifier could not re-read in the filing, books revenue mainly net, as agent . Segment revenue fell 17% in the fourth quarter of 2025 and 21% in the second quarter of 2026, which Criteo attributes to "previously communicated scope changes with two specific Retail Media clients" it does not name .
Cardlytics' US$233.3 million of 2025 revenue is company-wide, with 13.0% from the United Kingdom and US$20.9 million from Bridg, sold in March 2026, and it is net of consumer cash back: billings of US$385.0 million less US$151.7 million of incentives . Bank of America, one of its top three bank partners, sent a non-renewal notice in April 2025, and the relationship ended on 16 February 2026 . Continuing revenue fell 36.5% to US$36.9 million in the second quarter of 2026, when the 10-Q ties a fall in qualified users to an unnamed US partner's exit, Bank of America by inference from the dates . Nasdaq's bid-price notice of 3 June 2026 set a compliance date of 30 November; a 1-for-10 reverse split followed, and no filing to 3 November states that compliance was regained .
The other lines each carry a limit. Expedia Group's US$758 million of "EG Advertising" revenue for 2025 mixes advertising for bookable travel suppliers with advertising for outside brands and excludes trivago . Ibotta's US$297.2 million of redemption revenue is a fee per verified redemption, also its billing unit, beside US$45.2 million of advertising sold at fixed prices . eBay's release gives a quarter only, US$544 million of advertising revenue in the fourth quarter of 2025 . Uber gives annualised run rates across Eats and rides, most recently above US$1.5 billion for the first quarter of 2025 . DoorDash's letter and release for the fourth quarter of 2025 give no amount, and its 10-K for 2025 recognises advertising products as part of Marketplaces revenue, again with no amount .
Kroger and Albertsons put media inside residual lines. Kroger's 10-K presents "Other sales" of US$1,346 million for fiscal 2025, mainly food-plant sales, other pharmacy services, third-party media revenue and data analytic services, and its audited revenue note an "Other" category of US$1,129 million, mainly third-party media, data analytics, specialty pharmacy and in-store clinics; neither quantifies media, and the figure draws both as separate rows, never as one series . Kroger's US$1.5 billion of 2025 operating profit from alternative profit streams spans several businesses . Albertsons' "Other" revenue, US$1,305.6 million for the 53 weeks to 28 February 2026, consists primarily of wholesale sales, commissions, rental income, media advertising revenue and miscellaneous revenue, while some media is booked as a reduction in cost of sales, so the line holds other revenue and misses part of media . The advertising dollar figures in both 10-Ks, US$1,180 million and US$524.7 million, are the grocers' own advertising expense .
Measurers' revenue differs again: Kantar's Numerator division, which took in the former Worldpanel business during the first half of 2026, reported US$332.2 million of gross revenue for that half . No series sizes the enablement market, and Pacvue's claim to manage 12% of retail media spend worldwide outside China names no denominator .
Why the lines cannot be combined
The lines differ on every axis that would let them be combined. They are global (Amazon, Walmart, Criteo, Expedia, eBay), company-wide and not confined to the US (Instacart, Cardlytics) or a US survey (IAB/PwC) . They are net as agent, net of cash back, partly netted against cost of sales or annualised . Walmart's, Kroger's and Albertsons' fiscal years end in January or February, the others' in December .
They also overlap in ways no source measures. Spend bought through Criteo can appear in a retailer's revenue and in Criteo's; Instacart's line includes, by inference, what Carrot Ads earns from retailers with media lines of their own; and IAB/PwC's aggregate does not show which companies sit inside it . Subtracting Amazon and Walmart from IAB/PwC's total mixes global and US figures, adding company lines counts spend twice, and ranking networks by these lines ranks perimeters. The instrument lists every line and series point read with its unit, period, geography, audit status, contents and exclusions; no row with a value is marked combinable.
Market-size reconciliation
Which estimates include which formats, surfaces and markets, and which cannot be combined. Values are as published; no estimate here is the paper’s own.
Machine-readable copy: data/estimates.csv.
| Estimate | Published | Metric | Value | Year | Geography | Includes | Excludes | Combinable with | Src |
|---|---|---|---|---|---|---|---|---|---|
| Amazon.com, Inc. 10-K FY2025 | 2026-02-06 | Advertising services (net sales) | 46906 | 2023 | global | sponsored ads, display and video advertising sold to sellers, vendors, publishers, authors and others (10-K footnote 4) | subscription and other lines | no (company-wide, overlaps every network and market series) | |
| Amazon.com, Inc. 10-K FY2025 | 2026-02-06 | Advertising services (net sales) | 56214 | 2024 | global | sponsored ads, display and video advertising sold to sellers, vendors, publishers, authors and others (10-K footnote 4) | subscription and other lines | no | |
| Amazon.com, Inc. 10-K FY2025 | 2026-02-06 | Advertising services (net sales) | 68635 | 2025 | global | sponsored ads, display and video advertising sold to sellers, vendors, publishers, authors and others (10-K footnote 4) | subscription and other lines | no | |
| Amazon.com, Inc. 10-Q Q3 2026 | 2026-10-30 | Advertising services (net sales), quarter | 19637 | Q3 2026 | global | sponsored ads, display and video advertising sold to sellers, vendors, publishers, authors and others (10-K footnote 4) | no | ||
| Amazon.com, Inc. 10-Q Q3 2026 | 2026-10-30 | Advertising services (net sales), quarter | 17703 | Q3 2025 | global | sponsored ads, display and video advertising sold to sellers, vendors, publishers, authors and others (10-K footnote 4) | no | ||
| Amazon.com, Inc. 10-Q Q3 2026 | 2026-10-30 | Advertising services (net sales), nine months | 56689 | 9M 2026 | global | sponsored ads, display and video advertising sold to sellers, vendors, publishers, authors and others (10-K footnote 4) | no | ||
| Amazon.com, Inc. 10-Q Q3 2026 | 2026-10-30 | Advertising services (net sales), nine months | 47318 | 9M 2025 | global | sponsored ads, display and video advertising sold to sellers, vendors, publishers, authors and others (10-K footnote 4) | no | ||
| Walmart Inc. Q4 FY24 earnings release (8-K ex. 99.1) | 2024-02-20 | Global advertising business | 3400 | FY2024 | global | Walmart Connect US, Walmart International (Flipkart, Walmex), Sam's Club MAP | no | ||
| Walmart Inc. Q4 FY25 earnings release (8-K ex. 99.1) | 2025-02-20 | Global advertising business | 4400 | FY2025 | global | as above | no | ||
| Walmart Inc. Q4 FY26 earnings release (8-K ex. 99.1) | 2026-02-19 | Global advertising business, including VIZIO | 6400 | FY2026 | global | as above plus VIZIO (acquired December 2024) | no | ||
| Walmart Inc. earnings releases | 2024-02-20 / 2025-02-20 / 2026-02-19 | Walmart Connect in the U.S., fourth-quarter growth rate | 22% / 24% / 41% | Q4 FY2024 / Q4 FY2025 / Q4 FY2026 | US | Walmart Connect US | no | ||
| Maplebear Inc. 10-Q Q3 2026 | not filed by 2026-11-03 | Advertising and other (revenue), quarter | Q3 2026 | company-wide (not shown to be US-only; the 10-K mentions Canada) | no | ||||
| Criteo S.A. 10-K FY2025 | 2026-02-26 | Retail Media segment revenue | 209.007 | 2023 | global | Criteo's retailer-side (Commerce Yield) and buyer-side (Commerce Max, Commerce Grid) retail media business | Performance Media segment | no (enablement/buying revenue overlaps networks' revenue) | |
| Criteo S.A. 10-K FY2025 | 2026-02-26 | Retail Media segment revenue | 258.303 | 2024 | global | as above | Performance Media segment | no | |
| Criteo S.A. 10-K FY2025 | 2026-02-26 | Retail Media segment revenue | 263.872 | 2025 | global | as above | Performance Media segment | no | |
| Criteo S.A. 10-K FY2025 | 2026-02-26 | Retail Media Contribution ex-TAC | 203.46 | 2023 | global | revenue less traffic acquisition costs (non-GAAP) | no | ||
| Criteo S.A. 10-K FY2025 | 2026-02-26 | Retail Media Contribution ex-TAC | 253.846 | 2024 | global | as above | no | ||
| Criteo S.A. 10-K FY2025 | 2026-02-26 | Retail Media Contribution ex-TAC | 259.684 | 2025 | global | as above | no | ||
| Cardlytics, Inc. 10-K FY2025 | 2026-03-04 | Revenues (company-wide) | 309.204 | 2023 | US and UK | Cardlytics Direct (285.425) and Bridg Acquisition (23.779) segments | no | ||
| Cardlytics, Inc. 10-K FY2025 | 2026-03-04 | Revenues (company-wide) | 278.298 | 2024 | US and UK | Cardlytics Direct (255.615) and Bridg Acquisition (22.683) | no | ||
| Cardlytics, Inc. 10-K FY2025 | 2026-03-04 | Revenues (company-wide) | 233.273 | 2025 | US and UK | Cardlytics Direct (212.326) and Bridg Acquisition (20.947); United States 202.974, United Kingdom 30.299 (13.0%) | no | ||
| Expedia Group, Inc. 10-K FY2025 | 2026-02-13 | Expedia Group ("EG") Advertising (revenue by services) | 483 | 2023 | global | whole Expedia Group Advertising business (endemic travel suppliers and outside brands) | trivago Advertising (417 in 2025); 'Advertising, media and other' sales-channel line (1,294 in 2025) is broader | no | |
| Expedia Group, Inc. 10-K FY2025 | 2026-02-13 | Expedia Group ("EG") Advertising (revenue by services) | 639 | 2024 | global | as above | as above | no | |
| Expedia Group, Inc. 10-K FY2025 | 2026-02-13 | Expedia Group ("EG") Advertising (revenue by services) | 758 | 2025 | global | as above | as above | no | |
| Uber Advertising press page | 2024-03-22 item | Uber Advertising annual revenue run rate | 900 | Q4 2023 | global | Uber Eats and rides advertising (cohort A and B rows) | no | ||
| Uber Advertising press page | 2025-06-17 item | Uber Advertising annual revenue run rate | 1500 | Q1 2025 | global | as above | no | ||
| DoorDash, Inc. Q4 2025 shareholder letter (8-K ex. 99.2) | 2026-02-18 | advertising revenue | 2025 | global | n/a | ||||
| EMARKETER (free article, browser pane) | 2026-09-23 | Financial media network ad spending, forecast ('will exceed $3 billion by 2028') | 3000 | 2028 | unstated (US presumed, not verified) | financial media networks (eMarketer definition not disclosed on the free page) | no | ||
| EMARKETER (free article, browser pane) | 2026-09-23 | Financial media network ad spending CAGR 2023-2028 ('77.3%') versus travel media network segment ('21.2%') | 77.3% / 21.2% | 2023-2028 | unstated | as above | no | ||
| Kantar H1 2026 interim results | 2026-08-20 | Numerator divisional Gross Revenue (APM), six months | 332.2 | H1 2026 | global | Numerator CGU including the former Worldpanel business (combined 'during the period'; the report's 1 January 2026 date applies to the Insights-Profiles combination) | no | ||
| Kantar H1 2026 interim results | 2026-08-20 | Numerator divisional Gross Revenue (APM), six months (comparative) | 333.4 | H1 2025 | global | reclassified comparative on the 2026 divisional basis | no | ||
| Maplebear Inc. 10-K FY2025 | 2026-02-26 | Advertising and other (revenue) | 871 | 2023 | company-wide (not shown to be US-only; the 10-K mentions Canada) | Sponsored Product ads, display ads, coupons, other online advertising services; software subscription services (immaterial) | transaction revenue | no (spans cohort A and C rows) | |
| Maplebear Inc. 10-K FY2025 | 2026-02-26 | Advertising and other (revenue) | 958 | 2024 | company-wide (not shown to be US-only; the 10-K mentions Canada) | as above | transaction revenue | no | |
| Maplebear Inc. 10-K FY2025 | 2026-02-26 | Advertising and other (revenue) | 1065 | 2025 | company-wide (not shown to be US-only; the 10-K mentions Canada) | as above | transaction revenue | no | |
| eBay Inc. Q4 2025 earnings release (8-K ex. 99.1) | 2026-02-18 | Total advertising offerings revenue, quarter ('2.6% of GMV') | 544 | Q4 2025 | global | not defined in the release (Promoted Listings named only in forward-looking statements) | no | ||
| eBay Inc. Q4 2025 earnings release (8-K ex. 99.1) | 2026-02-18 | First-party advertising products revenue, quarter ('up 19% as reported, 17% FX-neutral') | 517 | Q4 2025 | global | not defined in the release (Promoted Listings named only in forward-looking statements) | third-party advertising | no | |
| IAB/PwC Internet Advertising Revenue Report FY2025 | 2026-04 (PDF created 2026-04-15; copy served by IAB Hong Kong) | Commerce media advertising revenue (survey) | 37500 | 2022 | US | 'Commerce Media Network ads that allow a retailer to promote/ sponsor their own white-label products and/or the products they sell from their vendors directly on their own eCommerce site or on other sites' (p. 32) | no (never splice with another publisher) | ||
| IAB/PwC Internet Advertising Revenue Report FY2025 | 2026-04 | Commerce media advertising revenue (survey) | 43700 | 2023 | US | as above | no | ||
| IAB/PwC Internet Advertising Revenue Report FY2025 | 2026-04 | Commerce media advertising revenue (survey) | 53700 | 2024 | US | as above | no | ||
| IAB/PwC Internet Advertising Revenue Report FY2025 | 2026-04 | Commerce media advertising revenue (survey) | 63400 | 2025 | US | as above | no | ||
| IAB/PwC Internet Advertising Revenue Report FY2024 | 2025-04 (PDF created 2025-04-09) | Retail media network advertising revenue (survey) | 37500 | 2022 | US | 'Retail Media Network ads (Retail Media / Shopper)' with the same definition sentence (p. 27) | no | ||
| IAB/PwC Internet Advertising Revenue Report FY2024 | 2025-04 | Retail media network advertising revenue (survey) | 43700 | 2023 | US | as above | no | ||
| IAB/PwC Internet Advertising Revenue Report FY2024 | 2025-04 | Retail media network advertising revenue (survey) | 53700 | 2024 | US | as above | no | ||
| Uber Advertising press page | 2022-02-09 item | Uber Advertising annual revenue run rate | 500 | 2022 item (period annualised not stated) | global | Uber Eats and rides advertising (cohort A and B rows) | no |
Who owns what: consolidation since 2023
Q3 follows shared rule 11: announcement, agreement, expected close and completion are separate events, each dated from the source that settles it, and a rename is an alias, never a retirement (chapter 14). No event is read as the failure of a format or metric . The figure plots 35 dated events from 1 January 2023 to 3 November 2026, two of them resting on reporting alone; after that date it shows two scheduled dates and one expected window.
Eleven changes of ownership completed: eight acquisitions, an initial public offering, a take-private and a divestiture. Neptune Retail Solutions completed its purchase of Quotient Technology on 5 September 2023 . Omnicom closed its purchase of Flywheel Digital, under which Perpetua sits, on 2 January 2024 . The author's past employer, Hearts & Science, was an Omnicom Media network, combined with Mediahub into Hearts United on 28 August 2026; Flywheel and Perpetua are sister Omnicom businesses, so this disclosure is voluntary .
VIZIO shows the events apart. It signed a merger agreement with Walmart on 19 February 2024 at US$11.50 a share in cash, Walmart announced the deal the next day as a way to accelerate Walmart Connect, and the purchase completed on 3 December 2024 for net cash of US$1.9 billion . Circana agreed on 26 August 2024 to buy NCSolutions, the Nielsen and Catalina joint venture, and Nielsen's marketing-mix modelling business, expecting to close in the fourth quarter of 2024; it completed them on 2 June and 21 August 2025 . NIQ Global Intelligence priced its initial public offering on 22 July 2025 and completed it on 24 July, with net proceeds of US$985.1 million; its first trading day is stated only as expected .
Walgreens Boots Alliance became a wholly owned subsidiary of a Sycamore Partners vehicle on 28 August 2025 . Cardlytics agreed on 23 January 2026 to sell Bridg to PAR Technology, announced it on 26 January and completed it on 24 March for 1,810,222 PAR shares; neither document names Rippl, whose page still carries a Bridg contact, so who runs it now is unsettled .
Three completions entered the record on 3 November 2026, while the final review's findings were checked: one the consolidation search had missed and two it had read as open. DoorDash's 10-K for 2025 records its purchase of Symbiosys, "a retail media platform company", on 28 May 2025 for US$121 million, to expand its off-site advertising . Infillion announced its Catalina purchase on 23 February 2026 with no closing date, and Catalina's Italian site confirmed on 26 February that "from today" Catalina had become part of Infillion. That is the Italian unit's confirmation date, not the group's legal closing date, which has not been published . Publicis Groupe agreed on 17 May 2026 to buy LiveRamp at an enterprise value of US$2.2 billion. On 19 October the FTC proposed a consent order under which Publicis could complete it with five years of non-discriminatory access and data firewalls, and Publicis announced the closing on 2 November . Walmart Connect and Roundel both describe measuring Meta campaigns through LiveRamp's clean room (chapter 6) .
Three deals were agreed or announced with no dated close by 3 November 2026. DoorDash and Maplebear, Instacart's parent, signed a merger agreement on 28 October 2026 at US$65.00 a share, an equity value of about US$17 billion, and expect to close at the end of 2027 or in the first half of 2028, subject to stockholder and antitrust approvals . The outside date is 12 months with up to four three-month antitrust extensions, and the termination fees are US$650 million from Instacart and US$1.25 billion from DoorDash . "Until the transaction closes, DoorDash and Instacart will continue to operate as separate and independent companies", so their rows stay apart .
Citi's U.S. Consumer Cards business announced an agreement to acquire Kard Financial on 13 August 2026, subject to customary closing conditions, and its launch release for Citi Commerce Media on 23 September still describes an agreement . Reporting that called Kard acquired in August overstated its status, and no completion record was found by 3 November .
Infillion's Foursquare release of 18 September 2026 says Foursquare's leaders will join after the transaction closes . Foursquare's home page carried an undated banner on 3 November saying Foursquare "has joined Infillion". The close may have happened, but no source dates it, so the deal is counted here, as one with no dated close, and not as completed .
One deal was blocked. The Kroger–Albertsons merger agreed on 13 October 2022 was enjoined by the US District Court for the District of Oregon on 10 December 2024, and each party then sent a termination notice, Kroger calling Albertsons' letter "not an effective termination" . Kroger still lists litigation over the terminated deal among its risks .
Five renames changed names, not rows. IRI and NPD, merged in August 2022, became Circana in a release dated 7 March 2023 on the site of Hellman & Friedman, the majority investor; Circana's own post could not be read . Assembly took the Pacvue name on 27 March 2023 , and Lowe's One Roof Media Network became Lowe's Media Network on 8 August 2024 . Walmart said on 22 June 2026 that Sam's Club Member Access Platform "will become" Sam's Club Connect, with no effective date . On 29 September 2026 Amazon named its unified platform Amazon Ads Agent, and Sponsored Display now sells as display ads .
Four exits are recorded, two on reporting alone. PromoteIQ's wind-down has no Microsoft notice behind it: Digiday reported it in October 2024, and Microsoft's own post of 11 July 2024 names Criteo as its planned preferred onsite partner without mentioning PromoteIQ . Shopkick's chapter 7 filing of 29 May 2026 rests on a law firm's bankruptcy digest; the docket was not opened . Cardlytics records its Dosh app as decommissioned on 28 February 2025 . On 25 October 2026 Maplebear's board decided to wind down Caper Cart operations, substantially by the end of the second quarter of 2027, at pre-tax charges of about US$55 million to US$80 million; the 8-K does not mention advertising, so the loss of the carts' Instacart display and Carrot Ads placements is an inference .
Five restructurings complete the record. Pentaleap, a self-described spin-off from Crealytics, was live by 3 September 2023, with no completion date stated . Kantar combined its Worldpanel and Numerator units during the first half of 2026, while Numerator's page says it is "not owned or operated by Kantar" and a Bain Capital fund controls the group: three corporate levels, not a separation . Walmart said on 22 June 2026 that Walmart Connect U.S., Walmart Connect International and Sam's Club Connect would operate separately . Criteo moved its domicile from France to Luxembourg on 29 July 2026 and agreed on 5 August to merge into a Delaware subsidiary on 1 January 2027, a step still awaiting an effective registration statement and a shareholder vote at the end of October .
What would settle it
Three disclosures would change this chapter: a US series splitting revenue by placement on one definition from 2023 to 2026, a revenue line scoped to one network in a retailer's filing, and IAB/PwC's reason for its new label. Walmart's and Criteo's third-quarter releases, Instacart's and Cardlytics' 10-Qs and the three open deals will be rechecked on 20 November 2026, the registered freeze date, and any change logged as a dated correction . A buyer quoted a market size can ask whose count it is, its perimeter and period, and whether it is gross or net, audited, estimated or annualised (chapter 13).
Chapter 3
The number a buyer gets by default
By default, a buyer of a US retailer or marketplace ad service receives attributed sales. On documentation read by 3 November 2026, 36 of the 37 services that say what their standard report delivers name attribution with no counterfactual, as do all 20 networks whose representative service documents a default, so H1a is supported. Controlled lift is offered beside the default at 22 services, but 21 of those offers do not say how shoppers, households or stores were allocated to test and control; the one that discloses random assignment, Instacart's Sponsored Products holdout, does so only on pages from 2022 and 2023.
Of 49 eligible retailer and marketplace services, 37 document a default and 36 of those report attribution with no counterfactual; the exception, Albertsons' in-store screens, reports lift against matched stores . At network level the count is 20 of 20, and the verdict holds under a lenient reading, three sensitivities and both disclosure gates . Amazon's Sponsored Products default counts sales 7 days after a click for sellers and 14 for vendors, and reports zero conversions from views . Walmart moved Sponsored Search to one deduplicated last-touch model on 1 June 2026 and did not restate earlier figures . Of 22 documented lift offers, two say a request is needed, two state an eligibility rule and one discloses random assignment . Of 129 outcome metrics the services name, 89 are attributed with no counterfactual; of 34 lift and brand-lift metrics, 2 disclose random assignment, both on Instacart's 2022 and 2023 pages .
What the three questions ask
H1a asks whether, among cohort A services whose default outcome method is documented, at least two-thirds report attributed sales as the default: an exposure or click matched to a later purchase, with no counterfactual . Cohort A is the retailer and marketplace services, 49 rows across 31 networks on the register frozen on 3 November 2026 . The verdict concerns documented defaults only, and it is unresolved if fewer than half of the eligible rows document one . Each network is also counted once, through one service chosen before extraction by a fixed priority, sponsored products first, so that splitting a network into rows cannot tilt the result .
H1b carries no verdict. For every cohort A row it records, as separate flags coded from that offer's own disclosure, whether a controlled-lift offer is documented, whether it must be requested, its stated eligibility, whether lift appears in the default report, how its control is built and assigned, and any minimum spend, sample or significance rule . Q1 asks what kind of number each reported metric is: its target contrast (attributed, incremental or descriptive), its unit, and how any comparison group was assigned.
The terms are those of chapter 1. A return on ad spend (ROAS) built on attributed sales is a ratio, not a return. A holdout, a matched group or a test-and-control design is not randomised unless the design says how units were allocated. "Closed loop" is the sellers' term for the match, and brand lift measures survey answers about a brand, not sales.
The coding rules, and when they were set
The registered rule says "documented" without saying what counts. Three coding decisions, taken on 3 November 2026 after extraction had begun, fill that gap; the thresholds did not change, and chapter 14 logs all three . They were needed because the first codes were uneven. The verifier of workstream E2 found Chewy Ads, which names lifetime-value ROAS and new-to-brand metrics, coded undocumented while DICK'S Media, with one "closed-loop measurement" sentence, was coded attribution; Leap Media Group, which names ROAS, was coded undocumented while Petco, whose dashboard shows ROAS and sales, was coded attribution .
The first decision defines a documented default: the row's own documentation must say what the standard report delivers, either an attributed metric (attributed sales, ROAS, conversions tied to ads) or a lift design . A bare closed-loop claim naming no reported metric is a term, not a method; the primary reading counts it as undocumented, and a lenient reading counts it as attribution .
The second, ruling 8 in the session's log, settles lift labels. A row whose documentation names an attributed metric in standard reporting and no campaign counterfactual for the default is coded attribution; a lift label with no design, such as incrementality or iROAS (incremental return on ad spend) named on a page, or a study on request, neither moves it out of attribution nor counts for anything by itself . Under it Orange Apron Media, Petco and UB Media count as attribution, while DICK'S Media stays term-only . The ruling was first applied to Chewy, Gulp Media Network and Leap as well. The final review of 3 November 2026 pointed out that their pages list attributed metrics among what they offer without naming a report or a default, so on the ruling's own words they are undocumented, as workstream E2 had coded them; their dimension-1 scores went to n/e with them .
The third, ruling 3, covers products sold side by side at different rungs with no default named: only the rung every product meets counts. Both of Kroger Precision Marketing's routes report attributed sales and only the managed route adds a matched comparison, so its rows are coded attribution . Walmart's help centre puts only the self-checkout formats of its store ads inside its attribution model, and one format cannot carry the row .
Before the decisions, the E2 verifier had found that on the strict reading only 10 of that stream's 21 networks documented a default, which would fail the gate had they stood alone; the gate applies to the pooled population, where it passes under both readings . By the registered rule, a lift offer counted under H1b never changes how H1a codes the row's default .
H1a: attribution is the default
Of the 49 eligible rows, 37 document a default under the primary reading (75.5%), so the disclosure gate passes, and 36 of the 37 (97.3%) report attribution with no counterfactual against a threshold of two-thirds: H1a is supported . At network level, 20 of 31 representative services document a default (64.5%) and all 20 are attribution . The lenient reading gives 41 of 42 services and 24 of 24 networks, with gates of 85.7% and 77.4% .
The one exception is Albertsons Media Collective in-store. Its release of 6 January 2026 compares test stores with "a rigorously selected set of control stores with no media exposure", matched on nearly 60 store-level variables . It does not say how stores were allocated, so assignment is unknown, and it calls the result statistically validated without naming a test or threshold .
Five rows are term-only. DICK'S Media, Lowe's Media Network, Walgreens Advertising Group and Kohl's Media Network assert a closed loop and name no metric, Kohl's in a release of 21 July 2022 . The fifth, Walmart's store ads, has a page that describes brand and sales lift with no design, and its 80% confidence footnote qualifies one proof-of-concept figure, not the reports . Seven rows document no default: AD Retail Media, whose launch release for its Edge platform names no reported metric, Wayfair, Gopuff and Staples , and Chewy, Gulp Media Network and Leap, whose pages list the metrics they offer without saying what the standard report delivers .
Three sensitivities test the closest calls. Taking the three rows coded at the rung that two routes or rails share (Kroger's two rows and Albertsons offsite) out of attribution gives 33 of 37 (89.2%). Treating the six rows that list or conditionally offer lift beside attribution (Walmart Connect Onsite Display, Kroger's two rows, Albertsons onsite and offsite, and DG Media Network) as not attribution gives 30 of 37 (81.1%). Coding Albertsons onsite as undocumented gives 35 of 36 . All three stay above two-thirds at both levels, as do their lenient versions .
Albertsons onsite is the closest single call, and the evidence files disagree on it. The verifier of Albertsons' profile kept its default undocumented, because the only statement of what onsite advertisers receive is that they will see matched reporting on display once programming reaches an undefined scale . The cell verifier coded it attribution: Albertsons' release of 11 December 2025 ranks its award winners on ROAS and guarantees each winner's ROAS for the first quarter of 2026, and its Sponsored Search case reports ROAS . The figure follows the cell verifier; the third sensitivity is the other reading.
Four rows carry the paper's interest mark (†): Walmart Connect Sponsored Search, the Walmart DSP (demand-side platform), Walmart Connect Offsite Display and Amazon DSP. The author works at Samba TV, which Walmart Connect names among its reach and frequency partners and lists among approved off-site measurement partners, and which Amazon lists among about sixty certified measurement technologies . The marks change no code.
The verdict is about pages: a row coded attribution may run lift studies on request, a delivered report may show more than its documentation, and several rows rest on one or two marketing pages .
What the default counts
The construct is shared; the parameters are not. The table sets out what eight services' own documentation says their standard report counts.
| Service | Default outcome | Window, credit and scope |
|---|---|---|
| Amazon Sponsored Products | Purchases, sales, units, ROAS | 7 days after a click for sellers, 14 for vendors; zero conversions from views; halo products inside base metrics |
| Walmart Connect Sponsored Search † | Attributed sales; default ROAS online only, Omni ROAS adds stores | 3, 14 or 30 days after a click, 14 by default; one deduplicated last-touch model since 1 June 2026 |
| Instacart Sponsored Products | Attributed sales, ROAS | 14 days after a click; credit split evenly across ad interactions, "primarily" |
| Kroger Precision Marketing onsite (self-serve) | Attributable sales | 14 days after a click, or a view for one format; last interaction; click over view; brand halo |
| DoorDash Ads, consumer packaged goods (CPG) | Ad-attributed sales, ROAS | First order within 14 days of a click; most recent ad interaction |
| Uber Eats sponsored listings | Attributed orders, sales, ROAS | 7 days, clicks and impressions as separate series |
| eBay Promoted Listings General | Attributed sale, which is also the fee | 30 days after any click on the ad, by any buyer |
| Etsy Ads | Attributed orders and sales, ROAS | Any item in the shop within 30 days of a click |
Credit rules differ: last touch at Amazon and Walmart, an even split at Instacart, the last interaction at Kroger, the most recent interaction at DoorDash . New-to-brand is not one metric either: its lookback is 12 months at Amazon, one year at Walmart, 26 weeks at Instacart and 180 days or one year at DoorDash, and Uber states none .
The match itself varies. eBay credits an attributed sale when any buyer purchases the promoted item within 30 days of any click, "regardless of whether the buyer themselves clicked on the ad", and charges its fee on that sale . eBay's own developer overview describes the trigger as the clicking buyer's purchase . Etsy credits any item bought in the shop within 30 days of a click and calls ROAS a measure of how profitable an ad is; this paper reads it as attributed revenue over ad cost .
What counts as a sale is rarely written down. Amazon's report columns do not say whether sales are gross or net of returns, cancellations, discounts, taxes or shipping . Since 1 June 2026 Walmart has removed online orders cancelled within 72 hours and does not deduct returns . Albertsons documents no value for any of the eight attribution parameters on any of its three rows .
The definitions also move. On 1 January 2026 Amazon put Sponsored Brands and display ads billed on viewable impressions, and Amazon DSP ads in its store, on a "shopping-signal enhanced last-touch attribution model" with a shorter window it does not quantify, keeping 14-day all-views figures beside it . Walmart's change of 1 June 2026 is not restated backwards, and Walmart advises against comparing later figures with earlier ones .
84.51°, which powers Kroger Precision Marketing, defines attributable ROAS (aROAS) as the targeted group's sales over campaign cost, and its managed-service page says the metric "doesn't factor in whether the ad truly affected customer behavior" .
Lift beside the default
Twenty-two cohort A rows document a controlled-lift offer, 23 document none, two name lift only as a label, one rests on trade press and one is unknown . Of the 22, two say a request is needed, one needs none and 19 do not say; two state an eligibility rule and nine name eligibility without a threshold . Lift is in the default report for one (Albertsons in-store), listed beside it for four, conditional on scale or qualification for four, outside it for five and unstated for eight . Fifteen name a comparison design; one discloses random assignment, none documents a non-random allocation, and 21 leave assignment unknown .
The random design is Instacart's, and it is dated. A post of 2 September 2022 launched a pilot using "randomized test and holdout groups", and the engineering post of 30 June 2023 randomises each user and advertiser pair into an ads group or a holdout, uses ghost ads to simulate the holdout's impressions and limits continual measurement to Sponsored Products . The 2023 post states no significance rule, minimum sample or interval, and a 2023 case page sends advertisers to their account team because lift testing may not be available to all partners . In 2026 the offer appears as a sales lift testing card with no design, format or eligibility, so whether today's tests use the 2023 design is not stated .
Kroger Precision Marketing sells lift in three forms, all with assignment unknown. Its managed-service campaigns report Sales Uplift against tightly matched control households by default . PrecisionView 360 is a paid, opt-in study for campaigns that may qualify with at least US$400,000 across three or more channels, built on "marketing mix / precision Test vs. Control modeling" . Off site, qualified self-service campaigns through The Trade Desk get incremental sales against a control audience built to mirror the exposed one, in the standard wrap report, subject to an unstated minimum spend .
Albertsons matches by default in stores and conditionally on site. Its vice president of media and measurement wrote on 23 March 2026 that it offers "a randomized testing methodology upon request"; no unit, holdout share, sample, significance rule or randomised result is published . The study of matching methods its staff co-wrote stays with that study and fills none of its defaults (chapter 8).
DoorDash names ghost ads, a technique for reading randomised experiments, without saying its own allocation is random. Its Sales Lift Measurement for consumer-goods Sponsored Products is "based on Ghost Ads", offered to select advertisers through its managed service, against a holdout whose assignment the release does not state . Its restaurant resource states random assignment for the intent-to-treat design that ghost ads replaced, while its ghost-ads section describes an 80/20 split without stating allocation, so assignment is coded unknown . The Alliance for Audited Media certified DoorDash's incrementality method for Sponsored Products and Sponsored Brands (chapter 7); lift remains an offer to select advertisers, not the default .
The other offers name a design, never an allocation. Walmart Connect Onsite Display reports lift against unexposed households of "similar digital & purchase propensity", with group sizes and a significance level but no threshold, minimum or eligibility rule . Roundel lists synthetic control, matched market and randomised trials for eligible campaigns, judged on objectives, investment level and measurement needs, without saying which campaign receives which . Best Buy Ads uses customer-level holdouts on site and geographic experiments across similar regions off site . Amazon's Brand Lift surveys an exposed audience against an unexposed control group for sponsored ads other than Sponsored Products and for display, video and audio campaigns, without disclosing assignment; no sales-lift offer for Sponsored Products was found on the pages screened .
Seven offers name no design. Walmart announced test-and-control incrementality for Sponsored Search on 28 April 2025 for a roll-out to managed advertisers, and no general-availability notice was found . Sam's Club Connect's cross-channel incrementality cites a 36% average from a Circana report on 1,440 campaigns . Uber's cross-channel incrementality, launched on 5 October 2026, pairs its test-and-control experiments with card-transaction data and states no eligibility, cost, control construction or sample . Orange Apron Media lists incrementality, geo lift and brand lift studies, DG Media Network lists Sales Lift and iROAS in its dashboard with no comparison, Gulp describes control and exposed reporting in the future tense, and Walmart's store ads describe sales lift with no design .
Minimum spends appear, but none of the sales-lift offers states a minimum sample or a significance threshold on the pages read by 3 November 2026 .
What kind of number
The 41 rows that name an outcome metric name 129 of them, counted once per row and kind of metric; 8 rows name none . Eighty-nine are attributed with no counterfactual: 32 attributed sales, 20 attributed orders or units, 24 ROAS and 13 new-to-brand counts . Twenty-three are incremental against a named comparison, 16 lift figures and 7 iROAS; 21 leave assignment unknown, and the 2 that disclose random assignment are Instacart's, from 2022 and 2023 . Five more name lift with no comparison, six are brand lift and six are modelled .
Seventeen rows report only attributed metrics and 24 report an incremental, lift or brand-lift metric; only Albertsons in-store reports lift with no attributed metric beside it . The unknown column holds every matched, holdout and ghost-ads design whose allocation is not disclosed, and the non-random column is empty because no row documents a non-random allocation . A buyer reading an iROAS usually cannot tell from the documentation whether its comparison group was drawn at random .
The modelled numbers carry the word incremental without a counterfactual. Amazon's long-term sales metrics "estimate the incremental sales value your brand can expect to generate" over 12 months, and are listed for Sponsored Brands, display, Sponsored TV and DSP reports . Walmart Ad Mix Modeling, which Walmart runs and calls a directional proxy estimate, and Kroger's panel-projected market impact are models too .
The six brand-lift metrics, at three Amazon rows, Walmart Onsite Display, Walmart's store ads and Orange Apron Media, measure answers about a brand, not purchases . Eight measurement records sit outside the grid because they are not report metrics, among them Amazon's internal calibration experiments, which nothing shows an advertiser receives, its multi-touch attribution beta and Uber's global holdout of 5% of winning ads .
What would change the reading
H1a would move only if many networks documented a lift design as what standard reports deliver, as Albertsons does for its stores; H1b would change if networks published allocation, sample and significance rules for their offers. A buyer can ask now for a report's kind of number, window, credit rule, sales basis and new-to-brand lookback, and how any lift study's control was allocated (chapter 13).
Chapter 4
Who holds the pen
In every retailer and commerce media service whose outcome reporting is documented, the default number a buyer receives is produced by the company that sold the media, on its own definitions, and at several of them the seller's count also sets the invoice. On the pages and terms screened by 3 November 2026, no service documents a separation between selling the media and measuring it, and the correction rules found are the seller's own: restatement schedules at Amazon, a no-restatement notice at Walmart, dispute windows of eight to 60 days, and clauses that make the seller's measurement the basis of the bill. Only 7 of 69 rows could be scored on change and correction governance, none above 3 of 5; the rest are unscored because the screen of their documentation and terms could not be completed.
Wherever a retailer or commerce media service documents its outcome reporting, the operator that sold the media produces the default number on its own definitions, and no row documents a separation between sales and measurement . Named third parties such as Nielsen, Circana and KPMG appear in studies the operator offers, engages or publishes . Amazon's Advertising Agreement bases fees "solely on our measurements" and waives fee claims not made within 60 days . Amazon restates conversions 1, 7 and 28 days after the event and gives three horizons for when its data settle . Walmart moved its media to one deduplicated last-touch model on 1 June 2026 and did not restate earlier figures . Albertsons', Macy's and Cardlytics' terms make the seller's own measurement the basis of the bill . Of 69 rows, 7 were scored on change and correction governance, Instacart's two highest at 3 of 5 .
The question and the coding
Q2 asks, for every cohort A (retailer and marketplace) and cohort B (commerce media) service on the register, who produces the outcome number a buyer receives, which metric definitions it follows, what restatement or correction policy exists, and whether a separation between sales and measurement is documented. It is a question, answered by records and counts, with silence coded unknown and never as an absence of controls (chapter 14). The reporting party is coded per claim, as defined in chapter 1: the operator or its enablement platform, a named third party computing on the operator's data on its terms, or a party computing on data it controls. A third party is not thereby independent; funding, data provenance and analytical control decide that.
Dimension 11 of the scoring rules, change and correction governance, looks for five items: metric versioning or change logs, a restatement or correction policy, documented reporting revisions, a challenge or dispute procedure, and methodology documents with version history. A row is scored only after a completed screen of its documentation and its terms; during scoring the session ruled that a screen which had not read the service's advertiser, insertion-order (IO) or platform terms was incomplete. An incomplete screen is n/e, and a score of 1 means none of the five turned up in a completed screen, a bounded absence and not proof that no controls exist. The final review asked two questions of the rule, and both were ruled on 3 November 2026 for every row. A dispute procedure for invoices counts as a challenge procedure, because the definition covers how a service governs its own numbers and the counts it bills on are among them. A schedule saying when numbers stop moving does not count as a documented revision of reported figures.
The seller computes the number
Amazon Ads reports attributed sales and return on ad spend (ROAS, attributed sales divided by ad spend) on its own definitions, including a 12-month new-to-brand rule . Roundel reports its own attributed outcomes and runs the incrementality designs for eligible campaigns on Target data . Kroger Precision Marketing (KPM), powered by Kroger's analytics subsidiary 84.51°, computes both the attributable and the incremental sales it reports . DoorDash and Uber compute their results on their own metric definitions . Licensed platforms, such as Koddi for Kohl's and Vantage for Orange Apron Media, are enablers and are not shown to validate an outcome number . In cohort B the operator produces the default number in every row whose outcome reporting is documented, the network operator in Rokt's case, and Lyft's only documented outcome measurement is a list of partner studies on terms not stated .
Third parties appear as studies the operator offers, engages or publishes, mostly on the operator's data. Cardlytics offers Nielsen Sales Lift Measurement as "an independent, third-party process", based on purchase data seen through Cardlytics' platform, without saying who pays . DoorDash and Circana announced on 2 September 2026 a partnership to bring third-party validation to DoorDash's measurement for select advertisers, and the headline result carries no design, panel, period or payer . Instacart's Circana lift figures come from a study Instacart says it collaborated on, with funding and control construction not stated . Fetch engaged KPMG, which Fetch says replicated the outcomes it had reported to partners, and Ibotta publishes Circana's meta-study of its campaigns . In each case the operator supplied the data, engaged the third party or published the result, and none states who paid, so none is coded independent.
On the pages screened, no cohort A or B row documents a separation between those who sell its media and those who measure it . The nearest statement is reported, not documented. Marketing Dive quoted the vice president and general manager of Lowe's Media Network on 13 October 2026, on a partnership with Ipsos for incrementality measurement: "I don't want an advertiser grading their own work" . No Lowe's source opened states the partnership, and who pays Ipsos is not stated .
The standards do not ask for separation. The Retail Media Measurement Guidelines that the Interactive Advertising Bureau (IAB) wrote with the Media Rating Council (MRC) apply to networks or other parties measuring retail media and rely on audit, validation and disclosure, with no passage on separating measurement from sales . IAB Europe's commerce media standards (body V2 of January 2026, cover V2.1 of May 2026) make whichever party provides the metrics to the brand responsible for adhering to them . A party study describes the practice: the Ovative paper written with Albertsons Media Collective, two of whose six authors work for the network, says the incrementality models of omni-channel networks are "typically done in-house by the RMN and aggregated results are provided to the advertiser", RMN meaning retail media network . That is the study's general characterisation, not a count of networks .
Definitions are the seller's own
Each network defines its outcome metrics, and the definitions differ in what they count. 84.51°'s glossary defines aROAS (attributable ROAS) as the targeted test group's total sales over total spend and iROAS (incremental ROAS) as test-group sales minus control-group sales over cost, and KPM's managed-service page says aROAS "doesn't factor in whether the ad truly affected customer behavior" . Walmart's default ROAS excludes in-store sales, and a separate Omni ROAS adds them . Cardlytics computes revenue and ROAS on purchases by users who activated an offer . Etsy describes ROAS, attributed revenue over ad cost, as a measure of how profitable an ad is ; under this paper's rules ROAS is never profit. eBay's own pages describe its attribution trigger in two ways .
Where a standard is named, the document usually is not. CVS Media Exchange (CMX) says it aligns its methodology with the IAB's measurement guidelines and DG Media Network calls its reporting IAB-aligned, neither naming a document . KPM says its off-site incremental methodology "adheres to IAB standards and best practices", a self-attestation that chapter 7 reads beside its MRC pre-audit . Roundel's and Walmart Connect's measurement pages do not use the words IAB, MRC, standard, glossary or definition , although Walmart's reporting API credits video view-through sales after an MRC-defined view of two seconds . No cohort B row cites the IAB/MRC definitions or IAB Europe's .
The seller's definitions also travel downstream. Instacart's Ads API terms require agencies that build reporting on its data to "Show ROAS Metric and associated calculation" and to give Instacart the list of metrics they display with their calculation methodology . Intentwise, an optimisation platform, shows Amazon's click and conversion validation as negative values in its hourly feed .
The seller's count is the bill
Several contracts make the seller's measurement the basis of the invoice. Amazon's Advertising Agreement, posted 23 July 2026, says "Fees are based solely on our measurements for the Ad Services and the applicable billing metrics (e.g., impressions or clicks)", and that the customer accepts the risk that third parties generate impressions or clicks for improper purposes . Albertsons Media Collective's standard terms make its ad server the measurement used for invoicing and the controlling measurement, in a section that applies where a third-party ad server is used, and let the terms of an IO prevail where the two conflict . Macy's Media Network (MMN) bases reporting and the advertiser's balance solely on its own ad server's measurements . Cardlytics' US terms for Cost per Served Sale make its data the definitive data used to calculate the fees, have the client pay fees above the IO budget, and let an IO prevail over the terms .
Klarna's Search & Compare terms, revised on 23 July 2025, separate the statistics from the bill. They state that "in no event ... will data provided by Klarna constitute final billing numbers"; only invoices are representative of billable amounts . In a dispute over amounts due, Klarna's tracking count applies, and an invoice not objected to in writing within eight days is deemed valid and payable .
For some services the attributed outcome is itself the billing event, so the party that writes the attribution rule also sets the fee base. eBay charges its general-strategy ad fee when any buyer purchases a promoted item within 30 days of any click on its ad, Etsy charges 15% (12% for larger shops) on orders it attributes to Offsite Ads, and Vrbo charges per night booked after a click . When eBay applied its any-buyer rule to its North American sites on 13 January 2026, reported sales and the fee base changed together . In cohort B, Chase charges brands when a customer views a campaign and makes a purchase , Ibotta charges per verified redemption , and Upside's fee is a share of incremental gross profit, net of promotional costs, measured by its own comparison with matched non-users .
On 31 August 2026 the Federal Trade Commission (FTC) and 22 states sued Amazon over alleged undisclosed reserve prices in its sponsored ads auctions . The complaint also alleges that Amazon's invoices are not itemised by click or other charge and present only averages . These are allegations in a pending case .
Restatement and revision
The IAB/MRC guidelines ask that "Known errors should be disclosed to measurement users as soon as is possible through client notification and quantification of impact or data restatement where applicable" . What the record holds is schedules on which numbers settle, dated changes of method and a few contractual correction rules .
Amazon publishes the fullest settlement schedule. Click data arrive within 12 hours and may change for up to three days through traffic validation; conversions arrive within 24 hours and are restated 1, 7 and 28 days after the conversion event, so a 14-day window can move up to 42 days after the report date . Its reporting overview says conversion data can change for up to 60 days and invalid-traffic detection runs for up to 30 days, and its Marketing Stream feed backfills up to 70 days as attribution validation completes . The overview ties click validation to charging, but neither page says whether fees are adjusted after the validation windows .
Amazon also dates its changes of method. Its view-attribution model of 1 January 2026, for Sponsored Brands, display ads billed on viewable impressions and Amazon demand-side platform (DSP) ads in its store, was published on a page dated that day, kept the 14-day all-views metrics as a parallel series and says nothing about restating history . Its newer Reporting API dates DSP conversions to the ad interaction rather than the purchase and adds halo conversions to base metrics, which "may return higher totals than in Version 3" . Version 3 reporting is to be shut off after 30 June 2027 .
Walmart Connect changed method on 1 June 2026. Sponsored Search moved from a last-click model not deduplicated with display to one deduplicated last-touch model shared with Onsite Display, the Walmart DSP and self-checkout store ads; display moved from fair-share credit over all delivered impressions to last touch over viewable impressions and clicks, with fair share fully deprecated from 1 September 2026; and online orders cancelled within 72 hours are now deducted, returns are not . Walmart's help centre says "Historical performance data will not be restated" under the new model and advises against comparing figures across the change . It also says attributed metrics keep updating to the end of the attribution window, so conversions can be added after a campaign ends . A methodology description for Sponsored Search delivery metrics was last updated on 9 September 2026, with no version history seen .
Instacart switched sponsored product impressions from served to viewable from 18 January 2022, applied the change to reporting from 1 January 2022, left earlier data unaltered and states that clicks, cost per click, attributed sales and ROAS were unaffected . Linear attribution applied to sales reporting from 1 August 2022 and clicks-only credit from 1 April 2025, and its Ads Manager changelog carries nine dated entries from 25 March to 29 October 2026 .
84.51° keeps a dated public release log for its Prism platform, 33 entries from January to September 2026, none about a restatement, a data correction or a change of attribution method . Its one written correction rule covers legacy Direct Connect campaigns, a topic marked for deprecation from 1 June 2026 and still served on 3 November: advertisers must notify KPM if data has not flowed into the portal "after five (5) business days", or a makegood or restatement will not be submitted for approval, and no deadline for that notice is stated .
In cohort B, numbers move without being called restatements. Cardlytics says purchase data can take 7 to 14 days to complete and that performance data update hourly after an initial four-day delay . Rokt says conversion counts by referral time may keep accumulating until the attribution window closes, while counts by conversion time do not change . Upside may update its billing methodology in its reasonable discretion .
Disputes
Amazon's Agreement routes fee disputes through the console's "Contact Us" feature and waives any claim about fees not made within 60 days after the date charged . It disclaims guarantees about the accuracy of ad services data, targeting and reporting, and sends US disputes under the Agreement to binding arbitration . Its definition of Performance Data includes whether a user viewed or clicked an ad, which it calls user-level data, while no route delivering such data for Sponsored Products is documented . No procedure for challenging an attributed outcome was found .
Instacart's ads terms, updated 6 May 2026, require an invoice dispute within 60 days of receiving the invoice and add: "Instacart shall not adjust invoices retroactively if Instacart, in its sole discretion, finds an inaccuracy in the billing amount of +/- 1% of the initial invoiced amount." . Read in full, the clause sets a threshold for small billing errors, and no dispute procedure for reported outcomes was found . Walmart's help centre lists billing-dispute and support-case routes; neither those pages nor Walmart's advertising terms were read .
Albertsons' standard terms oblige the network to cure an incomplete or inaccurate report within five business days of notice, and failure to cure "may result in nonpayment for all activity for which data is incomplete or missing until Media Company delivers reasonable evidence of performance" . The clause covers delivery reports and invoices, not outcome or lift figures, and an IO can override it . The published terms also carry an unfilled drafting placeholder under the heading Compliance with Law .
MMN's self-serve terms state that reporting from its ad server "will be final and not subject to challenge unless otherwise required by applicable law", and its make-good concerns product identifiers rejected through its own error . Cardlytics' terms contain no billing-dispute window or audit right, and their one dispute clause sets Georgia law and a Fulton County venue . On the pages screened, DoorDash, Uber Eats and Roundel document no dispute procedure .
What the governance screens found
Of the 69 cohort A and B rows on the register, 7 carry a score on change and correction governance and 62 are n/e: 44 of 49 in cohort A and 18 of 20 in cohort B (chapter 12). Instacart Sponsored Products and Instacart display and video ads score 3, the highest, for a changelog and API versioning, dated reporting revisions and an invoice dispute rule; a metric restatement policy and a versioned methodology were not found . Albertsons' onsite and off-site rows score 2 on the notice-and-cure clause . Klarna Search & Compare scores 2 on its dispute rule . Cardlytics and Macy's Media Network score 1: none of the five items turned up in a completed screen. Cardlytics' data delays say when its numbers settle, not how a reported figure is revised, and its cell scored 2 on them until the final review ; Macy's finality clause records the absence of a challenge route . No row reaches 4 or 5, and none scores 0.
Most rows are n/e because a screen could not be completed. Amazon's partial screen found four of the five items, but its API release notes and deprecations page did not load, its console help centre was unreadable and the Agreement's change log was not opened . Walmart's documentation yields change logs, a dated revision and support routes, but its advertising terms were not read . KPM's Kroger Ad Platform documentation sits behind a sign-in and its specifications and policies page refused the tools used . For DoorDash, Uber Eats, Roundel and Chase Media Solutions, help and marketing pages were read but no advertiser terms . A policy missing from the pages screened may sit in a console, an IO or a contract this review could not open.
The Amazon, Walmart and Instacart rows carry interest marks, which change no coding or score. The author works at Samba TV, which Amazon lists among its certified measurement technologies and Walmart Connect among its reach and frequency partners . Instacart names Omnicom's Flywheel as an agency partner, and the author's past employer, Hearts & Science, is an Omnicom agency; this is a voluntary disclosure .
What would change the reading
The reading would change if a network published a restatement policy for its outcome metrics with notice to buyers, a versioned methodology, a route to challenge an attributed figure or a documented separation of sales from measurement, or if the unread terms of the n/e rows were opened. A buyer can ask now who computed a figure, on whose definitions, when it stops moving, how a correction is notified and whether the invoice follows it. Chapter 13 turns these into questions.
Chapter 5
Can the buyer check it
Of 91 scored services, only Amazon's four document a way for the buyer to run its own outcome analysis under stated output rules: in Amazon Marketing Cloud, sponsored-ads advertisers write queries over Amazon's impression and click records and take out aggregated results. On documentation read by 3 November 2026, that one route contradicts H3 at the first step of its rule, but none of the six pre-registered on-site services documents event-level exposure data for the buyer or an independent measurer. Everywhere else the buyer receives the seller's aggregates, through dashboards, exports, APIs or a clean room the seller runs.
H3 is contradicted at the first step of its decision order: Amazon Sponsored Products documents buyer-controlled outcome analysis in Amazon Marketing Cloud (AMC), and none of the six pre-registered services documents event-level exposure data for the buyer or an independent measurer . AMC returns only aggregated, anonymous outputs, labels each column's aggregation threshold from NONE to VERY_HIGH with no number, and bars user, household, IP address and event identifiers from a query's final output . The conversions a buyer queries there are Amazon's own 14-day attributions, so the analysis re-cuts Amazon's attribution and supplies no counterfactual . Walmart gives Sponsored Search buyers its own aggregate reports and a mix-modelling export of weekly cost, impressions and clicks with no Walmart sales . Instacart's Data Hub clean room runs Instacart's templates on restricted listings, is in pilot and states no output rule . Kroger's and Albertsons' searches are incomplete, stopped by sign-in pages and an unread client hub . Verification tags, where allowed, count delivery: Walmart approves DoubleVerify and IAS tags on Onsite Display for viewability, and Amazon's verification policy covers its DSP, not sponsored ads .
The question and the ladder
H3 states that, for a pre-registered subset of major US networks, no on-site service documents (a) event-level exposure data available to the buyer or to an independent measurer, or (b) buyer-controlled outcome analysis of on-site exposures, meaning buyer-run queries in a clean room or analysis on permitted exports . The subset was frozen on 3 November 2026, before extraction, on scale and depth of public documentation: Amazon Sponsored Products, Walmart Connect Sponsored Search, Instacart Sponsored Products, Roundel (Target's network), Kroger Precision Marketing (KPM) onsite search and display, and Albertsons Media Collective onsite . Each axis is coded documented, not documented in the screened record, or source inaccessible. H3 is contradicted if any subset service documents either axis at its exact scope, whatever the other searches show; otherwise it is unresolved if any search is incomplete, and supported if every search is complete and neither axis is documented. Verification tags, API reporting and clean-room output rules are recorded beside the axes and decide nothing. Not documented means unknown, never that a buyer cannot obtain something.
Dimension 3 of the scoring rules asks every scored row a wider question: can a buyer, or an independent party acting for it, run its own outcome analysis (chapter 1)? Rung 1 is dashboards only; 2 adds a named third party's documented access to the operator's outcome data; 3 is aggregates through an API or clean room on the operator's queries; 4 is buyer-run analysis with output rules; 5 adds lineage and a named independent replication; 0 needs terms prohibiting any third-party analysis. For cross-network optimisers (cohort D) it reads the provenance of their outcome feeds.
Amazon contradicts H3, on one axis
Amazon Sponsored Products documents axis (b) at the service's exact scope, so H3 is contradicted at step 1 . On axis (a), no subset service documents event-level exposure data: four are not documented and two are source inaccessible, because the Kroger and Albertsons searches are incomplete. On axis (b), Amazon is documented, three services are not and the same two are inaccessible .
Amazon says sponsored-ads advertisers can reach AMC, a clean room built on AWS Clean Rooms, by self-service, at no cost to eligible advertisers, through a web interface and an API . Its sponsored-ads traffic table holds one record per Sponsored Products impression and one per click, and its conversion tables pair each conversion with the impression or click Amazon attributed it to, within 14 days . The documentation joins the two on an event identifier and publishes a worked Sponsored Products query that does so . The buyer writes its own SQL over Amazon's records of its own campaigns.
Output rules fix what leaves. Amazon's product page says "You can only access aggregated, anonymous outputs from AMC" . Each column carries an aggregation threshold labelled NONE, LOW, MEDIUM, HIGH or VERY_HIGH, and VERY_HIGH fields, among them the user, household, IP address and event identifiers, cannot appear in a query's final output, though they can be used in its intermediate steps . No numeric value stands behind any label on the pages read, and the page that should describe the thresholds rendered empty . These are still output rules, and the paper codes them so: Amazon states by class which fields may leave, not how many users or households a row must aggregate before it can.
The same pages set limits. Access is for eligible advertisers, a term they do not define; dataset availability varies by marketplace; and the traffic table mixes on-Amazon and off-Amazon placements, so on-site analysis needs a placement filter . The conversions carry Amazon's own attribution, and no buyer-built holdout is documented, so a query in AMC re-cuts Amazon's attribution and supplies no counterfactual . It shows where Amazon credited sales, not how many the ads caused.
Axis (a) is not documented for Sponsored Products. Queries run over event-level rows, but no event-level export is documented, and Amazon Marketing Stream, which delivers datasets to the buyer's own storage as often as every 15 minutes, carries aggregated performance data . Amazon's Ads API reporting pages for Sponsored Products were not readable to the tools used, which bounds the absence . The Advertising Agreement counts whether a user viewed or clicked an ad as performance data and limits its use to running campaigns, reporting to the advertiser and evaluating performance "only for its internal purposes"; no route that delivers such user-level data for Sponsored Products is documented on the pages read . Where AMC results go, and what the AMC agreement allows, belong to the series' data-collaboration paper, What Can Leave the Room? (cutoff 9 October 2026); this paper did not re-open those pages .
Amazon's other three rows reach rung 4 through the same clean room: Sponsored Brands and display ads sit in the same traffic table , and Amazon DSP advertisers request access or work through partners, and can also commission third-party measurement studies from more than 50 vendor products in a self-service workflow .
The other five services
| Service | Axis (a) | Axis (b) | What the pages read document | Beside the axes: verification tags |
|---|---|---|---|---|
| Amazon Sponsored Products † | Not documented | Documented | Buyer-written SQL in AMC over Sponsored Products impressions and clicks joined to Amazon-attributed conversions; aggregated outputs only | None on the Sponsored Products pages; Amazon's verification policy covers its DSP |
| Walmart Connect Sponsored Search † | Not documented | Not documented | Walmart-defined reports in Ad Center and the API; a mix-modelling export of weekly cost, impressions and clicks, no sales | No route; DoubleVerify and IAS approved for Onsite Display |
| Instacart Sponsored Products † | Not documented | Not documented | An agency API, and the Data Hub clean room in pilot with templates and restricted listings | DoubleVerify integration announced for video and display ads in 2023 |
| Roundel | Not documented | Not documented | Login-based reporting; secure data collaboration environments asserted without rules | Not documented |
| Kroger Precision Marketing onsite | Source inaccessible | Source inaccessible | Learning Center and API catalogue behind sign-in; dashboards and chart-level export on the pages read | Not documented |
| Albertsons Media Collective onsite | Source inaccessible | Source inaccessible | Client Hub unread; an API sends campaign performance data to the advertiser's models, fields unstated | By default, only its own tags on its properties |
The marked rows carry the paper's interest marks, which change no code. The author works at Samba TV, which Amazon lists among its certified measurement technologies and Walmart Connect among its reach and frequency partners . Instacart names Omnicom's Flywheel as an agency partner, and the author's past employer, Hearts & Science, is an Omnicom agency; this disclosure is voluntary . The verdict depends on a marked row. Set aside the three marked services, Amazon Sponsored Products, Walmart Connect Sponsored Search and Instacart Sponsored Products, and the counterexample goes with them; of the three services left, Kroger's and Albertsons' searches are incomplete, so H3 would read unresolved .
Walmart's coding moved during the work. Stream E1 coded both axes source inaccessible because Walmart's help centre served its fetch tool an empty shell; the verifier of Walmart's profile read it in a browser and kept the absence, so both axes code not documented . Buyers get downloadable Ad Center dashboards and, through API partners, snapshot reports kept for 62 days . The Marketing Mix Modeling Data Portal exports weekly, DMA-level cost, impressions and clicks for up to two years, for the advertiser's own mix model, with no Walmart sales . The paper codes it as a media-input export: a buyer can model its own sales against Walmart's delivery, but Walmart's outcome data do not leave. Read as analysis on permitted exports, it would make Walmart a second axis-(b) case; under stream E1's coding Walmart would be a third incomplete search; neither changes the verdict . Walmart also names measurement partners connected server to server by its own team, among them TransUnion, Upwave, Crossix and IQVIA, and a partner directory that includes Samba TV, without stating what data any partner receives or on what terms . A partner's name without its data terms earns nothing on the ladder.
Instacart's Data Hub is a Snowflake clean room in which partners set up clean rooms with SQL commands and run Instacart's pre-built templates; Instacart defines a clean room as "a data listing that features datasets and queries where partners are restricted on what they can run" . Access is not guaranteed, and no page states an output threshold, export rule or destination, or names a Sponsored Products exposure dataset . The launch release of 6 January 2026 describes a pilot with select agencies and consumer-goods brands, and its promise of flexible attribution approaches stays a candidate . Agency API terms require tools to show ROAS with its calculation, and the ads terms make reporting confidential and say nothing about third-party measurement .
Roundel documents login-based reporting and no API, and asserts secure data collaboration environments without output rules or buyer query rights . The clean room it is on record using, LiveRamp's for Meta campaigns, is measurement Roundel runs: its vice president of product management says it lets Roundel "combine Target’s first-party guest insights with Meta’s platform ads data in a privacy-safe way" .
KPM's help documentation sits behind a sign-in: the Kroger Ad Platform Learning Center redirects to a login, and the 84.51 API Catalog shows only a sign-in and a contact address . The site of 84.51°, the Kroger subsidiary that runs KPM's measurement, refused the fetch tool, so the search is coded incomplete; a sign-in gate is not evidence that an API or export is absent . 84.51°'s Collaborative Cloud gives consumer-goods data-science teams transaction-level Kroger data, but no page read places KPM campaign exposure or outcome data in it .
Albertsons Media Collective links a Client Hub it does not describe, and the hub could not be read . Its API, launched on 7 January 2025, sends campaign performance data to the advertiser's chosen destination for its own measurement models, with no fields, granularity or sales content stated; the release also says "there is diminishing ability to match impression data to sales data at an individual level at scale" . Its standard terms allow only its own tags on its properties, make its ad server the controlling measurement where a third-party ad server is used, and yield to a negotiated insertion order (IO) .
A clean room is a technology
"Clean room" names where data are joined, not who writes the query or what leaves. Among the clean rooms retailer networks name on the pages read, only Amazon's puts queries in the buyer's hands under stated output rules, which is why only Amazon's rows reach rung 4 . Walmart's article of 8 April 2026 says "Closed-loop measurement via LiveRamp's data clean room is now available for Meta", with no buyer query or output rule documented and an API for technology partners planned . Albertsons' multi-touch attribution, announced on 17 September 2026 as a launch with one pilot partner, joins impression-level exposure with Albertsons transactions in an environment LiveRamp powers, and describes no buyer query . Chapter 6 reads these joins for off-site media.
Other networks name a clean room and stop. Chewy names LiveRamp-powered clean-room solutions for optimisation and performance, Best Buy a LiveRamp clean room for activating audiences with it, and Orange Apron Media clean-room environments for deeper insights; none states who writes the queries or what comes out . Amazon Retail Ad Service measures inside AWS Clean Rooms "to generate aggregated, anonymized reports for the retailer and its advertisers" , and Circana's NCS Sales Effect delivers results in a dashboard or clean room . None of these pages puts the query in the buyer's hands. A restriction on what leaves is recorded as a restriction, never as proof that the buyer depends on the seller's analysis.
Dashboards, exports and APIs
Across the 91 scored rows, none reaches rung 5 or rung 0, and only Amazon's four reach rung 4 . In cohort A, 27 of the 47 rows document dashboard reports only, 13 deliver the seller's aggregates through an API or a clean room, and 3 are n/e. Of the 17 commerce media rows in cohort B, 10 sit at rung 1, Cardlytics and Rokt at rung 3, and 5 are n/e. Of the 15 enablement platforms in cohort C, 6 sit at rung 3, 4 at rung 1 and 5 are n/e; of the 7 measurement providers in cohort E, 3 sit at rung 3 and 4 at rung 1. Of the 16 eligible rows with too few counted cells to be admitted, 15 are n/e here.
A dashboard is not buyer-run analysis: the buyer reads the seller's numbers on the seller's definitions, cut as the interface allows. Roundel, KPM onsite, the Walmart DSP and Sam's Club Connect sit at rung 1 , the last despite its "Flexible data exports" . An export of the seller's aggregates lets a buyer re-read them, not rebuild them. Rung 3 delivers the same aggregates by machine: DoorDash lets consumer-goods brands pull up to a year of campaign performance through a direct API connection , eBay's Marketing API creates five kinds of Promoted Listings report , and Uber Eats offers report endpoints in its Ads API . None documents a buyer query, an output rule or an exposure log.
On the shared ladder no row sits at rung 2, the rung for a named third party with documented access to the operator's outcome data. Named partners exist: Walmart's lists, DoorDash's partnership with Circana, which both firms describe as independent validation for select advertisers, and Lowe's reported use of Ipsos (chapter 4) . None of the pages read says what data the partner receives or on what terms .
No row scores 0, because no terms read prohibit all third-party analysis; several restrict it. CVS Media Exchange's ad policy says the advertiser "agrees not to use any data obtained from CVS for measurement or analytics purposes" without its prior written consent . Macy's Media Network confines click and view data to the advertiser's own internal purposes and forbids passing them to ad networks, exchanges or data brokers . Cardlytics bars disclosure of its Insights Portal data to any consultant or third party without its written consent, limits transaction-level or individual activation data to internal verification and evaluation with deletion within 45 days, and lets an IO prevail . Each clause leaves consent, internal evaluation or contractors open.
Event-level data appear only in fragments. Outside the subset, stream E1 codes axis (a) documented for one row, Amazon DSP, in a narrow scope: third-party impression tracking on the Amazon Mobile Shopping site and app, by 1x1 pixel only . eBay reports per-sale attributed-sale rows but no exposure log , and Rokt exports event-level click data with no impression-level export . Cardlytics' hourly Redemption Feed carries one row per redeemed transaction and no exposures; inference: redemptions alone cannot support a counterfactual . Criteo's Commerce Yield, a retailer platform, shows brands offline sales only if the retailer shares them .
For optimisers the dimension reads the outcome feed. Criteo Commerce Max reaches rung 4 on that ladder by documenting origin, granularity, delays and limits for its one feed, Criteo's own retailer network, including that dashboard figures "can change and reflect new events, even if your campaign has stopped" . Intentwise and Perpetua sit at rung 2, and of the 47 network feeds the optimisers name, 32 document nothing beyond the network's name (chapter 9) . Perpetua sits under Omnicom's Flywheel, and the author's past employer, Hearts & Science, was an Omnicom agency; this disclosure is voluntary . Among measurement providers, Numerator sells data feeds for buyers' own models and Incremental delivers model outputs by API, neither with output rules or a named replication .
Tags count delivery
Verification tags verify what was delivered, not what was sold. Amazon's approved-providers page, last updated 1 June 2026, makes DoubleVerify, IAS and Adloox available on Amazon DSP and says nothing about sponsored ads on the retail site . Walmart's accreditation post names DoubleVerify and IAS as onsite partners without scope ; its ad policy approves their tags on Onsite Display as viewability trackers for monitoring only, on desktop and mobile web, with no tag route for Sponsored Search . DoubleVerify scoped its Instacart integration in October 2023 to viewability and invalid traffic on video and display ads , and in November 2024 said it had launched a platform-wide solution there, naming no formats . CVS Media Exchange approves 1x1 tags from CM360, Flashtalking, Innovid, IAS and DoubleVerify . Inference: a tag does not observe the purchase that attribution joins to an impression, so it can check what was delivered and billed, not the sales credited to it.
What would change the reading
One documented case decides H3, so the verdict would move only if Amazon withdrew AMC's self-service route for sponsored ads. Axis (a) would be documented if a subset network published an event-level exposure export or a log-level route for an independent measurer, and the Kroger and Albertsons searches would complete if their help pages and client hub were readable without a client login. AMC would reach rung 5 with documented lineage and a named independent replication of an analysis run inside it. A buyer can ask now which records it may query, which thresholds govern what leaves, whose attribution the conversions carry, whether a holdout can be built, and what any named partner receives. Chapter 13 turns these into questions.
Chapter 6
Off the retailer's site and in its stores
No US market series read by 3 November 2026 splits retail media into on-site, off-site and in-store spending on one definition from 2023 to 2026, so H4a, that the off-site share is rising, is unresolved. Of the 13 retailer and marketplace services whose documentation ties outcome measurement to an off-site channel, 3 (23.1%) document that the join between an exposure and a purchase runs through an identity match outside the retailer's own logged-in surface. That share is at or below one-third, so H4b, a proposition about the public record, is contradicted, and every stricter reading agrees; only the loosest reading, at 8, would leave it unresolved. In stores, no seller opened publishes the impression formula the in-store standard asks for, and most published lift figures leave out how test and control stores were chosen.
The one US series read in full, the Interactive Advertising Bureau (IAB) and PwC's commerce media revenue for 2022 to 2025, has no placement split, and the other forecasters' series were gated or returned errors, so H4a is unresolved . Of 28 cohort A services that name an off-site offer, 13 tie outcome measurement to an off-site channel . Three of the 13 document a join of ad to purchase outside their own surface, all through LiveRamp: Walmart Connect Offsite Display, Albertsons Media Collective and CVS Media Exchange . Kroger Precision Marketing and Roundel name a clean room for Meta campaigns but no purchase outcome . At 3 of 13 (23.1%), at or below one-third, H4b is contradicted, and every stricter reading gives 1 or 2 of 13 . A row coded not documented may still match identities in practice. In stores, Walmart projects its impressions and its 80% confidence figure qualifies one proof-of-concept result , and Albertsons matches test with control stores without saying how they were allocated . The same screen is retail media on the seller's page and digital out-of-home (DOOH) in IAB's guide .
What the two tests ask
Off-site media, in this paper's definitions, is third-party inventory on the open web, social platforms or connected TV (CTV) bought with a retailer's data or audiences; in-store media is screens, audio and shelf media inside its stores (chapter 1). H4a asks whether, in US series that split retail media into on-site, off-site and in-store on a constant definition across 2023 to 2026, the off-site share in 2026 exceeds the 2023 share. It is unresolved if no eligible series exists or the series disagree .
H4b asks how the off-site loop is closed. Among cohort A services that document off-site outcome measurement, it holds when at least two-thirds document that the exposure-to-transaction join relies on an identity match outside the operator's logged-in surface: a clean room, a partner graph or a device graph. It is contradicted at one-third or fewer, and unresolved between the thresholds, below four documented rows, or if any row's linkage search is incomplete . Each row is coded on five fields (exposure source, identity linkage, transaction source, modelled components, validation), and an operator's own exposure log never counts against an identity join . The author's essay of 22 August 2026 argued that the measurement seam returns once a retailer's audience is bought through a third-party demand-side platform (DSP) against an outcome on a third-party site ; that is the rival hypothesis, not evidence.
H4b is about documentation, not practice. A row coded "not documented" may still match identities when it measures; the code says only that its public pages, read by 3 November 2026, do not state how the join is made . Nothing about measurement quality follows from where a placement runs.
No series splits the market
Workstream U screened six candidate series before extraction and found none eligible, so the eligible set was frozen empty on 3 November 2026 and the planned off-site share figure is not drawn .
The only US series read in full is IAB/PwC's. Its report for 2025 (April 2026) gives "Commerce media advertising revenue" for 2022 to 2025 as one line . Its narrative says commerce media "now encompasses media powered by commerce data across on-site, off-site, and in-store environments", but a full-text screen of the 43-page PDF found no revenue split by placement . Its one forward statement, that commerce media will grow 30% faster than the advertising market on off-site and in-store opportunities, comes from IAB's own outlook survey and gives no share or base .
The other series could not be read. eMarketer's series page refused the fetch tool and its forecast tables were not read . WPP Media's pages returned errors . dentsu's report sits behind a form, MAGNA's page returned an error, no 2026 Advertiser Perceptions or Coresight report surfaced, and no gated form was submitted . Vendor series from Skai, Pacvue, Criteo and Mediaocean were not opened for a split . A gated forecast may hold such a split; one found after the freeze enters only through a versioned amendment . Unresolved does not mean a flat share. Chapter 2 reads the denominators.
Thirteen services document an off-site loop
Of the cohort A services on the eligible register, 28 name an off-site offer, and 13 tie outcome measurement to an off-site channel . They are the denominator. Nine more, among them Amazon DSP, Orange Apron Media (The Home Depot), DG Media Network and Lowe's Media Network, make only a network-wide statement not tied to an off-site channel . Five name an off-site service with no measurement documented , and AD Retail Media's off-site service could not be verified .
Three say how the join is made
Walmart Connect Offsite Display documents two outside joins. The row is the frozen register's, defined from Walmart's offsite page, whose channels include social media on Meta, TikTok and Pinterest; Walmart's 2026 article presents Social as its own tactic within that offer . Disney Advertising's release of 8 May 2024 says Walmart's measurement on Hulu and Disney+ is enabled by "a sync to Disney's Audience Graph via clean room technology", a proof of concept launching that quarter with select advertisers; no later page on it was opened . On 8 April 2026 Walmart wrote that closed-loop measurement through LiveRamp's data clean room was available for Meta, covering online and in-store sales, and was also available for TikTok and Pinterest with no mechanism named for those two .
Walmart's other rails state no mechanism: Walmart and VIZIO, which it bought in December 2024, "securely connect ad engagement across CTV to purchases" . Its social results, a 2.5 times stronger incremental return on ad spend (iROAS) and 1.8 times higher sales lift than social benchmarks, come from a Circana benchmark analysis of campaigns that qualified for a sales-lift study in four categories, with no design, sample or interval; who paid for it is not stated .
Albertsons Media Collective's credit rests on its multi-touch attribution, announced on 17 September 2026 as a launch built with LiveRamp, with PepsiCo as an early pilot partner . It joins impression-level exposure from Albertsons' own channels, off-site media and partners including Meta, Display & Video 360, Google Ad Manager and Perion "with Albertsons transaction data in a neutral, privacy-conscious environment" . The release never says clean room or identity and states no method, window or general availability; "neutral" is the word of the operator and of LiveRamp, which sells the environment .
CVS Media Exchange's credit rests on its announcement, reprinted on 29 February 2024, that pseudonymised ExtraCare loyalty data and Pinterest platform data meet in a LiveRamp-provided clean room that reveals their overlap, so that it can measure attributed Pinterest sales and report them to suppliers; Pinterest's programme was then a beta, and the keys are not named . Its current Social page still offers attributed sales on Pinterest but ties the clean room to "a data-safe targeted approach", not to the measurement .
Two name a clean room but no sale
Kroger and Roundel were credited on 3 November 2026 and lost the credit the same day, after the final review; the coding section below explains why. Both name a clean room for Meta campaigns; neither says the purchase is what it joins.
Kroger Precision Marketing (KPM) buys off-site media through The Trade Desk, Yahoo DSP and Google's Display & Video 360 (DV360), and runs Meta campaigns through its managed service and, since the third quarter of 2025, through Direct Connect in its ad platform . Its strongest statement is 84.51°'s roadmap, last updated 9 September 2026, which marks Meta Advanced Analytics as completed in the first quarter of 2025 and says it lets KPM "measure the impact of media on Meta in a secure clean room environment" . The entry names no purchase outcome, no clean-room operator and no keys, and the roadmap disclaims any guarantee that listed functionality is delivered, though this item is marked completed . Its other statements are weaker: its FAQ cites hashed identifiers and clean rooms where appropriate at network level, and its DV360 measurement runs on a conversions feed with no stated key .
Roundel's own pages describe secure data collaboration environments at network level and closed-loop sales reporting for its programmatic and CTV rails, naming no partner, window or control . Its strongest statement is in a LiveRamp release of 23 October 2025, in which Roundel's vice president of product management says clean-room measurement of Meta ads lets it "combine Target’s first-party guest insights with Meta’s platform ads data" . His words name guest insights, not sales; the sales and ROAS in the release are LiveRamp's description of its product, and Albertsons' executive, quoted in the same release, does not name a clean room .
All three credited rows name LiveRamp, as does Roundel's statement; Kroger's entry names no operator . On 19 October 2026 the Federal Trade Commission (FTC) proposed a consent order under which Publicis Groupe, an agency group, could complete its US$2.2 billion acquisition of LiveRamp with access and data-firewall conditions , and Publicis announced the closing on 2 November .
Eight do not say
Instacart's release of 10 June 2025 says agency partners such as Omnicom's Flywheel can access closed-loop measurement inside The Trade Desk, including attributed sales and return on ad spend (ROAS), built on "Instacart's deterministic purchase-based data"; Flywheel's quote names a conversion route between the two, with no matching key or match location . Flywheel is an Omnicom business, and the author's past employer, Hearts & Science, is an Omnicom agency; this is a voluntary disclosure. Instacart's off-platform page (captured 1 March 2026) offers closed-loop measurement with NBCUniversal, Pinterest, Roku and TikTok and describes no join ; its Data Hub clean room, announced in January 2026 and in pilot, does not say that off-site exposure enters it ; and its 7 October 2025 release promises select advertisers TikTok measurement with Instacart conversion data, naming no identifier .
Chewy Ads names "Clean Room solutions powered by LiveRamp" at network level, for optimisation rather than as its off-site join . Best Buy Ads reports closed-loop online and in-store sales for off-site display and video, names a LiveRamp clean room for audiences, and tests off-site lift by geographic experiment, allocation not stated; a geo experiment needs no person-level match .
Three rows state no mechanism. Walmart announced its DSP with The Trade Desk on 28 January 2021 and promised "Closing the measurement loop across digital and in-store sales" in August 2021 ; its help centre now documents the DSP's windows, touchpoints and deduplicated last-touch model, but no match method . Walgreens Advertising Group's off-site page, footered 2024 and older than the August 2025 take-private, promises closed-loop reporting , and DICK'S Media uses one closed-loop sentence on its on-site and off-site pages .
Two rows close the loop on their own surface. eBay credits a Promoted Offsite sale when the buyer who clicked buys the same item, or another from the same seller, within 30 days . Etsy attributes an order when a buyer clicks its off-site ad and buys from the shop within 30 days, and the attribution is the invoice, a 15% or 12% fee . A click that lands on the marketplace needs no outside match, so neither is credited.
The author works at Samba TV. Walmart Connect lists Samba TV among its reach and frequency partners without saying what data any partner receives , and Amazon lists it among about sixty certified measurement technologies . Samba's purchase-outcome partnerships in the releases screened, with Catalina and Affinity Solutions, do not describe measuring a retail network's media .
H4b, coded by the rule
With 13 rows, the rule reads contradicted at 4 or fewer documented matches, supported at 9 or more and unresolved in between . Every linkage search is complete: the H stream's verifier found three incomplete, for Roundel, CVS Media Exchange and Instacart, and stream X read the outstanding leads on 3 November 2026 .
Two coding decisions and two corrections shape the count; the rule itself did not change. Neither decision is in the registered rule: the session took both on 3 November 2026, after the H stream raised the boundary (chapter 14). First, a row earns credit only when its own documentation states that the exposure-to-purchase join for its off-site measurement runs through a match outside the operator's logged-in surface: a clean room, a partner or device graph, or a hashed-identifier match inside a partner platform. Generic privacy statements, match keys without a stated join and network pages that do not cover the off-site service do not count. Second, an operator's executive quoted on record in a partner's release counts as the operator's documentation . It carries CVS, whose announcement survives in a trade reprint . Two corrections followed the same day. Kroger was first credited on the roadmap entry, which the Kroger profile found and the H stream's tally did not, as a sentence of the same kind as Walmart's credited Meta sentence . The final review then pointed out that the entry measures "the impact of media" and names no purchase, while Walmart's sentence names closed-loop measurement of online and in-store sales. Applied as written, the first decision withdraws Kroger's credit, and Roundel's with it, since its executive's words name guest insights, not sales .
Under those decisions 3 of the 13 rows document an outside match, 23.1%, at or below one-third, so H4b is contradicted under its registered rule . Before the corrections the count was 5 of 13 (38.5%), between the thresholds, which would have left H4b unresolved; both moves are logged (chapter 14). Read on the operators' own pages only, without CVS's reprint, the count is 2 of 13 (15.4%), Walmart and Albertsons .
The other strict readings fall lower still. Counting only current statements drops CVS's 2024 announcement, 2 of 13 (15.4%); requiring the words clean room or identity drops Albertsons, 2 of 13; requiring both leaves Walmart alone, 1 of 13 . The loosest reading, which counts network-level clean-room statements and implied matches inside a partner platform such as Instacart's TikTok measurement, adds Kroger, Roundel, Instacart, Chewy and Best Buy, 8 of 13 (61.5%), which would leave H4b unresolved . Over the broad denominator of 22, which adds the nine network-level rows, the count is 3 (13.6%) .
Reading LiveRamp's release as a whole, whose sentence on first-party sales is about networks in general while Roundel appears only in its executive's quote, adds Roundel: 4 of 13 (30.8%), still contradicted . Walmart's credit would also stand on the Disney release alone, a 2024 proof of concept with select advertisers; without both of Walmart's statements the count would be 2 of 13 . Two more credits would move the registered verdict to unresolved, and no reading reaches the 9 of 13 that would support H4b.
On any reading the verdict is a statement about pages. It neither confirms nor refutes the rival line that off-site loops are closed by identity matches outside the retailer's surface: most off-site rows do not say how they close the loop, and no credited row names its match keys .
In the store
The IAB and IAB Europe in-store standard, final in December 2024, asks networks to report ad plays and gross impressions, defines "Ad impression = audience x ad play", wants the formula and the source of store-traffic data disclosed, points to the Media Rating Council's (MRC) digital place-based standards, and excludes stores where the media did not run from sales lift . Chapter 7 reads IAB Europe's 2026 draft revision. On the in-store pages opened by 3 November 2026, no seller publishes its impression formula .
Walmart's store ads page says "Gross impressions are estimated projections based on historical and aggregated data", footnoted to ComScore data for a standard TV Wall campaign, with no formula or store-traffic method . It says store ads measurement includes brand and sales lift, and that a proof of concept on Deli and Bakery screens, food suppliers only, September 2024 to January 2025, achieved up to 5% sales lift, a maximum, not an average . The footnote that this lift was validated at the 80% confidence level or above qualifies that one statistic; it is not a significance rule for store ads reports, and no design, control or sample is stated .
Albertsons Media Collective compares test stores with "a rigorously selected set of control stores with no media exposure", matched on nearly 60 store-level variables, and calls the result statistically validated without naming a test or threshold . It does not say whether stores are allocated at random, so assignment is unknown, and its release of 6 January 2026 calls closed-loop measurement a future capability .
Albertsons' Mondelēz case, its "first in-store Matched Market campaign", reports an iROAS of $2.41 and a 14% in-store sales lift with screens in 116 Albertsons banner stores; it does not say how many stores entered the analysis as test or control, how they were allocated, when the campaign ran or how uncertain the estimate is . The Primo Water case belongs to Albertsons' off-site row: a DOOH campaign Albertsons ran through Perion, April to May 2025, reporting "a 5.5% incremental sales lift" with no design, control, sample or interval . The case places the screens in transit hubs, gyms and offices without saying they were the near-store screens Albertsons sells, and says mid-flight optimisation concentrated on high-performing stores, so the 5.5% cannot be read as a pre-specified test against control .
Elsewhere the figures lack a usable design. Stingray, Albertsons' in-store audio partner, reports a 20.3% average sales lift across 100 matched panel tests on Nielsen Scantrack data, saying its method "relies on random assignment" while comparing stores with similar stores, and gives no interval or period . Grocery TV, a multi-retailer network outside cohort A, states a 14% average sales lift and gives average brand awareness as 27% on one page and +22% in its media kit, naming no design or measurer . Gulp Radio claims average sales lifts of 5 to 9% with no design .
The same screen carries two labels. Sellers call it store ads or in-store retail media, while IAB's DOOH Measurement Guide (July 2025) files retail stores under "Retail and Venue-Based DOOH" . Walmart's own 2021 release described its screens as a digital out-of-home platform . The guide calls a matched-store test closed-loop, a test-versus-control design; the networks use the term for an exposure-to-purchase match with no counterfactual . Both labels are kept, and no market totals are added across them.
What would change the reading
H4a would move if a US series split on one definition across 2023 to 2026 were published, or found behind a gate and registered. H4b would move with two more credits to unresolved, and with six more to supported; Kroger and Roundel would each gain one by naming the purchase outcome their clean-room measurement joins. Walmart's identity graph, announced on 3 April 2026 to join in-store transactions, online activity and media engagement and to reach its DSP, social and VIZIO over time, states no match method yet . In stores, the reading would change with a published impression formula and a stated allocation of test and control stores. A buyer can ask for both, and for the join behind any off-site sales number (chapter 13).
Chapter 7
Who checks the numbers
On the Media Rating Council's (MRC) lists as enumerated on 3 November 2026, every completed accreditation that can be mapped to an eligible retail or commerce media service covers delivery metrics, and none covers attributed sales, conversions or return on ad spend. One entry cannot be mapped on current evidence: MRC's only outcome-type entry belongs to a service named "Epsilon", which this paper ties to Epsilon's programmatic business, not its retail media platform, only by inference from a 2021 release. Under the registered decision order an unmapped letter blocks support, so H2 is unresolved; on the inference it would be supported. Of 89 eligible service rows, 12 hold any instrument from MRC, the Alliance for Audited Media (AAM) or IAB Europe, and the only completed one that touches outcomes, AAM's certification of DoorDash's incrementality method, sits outside MRC and outside H2. The guidance itself softened between versions, in the IAB/MRC guidelines and in IAB Europe's in-store draft.
Seven eligible service rows hold completed MRC accreditation through five list entries, all for delivery metrics . The list's only outcome-type metric belongs to a service named "Epsilon", which Epsilon's release of 2021 ties to its programmatic business, not to a register row . Kroger's onsite formats are in pre-audit for attributed and incremental sales and their returns on ad spend, and Kevel's for conversions and incrementality; neither is accredited . H2 is unresolved under its decision order, because the Epsilon entry is mapped only by inference; on the inference it would be supported . AAM certified DoorDash Ads' incrementality measurement and reporting for Sponsored Products and Sponsored Brands in September 2026, outside H2's MRC test . None of the six MRC letters for register services cites the IAB/MRC Retail Media Measurement Guidelines . IAB Europe's certification makes incrementality optional , and its in-store draft deletes its predecessor's test-and-control section . Instacart and Criteo describe their accreditations more broadly than MRC's letters do .
The test
H2 asks whether any MRC accreditation held by an eligible retail or commerce media service covers an outcome metric: attributed sales, incremental sales, conversions, return on ad spend (ROAS) or incremental ROAS (iROAS). The population is every eligible cohort A, B and C row on the frozen register, scored or not: 89 rows, 49 of them retailer and marketplace services, 20 commerce media services and 20 enablement platforms (chapter 1). The hypothesis is about the public record, so absence from MRC's lists is the test. The rule decides in order: contradicted if any letter in force is verified as accrediting an outcome metric of an eligible service at the exact service, metric, environment and date; otherwise unresolved if the screen is incomplete or a potentially contradicting letter is unmapped to the register; otherwise supported. AAM and IAB Europe instruments are scored under independent assurance, dimension 10, never under H2 (chapter 14). The author's own standards page says every MRC accreditation in retail media covers delivery metrics ; that is a rival hypothesis, not evidence.
Accreditation attests the service, metrics and environments a letter names, and nothing beyond them. A pre-audit or an audit in process is a status. A product built to or aligned with a guideline carries only its vendor's word.
What MRC has accredited
MRC's accredited digital list named 41 services when the session enumerated it on 3 November 2026 . Five entries concern eligible rows, and seven rows hold completed accreditation through them, because one Instacart letter covers Instacart Sponsored Products, Instacart display and video ads and Carrot Ads, the platform Instacart licenses to other retailers . Every accredited metric in those letters is a delivery metric: impressions, clicks, click-through rate (CTR), viewable impressions or invalid-traffic filtration .
| Service row | Body | Scope | Stage | Evidence |
|---|---|---|---|---|
| Amazon Sponsored Products | MRC | Clicks with sophisticated invalid-traffic (SIVT) filtration; total only | Continuing, 8 Jul 2026 | |
| Amazon DSP | MRC | Impressions, click-throughs, viewable impressions by supply source | Grants Jan and May 2026; letter 29 May 2026 | |
| Walmart Connect Sponsored Search | MRC | Impressions, clicks, CTR, invalid metrics | Initial, 26 Sep 2025 | |
| Instacart Sponsored Products, display and video ads; Carrot Ads | MRC, one letter | Impressions, clicks, CTR, viewability | Continuing, 1 Jun 2026 | |
| Criteo Commerce Yield | MRC | Display impressions, clicks, invalid counts; three on-site formats | Initial, 15 Mar 2024 | |
| Walmart Connect Onsite Display | MRC | Display impressions, viewability | Audit in process as of 11/2025 | |
| DoorDash Ads Sponsored Products and Sponsored Brands | MRC | Impressions, clicks, CTR | Pre-audit reviewed 1/2026, Audit not begun | |
| DoorDash Ads, same formats | AAM | Incrementality method and reporting | Certified, Sep 2026 | |
| Kroger Precision Marketing onsite search and display | MRC | Delivery metrics; attributed and incremental sales, aROAS, iROAS | Pre-audit in process as of 6/2026 | |
| Kevel Retail Media Cloud | MRC | Delivery metrics; conversions and incrementality | Pre-audit reviewed 9/2026 | |
| Moloco Commerce Media | IAB Europe | Platform capability, on-site display and sponsored products | Certified 7 Oct 2026 |
The scopes are narrower than the product names. Amazon's Sponsored Products letter covers clicks only, at a total level with no desktop, mobile web or in-app split, and no letter names Sponsored Brands or display ads . Amazon's own accreditation for its demand-side platform (DSP) filters general invalid traffic only; MRC separately accredited Integral Ad Science's third-party reporting of Amazon DSP in November 2025, Fire TV and Twitch excluded . Walmart's enclosure scopes its metrics to Ad Center's interface reports and invalid-traffic report and does not name the reporting API . Criteo's letter names its Retail Media Platform, mapped to Commerce Yield, and its grant does not cover the Albertsons sponsored products Criteo powers .
None of the six letters in force for register services or their operators, Epsilon's included, cites the Retail Media Measurement Guidelines MRC wrote with the Interactive Advertising Bureau (IAB); each applies MRC's Minimum Standards, its invalid-traffic addendum and IAB's delivery guidelines . No letter for an eligible row accredits attributed sales, ROAS, new-to-brand sales or lift . Delivery accreditation tells a buyer that impressions and clicks were counted to a standard, and nothing about the sales they were matched to.
Epsilon's correlated outcomes
MRC's letter of 1 October 2025 continues the accreditation of a service named only "Epsilon" for display and video impressions and correlated outcomes, online and offline, in desktop, mobile web and mobile in-app environments, and names no product, retailer or retail environment . It is the only service on MRC's list with an outcome-type metric .
Epsilon's release of 11 May 2021 names the accredited offering as Epsilon PeopleCloud Digital Media Solutions, formerly Conversant, defines a correlated outcome as an event such as a purchase connected to an ad served, and calls it "the only offering in the integrated suite to be accredited by the MRC"; MRC's chief executive, quoted there, names the same product . The 2025 letter keeps the 2021 metrics and environments and, unlike the other five letters, names no placement type . The paper's working mapping puts the entry with an Epsilon service outside the register, not with Epsilon Retail Media. It is an inference: the current letter names no product, no MRC record does, and an earlier MRC letter for Epsilon returned MRC's 404 page on 3 November .
Even an Epsilon Retail Media entry would not have been lift: a correlated outcome associates an exposure with a later event, with no counterfactual . Until a record names the product, H2 cannot be supported, because an unmapped letter blocks support .
Under review is not accredited
MRC's digital under-review list held 14 entries on 3 November 2026, five of them retail or commerce media services, and three of those carry purchase- or conversion-type metrics . Kroger's covers Targeted Onsite Ads, Product Listing Ads, Onsite Video and Promoted Product Carousels for attributed sales, incremental sales, aROAS and iROAS (attributed and incremental return on ad spend), at the stage "Pre-audit in process as of 6/2026" . Kevel's lists conversions and incrementality and names no retail environment; the register, not MRC, counts it as retail media . Schwarz Media's, for shoppers reached and units purchased in the Lidl and Kaufland apps, is not a US service or a register row . DoorDash Ads and Walmart Connect Onsite Display are listed for delivery metrics only .
A reviewed pre-audit is dated by its review, not its start, and MRC's quarterly updates record no accreditation decision for any of the five from mid-2025 to the third quarter of 2026 . Kroger's entry names no standard to be applied and no method behind its incremental sales . Kroger's and Kevel's pre-audits are the live candidates for a future counterexample; on 3 November neither had a letter .
H2, coded by the rule
Step one finds no counterexample: no letter in force on 3 November 2026 is verified as accrediting an outcome metric of an eligible service at the exact service, metric, environment and date . Step two needs a complete screen with every applicability check resolved. Workstream A stopped there, because its reads of MRC's lists were keyword reads that left 8 of the 14 entries under review unnamed . The session enumerated both lists in full the same day , and the one applicability check, Epsilon's, rests on an inference from a 2021 release, not on a record that names the product . Read as the decision order is written, the check is unresolved and so is H2; accepting the inference, H2 would be supported . The first draft led with the supported reading; the final review of 3 November 2026 pointed out that the rule does not allow it, so the strict reading now leads (chapter 14). Neither list carries a last-updated date; both are re-read on 20 November 2026, the registered freeze date, and a change will be logged as a dated correction (chapter 14) .
Across the 89 eligible rows, 12 hold any MRC, AAM or IAB Europe instrument at any stage and 77 hold none: 41 of 49 in cohort A, all 20 in cohort B and 16 of 20 in cohort C . The 12 hold 13 records: seven completed MRC accreditations of delivery metrics, AAM's DoorDash certificate, IAB Europe's Moloco certificate, Walmart's audit, and pre-audits for DoorDash's delivery metrics and Kroger's and Kevel's outcome metrics . No eligible row holds a completed MRC accreditation of an outcome metric. AAM's Platform Certification listing was not screened, so absences are bounded to the lists opened .
Outside MRC: AAM and IAB Europe
AAM launched its Incrementality Certification for Commerce Media in June 2026 to review incrementality methods for their data sources, reporting and disclosure . On 10 September 2026 it said it had evaluated "incrementality measurement and reporting for DoorDash Ads' Sponsored Products and Sponsored Brands", including methodology, data governance, causal testing framework, metric calculations and customer disclosures, against industry guidelines including the IAB and IAB Europe incrementality guidelines . DoorDash was the only company listed under the certification on 3 November .
The certificate attests a review of how DoorDash measures, not a campaign result, and does not say whether its causal testing framework assigns customers at random . AAM's compliance guide recognises hybrid and proxy approaches among its method families and states no randomisation or holdout requirement . DoorDash's own release describes its Sponsored Products lift measurement as "based on Ghost Ads", for select managed-service advertisers, against a holdout whose assignment it does not state . Restaurant Sponsored Listings are outside the certificate, and the same formats hold MRC's pre-audit for delivery metrics, so one row carries two instruments at different stages for different metrics . The certificate does not bear on H2, which tests MRC accreditation of an outcome metric; it is another body's review of a method, scored under dimension 10 .
IAB Europe certifies networks and ad tech companies after an audit by AAM, ABC or CESP, for two years, for a whole offering or named ad products . Its compliance table (v2.1, May 2026) makes a sales-uplift row mandatory for any network that reports the metric, requiring a "Statistically significant A/B Test (not before and after test)" without saying how groups are formed; incrementality is optional, and auditors check 10 campaigns chosen at random from the previous six months . Its list held three certificates, all audited by ABC and none stating a standards version: Albert Heijn (September 2025), Nectar 360 (June 2026) and Moloco (October 2026) . Moloco's, the only one on a register row, certifies platform capability for on-site display and sponsored product ads and names no metric list, retailer or geography .
What the guidelines ask for
The IAB/MRC Retail Media Measurement Guidelines went out as a draft on 13 September 2023 and were finalised in January 2024, with no version number . They call the common practice "event-level attribution (also sometimes referred to as closed-loop attribution)" . They let the network itself be the measurer, and rely on audit and disclosure rather than on separating measurement from sales; outcomes used for pricing or optimisation are encouraged to be audited and validated by a third party . The document recommends compliance with MRC requirements and, ultimately, accreditation .
On incrementality they say that "When feasible, marketers are recommended to use RCTs", randomised controlled trials, and that holdouts are not always available or appropriate; they require no holdout, minimum sample or significance rule . Between draft and final, user-level attribution became event-level attribution, experimental design became experimental design and non-experimental approaches, and marketers who should use randomised trials became marketers who are recommended to . Across the text, must fell from 49 occurrences to 35 and should from 126 to 86, while is recommended or are recommended rose from 1 to 45, or 49 counting line breaks; the counts show direction, not the status of any clause . Operators helped write it, Albertsons Media Collective among the listed participants, which is participation, not assurance .
The IAB and IAB Europe Guidelines for Incremental Measurement in Commerce Media (3 November 2025) define incrementality as the causal impact of marketing against what would have happened without it, and class platform-reported incrementality and baseline-versus-exposed analysis as proxies giving "directional evidence of lift" rather than an independent control . They are a method framework, not a certification standard .
IAB Europe's Retail Media Measurement Standards (V1, April 2024) became the Commerce Media Measurement Standards V2 on 22 January 2026, adding a formal incrementality definition, gross and net sales, five new-to-brand timeframes and a 30-day default lookback, with certification only under V2 after July 2026 . The PDF now served has a V2.1 cover dated May 2026 over a body that says V2, and no V2.1-specific change is stated anywhere opened .
In store: the controlled design drops out
The joint IAB and IAB Europe In-Store Retail Media Definitions and Measurement Standards, final on 3 December 2024, define five store zones, have networks report ad plays and gross impressions, and recommend 30-day sales windows before and after exposure . Beside Sales Lift and a before-and-after Sales Variance, to be used only when incremental studies are not possible, they carry an Incremental Sales/Brand Lift section: test versus control, consumers ideally sorted into groups at random, group sizes and statistical significance reported .
IAB Europe's Version 2, a draft it released alone on 17 September 2026 with comments due by 23 October, re-anchors the impression on footfall and moves the sales windows to 28 days . It drops Version 1's Sales Lift and Brand Sales Lift metrics and the whole Incremental Sales/Brand Lift section, keeps Sales Variance, and leaves one line pointing to the incrementality guide . Its change table records neither deletion: it says "All other definitions carry over from Version 1 unchanged" and that incrementality continues under Incremental Sales/Brand Lift, a section the draft no longer contains . IAB Europe's release, as reprinted by ExchangeWire, calls the recommendations proposals and does announce the removal of the sales lift metrics . Version 1 remains the standard; adopted as written, the draft would remove the only in-store text that described a controlled design and asked for group sizes and significance . Chapter 6 covers in-store measurement in practice.
What buyers asked for
The Association of National Advertisers (ANA) publishes buyer research, not standards. Its 2023 report, from 138 members surveyed in summer 2022, found that "a lack of measurement standardization and transparency are preventing advertisers from deriving full value" . Its 2024 release ranked the lack of standardisation across platforms first among challenges, at 55%, and said its measurement council was working with MRC on must-haves for every network . Its August 2026 report sets out an initial framework for outcome comparability behind a log-in . Across the timeline, eight MRC decisions since 2024 concern register services, and none covers an outcome metric .
What the services say about themselves
Instacart's methodology page of 12 August 2026 says "All of these impression metrics are accredited by the Media Rating Council" on a page whose platform list includes Caper carts and Rosie . Its Caper Carts page says eligible shoppable display campaigns extend to the carts automatically and Ads Manager reporting pools every surface, and its November 2025 release promised verified metrics on more than 240 e-commerce sites . MRC's letter names the Instacart Marketplace and retailer properties reached through Storefront, Storefront Pro and Connect API integrations, and not Caper, Rosie or retailers' own storefronts . Inference: Ads Manager display totals that include Caper impressions are not shown by the letter to be accredited as reported .
Criteo's 10-K says "Our MRC-accredited technology provides our clients with the unique ability to measure against product sales at the product SKU level", though its letter covers display impressions, clicks and invalid counts only . The sentence stops short of calling sales measurement accredited, but Criteo's 2024 release tied the accreditation to commerce outcomes, and its Commerce Yield note says MRC certifies systems meeting the IAB/MRC guidelines, which the letter does not cite .
Kroger Precision Marketing's FAQ says it is "proudly certified by the Interactive Advertising Bureau (IAB) and Merchant Risk Council (MRC)"; the Merchant Risk Council is a payments-fraud body, no certificate from either was found, and the Media Rating Council lists Kroger only in pre-audit . It also says its off-site incremental methodology adheres to IAB standards, a self-attestation . Albertsons Media Collective, on no list, says its in-store measurement complies with and is aligning to IAB standards, with no attestation behind it . Under the scoring rules such statements earn the lowest non-zero rung of independent assurance, and a listing in an accreditor's process the rung above (chapter 12).
What assurance covers
In the record screened on 3 November 2026, independent assurance of retail and commerce media comes to delivery accreditation on seven rows, a delivery audit or pre-audit on two more, outcome pre-audits on two, one review of DoorDash's incrementality method and one European platform certificate . For no eligible service that can be mapped has MRC accredited the attributed or incremental sales number a buyer receives, and the one unmapped outcome-type entry, Epsilon's correlated outcomes, keeps H2 unresolved . A buyer can check a letter's date, service, metrics, environments and named reports against MRC's current list, and ask of any lift figure whether a certified method produced it and how its groups were assigned. Chapter 12 scores these instruments and chapter 13 turns them into questions.
Chapter 8
What the studies show, and what buyers are shown
Estimates made without a randomised comparison can miss the experimental effect by multiples, usually upward, and experiments precise enough to settle a return need very large samples: on that, the designed studies agree. None measures what a brand gains from a US retail media network, and every study of advertiser outcomes inside a network found by 3 November 2026 was written by the network or with it. Buyers are shown figures published by the seller of the media or the measurement; of thirteen examined here, one states random assignment and none prints an interval.
In 15 randomised Facebook experiments, observational methods missed the experimental lift by a factor of three in half the studies . In 663 later experiments, median lifts of 29%, 18% and 5% compared with 83%, 58% and 24% from double machine learning . Switching off eBay's non-brand search ads in randomised regions gave a return on investment of -63%, against over 1,400% from regression . No refereed study from 2024 to 2026 measuring a brand's lift from a US retail media network was found . In Ovative and Albertsons Media Collective's own study of 42 onsite display campaigns, the gap in iROAS (incremental sales per ad dollar) between matching choices averaged 6.5 times (median 2.5 times), and 83% of campaigns flipped sign, worded elsewhere in the report as "could flip" . Amazon's attribution paper calibrates on randomised experiments but reports no accuracy figure for the model behind advertisers' numbers . Association of National Advertisers (ANA) member surveys found 88% (fielded 2022) and 63% (fielded 2023) reporting retailer influence, on different wordings . No survey opened measures spend tied to joint business plans .
This chapter reads three records: published studies with a stated design, coded by who ran and paid for them, how units were assigned and what they can be generalised to; the outcome figures networks and vendors publish; and the trade surveys on retailer influence and joint business plans (Q5). Under the rules of chapter 1, an attributed figure is never a return, a holdout is not randomised unless its design says so, and a third party is not independent unless the funding and data fields show it. What buyers receive by default is coded in chapter 3.
What the experiments measured
The clearest evidence comes from platforms that ran randomised controlled trials (RCTs) and then re-estimated the same campaigns without them. In 15 US Facebook experiments from 2015, Gordon, Zettelmeyer, Bhargava and Chapsky found that observational methods generally overestimated the effect, and in half the studies were "off by a factor of three" . Two of the four authors worked for Facebook, which supplied the data; the other two state that Facebook did not pay them .
The follow-up by Gordon, Moakler and Zettelmeyer used 663 treatment-control pairs from 563 US Facebook experiments run from November 2019 to March 2020 . Median experimental lifts of 29%, 18% and 5% by funnel stage compared with 83%, 58% and 24% from double machine learning and 173%, 176% and 64% from stratified propensity-score matching . All three authors were Facebook or Meta employees for the paper, two part-time as a condition of data access . Both studies measure pixel-tracked conversions on a social platform, not purchases in a retailer's transaction log .
Blake, Nosko and Tadelis tested paid search from the buyer's side. In 2012 eBay stopped non-brand search ads for 60 days in 68 US designated market areas chosen by restricted randomisation (random candidates, poorly balanced draws discarded, a coin flip to pick the group), with 142 areas as controls . Paid search added 0.66% to sales, an experimental return on investment of -63% (95% interval -124% to -3%), against over 4,100% and over 1,400% from ordinary regression . All three authors were at eBay Research Labs, and the evidence comes from one advertiser with a strong brand buying search, not from media a retailer sells .
Lewis and Reiley randomised 1.6 million customers of an anonymous retailer on Yahoo!: display ads raised purchases by about 5% (a log difference-in-differences estimate, standard error 2.3%), 93% of the increase came in stores and 78% from people who never clicked . A cross-sectional estimator on the same data understated the effect, with a bias three times its size . Yahoo! funded the work and guaranteed academic independence, and both authors were Yahoo! Research employees . Lewis and Rao pooled 25 randomised display experiments for 19 retailers and 6 financial firms and found the median confidence interval on return on investment over 100 percentage points wide and that informative tests can need more than 10 million person-weeks; many of those experiments came from the same Yahoo! work .
Ghost ads matter here because networks cite them. A 2016 working paper by authors then at the University of Rochester, Netflix and Google sets out the method, which identifies the control-group users who would have been exposed in a randomised experiment . Instacart and DoorDash both name it . The one marketplace experiment found, at Flipkart in India, reports category-dependent effects of sponsored listings; its abstract states neither allocation nor funding, and it measures the platform's outcome, not a brand's lift .
Read together, the experiments show advertising effects that are small beside the noise in individual sales, and non-experimental estimates that miss by multiples in either direction . Of the eight experimental studies in the table that no retail network wrote, six have at least one author employed by the company whose data were used, and the other two do not state their funding . In three targeted searches, a verifier's search and the pages screened by 3 November 2026, this review found no refereed study from 2024 to 2026 that measures a brand's incremental sales from a US retail media network with a published design . The table lists the thirteen studies coded.
The retail media studies are the networks' own
Three documents describe incrementality measurement for advertisers inside a retail media network with a stated design, and a network wrote or co-wrote each . The most detailed, "Retail Media iROAS Demystified" (March 2026), is by two Albertsons Media Collective staff, two Ovative Group staff and two Northwestern Kellogg professors, Eric T. Anderson and Brett R. Gordon, who are authors, not outside reviewers . It carries no funding, compensation or competing-interest statement . Gordon is also an author of Amazon's attribution paper, listed there with an Amazon affiliation; whether that affiliation held in March 2026 is unresolved .
It is a retrospective sensitivity analysis of 42 of the Collective's onsite display campaigns on Albertsons websites, re-estimated under different matching choices, with no randomised benchmark . Its headline comes from the variations without pre-filtering: within a campaign the gap between the highest and lowest iROAS averaged 6.5 times, with a median of 2.5 times, and 83% of campaigns flipped from positive to negative depending on method, which page 12 and the executive summary word as campaigns that "could flip" . The report does not say how it computed a ratio for campaigns whose iROAS changed sign .
Each choice moved the result. Restricting the analysis to past brand buyers cut the average iROAS from $2.27 to $0.22 and shrank the test group to 17% of its size; the filter also changes the question asked . Cluster matching gave $1.80 against $0.73 for one-to-many propensity-score matching, which balanced the groups about twelve times better on the authors' own measure . The variants are therefore not equally credible . The study shows that matched-control iROAS in one network's onsite display campaigns is sensitive to analyst choices; it does not show how accurate any method is or what Albertsons reports by default, and its line that the Collective "cannot always rely on RCTs" stays with the study .
Albertsons' own communications drop the qualifiers. Its release of 16 April 2026 reports a single $7.45 iROAS for S. Martinelli & Co. with no period, spend or interval, and cites the research as showing that iROAS "can vary by 6.5x", with no median, hedge or scope . Its vice president of media and measurement wrote on 23 March 2026 that clean experimental isolation is not always possible and that a randomised design is offered on request .
Amazon's multi-touch attribution paper (arXiv, 11 August 2025) has eight authors: six list an Amazon affiliation and two are marked as work done at Amazon . Its calibration experiments are randomised, with a group of shoppers eligible to see a campaign's ads and a holdout that is not, drawn from what it calls hundreds of thousands of RCTs; no sample, holdout share, period, interval or result is reported . The number for advertisers is a model: it predicts experimental conversions from attributed ones, extends the effects to campaigns "that were not run as RCTs" and normalises them to touchpoint shares, with no accuracy figure .
The paper names no ad product , and Amazon presents multi-touch attribution as a beta whose reach to Sponsored Products is unconfirmed . The public accuracy evidence found for the method it extends comes from Meta data and authors employed, at least part-time, by Facebook or Meta . Its description of the 2023 Facebook study as almost 2,000 representative campaigns matches no opened version of that study, whose latest arXiv text reports 563 experiments; the journal version was not opened .
Instacart's engineering post of 30 June 2023 is the most complete design a network has published in the sources screened. Users are randomised into an ads group and a holdout per user-advertiser pair, ghost ads simulate the holdout's impressions, and the outcome is total sales of the advertised products, for Sponsored Products only . The post states no significance rule, minimum sample or interval, and does not say whether advertisers receive the results . In 2026 the offer appears as a sales lift testing card with no design or eligibility, and a 2023 case page says lift testing may not be available to all partners .
Study evidence table
Outcome measures are deliberately not pooled: different outcomes are different quantities. Sort by any column; sponsorship is shown so a vendor-funded result is never read as independent.
Full records with limitations and source passages are in data/study-evidence.csv.
| Study | Year | Design | Sample | Setting | Format / channel | Outcome measure | Headline result | Uncertainty | Sponsor | Generalisability | Src |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Ovative Group with Albertsons Media Collective, Retail Media iROAS Demystified | 2026 | retrospective sensitivity analysis of matched-control iROAS; non-random; 54 iROAS variants per campaign (executive summary p. 3; not reproducible from the option list on p. 14); no RCT benchmark | 42 campaigns; customer counts indexed only | Albertsons Media Collective web properties (US grocery retail media) | onsite display | iROAS (incremental revenue ÷ spend) | no-filter variations (Exhibit 12): highest-to-lowest iROAS gap averaged 6.5x (median 2.5x); 83% of campaigns 'flipped' / 'could flip' sign; e.g. brand-purchaser filter $2.27 to $0.22; PSM 1:many $0.73 vs clustering $1.80; BSTS $0.97 vs observed $1.56 | none reported per campaign; t-test and BSTS credible intervals described as methods | Ovative Group and Albertsons Companies (party study; operator staff co-authors); no funding, compensation or competing-interest statement; co-author Brett R. Gordon listed with Amazon on an August 2025 paper (open interest question) | one network's onsite display; sensitivity of matched designs, not accuracy | |
| Amazon Ads Multi-Touch Attribution (arXiv 2508.08209), calibration experiments | 2025 | randomised treatment-eligible vs holdout shoppers; 'hundreds of thousands of RCTs'; method description | not disclosed | Amazon Ads (on-Amazon conversions) | all Amazon Ads touchpoints | conversions (RCT conversions) | no results reported | none reported | Amazon (authors employed by Amazon; Gordon and Zettelmeyer also Northwestern) | Amazon only; nothing disclosed to generalise | |
| Amazon Ads Multi-Touch Attribution (arXiv 2508.08209), reported attribution model | 2025 | ensemble prediction of RCT conversions from last-touch and model-driven attributions; touchpoint credits normalised to shares; no per-campaign counterfactual | not disclosed | Amazon Ads | all Amazon Ads touchpoints within a lookback window | attribution credits and shares (conversions) | no accuracy figures reported | none reported | Amazon | Amazon only | |
| Instacart, The Methodology of Ad Incrementality (engineering post) | 2023 | randomised holdout per user-advertiser pair (hash of user_id, advertiser_id) with ghost ads, assigned before the auction | not stated | Instacart marketplace (US grocery) | Sponsored Products only | total incremental sales; incremental sales lift %; incremental sales ÷ ad spend | none reported in the post (system-level spend-weighted iROAS described) | none stated | Instacart (operator) | one operator's framework; not shown to be a default report | |
| Johnson, Lewis and Nubbemeyer, Ghost Ads (JMR 54, 867-884) | 2016 WP / 2017 | randomised experiment; ghost ads identify the control-group counterparts of exposed consumers (2016 working paper; JMR version not opened) | not read | Google display network; an online sports and outdoors retailer's retargeting campaign (working paper) | display | website visits and purchases (working paper) | visits +17%, purchases +11% (WP abstract; WP intro gives visitors +26.6%, transactions +12.0%, sales +10.8%) | WP intro t-statistics: visitors t=40.38, transactions t=5.42, sales t=3.45; JMR version not read | unknown (no funding statement in the WP); authors at Rochester, Netflix and Google (Feb 2016); thanks 'many Google employees and advertisers' | design reference for Instacart, Facebook and the IAB/MRC guidelines | |
| Gordon, Zettelmeyer, Bhargava and Chapsky, A Comparison of Approaches to Advertising Measurement | 2018 WP / 2019 | 15 US RCTs vs observational re-analyses (matching, stratification, regression) | 500 million user-experiment observations; 1.6 billion impressions; 2-140 million users per study | Facebook (advertiser conversion pixels) | Facebook ads | conversions (checkout in 14 studies; registration 5; page view 3) | observational estimates 'off by a factor of three' in half the studies; generally overestimate; 6 of 14 checkout RCTs not significant at 5% | 95% bootstrap CIs per study | academic authors uncompensated; two Facebook co-authors; Facebook data | social platform, not retail media; data richer than most advertisers hold | |
| Blake, Nosko and Tadelis, Consumer Heterogeneity and Paid Search Effectiveness | 2014 WP / 2015 | DMA-level switch-off; restricted randomisation (random candidates, re-randomised 68/65 split, coin flip); difference-in-differences; 2SLS | 68 test DMAs, 142 control DMAs (210 US DMAs); 30% of eBay US traffic | eBay as advertiser on Google and MSN paid search | paid search (brand and non-brand keywords) | sales (US$); ROI | paid search adds 0.66% to sales; experimental ROI -63% vs OLS over 4,100% / over 1,400%; brand keywords no measurable short-term benefit | ROI 95% CI [-124%, -3%]; DMA-clustered SEs | all three authors at eBay Research Labs (advertiser, not media seller) | one large marketplace brand buying search; not retailer-sold media | |
| Lewis and Rao, The Unfavorable Economics of Measuring the Returns to Advertising | 2015 | synthesis of 25 randomised field experiments (random holdouts of targeted users from online display ads); statistical-power argument | 25 RCTs (19 retailers, 6 financial service firms); 'most reaching millions of customers'; $2.8 million spend | US retailers and financial service firms (display, on Yahoo!) | display | sales; ROI | 'The median confidence interval on return on investment is over 100 percentage points wide'; 'informative advertising experiments can easily require more than 10 million person-weeks' | the paper's subject | authors were Yahoo! employees when the experiments ran (Yahoo! sold the media); funding not stated | general to small per-person effects; not retail-media specific; many experiments come from Yahoo! Labs work incl. Lewis and Reiley (2014), so not an independent replication | |
| Lewis and Reiley, Online ads and offline sales | 2014 | randomised customer-level holdout on Yahoo! for an anonymous major retailer | 1.6 million customers; 75,000 with nonzero purchases | Yahoo! portal display for a retailer (online and in-store sales matched) | display | purchases (R$ as reported, online and in-store) | +5.0% purchases (log difference-in-differences estimate E[Δ], midway between the R$0.083 experimental and R$0.102 DID level estimates); 93% of the increase in stores; 78% from non-clickers; cross-sectional estimator biased by three times the effect | SE 2.3% on the log-DID percentage estimate | Yahoo! (media seller) funded, with independence guaranteed; authors Yahoo! Research employees | retailer advertiser, portal media; closest analogue to offline sales matching; the retailer chose the customer subset by an undisclosed rule | |
| Sahni and Zhang, Are consumers averse to sponsored messages? | 2024 | randomised field experiment on search-ad prominence (status quo vs ads sidelined) | 3.3 million US users | an unnamed search engine | search ads | search usage, revenue, website traffic | lower advertising reduced usage; newest decile of websites lost 10% traffic | not in the abstract | unknown (funding not stated; an unnamed search engine ran the experiment) | search engine, not a marketplace; platform-side estimand | |
| Gordon, Moakler and Zettelmeyer, Close Enough? | 2022 arXiv / 2023 | 663 US Facebook RCT pairs vs DML and SPSM with 5,000+ features | 563 experiments (663 pairs); ~7.9 billion user-experiment observations; >38 billion impressions; median 7.3 million users | Facebook ads | conversions (page view, add-to-cart, purchase) | median RCT lifts 29/18/5% vs DML 83/58/24% and SPSM 173/176/64% | medians; distributions of relative error | all three authors Facebook/Meta employees for this paper (Gordon and Zettelmeyer part-time, as a condition of data access; Moakler full-time with Meta stock); Facebook data | social platform; richest observational data, still fails | ||
| Einav, Leibtag and Nevo, validation of Nielsen Homescan | 2008 WP / 2010 | transaction-level match of panel records to one grocer's register records | thousands of transactions (count not on the page) | US grocery (one large chain) | n/a (panel data) | recording accuracy (trips, products, prices, quantities) | unreported trips and recording errors present; price errors most prevalent, partly from Nielsen's price construction; corrections feasible | not on the page | USDA ERS cooperative agreement with Northwestern | scanner panel in 2008; not receipt or loyalty panels | |
| Abhishek, Jerath and Sharma, The Impact of Retail Media on Online Marketplaces | 2025 | large-scale field experiment varying sponsored-listing share (allocation not disclosed in the abstract) | not in the abstract | Flipkart (India marketplace) | sponsored product listings in search | clicks and conversions; marketplace revenue | category-dependent: ads worse than displaced organic in electronics, better in clothing and footwear; +~10 pp sponsored share did not change clicks or conversions | not in the abstract | unknown (funding not stated); Flipkart data | India; platform-side estimand |
What buyers are shown
The table sets thirteen published outcome figures from cohorts A and B beside what each publisher discloses about the design (ROAS is attributed sales divided by ad spend; GMV is gross merchandise value; CPG is consumer packaged goods).
| Figure (publisher, date) | Design as disclosed | Kind of number |
|---|---|---|
| Over 30% average sales lift from 30 CPG Sponsored Products tests in H1 2025, headline "Up to 30%" (DoorDash, 24 March 2026) | Non-ad-exposed holdout; based on ghost ads; select advertisers through managed service | Lift; holdout, allocation not stated |
| $0.59 incremental in-store sales per household, three times a beverage benchmark (DoorDash and Circana, 2 September 2026) | A household-level lift study; who ran it, period, control and panel not stated | Lift claimed; no design |
| 16.5% incremental sales lift for Reckitt, Q3 and Q4 2022 (Instacart, May 2023) | Two randomly selected groups; no sample, interval or spend | Lift; random assignment stated |
| 21% and 15% in-store sales lift for two unnamed brands, April to June 2024 (Instacart, 17 December 2024) | Against a "Circana control group"; construction and assignment not stated; Instacart collaborated | Lift; allocation not stated |
| $7.45 iROAS and 33% sales lift, S. Martinelli & Co. (Albertsons Media Collective, 16 April 2026) | Test and control frameworks; no period, spend or interval | Lift; allocation not stated |
| $2.41 incremental return on spend and 14% in-store sales lift, Mondelēz (Albertsons Media Collective, 6 January 2026) | Control stores matched on nearly 60 variables; allocation, store counts and dates not stated | Lift; matched stores, allocation not stated |
| 2.5x incremental ROAS and 1.8x sales lift against social industry benchmarks (Walmart Connect, 8 April 2026) | Aggregate over campaigns that qualified for a lift study in four categories; Circana analysed both sides | Benchmark comparison |
| $2.15 incremental total GMV per Search dollar, one brand, one week in August 2026 (Walmart Connect with QBR Media, 30 October 2026) | Brand cut spend by about 95%; seasonally adjusted baseline, placebo outcome; interval not printed | Pre/post; no assignment |
| 20% rise in Ulta's spend via PayPal against prior weeks, June to July 2025 (PayPal, 6 January 2026) | Before against during; no control | Pre/post |
| One campaign over-attributed by over $2.8MM against $98K (Kroger Precision Marketing, 19 July 2024) | All households against viewable-exposed households | Attribution filters; no counterfactual |
| 280% click-through lift and 10% ROAS lift, pet brand (Roundel, undated) | None; the linked case reports 437% ROAS on 14-day click and view attribution in 2021 | Attributed ratio |
| $509K incremental sales of $878K revenue and 10:1 ROAS, LensDirect, 90 days from 15 September 2024 (Chase Media Solutions, 7 July 2025) | None; the offer targeted cardholders with no brand purchase in the past year | Incremental label; no comparison |
| 5-9% average sales lifts, Gulp Radio (7-Eleven, 12 November 2024) | None; Gulp's site describes control and exposed reporting in the future tense | Lift claimed; no design |
One of the thirteen states random assignment beside its number, and none prints an interval . Four describe a holdout, control group or matched stores without saying how units were allocated . Two compare a period with an earlier one, two are attributed, one sets qualifying campaigns against a benchmark, and three state no design . An attributed ROAS is not a return, and Roundel's tile figures do not appear on the case page they link to .
Labels run ahead of designs. DoorDash's average is a floor in the body and a ceiling in the headline, and its restaurant resource states random assignment only for the intent-to-treat design that ghost ads replaced . Circana's DoorDash release calls a descriptive buyer-overlap analysis "proof" of incremental demand, and its one footnoted period belongs to that analysis, not to the $0.59 . Instacart presented earlier Circana buyer-overlap analyses as third-party measurement and validation . Best Buy's report that its ads beat a national media ROAS average by 17% comes from a Circana study whose outputs point to a modelled mix study . Walmart Connect announced test-and-control incrementality for Sponsored Search in April 2025, and no general-availability notice was found on the pages read .
Third party, independence unknown
Several of these figures carry an outside firm's name, and in each the payer is unstated or is the seller. Walmart calls Circana independent and not affiliated, but Circana ran both sides of the benchmark and the page names no payer . DoorDash and Circana call their partnership "independent, third-party validation", offered to select advertisers, with no payer stated . Cardlytics' Nielsen sales lift runs on purchase data seen through Cardlytics' own platform . Fetch enlisted KPMG for what it calls an independent audit, publishes no report, standard or period, and divides incremental revenue by the amount invoiced to the brand . Ipsos MMA names DoorDash as the funder of its marketing mix study, whose 42% carry-over figure appears only in DoorDash's release .
A named firm computing on the seller's data with the payer unstated is coded independence unknown. On that basis, in the sources screened by 3 November 2026, this review found no independent validation of the outcome numbers of Instacart, Kroger Precision Marketing or Cardlytics, and no Circana page among those read that describes how Media Lift allocates households or validates its own numbers . Receipt and loyalty panels have little public validation either: in one search and the pages screened, the nearest found are a 2008 match of Nielsen Homescan records to one grocer's transactions and a 2016 US Department of Agriculture comparison of IRI's household panel, now Circana's, with government benchmarks . The Alliance for Audited Media's certification of DoorDash's incrementality measurement, the only entry on its list in that category, is read in chapter 7 .
Retailer influence and joint business plans
The ANA's two surveys are the only ones opened with a measured item on retailer influence . The 2023 report rests on an online survey fielded in summer 2022: 138 members took part, 80 of them using retail media networks, from a membership that includes nonprofits and marketing solutions providers as well as marketers . Its release says 88 percent "believe they are somewhat or heavily influenced by retailers" to buy on their networks, without saying whether the base is 138 or 80 . WARC's summary of the gated report records that 85% of brands surveyed feel pressured by retailers to support the networks; whether that is the same item is unresolved . The release says spending shifted to networks from shopper, brand and trade budgets, a statement about where money came from, not about trade terms .
The 2024 report, a survey of client-side marketers fielded in November 2023, found 63% saying retailers have "a lot of or complete influence" on the decision to use networks; sample size and item base are not on the open pages . The items are worded differently, so 88% and 63% are not a trend . Both are member samples from 2022 and 2023, which this paper does not read as the view of US advertisers in 2026, and influence as worded is not pressure.
On joint business plans (JBPs) between brands and retailers the record is thinner. Microsoft Advertising's 2024 study with the Digital Shelf Institute counts 18 executives in Microsoft's post and 14 brands and retailers on the institute's page, with no dates, frame or method on either, and the post's claim that improving the process can raise return on investment by 10% carries no source . Trade reporting is anecdote. In March 2025 Digiday quoted an anonymous agency executive whose two clients walked away from Walmart JBPs because they "felt that they got so shook down" . In April 2025 it quoted five named agency sources seeking flexible commitments, while reporting that brands rarely upend them .
This review found no survey of US advertisers with a stated frame, field dates and sample that measures retailer influence or JBP-linked spend for 2025 or 2026, in the searches run by 3 November 2026, so the paper gives no estimate of retail media spend inside trade terms . Inference: the retailer whose influence the surveys describe also sells the media, and every figure in the table above was published by the seller of the media measured, alone or jointly . Chapter 4 reads who holds the pen.
What would change the reading
The reading would change if a network published randomised lift, with allocation, sample and interval beside the attributed figure; if a party with no stake in the media replicated a network's numbers on data it controls; or if a survey with a stated frame measured JBP-linked spend. Until then a buyer can ask who computed a figure, on whose data and at whose expense, how the comparison group was allocated and what the interval is. Chapter 13 turns these into questions.
Chapter 9
The software underneath the networks
The platforms retailers license to run a network, the tools brands use to buy across networks and the firms that sell outcome measurement document what the networks document: attribution by default, and lift claimed more often than designed. In the pages screened by 3 November 2026, Criteo Commerce Yield is the only enablement platform that states default attribution windows, and Moloco the only one that documents a randomised holdout for clients' campaigns. The instruments enablement platforms hold cover delivery metrics, platform capability or a pre-audit, and none of the seven outcome measurement providers says how its control groups are assigned.
Kevel, Koddi, Topsort and Moloco document last-touch attribution with configurable windows between 1 and 30 days, and none states a default . Commerce Yield applies 30 days after a click and 1 day after a view unless a campaign changes them, and applies a change retroactively . Moloco describes a randomised ghost-bidding holdout in an undated blog with no sample or significance rule , Osmos an early-access holdout of unstated assignment , and Koddi, Swiftly and Zitcha assert incrementality without a design . Carrot Ads, the platform Instacart licenses to other retailers, names six US networks with dated use, Schnucks since 2021 . Criteo's Media Rating Council (MRC) accreditation of 15 March 2024 covers display impressions and clicks for three on-site formats , and Instacart's letter covers Carrot Ads placements reached through Instacart's integrations, not Caper carts or retailers' own storefronts . No cross-network optimiser documents incrementality testing with an assignment or a counterfactual method . Circana, Nielsen, Numerator and Attain describe tests or exposed-versus-control comparisons without saying how the control group is formed , and Incremental says it is independent of media sellers while naming them among its customers .
Three kinds of software
Cohort C holds the enablement platforms a retailer licenses to run a network: ad servers, sponsored-product engines, self-serve portals and measurement modules. Cohort D holds the tools brands and agencies use to plan, bid and report across networks, and cohort E the services that measure retail media outcomes and sell the result (chapter 1). The register holds 20, 11 and 7 rows. A platform is coded on what its own documentation says it supports, never on what one client deployed, so its options describe what a network built on it could report, not what any network does. Membership of cohort E earns no presumption of independence, which is coded claim by claim (chapter 4).
What a retailer can set
Kevel, Koddi, Topsort, Moloco and Streaem document configurable attribution windows , and Criteo documents them for Commerce Yield, its retailer-side platform, with defaults.
| Platform | Attribution documented | Default stated | Lift design for clients' campaigns |
|---|---|---|---|
| Criteo Commerce Yield | Post-click 7, 14 or 30 days; post-view none to 30 days ; one event per sale, clicks over views | 30 days post-click, 1 day post-view | None; an A/B test results report named without a design |
| Kevel Retail Media Cloud | Opt-in last touch; 1, 7, 14 or 30 days per flight and ad | None | None |
| Koddi Ads | Last touch, not configurable; any window under 30 days; halo variants; loyalty matchback in store | None | Asserted, no design |
| Topsort | Last click or last impression per format; 1 to 30 days | None; values shown are an example | None on the pages opened |
| Moloco Commerce Media | Last touch, clicks only; 1, 7 or 14 days for the whole platform; returns and cancellations not deducted | None | Randomised ghost-bidding holdout, vendor blog |
| Osmos Retail Media OS | Off-site attribution asserted, no parameter | 95/5 split, 8-week minimum for lift | In-campaign holdout, assignment unstated, early access |
| Carrot Ads | No window or crediting rule; orders count only if ingested within three days | None | None on the developer pages opened |
| Amazon Retail Ad Service | Attributed sales and return on ad spend (ROAS) measured in AWS Clean Rooms, no parameter | None | None |
The help centre of Commerce Max, Criteo's buyer-side product, which workstream F read after Commerce Yield's settings page refused its tool, offers the same windows with no default . Commerce Yield and Moloco each document six of the eight attribution parameters this paper records: Commerce Yield states no sales basis and takes in-store sales from retailers' store-sales files for on-site campaigns only , and Moloco documents no in-store conversion . Streaem's owner and status could not be confirmed .
The other eleven platforms document no attribution parameter in the pages screened: Criteo Commerce Grid, Epsilon Retail Media, Pentaleap, Vantage, Mirakl Ads, NIQ Activate, Rippl, STRATACACHE, Swiftly, Zitcha and ADvendio .
These are menus. Inference: two networks on Kevel can report a one-day and a 30-day window for the same kind of campaign and both be using the platform as documented . Part of the sales figure comes from the client: Kevel's ROAS divides a gross merchandise value the customer supplies by Kevel's revenue metric, and does not address returns, taxes or shipping . Kevel accepts purchase events up to 15 days after a window closes , and Commerce Yield's changed settings are "retroactively applied to the full extent of the campaign" .
Lift is claimed more than designed
Moloco is the only enablement platform in the pages screened that documents a randomised design for clients' campaigns. Its undated blog describes ghost bidding: where the advertiser's ad would have won, Moloco withholds it "from a randomized subset of users", counting every purchase in both groups . The test is offered above minimum traffic thresholds it does not state, and the incremental ROAS of 253% to 1,609% it reports comes from unnamed advertisers with no period, sample or significance rule .
Osmos describes an early-access product, limited at first to on-site sponsored display, that holds out a small share of eligible users inside a campaign, 5% by default, for at least eight weeks. It says the control "matches the exposed profile" but not how users are assigned . Pentaleap's randomised A/B tests measure the ad revenue a retailer gains from its ranking layer, put at 80% to 140%, not an advertiser's outcome .
The rest is assertion. Koddi's pages mention incrementality testing with no design . Swiftly carries a PepsiCo quote of more than $100 incremental ROAS (iROAS), a dollar figure as written, and over 39% incremental sales lift, with no method . Zitcha offers to "Measure the sales your media actually caused" without saying how . Criteo's user-split and geo-split framework is described for Commerce Growth, part of its performance media business, where a real-time incrementality metric is in pilot .
Kevel shows the gap between status and documentation. Its guide describes last-touch attribution and no holdout, A/B or incrementality design , while MRC's under-review list names "Clicks, Conversions and incrementality" among the metrics in its pre-audit . Inference: outside Moloco and Osmos, a network that wants controlled lift for its advertisers gets no documented design from its platform and must build or buy one.
Where the platforms run
Platforms rarely say which US networks run on them, or since when. Counting named US networks with a dated record of production use, Carrot Ads documents six: Schnucks, live since 2021 according to Instacart's blog of 19 August 2025, Hy-Vee, Sprouts, Thrive Market and Uber Eats (6 November 2025) and Associated Food Stores (11 November 2025) . Gopuff and Vroom Delivery were announced as future users, and Instacart says "more than 310 partners" use the platform .
Criteo says it powers the networks of "approximately 235 retailers" worldwide, but its dated statements name few US ones for Commerce Yield: Belk's network launched on it on 8 October 2024, while the releases on Albertsons (January 2024) and Dollar General (31 May 2024) announce selections with no go-live date . Albertsons described sponsored products served through Criteo in its app on 23 June 2026 ; tying that to Commerce Yield is this paper's inference.
Kevel's dated record is Dollar General's network (20 May 2025) and The Home Depot's Orange Access (3 October 2024), which runs on Vantage with Pentaleap and Kevel on its ad server . Pentaleap adds Macy's selection of 22 January 2026, with no go-live date . Amazon Retail Ad Service, a US beta, named iHerb, Oriental Trading Company and Weee! on 9 January 2025 . STRATACACHE powers the in-store screens Albertsons Media Collective debuted in January 2026 . Moloco's one dated US client is Wayfair, in a statistic for the fourth quarter of 2024 , and Epsilon's is Gopuff's off-site launch on CitrusAd, Epsilon Retail Media's former name, in August 2022 .
Koddi, Topsort, Osmos, Zitcha and Rippl name US retailers only in undated pages or logos, or not at all, and Swiftly's three grocers carry only a page timestamp . These counts are documented floors: no publisher series sizes the enablement market .
What the instruments cover
Four enablement platforms hold an assurance instrument, and sixteen hold none on the lists of MRC, the Alliance for Audited Media (AAM) or IAB Europe . MRC's letter of 15 March 2024 grants initial accreditation to "Criteo's Retail Media Platform (RMP)" for display impressions and clicks, gross and net of general invalid traffic, for on-site sponsored, commerce display and standard banner ads; it accredits no outcome metric and does not cite the IAB/MRC Retail Media Measurement Guidelines . Only Criteo's release maps the platform to Commerce Yield , whose methodology note puts off-site inventory outside the scope . Criteo's 10-K goes further than the letter, saying its accredited technology lets clients measure against product sales at item level .
Instacart's letter of 1 June 2026 continues accreditation of display and video impressions, clicks, click-through rate and viewability "across Instacart Ads and Carrot Ads" for placements on retailer properties reached through Storefront, Storefront Pro and Connect API integrations; MRC granted the Carrot Ads scope in October 2025 . Instacart also offers Carrot Ads on retailers' own storefronts and on Caper carts in stores, which the letter does not name, and Maplebear decided on 25 October 2026 to wind down Caper Cart operations, substantially by mid-2027 .
Kevel's pre-audit names no retail environment, and on 3 November Kevel held no letter . IAB Europe's certificate for Moloco, created 7 October 2026 and valid to 1 October 2028, certifies the platform's capability to support measurement and reporting for on-site display and sponsored product ads, and names no standards version, metric list, retailer or geography ; IAB Europe's compliance table makes incrementality optional . Epsilon Retail Media is on no list under its own name, and MRC's correlated-outcomes entry for Epsilon maps, by this paper's inference, to another Epsilon business (chapter 7) .
None of these instruments covers an attributed sales figure, and H2, whose population includes every enablement platform, is unresolved only because the Epsilon entry is mapped by inference . Accreditation does not travel down the stack: Albertsons Media Collective, whose sponsored products Criteo powers, is on no MRC, AAM or IAB Europe list . The SOC 2 security attestations Vantage, Osmos and Attain cite say nothing about measurement .
The optimisers relay the networks' numbers
Flywheel Commerce Cloud and Perpetua, two of the eleven optimisers, are Omnicom businesses, Perpetua under Flywheel . The author's past employer, Hearts & Science, was an Omnicom agency; this disclosure is voluntary .
An optimiser's outcome number is the network's, carried through an application programming interface (API) or a login. Of 47 product-by-network feeds the optimisers name, 5 document the feed's origin and at least two of its granularity, constraints and limitations, 5 the origin and one more, 5 the origin alone, and 32 nothing beyond the network's name . Intentwise, Perpetua and Xnurta each document API connections to two networks ; Pacvue's documentation could not be read, because its site refused the fetch tool and its help centre's certificate had expired .
The documentation that exists shows the numbers moving. Perpetua says its Walmart data lag Walmart's by one to two days and that "Conversions may be added retroactively" as purchases arrive inside Walmart's windows . Intentwise applies a 14-day window to Walmart in-store and online figures while, in its words, Walmart experiments with 3, 14 and 30 days , and shows Amazon's corrections as negative hourly values . Skai offers to bridge Instacart's, Walmart's and Amazon's attribution models without stating the method . Inference: two dashboards showing one campaign on different dates or windows can disagree without either being wrong.
Criteo Commerce Max reports one feed, Criteo's own retailer network, and documents it most fully: windows set per campaign with no default, one interaction credited per sale with clicks first, deduplication between on-site and off-site, deterministic matching and in-store sales from some retailers , and a new-to-brand lookback of 180 days . Its dashboard figures "can change and reflect new events, even if your campaign has stopped" , and since February 2026 off-site attribution reporting waits 72 hours instead of 9 .
No optimiser documents incrementality testing with an assignment, a counterfactual method, a minimum sample or a significance rule . Flywheel offers to "Measure causality by comparing sales when your ads are displayed versus when they're not", with no holdout or assignment stated . Stackline claims to read the net new customer impact of every impression without a method , and CommerceIQ optimises to an incremental ROAS it does not define .
What the measurers disclose
The author works at Samba TV. Samba announced a TV-to-purchase measurement arrangement with Catalina in August 2020 and an integration with Affinity Solutions' card data in October 2023; neither describes measuring a retail network's media, and Catalina itself sells in-store and digital media .
None of the seven cohort E services documents random assignment, a minimum sample or a significance rule for its standard design . Circana's Household Lift offers "household-level A/B testing" and controlled household tests with no allocation rule, minimum sample or significance rule , and Market Lift a matching process on point-of-sale data whose variables are unstated . NCS Sales Effect, which Circana sells to brands, publishers and retail media networks, names no design beyond machine learning .
Nielsen describes Sales Lift in one sentence, an "industry-standard control/exposed methodology" on retailer-verified purchase data, and the same page calls the output the sales attributed to marketing . Numerator compares exposed and non-exposed consumers on its receipt panel without saying how the unexposed group is built . Attain sells Sales Attribution, with no counterfactual, beside an Incrementality product on holdout methods it does not describe . Incremental decomposes sales with a causal model retrained daily and runs no A/B tests itself, calibrating on tests run on retail platforms . Catalina's HUB360 attributes media to purchases through its pixel and in-store integration and lists an incrementality report without describing it .
The only measurement services screened that state random assignment are geography-level experiment services the register lists outside cohort E, because an aggregate geographic test has no attribution window or household panel to score (chapter 1). Haus's GeoLift assigns regions by random stratified sampling , and LiftLab offers stratified random sampling as one option .
Few providers say which networks they measure. Circana names one with the service attached, the Ibotta Performance Network, a commerce media network whose campaigns get Circana Household Lift studies under a joint release of 30 September 2025 . Its DoorDash partnership of 2 September 2026, third-party validation of DoorDash's own measurement, names no Circana service, and its $0.59 of incremental sales per household is presented as early results with no period, so it is not counted for Media Lift . Nielsen, Attain, Catalina and Incremental publish no list of networks measured on the pages opened .
Sellers are often the customers. Numerator pitches its lift studies to media companies as proof of their advertisers' results, so a study's commissioning party may be the seller . Attain says it has "No media inventory" while serving retailers and media platforms ; Incremental says it operates independently of media sellers and calls itself a platform for them ; and Catalina's site sells display, out-of-home, in-store print and shoppable media beside its measurement . Networks cite Circana for their own results without a design or a payer (chapter 8) . No cohort D or E service holds an instrument on the MRC digital, AAM or IAB Europe lists read on 3 November 2026 .
What it means for the networks built on them
A network's number is partly made by software it licenses. The platform supplies the window, the crediting rule, the match types and the halo options, and its documentation shows the menu, not the choice . A buyer comparing networks on one platform cannot assume they report on the same terms, and a buyer comparing one network's figure across dates should know that a setting can change for the whole campaign after the fact . Assurance attaches to the scope a letter names: Criteo's to three on-site formats, leaving out the off-site inventory Commerce Yield also sells, and Instacart's to Carrot Ads placements through named integrations .
The stack also moves. Gopuff launched off-site media on CitrusAd in 2022 and announced in September 2026 that Carrot Ads will power its US storefront advertising . Microsoft said in July 2024 that Criteo would be its preferred partner for its retailer clients' on-site needs, without naming PromoteIQ, its own platform . DoorDash's agreement to acquire Instacart, expected to close at the end of 2027 or in the first half of 2028, would put Carrot Ads under the owner of DoorDash Ads . Criteo attributed a US$21 million retail media headwind in the second quarter of 2026 to scope changes with two clients it has not named .
Change governance is the thinnest part of the record. No governance cell in the three cohorts is scored, because no screen of a service's documentation and terms was completed (chapter 12). The items found sit beside the cells: Carrot Ads keeps a dated changelog that on 10 October 2025 clarified how impressions are counted , Criteo describes a backward-looking correction route for invalid traffic , and Commerce Max dates its reporting changes in monthly release notes .
Chapter 12 scores these services with cohort anchors. Cohort C has 9 band-eligible rows and its bands are reported there; cohorts D and E have 2 and 2 against a minimum of 6, too few for bands. For a buyer, the questions follow from the record: which platform runs the network, which settings the network chose, which surfaces the platform's instrument names and, where lift is offered, who designed it, who assigned the groups and who paid (chapter 13).
Chapter 10
Who may use the receipt
The US rules read for this paper govern purchase and loyalty records in advertising mainly through opt-outs of sale, sharing and targeted advertising, and through consent, or in Maryland strict necessity, once a purchase reveals health. The federal record that reaches retail shopper data directly is about prices: Federal Trade Commission (FTC) study orders of 2024 and of 28 October 2026, the second naming Instacart, Target and Walmart, and a proposed policy statement. In the networks' documentation read on 3 November 2026, consent is stated at company level, and no shopper notice names the retail media service or an independent privacy audit.
California treats disclosure for cross-context behavioural advertising as sharing that a shopper can stop, paid or not . Washington counts the purchase of prescribed medication as consumer health data, collected only with consent or as needed for a requested product, and shared only with a separate consent or to that same extent . FTC staff's 2025 "initial findings" on surveillance pricing rested on documents from six intermediary firms ; in October 2026 a differently composed Commission ordered nine companies, Instacart, Target and Walmart among them, to report, saying it had learned nothing about actual pricing decisions from the first round . Instacart's FTC order of 13 January 2026 concerns fees, subscription charges and delivery claims . California's data-broker registry lists Criteo, Circana, LiveRamp and Walmart's VIZIO unit but none of the profiled retailers . The searches logged found no order, settlement, court finding or breach notice about a network's use of shopper data for retail media .
Retail media sells the purchase record; privacy law decides who may use it, on whose consent. Albertsons Media Collective wrote in January 2025 that "with increased privacy regulations, there is diminishing ability to match impression data to sales data at an individual level at scale" . This chapter reads the US rules on purchase, loyalty and payment data in ad targeting and measurement, and the networks' privacy documents against the checklist of dimension 12: consent basis, partner-matching notice, sensitive purchase categories, state-law handling, data-broker registration and independent assurance. Adverse events are recorded apart, and chapter 12 applies the ladder. A law's requirement is not evidence that any network breaks it, and nothing here assesses compliance.
Opt-outs for the record, consent for health
California's Consumer Privacy Act (CCPA), as in force on 1 January 2026, lets a consumer opt out of the sale or sharing of personal information, and defines sharing as disclosure "for cross-context behavioral advertising, whether or not for monetary or other valuable consideration" . Its regulations make an opt-out preference signal such as Global Privacy Control (GPC) a valid opt-out, and consumers may limit the use of sensitive information, which includes health information, except where it is processed without the purpose of inferring characteristics .
Texas requires consent to process sensitive data, including data revealing a health diagnosis, and a posted notice from any business that sells it; its definition of targeted advertising excludes ads based on activity on a controller's own sites or apps, and processing solely to measure or report ad performance, reach or frequency . That first exclusion covers on-site activity and does not address in-store purchase data used for ads elsewhere . Colorado's Attorney General says controllers need consent before processing sensitive data, and before using bona fide loyalty-programme data for targeted advertising or sale, which the page calls "a secondary use" . The verifier could not reopen that page; the saved copy matches word for word .
Minnesota (Minn. Stat. §§ 325M.16 and 325M.17, effective 31 July 2025) bars processing sensitive data without consent, and targeted advertising to, or sale of the data of, consumers known to be 13 to 16 without consent . Virginia requires consent for sensitive data and limits collection to what is adequate, relevant and reasonably necessary . Connecticut's act was rewritten from 1 July 2026 by Public Act 25-113: sensitive data, which includes a health condition, diagnosis, disability or treatment and consumer health data, may be processed only where reasonably necessary for the disclosed purpose and with consent, and sold only with consent; targeted advertising to, or sale of the data of, consumers known or wilfully disregarded to be 13 to 17 is barred; and collection is limited to what is reasonably necessary and proportionate . Oregon's Department of Justice says its act requires consent for any sensitive data, health conditions included, and from 1 January 2026 requires controllers to accept universal opt-out mechanisms .
Maryland is the outlier: its Online Data Privacy Act, effective 1 October 2025, allows collecting, processing or sharing sensitive data, consumer health data included, only where strictly necessary for a product or service the consumer requested; the enrolled text strikes the words that would have allowed it with consent . The act also bars selling sensitive data, limits collection to what is reasonably necessary and proportionate, and bars targeted advertising and sale where the controller knew or should have known the consumer is under 18; Section 2 bars applying § 14-4612 as enacted to processing before 1 April 2026 . So consent, the gate for sensitive data in Colorado, Texas and Minnesota, would not on its own satisfy Maryland if its definition reaches health-related purchases, which this paper does not decide .
Washington's My Health My Data Act names purchases: consumer health data includes "Use or purchase of prescribed medication" and health data inferred from other information, including by algorithms or machine learning . It may be collected only with consent for a specified purpose or to the extent needed to provide a product or service the consumer requested, shared only with a separate consent or to that same extent, and sold only with a valid authorisation; accepting general terms of use is not consent . These duties apply from 31 March 2024, and a separate section, undated in the text read, bars geofences within 2,000 feet of in-person health-care providers used to track people seeking care, collect their health data or send them notifications, messages or advertisements . Whether a grocer's record of a health-product purchase is consumer health data in a given case is not decided here . For pharmacies, the Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule already requires an individual's authorisation for marketing uses of protected health information, with narrow exceptions, and for any sale of it .
Pricing, not advertising: the FTC and the states
The federal record that touches retail purchase data is a pricing study: on 23 July 2024 the FTC voted 5-0 to send Section 6(b) study orders to Mastercard, Revionics, Bloomreach, JPMorgan Chase, Task Software, PROS, Accenture and McKinsey & Co. . JPMorgan Chase owns Chase Media Solutions, a commerce media row in this paper, but the orders concerned pricing products . They defined those products by reference to purchase history, among other variables . On 17 January 2025 staff released research summaries of their "initial findings", allowed by a 3-2 vote over the dissents of Commissioners Ferguson and Holyoak, resting on six respondents' documents (not JPMorgan Chase's) and disclaiming any assessment of illegal conduct . The release calls the examples hypothetical, yet some footnotes report what respondents' documents indicated: one says targeted customers can include the coupon-inclined and those with perceived loyalty or willingness to pay . No company or retail media network is named .
On 19 August 2026 the Commission voted 2-0 to propose an enforcement policy statement: where consumers reasonably expect prices not to vary with their personal data, a business that personalises prices should disclose that, the basis for it and the types of data used, and failing to do so is likely unfair or deceptive under Section 5 of the FTC Act . The statement binds neither the FTC nor the public; comments closed on 25 September, and no final version was found by 3 November .
On 28 October 2026 the Commission (2-0) ordered Delta, United, JetBlue, Lyft, Uber, DoorDash, Instacart, Target and Walmart to report whether pricing "varies from one person to the next based on factors such as participation in loyalty programs or analysis of personal data by superintelligence algorithms" . The template order, under the 2024 matter number P246202, covers 1 January 2023 onward and defines personal information to include purchases and reactions to advertising . The release says the agency learned nothing about actual pricing decisions from the 2024 study, which it says examined only middlemen consultancies . Such orders are study demands the Commission can enforce in federal court, not complaints . The recipients operate this paper's Instacart, Carrot Ads, Roundel, Walmart Connect, Sam's Club Connect, DoorDash Ads, Uber, Lyft Ads and Kinective Media rows; the orders name no media service .
Two states opened pricing inquiries: New York's Attorney General, under the Algorithmic Pricing Disclosure Act in effect since 10 November 2025, wrote to Instacart on 8 January 2026 demanding information on its price-setting experiments and on how consumer data may be used for price changes; the release says Instacart had announced it was ending its item price tests, and the letter itself was not opened . On 27 January 2026 California's Attorney General announced a sweep of businesses' use of personal information to set individualised prices, with letters to unnamed retail, grocery and hotel businesses . Neither inquiry concerns shopper data used for advertising .
Instacart's FTC order concerns fees: the FTC sued Maplebear Inc., doing business as Instacart, on 18 December 2025 (N.D. Cal., No. 4:25-cv-10783-JST), alleging false "free delivery" and satisfaction-guarantee claims, undisclosed service fees and Instacart+ charges without express informed consent; Judge Jon S. Tigar entered the stipulated order on 13 January 2026, with a US$60 million judgment and a ten-year term . The FTC's case page still showed the matter as pending when opened .
The FTC's first Health Breach Notification Rule action, announced on 1 February 2023 against GoodRx, concerned its alleged disclosure of users' prescription and health information to advertising platforms including Criteo, with a US$1.5 million civil penalty and a proposed order; Criteo was a named recipient, not a respondent, and the release makes no finding about it . Criteo's 2025 annual report adds that the court refused in January 2026 to approve its US$7.0 million settlement of Doe v. GoodRx Holdings, and that a putative class action against Criteo, CVS and others over sensitive health and personal information was dismissed as to Criteo, refiled, and in mediation after the plaintiff filed a stipulation to dismiss on 30 January 2026 . The filing ties neither matter to Criteo's retail media products . The FTC's two pending suits against Amazon (No. 2:23-cv-01495, with 17 states, and No. 2:26-cv-03097, with 22) allege monopolisation and overcharging in sponsored-ads auctions, and neither concerns shopper data .
What the networks' notices say
Apart from one Kroger Precision Marketing FAQ answer, every network's privacy record read is company-level: the shopper notices of Amazon, Walmart, Target and Instacart name no retail media service, and Walmart's, Target's and Instacart's never mention a clean room . Kroger's and Albertsons' shopper notices refused automated access , and the network-branded policies of Walmart Connect, Sam's Club's ad platform and Albertsons Media Collective do not cover shopper data from the retailers' own sites . Amazon's pages refused automated fetches: the passages used come from the researcher's saved copies, which the verifier matched word for word but could not reopen, and the E1 stream coded the same pages inaccessible .
Where a consent basis is stated, it is general: Amazon treats use of its services as consent to its notice, and says it does not share "specific shopping actions like purchases, product views, or searches with ad companies", though it may share an advertising identifier, such as a code derived by irreversible cryptography from an email address . Walmart lists targeted advertising among its uses and seeks consent for sensitive information where the law requires it . Target's Colorado section says it uses personal information for targeted advertising and sells it in that law's sense . Instacart discloses data to advertising partners to help serve and measure its own and other companies' ads, and lists loyalty programme numbers among the identifiers it may have sold or shared for targeted advertising .
Walmart's notice discloses data to ad-technology vendors including identity management companies and to measurement vendors that provide aggregate reporting, and keeps VIZIO OS television data apart from Walmart account data "unless and until you opt in to link and combine it", with identifiers encrypted or hashed . On 3 April 2026 Walmart Connect said its new identity graph, joining in-store transactions, online activity and media engagement, would extend to connected TV through VIZIO in the future; it stated no match method, and neither document says how the opt-in applies to that extension .
Other joins are described to advertisers in trade copy: in a LiveRamp release of 23 October 2025, Roundel's vice president of product management says clean-room measurement with LiveRamp combines Target's first-party guest insights with Meta's ads data "in a privacy-safe way" . CVS Media Exchange's 2024 announcement says ExtraCare loyalty data are pseudonymised before entering a LiveRamp clean room that reveals their overlap with Pinterest's data, without naming the keys . Kroger Precision Marketing's FAQ cites hashed identifiers and clean rooms where appropriate . Criteo's privacy policy, which never mentions retail media, describes matching purchases, online and in shops, with ad clicks and displays, and receiving loyalty card numbers from advertisers such as brands and retailers ; Commerce Yield's help centre says a hashed email enables cross-retailer identity resolution across Criteo's network . Amazon Marketing Cloud accepts only pseudonymised information, returns only aggregated outputs and keeps user, household and IP identifiers out of final query results .
Sensitive purchase categories are handled, where at all, in company health policies: in states with consumer-health-data laws, Target counts the purchase of health-related items as health data, used for marketing only with consent and not shared for cross-site advertising without it . Amazon's Washington and Nevada health disclosure lists no advertising use, but its categories do not include ordinary purchases of health products, and Amazon's advertising policies could not be opened . Instacart treats prescription-delivery data under HIPAA . Walmart refers pharmacy data to separate notices , while Walmart Connect's ad policy lists Crossix and IQVIA as approved partners for "Rx sales lift measurement" by server-to-server integration, with data scope and terms unstated . Albertsons sells deli and pharmacy screen placements with no stated handling of health data , and no sensitive-category rule was found on Kroger Precision Marketing's pages .
Among the enablers and measurers, Criteo says it does not process sensitive personal information . Circana's group policy collects health and fitness data from panelists, keys media-product data to "a unique ID that is made available to all data contributors", and allows sale or sharing of some CCPA sensitive information with media-product partners . Cardlytics says banks send it de-identified data and that it has no insight into sensitive categories such as healthcare; its advertiser redemption feed still carries the card's last four digits where available, which the verifier recorded as a tension with Cardlytics' own FAQ, not a breach .
Contracts also fence shopper-derived data off from buyers: Amazon's Advertising Agreement makes all Ad Services Data Amazon's exclusive property, usable by the buyer only to run campaigns, evaluate them internally and report to the advertiser, and the performance data it may pass on include whether a user viewed or clicked an ad . Cardlytics' Cost per Served Sale terms confine transaction-level or individual activation data to internal evaluation and require its deletion within 45 days . Macy's Media Network forbids passing click and view data to any ad network, exchange or data broker . Chapter 5 reads what such clauses leave a buyer able to check.
Walmart, Target, Instacart and Amazon say they honour GPC, Target as an opt-out of both sale and targeted advertising . Amazon reports about 349.5 million cross-context behavioural advertising opt-out requests in 2025, most from visits with GPC already enabled, and opts under-18s out of such ads by default; these are Amazon's own counts of requests . A "no sale" statement does not say that shopper data stay with the operator: Amazon reports no sale in the prior twelve months and Albertsons Media Collective none under the CCPA, while both process or share data for targeted or cross-context behavioural advertising .
Registries, assurance and adverse events
California's data-broker file, on a page updated on 30 October 2026, lists among others Catalina, Bridg (still "a division of Cardlytics, Inc." after its sale to PAR Technology on 24 March 2026), Circana, Criteo Corp., Fetch, LiveRamp, Epsilon Data Management and VIZIO Services, a Walmart subsidiary . None of the profiled retailers appears by name (Kroger and 84.51°, Albertsons, Walmart, Amazon, Instacart, Target), nor Cardlytics itself, Attain, Numerator, Ibotta or NIQ; the Delete Act covers businesses selling data on consumers with whom they have no direct relationship, so a retailer's absence says nothing about its practices . NCSolutions, merged into Circana on 30 September 2025, sits in the 2025 file under a misspelt name that an exact-name search misses . Criteo says it is also registered in Texas; the Texas, Oregon and Vermont registries could not be read .
None of the privacy documents read for the profiled networks and providers names an independent privacy certification or audit covering a retail media product, and Walmart's and Target's adherence to the Digital Advertising Alliance's principles is self-attested . Kroger Precision Marketing's FAQ claims certification by the IAB and a "Merchant Risk Council (MRC)"; neither is a privacy audit, and the Media Rating Council lists Kroger's outcome metrics only as in pre-audit .
In the searches logged by 3 November 2026, no FTC or state attorney-general order, settlement or court finding, and no breach notice, concerned a register network's use of shopper data for retail media; the official matters touching register companies are inquiries, Criteo's place in the GoodRx matter and Instacart's fee settlement . The FTC's case list by company, state breach databases and SEC cybersecurity filings were not screened .
One matter bears on partner matching: on 19 October 2026 the FTC announced a complaint and proposed consent order under which Publicis Groupe, which also owns Epsilon, could complete its US$2.2 billion purchase of LiveRamp subject to five years of non-discriminatory access and firewalls against Publicis's access to LiveRamp customer data , with an independent auditor for two years after closing . LiveRamp runs the Roundel and CVS clean rooms described above and is a registered California data broker . The FTC's release treated the deal as pending; Publicis Groupe announced the closing on 2 November 2026 . Whether the FTC has made the order final was not read by 3 November and is rechecked on 20 November.
The rules read
The instrument lists the laws, regulations and regulator actions read for this chapter, each as its primary text was read; New York's row rests on the Attorney General's release and Oregon's on the Department of Justice's FAQ, because neither statute was opened . None of it is legal advice.
Rules and governance instruments
Official sources only for the rule text; industry guidance is marked as practice, not law. Nothing here is legal advice.
Machine-readable copy: data/rules.csv.
| Jurisdiction | Instrument | What it requires | Effective | Scope | Law or practice | Src |
|---|---|---|---|---|---|---|
| California | CCPA, Civ. Code § 1798.120 and § 1798.140(ad), (ah) (text effective 1 Jan 2026) | Right to opt out of sale or sharing; 'share' = disclosure for cross-context behavioral advertising whether or not for consideration; no sale/sharing of under-16s' data without affirmative authorization | consolidated text effective 2026-01-01 (right pre-dates it) | businesses that sell or share personal information | statute | |
| California | CCPA, Civ. Code § 1798.121 and § 1798.140(ae); Regs § 7027 | Right to limit use of sensitive personal information (includes 'Personal information collected and analyzed concerning a consumer's health') to necessary uses; 'Limit the Use of My Sensitive Personal Information' link or Alternative Opt-out Link; not applicable where SPI is processed without the purpose of inferring characteristics | regulations effective 2026-01-01 | businesses using SPI beyond necessary purposes | statute and regulation | |
| California | CCPA Regulations § 7025 | Opt-out preference signals (e.g. Global Privacy Control) must be treated as valid requests to opt out of sale/sharing for the browser or device and associated profiles, including pseudonymous profiles | 2026-01-01 (consolidated text) | businesses that sell or share and collect online | regulation | |
| California | Delete Act data broker registration (CPPA registry) | Data brokers register with CalPrivacy annually between January 1-31; registry published with collection flags and request metrics | annual (page last updated 2026-10-30) | data brokers (businesses selling information about consumers with whom they have no direct relationship) | registration | |
| Washington | My Health My Data Act, RCW 19.373 | Consent to collect consumer health data for a specified purpose, or collection to the extent necessary to provide a product or service the consumer requested; a separate consent to share it, or sharing to that same extent; valid authorization to sell it; consumer health data includes 'Use or purchase of prescribed medication' and data inferred from non-health information; acceptance of general terms of use is not consent; geofences within 2,000 feet of in-person health-care providers barred when used to track consumers seeking care, collect their health data, or send them notifications, messages or advertisements; violations enforceable under the Consumer Protection Act | 2024-03-31 for RCW 19.373.020 to .070 (small businesses 2024-06-30); the geofence ban (RCW 19.373.080) carries no date in the text read | regulated entities conducting business in Washington | statute | |
| Colorado | Colorado Privacy Act (C.R.S. 6-1-1301 et seq.) and 4 CCR 904-3, per the Attorney General's page | Affirmative consent before collecting and processing sensitive data; opt-out of sale, targeted advertising and profiling; universal opt-out mechanisms must be accepted from 1 July 2024; AG FAQ: sale or targeted-advertising use of bona fide loyalty-program data is a secondary use requiring consent | 2023-07-01; UOOM 2024-07-01 | controllers meeting the CPA thresholds | statute, rules and regulator FAQ | |
| Texas | Texas Data Privacy and Security Act, Bus. & Com. Code ch. 541 (HB 4) | Consent before processing sensitive data ('mental or physical health diagnosis' included); 'NOTICE: We may sell your sensitive personal data.' where sensitive data is sold; targeted advertising excludes first-party on-site ads and processing solely for measurement or reporting | 2024-07-01 | controllers conducting business in Texas (small-business exclusion in § 541.002) | statute | |
| Texas | Data broker registration, Bus. & Com. Code ch. 510 (formerly 509), Secretary of State | Data brokers register with the Secretary of State with a $300 fee; certificate effective one year, renewable; searchable registry maintained by the SOS | per SOS page (statute date not on page) | data brokers | registration | |
| Minnesota | Minnesota Consumer Data Privacy Act, Minn. Stat. ch. 325M, §§ 325M.16-325M.17 (2024 ch. 121 art. 5) | Consent before processing sensitive data; consent for targeted advertising or sale where the consumer is known to be 13-16; small businesses must not sell sensitive data without prior consent; privacy notice must disclose sale and targeted advertising | 2025-07-31 | controllers meeting the act's thresholds | statute | |
| Maryland | Maryland Online Data Privacy Act of 2024 (ch. 455, SB 541), § 14-4607 | No collection, processing or sharing of sensitive data (includes consumer health data) except where strictly necessary to provide or maintain a specific product or service the consumer requested (consent clause struck in the enrolled text); no sale of sensitive data; collection limited to what is reasonably necessary and proportionate; no targeted advertising, and no sale of personal data, where the controller knew or should have known the consumer is under 18 (the sale consent clause is struck too); opt-out preference signals by 1 Oct 2025 | 2025-10-01; Section 2: § 14-4612 as enacted may not be applied to processing activities before 2026-04-01 | controllers meeting the act's thresholds | statute | |
| US (federal) | HIPAA Privacy Rule, 45 CFR 164.508(a)(3)-(4) | Covered entities need the individual's authorization for marketing uses or disclosures of PHI (exceptions: face-to-face communications, nominal gifts), stating any third-party financial remuneration; authorization required for any sale of PHI, stating remuneration | CFR 2024 edition | HIPAA covered entities (e.g. pharmacies) and business associates | regulation | |
| US (federal) | FTC Health Breach Notification Rule enforcement (GoodRx) | First HBNR action: disclosure of prescription and health data to advertising platforms (Facebook, Google, Criteo) treated as breaches; $1.5m civil penalty; proposed order bans sharing health data for advertising; referred to DOJ 4-0; court approval required | 2023-02-01 (proposed) | vendors of personal health records (GoodRx); Criteo a named recipient | enforcement action | |
| US (federal) | FTC Section 6(b) surveillance pricing study | Orders to eight companies (5-0) for information on pricing solutions using location, purchase history, browsing and demographics; staff perspective of 'initial findings' (3-2; 'not an exhaustive declaration'): hypothetical examples plus footnotes reporting what respondent documents indicated; no assessment of illegal conduct; RFI with comments due 17 April 2025 | 2024-07-23 orders; 2025-01-17 staff perspective | eight recipients including JPMorgan Chase; study, not enforcement; the FTC's 28 October 2026 release says the 2024 study 'examined only middlemen consultancies' | inquiry | |
| Virginia | Consumer Data Protection Act, Va. Code § 59.1-578 | Consent before processing sensitive data; collection limited to what is adequate, relevant and reasonably necessary; clear disclosure of sale and targeted-advertising processing and of the opt-out | history 2021, Sp. Sess. I; amended 2024 and 2026 (effective date not on the page) | controllers meeting the act's thresholds | statute | |
| Connecticut | Connecticut Data Privacy Act, Gen. Stat. §§ 42-515(39), 42-520(a)(1), 42-526, as amended by P.A. 25-113 | Sensitive data (includes 'a mental or physical health condition, diagnosis, disability or treatment' and consumer health data) processed only where reasonably necessary for the disclosed purpose and with consent (§ 42-520(a)(1)(D)); no sale of sensitive data without consent; no targeted advertising or sale for consumers known (or wilfully disregarded) to be 13-17; collection limited to what is reasonably necessary and proportionate; no sale of consumer health data without consent (§ 42-526, 2025 revision; any 2025 amendment not read) | 2023-07-01 (P.A. 22-15); amended 2026-07-01 (P.A. 25-113) | controllers meeting the act's thresholds; § 42-526 per its own scope | statute | |
| US (federal) | FTC proposed enforcement policy statement on personalized pricing (Section 5) | Where consumers expect non-personalized prices, clear and conspicuous disclosure that a price is personalized, its basis and the types of data used; failure 'likely' unfair or deceptive; non-binding | proposed 2026-08-19; comments closed 2026-09-25; final not found by 2026-11-03 | businesses setting prices with personal data | proposed policy statement | |
| US (federal) | FTC Section 6(b) orders on personal data in pricing and compensation (Matter P246202) | Special reports on methodologies using personal information (incl. purchases, browsing, 'reactions to advertising', loyalty participation) to set prices or compensation since 1 January 2023, and on disclosures to consumers | orders 2026-10-28; responses due 45 days after service | nine recipients incl. Instacart, Target, Walmart, DoorDash, Uber, Lyft, United; study, not enforcement | inquiry | |
| Oregon | Oregon Consumer Privacy Act (ORS 646A.570 et seq.), per the Oregon DOJ's business FAQ | Consent to collect, store or otherwise process all categories of sensitive data (includes 'mental or physical health conditions or diagnoses'); from 1 January 2026, universal opt-out mechanisms must be accepted and no sale of precise geolocation data or of data of consumers under 16 | 2024-07-01; opt-out mechanisms and the two sale bans 2026-01-01 | controllers subject to the OCPA (thresholds not quoted; statute text not opened) | statute (enforcer's FAQ) | |
| New York | Algorithmic Pricing Disclosure Act (per an AG release; statute not opened) | Disclosure duty for prices set with personal data (verifier's summary of the AG release of 8 January 2026); the AG's letter asks Instacart about its price experiments and use of consumer data | 2025-11-10 (per the NY Attorney General) | businesses setting prices with consumers' personal data | statute (AG release) |
Chapter 11
Agents and protocols: what the advertiser gets back
Coded from the operators' pages as read on 3 November 2026, the register holds eleven sponsored formats on agent surfaces in the US. The two stated generally available, Amazon's Sponsored Products and Sponsored Brands prompts, come with Amazon's own report of attributed sales, so H6a is supported. None of the eleven documents explicit general availability of third-party attribution, causal measurement or ad-quality verification, and those functions are mostly undocumented rather than shown absent, so H6b to H6d are unresolved. No protocol read goes beyond schema representation for a sponsored placement and its outcome.
The frozen register holds eleven records, one per agent surface, sponsored format and US market, none excluded . Two are explicitly generally available, five available with no stage stated, three pilots under way and one only announced . Amazon's Prompts report gives impressions, clicks, spend, sales, ACOS, ROAS and 7-day orders and units, all computed by the seller of the media . Both generally available records report, so H6a is supported . Only ChatGPT's standard ads name attribution partners, with no stage stated; OpenAI calls its incrementality work early and its brand-suitability work pilots . Every other record is unknown on those axes, so H6b, H6c and H6d are unresolved . AdCP's schema represents a sponsored placement and its attributed outcomes, UCP carries an ad attribution context, and ACP, AP2 and AAMP carry neither . AdCP's own documentation keeps that format experimental until at least two retail media networks ship evidence .
The test
H6 asks whether an advertiser who buys a sponsored placement inside an AI agent receives anything beyond the operator's own report. One record per agent surface, sponsored format and geography (US) enters only where the operator documents a shopping use for that format; the eleven were frozen on 3 November 2026, before extraction (chapter 14). Each is coded on availability (announced; pilot or test started; available, stage unstated; explicitly generally available; unknown) and on four further axes: operator reporting, third-party attribution, causal outcome measurement and ad-quality verification by a named third party, each none, pilot, available with stage unstated, generally available or unknown. An announced test stays announced until a source shows it started, and a platform-wide statement is never promoted to a format.
H6a is supported if every record stated generally available (the set A) has operator reporting, contradicted if any documents none, and unresolved if A is empty or holds a pilot or unknown reporting state. For H6b to H6d, with N records, G generally available on the axis and U unknown or stage unstated, a proposition is supported when (G + U) ÷ N is below 0.5, contradicted when G ÷ N is at least 0.5, and unresolved otherwise.
The rival line is the author's: his essay Measurement for Agentic Commerce (13 July 2026) speculated that networks would ship "agentic iROAS" (incremental return on ad spend) before any neutral standard exists . It is a hypothesis, not evidence. The series' shopping paper, Who Does the Shopping? (cutoff 9 October 2026), owns how agents find and buy. This chapter covers what an advertiser receives for sponsored placement inside them and notes where its primary reads supersede that paper's records.
Eleven records
Amazon holds three records: Sponsored Products and Sponsored Brands prompts on its store, where they appear in shopping results and product detail pages and may open a dialog in Rufus, and both formats on Alexa for Shopping, the assistant's name after 13 May 2026 . OpenAI holds three: standard ads at the bottom of ChatGPT answers, a visual format during image generation, and Sponsored Agents . Amazon-sold advertising in ChatGPT entered on Amazon's page of 10 September 2026 . Google holds ads and sponsored product list items in AI Overviews and AI Mode, and Direct Offers in AI Mode . Product listing ads from Kroger Precision Marketing (KPM) in Kroger's AI Shopping Assistant are the only record from another retailer, and Microsoft's shopping ads in Copilot the eleventh . Perplexity has none: the register excluded it on its 2024 statement that its shopping cards are unsponsored, a page this review could not open .
Two records are explicitly generally available, five available with no stage stated, three pilots under way and one only announced . Operators rarely use the coding's words. OpenAI called its ChatGPT ads a pilot as late as 5 May 2026 and a launch from 31 August, and Google's help page says its AI ads are available in English in the United States and select English-speaking markets; neither uses a stage word . OpenAI's visual format, announced on 5 October 2026, was to begin testing later that month, and no OpenAI page read by 3 November says it started .
Two records rest on shopping use stated at platform level: ChatGPT's standard ads, and the Amazon-sold placement, whose page speaks of customers who "explore options, compare choices, and make decisions on ChatGPT" . That page says select US advertisers are testing in a pilot, ties the placement to Amazon's demand-side platform (DSP) only through a quotation and a link, and names one participant, the travel advertiser Delta Vacations . Microsoft's record is inferred from a quick-back comparison for shopping ads on global data and from Offer Highlights, described in April 2026 as available for retail use cases while also in live pilots; no Microsoft page read names the format with a stage or US availability .
Amazon's prompts: the seller's own report
Amazon's page of 10 March 2026 says the prompts entered open beta in November 2025 and became generally available in the US on 25 March 2026, billed within cost-per-click (CPC) bidding, with existing campaigns enrolled automatically . Its subtitle says both formats and their reporting are now generally available, and that sentence, not the planned date, puts the two records in A . The page supersedes the shopping paper's record of the prompts, read from search summaries .
The advertiser receives a Prompts report of prompt text, associated ad, impressions, clicks, click-through rate, cost per click, spend, sales, advertising cost of sales (ACOS), return on ad spend (ROAS) and 7-day orders and units, in the console and through the application programming interface (API) . Amazon computes each outcome from purchases in its own store, as the seller of the media, and describes no counterfactual . Its API counts prompt sales within 1, 7, 14 or 30 days of an ad click and splits the advertised stock-keeping unit (SKU) from others, and no column read says whether sales are gross or net of returns . The page's "7-day orders and units" is not reconciled with Amazon's documented 14-day default for vendors .
On Alexa for Shopping, Amazon's page of 11 June 2026 makes existing campaigns eligible and promises "closed-loop measurement that tracks impressions all the way to conversions", but lists no metric for the surface and no holdout or control . Its figures, that nearly 20% of shoppers who interact with a prompt continue the conversation about that brand and that adding prompts to a Sponsored Brands ad raises conversions by 6%, carry no design, period or sample .
"Closed loop" is Amazon's term, coded here as attribution with no counterfactual (chapter 1) . Neither Amazon page names an attribution partner, a verification vendor or a holdout for any of the three records, so those axes are unknown, not none . The Media Rating Council's (MRC) accreditation of Sponsored Products clicks, continued on 8 July 2026, covers clicks and invalid traffic, not outcomes, and does not say whether prompts are in its scope . Amazon documents a survey-based Brand Lift product for sponsored ads except Sponsored Products; under the coding rule a product-wide offer is not promoted to a prompt format .
ChatGPT: the partners are platform-wide
OpenAI's ads sit at the bottom of ChatGPT answers on the Free and Go plans and were tested in the US from 9 February 2026, when advertisers "only receive aggregate information about how their ads perform such as number of views or clicks" . By 5 May OpenAI had added a Conversions API and pixel, and its developer pages describe catalog ads reported by product . These conversions come from the advertiser's own tags, not a store's transaction log, and no post read states a window or model .
The standard ads are the only record with anything on the partner axes, all of it platform-wide. On 5 October 2026 OpenAI named ten attribution partners, among them AppsFlyer, DV Rockerbox and Northbeam, with no stage label . It said its "work on incrementality is still in its early stages" and that it is partnering with Haus, Measured and WorkMagic to explore geo-based experiments; the one result shown is a partner-reported lift figure with no design, sample or period . It is developing brand-suitability pilots with DoubleVerify and Integral Ad Science (IAS); their start is not stated, and viewability is not mentioned . None of this reaches OpenAI's other formats. Sponsored Agents, tested with select US advertisers since 16 September 2026, have no reporting or partner described, and neither has the Amazon-sold pilot .
On 29 September 2025 OpenAI called its product results "organic and unsponsored, ranked purely on relevance to the user" . When several merchants sell the same product, ChatGPT considers availability, price, quality, primary-seller status and whether Instant Checkout is enabled: one listed ranking factor in ChatGPT's implementation, not a tie-break and not a rule of the Agentic Commerce Protocol (ACP) . On 24 March 2026 OpenAI let merchants use their own checkout while it focused on discovery . The shopping paper held both facts at search level or as reporting; OpenAI's own posts, read for this paper, supersede those records, and the March post does not say every Instant Checkout integration ended .
Google, Kroger and Microsoft
Google's help page says metrics for ads in AI Overviews and AI Mode "are included in aggregate" in standard campaign and search-terms reports and does not say they can be reported separately; the conversions come from the advertiser's Google Ads tags . Its post of 28 October 2026 says click-through rates are already similar for sponsored and organic product list items in AI Mode, footnoted on the US copy to internal data from 20 August to 3 September 2026: an engagement comparison, not outcome reporting . No Google page read names an attribution partner, a verification vendor or a causal design . Direct Offers, a pilot since 11 January 2026 showing exclusive offers to shoppers in AI Mode, has no reporting described on the two Google posts read .
KPM put its Search & Browse product listing ads into Kroger's AI Shopping Assistant on 3 August 2026, where sponsored placements from active campaigns can appear, labelled and included automatically . Where Google's help page offers only aggregate figures, KPM says advertisers may analyse the assistant placements separately in its report builder, though it lists no metrics . Its network-level statement that its incrementality and "closed-loop" attribution methods compare exposed and unexposed groups is not tied to the placement, and its Trustworthy Accountability Group (TAG) Platinum certification covers the operator, not this placement . Microsoft's posts read describe no reporting for shopping ads in Copilot; its figure of 53% more purchases concerns journeys that include Copilot, not ads .
H6, coded by the rule
H6a is supported . A holds Amazon's two store prompt formats, both with operator reporting generally available in Amazon's words . No record documents the absence of reporting; four more have it with stage unstated and five are unknown, but records outside A do not decide H6a .
The verdict has one point of failure. Without the subtitle of Amazon's page, the general availability date is a plan stated in advance, A is empty and H6a is unresolved . Dropping Amazon's three records does the same; dropping any other operator, or the announced, Amazon-sold and inferred records together (N = 8), leaves it supported . So does the stricter screen the final review proposed, which reads the shopping and US conditions format by format and drops ChatGPT's standard ads, whose shopping use OpenAI states for the platform, the Amazon-sold ChatGPT ads, whose shopping wording is generic, and Copilot's shopping ads, offered in English-speaking markets with the US not named: N is again 8, and H6b to H6d stay unresolved . H6a shows only that the one operator selling a generally available sponsored format inside an agent reports on it itself.
H6b, H6c and H6d are unresolved . No record documents explicit general availability of third-party attribution, causal measurement or ad-quality verification, so G is 0 of 11 and none is contradicted, but nor are most records documented as none or pilot. On attribution, ChatGPT's standard ads are stage unstated and ten records unknown, so (G + U) ÷ N is 11 of 11; on the other two axes the standard ads are a pilot and ten records unknown, 10 of 11 . Extending OpenAI's platform-level pilots to its other ChatGPT records still leaves 7 of 11 in U on causal measurement and verification, and no sensitivity tried moves the three verdicts .
The gap is one of disclosure: on the pages read, no operator says whether anyone else can attribute, test or verify a given format, and none says no one can. Inference: Amazon reports outcomes from purchases in its own store, while OpenAI and Google report clicks and the conversions advertisers' own tags send back, which is where OpenAI's attribution partners connect .
Five protocols, one use case
The author co-leads the Signals & Measurement working group of AgenticAdvertising.org, which maintains AdCP, the Ad Context Protocol, one of the five protocols read here, and his firm, No Fluff Advisory, operates agents listed in that organisation's registry . Neither earns AdCP a state, no such agent is cited, and the protocol checklist was frozen before any protocol was read.
Shared rule 12 records five states per protocol at a pinned release: schema representation, which shows only that a thing can be expressed; an implemented connector; a demonstrated interoperability test; documented production use; and a demonstrated transaction. One use case was applied to all: representing a sponsored placement inside a shopping surface and reporting its outcome. A protocol's own task earns no state for it, and membership and announcements earn none (chapter 14). IAB Tech Lab's Agentic Advertising Management Protocols (AAMP) are read at 2.3 and the 3.0 draft, so the grid has six rows .
ACP, maintained by OpenAI and Stripe and in beta, is pinned at specification 2026-04-17 on commit 7fdd78df of 18 July 2026, with no releases or tags; its README describes "an interaction model and open standard for connecting buyers, their AI agents, and businesses to complete purchases seamlessly" . Nothing in the files read represents a sponsored placement or its outcome, and the flag in OpenAI's separate product-feed specification that marks catalog items for its own ads processing represents neither . Its checkout has operator-documented production use from 29 September 2025 (Etsy sellers) and 8 December 2025 (Instacart), with no independent volume, and no page read says which integrations remain live after 24 March 2026 .
The Universal Commerce Protocol (UCP), announced by Google on 11 January 2026 and co-developed, Google says, with partners including Shopify, Etsy, Wayfair, Target and Walmart, is pinned at release v2026-08-25 (commit 87dd4b1f) . It partly represents the use case: that release adds "platform-driven referral and ad attribution context", and no release uses advertising, retail media or measurement vocabulary . Its checkout was still coming soon in May 2026, and Microsoft's UCP-ready Merchant Center feeds, generally available in the US since 21 April 2026, are a feed format, not a placement . The shopping paper paired the release with a later main-branch commit, f2206e68 . The Agent Payments Protocol (AP2), at v0.2.0 of 28 April 2026, carries payment mandates, so every state is n/a .
AdCP is pinned at release v3.2.8 of 28 October 2026; its documentation introduction still names 3.2.1, and the shopping paper's 3.2.3 is superseded . Its schema can represent the use case: a retail_media channel, a canonical sponsored_placement format marked experimental at 3.1, and a sponsored-placement-type enum that the schema calls "seller-declared discovery metadata, not independent verification of inventory quality or delivery context" .
Its commerce-media page calls the outcome basis "closed-loop attribution". Retailers match exposure to transactions through loyalty cards, login state and point-of-sale records, products declare metrics from conversions and ROAS to incremental_sales_lift, with no design specified for the last, and the retailer platform, not a third party such as IAS or DoubleVerify, is named as commerce media's measurement owner . On this paper's reading, the one protocol that can carry the use case encodes what H6a found: the seller reports its own outcomes.
Beyond schema AdCP earns no state. Its retail media seller specialism is a preview at version 0.1.0 whose storyboard uses a fictional buyer and retailer, a conformance kit of the maintaining body, not a test between named parties . The sponsored_placement format stays experimental under a gate of format_schema evidence from at least two retail media networks, and no third-party connector, production use or transaction was found on the pages opened .
AAMP carries no sponsored placement. OpenProposal, the 3.0 core object, is a draft whose README says nothing in it is a ratified standard; its comment period was extended to 10 November 2026 on IAB Tech Lab's page while the README still shows 22 October . No retail, commerce, sponsored or shopping vocabulary appears in its README or release, and outcome reporting is left to existing standards. A keyword absence does not show that a generic proposal line cannot describe a sponsored placement, which was not tested . AAMP 2.3, released on 30 July 2026, has none of those terms on its release page .
Across the six releases, one schema represents the use case (AdCP), one partly (UCP), three do not (ACP and both AAMP releases) and one has nothing to represent (AP2); none reaches a connector, a test, production use or a transaction . Where a sponsored format touches a protocol on the pages read, the link is a flag, a feed format or a plan: OpenAI's catalog ads use its own feed specification outside the ACP repository, and Google says UCP-integrated brands will soon let shoppers buy from Direct Offers .
What would move the verdicts
H6a would fall to unresolved if Amazon qualified its general-availability statement, and would be contradicted only by a generally available format documented with no operator reporting . H6b to H6d need partner axes documented format by format: six of the eleven records as none or pilot for support, six as generally available for contradiction. A format-level page naming an attribution partner, a holdout or a verification vendor for Amazon's prompts, Google's AI ads or KPM's placements would move the counts fastest, as might OpenAI's help-centre pages, which refused the tools used here . A protocol would move past schema with a third-party connector, a test between named parties or a production buy that carries a sponsored placement and returns its outcome .
A buyer offered sponsored placement inside an agent can ask who computes the outcome, from whose data, over which window, whether the agent surface is reported apart, and whether anyone but the seller can attribute, test or verify it. Chapter 13 turns these into questions. The register below lists all eleven records with each state's date, observable and source.
AI and agentic capability map
Sponsored formats and capabilities on agent surfaces, coded from the operators' own pages read on 3 November 2026; "announced" is not live, and a platform-wide statement is never promoted to a format.
Machine-readable copy: data/agentic-map.csv.
| Platform / product | Capability | Status | Date | What the record shows | What the record supports | Src |
|---|---|---|---|---|---|---|
| Amazon store (shopping results and product detail pages; a prompt may open a dialog in Rufus) | R01 Sponsored Products prompts | availability | live | 2026-03-25 | operator subtitle 'Sponsored Products prompts and Sponsored Brands prompts and reporting are now generally available' (10 March 2026); GA in the US from 25 March 2026 | operator: Amazon Ads; record R01; finding RTL-K-F01; verify/K.json RTL-K-F01: GA rests on the operator subtitle | |
| Amazon store (shopping results and product detail pages; a prompt may open a dialog in Rufus) | R01 Sponsored Products prompts | operator reporting | live | 2026-03-10 | Prompts report: impressions, clicks, CTR, CPC, spend, sales, ACOS, ROAS, 7-day orders and units; console and API; 'prompts and reporting are now generally available' (operator subtitle) | operator: Amazon Ads; record R01; finding RTL-K-F02; verify/K.json RTL-K-F02 and TABLE:agent_surfaces correction 1: reporting is GA in the operator's words | |
| Amazon store (shopping results and product detail pages; a prompt may open a dialog in Rufus) | R01 Sponsored Products prompts | third-party attribution | unconfirmed | 2026-03-10 | no attribution partner named for the format on the two Amazon pages | operator: Amazon Ads; record R01; finding RTL-K-F03; verify/K.json TABLE:agent_surfaces correction 2: the 2026-03-10 date rests on RTL-U-S312, now cited with RTL-K-S15 | |
| Amazon store (shopping results and product detail pages; a prompt may open a dialog in Rufus) | R01 Sponsored Products prompts | causal outcome measurement | unconfirmed | 2026-06-11 | no holdout, control, incrementality or lift wording on either Amazon page | operator: Amazon Ads; record R01; finding RTL-K-F04 | |
| Amazon store (shopping results and product detail pages; a prompt may open a dialog in Rufus) | R01 Sponsored Products prompts | ad-quality verification | unconfirmed | 2026-06-11 | no third-party verification named for the format | operator: Amazon Ads; record R01; finding RTL-K-F05 | |
| Amazon store (shopping results and product detail pages; a prompt may open a dialog in Rufus) | R02 Sponsored Brands prompts | availability | live | 2026-03-25 | operator subtitle 'Sponsored Products prompts and Sponsored Brands prompts and reporting are now generally available' (10 March 2026); GA in the US from 25 March 2026 | operator: Amazon Ads; record R02; finding RTL-K-F06; verify/K.json RTL-K-F06: GA rests on the operator subtitle | |
| Amazon store (shopping results and product detail pages; a prompt may open a dialog in Rufus) | R02 Sponsored Brands prompts | operator reporting | live | 2026-03-10 | Prompts report: impressions, clicks, CTR, CPC, spend, sales, ACOS, ROAS, 7-day orders and units; console and API; 'prompts and reporting are now generally available' (operator subtitle) | operator: Amazon Ads; record R02; finding RTL-K-F07; verify/K.json RTL-K-F07 and TABLE:agent_surfaces correction 1: reporting is GA in the operator's words | |
| Amazon store (shopping results and product detail pages; a prompt may open a dialog in Rufus) | R02 Sponsored Brands prompts | third-party attribution | unconfirmed | 2026-03-10 | no attribution partner named for the format on the two Amazon pages | operator: Amazon Ads; record R02; finding RTL-K-F08; verify/K.json TABLE:agent_surfaces correction 2: the 2026-03-10 date rests on RTL-U-S312, now cited with RTL-K-S15 | |
| Amazon store (shopping results and product detail pages; a prompt may open a dialog in Rufus) | R02 Sponsored Brands prompts | causal outcome measurement | unconfirmed | 2026-06-11 | no holdout, control, incrementality or lift wording on either Amazon page | operator: Amazon Ads; record R02; finding RTL-K-F09 | |
| Amazon store (shopping results and product detail pages; a prompt may open a dialog in Rufus) | R02 Sponsored Brands prompts | ad-quality verification | unconfirmed | 2026-06-11 | no third-party verification named for the format | operator: Amazon Ads; record R02; finding RTL-K-F10 | |
| Alexa for Shopping (Amazon's assistant after 13 May 2026; Rufus and Alexa+ merged) | R03 Sponsored Products and Sponsored Brands prompts on Alexa for Shopping | availability | live | 2026-06-11 | 'available today' / 'current state'; no stage word | operator: Amazon Ads; record R03; finding RTL-K-F11; verify/K.json RTL-K-F11 and TABLE:agent_surfaces correction 3: '(plus sponsored ads)' removed from the R03 label | |
| Alexa for Shopping (Amazon's assistant after 13 May 2026; Rufus and Alexa+ merged) | R03 Sponsored Products and Sponsored Brands prompts on Alexa for Shopping | operator reporting | live | 2026-06-11 | 'closed-loop measurement ... impressions all the way to conversions'; no metric list for the surface | operator: Amazon Ads; record R03; finding RTL-K-F12; verify/K.json RTL-K-F11 and TABLE:agent_surfaces correction 3: '(plus sponsored ads)' removed from the R03 label | |
| Alexa for Shopping (Amazon's assistant after 13 May 2026; Rufus and Alexa+ merged) | R03 Sponsored Products and Sponsored Brands prompts on Alexa for Shopping | third-party attribution | unconfirmed | 2026-06-11 | no attribution partner named | operator: Amazon Ads; record R03; finding RTL-K-F13; verify/K.json RTL-K-F11 and TABLE:agent_surfaces correction 3: '(plus sponsored ads)' removed from the R03 label | |
| Alexa for Shopping (Amazon's assistant after 13 May 2026; Rufus and Alexa+ merged) | R03 Sponsored Products and Sponsored Brands prompts on Alexa for Shopping | causal outcome measurement | unconfirmed | 2026-06-11 | no counterfactual wording | operator: Amazon Ads; record R03; finding RTL-K-F14; verify/K.json RTL-K-F11 and TABLE:agent_surfaces correction 3: '(plus sponsored ads)' removed from the R03 label | |
| Alexa for Shopping (Amazon's assistant after 13 May 2026; Rufus and Alexa+ merged) | R03 Sponsored Products and Sponsored Brands prompts on Alexa for Shopping | ad-quality verification | unconfirmed | 2026-06-11 | no third-party verification named | operator: Amazon Ads; record R03; finding RTL-K-F15; verify/K.json RTL-K-F11 and TABLE:agent_surfaces correction 3: '(plus sponsored ads)' removed from the R03 label | |
| ChatGPT (Free and Go plans) | R04 standard ads at the bottom of answers | availability | live | 2026-02-09 | test started 9 Feb 2026 (US); still 'pilot' wording on 5 May 2026 ('In the first phase of the pilot'); 'launch' and 'available' wording later (31 Aug and 21 Oct 2026); no GA statement found | operator: OpenAI; record R04; finding RTL-K-F16; verify/K.json RTL-K-F16: pilot wording to 5 May 2026 | |
| ChatGPT (Free and Go plans) | R04 standard ads at the bottom of answers | operator reporting | live | 2026-05-05 | views, clicks; Conversions API and pixel; aggregated insights; product-segmented Insights for catalog ads (developer docs, undated) | operator: OpenAI; record R04; finding RTL-K-F17; verify/K.json omission (RTL-K-F16 note): OpenAI Ads product-feeds docs, RTL-VK-S02 | |
| ChatGPT (Free and Go plans) | R04 standard ads at the bottom of answers | third-party attribution | live | 2026-10-05 | ten named attribution partners, no stage label | operator: OpenAI; record R04; finding RTL-K-F18 | |
| ChatGPT (Free and Go plans) | R04 standard ads at the bottom of answers | causal outcome measurement | beta | 2026-10-05 | 'still in its early stages'; partnering with Haus, Measured and WorkMagic 'to explore' geo-based experiments; one partner-reported lift (WorkMagic for Dose; no design, sample or period) | operator: OpenAI; record R04; finding RTL-K-F19; verify/K.json RTL-K-F19 | |
| ChatGPT (Free and Go plans) | R04 standard ads at the bottom of answers | ad-quality verification | beta | 2026-10-05 | DV and IAS suitability 'evaluation pilots' the operator says it is 'developing'; start by 5 Oct 2026 not stated | operator: OpenAI; record R04; finding RTL-K-F20; verify/K.json RTL-K-F20: 'pilot or test' is the nearest code | |
| ChatGPT (image generation, Free and Go plans) | R05 visual ad format | availability | announced | 2026-10-05 | announced 5 Oct 2026; planned start 'later this month'; no evidenced start by 21 Oct 2026 | operator: OpenAI; record R05; finding RTL-K-F21 | |
| ChatGPT (image generation, Free and Go plans) | R05 visual ad format | operator reporting | unconfirmed | 2026-10-05 | no format-specific reporting documented | operator: OpenAI; record R05; finding RTL-K-F22 | |
| ChatGPT (image generation, Free and Go plans) | R05 visual ad format | third-party attribution | unconfirmed | 2026-10-05 | no format-specific evidence | operator: OpenAI; record R05; finding RTL-K-F23 | |
| ChatGPT (image generation, Free and Go plans) | R05 visual ad format | causal outcome measurement | unconfirmed | 2026-10-05 | no format-specific evidence | operator: OpenAI; record R05; finding RTL-K-F24 | |
| ChatGPT (image generation, Free and Go plans) | R05 visual ad format | ad-quality verification | unconfirmed | 2026-10-05 | no format-specific evidence | operator: OpenAI; record R05; finding RTL-K-F25 | |
| ChatGPT | R06 Sponsored Agents | availability | beta | 2026-09-16 | 'now being tested with select advertisers in the United States' | operator: OpenAI; record R06; finding RTL-K-F26 | |
| ChatGPT | R06 Sponsored Agents | operator reporting | unconfirmed | 2026-09-16 | no format-specific reporting | operator: OpenAI; record R06; finding RTL-K-F27 | |
| ChatGPT | R06 Sponsored Agents | third-party attribution | unconfirmed | 2026-09-16 | no format-specific evidence | operator: OpenAI; record R06; finding RTL-K-F28 | |
| ChatGPT | R06 Sponsored Agents | causal outcome measurement | unconfirmed | 2026-09-16 | no format-specific evidence | operator: OpenAI; record R06; finding RTL-K-F29 | |
| ChatGPT | R06 Sponsored Agents | ad-quality verification | unconfirmed | 2026-09-16 | no format-specific evidence | operator: OpenAI; record R06; finding RTL-K-F30 | |
| ChatGPT (Amazon-sold placements) | R07 Amazon Ads-sold conversational ads in ChatGPT | availability | beta | 2026-09-10 | 'Select advertisers in the U.S. are currently testing ChatGPT Ads as part of a pilot.'; the page body names neither the Amazon DSP as buying route nor first-party data; shopping-use wording is platform-level; the only named participant is a travel advertiser (Delta Vacations) | operator: OpenAI (surface) / Amazon Ads (seller); record R07; finding RTL-K-F31; verify/K.json RTL-K-F31 | |
| ChatGPT (Amazon-sold placements) | R07 Amazon Ads-sold conversational ads in ChatGPT | operator reporting | unconfirmed | 2026-09-10 | no reporting described | operator: OpenAI (surface) / Amazon Ads (seller); record R07; finding RTL-K-F32 | |
| ChatGPT (Amazon-sold placements) | R07 Amazon Ads-sold conversational ads in ChatGPT | third-party attribution | unconfirmed | 2026-09-10 | not mentioned | operator: OpenAI (surface) / Amazon Ads (seller); record R07; finding RTL-K-F33 | |
| ChatGPT (Amazon-sold placements) | R07 Amazon Ads-sold conversational ads in ChatGPT | causal outcome measurement | unconfirmed | 2026-09-10 | not mentioned | operator: OpenAI (surface) / Amazon Ads (seller); record R07; finding RTL-K-F34 | |
| ChatGPT (Amazon-sold placements) | R07 Amazon Ads-sold conversational ads in ChatGPT | ad-quality verification | unconfirmed | 2026-09-10 | not mentioned | operator: OpenAI (surface) / Amazon Ads (seller); record R07; finding RTL-K-F35 | |
| Google AI Overviews and AI Mode | R08 ads (Search and Shopping ads; sponsored product list items) | availability | live | 2025-05-21 | 'available in English in the United States and select English-speaking markets'; multiturn ads launching 'in AI Mode' in the US (US copy, 28 Oct 2026) | operator: Google; record R08; finding RTL-K-F36; verify/K.json RTL-K-F36 (US copy found, RTL-VK-S01) | |
| Google AI Overviews and AI Mode | R08 ads (Search and Shopping ads; sponsored product list items) | operator reporting | live | undated (accessed 2026-11-03) | impressions, clicks, conversions, cost 'in aggregate'; no separate AI-surface segment | operator: Google; record R08; finding RTL-K-F37; verify/K.json TABLE:agent_surfaces correction 4: access date is not a publication date | |
| Google AI Overviews and AI Mode | R08 ads (Search and Shopping ads; sponsored product list items) | third-party attribution | unconfirmed | 2026-10-28 | not mentioned | operator: Google; record R08; finding RTL-K-F38 | |
| Google AI Overviews and AI Mode | R08 ads (Search and Shopping ads; sponsored product list items) | causal outcome measurement | unconfirmed | 2026-10-28 | CTR comparison footnoted 'Google Internal Data, US, August 20, 2026- September 3, 2026' on the US copy; internal engagement comparison with no sample or causal design | operator: Google; record R08; finding RTL-K-F39; verify/K.json RTL-K-F39 | |
| Google AI Overviews and AI Mode | R08 ads (Search and Shopping ads; sponsored product list items) | ad-quality verification | unconfirmed | undated (accessed 2026-11-03) | not mentioned | operator: Google; record R08; finding RTL-K-F40; verify/K.json TABLE:agent_surfaces correction 4 (same undated help page): access date is not a publication date | |
| Google AI Mode | R09 Direct Offers ('Sponsored deal') | availability | beta | 2026-01-11 | 'This new Google Ads pilot' | operator: Google; record R09; finding RTL-K-F41 | |
| Google AI Mode | R09 Direct Offers ('Sponsored deal') | operator reporting | unconfirmed | 2026-05-20 | not described | operator: Google; record R09; finding RTL-K-F42 | |
| Google AI Mode | R09 Direct Offers ('Sponsored deal') | third-party attribution | unconfirmed | 2026-05-20 | not mentioned | operator: Google; record R09; finding RTL-K-F43 | |
| Google AI Mode | R09 Direct Offers ('Sponsored deal') | causal outcome measurement | unconfirmed | 2026-05-20 | not mentioned | operator: Google; record R09; finding RTL-K-F44 | |
| Google AI Mode | R09 Direct Offers ('Sponsored deal') | ad-quality verification | unconfirmed | 2026-05-20 | not mentioned | operator: Google; record R09; finding RTL-K-F45 | |
| Kroger AI Shopping Assistant (Kroger Family of Companies websites and apps) | R10 Shopping Assistant PLAs | availability | live | 2026-08-03 | 'now available' (assistant); PLAs 'may appear' | operator: Kroger Precision Marketing; record R10; finding RTL-K-F46 | |
| Kroger AI Shopping Assistant (Kroger Family of Companies websites and apps) | R10 Shopping Assistant PLAs | operator reporting | live | 2026-08-03 | report builder, separable from other Search & Browse inventory; no metric list | operator: Kroger Precision Marketing; record R10; finding RTL-K-F47 | |
| Kroger AI Shopping Assistant (Kroger Family of Companies websites and apps) | R10 Shopping Assistant PLAs | third-party attribution | unconfirmed | 2026-08-03 | not mentioned | operator: Kroger Precision Marketing; record R10; finding RTL-K-F48 | |
| Kroger AI Shopping Assistant (Kroger Family of Companies websites and apps) | R10 Shopping Assistant PLAs | causal outcome measurement | unconfirmed | 2026-08-03 | not format-specific | operator: Kroger Precision Marketing; record R10; finding RTL-K-F49 | |
| Kroger AI Shopping Assistant (Kroger Family of Companies websites and apps) | R10 Shopping Assistant PLAs | ad-quality verification | unconfirmed | 2026-08-03 | not mentioned; TAG certification is operator-level | operator: Kroger Precision Marketing; record R10; finding RTL-K-F50 | |
| Microsoft Copilot | R11 shopping ads in Copilot (Product Ads on product detail pages; Offer Highlights) | availability | live | 2026-04-21 | inferred from a performance footnote and a feature post: shopping ads in Copilot evidenced by a Feb-May 2025 quick-back footnote; Offer Highlights 'now available for retail use cases' on Copilot product detail pages, also in 'live pilots'; US availability inferred | operator: Microsoft Advertising; record R11; finding RTL-K-F51; verify/K.json RTL-K-F51 and TABLE:agent_surfaces correction 5: inference, low confidence | |
| Microsoft Copilot | R11 shopping ads in Copilot (Product Ads on product detail pages; Offer Highlights) | operator reporting | unconfirmed | 2026-04-21 | no Copilot-specific shopping-ads reporting documented | operator: Microsoft Advertising; record R11; finding RTL-K-F52 | |
| Microsoft Copilot | R11 shopping ads in Copilot (Product Ads on product detail pages; Offer Highlights) | third-party attribution | unconfirmed | 2026-04-21 | not mentioned | operator: Microsoft Advertising; record R11; finding RTL-K-F53 | |
| Microsoft Copilot | R11 shopping ads in Copilot (Product Ads on product detail pages; Offer Highlights) | causal outcome measurement | unconfirmed | 2026-04-21 | internal comparisons only | operator: Microsoft Advertising; record R11; finding RTL-K-F54 | |
| Microsoft Copilot | R11 shopping ads in Copilot (Product Ads on product detail pages; Offer Highlights) | ad-quality verification | unconfirmed | 2026-04-21 | not mentioned | operator: Microsoft Advertising; record R11; finding RTL-K-F55 |
Chapter 12
The Receipt Quotient
The Receipt Quotient (RQ) summarises how much of a service's measurement capability is documented in public, within its cohort, on records read by 3 November 2026. Two rows reach the Most documented band, DoorDash Ads' CPG Sponsored Products and Sponsored Brands among retailer and marketplace networks and Moloco Commerce Media among enablement platforms, and each owes its Standing mean of 4.0 to a single counted assurance cell. Most rows cannot be banded at all, because the record is thinnest where a buyer most needs it: assurance, governance and privacy.
91 service rows are scored in five cohorts: 39 carry a band, 48 are thin records, and the other 4, the eligible rows of cohorts D and E, carry none because those cohorts have too few. In cohort A, 24 of 47 rows are banded, with DoorDash Ads' CPG Sponsored Products and Sponsored Brands Most documented and Walmart Connect Sponsored Search, Instacart Sponsored Products and Amazon Sponsored Products next. In cohort B, 6 of 17 are banded and Cardlytics comes first; in cohort C, 9 of 15, with Moloco Commerce Media ahead of Criteo Commerce Yield and Kevel. DoorDash's Standing rests on one cell, the Alliance for Audited Media's certification of its incrementality method , and Moloco's on one IAB Europe capability certificate . Of the 48 thin records, 44 have no counted Standing cell. Under four other weightings 32 of the 39 banded rows keep their band, both top rows among them; filled with 1s, their unscored cells would drop each top row to Well documented. The RQ measures documented capability within a cohort, never quality or performance, and it never ranks across cohorts.
What the quotient measures
The RQ summarises a row's counted cells within its cohort and places the row in a band against the cohort's field. It measures how much measurement capability a service documents in public, and how specifically. It does not measure whether the numbers are accurate, whether a lift is causal, or whether the service performs or sells well. A score is read only against scores in the same cohort, and a network's rows are never pooled into a network score. The index was fixed with the scoring rules on 3 November 2026, before any service was scored (chapter 1).
Each row is scored from 0 to 5 on twelve dimensions, against anchors that differ by cohort where the tasks differ. Missing evidence is n/e, never zero, and n/a comes only from a cohort's mask. Only cells a cell verifier confirmed or qualified count, and none of the scored cells is unverified. Of the 1092 cells in the matrix, 660 are scored, 427 are n/e and 5 are n/a.
One property shapes every number that follows. A sum would count an unscored dimension as zero and so punish a thin public record; a plain average of scored cells rewards missing evidence. The RQ's group means and shrinkage limit both effects but do not remove the second: a row that leaves a documented weakness unrecorded can score higher than one that discloses it. That effect decides the top of cohort A.
How it is built
The twelve dimensions fall into three groups. Depth is what the number is: the default outcome method and controlled-lift availability, attribution definition disclosure, buyer-controlled outcome analysis, identity and data-basis disclosure, and validation. Reach is where it works: placement measurement coverage, activation and integration routes, scale and US availability. Standing is whether a buyer can trust and trade it: independent assurance, change and correction governance, and privacy and consent posture. A group score is the mean of the row's counted cells in that group, and the row's Raw score is 0.45 × Depth + 0.25 × Reach + 0.30 × Standing, renormalised over the groups present.
Raw is pulled toward the cohort's field mean in proportion to how thin the record is: Shrunk = (k × Raw + K × F) / (k + K), where k is the row's counted cells, F the mean Raw of the cohort's admitted rows and K, the prior strength, 4. The RQ is 100 × Shrunk / 5, kept at full precision and rounded only for display. K is an author-set heuristic, not a statistical estimate, since the dimensions are different capabilities rather than repeated measurements of one. Bands come from the unrounded distance between a row's RQ and its cohort's field RQ: Most documented at 12 points or more above it, Well documented at 1 or more, Partly documented down to 8 below, and Least documented beneath that.
Three thresholds decide what appears. A row needs four counted cells to be admitted: 91 rows were and 16 eligible rows were not. A row needs 6 counted cells, at least one in each group, to carry a band; otherwise it is a thin record, shown with its RQ and the missing group named. A cohort shows bands only when 6 or more of its rows qualify. Cohorts A, B and C do, and B clears the minimum exactly, after cell verification cut its eligible rows from nine to six. Cohorts D and E have 2 and 2 eligible rows and carry quotients without bands.
The field RQ is 39 in cohort A, 37 in B, 42 in C, 39 in D and 38 in E. Most documented starts at 50.6 in A, 49.5 in B and 54.2 in C, and Well documented at 39.6, 38.5 and 43.2. At these field means a row scoring 5 on every dimension would reach 85 in cohort A, because of the shrinkage. The RQ borrows the band offsets of the author's Contextual Quotient, but its field mean is the cohort's mean Raw score rather than a pooled group mean, so the two do not read alike, and a score in one cohort says nothing about a score in another.
The results
Across the three banded cohorts, 2 rows are Most documented, 19 Well documented, 17 Partly documented and 1 Least documented.
In cohort A, 24 of 47 rows are band-eligible. DoorDash Ads' CPG Sponsored Products and Sponsored Brands is the one Most documented row. Walmart Connect Sponsored Search, Instacart Sponsored Products and Amazon Sponsored Products lead the eleven Well documented rows, which include Kroger Precision Marketing's two, Instacart's display and video ads and five more from Walmart and Amazon. Twelve rows are Partly documented, from Amazon display ads, Roundel and Orange Apron Media down to Macy's Media Network, and none is Least documented. Two eBay Promoted Listings rows score level with Walmart's and Instacart's leaders but have no counted Standing cell, so they carry no band; both Etsy rows, Uber Eats and DoorDash's restaurant Sponsored Listings are thin for the same reason.
DoorDash's CPG row reports attributed sales by default with no counterfactual: sales from the first order within 14 days of a click, credited to the most recent ad interaction . It documents six of the eight attribution parameters, offers exports and a direct API connection , and scores 1 on validation, where the record holds a Circana partnership announced on 2 September 2026 to bring "independent, third-party validation" to DoorDash's measurement for select advertisers . Its Reach is among the highest in the cohort, with no scale figure counted.
In cohort B, 6 of 17 rows are eligible and none is Most documented. Fetch Special Offers comes first, then Cardlytics' card-linked offers, the Ibotta Performance Network's offers and Upside's merchant promotions in Well documented, with Klarna Search & Compare and Chase Media Solutions Partly documented. No cohort B service is on MRC's, AAM's or IAB Europe's lists as read on 3 November 2026, so independent assurance is n/e on every row . Fetch's Standing mean also rests on one cell, a privacy score of 4 for a policy that treats participation as consent to sharing with brand partners and keeps sensitive data out of it . Cardlytics' rests on two, a 1 on governance and a 3 on privacy. Its governance cell scored 2 until the final review, on data delays that say when its numbers settle rather than how a reported figure is revised ; at 2 Cardlytics was first at 47.99, and at 1 it is 45.74, behind Fetch at 46.77, both Well documented.
In cohort C, 9 of 15 rows are eligible. Moloco Commerce Media is Most documented; Criteo Commerce Yield, Kevel Retail Media Cloud, Koddi Ads and Carrot Ads are Well documented; Amazon Retail Ad Service, Osmos and Vantage are Partly documented. Swiftly, whose page asserts IAB-standard attribution and sales lift with no method , is the paper's one Least documented row. Commerce Yield documents its attribution settings with defaults, 30 days after a click and 1 day after a view , and holds MRC accreditation of on-site display impressions and clicks from 15 March 2024 . Kevel's assurance cell is a listing: MRC's under-review list carries it in a reviewed pre-audit that includes clicks, conversions and incrementality, a status rather than an accreditation .
In cohorts D and E, which carry no bands, the eligible rows are Criteo Commerce Max and Perpetua, and Numerator's sales lift and Circana Media Lift.
The capability matrix
One row per purchasable service, scored within its cohort on documented measurement capability, never on measurement quality or effectiveness. Click a cell for its anchor, rationale, sources and verification status; n/e means the opened record gave no basis to score, n/a that the dimension is outside the cohort's mask. A network's rows sit together; there is no pooled network score.
| Product | RQ | Method | Attribution | Access | Identity | Validation | Placements | Routes | Scale | US | Assurance | Governance | Privacy | Counted |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| A · Retailer and marketplace media networks | ||||||||||||||
| banded, by RQ | ||||||||||||||
| DoorDash AdsDoorDash Ads CPG Sponsored Products and Sponsored Brands · Marketplaces | 56±Most documented | 9/12 | ||||||||||||
| Walmart Connect ◆Walmart Connect Sponsored Search · Retailers · operating | 48Well documented | 11/12 | ||||||||||||
| Instacart Ads ◆Instacart Sponsored Products · Marketplaces · operating; acquisition agreement pending | 48Well documented | 12/12 | ||||||||||||
| Amazon Ads ◆Amazon Sponsored Products · Marketplaces · active; defendant in two pending FTC actions (allegations, no findings) | 47±Well documented | 9/12 | ||||||||||||
| Instacart Ads (Maplebear Inc.) ◆Instacart display and video ads · Marketplaces | 46Well documented | 11/12 | ||||||||||||
| Walmart Connect (Walmart Inc.)Walmart Connect Onsite Display · Retailers | 45Well documented | 11/12 | ||||||||||||
| Kroger Precision MarketingKroger Precision Marketing onsite search and display · Retailers · active (both register rows live) | 44Well documented | 11/12 | ||||||||||||
| Amazon Ads ◆Amazon DSP · Marketplaces | 43±Well documented | 9/12 | ||||||||||||
| Amazon AdsAmazon Sponsored Brands · Marketplaces | 43±Well documented | 8/12 | ||||||||||||
| Kroger Precision Marketing (84.51° LLC)Kroger Precision Marketing offsite programmatic · Retailers | 43Well documented | 11/12 | ||||||||||||
| Walmart Connect (Walmart Inc.) ◆Walmart DSP · Retailers | 42±Well documented | 10/12 | ||||||||||||
| Walmart Connect (Walmart Inc.) ◆Walmart Connect Offsite Display · Retailers | 41±Well documented | 10/12 | ||||||||||||
| Amazon AdsAmazon display ads · Marketplaces | 39±Partly documented | 8/12 | ||||||||||||
| Roundel (Target Corporation)Roundel · Retailers | 39±Partly documented | 8/12 | ||||||||||||
| Orange Apron Media (The Home Depot, Inc.)Orange Apron Media · Retailers | 39±Partly documented | 8/12 | ||||||||||||
| Albertsons Media CollectiveAlbertsons Media Collective onsite · Retailers · live | 38±Partly documented | 11/12 | ||||||||||||
| Albertsons Media CollectiveAlbertsons Media Collective offsite · Retailers | 37±Partly documented | 11/12 | ||||||||||||
| Best Buy Ads (Best Buy Co., Inc.)Best Buy Ads Sponsored Products · Retailers | 37±Partly documented | 7/12 | ||||||||||||
| Best Buy Ads (Best Buy Co., Inc.)Best Buy Ads display, video, search and CTV media · Retailers | 37±Partly documented | 7/12 | ||||||||||||
| CVS Media Exchange (CVS Health Corporation)CVS Media Exchange · Retailers | 36±Partly documented | 10/12 | ||||||||||||
| Albertsons Media CollectiveAlbertsons Media Collective in-store · Retailers | 35±Partly documented | 11/12 | ||||||||||||
| Walmart Connect (Walmart Inc.)Walmart Connect Store Ads · Retailers | 34±Partly documented | 7/12 | ||||||||||||
| DG Media Network (Dollar General Corporation)DG Media Network · Retailers | 34±Partly documented | 9/12 | ||||||||||||
| Macy's Media Network (Advertex Communications Inc. d/b/a Macy's Media Network; Macy's, Inc.)Macy's Media Network · Retailers | 32±Partly documented | 8/12 | ||||||||||||
| not banded, A to Z | ||||||||||||||
| Chewy Ads (Chewy, Inc.)Chewy Ads · Marketplaces | 37Thin record · no Standing | 5/12 | ||||||||||||
| DICK'S Media (DICK'S Sporting Goods, Inc.)DICK'S Media · Retailers | 32Thin record · no Standing | 5/12 | ||||||||||||
| DoorDash AdsDoorDash Ads restaurant Sponsored Listings · Marketplaces | 35Thin record · no Standing | 7/12 | ||||||||||||
| eBay Advertising (eBay Inc.)eBay Promoted Listings General · Marketplaces | 48Thin record · no Standing | 7/12 | ||||||||||||
| eBay Advertising (eBay Inc.)eBay Promoted Listings Priority · Marketplaces | 48Thin record · no Standing | 7/12 | ||||||||||||
| eBay Advertising (eBay Inc.)eBay Promoted Offsite · Marketplaces | 44Thin record · no Standing | 6/12 | ||||||||||||
| eBay Advertising (eBay Inc.)eBay Promoted Stores · Marketplaces | 40Thin record · no Standing | 6/12 | ||||||||||||
| EtsyEtsy Ads · Marketplaces | 44Thin record · no Standing | 6/12 | ||||||||||||
| EtsyEtsy Offsite Ads · Marketplaces | 46Thin record · no Standing | 6/12 | ||||||||||||
| Expedia Group AdvertisingExpedia Group Advertising sponsored listings · Marketplaces | 38Thin record · no Standing | 7/12 | ||||||||||||
| Expedia Group Advertising (Vrbo)Vrbo Sponsored Listings · Marketplaces | 38Thin record · no Standing | 6/12 | ||||||||||||
| Gopuff Ads (GoBrands, Inc.)Gopuff Ads · Marketplaces | 32Thin record · no Standing | 4/12 | ||||||||||||
| Gulp Media Network (7-Eleven)Gulp Media Network · Retailers | 36Thin record · no Standing | 6/12 | ||||||||||||
| Hy-Vee RedMedia (Hy-Vee, Inc.)Hy-Vee RedMedia · Retailers | 33Thin record · no Standing | 7/12 | ||||||||||||
| Kohl's Media Network (Kohl's Corporation)Kohl's Media Network · Retailers | 34Thin record · no Standing | 5/12 | ||||||||||||
| Leap Media Group (Giant Eagle, Inc.)Leap Media Group (Giant Eagle) · Retailers | 28Thin record · no Standing | 5/12 | ||||||||||||
| Lowe's Media Network (Lowe's Companies, Inc.)Lowe's Media Network · Retailers | 34Thin record · no Standing | 5/12 | ||||||||||||
| Petco RMN (Petco Health and Wellness Company, Inc.)Petco RMN · Retailers | 32Thin record · no Standing | 5/12 | ||||||||||||
| Sam's Club Connect (Walmart Inc.)Sam's Club Connect · Retailers | 33Thin record · no Standing | 7/12 | ||||||||||||
| Staples Media NetworkRetailers | 32Thin record · no Standing | 4/12 | ||||||||||||
| UB Media (Ulta Beauty)UB Media · Retailers | 28Thin record · no Standing | 5/12 | ||||||||||||
| Uber Advertising (Uber Eats)Uber Eats sponsored listings · Marketplaces | 46Thin record · no Standing | 7/12 | ||||||||||||
| Walgreens Advertising Group (Walgreen Co.)Walgreens Advertising Group · Retailers | 33Thin record · no Standing | 7/12 | ||||||||||||
| B · Commerce media networks | ||||||||||||||
| banded, by RQ | ||||||||||||||
| Fetch (Fetch Rewards LLC)Fetch Special Offers · Commerce-data operators | 47±Well documented | 8/12 | ||||||||||||
| CardlyticsCardlytics card-linked offers · Commerce-data operators · live (operating, publicly listed) | 46Well documented | 11/12 | ||||||||||||
| IbottaIbotta Performance Network offers · Commerce-data operators | 43±Well documented | 9/12 | ||||||||||||
| UpsideUpside merchant promotions · Commerce-data operators | 42±Well documented | 9/12 | ||||||||||||
| KlarnaKlarna Search & Compare sponsored placements · Commerce-data operators | 37±Partly documented | 8/12 | ||||||||||||
| Chase Media Solutions (JPMorgan Chase Bank, N.A.)Chase Media Solutions · Commerce-data operators | 37±Partly documented | 9/12 | ||||||||||||
| not banded, A to Z | ||||||||||||||
| American ExpressAmex Offers · Commerce-data operators | 33Thin record · no Standing | 7/12 | ||||||||||||
| Citi (U.S. Consumer Cards)Citi Commerce Media · Commerce-data operators | 31Thin record · no Standing | 6/12 | ||||||||||||
| Expedia Group AdvertisingExpedia Group Advertising for non-endemic brands · Marketplaces | 33Thin record · no Standing | 7/12 | ||||||||||||
| IbottaIbotta D2C ad products · Commerce-data operators | 39Thin record · no Depth, Standing | 4/12 | ||||||||||||
| Lyft Ads (Lyft, Inc.)Lyft Ads · Commerce-data operators | 33Thin record · no Standing | 5/12 | ||||||||||||
| PayPal Ads (PayPal Holdings)PayPal Ads · Commerce-data operators | 34Thin record · no Standing | 8/12 | ||||||||||||
| RoktRokt Ads · Commerce-data operators | 45Thin record · no Standing | 7/12 | ||||||||||||
| Uber Advertising (Uber Technologies)Uber Journey Ads · Commerce-data operators | 34Thin record · no Standing | 4/12 | ||||||||||||
| United AirlinesKinective Media by United Airlines · Commerce-data operators | 40Thin record · no Depth | 4/12 | ||||||||||||
| Walmart Connect (Walmart Inc.)Walmart Connect non-endemic advertising · Retailers | 35Thin record · no Depth | 5/12 | ||||||||||||
| ZillowZillow Elevate Media Manager · Commerce-data operators | 34Thin record · no Standing | 4/12 | ||||||||||||
| C · Enablement platforms | ||||||||||||||
| banded, by RQ | ||||||||||||||
| MolocoMoloco Commerce Media · Ad tech | 58±Most documented | 9/12 | ||||||||||||
| CriteoCriteo Commerce Yield · Ad tech · active; redomiciled from France to Luxembourg (completed by 29 July 2026); merger into Criteo Holdings, Inc. (Delaware) agreed 5 August 2026 with an Effective Time of 1 January 2027 unless otherwise agreed, pending shareholder approval and other conditions at 3 November 2026 | 53±Well documented | 10/12 | ||||||||||||
| KevelKevel Retail Media Cloud · Ad tech | 50±Well documented | 10/12 | ||||||||||||
| KoddiKoddi Ads · Ad tech | 49±Well documented | 8/12 | ||||||||||||
| InstacartCarrot Ads · Ad tech | 44±Well documented | 10/12 | ||||||||||||
| Amazon AdsAmazon Retail Ad Service · Ad tech | 42±Partly documented | 8/12 | ||||||||||||
| OsmosOsmos Retail Media OS · Ad tech | 41±Partly documented | 6/12 | ||||||||||||
| VantageAd tech | 35±Partly documented | 6/12 | ||||||||||||
| SwiftlyAd tech | 34±Least documented | 6/12 | ||||||||||||
| not banded, A to Z | ||||||||||||||
| EpsilonEpsilon Retail Media · Ad tech | 33Thin record · no Standing | 5/12 | ||||||||||||
| PentaleapAd tech | 41Thin record · no Depth, Standing | 4/12 | ||||||||||||
| RipplAd tech | 36Thin record · no Standing | 4/12 | ||||||||||||
| STRATACACHESTRATACACHE in-store retail media · Ad tech | 43Thin record · no Depth | 4/12 | ||||||||||||
| TopsortAd tech | 47Thin record · no Standing | 7/12 | ||||||||||||
| ZitchaAd tech | 37Thin record · no Standing | 6/12 | ||||||||||||
| D · Cross-network optimisation platforms | ||||||||||||||
| no bands in this cohort, A to Z | ||||||||||||||
| CriteoCriteo Commerce Max · Ad tech | 50Cohort too small | 7/11 | ||||||||||||
| IntentwiseIntentwise Optimize · Ad tech | 40Thin record · no Standing | 8/11 | ||||||||||||
| Perpetua (Flywheel Digital, Omnicom) ◆Perpetua · Ad tech | 36Cohort too small | 7/11 | ||||||||||||
| StacklineStackline Ad Manager · Ad tech | 37Thin record · no Standing | 4/11 | ||||||||||||
| XnurtaAd tech | 36Thin record · no Standing | 4/11 | ||||||||||||
| E · Outcome measurement providers | ||||||||||||||
| no bands in this cohort, A to Z | ||||||||||||||
| AttainMeasurers | 39Thin record · no Standing | 7/12 | ||||||||||||
| CatalinaCatalina HUB360 · Measurers | 39Thin record · no Standing | 7/12 | ||||||||||||
| CircanaCircana Media Lift · Measurers · active; private | 33Cohort too small | 9/12 | ||||||||||||
| Circana (NCSolutions)NCS Sales Effect · Measurers | 37Thin record | 5/12 | ||||||||||||
| IncrementalIncremental — Measurement & Reporting · Measurers | 45Thin record · no Standing | 6/12 | ||||||||||||
| NielsenNielsen Sales Lift · Measurers | 37Thin record · no Standing | 7/12 | ||||||||||||
| NumeratorNumerator sales lift · Measurers | 36Cohort too small | 9/12 | ||||||||||||
Read the band, not the rank. A · Retailer and marketplace media networks (47 rows, field RQ 38.6). Band lines: Most documented 50.6, Well documented 39.6, Partly documented 30.6, Least documented below. Under four reweightings (equal thirds, depth-led, reach-led, standing-led), with Raw and the field mean recomputed, the product in the top band keeps it, while positions move by up to 11 places and by 1.7 on average. Of the 24 banded products, 20 keep their band under every weighting (Albertsons Media Collective, Albertsons Media Collective, Albertsons Media Collective, Amazon Ads, Amazon Ads, Best Buy Ads (Best Buy Co., Inc.), Best Buy Ads (Best Buy Co., Inc.), CVS Media Exchange (CVS Health Corporation), DG Media Network (Dollar General Corporation), DoorDash Ads, Instacart Ads, Instacart Ads (Maplebear Inc.), Kroger Precision Marketing, Kroger Precision Marketing (84.51° LLC), Orange Apron Media (The Home Depot, Inc.), Walmart Connect, Walmart Connect (Walmart Inc.), Walmart Connect (Walmart Inc.), Walmart Connect (Walmart Inc.), Walmart Connect (Walmart Inc.)); Amazon Ads, Amazon Ads, Macy's Media Network (Advertex Communications Inc. d/b/a Macy's Media Network; Macy's, Inc.), Roundel (Target Corporation) cross a band line under at least one. With a prior strength of 0 (Raw, no shrinkage), 2 and 8 instead of 4, positions move by up to 4, 2 and 2 places. Deleting one counted cell at a time moves a score by up to 6 points up (Amazon Ads, privacy and consent posture) or 6 down (Walmart Connect, independent assurance); 8 deletions would drop a row below admission and 10 would cost a row its band eligibility. In the missing-evidence scenarios (every n/e cell of a banded row filled with 1, then with 4, with the field and every other row frozen), 18 of the 23 banded rows with n/e cells change band (Albertsons Media Collective, Albertsons Media Collective, Albertsons Media Collective, Amazon Ads, Amazon Ads, Amazon Ads, Amazon Ads, Best Buy Ads (Best Buy Co., Inc.), Best Buy Ads (Best Buy Co., Inc.), CVS Media Exchange (CVS Health Corporation), DG Media Network (Dollar General Corporation), DoorDash Ads, Macy's Media Network (Advertex Communications Inc. d/b/a Macy's Media Network; Macy's, Inc.), Orange Apron Media (The Home Depot, Inc.), Roundel (Target Corporation), Walmart Connect (Walmart Inc.), Walmart Connect (Walmart Inc.), Walmart Connect (Walmart Inc.); marked ±). B · Commerce media networks (17 rows, field RQ 37.5). Band lines: Most documented 49.5, Well documented 38.5, Partly documented 29.5, Least documented below. Under four reweightings (equal thirds, depth-led, reach-led, standing-led), with Raw and the field mean recomputed, no product sits in the top band (Most documented) (Fetch (Fetch Rewards LLC) joins it under standing-led weights), while positions move by up to 6 places and by 0.7 on average. Of the 6 banded products, 5 keep their band under every weighting (Cardlytics, Chase Media Solutions (JPMorgan Chase Bank, N.A.), Ibotta, Klarna, Upside); Fetch (Fetch Rewards LLC) crosses a band line under at least one. With a prior strength of 0 (Raw, no shrinkage), 2 and 8 instead of 4, positions move by up to 3, 2 and 2 places. Deleting one counted cell at a time moves a score by up to 9 points up (Walmart Connect (Walmart Inc.), privacy and consent posture) or 9 down (Ibotta, us availability and coverage); 16 deletions would drop a row below admission and 4 would cost a row its band eligibility, 4 of them leaving the cohort below the 6 eligible rows bands need, which would withdraw every band in it (Chase Media Solutions (JPMorgan Chase Bank, N.A.), privacy and consent posture; Fetch (Fetch Rewards LLC), privacy and consent posture; Ibotta, privacy and consent posture; Upside, privacy and consent posture). In the missing-evidence scenarios (every n/e cell of a banded row filled with 1, then with 4, with the field and every other row frozen), 5 of the 6 banded rows with n/e cells change band (Chase Media Solutions (JPMorgan Chase Bank, N.A.), Fetch (Fetch Rewards LLC), Ibotta, Klarna, Upside; marked ±). C · Enablement platforms (15 rows, field RQ 42.2). Band lines: Most documented 54.2, Well documented 43.2, Partly documented 34.2, Least documented below. Under four reweightings (equal thirds, depth-led, reach-led, standing-led), with Raw and the field mean recomputed, the product in the top band keeps it, while positions move by up to 1 places and by 0.4 on average. Of the 9 banded products, 7 keep their band under every weighting (Criteo, Instacart, Kevel, Koddi, Moloco, Osmos, Vantage); Amazon Ads, Swiftly cross a band line under at least one. With a prior strength of 0 (Raw, no shrinkage), 2 and 8 instead of 4, positions move by up to 1, 1 and 0 places. Deleting one counted cell at a time moves a score by up to 8 points up (Rippl, identity and data-basis disclosure) or 5 down (Moloco, independent assurance); 12 deletions would drop a row below admission and 21 would cost a row its band eligibility. In the missing-evidence scenarios (every n/e cell of a banded row filled with 1, then with 4, with the field and every other row frozen), 9 of the 9 banded rows with n/e cells change band (Amazon Ads, Criteo, Instacart, Kevel, Koddi, Moloco, Osmos, Swiftly, Vantage; marked ±). D · Cross-network optimisation platforms (5 rows, field RQ 39.3). Only 2 rows are band-eligible, under the 6 a cohort needs, so its scores and coverage are shown without bands. Deleting one counted cell at a time moves a score by up to 11 points up (Stackline, validation) or 4 down (Criteo, attribution definition disclosure); 8 deletions would drop a row below admission and 2 would cost a row its band eligibility. E · Outcome measurement providers (7 rows, field RQ 38.2). Only 2 rows are band-eligible, under the 6 a cohort needs, so its scores and coverage are shown without bands. Deleting one counted cell at a time moves a score by up to 5 points up (Numerator, privacy and consent posture) or 4 down (Circana (NCSolutions), activation and integration routes); 0 deletions would drop a row below admission and 2 would cost a row its band eligibility. Missing evidence is not neutral: every deletion case is in data/rq-missing-evidence.json, and every row’s scenario scores, including a hypothetical 0-to-5 envelope reported separately and never used for the flag, are in data/rq.csv. These tests cover weights, prior strength and missing cells, not evidence grades, the scorer or the product sample. Every input is in the CSVs in the appendix, so the index can be recomputed on different weights.
Only verified or qualified cells count. Unverified cells are shown dashed and treated as n/e. Interest marks (the author's employer, the AdCP working-group role, past employers) are disclosed and never alter a score.
What the two top bands rest on
Each Most documented row has a Standing mean of 4.0, and in each case that mean is one counted cell, independent assurance. Governance and privacy are n/e for both.
DoorDash's cell is the Alliance for Audited Media (AAM) incrementality certification. AAM's post of 10 September 2026 says it evaluated "incrementality measurement and reporting for DoorDash Ads' Sponsored Products and Sponsored Brands", including methodology, data governance, causal testing framework, metric calculations and customer disclosures, against industry guidelines that include the IAB and IAB Europe incrementality guidelines; it names no expiry, renewal or excluded format . On 3 November 2026 AAM's list filtered to that certification returned one entry, DoorDash, with no date . The cell scores 4, outcome assurance of a narrower scope than sold, because the lift product it reviewed is offered to select advertisers through managed service while every buyer receives the attributed default . Beside it, MRC lists DoorDash Ads in a reviewed pre-audit for impressions, clicks and CTR . Governance is n/e because no advertiser terms were read , privacy because no DoorDash privacy notice was screened.
Moloco's cell is IAB Europe's certificate of 7 October 2026 for "TECHNOLOGY PLATFORM CAPABILITY TO SUPPORT MEASUREMENT AND REPORTING FOR ONSITE STANDARD DISPLAY AND SPONSORED PRODUCT ADS", valid to 01/10/2028, which names no standards version, metric list, retailer deployment or geography . It certifies a platform's capability under a European programme and says nothing about what any US retailer running the platform reports; it scores 4 as outcome-method assurance narrower than what Moloco sells. The outcome-method reading comes from the programme's own rules, not from the certificate: certified organisations must meet mandatory requirements that include sale definitions, ROAS and lookback windows, and a random audit of ten campaigns checks attribution . Moloco's Depth is the highest in its cohort because it documents a randomised ghost-bidding holdout, in an undated vendor blog with no sample or significance rule . Its scale cell is a 1: one named US network, Wayfair, with an analysis period of the fourth quarter of 2024 .
Each top band therefore rests on how one cell is read, and the final review asked for the test. With each cohort recomputed, field included, Moloco scored one rung lower, as assurance of a narrower delivery scope, would be Well documented at 53.55, and without the cell a thin record at 52.75. DoorDash's AAM record names its object in the certifier's words, so its 4 rests on less inference; one rung lower it stays Most documented at 51.94, and without the cell it too is a thin record, at 50.85 .
The order at the top of cohort A follows what was screened. Walmart Connect Sponsored Search, Instacart Sponsored Products and Amazon Sponsored Products each carry a privacy score of 1 for a generic company policy that a completed screen found, while DoorDash's privacy cell was never screened. In the deletion test below, removing that 1 would move Walmart Connect Sponsored Search and Instacart Sponsored Products into Most documented and Amazon Sponsored Products to 50.59, a hundredth below the line, and removing either top row's assurance cell would leave it with no Standing cell at all, a thin record.
The thin records show the gap from the other side: 44 of the 48 have no counted Standing cell, so nothing countable is recorded about who assures their numbers, how they correct them or how they handle shopper data.
DoorDash agreed on 28 October 2026 to acquire Instacart, with closing expected at the end of 2027 or in the first half of 2028; until then the two operate as separate companies, and the agreement changes no row .
What the cells say beyond the bands
Change and correction governance is scored on 7 of the 91 rows, all in cohorts A and B. Under the frozen definition it is scored only after a screen has read both a service's documentation and its terms, and for most rows the advertiser terms, insertion-order terms or agreement were not read (chapter 4). An n/e here records an unfinished screen, not missing controls.
Independent assurance is n/e on most rows: it is scored on 12 of the 47 rows in cohort A, 5 of the 15 in cohort C and none in cohorts B, D and E. No row reaches 5, which needs completed assurance of the outcome metrics or method across the environments sold. The 4s are delivery accreditation in the environments sold, for Walmart Connect Sponsored Search and Instacart Sponsored Products, and the two narrower outcome instruments above; a self-attested alignment with a guideline scores 1 and a listing in an accreditor's process 2 (chapter 7).
Validation is the weakest Depth dimension. No validation cell scores above 2, every scored validation cell in cohort A is a 1 for accuracy claims, and no cohort C platform has one. The default-method cells repeat chapter 3: 53 of the 64 scored are 1, attribution with no counterfactual, and only Moloco and Osmos reach 3. On buyer-controlled analysis only Amazon's four rows reach 4 in cohort A, for Amazon Marketing Cloud, where eligible sponsored-ads advertisers run their own queries over event-level impression and click records and receive aggregated outputs .
Scale is capped by broader parents. In cohort A it is scored on 13 rows, all at 3, and twelve of them rest on revenue lines wider than the service. Walmart's figure, "nearly $6.4 billion, including VIZIO", is global, partly netted against cost of sales and stands on all five Walmart rows in the cohort . Kroger's audited Other revenue line, US$1,129 million in fiscal 2025, mixes third-party media revenue with data analytic services, specialty pharmacy and in-store clinics . Instacart's Advertising and other line, US$1,065 million in 2025, includes coupons , and Albertsons' Other line is led by wholesale sales, commissions and rental income . One figure on several rows of a network is one piece of evidence. Amazon's Advertising services line is company-wide and unallocated, so it scores no Amazon row .
Three cells in the matrix score 5, all US availability in cohort B, each resting on a filing: Cardlytics' card-linked offers and Ibotta's two rows.
How stable the bands are
Under equal thirds, Depth-led (60/20/20), Reach-led (30/45/25) and Standing-led (30/20/50) weights, recomputing Raw and the field each time, positions move by up to 11 places in cohort A, 6 in B and 1 in C, and 32 of the 39 banded rows keep their band under all four. Both Most documented rows hold under every weighting. The rows that move are Amazon Sponsored Brands, Amazon display ads, Roundel and Macy's Media Network in A; Fetch in B, which would be Most documented under the Standing-led weights; and Amazon Retail Ad Service and Swiftly in C.
Removing the shrinkage (K = 0, which reports Raw) changes 3 bands in cohort A, 1 in B and 2 in C. Walmart Connect Sponsored Search, Instacart Sponsored Products and Amazon Sponsored Products would join DoorDash in Most documented, as would Cardlytics and Fetch in B and Criteo Commerce Yield in C. At K = 2 the changes are 1, 0 and 2. At K = 8 the count in cohort A is 0, while in C Moloco falls to Well documented and Swiftly rises to Partly documented. DoorDash holds at every K.
Deleting one counted cell at a time, with the field held fixed, changes a band in 37 of 362 cases in cohort A, 22 of 115 in B and 41 of 103 in C, often because the deleted cell was a row's only one in a group. Leaving out a low score raises a quotient. The largest rise in cohort A comes from deleting Amazon Sponsored Brands' privacy cell, a 1, which would lift its RQ from 43 to 49 and, since it is the row's only Standing cell, make it a thin record. The test also asks what a cohort would show. Cohort B has exactly the six eligible rows bands need, so deleting the privacy cell of Chase Media Solutions, Fetch, Ibotta or Upside, each that row's only Standing cell, would leave five, and every band in the cohort would be withdrawn; the first version of this test held the cohort's bands fixed and missed it, as the final review found.
The missing-evidence scenarios test the opposite case. Each banded row's n/e cells are filled with 1, then 4, one row at a time with the field and every other row frozen; these are scenarios, not bounds. Of the banded rows with n/e cells, 18 of 23 in cohort A change band under one of them, as do 5 of 6 in B and 9 of 9 in C, and the matrix marks them ±. With 1s in their unscored cells, DoorDash and Moloco would both be Well documented.
Eight scored rows carry an interest mark (◆ in the matrix). The author works at Samba TV, which Amazon Ads' approved third-party providers page lists among certified measurement technologies without tying the list to a product , which Walmart Connect names among its reach and frequency partners , and which Walmart's ad policy lists among approved off-site measurement partners . Those marks sit on Amazon Sponsored Products, Amazon DSP, Walmart Connect Sponsored Search, the Walmart DSP and Walmart Connect Offsite Display. The author previously worked at Hearts & Science, an Omnicom agency, and Omnicom owns Flywheel and Perpetua; as a voluntary disclosure of sister units, Perpetua is marked , as are Instacart's two rows, whose route evidence rests partly on Flywheel and Perpetua documents . The marks alter no score. Neither Most documented row carries one; seven of cohort A's eleven Well documented rows do. Excluding every marked row and recomputing each cohort moves three bands: Roundel, Orange Apron Media and Amazon display ads go from Partly to Well documented, as cohort A's field falls from 38.60 to 37.04. The same exclusion removes Amazon Sponsored Products, Walmart Connect Sponsored Search and Instacart Sponsored Products from the H3 subset, among them H3's one counterexample, and two of the three services left have incomplete searches, so H3 would read unresolved (chapter 5).
The coding rulings
Several cells moved under rulings taken on 3 November 2026, after cell verifiers found anchors applied unevenly across scoring slices; each ruling applies to every row, and chapter 14 logs them. Ruling 1 scores governance only after a screen of both documentation and terms. Ruling 3 codes products sold side by side at different rungs, with no default named, at the rung every product meets. Ruling 5 dates an undated page to the month it was read. Ruling 6 adopts as qualified cells the evidence a cell verifier found and quoted where the scorer had left n/e. Ruling 8 puts a row on the first rung of the default-method dimension when its documentation names an attributed metric in standard reporting and no counterfactual, and leaves a bare "closed loop" n/e. Ruling 10 gives US availability its top rung only when a filing names the row's own service as offered; Cardlytics keeps its 5 on 10-K passages that offer the Cardlytics platform to marketers and maintain it in the US. After the final review two more points were ruled for every row: ruling 8's condition, a metric named in standard reporting, was applied as written, which moved Chewy, Gulp Media Network and Leap to n/e on the default method; and on governance an invoice dispute procedure counts as a challenge route, while a schedule of when numbers settle does not count as a revision.
A scale decision closed the last disagreement. Albertsons' profile proposed n/e for its Other revenue line, which omits media revenue booked against cost of sales ; the session coded it, like Kroger's line, as a broader parent proxy capped at 3, and logged n/e as the sensitivity. With all five residual-line scale cells at n/e and cohort A recomputed, no band moves; the largest change is 0.80 points, Kroger Precision Marketing offsite programmatic rising from 42.52 to 43.33.
What the quotient is not
The RQ does not say whose attributed sales are accurate or whose lift is causal, and a band is not a verdict on a service's quality: a Partly documented row may measure well and publish little, and each Most documented row's single assurance instrument covers less than the service sells. The index rests on one scoring pass under published rules and a check of every counted cell by a separate AI verifier told to refute it; no second scorer re-scored any row, and no inter-rater statistic exists (chapter 14).
The profiles
The paper profiles 8 companies in depth: five cohort A networks (Amazon Ads, Walmart Connect, Instacart Ads, Kroger Precision Marketing and Albertsons Media Collective), one commerce media network (Cardlytics), one enablement platform (Criteo) and one measurement provider (Circana). A profile shows a network's rows together with no pooled score. Each carries a SWOT in which every line cites its sources and is tagged record, vendor documentation, vendor claim or inference.
Deep profiles — select a company
Each card shows the matrix scores as a bar strip. Open a company for its SWOT, what it sells, who buys it, how it earns, its verified capabilities with availability and evidence grade, performance evidence with sponsorship, limits, risks and fit. Pick a SWOT quadrant to isolate it and compare every profiled company on it. Each SWOT line is tagged record, vendor documentation, vendor claim or inference, and opportunities and threats are the author’s reading of dated facts.
Albertsons Media Collective
liveAlbertsons Companies, Inc. (registrant); Albertsons Media Collective is its retail media business ('the retail media arm for Albertsons Companies, Inc. (NYSE: ACI)'), with no separate legal entity named on the pages opened
ROAS is the network's named standard metric and was guaranteed to award winners, matched lift arrives only at scale, no MRC, AAM or IAB Europe listing was found, and the filings give no media revenue figure.
Strengths
- Its staff co-authored a published method-level study of how matched-control iROAS shifts with analyst choices across 42 onsite display campaigns; a party study, but public. vendor documentation
- Names a design per placement class: matched audiences onsite, nearly-60-variable matched stores in-store, and a LiveRamp multi-touch pilot across owned and off-site channels. vendor claim
- Offers randomized testing on request and a reporting API that sends campaign performance data to an advertiser's chosen system, stated as available to all clients. vendor claim
- Parent 10-K: 2,244 US stores in 35 states and DC and 51.2 million loyalty members behind the network's first-party data. record
Weaknesses
- No attribution window, touch model, new-to-brand lookback or sales basis is published for any row, although its 2023 framework urged networks to disclose them. vendor documentation
- No MRC, AAM or IAB Europe listing for any Albertsons service on 3 November 2026; only a self-attested IAB alignment statement for in-store. record
- Media revenue is undisclosed: the 10-K places it inside a US$1,305.6 million 'Other' line and partly in cost of sales. record
- Published case figures ($7.45 and $2.41 iROAS, 5.5% DOOH lift) carry no period, sample, control construction or interval. vendor claim
Opportunities
- Publishing its matching variables, balance thresholds and randomized-test terms would turn the March 2026 study's warning into a documented default buyers can check. inference
- The September 2026 LiveRamp pilot already names Meta, DV360, Google Ad Manager and Perion exposures; publishing its join and method would document an off-site identity loop. inference
- About 800 additional in-store screen stores planned for 2026 would enlarge the pool of matched control stores for in-store tests. inference
Threats
- Its own study shows matched iROAS can change sign with analyst choices, which undercuts the single-number iROAS cases it publishes. inference
- Four differently dated statements on when onsite incrementality began weaken the audit trail buyers need to compare reports over time. inference
- Screens at pharmacy placements alongside an undocumented consent basis for shopper data leave a privacy question buyers and regulators may raise. inference
What is sold
Three placement classes, one register row each. Onsite: display and native ads on Albertsons banner sites and apps, Sponsored Search (S17), sponsored products including inside the app's AI-powered conversational search through Criteo (23 June 2026, S16), launched in 2021 as 'native display and sponsored product inventory' across homepage, search, app and pharmacy placements (S10). Offsite: open-web display through Google DV360, Collective TV (streaming and CTV through DV360 with iSpot, FreeWheel and Clinch, and a DIY channel on The Trade Desk, S13), social (Meta, Pinterest; H-S26, S06) and near-store digital out-of-home on '120,000 digital screens near stores' (H-S25). In-store: digital screens built with STRATACACHE, including deli and pharmacy placements, plus audio (Stingray), signage (Neptune Retail Solutions), sampling (Advantage) and entrance media (Terraboost) (S15, RTL-U-S033, RTL-H-S24). Measurement offers sold or announced: onsite and in-store incrementality (matched designs; randomized testing 'upon request'), a LiveRamp-built multi-touch attribution (pilot, 17 September 2026), a reporting API and a TransUnion marketing-mix feed (S06, S14, RTL-H-S26). A branded-entertainment model with Procter & Gamble was announced on 17 June 2026 (headline only, S07).
Who buys it
Consumer packaged goods brands and their agencies; the 10-K calls them 'vendor partners' (S01). Named in the record: S. Martinelli & Co. (onsite case), Bob's Red Mill (Sponsored Search), Mondelēz (in-store launch brand and case), Primo Water (DOOH), PepsiCo (multi-touch pilot), McCormick (conversational search quote), Procter & Gamble (branded entertainment headline) (RTL-U-S032, S17, S15, S08, S18, RTL-H-S26, S16, S07). The in-store network reported 'more than 50 unique advertising partners' by January 2026 (RTL-U-S033).
How it earns
Media fees under insertion orders (the standard terms define CPM, CPC, CPA and CPL deliverables, S09) and DSP-based buying for off-site. Revenue is recognised either within net sales and other revenue (inside the 'Other' line, with wholesale, commissions and rental income) or as a reduction of cost of sales 'depending on the nature of the media advertising arrangement' (S01); no amount, growth rate or margin is disclosed in the 10-K, the latest 10-Q or the two earnings releases screened (S01 to S04). Prices are not published. A home-page block (inside an HTML comment, not rendered) describes a 'custom ROAS guarantee' for 2026 award winners; it is not used as evidence.
Best fit
- Grocery and CPG brands that want matched-store in-store lift reports alongside onsite and off-site media from one US grocer [RTL-U-S033, RTL-ALBERTSONS-S08].
- Advertisers with their own marketing-mix or measurement models who will take campaign data through the API or TransUnion [RTL-ALBERTSONS-S14].
- Buyers able to negotiate a randomized test and agree its design in writing [RTL-ALBERTSONS-S06].
- Brands buying CTV through The Trade Desk or DV360 that want Albertsons purchase data for targeting and closed-loop reporting [RTL-ALBERTSONS-S13].
Poor fit
- Buyers who need published attribution windows, sales-basis rules or deduplication rules before buying [RTL-ALBERTSONS-S12, RTL-ALBERTSONS-S13].
- Buyers who require MRC- or AAM-assured delivery or outcome metrics [RTL-A-S01, RTL-A-S54].
- Buyers who need random-assignment lift by default with published samples, significance rules and intervals [RTL-ALBERTSONS-S06, RTL-U-S032].
- Buyers who must verify on-site delivery with their own third-party ad server tags [RTL-ALBERTSONS-S09].
Capabilities by dimension, performance evidence, financial record and timeline for this company are in the web edition and in data/deep-profile-records.csv.
Amazon Ads
active; defendant in two pending FTC actions (allegations, no findings)Amazon.com, Inc. (owner; SEC registrant, CIK 1018724). For the United States the Amazon Advertising Agreement names the contracting party as 'Amazon Advertising LLC or, if Customer is a seller under the Amazon Services Business Solutions Agreement, Amazon.com Services LLC' (RTL-AMAZON-S07); Amazon's state privacy disclosure lists 'Amazon Advertising LLC' among controller entities (RTL-J-S32). 'Amazon Ads' is the trading name on every product page opened.
Click-based last-touch attribution by default, buyer-run queries in Amazon Marketing Cloud, and an MRC grant that covers clicks and invalid-traffic filtration only; brand lift is offered for every sponsored format except Sponsored Products.
Strengths
- Documents Sponsored Products windows by account type (7-day sellers, 14-day vendors), halo splits and a 1-, 7-, 28-day restatement schedule in its API documentation. vendor documentation
- Self-service AMC lets sponsored-ads advertisers query event-level Sponsored Products impressions and clicks paired with attributed conversions, with aggregated outputs. vendor documentation
- Only sponsored-products service in the register with MRC accreditation: Display and Search Clicks with SIVT filtration, continued 8 July 2026. record
- Sponsored prompts are explicitly generally available in the US since 25 March 2026 with an itemised report in console and API. vendor documentation
Weaknesses
- Default Sponsored Products outcome is last-touch click attribution with no counterfactual; no controlled-lift offer for advertisers was found on the pages screened. vendor documentation
- Match basis, in-store inclusion, new-to-brand lookback and sales basis are undocumented for Sponsored Products; AMC thresholds carry labels, not numbers. vendor documentation
- Accreditation covers clicks at total level only; impressions and every outcome metric are unaccredited, and Sponsored Brands and display ads hold none. record
- Terms bill on Amazon's own counts, waive fee claims after 60 days and disclaim guarantees about reporting accuracy. vendor documentation
Opportunities
- Its published RCT-calibrated attribution method could support delivered, campaign-level lift if Amazon exposed experiment results to advertisers (inference from a method paper without results). inference
- Extending MRC scope from clicks to outcome metrics would answer the IAB/MRC guidelines' encouragement of third-party audit of outcomes (inference). inference
- The 2027 retirement of version 3 reporting is a natural point to publish sales-basis and match-basis definitions (inference from the scheduled shutoff). inference
Threats
- FTC and 22 states allege undisclosed reserve-price surcharges in Sponsored Products auctions tuned to advertisers' ROAS; filed 31 August 2026, pending. record
- The 2023 FTC monopolisation case alleges search was degraded by paid ads; an adverse finding could constrain ad load (inference; allegations only). inference
- Shorter, unquantified view windows since 1 January 2026 and v1 metric redefinitions make Amazon's numbers harder to compare across time and formats (inference). inference
What is sold
Four scored cohort A rows and two described services. (1) Amazon Sponsored Products (the profile row): cost-per-click ads promoting individual product listings in the Amazon store, for 'professional sellers, vendors, book vendors' plus Kindle Direct Publishing authors and agencies, with 'no monthly or upfront fees' (RTL-AMAZON-S08, RTL-U-S001); since 25 March 2026 also served as 'prompts' in US shopping results and product detail pages that may open a dialog in Rufus, and, per Amazon's page of 11 June 2026, in Alexa for Shopping (RTL-U-S312, RTL-E1-S07). (2) Amazon Sponsored Brands: brand placements in the store billed by CPC, viewable CPM or fixed share of voice (RTL-U-S002). (3) Amazon display ads, formerly Sponsored Display, on Amazon properties and the open internet (RTL-U-S003); from late October 2026 DVA+ combines 'Sponsored Display, Sponsored TV, and programmatic campaigns' (RTL-U-S310). (4) Amazon DSP, programmatic buying of Amazon and third-party inventory with self-service and managed-service options, now 'part of Amazon Ads Agent' (RTL-U-S004). Described, not scored here: Amazon Ads sponsored products across retailers (beta), a listed buyer-side service placing brand campaigns on participating US retailers 'powered by Amazon Retail Ad Service' (RTL-U-S153); and Amazon Retail Ad Service (beta), the cohort C enablement row licensed to multi-brand US retailers and scored separately (RTL-U-S152, RTL-F-S36). Measurement tools that are not rows: Amazon Marketing Cloud (a clean room recorded under dimension 3 for the rows it serves, RTL-X-S03), Amazon Attribution (a 14-day measurement product for non-Amazon media, RTL-E1-S01) and the multi-touch attribution beta (RTL-AMAZON-S06).
Who buys it
Sellers, vendors, book vendors, KDP authors and their agencies for the sponsored ads (RTL-AMAZON-S08, RTL-U-S002); endemic and non-endemic advertisers for Amazon DSP, sold 'to both advertisers who sell on Amazon and those who do not' with no documented split (register; RTL-U-S004). The FTC's 2026 complaint alleges 'approximately 1.2 million U.S. advertising customers, which include over 500,000 small and medium-size businesses' (an allegation, RTL-I-S15 ¶4). No named US buyer with a dated, Amazon-documented production campaign appears on the Sponsored Products page beyond unattributed case-study titles (RTL-AMAZON-S08).
How it earns
Sponsored Products charges per click: advertisers 'pay only for clicks' and bid 'the maximum amount that you're willing to pay when a shopper clicks an ad' (RTL-AMAZON-S08); the Advertising Agreement states 'Fees are based solely on our measurements for the Ad Services and the applicable billing metrics (e.g., impressions or clicks).' (RTL-AMAZON-S07). Sponsored Brands adds vCPM and fixed share-of-voice pricing (RTL-U-S002); prompts are charged 'as part of your CPC bidding and billing parameters' from US general availability (RTL-U-S312). Amazon Marketing Cloud is 'available at no cost to eligible advertisers' (RTL-E1-S02). Revenue is reported only as Amazon.com, Inc.'s global 'Advertising services' net sales line (US$68,635 million in 2025), which also covers display and video advertising sold to publishers and authors and cannot be allocated to any row (RTL-D-S01). The FTC's 2026 complaint alleges that Amazon added undisclosed 'soft reserve' surcharges to its nominally second-price Sponsored Products, Sponsored Brands and Sponsored Display auctions since 2018/2019 (allegation, RTL-I-S15).
Best fit
- Endemic sellers and vendors who need on-site search placement with click-attributed sales and fine-grained reports by search term and product [RTL-AMAZON-S01, RTL-AMAZON-S08].
- Analysts who want to run their own aggregate analyses of on-site exposures and Amazon-attributed conversions in a clean room at no cost [RTL-X-S03, RTL-X-S05].
- Buyers who need an accredited click count and documented invalid-traffic filtration for a search ad product [RTL-A-S07].
- Advertisers testing sponsored formats inside an AI shopping assistant with operator reporting [RTL-U-S312, RTL-E1-S07].
Poor fit
- Buyers who need a documented, randomised incrementality result delivered for their Sponsored Products campaigns [RTL-AMAZON-S06, RTL-AMAZON-S04, RTL-E1-S07].
- Buyers who need event-level exposure data to leave the platform or an independent measurer working on its own copy of the data [RTL-X-S03, RTL-AMAZON-S07].
- Buyers who need a documented sales basis (net of returns and cancellations) or a numeric matchable share [RTL-AMAZON-S01].
- Buyers who need accredited outcome metrics rather than accredited clicks [RTL-A-S07].
Capabilities by dimension, performance evidence, financial record and timeline for this company are in the web edition and in data/deep-profile-records.csv.
Cardlytics
live (operating, publicly listed)Cardlytics, Inc. (UK operations through Cardlytics UK Ltd., RTL-CARDLYTICS-S04)
Card-linked offers are reported on the purchases of shoppers who activated an offer; the network lost Bank of America in February 2026 and received a Nasdaq notice in June 2026.
Strengths
- Matches served offers to partner banks' card and account transactions, in-store and online, at account level rather than through receipts or modelled panels. record
- Filings disclose the network's economics: billings, Consumer Incentives, Partner Share, revenue by pricing model and FI-partner concentration. record
- Three buyer routes are documented: an Ads Manager console, account-team managed campaigns and an Advertiser API with redemption feed and aggregate reporting. vendor documentation
Weaknesses
- Default reporting and served-sale billing are attributed: post-activation ROAS and CPS fees on all purchases by served consumers, with no counterfactual. record
- The test-versus-control report is on request, outside Ads Manager, with no published allocation rule, control size, minimum spend or significance test. vendor documentation
- Terms make Cardlytics' data definitive for fees, bar disclosure to third parties without consent and require deletion of campaign data within 45 days. vendor documentation
- No MRC, AAM or IAB Europe instrument and no published validation cover any Cardlytics outcome metric. record
Opportunities
- Moving the iROAS framework into Ads Manager, as its FAQ plans, would place a counterfactual beside the default attributed report. inference
- Publishing the control design behind 'withholds a representative control group' would let buyers grade its lift claims. inference
- Weekly sales by geography already feed buyers' mix models; documenting that feed could support buyer-run checks of attributed results. inference
Threats
- Revenue fell 36.5% year on year in Q2 2026 after a US partner's exit; Bank of America's relationship ended on 16 February 2026. record
- Top three FI partners took over 85% of Partner Share in H1 2026; Chase can end its agreement on 90 days' notice. record
- Chase and American Express also sell card-linked media themselves, so partners may compete for the same bank channels. inference
- A Nasdaq bid-price notice, US$172.5 million of 2029 notes and Bridg indemnity costs strain the company's finances. record
What is sold
One purchasable service (the register row 'Cardlytics card-linked offers'): cash-back offers and ads served to customers inside partner banks' and other publishers' digital channels (online banking and mobile apps), which the 10-K calls 'the financial media network that we run within our partners' digital channels' and the 10-Q 'our proprietary native bank advertising channel' (RTL-CARDLYTICS-S01, RTL-CARDLYTICS-S03). Advertisers buy through insertion orders with the account team or through Ads Manager, a campaign and reporting console whose accounts the Cardlytics contact creates on request (RTL-CARDLYTICS-S07, RTL-U-S110); demand partners with card-linked offers for local merchants can push merchants and offers and pull reports through the Advertiser API (RTL-CARDLYTICS-S08). Measurement products are child records of the same service: the Ads Manager dashboard (attributed, post-engagement), post-campaign Test vs. Control, Long-Term Value and Share Shift reports, the Nielsen Sales Lift option and weekly impacted-sales data for marketing-mix models (RTL-U-S111, RTL-CARDLYTICS-S09). Until 24 March 2026 the company also sold the Bridg customer-data platform (POS data from merchant data partners), now PAR Technology's (RTL-CARDLYTICS-S01, RTL-U-S116).
Who buys it
Marketers (brands) and their agencies; the 10-K says 'our top five marketers accounted for 20%, 16% and 15% of our revenue' in 2025, 2024 and 2023, with no long-term commitments from most of them (RTL-CARDLYTICS-S01). The API guide adds demand partners 'such as an advertiser, agency, or content aggregator' with card-linked offers for local merchants (RTL-CARDLYTICS-S08). Advertisers are non-endemic to the channel: offers are redeemed at the advertiser's own stores and sites, in-store and online (register inclusion reason; RTL-U-S111). Services are offered in the US and the UK (RTL-CARDLYTICS-S09).
How it earns
Two pricing families (10-Q Note 5, RTL-CARDLYTICS-S03). Cost per Served Sale (CPS): a percentage ('CPS rate') 'of all purchases from the marketer by consumers who (1) are served marketing and (2) subsequently make a purchase' during the campaign period, 'regardless of whether consumers engage'; Cardlytics sets the rate on its 'expectation of the marketer's return on spend' and funds the Consumer Incentive itself. Engagement-based: Cost per Redemption (the marketer funds the incentive and pays a fixed fee) or Cost per Transaction (a fixed fee out of which Cardlytics funds the incentive). Revenue is reported net of Consumer Incentives and gross of Partner Share, a 'negotiated and fixed percentage' of billings less incentives paid to the banks (RTL-CARDLYTICS-S01). Fiscal 2025: billings US$385.0 million, Consumer Incentives US$151.7 million, revenue US$233.3 million (company-wide, including the UK and Bridg), Partner Share and other third-party costs US$102.9 million, Adjusted Contribution US$130.3 million (RTL-CARDLYTICS-S01, RTL-U-S113). In the second quarter of 2026 revenue was US$36.9 million, down 36.5% year on year, of which served-based pricing was 32.0% (RTL-CARDLYTICS-S03; derived). Other revenue includes 'hosting fees that we charge our FI partners' (RTL-CARDLYTICS-S03). Fees are computed on 'Cardlytics' proprietary data', which the CPS terms make 'the definitive data used to calculate the fees' (RTL-CARDLYTICS-S14).
Best fit
- Brands sold in stores and online that want offers placed inside customers' banking apps and paid on purchases matched to cards [RTL-CARDLYTICS-S03, RTL-CARDLYTICS-S09].
- Advertisers who will request the Test vs. Control or Nielsen report and judge it on a disclosed design before relying on it [RTL-CARDLYTICS-S09, RTL-U-S111].
- Buyers running their own marketing-mix models who can use weekly sales by geography with spend [RTL-CARDLYTICS-S09].
- Local-merchant aggregators with card-linked offers who need an API for merchants, offers and redemptions [RTL-CARDLYTICS-S08].
Poor fit
- Buyers who need incremental results by default inside the self-serve report [RTL-U-S110].
- Buyers who need a published randomisation rule, minimum sample or significance test for lift [RTL-CARDLYTICS-S09, RTL-U-S111].
- Buyers who need exposure-level data or independent re-analysis without seller consent [RTL-CARDLYTICS-S14, RTL-E3-S06].
- Buyers who require MRC, AAM or IAB Europe assurance of outcome metrics [RTL-A-S01, RTL-A-S53].
- Advertisers who need product-level (SKU) sales rather than merchant-level card purchases (inference) [RTL-CARDLYTICS-S03].
Capabilities by dimension, performance evidence, financial record and timeline for this company are in the web edition and in data/deep-profile-records.csv.
Circana
active; privateCircana, LLC (completion posts RTL-CIRCANA-S02 and S07; the privacy policy names Circana, LLC as the entity the Circana Group refers to, RTL-J-S27)
Household and market lift studies sold to brands and to the networks they measure, with assignment undisclosed; the DoorDash study cannot be tied to a named Circana service.
Strengths
- Sells household-level tests and market or store tests on point-of-sale data under one Media Lift service. vendor documentation
- Named as measurer or validator by DoorDash, Ibotta, Walmart Connect, Sam's Club Connect and Best Buy Ads on pages current in 2026. vendor claim
- Completed acquisitions of NCSolutions (2 June 2025) and Nielsen's Marketing Mix Modeling business (21 August 2025) put lift, mix modelling and targeting in one team. record
- An independent 2016 USDA report documents the predecessor IRI panel's projection: iterative proportional fitting to Census targets, weights capped at 20. record
Weaknesses
- No allocation rule, minimum sample, significance rule or uncertainty is published for Household Lift or Market Lift; 'A/B testing' is a label. vendor documentation
- No household count, coverage or projection method is published for Media Lift's data; the 200,000-household panel belongs to Complete Consumer. vendor documentation
- Absent from both MRC digital lists, accredited and under review, as screened on 3 November 2026. record
- Networks that sell the media publish its studies (DoorDash, Ibotta, Sam's Club Connect) without naming the payer or the design. vendor claim
Opportunities
- Networks now buy outside validation of their own incrementality (DoorDash, 2 September 2026); a published Household Lift method would let buyers judge it. inference
- MRC's digital under-review list carries Kroger's and Kevel's 2026 pre-audits with incremental-sales metrics but no sales-lift measurer: room for a first audited lift method. inference
- NCS Sales Effect already offers dashboard and clean-room delivery; extending it to Media Lift would give buyers a documented access route. inference
Threats
- Networks certify their own incrementality: DoorDash announced AAM certification beside its Circana partnership, which can narrow the role of outside validators. inference
- Rival measurers publish household counts (Numerator 1 Million+; Catalina 130 million households) while Circana publishes none for Media Lift. inference
- Group privacy policy allows sharing CCPA-sensitive data with media partners; tighter state rules on sensitive purchase data would bear directly on its media products. inference
What is sold
Scored row: Circana Media Lift, two outcome-measurement services on one page: Household Lift ('household-level A/B testing' linking media exposure to household purchases, in-store) and Market Lift (market- or store-level tests on point-of-sale data with 'a thorough matching process', reporting 'sales uplift', 'projected sales impact' and 'ROAS') (RTL-CIRCANA-S01). Circana also sells it as third-party validation to networks (DoorDash, 2 September 2026, RTL-C-S02). Described, not scored (no register row of their own): NCS Sales Effect, the NCSolutions measurement service now sold on circana.com ('incremental sales lift and ROAS'; managed, dashboard or clean-room delivery), with NCS Targeting, NCS Optimization and NCS Enablement in the same portfolio (RTL-CIRCANA-S03); Marketing Mix, an enterprise marketing-mix model on POS and causal data, now including Nielsen's former MMM business (RTL-CIRCANA-S06, RTL-CIRCANA-S07); Liquid Mix, a self-serve marketing-mix platform (RTL-CIRCANA-S05); a Retail Media solution for retailers, brands and advertisers that pairs targeting with 'fast lift measurement' and 'closed-loop measurement' (RTL-CIRCANA-S04); Complete Audiences, purchase-based targeting activated on 'more than 20 partner platforms' (RTL-CIRCANA-S13); and the panel and market data products behind them (Complete Consumer, RTL-CIRCANA-S12).
Who buys it
Brands and manufacturers ('For Manufacturers'), digital advertisers and social-media brand managers (RTL-CIRCANA-S01); publishers and retail media networks (NCS Sales Effect, RTL-CIRCANA-S03); media sellers (Marketing Mix, RTL-CIRCANA-S06); retailers building or evaluating their networks (Retail Media, RTL-CIRCANA-S04); networks offering Circana validation to their advertisers (DoorDash, RTL-C-S02). Named clients in the record: Mars (RTL-CIRCANA-S17) and Chomps via Ibotta (RTL-CIRCANA-S08); network partners DoorDash, Ibotta, Walmart Connect, Sam's Club Connect and Best Buy Ads (findings F11, F12, F15 to F18).
How it earns
Fees for data, studies and software from brands, agencies, retailers, publishers and networks; no price is published (the Media Lift page offers 'flexible plans' and a free trial without terms, RTL-CIRCANA-S01). Circana is private and publishes no revenue; no filing or audited figure was found. Who paid for any network-published Circana study (DoorDash, Ibotta, Walmart Connect, Sam's Club Connect, Best Buy Ads) is not stated.
Best fit
- CPG brands that want an outside party's household- or store-level sales read on a network campaign and accept an undisclosed design [RTL-CIRCANA-S01, RTL-C-S02].
- Networks that want a named third party to validate their own incrementality reporting for selected advertisers, as DoorDash did [RTL-C-S02].
- Brands testing non-addressable or regional retailer campaigns on point-of-sale data by market or store [RTL-CIRCANA-S01, RTL-CIRCANA-S09].
- Buyers who want lift, marketing-mix modelling and purchase-based audiences from one supplier's data [RTL-CIRCANA-S02, RTL-CIRCANA-S07, RTL-CIRCANA-S13].
Poor fit
- Buyers who require random assignment with documented samples, significance rules and intervals [RTL-CIRCANA-S01].
- Buyers who need an accredited outcome metric [RTL-G2-S24, RTL-U-S022].
- Buyers who want to rerun the analysis or inspect the household data themselves, which no Media Lift route documents [RTL-CIRCANA-S01].
- Buyers who need a measurer whose commercial ties to the measured network are disclosed per study [RTL-C-S02, RTL-E3-S23, RTL-CIRCANA-S04].
Capabilities by dimension, performance evidence, financial record and timeline for this company are in the web edition and in data/deep-profile-records.csv.
Criteo
active; redomiciled from France to Luxembourg (completed by 29 July 2026); merger into Criteo Holdings, Inc. (Delaware) agreed 5 August 2026 with an Effective Time of 1 January 2027 unless otherwise agreed, pending shareholder approval and other conditions at 3 November 2026Criteo S.A. (a Luxembourg société anonyme since the redomiciliation completed in July 2026; incorporated in France in 2005; US operating subsidiary Criteo Corp.; planned successor Criteo Holdings, Inc., Delaware)
Commerce Yield documents six of eight attribution parameters and holds an MRC grant for on-site display delivery metrics; Criteo reports retail media as one global segment.
Strengths
- MRC accreditation since 15 March 2024 for on-site display impressions and clicks, including invalid traffic, across desktop, mobile web and in-app. record
- Commerce Yield states default attribution windows (post-click 30 days, post-view 1 day) and default product-match rules, not just the options. vendor documentation
- Identity basis spelled out: four signals, a match order and the undercounting that follows when retailers omit them. vendor documentation
- Wide demand routes: own DSP, 60+ third-party DSPs, an API program with 14 buying partners and Google Search Ads 360. vendor claim
Weaknesses
- No control design is documented for retailers' advertiser campaigns; 'closed-loop reporting' is attribution without a counterfactual. vendor documentation
- Accreditation excludes off-site, video and every outcome metric, yet the 10-K links MRC-accredited technology to measuring 'against product sales'. record
- No match rate, coverage report or sales basis is published, and changed attribution settings rewrite a campaign's whole history. vendor documentation
- Retail Media revenue fell 21% in Q2 2026, including a US$21 million headwind from two unnamed clients' scope changes. record
Opportunities
- Criteo already runs random-split ghost bidding and holdout tests elsewhere; extending them to Commerce Yield would fill its documented gap. inference
- Microsoft's planned preferred-partner role and Google's onsite integration could route new demand and retailers to the platform. inference
- Its published correction route for delivery data could be extended to attributed sales, which no screened page covers. inference
Threats
- Large retailers may in-house retail media technology, as the 10-K warns; the 2025 scope changes show how fast segment revenue can fall. inference
- Rivals are entering outcome assurance (Kevel's MRC pre-audit for conversions and incrementality), while Criteo's accreditation stops at delivery. inference
- A second domicile change, a merger into a Delaware company due 1 January 2027, still awaited shareholder approval on 28 October 2026. record
- Health-data litigation naming Criteo with CVS, a retail media operator, was unresolved at the 10-K date. inference
What is sold
Three register rows. Commerce Yield (the profile row): a retailer- and marketplace-licensed platform that serves Sponsored Products on a CPC model and onsite display and video on a CPM model, extends retailer audiences off-site on the open web, monetizes first-party data, and reports to retailers and their advertisers through dashboards, emailed reports and API endpoints, with an optional white-label private market and professional services (RTL-U-S147, RTL-CRITEO-S03). Commerce Max: a self-service demand-side platform for brands and agencies buying Criteo's retailer network on-site and off-site, with attributed sales, ROAS and new-to-brand reporting (RTL-CRITEO-S16, RTL-G1-S40, RTL-G1-S48). Commerce Grid: a supply-side platform for media owners that 'powers Commerce Growth' (RTL-CRITEO-S03, RTL-U-S148). Criteo also sells Performance Media products (Commerce Growth, GO) outside this paper's rows.
Who buys it
Commerce Yield: retailers, marketplaces and commerce companies running media networks (named US examples in Criteo's releases: Albertsons Media Collective, Dollar General, Belk, BJ's Wholesale Club, Weis Markets, Winn-Dixie, Harveys Supermarkets; RTL-CRITEO-S10 to S12, S15). Commerce Max: brands and agencies (Criteo reports about 33% of Retail Media revenue from agencies; RTL-CRITEO-S03). Commerce Grid: publishers and media owners (RTL-U-S148).
How it earns
'We typically charge retailers a negotiated SSP fee and sometimes a technology fee, while brands pay us a negotiated DSP fee' (RTL-CRITEO-S03). Retail Media revenue is recognised 'primarily ... on a net basis, as we act as an agent', so the segment's US$263.9 million (2025) is Criteo's take, not the media spend passing through; professional services are also billed (RTL-CRITEO-S03). Price lists are not published.
Best fit
- US retailers that want a ready-made on-site sponsored products and display stack with Criteo-sourced brand demand and self-service tools [RTL-U-S147, RTL-CRITEO-S10, RTL-CRITEO-S12].
- Networks whose brand partners ask for an MRC-accredited count of on-site impressions and clicks [RTL-CRITEO-S01, RTL-CRITEO-S21].
- Retailers that can pass hashed customer IDs and emails on every call and want documented attribution defaults [RTL-CRITEO-S22, RTL-CRITEO-S23].
Poor fit
- Networks or brands that need randomised lift as a documented platform feature for on-site retail media [RTL-CRITEO-S20, RTL-CRITEO-S22, RTL-CRITEO-S19].
- Buyers who need accredited outcome metrics or accredited off-site delivery [RTL-CRITEO-S01, RTL-CRITEO-S21].
- Buyers who need buyer-run analysis in a clean room with documented output rules [RTL-U-S147, RTL-F-S45].
Capabilities by dimension, performance evidence, financial record and timeline for this company are in the web edition and in data/deep-profile-records.csv.
Instacart Ads
operating; acquisition agreement pendingMaplebear Inc. (d/b/a Instacart)
Attributed sales within 14 days of a click by default; a randomised lift design is documented only for 2022 and 2023, and Instacart's own accreditation wording reaches beyond its MRC letter.
Strengths
- Published a randomised holdout design for Sponsored Products lift: assignment per user-advertiser pair, ghost ads, limited to Sponsored Products (June 2023 engineering post). vendor documentation
- Documents a 14-day click window, linear credit, a 26-week new-to-brand lookback, halo sales and dated reporting revisions shown to advertisers with notices. vendor documentation
- One MRC letter (1 June 2026) accredits delivery metrics on its marketplace and on Carrot Ads retailer properties through Storefront, Storefront Pro and Connect API. record
- Advertising and other revenue was US$1,065 million in 2025 (audited) and US$331 million in Q3 2026, up 23% year over year. record
Weaknesses
- Default reporting is attribution with no counterfactual; lift is a separate test on request, 'not available to all partners', with no published sample or significance rule. vendor documentation
- No outcome metric of any Instacart row has completed independent assurance or validation; MRC's letter covers impressions, clicks, CTR and viewability only. record
- Sales basis, deduplication, in-store inclusion and matchable share are undocumented, and shoppable video and off-site outcome metrics have no published definitions. vendor documentation
- Revenue is disclosed only as company-wide 'Advertising and other', blending Sponsored Products, display, coupons and Carrot Ads, with no service or US figure. record
Opportunities
- Carrot Ads named five US networks as users by November 2025 and claims 310+ partners; Gopuff's storefront would add Instacart-sourced demand under the same accreditation. inference
- Its documented randomised design could become a default report or an audited incrementality claim, as DoorDash's AAM certification shows is attainable for a peer. inference
- DoorDash says it expects to keep Instacart's brand, enterprise technology and app after closing, which could widen inventory sold to the same CPG advertisers. inference
Threats
- The pending DoorDash acquisition, expected to close at end-2027 or in early 2028, has already suspended guidance and could reshape products and measurement before then. inference
- The Caper Cart wind-down, due by mid-2027, removes the in-store ad surface Carrot Ads offers retailers. inference
- An FTC 6(b) pricing inquiry (October 2026) and a 2025-2026 subscription settlement keep Instacart's data practices under scrutiny, though neither concerns advertising data. inference
What is sold
Three register rows. (1) Instacart Sponsored Products: sponsored product placements on the Instacart Marketplace, billed per click, reported as attributed sales within 14 days of a click with linear credit, new-to-brand sales and ROAS, with a separately requested randomised sales-lift test (S08, RTL-E1-S14, S02). (2) Instacart display and video ads: display, shoppable display, shoppable video and brand-page placements on the marketplace, reported as direct and halo sales after a view or click within 14 days, plus off-platform audiences in The Trade Desk with 'closed-loop measurement' (S09, RTL-E1-S42, RTL-H-S40). (3) Carrot Ads: Instacart's white-label retail media technology licensed to other retailers and platforms, with Instacart-sourced brand demand, on-site placements and in-store Caper cart surfaces (RTL-U-S308, RTL-U-S169). Also sold: coupons and promotions (inside the same revenue line), the Insights Center and the Data Hub clean room (RTL-D-S04, RTL-E1-S13, RTL-E1-S11).
Who buys it
Brands and their agencies, mainly consumer packaged goods brands ('over 9,000 active brands', FY2025 10-K, S17), through the self-serve Ads Manager, the agency API and tools such as Perpetua, Intentwise and Skai (rows 1 and 2); retailers and commerce platforms that license Carrot Ads (row 3), among them Hy-Vee, Sprouts, Thrive Market, Uber Eats and Associated Food Stores (S14, S15).
How it earns
Advertising sold to brands on the marketplace and off-platform, coupons and other online advertising, and Carrot Ads (revenue arrangements with retailers not described), reported together as 'Advertising and other' revenue: US$1,065 million in 2025, 28.5% of total revenue (derived), and US$331 million in Q3 2026, 3.1% of GTV (RTL-U-S026, RTL-INSTACART-S05). Third-party arrangements are recorded gross or net by contract (RTL-D-S04). Pricing and take rates are not published in the sources opened.
Best fit
- CPG brands selling through Instacart that want click-based attributed sales with a documented 14-day window and new-to-brand reporting in a self-serve console (RTL-INSTACART-S08).
- Brands able to request a randomised sales-lift test for sponsored products through their account team (RTL-INSTACART-S02, RTL-U-S309).
- Regional grocers and commerce platforms that want MRC-accredited delivery measurement and Instacart-sourced brand demand without building ad technology (RTL-A-S06, RTL-INSTACART-S14).
Poor fit
- Buyers who need incrementality as the default report, with published minimum samples, significance rules and intervals (RTL-INSTACART-S08, RTL-INSTACART-S02).
- Buyers who need independently validated or assured outcome metrics; only delivery metrics are accredited (RTL-A-S06).
- Buyers who need row-level outcome exports or documented clean-room output rules for their own analysis (RTL-E1-S11, RTL-INSTACART-S08).
- Buyers who need in-store sales included in attributed outcomes; inclusion is not documented (RTL-INSTACART-S08).
Capabilities by dimension, performance evidence, financial record and timeline for this company are in the web edition and in data/deep-profile-records.csv.
Kroger Precision Marketing
active (both register rows live)84.51° LLC operates Kroger Precision Marketing (KPM, 'powered by 84.51°'); KPM is a business, not a separately named legal entity in the sources opened
Attribution by default on self-serve campaigns and matched-control uplift on managed ones; MRC's under-review list names Kroger outcome metrics, and its filings disclose no media revenue figure.
Strengths
- Kroger's 10-K quantifies the match base: over 95% of customer transactions are tethered to a loyalty card, about 63 million households a year. record
- MRC lists Kroger's onsite formats in a pre-audit covering attributed sales, incremental sales, aROAS and iROAS, the only such US retail network entry. record
- Publishes onsite attribution rules: 14-day window, last interaction, click over view, halo by brand and commodity, reporting on the exposure day. vendor documentation
- Documents paths to incremental measurement on both rows: PrecisionView 360 for onsite and cross-channel, mirrored-control incremental sales for Trade Desk campaigns. vendor documentation
Weaknesses
- Default onsite reporting is attribution; lift needs PrecisionView 360, paid and opt-in at US$400K across three channels, with no assignment or uncertainty rule. vendor documentation
- KAP documentation and the 84.51 API Catalog sit behind sign-in; no public API, export or buyer-run analysis of onsite outcomes is documented. vendor documentation
- Own documentation conflicts: two halo scopes, three attribution-model statements, two lapsed-household definitions, and Direct Connect pages 'removed' yet served. vendor documentation
- KPM's FAQ claims certification by the 'Merchant Risk Council (MRC)'; the Media Rating Council lists only a pre-audit in process. record
Opportunities
- Completing MRC's audit would give Kroger the first accredited attributed and incremental sales metrics for a US retail network's onsite formats. inference
- The roadmap's 'Audience A/B testing' (in development, Q4 2026) could support random splits for lift if applied to measurement; nothing says it will. inference
- Kroger told investors on 20 October 2026 it expects to grow retail media revenue; disclosing that revenue would let buyers size KPM. inference
Threats
- KPM computes and reports its own outcomes while its owner calls retail media a high-margin profit pillar; external checks rest on an unfinished pre-audit. inference
- Kroger's October 2026 release still lists litigation over the terminated Albertsons merger among its risks. record
- New surfaces (Shopping Assistant PLAs, KAP Direct Connect) report through tools documented only as 'report builder' or behind sign-in, while legacy reports stay in Prism. inference
What is sold
Two register rows. (1) Onsite search and display (profile row): Product Listing Ads, Targeted Onsite Ads, Promoted Product Carousels and Premium Placements on Kroger Family of Companies sites and apps, bought self-serve in the Kroger Ad Platform (KAP, a 'second-price auction system') or through the managed service; Shopping Assistant PLAs (3 August 2026) extend PLA ad groups into Kroger's AI assistant [RTL-U-S297, RTL-U-S299, RTL-U-S301]. (2) Offsite programmatic: display, streaming, audio and online video activated with Kroger audiences and measurement in partner DSPs (The Trade Desk, Yahoo DSP, DV360 via Direct Connect in KAP) or managed by KPM [RTL-U-S298, RTL-U-S302, RTL-U-S304]. Measurement products: KAP onsite attribution (14-day halo), offsite in-flight attributable sales, incremental sales for qualified Trade Desk campaigns, PrecisionView 360 (paid cross-channel incremental study) and PrecisionView Lite (free attribution-only cross-channel report) [RTL-KROGER-S05, RTL-KROGER-S08]. Described, not rows: managed owned channels (email, push, targeted digital coupons), social and influencer campaigns, in-store digital screens (reported) and 84.51°'s Collaborative Cloud for data scientists [RTL-KROGER-S04, RTL-H-S44, RTL-KROGER-S12].
Who buys it
Brands and agencies advertising products sold at Kroger, which the 10-K calls 'advertising partners'; 84.51°'s boilerplate names 'consumer packaged goods companies, agencies, publishers and affiliated partners' [RTL-KROGER-S11, RTL-KROGER-S12]. KAP is 'Now available to all advertisers' [RTL-U-S299]; managed-service campaigns run through an account team [RTL-KROGER-S04]; KPM's FAQ says non-endemic brands can activate KPM audiences through DSPs, with no separate service documented [RTL-U-S300]. No named buyer was read.
How it earns
Onsite media sold by auction in KAP ('second-price auction system') or managed [RTL-U-S299, RTL-KROGER-S04]; offsite media through DSPs or managed [RTL-U-S296]; PrecisionView 360 is a paid study priced through the account representative, PrecisionView Lite is free [RTL-KROGER-S05, RTL-KROGER-S08]. Kroger books 'third-party media revenue' inside 'Other sales' (US$1,346 million in fiscal 2025, with plant, pharmacy-service and data-analytics sales) and counts retail media among 'alternative profit streams' that 'contributed $1.5 billion of operating profit in 2025'; KPM's own revenue is not disclosed [RTL-KROGER-S11].
Best fit
- Brands sold at Kroger that want onsite search and display bought self-serve in KAP with loyalty-household attribution and a published halo rule [RTL-U-S299, RTL-KROGER-S05, RTL-KROGER-S11].
- Advertisers with US$400K or more across three or more channels who want a cross-channel incremental study and accept modelled test-versus-control [RTL-KROGER-S05].
- Buyers who value a network whose outcome metrics are in an MRC process, accepting that it is a pre-audit [RTL-X-S02].
Poor fit
- Buyers who need randomised lift as the default report, with documented assignment and uncertainty [RTL-KROGER-S05, RTL-U-S304].
- Buyers who need buyer-run, event-level analysis of onsite exposures or public API documentation [RTL-KROGER-S02, RTL-KROGER-S18].
- Buyers who need completed accreditation of outcome metrics now [RTL-X-S02, RTL-X-S01].
- Small budgets that want an incremental read: PrecisionView 360 needs US$400K across three channels and legacy Direct Connect iROAS US$150K [RTL-KROGER-S05, RTL-KROGER-S01].
Capabilities by dimension, performance evidence, financial record and timeline for this company are in the web edition and in data/deep-profile-records.csv.
Walmart Connect
operatingWalmart Inc. (Walmart Connect is the name of its U.S. retail media business; no separate legal entity for Walmart Connect is named on the pages opened)
One deduplicated last-touch model across Walmart's media since 1 June 2026, with history not restated; MRC accredits Sponsored Search delivery metrics, and none of its 137 API report elements is a lift metric.
Strengths
- Developer documentation defines Sponsored Search attribution element by element: 3-, 14- and 30-day click windows, separate in-store metrics, same-brand halo, one-year new-to-brand lookback. vendor documentation
- MRC accredited Sponsored Search impressions, clicks, CTR and invalid-traffic filtration across desktop, mobile web and in-app on 26 September 2025. record
- Onsite Display's API reports test-versus-matched-control lift with group sizes and significance levels beside fair-share or last-touch attribution. vendor documentation
- Four documented Sponsored Search routes: self-serve Ad Center, full-service and reporting-only APIs, partner network and managed service. vendor documentation
Weaknesses
- Sponsored Search's default is click-based attribution; its API lists no lift metric, and the 2025 test-and-control offer documents no assignment, eligibility or result. vendor documentation
- Touch model, match basis, deduplication and returns treatment for Sponsored Search attributed sales are not stated in the developer documentation read. vendor documentation
- Advertising revenue appears only as a global, VIZIO-inclusive earnings-release approximation; no 10-K or 10-Q carries the phrase. record
- No product-specific privacy notice: the shopper notice is company-wide and the Walmart Connect policy covers business users only. vendor documentation
Opportunities
- The June 2026 attribution upgrades and an announced unified model across Search, Display and DSP could align parameters across rows if Walmart publishes the rules. inference
- Publishing the Search Incrementality design would move the profile row past attribution; Walmart already published a third party's incrementality method in October 2026. inference
- Completing Onsite Display's MRC audit, in process since November 2025, would give a second Walmart row accredited delivery metrics. inference
Threats
- Publicis's pending LiveRamp acquisition, under a proposed FTC order of 19 October 2026, could alter terms of the clean room Walmart uses for Meta measurement. inference
- FTC 6(b) orders to Walmart on personal data in pricing (28 October 2026) put loyalty and purchase-data use under study; an inquiry, not an allegation. inference
- Identity-graph migration and June 2026 attribution changes may break comparability of Walmart outcome series, with change details behind an unreadable help article. inference
What is sold
Six purchasable services in the register, profiled here without pooling: (1) Walmart Connect Sponsored Search, the profile row: Sponsored Products (cost per click, auction on relevancy and bid), Sponsored Brands and Sponsored Videos on Walmart's site and app, reported on click-based attribution with separate online and in-store metrics (RTL-WALMART-S01, S10); (2) Onsite Display (Gallery, Marquee, Brand Box, Skyline, Tile; auction pricing; fair-share or last-touch attribution and test-versus-matched-control lift reports) (RTL-U-S009, RTL-WALMART-S15); (3) the Walmart DSP, a self-serve seat 'Powered by The Trade Desk' for open web, social, CTV, audio and gaming, plus private-marketplace live linear TV and Spotify deals (RTL-U-S010, RTL-H-S08); (4) Offsite Display, Walmart's managed off-site media on VIZIO Home Screen, CTV and social (Meta self-serve; TikTok and Pinterest managed), measured through a LiveRamp clean room for Meta (RTL-H-S06); (5) Store Ads: digital screens, TV Wall, self-checkout, Walmart Radio, demos and sampling in U.S. stores (RTL-U-S011, RTL-WALMART-S06); and (6) a non-endemic offer (cohort B) selling the DSP, Offsite Display audiences and in-store placements to brands that do not sell at Walmart (RTL-U-S015). VIZIO placements are child records of Offsite Display. Sam's Club Connect (formerly Sam's Club Member Access Platform) is a separate Walmart-owned network with its own register row and is described here only in narrative (RTL-U-S019, RTL-U-S020).
Who buys it
Walmart suppliers (first party) and Walmart Marketplace sellers (third party), who sign in to Ad Center with Supplier Center or Seller Center credentials, and their agencies and technology partners in the Walmart Connect Partner Network (API partners with full-service or reporting-only access) (RTL-WALMART-S01, S12); managed advertisers (RTL-C-S06); and, for the non-endemic row, brands in 'automotive, financial services, insurance, quick-service restaurants and travel' (RTL-U-S015). No buyer is named here beyond those on Walmart's own pages.
How it earns
Media sales: Sponsored Products on cost per click through an auction that weighs relevancy and bid, with minimum bids published for Sponsored Brands ($0.50 from 19 May 2025) and Sponsored Videos ($0.50 from 23 July 2026); Onsite Display on 'a new, programmatic buying model with auction pricing'; DSP, off-site, store and non-endemic placements on terms not published (RTL-WALMART-S01, S14, S04). Walmart reports a global advertising business of 'nearly $6.4 billion, including VIZIO' for fiscal 2026, recorded 'in either net sales or as a reduction to cost of sales' depending on the arrangement (RTL-U-S017); no service-level revenue is disclosed. Measurement products (Search Incrementality, lift reports, brand lift) carry no published price.
Best fit
- Suppliers and sellers who need on-site search ads reported on Walmart's own online and in-store transactions with documented click windows and new-to-brand lookback [RTL-WALMART-S10].
- Buyers who want MRC-accredited impression and click counts for Sponsored Products, Brands and Videos [RTL-A-S05].
- Technology partners and agencies that automate bidding and reporting through a documented API with full-service or reporting-only access [RTL-WALMART-S12, RTL-WALMART-S13].
- Display buyers who want a per-campaign test-versus-matched-control lift readout alongside attribution [RTL-WALMART-S15].
Poor fit
- Buyers who need randomised, holdout-based incrementality for Sponsored Search by default, with a published design and significance rule [RTL-WALMART-S10, RTL-C-S06].
- Buyers who want to run their own outcome queries on Walmart exposure and sales data in a clean room for on-site search [RTL-WALMART-S13].
- Non-endemic brands that need sales outcomes for their own products, which Walmart's documented closed loop does not measure [RTL-U-S015, RTL-E3-S26].
- Analysts who need a disclosed, audited revenue line for Walmart's U.S. retail media or for any single service [RTL-U-S017, RTL-WALMART-S18].
Capabilities by dimension, performance evidence, financial record and timeline for this company are in the web edition and in data/deep-profile-records.csv.
Chapter 13
What to ask before you buy
On the public record read by 3 November 2026, retail media sells a number its seller computes: attributed sales by default, controlled lift as an offer whose allocation is rarely disclosed, and accreditation that covers delivery metrics wherever it can be mapped to a service. Of the nine hypotheses registered with decision rules, two are supported, two are contradicted and five are unresolved, most of them because the record is silent. A buyer can test any network with seven questions before the money moves.
Attribution is the default: 36 of the 37 retailer and marketplace services that document one report attributed sales with no counterfactual, so H1a is supported . Of 22 documented lift offers, 21 leave the assignment of test and control groups unknown; H1b is descriptive and carries no verdict . Every Media Rating Council (MRC) accreditation mapped to an eligible service covers delivery metrics, but MRC's one outcome-type entry, "Epsilon", is tied to a service outside the register only by inference, so H2 is unresolved under its decision order and supported only on that inference . H3 is contradicted at its first step: Amazon Sponsored Products advertisers can run their own queries in Amazon Marketing Cloud over impression and click records and take out only aggregates . No US series splits spending by placement, so H4a is unresolved . Three of 13 off-site services document that the ad is matched to the purchase outside the retailer's own surface, so H4b is contradicted, as it is on every stricter reading . Amazon itself reports on its two prompt formats that are generally available on agent surfaces, so H6a is supported ; whether anyone else can attribute, test or verify such formats is mostly undocumented, so H6b to H6d are unresolved .
Seven questions for a network
-
What does the default report count, and over what window? Ask for the touch model, the window after a click or a view, the sales basis and the new-to-brand lookback. Amazon Sponsored Products counts sales 7 days after a click for sellers and 14 for vendors, and reports no conversions from views ; no Amazon column read says whether sales are net of returns or cancellations . Ask whether the method changed within your comparison period: Walmart moved Sponsored Search to one deduplicated last-touch model on 1 June 2026 and did not restate history (chapter 3).
-
Is lift offered, on what terms, and how were the groups assigned? Of the 22 documented lift offers, two say a request is needed, two state an eligibility rule and one discloses random assignment . That one is Instacart's per-user holdout with ghost ads, documented on pages from 2022 and 2023, with no significance rule, minimum sample or interval . Ask who qualifies, at what spend, how units were allocated and how uncertain the result is. The method moves the number: in Albertsons Media Collective's study with Ovative Group, a campaign's highest and lowest iROAS (incremental sales per ad dollar) under different matching choices differed by 6.5 times on average (median 2.5 times) (chapter 3, chapter 8).
-
Who produces the number, and how is it restated or disputed? Wherever outcome reporting is documented, the company that sold the media produces the default number on its own definitions, and no service documents a separation between selling and measuring . The seller's count can also be the bill: Amazon's Advertising Agreement says "Fees are based solely on our measurements" and requires fee claims within 60 days . Ask when a figure stops moving, how a correction is notified and whether the invoice follows it; Amazon restates conversions 1, 7 and 28 days after the event . Ask how to challenge an attributed figure: Macy's Media Network's terms make its ad server's reporting final and not subject to challenge (chapter 4).
-
What can I query, and what leaves the clean room? H3 was contradicted by Amazon Sponsored Products, whose advertisers can write their own SQL in Amazon Marketing Cloud over one record per impression and per click, paired with attributed conversions . The loop is narrow: only aggregated outputs leave, under per-column thresholds whose values are not published, and the pairing is Amazon's own attribution, so a buyer's query adds no counterfactual . None of the six pre-registered services documents event-level exposure data for the buyer or an independent measurer . Ask which tables you can query, at what thresholds, what you may export and to whom, and whether the contract lets an outside measurer use it; Amazon's agreement makes all ad services data Amazon's property (chapter 5).
-
What is accredited, by whom, for which metric and environment? On MRC's lists read on 3 November 2026, the seven eligible services with completed accreditation hold it for delivery and invalid-traffic metrics, and the one outcome-type entry, "Epsilon", names no product . Scopes are narrower than product names: Amazon's Sponsored Products letter covers clicks at total level , and Instacart's pages describe its accreditation more broadly than MRC's letter does . Kroger's attributed and incremental sales are in pre-audit, a status, not an accreditation . Ask for the letter and check its date, service, metrics, environments and reports against MRC's current list. Of a certified lift method, such as DoorDash's under the Alliance for Audited Media, ask whether it assigns customers at random; the certificate does not say (chapter 7).
-
How does off-site exposure meet the purchase? Of 13 retailer and marketplace services that tie outcome measurement to an off-site channel, 3 document that the ad is matched to the purchase through a clean room or another match outside the retailer's logged-in surface , and none of the three names its match keys . Kroger and Roundel name a clean room for Meta campaigns but no purchase outcome . Ask where the join runs, on which keys, what share of your sales could be matched and whether any off-site lift used a control. In stores, ask for the impression formula and how test and control stores were chosen: Walmart calls its store-ad impressions estimated projections and gives no formula , and Albertsons does not say how its control stores were allocated (chapter 6).
-
What does the agent surface report? Of eleven sponsored formats on agent surfaces, the two stated generally available, Amazon's prompts, come with Amazon's own report of sales, return on ad spend (ROAS) and 7-day orders and units . None documents explicit general availability of third-party attribution, causal measurement or ad-quality verification, and most say nothing either way . Google folds its AI Overviews and AI Mode ads into aggregate campaign reports . Ask whether the agent placement is reported apart, which window applies, and whether anyone but the seller can attribute, test or verify it (chapter 11).
The author works at Samba TV, which Walmart Connect and Amazon list among their measurement partners . The Amazon Sponsored Products, Walmart Connect Sponsored Search and Walmart DSP (demand-side platform) rows named here carry the interest mark. Instacart names Omnicom's Flywheel as an agency partner, and his past employer, Hearts & Science, was an Omnicom agency; that disclosure is voluntary . The marks change no code or score (chapter 14).
What would settle the open questions
- H2. An MRC record naming the product behind the "Epsilon" entry would settle it: a service outside the register would make H2 supported, while Epsilon Retail Media, a register row, would make the entry a candidate counterexample . A letter accrediting the outcome metrics in Kroger's or Kevel's pre-audit would contradict H2 . Both MRC lists are re-read on 20 November 2026 (chapter 7).
- H4a. A US series that splits retail media into on-site, off-site and in-store spending on one definition from 2023 to 2026, published or found behind a gate and registered by amendment .
- H4b. Two more credits would leave it unresolved and six more would support it ; Kroger or Roundel would gain one by naming the purchase outcome their clean-room measurement joins . The Walmart DSP is one of the ten rows that do not document the join, and Walmart's identity graph, announced on 3 April 2026 to reach it over time, states no match method yet (chapter 6).
- H6a to H6d. H6a rests on the subtitle of one Amazon page and would fall to unresolved if Amazon qualified it . H6b to H6d need the partner axes documented format by format: six of the eleven records shown as none or pilot would support each, six generally available would contradict it (chapter 11).
- H1a, H1b and H3. H1a would move only if many networks made a lift design what their standard report delivers . H1b carries no verdict, and one counterexample decided H3. What stays open is disclosure of the allocation, sample and interval of lift offers, and of an event-level export at any of the six pre-registered services .
The author's positions, tested
Three positions in the author's own writing were registered as rival hypotheses, not evidence (chapter 1). His standards page says every MRC accreditation in retail media covers delivery metrics ; on the 3 November screen every accreditation mapped to a service does, and the unmapped "Epsilon" entry leaves H2 unresolved . His essay of 22 August 2026 argued that the measurement seam returns when a retailer's audience is bought through a third-party DSP against an outcome on a third-party site . Ten of the 13 off-site services do not document how exposure meets the purchase, so the record neither confirms nor refutes it . His essay Measurement for Agentic Commerce speculated that networks would ship "agentic iROAS" before any neutral standard exists . Amazon's prompt reports, the only generally available ones found, are attributed, and no format has causal measurement generally available, so the speculation is untested .
Chapter 14
How this was done
AI agents gathered the evidence, separate AI verifiers tried to refute it, and one scoring agent per slice scored the services, under hypotheses, a product register and scoring rules frozen on 3 November 2026 before any claim was extracted. Every later change, including the coding rulings taken once extraction had begun, is dated and logged. The process is published so that a reader can check the paper rather than trust it; it is not an audit.
AI research agents wrote 15 workstreams, 1 vendor-universe sweep and 8 deep profiles under registers frozen on 3 November 2026. A separate AI verifier, told to refute, checked each workstream and profile file: of 1123 checks, 460 needed a correction and 29 found the item wrong. Findings no verifier reached are labelled unresolved, not supported. One agent per slice scored 91 service rows, and only cells a separate cell verifier confirmed or qualified count. Rulings made after extraction began are dated and logged; one recoded a single row and moved H4b from contradicted to unresolved . The corrections ledger holds 609 entries. The author works at Samba TV, which Amazon Ads and Walmart Connect list as a measurement partner, and co-leads a working group of the body that maintains the AdCP protocol; interest marks never alter a score.
What was registered before the research
The hypotheses, the product register and the scoring rules were frozen on 3 November 2026, after the vendor-universe sweep had built the register and before any claim was extracted or any service scored. The hypothesis register (data/hypotheses.json) holds 15 entries: nine hypotheses with decision rules (H1a, H2, H3, H4a, H4b and H6a to H6d), one descriptive proposition with no verdict (H1b) and five coded questions (Q1 to Q5). Each hypothesis states its population, comparator, observable outcome and the rules for supported, contradicted and unresolved. Missing evidence reads unresolved, never contradicted, except in H2, H3 and H4b, which test what the public record documents. H1a also fixed in advance the one service that stands for each network, so that splitting a network into rows cannot tilt the network-level count .
The product register lists every service as scored, listed or historical, with its operating entity, owner, cohort, inclusion reason and US availability. Rows are purchasable services, not networks, and evidence never transfers between rows. A product found after the freeze could enter only through a versioned amendment. There was one, version 1.1 of 3 November 2026: NCS Sales Effect, which the Circana profile found offered under its own brand, became a cohort E row before any of its scores was seen . It changed no hypothesis population, and the frozen original is kept for a sensitivity run. The register holds 132 records, 22 of them listed and 3 historical.
The scoring rules (data/scoring-rules.json) fixed five cohorts by buying task, each with a mask over 12 dimensions: retailer and marketplace media networks (A), commerce media networks (B), enablement platforms (C), cross-network optimisation platforms (D) and outcome measurement providers (E). Every applicable cohort and dimension has a written ladder from 0 to 5, which a script confirmed complete before the freeze. The scale thresholds were set from the distribution the universe sweep reported, and every revenue line carries a perimeter code: a line that spans other networks or non-advertising revenue caps the cell at 3, and Amazon's company-wide "advertising services" line never scores. Another AI model reviewed the brief behind all three in three rounds on 2 and 3 November 2026, before any evidence or score existed.
How the evidence was gathered and checked
The research was AI-assisted throughout. AI research agents, each given the brief, the shared research rules and the frozen registers, wrote one structured file per workstream, 15 in all. A coordinating AI session briefed the agents, merged their files and took the coding decisions logged below; its own reads of pages that decide a hypothesis, such as the Media Rating Council's (MRC) accreditation lists, form workstream X . Eight deep profiles went further on Amazon Ads, Walmart Connect, Instacart Ads, Kroger Precision Marketing, Albertsons Media Collective, Criteo, Cardlytics and Circana.
Each file records sources with their dates, funding and access limits, and findings with the supporting passage, its locator, an evidence class, confidence and limitations. Outcome numbers carry a measurement record coded from what the source discloses, never from a method's name: a holdout is not randomised unless the design says how units were allocated. 3 registered sources are the author's own essays and pages, labelled and never counted as corroboration.
Opened has one meaning: the agent read the page itself, through the fetch tool, through curl with a browser user agent, in a browser, or as an Internet Archive capture or a syndicated reprint, and the source record names the route. A search snippet is a lead, never support, and an access error is never read as absence. Of 1058 source records, 222 are partial extracts or abstracts only, and claims on them are restricted to what was read. Hosts that refused automated fetches were read in the desktop app's built-in browser pane, among them Amazon Marketing Cloud's developer pages, the Federal Trade Commission's releases and Walmart Connect's help centre .
Every workstream and profile file then went to a separate AI verifier that had not seen the research and was told to refute it; the universe sweep had none. Verifiers opened the cited sources themselves. Across 23 verifier files, 1123 checks of findings, table rows and profile fields found 582 items supported as stated, 460 supported with a correction, 29 contradicted and 52 unresolved. A finding's final wording is its verifier's correction where one exists. Workstream E1, for example, read the 80% confidence footnote on Walmart's store-ads page as a significance rule; its verifier found that it qualifies one proof-of-concept figure .
Of the 1108 claims in the ledger, 737 were checked by a verifier. 437 stand as stated, including the ten hypothesis readings, which synthesise checked claims; 263 stand with a qualification; 11 were contradicted, and the text uses what the verifier found instead. The other 397, mostly the universe sweep's status records and findings no verifier reached, are labelled unresolved rather than counted as supported.
How the services were scored
Scoring agents each took one slice of rows and scored every applicable cell from the frozen ladders and the verified evidence; a contradicted finding could not support a cell. A cell is scored 0 to 5 against its cohort's ladder, then capped at 4 for vendor documentation and at 3 for vendor marketing. A 0 needs a source documenting absence or failure; missing evidence is n/e, and n/a comes only from the cohort's mask.
A separate cell verifier for each slice, told to refute, re-read each scored cell's sources against its ladder, marked it verified, qualified or unverified, and scanned the n/e cells for evidence the scorer had missed. Only verified or qualified cells count. Across the 91 admitted rows, 660 cells are counted scores, 427 are n/e and 5 are n/a, and 0 are left unverified. A further 16 rows had fewer than four counted cells and are listed, not admitted.
The Receipt Quotient
The Receipt Quotient (RQ) summarises, within one cohort, how much measurement capability a service documents in public. Its dimensions form three groups: Depth, what the number is (dimensions 1 to 5); Reach, where it works (6 to 9); and Standing, whether a buyer can trust and trade it (10 to 12). A group score is the mean of the row's counted cells in that group; n/e, n/a and unverified cells are skipped, never counted as zero. Raw = 0.45 Depth + 0.25 Reach + 0.30 Standing, renormalised over the groups present.
Each row is then shrunk toward its cohort's field mean F, the mean Raw score of the cohort's admitted rows: Shrunk = (k × Raw + K × F) ÷ (k + K), where k is the row's number of counted cells and the prior strength K is 4, so a row with few counted cells sits near the field. RQ is 20 × Shrunk, on a scale of 0 to 100, rounded only for display. Bands come from the unrounded distance to the field score, 20 × (Shrunk − F): Most documented at +12 or more, Well documented at +1 or more, Partly documented at −8 or more and Least documented below that.
A row is admitted with at least four counted cells and is band-eligible with at least 6 and one in every group its cohort's mask applies; a cohort shows bands only with 6 eligible rows. 3 of the 5 cohorts qualify and 39 rows carry a band; the optimisation and measurement cohorts, with 2 and 2 eligible rows, show quotients without bands. The weights and K are the author's choices, not estimates, so the build reruns the index under four other weightings and with K at 0, 2 and 8, deletes each counted cell in turn, and fills each row's n/e cells with 1, then 4; chapter 12 reports what moves.
RQ measures documented capability and nothing else: not measurement quality, product performance, data accuracy or effectiveness. It is computed and banded within a cohort, never ranked across cohorts, and a network receives no pooled score. Shrinkage limits a reward for omission without removing it: a service that leaves a weakness unrecorded can outscore one that discloses it, and the deletion test reports that effect row by row. RQ values do not compare with the author's other quotients, which score other cohorts on other dimensions.
Rulings made after extraction began
The coordinating session took the decisions below on 3 November 2026, after extraction had begun, because agents or verifiers had coded similar evidence differently or a later read found evidence an earlier tally had missed. None changed a threshold, a hypothesis or the register's inclusion criteria; each applies to every row it touches, is a ledger entry and keeps the session's number. The last of them followed the final AI review, described under Limits, later the same day.
The H1a rule. The first codes of documented defaults were uneven: the verifier of workstream E2 found Chewy Ads, which names lifetime-value ROAS and new-to-brand metrics, coded undocumented while DICK'S Media, with one closed-loop sentence, was coded attribution . A default now counts as documented when the row's own documentation says what the standard report delivers, either an attributed metric or a lift design. A bare "closed loop" claim naming no metric is a term only, undocumented in the primary reading and attribution in a lenient one. After the final review the rule was applied as written to three rows whose pages list attributed metrics among what they offer without naming a report: Chewy, Gulp Media Network and Leap are undocumented, their dimension-1 cells are n/e, and H1a reads 36 of 37 . H1a is supported under both readings .
Ruling 8 draws the same line for dimension 1: a row scores 1 when its own documentation names an attributed metric in standard reporting and no counterfactual for the default, while a lift label with no design neither raises nor blocks the cell and a bare term is n/e. It replaced an earlier ruling 2 that also required a screen finding no lift offer. Ruling 3 covers rows that sell products or routes at different rungs and name no default: only the rung all of them meet counts. Kroger Precision Marketing's self-serve onsite formats report attribution by default and its managed campaigns report uplift against matched control households, so the Kroger rows are coded at attribution .
The H4b rule. After the H workstream raised the boundary, the session fixed what documents an identity match outside the operator's logged-in surface, before the count. The row's own documentation, or its executive quoted on record in a partner's release, must state that the off-site exposure-to-purchase join runs through a clean room, a partner or device graph, or a hashed-identifier match inside a partner platform. Network-level privacy statements, match keys without a stated join and pages that do not cover the off-site service do not count .
Ruling 9 recoded one row under that rule. Later the same day, the chapter 6 draft surfaced a roadmap entry that the Kroger profile had recorded and the H workstream's tally had missed: 84.51° marks as completed in the first quarter of 2025 a feature that "Enables KPM Kroger Precision Marketing to measure the impact of media on Meta in a secure clean room environment" . It is the same kind of statement as Walmart's credited sentence on closed-loop measurement through LiveRamp's clean room for Meta , so Kroger's off-site programmatic row was credited; the rule did not change. The count moved from 4 to 5 of 13 rows and H4b from contradicted to unresolved . The final review then pointed out that the entry measures "the impact of media" and names no purchase, while Walmart's sentence names closed-loop measurement of online and in-store sales . Applied as written, the rule needs the purchase side of the join named, so Kroger's credit was withdrawn, and Roundel's with it, since its executive's words name guest insights, not sales . The count fell to 3 of 13 and H4b to contradicted, and every stricter reading agrees .
Ruling 1 scores governance (dimension 11) only after a completed screen of the service's documentation and its terms; a screen that skipped the terms leaves the cell n/e, even where it had been scored 1. Ruling 5 dates a cell resting on an undated page by the month it was read, November 2026. Ruling 6 adopts an n/e cell that a cell verifier documented with a quoted passage from an existing finding as qualified, never verified, marked as added by the verifier; eighteen counted cells entered the matrix this way. Ruling 10 gives the top rung of US availability (dimension 9) only when US production use and availability to new US buyers are documented for the row itself; a filing sentence about the network or the company leaves new-buyer availability to vendor pages, which cap the cell at 4. Cardlytics keeps its 5 on 10-K passages that offer the Cardlytics platform to marketers and maintain it in the US . After the final review two governance points were ruled for every row: an invoice dispute procedure counts as a challenge procedure, because the counts a service bills on are among its own numbers, and a schedule of when numbers settle is not a documented revision. Cardlytics' governance cell fell from 2 to 1 on the second point .
The residual revenue lines. Kroger and Albertsons report media revenue only inside residual lines that also hold non-advertising revenue, such as pharmacy services at Kroger and wholesale and rental income at Albertsons, which books part of its media revenue as a reduction in cost of sales . Where the Albertsons profile had coded scale n/e, the session applied the perimeter code literally to both companies, as broader parent proxies capped at 3 with no media figure prorated, and logged n/e as the sensitivity: with all five cells at n/e and cohort A recomputed, no band moves and the largest change is 0.80 points (chapter 12).
The Epsilon mapping. H2's decision order lets support stand only when every applicability check is resolved. The first draft treated the Epsilon entry's mapping, an inference from Epsilon's release of 2021, as resolved and reported H2 supported. The final review pointed out that MRC's current letter names no product, so the check is open and H2 is unresolved; the supported reading is kept as the sensitivity .
The consolidation record. Checking the final review's point about LiveRamp found Publicis Groupe's closing release of 2 November 2026, Catalina's dated post saying it had joined Infillion, and, in DoorDash's 10-K, a 2025 purchase of Symbiosys that the consolidation search had missed; Foursquare's undated "has joined" banner was not taken to date a close . The quotient's deletion test now also reports when deleting a cell would leave a cohort below the six eligible rows bands need (chapter 12).
What the checks changed
Every substantive change after a verification pass is an entry in the corrections ledger (data/corrections-ledger.json), recording what was questioned, by which check, the verdict and the change. It holds 609 entries: every qualification and contradiction a verifier returned, every cell a cell verifier changed or added, and every ruling above. 93 entries record an item found wrong, and 91 concern scores.
Limits
One scorer per slice. Each cell was scored once and then checked by a verifier that read the scorer's rationale. No second scorer worked blind, so no inter-rater statistic exists, and the quotient carries that limit.
A model checking a model is not an audit. The research agents, verifiers, scorers, cell verifiers and the coordinating session are all AI models. A verifier told to refute catches misreadings and overreach, but it can share the blind spots of the agent it checks, and two agents reading one page are not two measurements. On 3 November 2026 a separate AI review, run on a model from another provider and given the full package, spot-checked selected claims against primary sources, tested the scoring rules and inspected the print output. Every factual point it raised was checked at its primary source before anything changed, and each point and what was done about it is in the corrections ledger. That review is not an audit either: no independent party audited the evidence or the scores, and no journal has reviewed the paper.
Public documentation only. The paper scores what services document in public, not what they do. Contracts, prices, gated help centres and client-only documents were not read, every absence is bounded to the pages screened by 3 November 2026, and a service with a thin public record scores lower than its product may deserve.
Reads from Europe. The browser pane and some fetches reached the web from European network addresses, so some hosts served a European or localised site: Instacart's company pages redirected to its EU site and were read as Internet Archive captures , and Amazon's pages were read on their US English paths . The fetch tool returns at most 125 characters verbatim, so some passages are exact fragments.
Pages that changed. Each cell carries the date of its newest source, and a page whose displayed date changed between two reads was treated as undated. Some changes are on the record: Walmart moved Sponsored Search to a new attribution model on 1 June 2026 without restating earlier figures , and IAB Europe's pages now link only the May 2026 files of its commerce media standards, so what changed since January cannot be stated . Status-sensitive facts were read on 3 November 2026 and are dated in the text. The registered evidence freeze was 20 November 2026. The paper was published on 4 November, so its evidence runs to 3 November; the status-sensitive facts listed for the freeze are re-read on 20 November, and any change is logged as a dated correction.
Interests
The author has worked at Samba TV (TV measurement and advertising data) since May 2025. Amazon's approved third-party providers page lists Samba TV among about sixty measurement vendors, without tying the list to any ad product . Walmart Connect names Samba TV among its reach and frequency partners , and Walmart's ad policy lists it among approved off-site measurement partners . Under the conflict screen frozen before extraction, the rows tied to those listings carry the interest mark: Amazon Sponsored Products, Amazon DSP, Walmart Connect Sponsored Search, Walmart Connect Offsite Display and Walmart DSP. No Samba product is a row in the register, and Samba's purchase-outcome partnerships in the releases screened, with Catalina and Affinity Solutions, do not describe measuring a retail network's media .
He co-leads the Signals & Measurement working group of AgenticAdvertising.org, which maintains AdCP, the Ad Context Protocol, one of the five protocols assessed in chapter 11. AdCP's schema can represent a retail media sponsored placement , its commerce media page describes a reporting basis it calls closed-loop attribution , and its retail media seller specialism is at preview status . At release v3.2.8 the sponsored-placement format remains experimental, gated on evidence from at least two retail media networks . His firm, No Fluff Advisory, operates agents listed in that organisation's registry. The protocol checklist was frozen before any protocol was read and applied alike to AdCP, the Agentic Commerce Protocol (ACP), the Universal Commerce Protocol (UCP), the Agent Payments Protocol (AP2) and IAB Tech Lab's Agentic Advertising Management Protocols (AAMP). Nothing built by the author or his firm is cited or earns protocol credit, and the AdCP findings carry the interest mark.
His earlier employers are Verve Group, Hearts & Science and Lotame. Hearts & Science was an Omnicom Media network, combined with Mediahub into Hearts United on 28 August 2026 . Omnicom closed its acquisition of Flywheel Digital, under which Perpetua sits, on 2 January 2024 , and Instacart's release on off-site measurement names Omnicom's Flywheel as an agency partner . The Flywheel and Perpetua rows and the two Instacart rows therefore carry the interest mark; the marked findings concern sister Omnicom units, not Hearts & Science, so this disclosure is voluntary. Neither Verve Group nor Lotame appears in the product register or in any finding.
Scorers first set the marks unevenly; the session applied the frozen screen to every row on 3 November 2026. Marks are disclosed wherever a marked finding is used, marked rows are excluded in a sensitivity run, and a mark never alters a score. No vendor funded, reviewed or saw the paper before publication, and every vendor is scored under the same published rules. The author's own essays argue positions that this paper tests as rival hypotheses, not as evidence.
The claims
Every claim marked in the text, with its evidence class, review outcome, confidence and supporting passages.
Claim ledger
The 572 claims marked in the text, each with its evidence class, review outcome and confidence. Click an ID to see the supporting passage, scope, derivation and limitations. Headline claims are listed first. The full research ledger of 1108 claims, including those not used in the text, is in data/claim-ledger.csv.
Machine-readable copy with passages and limitations: data/claim-ledger.csv.
| ID | Claim (final wording) | Ch. | Evidence class | Review | Confidence | Sources |
|---|---|---|---|---|---|---|
| Albertsons Media Collective publishes audience figures in incompatible units and no matchable share: 'instant, intelligent access to 175M premium shoppers' (home page), '51+ M' 'loyalty members' (onsite page) and 36+ M customers per week (home page stat tile), none tied to the share of transactions matched to an identity. | 1 | vendor assertion | supported as stated | high | ||
| Inference: the closed loop Walmart Connect documents ties ad exposure to purchases at Walmart ('When a customer sees your ad and makes a related purchase at Walmart, we recognize and report it'), so for a brand that by this row's definition does not sell at Walmart it yields no advertiser-sales outcome; the measurement page does not mention non-endemic advertisers, and neither page opened documents what outcome (visits, brand lift, the advertiser's own sales) such a buyer receives. | 1 | inference | supported as stated | medium | ||
| The identity and transaction bases documented across cohort B fall into six kinds: card or bank-account transactions held by the operator or its bank partners (Chase, Cardlytics, Amex Offers, Citi by its own description, Upside through card digits and merchant loyalty identifiers); payment-account transactions (PayPal); receipts (Fetch, Ibotta's own app); publisher or retailer purchase feeds (Ibotta's third-party publishers); advertiser-supplied conversion events (Rokt, Klarna Search & Compare's tracking code); and loyalty or account audiences with no documented join to an outside brand's sales (MARRIOTT MEDIA, Kinective, Expedia's non-endemic offer, Uber, Lyft, Walmart Connect's non-endemic offer, Zillow). None of the 20 rows discloses the share of an advertiser's transactions it can match. | 1 | synthesis | supported with qualification | medium | ||
| Amazon Sponsored Products: dimension-8 scale figure 68635 US$ million for year ended 31 December 2025 (Advertising services (net sales), with 56,214 for 2024 and 46,906 for 2023; perimeter company_wide_unallocated; audited filing). | 1 | direct record | unresolved | high | ||
| Instacart Sponsored Products (retailer_network): status unconfirmed; operating entity Maplebear Inc. dba Instacart ('© 2026, Maplebear Inc. dba Instacart' on the .eu page served, RTL-U-S024; EDGAR company name 'Maplebear Inc.', RTL-U-S026); owner Maplebear Inc.; DoorDash, Inc. agreement to acquire dated 28 October 2026, pending (RTL-U-S027); US availability unconfirmed on the pages opened — US product pages not readable (geo-redirect); MRC lists 'Instacart' with 'Rendered Ad Impressions: Display. | 1 | direct record | unresolved | medium | ||
| Kroger Precision Marketing (retailer_network): status unconfirmed; operating entity unverified here (seed: 84.51° LLC); both own hosts refused the fetch tool (RTL-U-S030, RTL-U-S031); owner The Kroger Co. (seed; unverified here); US availability unconfirmed on the pages opened (403 ×2); MRC's listing states no geography (RTL-U-S022). | 1 | direct record | unresolved | medium | ||
| AD Retail Media (retailer_network): status unconfirmed; operating entity unverified here (seed: Ahold Delhaize USA); adretailmedia.com served an empty shell (RTL-U-S035) and adusa.com refused (RTL-U-S036); owner Koninklijke Ahold Delhaize N.V. (seed; unverified here); US availability unconfirmed on the pages opened. | 1 | direct record | unresolved | medium | ||
| Uber Eats sponsored listings (retailer_network): status live; operating entity Uber Eats / Uber Technologies Inc. ('© 2026 Uber Technologies Inc.', RTL-U-S042); Portier not named on the pages opened; owner Uber Technologies, Inc. (RTL-U-S042); US availability yes — RTL-U-S042 footnote 1: 'Based on an average of US SMB advertiser data between 3/1/23 and 5/31/23'; US merchant page (/us/en/). | 1 | vendor assertion | unresolved | medium | ||
| Uber Eats sponsored listings: dimension-8 scale figure 1500 US$ million annual revenue run rate (not annual revenue) for Q1 2025 run rate (Uber Advertising 'annual revenue run rate'; perimeter broader_parent_proxy; press release or product page). | 1 | vendor assertion | unresolved | medium | ||
| Uber Journey Ads (commerce_network): status live; operating entity Uber Advertising / Uber Technologies Inc. (RTL-U-S043); owner Uber Technologies, Inc. (RTL-U-S043); US availability yes — RTL-U-S044: 'programmatic clients in the US and global markets can now access' (11 June 2024); T-Mobile screens to expand JourneyTV 't. | 1 | vendor assertion | unresolved | medium | ||
| Chase Media Solutions (Chase Offers) (commerce_network): status live; operating entity JPMorgan Chase Bank, N.A. (RTL-U-S107: Chase Media Solutions marks "are trademarks of JPMorgan Chase Bank, N.A."; RTL-U-S106/RTL-U-S107: "JPMorgan Chase Bank, N.A. is a wholly-owned subsidiary of JPMorgan Chase & Co."); owner JPMorgan Chase & Co. (RTL-U-S105: "JPMorgan Chase & Co. (NYSE: JPM)"; RTL-U-S106 copyright "© 2026 JPMorgan Chase & Co."); US availability yes — RTL-U-S105 title "Connecting 80 Million U.S. Consumers" and "Chase is the U.S. consumer and commercial banking business of JPMorgan Ch. | 1 | vendor assertion | unresolved | medium | ||
| Cardlytics card-linked offers (commerce_network): status live; operating entity Cardlytics, Inc. (RTL-U-S116 "Cardlytics, Inc. (the \"Company\")"; RTL-U-S113); owner Cardlytics, Inc. (public company; CIK 1666071); US availability yes — RTL-U-S110 reporting schedule "US Accounts: ~9:00 AM Eastern Time (America/New_York)"; RTL-U-S111 "Through our partnerships with top b. | 1 | vendor assertion | unresolved | medium | ||
| Carrot Ads (enablement): status live; operating entity Maplebear Inc. dba Instacart (footer, S25; 'Copyright © 2026 Instacart', S26); owner Maplebear Inc. (DoorDash acquisition agreement of 28 October 2026 pending; per the brief's 8-K record, not re-opened); US availability unconfirmed from the pages opened here: the US product page geo-redirected to .eu (S24); S26 is global developer documentation; the brief's . | 1 | vendor assertion | unresolved | medium | ||
| Criteo Commerce Max (optimisation): status live; operating entity Criteo S.A. (seed; the page carries no legal entity; UK notices in footer); owner Criteo S.A. (listed; not re-verified from a filing in this sweep); US availability unconfirmed: RTL-U-S198 claims only 'We are live on 225+ retailers globally'; US retailers served through Criteo's retailer-side network are. | 1 | vendor assertion | unresolved | medium | ||
| NCSolutions (Nielsen Catalina Solutions) (historical): status retired; operating entity historical: NCSolutions, 'the joint venture between Nielsen and Catalina' (RTL-U-S241); owner Circana: agreement 26 August 2024 (RTL-U-S241); completion per Circana's page title 'Circana Officially Acquires NCSolutions' dated 2 June 2025 in a search snippet (body unread, RTL-U-S242); domain ncsolutions.com 301-redirects to circana.com on 2026-11-03 (RTL-U-S238); US availability not applicable (historical). | 1 | direct record | unresolved | medium | ||
| Shopkick (historical): status unconfirmed; operating entity Shopkick, Inc.; owner unverified (no readable operator or docket page); US availability unconfirmed: no operator page readable (RTL-U-S278). | 1 | direct record | unresolved | medium | ||
| Quotient Technology platform (historical): status unconfirmed; operating entity Neptune Retail Solutions (post-acquisition); Quotient Technology Inc. historically; owner Neptune Retail Solutions since 5 September 2023 (RTL-U-S270); Neptune's own owner unverified (no ownership statement on RTL-U-S269 or RTL-U-S270); US availability unconfirmed as a separately purchasable service; the login links establish that the platforms are still served (RTL-U-S269). | 1 | direct record | unresolved | medium | ||
| Kroger Precision Marketing onsite search and display (retailer_network): status live; operating entity 84.51° LLC (Kroger Precision Marketing, "powered by 84.51°"); owner The Kroger Co.; US availability yes: Kroger sites and apps, US; "62+ million verified households" (RTL-U-S300); KAP "Now available to all advertisers" (RTL-U-S299). | 1 | vendor assertion | unresolved | medium | ||
| MARRIOTT MEDIA (commerce_network): status live; operating entity Marriott International, Inc. (Marriott Media); owner Marriott International, Inc.; US availability unconfirmed: global figures ("nearly 10,000 properties worldwide", "271M+ Marriott Bonvoy members"); no US-specific availability statement o. | 1 | vendor assertion | unresolved | medium | ||
| Kinective Media by United Airlines (commerce_network): status live; operating entity United Airlines, Inc.; owner United Airlines Holdings, Inc.; US availability yes: "a U.S.-based traveler audience"; "more than 108 million unique flyers" (RTL-U-S125). | 1 | vendor assertion | unresolved | medium | ||
| Open name settled or bounded: Amazon Full-Funnel Campaigns and DVA+ (row rule): Neither. Both are campaign types inside Amazon Ads Agent with no separate default outcome method, access or assurance documented: Full-Funnel 'riunisce gli annunci sponsorizzati, display, video e per la TV in streaming in un'unica campagna' and is 'ora disponibile per tutti gli inserzionisti negli Stati Uniti'; DVA+ consolidates 'Sponsored Display, Sponsored TV e campagne programmatiche' from 'la fine di ottobre' (RTL-U-S005); RTL-U-S004: 'DVA+ brings the power of programmatic to all advertisers with simplified, outcome-based campaign creation.' Recorded as child records. | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: Walmart Connect VIZIO placements: Child record of Walmart Connect Offsite Display proposed: documented only as an offsite channel ('VIZIO Home Screen ads', RTL-U-S010) and as 'VIZIO's brand solutions suite' with 'closed-loop sales attribution' (RTL-U-S018); no separate default, access or assurance documented. | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: Target Product Ads vs the rest of Roundel: One row: RTL-U-S028 ties no measurement statement to any product and RTL-U-S029 names products without product-level defaults, access or assurance. | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: DoorDash CPG vs restaurant rows: Yes: AAM scope is Sponsored Products and Sponsored Brands only (RTL-U-S038); the managed-service Sales Lift Measurement is for Sponsored Products (RTL-U-S037); restaurant Sponsored Listings use the ghost-ads lift-study design (RTL-U-S039). | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: DoorDash–Instacart: Agreement 28 October 2026, 8-K signed 29 October 2026; pending; outside date 12 months with up to four three-month extensions for antitrust conditions; rows stay separate. | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: Costco — own-domain media service: No retail media or sponsored products page on costco.com found by a domain-restricted search; the only own-domain advertising page is 'Costco.com Ad Specs' (undated; filename 2021), documenting vendor placements sold by guaranteed size on Costco.com, in email and in print, with no service name, fees, measurement or reporting (RTL-U-S070). Costco stays listed. | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: eBay — are General, Priority, Stores and Offsite separate rows?: General differs from the rest (any-buyer click within 30 days, fee on sale); Priority and Offsite share one rule (same-buyer direct sale plus 30-day halo, CPC); Stores attributes any item from the store within 30 days of a click, CPC, no direct/halo split (RTL-U-S074). Four rows proposed; Offsite and Stores flagged as possible children of Priority under a strict reading. | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: DICK'S Media cohort B row: No: the offsite page documents no separate non-endemic product, package or measurement (RTL-U-S088); eligibility only on the homepage FAQ ('No. We work with all brands, on our shelves or not.', RTL-U-S086). The seed row 'DICK'S Media non-endemic offsite' should be withdrawn; the whole offsite service stays in A. | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: TikTok Shop Ads documentation: Recorded under the TikTok Shop Ads entity from RTL-U-S091 (attribution: last touch, clicks over views, 7-day click, 1-day view, unique Shop ID, no product ID) and RTL-U-S096 (gross revenue formula); both 'Last updated: August 2025'. | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: Ibotta: are offers and fixed-price display separate rows?: Yes, two rows: the FY2025 10-K reports "Redemption revenue" (US$297,236 thousand) and "Ad & other revenue" (US$45,153 thousand) with different recognition bases (fee per redemption vs marketing services performed), and the properties page sells "100% pay-per-sale offers" beside a "suite of high-impact ad products"; measurement is documented only for promotions. | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: Affirm: is a paid advertising service documented?: Not on the record read: the 10-K names "featured placements," "personalized advertisements and tailored offers" for merchants in the Affirm Marketplace without pricing, revenue line or non-merchant advertisers; kept listed. | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: Criteo Commerce Grid: separate row?: Yes, separate: different buyers (publishers/media owners, agencies, retailers for 'audience extension and programmatic monetization'), outside the MRC scope named in the 28 March 2024 release ('Commerce Max and Commerce Yield') and no attribution method documented; cohort C fit thin (SSP), session may list it instead | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: Google Commerce Media Suite: Documents named US retailers (Walmart, Costco, Kroger, Albertsons, Dollar General; '30+ leading retailers and marketplaces'), SKU-level reporting 'Available in Google Ads, Display & Video 360 and Search Ads 360', in-store reporting 'Available in Search Ads 360. Coming shortly to Google Ads and Display & Video 360'; no service retailers license from Google; editorial exclusion stands | 1 | direct record | unresolved | medium | ||
| Open name settled or bounded: Incoherent-task rule outcomes (cohort E): Descriptive profiles: NIQ MMT (store-level matched market), Ipsos MMA (MMM; In-Market Testing undocumented), Kantar LIFT ROI (MMM), Measured (geo/audience experiments + calibrated MMM), Haus (geo experiments; random stratified assignment documented), INCRMNTAL (observational continuous), WorkMagic (geo holdout), LiftLab (MMM + testing suite). Coherent E tasks: Circana household-level A/B testing, Numerator exposed/non-exposed on a US household panel, Catalina HUB360 purchase attribution, Nielsen Sales Lift control/exposed on retailer-verified purchase data, Attain exposure-to-transaction linkage; Affinity Consumer Purchase Lift would be coherent if admitted; Incremental undetermined (method undocumented). | 1 | direct record | unresolved | medium | ||
| Albertsons operated 2,244 stores across 35 states and the District of Columbia at 28 February 2026 under 22 banners, reported 51.2 million loyalty members and 'serving 36.5 million customers per week' in its 10-K; the Media Collective's own pages state '175M premium shoppers', '2,240 locations in 35 states', '36+ M customers per week' and '51+ M loyalty members', and its 2021 launch release cited 'over 100 million shoppers' and 'over 27 million members' of Just for U: these are different units and none is a share of matchable transactions. | 1 | direct record | supported as stated | high | ||
| Albertsons Companies announced Albertsons Media Collective on 11 November 2021, 'developed in partnership with CitrusAd and Merkle', opening 'native display and sponsored product inventory' on its websites and apps plus 'off-site targeted ad placements', with brand campaigns to begin 27 February 2022. | 1 | vendor assertion | supported as stated | high | ||
| Albertsons Media Collective's Collective TV, launched 13 June 2024, sells streaming and CTV through Display & Video 360 (with iSpot measurement, FreeWheel and Clinch) and a 'DIY CTV Channel' on The Trade Desk that includes 'item-level closed-loop measurement for on and offline sales', and 'offers flexible attribution windows via The Trade Desk' without stating values or a default. | 1 | vendor assertion | supported as stated | high | ||
| Criteo's Commerce Yield page (re-read 3 November 2026) sells a modular retailer platform: Sponsored Products on a CPC model, onsite display and video on a CPM model (standard IAB ads and 'Proprietary Commerce Display ads'), off-site display and video on the open web with 'Unified Measurement', reporting with 'flexible attribution windows', in-platform and emailed reports and 'API Reporting Endpoints', first-party data monetization through Criteo's OneTag or a CDP, a white-label 'private market' option and Professional Services; demand comes from 'Criteo-sourced demand from 2,500 brands and 185+ agencies' plus 'APIs and non-endemic demand via 60+ third-party DSP partnerships' (the page's FAQ says 3,800 brands). | 1 | vendor assertion | supported as stated | medium | ||
| Kroger's fiscal 2025 10-K states that 'Kroger serves approximately 63 million households annually' and that 'because of our rewards program, over 95% of customer transactions are tethered to a Kroger loyalty card'; KPM's FAQ (read 3 November 2026) cites '62+ million verified households' and its post of 19 July 2024 a '96% loyalty card capture rate'; the figures carry different definitions and dates, and none states the share of onsite ad-attributed purchases matched to a household or how ecommerce purchases are joined. | 1 | synthesis | supported as stated | high | ||
| Walmart Connect's non-endemic page sells the Walmart DSP, Offsite Display behavioural audiences built from 'omnichannel purchase history' and in-store TV Wall, self-checkout, Electronic Counter Screen and audio placements to brands in 'automotive, financial services, insurance, quick-service restaurants and travel', and its only results statement is 'Our data. Industry leading technology. Powerful results.'; inference: the closed loop Walmart documents measures purchases at Walmart, which such brands by definition do not have, and no page read says what outcome a non-endemic buyer receives. | 1 | inference | supported with qualification | medium | ||
| Among Walmart's EDGAR filings the advertising amount is given in the earnings-release exhibits (8-K exhibit 99.1): 'grew approximately 28% to reach $3.4 billion' (fiscal 2024, 20 February 2024), 'grew 27% to reach $4.4 billion' (fiscal 2025, 20 February 2025) and 'grew 46% to nearly $6.4 billion, including VIZIO' (fiscal 2026, 19 February 2026), each footnoted 'recorded in either net sales or as a reduction to cost of sales, depending on the nature of the advertising arrangement'; Walmart Connect in the US is given only as fourth-quarter growth (22%, 24%, 41%). The phrase 'global advertising business' appears in no 10-K or 10-Q (EDGAR full-text search, 3 November 2026) but does appear in Walmart's annual reports to shareholders (form ARS) and proxy statements, whose wording was not read. Only the fiscal 2024 and 2026 amounts carry 'approximately' or 'nearly'; the fiscal 2025 amount is a rounded figure with no qualifier. | 1, 2 | direct record | supported with qualification | high | ||
| Expedia Group, Inc.'s audited 'Expedia Group ("EG") Advertising' revenue was US$483 million (2023), US$639 million (2024) and US$758 million (2025), a network-level line with no endemic/non-endemic split that excludes 'trivago Advertising' (US$417 million in 2025). | 1, 2 | direct record | supported as stated | high | ||
| Walmart Connect Sponsored Search (scale record; applies equally to Walmart Connect Onsite Display, Walmart DSP, Walmart Connect Offsite Display, Walmart Connect Store Ads, Walmart Connect VIZIO placements and Walmart Connect non-endemic advertising): dimension-8 scale figure 6400 US$ million (approximation: 'nearly $6.4 billion') for fiscal year ended 31 January 2026 (Global advertising business, including VIZIO; perimeter broader_parent_proxy; unaudited filing disclosure or earnings release). | 1, 2 | direct record | unresolved | high | ||
| On 29 September 2026 Amazon renamed its unified ad platform 'Amazon Ads Agent' with three campaign types (sponsored ads, Full-Funnel Campaigns, and Display, Video, and Audio), made Full-Funnel Campaigns 'available to all advertisers in the United States' with pooled reporting 'across formats', and announced DVA+, combining 'Sponsored Display, Sponsored TV, and programmatic campaigns', rolling out from late October 2026: campaign types inside the platform, not new rows. | 1, 2 | vendor assertion | unresolved | high | ||
| IAB's DOOH Measurement Guide (July 2025) places 'Retail stores' under a 'Retail and Venue-Based DOOH' subcategory where 'exposure can be linked to real-world conversions via POS systems or loyalty programs', describes matched-store tests as providing 'a closed-loop, scalable way to attribute in-store sales impact to DOOH exposure', refers readers to IAB's 2024 in-store standard for 'Retail Media and Retail DOOH', and says POS data 'through Retail Media Networks' is increasing 'the ability to measure ROI based on exposures'. | 1, 2, 6 | direct record | supported as stated | high | ||
| On the current pages screened, the in-store sellers call the inventory store ads or in-store (retail) media and do not use a DOOH label, but Walmart's own release of 28 January 2021 described its in-store TV walls and self-checkout screens as making Walmart Connect one of the largest in-store and digital out-of-home ad platforms, so the dual label is used by a seller as well as by IAB's DOOH guide; separately, Albertsons sells 'Albertsons-powered Digital Out-of-Home' on '120,000 digital screens near stores' on a page that never says 'retail media', and Roundel sells 'billboards near Target stores', the retailer-data-targeted DOOH that IAB Europe's May 2026 definitions (per the brief) place under off-site. | 1, 2, 6 | synthesis | supported with qualification | medium | ||
| Roundel documents a multi-touch attribution default ('Using a multi-touch attribution model, we distribute credit across the customer touchpoints leading to a purchase') with metrics such as 'New guest ROAS', and a separate incrementality offer using synthetic control, matched market and randomized controlled trials that 'is available for eligible campaigns', with eligibility tied to 'campaign objectives, investment level, and measurement needs' and the Roundel team recommending the approach (thresholds unstated), and describes its Matched Market Model as 'a geo holdout approach'; design per campaign, assignment, minimum and significance rule are not stated. | 1, 3 | vendor assertion | supported with qualification | medium | ||
| 84.51° documents PrecisionView 360 as 'the premier omni-channel incremental sales measurement solution', 'A paid, cross-channel' study for which Cross-Channel Campaigns 'may qualify' if they meet 'a minimum investment of $400K across three (3) or more channels', with an opt-in request (eligibility and availability are discussed with the account representative, not guaranteed); it covers Kroger Ad Platform channels (Targeted Onsite Ad, Product Listing Ad, Promoted Product Carousels and Video Carousels), Prism managed onsite channels (email, push), Targeted Digital Coupons and managed offsite channels (display, online video, Pinterest, audio, CTV, Meta, Snapchat, Disney) as well as The Trade Desk and Yahoo, 'Includes richer methodology (marketing mix / precision Test vs. Control modeling)' and is delivered about five to nine weeks after the last exposure; whether TOA, PLA and carousels count as three 'channels' for a KAP-only campaign is not stated, and assignment, sample and uncertainty rules are not documented. | 1, 3 | vendor assertion | supported with qualification | medium | ||
| eBay Promoted Listings General attributes a sale when any buyer purchases the promoted item within 30 days of any click on its ad, 'regardless of whether the buyer themselves clicked on the ad', with the item promoted at click and at sale; the attributed sale is also the billing event (the ad rate in effect at the time of sale); the any-buyer definition took effect on ebay.com, ebay.ca and cafr.ebay.ca on 13 January 2026, having already applied on ebay.co.uk, ebay.fr, ebay.it, ebay.es, ebay.au and ebay.de (date of that earlier change not given on the announcement), so that from 13 January 2026 'all eBay sites' use it, while the attribution page itself carries no date and still says eBay reports and charges 'differently based on the product and listing site ID'; and such sales appear as 'Attributed sale' in the 'Sale type' reporting column. | 1, 3, 4 | vendor assertion | supported with qualification | high | ||
| DoorDash documented on 24 March 2026 a separate 'Sales Lift Measurement' for Sponsored Products 'based on Ghost Ads' that compares 'a brand's sales among ad-exposed customers versus a non-ad-exposed holdout group', 'currently available for select advertisers via Managed Service for Sponsored Products campaigns', and reports 'over 30% average sales lift for CPG advertisers' from 'DoorDash's 30 Sponsored Product Lift Measurement tests conducted in H1 2025'; random assignment is not mentioned in the release. | 1, 3, 7, 12 | vendor assertion | supported as stated | medium | ||
| IAB Europe's Commerce Media Measurement Standards V2 (body dated January 2026; cover: V2.1, May 2026) assign responsibility to the seller side and route assurance through certification: 'The party providing these metrics to the brand is responsible for adhering to these standards, that can be the Commerce Media Network or an Ad Tech provider that has a direct relationship with the brand. The only way to guarantee adherence to these standards is through IAB Europe's Retail Media Certification.' (p. 4); a CMN 'will need to disclose' which delivery route it uses (CMN reporting, CMN APIs with brand analysis, or clean rooms) (p. 11); 'A CMN doesn't have to provide incrementality measurement to be certified, but if they do, they have to use one of the methods specified in the Guidelines.' (p. 10); media metrics are defined by reference to 'The IAB/MRC definition' of impression and click and the IAB/MRC viewability thresholds (p. 6). | 1, 4 | direct record | supported with qualification | high | ||
| Who produces the Amazon outcome number: Amazon Ads computes and reports attributed sales, ROAS and the modelled long-term sales on its own metric definitions (its 'new methodology' of 1 January 2026; its 12-month new-to-brand rule), with no statement on the opened pages about a restatement or correction policy, a dispute procedure, or any separation between Amazon's sales organisation and its measurement function (silence coded unknown). | 1, 4, 13 | synthesis | supported as stated | medium | ||
| MRC's letter of 26 September 2025 grants Walmart Connect initial accreditation of its 'first-party measurement and reporting of Sponsored Search' for 'Display, Video and Search Clicks, Click-Through Rate (CTR), Impressions and Invalid metrics originating from Sponsored Product, Sponsored Brand, and Sponsored Video ad placement types, inclusive of General Invalid Traffic (GIVT) filtration within Desktop, Mobile Web, and Mobile In-App environments'; MRC's Q3 2025 quarterly update records the same September 2025 grant under the name 'Walmart Connect Sponsored Products service'; the retail media guidelines are not cited. | 1, 7 | direct record | supported as stated | high | ||
| The IAB/MRC Retail Media Measurement Guidelines are dated 'JANUARY 2024' on the cover, carry no version number, and describe the common practice on p. 28 as 'event-level attribution (also sometimes referred to as closed-loop attribution)', adding that 'Closed-loop attribution often involves models that involve the use of extrapolation techniques to show the full impact of a campaign on brand sales'. | 1, 7 | direct record | supported as stated | high | ||
| On p. 25 the guidelines state 'The MRC requires viewable impressions for attribution of outcomes to ad exposures, and this document recommends retail media measurement be compliant with MRC requirements and ultimately, accredited', followed by 'We acknowledge that the majority of retail media organizations may not reflect or consider viewability in attribution today, and we expect all retail media and publisher organizations will need to move in this direction', a sentence about using viewability in attribution, not about measuring viewable impressions; the same two sentences appear on p. 24 of the September 2023 draft. | 1, 7 | direct record | supported as stated | high | ||
| Chapter 4 of the January 2024 guidelines ('Guidelines and Best Practices When Measuring Incrementality') says 'Experimental design and non-experimental approaches form the foundation of any incrementality test', that MRC's Outcomes and Data Quality Standards 'emphasize the importance of randomized controlled trials (RCTs)' and that 'When feasible, marketers are recommended to use RCTs'; it adds (p. 33) that 'all attribution measurements attempt to capture this through a look-alike test and control matching technique' and that 'Holdouts are not always available or appropriate'. | 1, 7 | direct record | supported as stated | high | ||
| The IAB and IAB Europe Guidelines for Incremental Measurement in Commerce Media (IAB page dated 3 November 2025; the PDF body carries no date or version) define incrementality as 'the causal impact of marketing by identifying the additional business outcomes directly driven by a campaign or tactic, compared to what would have occurred in the absence of marketing activity', sort methods into experiment-based (RCTs, 'Hold out / Ghost Ads', matched markets), model-based counterfactual, econometric and 'Hybrid Proxies', list 'platform-reported incrementality' and 'baseline vs. exposed analysis' among the proxies that provide 'directional evidence of lift' rather than 'an independent, true control', and set three requirements for a causal claim: a credible counterfactual or intervention, control of bias, and separation of signal from noise. | 1, 7 | direct record | unresolved | high | ||
| MRC lists 'Walmart Connect Onsite Display (Audit in process as of 11/2025)' for 'Display Impressions, Viewable Impressions and related viewability metrics - Desktop, Mobile Web and Mobile In-App, inclusive of SIVT as performed by HUMAN'; Onsite Display holds no MRC letter, and no Walmart DSP, Offsite Display, Store Ads or non-endemic entry appears on MRC's digital lists read on 3 November 2026. | 1, 7 | direct record | supported as stated | high | ||
| Causal outcome measurement for the Alexa for Shopping prompts is unknown: the 11 June 2026 page promises closed-loop measurement 'that tracks impressions all the way to conversions' and contains no holdout, control, incrementality or lift wording (the fetch tool reported those words absent). | 1, 11 | vendor assertion | supported as stated | medium | ||
| Amazon's unaudited 'Advertising services' net sales were US$19,637 million in the quarter ended 30 September 2026 (US$17,703 million a year earlier) and US$56,689 million for the nine months (US$47,318 million), per the 10-Q filed 30 October 2026. | 2 | direct record | supported as stated | high | ||
| Maplebear Inc. (Instacart) reports audited 'Advertising and other' revenue of US$871 million (2023), US$958 million (2024) and US$1,065 million (2025), company-wide; its accounting policy says the advertising services 'include Sponsored Product ads, display ads, coupons, and a variety of other online advertising services', that third-party arrangements are recorded gross or net 'based on whether it acts as a principal or agent in the transaction, which is assessed on a contract by contract basis', and that other revenue from software subscriptions and other goods and services is 'not material'. | 2 | direct record | supported with qualification | high | ||
| Criteo S.A.'s audited Retail Media segment reported revenue of US$209.0 million (2023), US$258.3 million (2024) and US$263.9 million (2025) and Contribution ex-TAC of US$203.5 million, US$253.8 million and US$259.7 million, two different quantities from an enablement and buying business, not a retailer's network. | 2 | direct record | supported as stated | high | ||
| Cardlytics, Inc.'s audited revenue was US$309.2 million (2023), US$278.3 million (2024) and US$233.3 million (2025), split between the 'Cardlytics Direct' segment (US$285.4m, 255.6m, 212.3m) and the 'Bridg Acquisition' segment (US$23.8m, 22.7m, 20.9m), with US$30.3 million (13.0%) of 2025 revenue from the United Kingdom; the Bridg business was sold to PAR Technology on 24 March 2026, so the perimeter changes from fiscal 2026. | 2 | direct record | supported as stated | high | ||
| eBay Inc.'s fourth-quarter 2025 earnings release states that 'total advertising offerings generated $544 million of revenue in the fourth quarter of 2025, representing 2.6% of GMV' and that 'First-party advertising products on the eBay platform delivered $517 million of revenue in the fourth quarter of 2025, up 19% on an as-reported basis and up 17% on an FX-Neutral basis'; the release gives no full-year 2025 advertising figure. | 2 | direct record | supported as stated | high | ||
| Uber discloses no advertising revenue amount in the documents opened: its Q4 2025 earnings release (4 February 2026) contains no sentence on advertising (confirmed by EDGAR full-text search), its fiscal 2025 10-K mentions advertising but was not read, and the figures found are run-rate statements on Uber's advertising press page ('exceeding $500 million annual revenue run rate', February 2022; 'US$900m revenue run rate in Q4 2023'; 'surpassed $1.5 billion in annual revenue run rate (Q1'25)'), which annualise a quarter and span Uber Eats and rides advertising. | 2 | vendor assertion | supported with qualification | medium | ||
| DoorDash discloses no advertising revenue amount in the documents opened: its Q4 2025 shareholder letter (8-K exhibit 99.2, 18 February 2026) does not mention advertising, and the Q4 2025 press release (exhibit 99.1) mentions 'advertising revenue' only as a driver of growth in accounts receivable, with no amount; the fiscal 2025 10-K (filed 18 February 2026) mentions advertising but was not read. | 2 | direct record | supported with qualification | medium | ||
| Kantar's H1 2026 interim results (reporting entity Kantar Global Holdings S.à r.l.; ultimate controlling party Bain Capital Europe Fund V, SCSp) show Numerator divisional Gross Revenue of US$332.2 million for the six months to 30 June 2026 against a reclassified US$333.4 million (down 0.4%), 'with 9.3% growth in Numerator North America offset by a decline in Numerator International'; during the period the group combined the former Worldpanel and Numerator cash-generating units into 'the Numerator CGU' (the 1 January 2026 date in the report applies to the combination of the Insights and Profiles divisions into Kantar). | 2 | direct record | supported with qualification | high | ||
| Infillion's two 2026 acquisitions carry different wording and neither states a closing date: the Catalina release of 23 February 2026 is headed 'Infillion Acquires Catalina' and says 'today announced the acquisition of Catalina' (present-tense acquisition wording, no closing date, conditions or price), and Infillion's own boilerplate of 18 September 2026 lists Catalina among the technologies it was 'formed by unifying'; the Foursquare release of 18 September 2026 is headed 'Infillion to Acquire Foursquare' and says leadership 'will join Infillion after the transaction closes', so that deal was pending at announcement; no completion post for Foursquare was on the Infillion blog listing by 3 November 2026. | 2 | vendor assertion | supported with qualification | medium | ||
| NIQ Global Intelligence plc priced its IPO on 22 July 2025 (50,000,000 ordinary shares at US$21.00; pricing release) and completed it on 24 July 2025 with aggregate net proceeds of US$985.1 million (10-Q for the quarter ended 30 June 2025, Note 1); the first NYSE trading day (expected 23 July 2025) is stated only as expected in the documents read, though the FY2025 10-K contains the phrase 'began trading'. | 2 | direct record | supported with qualification | medium | ||
| Cardlytics' sale of the Bridg business to PAR Technology has three settled dates: agreement 23 January 2026 (8-K Item 2.01), announcement 26 January 2026 ('originally announced on January 26, 2026') and completion 24 March 2026 ('On March 24, 2026 (the "Closing Date"), the Company, PAR and Buyer completed the Bridg Sale'), for 1,810,222 PAR shares; Rippl is not named in the Item 2.01 text. | 2 | direct record | supported as stated | high | ||
| Two Instacart-related events of autumn 2026 are recorded with different types: the DoorDash-Maplebear Agreement and Plan of Merger of 28 October 2026 (8-K signed 29 October) is a pending acquisition with an outside date of 12 months extendable by up to four three-month periods for antitrust conditions, while the Gopuff-Instacart announcement of 22 September 2026 is a platform partnership (Gopuff to use Carrot Ads technology in the US), not a change of ownership. | 2 | direct record | supported as stated | high | ||
| Four Q3 events remain reported-only or bounded because no settling record could be opened: Microsoft's PromoteIQ wind-down (Digiday, 7 October 2024; a Kevel post asserts a July 2024 shutdown without a Microsoft source), Shopkick's chapter 7 filing (a law-firm digest row dated 5/29/26; docket not opened), Pentaleap's spin-off from Crealytics (brand live by 3 September 2023; no completion date or control perimeter stated) and Macy's Amazon Retail Ad Service pilot (undated Macy's item; Retail Dive, 27 August 2025). | 2 | synthesis | unresolved | low | ||
| Amazon display ads (formerly Sponsored Display) billed on a vCPM basis moved to the same shopping-signal enhanced last-touch view model on 1 January 2026, with 14-day 'all views' metrics retained and click attribution unchanged (the 1 January 2026 page names 'Sponsored Display', the former name, and does not itself say 'display ads'); the display ads product page itself states no window, touch model or counterfactual. | 2 | vendor assertion | supported with qualification | medium | ||
| Ibotta's D2C ad products are sold at fixed prices with no documented outcome method: the 10-K says 'We typically charge fixed dollar amounts for our ad products', defines them as 'Paid digital advertisements such as display ads, tiles, sponsored offers, newsletters', and attaches its measurement language ('Measurement is core to our promotions offering.') to promotions only; no ad-product report, metric or lift offer appears on the IPN pages or 10-K slices opened. | 2 | direct record | supported as stated | medium | ||
| Numerator's corporate levels, recorded as three fields from workstream U's sources without re-opening: operating business Numerator ('Numerator, LLC and Affiliates' in the site footer; 'Numerator is not owned or operated by Kantar' and 'independently managed by Bain Capital' on its own undated page); reporting group The Kantar Group ('consisting of Kantar and Numerator', 'two separate divisions'), whose H1 2026 interim results are issued by Kantar Global Holdings S.à r.l. 'as reporting entity for Numerator | Kantar'; ultimate control Bain Capital Europe Fund V, SCSp ('its ultimate controlling party is Bain Capital Europe Fund V, SCSp, a partnership established in Luxembourg'). | 2 | direct record | supported as stated | high | ||
| Ibotta Performance Network offers: dimension-8 scale figure 297236000 US$ for fiscal year ended 31 December 2025 (Redemption revenue (USD thousands 297,236; of which direct-to-consumer 94,785 and third-party publishers 202,451); perimeter service; audited filing amount). | 2 | direct record | unresolved | high | ||
| Ibotta D2C ad products: dimension-8 scale figure 45153000 US$ for fiscal year ended 31 December 2025 (Ad & other revenue (USD thousands 45,153; all direct-to-consumer); perimeter broader_parent_proxy; audited filing amount). | 2 | direct record | unresolved | high | ||
| Amazon Sponsored Products (scale record; applies equally to the seed rows Amazon Sponsored Brands, Amazon display ads and Amazon DSP): dimension-8 scale figure 68635 US$ million for 12 months ended 31 December 2025 (Advertising services (Net sales); perimeter company_wide_unallocated; audited filing). | 2 | direct record | unresolved | high | ||
| Open name settled or bounded: Sam's Club Connect effective date: Not published on the pages opened. Announced 22 June 2026 with 'will become' (RTL-U-S019); the brand is in use on samsclubconnect.com by 3 November 2026 while the footer privacy link still says 'Sam's Member Access Platform' (RTL-U-S020). | 2 | direct record | unresolved | medium | ||
| Open name settled or bounded: Lowe's previous name: 'Lowe's One Roof Media Network' launched 13 October 2021 (RTL-U-S052); rebranded 'Lowe's Media Network' 'with a new logo and a simplified name' 8 August 2024 (RTL-U-S053); the 2024 release does not state the former name. | 2 | direct record | unresolved | medium | ||
| Open name settled or bounded: Pentaleap spin-off date and control perimeter: Bounded: brand launched by 3 September 2023 (founder post header; page later edited); 29 July 2026 post calls it 'a spin-off from Crealytics' and 'effectively an employee-owned company' with origin 'roughly five years' back; no completion date, cap table or Crealytics stake stated on S13-S15 | 2 | direct record | unresolved | medium | ||
| Open name settled or bounded: Rippl: who operates it now: Unsettled by the parties: Cardlytics' closing release (24 March 2026) and PAR's 8-K release (26 January 2026) do not mention Rippl; the Rippl page carries a bridg.com contact and a cardlytics.com privacy link and no PAR mention. Bridg's 'employees and operations' 'have transitioned to PAR Technology' | 2 | direct record | unresolved | medium | ||
| Open name settled or bounded: Shopkick chapter 7 filing (case 26-10877, District of Delaware): Docket not opened. Reported: Pierce Atwood LLP's Weekly Bankruptcy Alert of 1 June 2026 lists 'Shopkick, Inc. (Boston, MA) | Information | Wilmington (DE) | $10,000,001 to $50 Million | $1,000,001 to $10 Million | 5/29/26' under chapter 7, with no case number (RTL-U-S281). shopkick.com returned no readable text (RTL-U-S278). | 2 | direct record | unresolved | medium | ||
| Event (completion), 2025-08-28: Walgreens Boots Alliance, Inc.; Blazing Star Parent, LLC (Sycamore Partners): Merger completed; the Company survived 'as a wholly owned subsidiary of Parent' ('Blazing Star Parent, LLC, a Delaware limited liability company'); shares converted into '$11.45 in cash, without interest thereon' plus one Divested Asset Proceed Right; common stock and listed notes 'ceased trading on the Nasdaq effective prior to the opening of trading on August 28, 2025' | 2 | direct record | unresolved | high | ||
| Event (announcement), 2024-02-20: Walmart Inc.: 'Walmart agrees to buy VIZIO HOLDING CORP. to further accelerate Walmart Connect in the U.S.' (fiscal 2024 Q4 earnings release, 8-K exhibit 99.1) | 2 | direct record | unresolved | high | ||
| Event (completion), 2023-09-05: Neptune Retail Solutions; Quotient Technology Inc.: 'Neptune Retail Solutions today announced the completion of its acquisition of Quotient Technology Inc.' | 2 | direct record | unresolved | high | ||
| The proposed Kroger–Albertsons merger was enjoined by the U.S. District Court for the District of Oregon on 10 December 2024 (FTC et al. v. The Kroger Company and Albertsons Companies, Inc., No. 3:24-cv-00347-AN); Albertsons sent a termination notice the same day, Kroger sent its own on 11 December 2024 calling Albertsons' letter 'not an effective termination', and Albertsons' release list carries a headline of 11 December 2024, 'Albertsons Files Lawsuit Against Kroger for Breach of Merger Agreement'. | 2 | direct record | supported as stated | high | ||
| Bank of America, which the 10-K calls one of Cardlytics' top three FI partners, sent a non-renewal notice on 22 April 2025; the agreements expired on 31 July 2025 with a wind-down extended to 16 February 2026, when the relationship ended; the undated cardlytics.com FAQ still lists Bank of America and BankAmeriDeals. | 2 | direct record | supported as stated | high | ||
| Cardlytics reports revenue net of the cash back it passes to consumers: in fiscal 2025 billings were US$385.0 million, Consumer Incentives US$151.7 million and revenue US$233.3 million (billings minus incentives, exactly), and Partner Share and other third-party costs of US$102.9 million left Adjusted Contribution of US$130.3 million; billings fell US$58.9 million from 2024. | 2 | direct record | supported as stated | high | ||
| Cardlytics' continuing revenue fell 36.5% to US$36.9 million in the second quarter of 2026 (US down 41.9% to US$29.6 million; UK up to US$7.3 million, 19.9% of the total), and its MQUs fell by 39.0 million to 185.4 million, which the 10-Q attributes primarily to 'an FI partner in the U.S. exiting'. | 2 | direct record | supported as stated | high | ||
| Cardlytics completed the Bridg sale on 24 March 2026 (agreement 23 January, announced 26 January) for 1,810,222 PAR shares, which it later sold for US$23.0 million net of fees; Bridg is presented as discontinued operations for all periods, and on 4 September 2026 Cardlytics agreed to pay Bridg's former CEO an aggregate US$6.4 million to settle his Delaware claim for indemnification linked to Bridg litigation. | 2 | direct record | supported as stated | high | ||
| Cardlytics' finances are strained: accumulated net losses reached US$1,404.1 million by the end of 2025, US$172.5 million of 4.25% convertible notes fall due in 2029, cash was US$28.0 million at 30 June 2026, Nasdaq notified it on 3 June 2026 that its bid price had closed below US$1.00 for 30 business days (compliance date 30 November 2026), and a 1-for-10 reverse split took effect on 5 June 2026. | 2 | direct record | supported with qualification | high | ||
| On 3 November 2026 EDGAR listed no Cardlytics 10-Q for the quarter ended 30 September 2026 (the prior year's was filed on 5 November 2025) and no 8-K after 11 September 2026, so the second-quarter 2026 10-Q of 5 August 2026 is the latest periodic report in this profile. | 2 | direct record | supported as stated | high | ||
| IRI and The NPD Group completed their merger on 1 August 2022 with 'H&F, a premier global private equity firm' as 'the majority investor in the combined company' and Vestar Capital Partners and New Mountain Capital, 'long-term investors in IRI', retaining 'significant investments'; Hellman & Friedman's own portfolio page, served on 3 November 2026, says H&F acquired The NPD Group in 2021 and in 2022 'merged NPD with Information Resources, Inc. (“IRI”), renaming the combined company Circana'. No opened source names Vista Equity Partners as an owner of Circana or its predecessors, and no filing or later ownership statement was found. | 2 | vendor assertion | supported with qualification | medium | ||
| Hellman & Friedman's post dated 7 March 2023 (dateline 'Chicago, IL') announces that IRI and NPD 'today unveiled a new combined company name and brand identity: Circana', and states that Circana 'tracks millions of products spanning 2,000+ categories across 500,000+ stores in 20 countries' and 'advises almost 7,000 brands and retailers worldwide'; Circana's own post of the rename served only its headline. | 2 | vendor assertion | supported as stated | medium | ||
| Circana's post dated 21 August 2025 announces 'the successful completion of its acquisition of Nielsen’s Marketing Mix Modeling (MMM) business', to be 'integrated into Circana's Media team alongside the recently acquired NCSolutions (NCS)'; Circana's California data-broker registration comment states that NC Ventures, LLC merged into Circana, LLC on 30 September 2025, and ncsolutions.com redirects to circana.com (3 November 2026). | 2 | vendor assertion | supported as stated | medium | ||
| Criteo S.A. completed its redomiciliation from France to Luxembourg by cross-border conversion effective 29 July 2026 (the ADS program was terminated and ordinary shares began trading directly on Nasdaq); on 5 August 2026 it entered a Merger Agreement to merge into its wholly owned Delaware subsidiary Criteo Holdings, Inc. at 12:00:01 a.m. New York time on 1 January 2027 unless the boards agree otherwise, subject to shareholder approval, an effective Form S-4 and a U.S. listing (Criteo expects to move to the NYSE); on 28-29 October 2026 the S-4 had not been declared effective and the vote had not been held. | 2 | direct record | supported with qualification | high | ||
| Criteo's audited Retail Media segment revenue was US$209.0 million (2023), US$258.3 million (2024) and US$263.9 million (2025), recognised 'primarily ... on a net basis' as agent, with segment traffic acquisition costs of US$5.5 million, US$4.5 million and US$4.2 million; the segment 'encompasses revenue generated from brands, agencies and retailers for the purchase and sale of retail media digital advertising inventory and audiences, and services', so it spans Commerce Yield and Commerce Max and is a broader parent proxy for either row; the 10-K attributes 2025 growth to Retail Media onsite 'in particular in the U.S. market' and says about 33% of Retail Media revenue comes from agencies. | 2 | direct record | supported with qualification | high | ||
| In its third-quarter 2026 earnings release (8-K, 29 October 2026) Maplebear reported 'Advertising and other revenue of $331 million, up 23% year-over-year, representing 3.1% of GTV', total revenue of US$1,095 million (US$939 million a year earlier), cancelled its 5 November call and suspended guidance because of the DoorDash agreement; no Form 10-Q for the quarter appeared on EDGAR's filing list read on 3 November 2026. | 2 | direct record | supported as stated | high | ||
| Kroger's fiscal 2025 10-K states that 'Alternative profit streams contributed $1.5 billion of operating profit in 2025', a profit figure for a group of businesses that the filing says includes 'data analytic services' and 'third-party media revenue' and, per its executive biography, Kroger Personal Finance; it is neither advertising revenue nor scoped to Kroger Precision Marketing. | 2 | direct record | supported as stated | high | ||
| At its Investor Update of 20 October 2026 (8-K exhibit 99.1), Kroger said it 'expects to grow eCommerce sales and Retail Media revenue' while 'accelerating growth in Kroger Precision Marketing', gave no retail media amount, and listed among forward-looking risks 'the outcome of litigation matters, including those relating to the terminated transaction with Albertsons Companies, Inc.'. | 2 | direct record | supported as stated | high | ||
| Walmart said on 22 June 2026 that 'Walmart Connect U.S., Walmart Connect International and Sam's Club Connect will operate separately' and that Sam's Club Member Access Platform (MAP) 'will become Sam's Club Connect' (no effective date); Sam's Club Connect is therefore a separate network row, and the Walmart Connect rows profiled here are the U.S. services. | 2 | vendor assertion | unresolved | high | ||
| An EDGAR full-text search for 'global advertising business' in Walmart Inc.'s filings returned 27 hits on 3 November 2026: 19 Form 8-K exhibit 99.1 earnings releases (17 February 2022 to 20 August 2026), four DEF 14A and one PRE 14A proxy filings and three annual reports to shareholders (Form ARS, fiscal 2024 to 2026), and no Form 10-K or 10-Q; the fiscal 2026 ARS could not be read (HTTP 403 to curl; over 10 MB for the fetch tool) and the 2026 proxy was readable only to 61,204 characters, so whether either states the dollar amount is not established here. | 2 | direct record | unresolved | high | ||
| Walmart's second-quarter fiscal 2027 earnings release (20 August 2026, quarter ended 31 July 2026) gives growth rates only: 'Global advertising business3 up 38%', Walmart U.S. advertising 'up 38% overall, including 43% increase in Walmart Connect (ex-VIZIO)' and Walmart International 'Advertising business3 grew 20%, led by Flipkart Ads'; no advertising amount is stated. | 2 | direct record | supported as stated | high | ||
| The Ibotta Performance Network's default outcome is the verified redemption, which is also the billing unit: a redemption is 'A verified purchase of one or more items qualifying for a client's offer on the IPN', Ibotta 'matches offers to the specific qualifying products that have been purchased', its platform 'records the number of offers redeemed in any given campaign' and charges a 'fee-per-redemption commission', and brands 'pay only for verified sales'; a redemption count carries no counterfactual. | 2, 4 | direct record | supported as stated | high | ||
| Walmart Connect's 8 April 2026 social article states that 'Closed-loop measurement via LiveRamp’s data clean room is now available for Meta' and 'is also available for TikTok and Pinterest' (no mechanism named for those two), that 'Self-serve access is currently available for Meta and TikTok self-service is planned to rollout this half' (so self-serve is live for Meta and planned for TikTok, resolving the reading flagged in RTL-H-F06), that 'An API for tech partners is also planned', that Add-to-Cart is 'in beta for managed-serve clients', and that its '2.5x stronger incremental ROAS and 1.8x higher sales lift' rest on Circana benchmarks of campaigns that 'qualified for a sales lift' in the Food, Consumables, Beverages or Alcohol categories only. | 2, 5, 6, 14 | vendor assertion | unresolved | medium | ||
| The remaining Q3 events are classified in the consolidation_events table from the sources workstream U opened: Walgreens Boots Alliance take-private completed 28 August 2025 (8-K); Kroger-Albertsons enjoined by preliminary injunction on 10 December 2024 and terminated by both parties' notices of 10 and 11 December 2024 (8-Ks); Walmart-VIZIO agreement 19 February 2024, announcement 20 February 2024, completion 3 December 2024 for net cash consideration of US$1.9 billion (8-K, release, 10-K); Omnicom closed the Flywheel Digital acquisition on 2 January 2024 (issuer release); Neptune Retail Solutions completed the Quotient Technology acquisition on 5 September 2023 (release); Assembly took the Pacvue name on 27 March 2023 (Pacvue post, Italian locale); Kenshoo became Skai on 8 June 2021 (Skai post); Lowe's One Roof Media Network launched 13 October 2021 and became Lowe's Media Network on 8 August 2024 (Lowe's releases); MARRIOTT MEDIA was introduced 12 June 2025 (release via PR Newswire); Sam's Club Member Access Platform 'will become' Sam's Club Connect per the 22 June 2026 Walmart release (effective date unpublished); Amazon Ads Agent was announced 29 September 2026 (Amazon Ads news page, browser pane); Edge launched 7 January 2026 (Ahold Delhaize USA release, browser pane); Kinective Media's 7 June 2024 launch rests on the brief's verification record (United release not re-opened). | 2, 5, 9, 14 | direct record | supported with qualification | high | ||
| The FY2025 IAB/PwC report gives no 2026 figure for commerce media; its only forward statement on it is qualitative, 'Commerce media ad spend is expected to grow 30% faster than the overall advertising market, driven by expanding off-site and in-store opportunities' (p. 15), sourced in endnote 32 to the IAB's own '2026 Outlook Study: A Snapshot of U.S. Ad Spend, Opportunities, and Strategies for Growth' (a buyer-side outlook survey by the report's commissioning publisher, not a third party). | 2, 6 | forecast | supported with qualification | high | ||
| eMarketer's US retail media series page returned HTTP 403 to the fetch tool (one attempt) and its forecast tables were not read; the only eMarketer forecast sentences read, in two free articles opened in the browser pane, are 'Financial media network ad spending will exceed $3 billion by 2028, per our forecast', growing 'at an explosive 77.3% compounded annually between 2023 and 2028' versus 'just 21.2% for the similarly sized travel media network segment' (23 September 2026), and that Amazon's ad business is expected 'to more than double from 2024 through 2028' (10 September 2026), each on an undisclosed definition. | 2, 6 | forecast | supported with qualification | medium | ||
| WPP Media's 'This Year Next Year' pages returned HTTP 404 to the fetch tool (hub page on 3 November 2026 in this stream; hub and mid-year 2026 page in U), and one search surfaced only secondary coverage (an Australian AUD 2.3 billion 2026 retail media figure; a global mid-year growth revision), so no WPP Media retail media definition or US figure was read. | 2, 6 | direct record | unresolved | high | ||
| Neither dentsu nor MAGNA yielded a readable retail media series: dentsu's mid-year 2026 page carries no sentence mentioning retail or commerce media and its report sits behind a form (the 12.3% for 2026 and 11.4% for 2027 growth rates exist only in secondary coverage), MAGNA's forecasts page returned 404 to U and one search in this stream found no MAGNA 2026 US retail media figure, and one search for Advertiser Perceptions or Coresight retail media reports surfaced no 2026 report page (only an unopened Advertiser Perceptions page from March 2024). | 2, 6 | direct record | unresolved | high | ||
| VIZIO entered a merger agreement with Walmart on 19 February 2024 at $11.50 per share in cash, Walmart announced completion on 3 December 2024, and its fiscal 2025 10-K states the Walmart U.S. segment completed the acquisition 'for net cash consideration of $1.9 billion'; VIZIO Home Screen ads are now sold as a Walmart Connect offsite channel. | 2, 6 | direct record | unresolved | high | ||
| No enablement-platform market share with a stated denominator was found: this stream and workstream U opened no publisher series that sizes the enablement (cohort C) market, and the only share claim seen, Pacvue's (cohort D) 'as of 2025, Pacvue manages 12% of total retail media ad spend worldwide, excluding China' (Italian-locale page, author's translation), names no denominator source. | 2, 9 | vendor assertion | supported as stated | medium | ||
| On 11 July 2024 Criteo announced that Microsoft Advertising 'intends to work with Criteo as its preferred onsite media partner', extending Criteo's monetization technology to Microsoft's retailer clients, with the demand integration and onsite collaboration 'expected to roll out in the second half of 2024'; Microsoft's own post the same day says 'we plan to work with Criteo as our preferred partner' for 'our retailer clients' onsite needs' and does not name PromoteIQ; no opened source records the rollout's completion or names a Microsoft retailer client that moved to Criteo. | 2, 9 | vendor assertion | supported as stated | medium | ||
| Among the releases opened, the wording 'previously communicated scope changes with two specific Retail Media clients' first appears in the Q3 2025 release (29 October 2025; the Q2 2025 release of 30 July 2025 carries no such sentence); Criteo then reported Q4 2025 Retail Media revenue down 17% (US$76.3 million against US$91.9 million) and full-year 2025 up 2% 'partially offset by the temporary impact' of those changes, and reported Q2 2026 Retail Media revenue down 21% (US$47.9 million against US$60.9 million) 'reflecting a $21 million headwind' from them, with first-half 2026 segment revenue of US$89.2 million against US$120.4 million; no opened source names the two clients. | 2, 9 | direct record | supported as stated | high | ||
| Maplebear's board decided on 25 October 2026 to 'wind down Caper Cart operations', substantially by the end of Q2 2027, with pre-tax charges of about US$55 million to US$80 million. Caper carts carry two ad offers: Instacart display and video ads, whose eligible shoppable display campaigns have extended automatically to Caper Carts since 14 April 2025 through Ads Manager, and Carrot Ads' in-store surface ('personalized ads on Caper carts'; Associated Food Stores' members 'using Storefront Pro and Caper Carts'). Inference (the 8-K does not mention advertising): both rows' in-store placements are being wound down. | 2, 9 | direct record | supported with qualification | medium | ||
| On 19 October 2026 the FTC announced a complaint and a proposed consent order for public comment on Publicis Groupe's US$2.2 billion acquisition of LiveRamp, requiring for five years non-discriminatory access to LiveRamp's standalone products and firewalls against Publicis's access to LiveRamp customer data; the release describes the acquisition as not yet completed. | 2, 10 | direct record | supported as stated | high | ||
| Amazon.com, Inc.'s audited 'Advertising services' net sales were US$46,906 million (2023), US$56,214 million (2024) and US$68,635 million (2025), a global line that footnote (4) defines as advertising 'to sellers, vendors, publishers, authors, and others, through programs such as sponsored ads, display, and video advertising'; unaudited Q3 2026 was US$19,637 million against US$17,703 million a year earlier (nine months US$56,689 million against US$47,318 million); the line is company-wide and unallocated, so it is context only and never a scale score for any Amazon row. | 2, 12 | direct record | supported as stated | high | ||
| Maplebear's audited 'Advertising and other' revenue was US$871 million (2023), US$958 million (2024) and US$1,065 million (2025), company-wide; the revenue note says the advertising services 'include Sponsored Product ads, display ads, coupons, and a variety of other online advertising services', records third-party arrangements gross or net by contract and calls software subscription and other revenue 'not material'. The note does not name Carrot Ads; that Carrot Ads revenue sits inside this line is an inference (Item 1 of the same 10-K calls Carrot Ads 'our enterprise ads offering'), so the line is read as spanning all three Instacart register rows plus coupons and other revenue (perimeter: broader parent proxy). | 2, 12 | direct record | supported with qualification | high | ||
| Albertsons Companies' fiscal 2025 10-K (53 weeks ended 28 February 2026, filed 27 April 2026) describes the Media Collective as its retail media business but discloses no advertising revenue amount: media advertising revenue is one named component of 'Other' revenue (US$1,305.6 million in fiscal 2025; US$1,203.0 million in fiscal 2024), and media advertising services are recorded either in net sales and other revenue or as a reduction in cost of sales. | 2, 12, 14 | direct record | supported as stated | high | ||
| Kroger discloses no retail media revenue amount in the parts of its fiscal 2025 10-K read (Items 1 and 7): third-party media revenue is one of four named components of 'Other sales', US$1,346 million in fiscal 2025 and US$1,177 million in fiscal 2024, a line that 'primarily relate[s] to external sales at food production plants, other pharmacy services, third-party media revenue and data analytic services', and Kroger attributes the 2025 increase 'primarily' to 'other pharmacy services and third-party media revenue'. | 2, 12, 14 | direct record | supported with qualification | high | ||
| Flywheel Commerce Cloud — Retail Media (Search, Display & Video, Measurement & Reporting) (optimisation): status live; operating entity Flywheel Digital (brand on flywheeldigital.com; privacy policy for its Perpetua product names 'Flywheel Digital Limited and/or its affiliates', RTL-U-S202); owner Omnicom Group Inc. — 'today announced it has closed the previously announced acquisition of Flywheel Digital', 2 January 2024 (RTL-U-S200); US availability unconfirmed: no US office, US client or US-locale statement on RTL-U-S199 ('30+ Global markets', '4.5k clients currently served worldwide');. | 2, 14 | vendor assertion | unresolved | medium | ||
| Catalina’s Italian site, in a post dated 26 February 2026, says "Da oggi Catalina è entrata a far parte di Infillion" ("From today Catalina has become part of Infillion"), three days after Infillion’s release of 23 February 2026 announcing the acquisition; the post confirms the completed state by 26 February but, as the Italian unit's notice, cannot fix the group's legal closing day, and neither source states a closing date, price or conditions. | 2, 14 | direct record | unresolved | medium | ||
| Foursquare’s home page carried, when read on 3 November 2026, an undated banner saying "Foursquare has joined Infillion", linking to Infillion’s announcement of 18 September 2026, which said Foursquare leaders "will join Infillion after the transaction closes"; no closing date was found. | 2, 14 | direct record | unresolved | medium | ||
| DoorDash’s 10-K for 2025 recognises its advertising products "as part of our Marketplaces revenue" and gives no advertising revenue amount; its acquisitions note records that "On May 28, 2025, the Company acquired Symbiosys Corp. (“Symbiosys”), a retail media platform company, to expand offsite advertising capabilities", for purchase consideration of US$121 million. | 2, 14 | direct record | unresolved | high | ||
| Sam's Club Connect documents attribution as its default ('Our membership model links ad exposure to sales across every touchpoint') in a self-service dashboard that lists 'Attributed sales' and 'ROAS', and a separate 'Cross-channel incrementality' offer (Omni-Impact, which names 'Sales lift, including sponsored ads' and 'marginal iROAS') whose '36%' average sales lift, labelled 'incremental sales' on the measurement page, cites a 'Circana FY26 report (1,440 campaigns)'; validating partners are unnamed and no assignment, counterfactual, window or match basis is stated. | 3 | vendor assertion | supported with qualification | medium | ||
| Instacart Sponsored Products' documented default is linear attribution with no counterfactual: since 1 August 2022 Ads Manager and API reporting 'primarily use linear attribution', splitting credit evenly across ad interactions on the path, clicks only when clicks and views mix (from 1 April 2025), with direct and halo sales reported separately; no window in days is stated. | 3 | vendor assertion | supported as stated | high | ||
| Instacart offers sales-lift testing for Sponsored Products, introduced as a 'pilot testing experience' using 'randomized test and holdout groups' (page last updated 2 September 2022) and documented in its 30 June 2023 engineering post as user-level random assignment ('hash of (user_id, advertiser_id)') with ghost ads; Instacart states that 'Lift testing services may not be available to all partners. Please consult with your account team' (5 May 2023). H1b flags: offer documented; request via the account team; eligibility 'not available to all partners' (criteria unstated); lift outside the default report (a separate test); assignment random, user; no minimum, sample or significance rule documented. | 3 | vendor assertion | supported with qualification | medium | ||
| Kroger Precision Marketing offsite programmatic documents an 'in-flight attributable sales' default (since 2023) and, for qualified self-service campaigns through The Trade Desk, post-campaign incremental sales against 'a control audience that mirrors the exposed audience', included in the standard wrap report and subject to a minimum campaign spend (beta testing 'began in March', 2024 by inference from the post date; 'launched in December', December 2024 by inference; post of 14 January 2025); assignment is not disclosed. | 3 | vendor assertion | supported with qualification | high | ||
| Albertsons Media Collective in-store measures in-store campaigns with matched test and control stores: its 6 January 2026 release compares stores with ad exposure against 'a rigorously selected set of control stores with no media exposure' and reports 'incremental sales, orders, and unit lift' (confirmed on the live page only as 'sales, orders and unit lift'), claiming 'a statistically validated measure of true lift' with no threshold, interval or test named, so no significance rule is documented; allocation of stores (random or not), eligibility and whether every in-store campaign receives the study are not stated, and closed-loop measurement is described as a future capability. | 3 | vendor assertion | supported with qualification | medium | ||
| AD Retail Media's Edge platform, whose 'official launch' Ahold Delhaize USA announced on 7 January 2026, 'now powers Onsite Display, Sponsored Search and In-Store Digital Screens across Ahold Delhaize USA's brands' and is said to drive 'incremental return on ad spend' with no method; adretailmedia.com served only a title to the fetch tool, so the default outcome method, attribution parameters, access routes and governance of AD Retail Media are unknown in this stream. | 3 | vendor assertion | supported as stated | medium | ||
| Uber Eats sponsored listings report attributed orders, sales and ROAS on 7-day windows for clicks and, separately, impressions, with no counterfactual; new-to-brand means a first purchase from the brand (lookback unstated), existing-to-brand a purchase in the last 42 days, and customer metrics are not available at placement level; click and impression deduplication is not stated. | 3 | vendor assertion | supported as stated | medium | ||
| Uber's US merchant page bases its '20%' claim on an ongoing global holdout analysis (Uber internal data science, November 2025) that removes '5% of winning ads at random' and compares them with the 95% shown; a platform-wide, global average-effect experiment, not a US-specific or per-campaign lift report. | 3 | vendor assertion | supported with qualification | medium | ||
| Uber Advertising launched 'Sponsored Listings Cross-Channel Incrementality' on 5 October 2026 for US restaurant advertisers, pairing its test-and-control experiments with de-identified, aggregated card-transaction data from Affinity Solutions, and reported sales increases among exposed eaters of 40% on Uber Eats, 27% across measured channels and 5% at brick-and-mortar locations; eligibility, cost, control construction and sample are not stated. | 3 | vendor assertion | supported as stated | medium | ||
| Orange Apron Media documents one closed-loop construct for the network, linking advertising and media spend 'directly to both online and in-store sales and conversions', reported in real-time dashboards and a 'self-service, customizable reporting tool'; its 2024 release names 'single platform attribution and in-store sales for onsite', and no attribution window, touch model, match basis, new-to-brand lookback, sales basis, deduplication or halo rule is stated. | 3 | vendor assertion | supported as stated | high | ||
| Walgreens Advertising Group's pages (all footered 2024, served on 3 November 2026) assert 'closed loop reporting' for offsite media and 'Closed-loop reporting available' for coupons, and report a '20% average incremental sales lift' 'since wag's launch' footnoted on the home page to '68 wagDSP campaigns in 2022', beside a '$5' average ROAS and '3.4%' average conversion footnoted to the same 68 campaigns; no window, touch model, match basis, lift design, assignment or counterfactual is stated on the seven pages read. | 3 | vendor assertion | supported as stated | medium | ||
| UB Media documents one reporting construct for all channels, 'closed-loop reporting across channels' that 'measures campaign-level ROAS', delivered as 'aggregated brand reporting as well as campaign-level details', and its solutions page applies 'comprehensive closed-loop reporting' to campaigns across Ulta-owned properties and off-platform social, programmatic, CTV and streaming audio; no attribution parameter, in-store basis, lift design, verification tag, API or governance term appears on the three pages read. | 3 | vendor assertion | supported as stated | high | ||
| DG Media Network's on-site and off-site solution pages both offer 'Closed-loop reporting and incrementality with our Reporting Dashboard', whose listed sales metrics are Sales Lift, iROAS, Transactions/Buyer Lift, $/Transaction Lift, Conversion Rate and % New-to-Brand Shoppers; lift metrics are listed in 'our Reporting Dashboard' on both solutions pages (marketing pages); neither page says every campaign receives them or that they are the default, and the dashboard's listed sales metrics contain no attributed-sales or ROAS metric, so the attribution half of the construct rests on the term 'closed-loop reporting' and the releases; no assignment, counterfactual construction, window, match basis or significance rule is stated, and the network's releases describe the same construct as 'closed-loop 1:1 measurement' identifying 'both attribution and incrementality' with point-of-sale data integrated. | 3 | vendor assertion | supported with qualification | medium | ||
| Wayfair's supplier advertising page names Sponsored Products ('pay-per-click'), Sponsored Shops and display ads but documents no outcome metric, attribution rule, dashboard or measurement access; its only result claim is one partner's '715% more views' and '520% more orders' with no design, so the default method, lift offer and access record are coded unknown. | 3 | vendor assertion | unresolved | medium | ||
| Gopuff Ads documents one measurement sentence, attached to sampling ('closed-loop post sample customer analysis'), lists 'Data & Insights' without describing it, and poses reporting questions in its FAQ without answers; from the 22 September 2026 partnership its storefront ads are to run on Instacart's Carrot Ads technology, which Instacart describes as extending 'accredited measurements', with no date for the switch and no Gopuff-specific parameter. | 3 | vendor assertion | supported as stated | medium | ||
| DICK'S Media states on its onsite and offsite pages, in identical words, that it tracks 'every touchpoint with closed-loop measurement', promises 'transparent, verifiable results', shows case labels of '$12+' ROAS for On and '$4+' ROAS for Under Armour, and names its first-party loyalty data as the source; no window, touch model, match basis, lift design, dashboard, access route or verification policy is documented, and the service is sold to brands 'on our shelves or not' without a separate non-endemic measurement. | 3 | vendor assertion | supported as stated | medium | ||
| Petco's retail media page offers self-service CPC sponsored product, display banner and offsite ads with 'a custom reporting dashboard' showing 'ROAS, Sales, CTR, ROI and more', and lists 'ad influenced store sales' as a roadmap measurement capability, so in-store sales are not documented as included today; no window, match basis or lift design is stated. | 3 | vendor assertion | supported as stated | medium | ||
| Staples Media Network's page (read on 3 November 2026; it had refused the fetch tool on 2 November) promises only 'sophisticated measurement, brand & business insights, quarterly business reviews/monthly reports' across nine placements on Staples, Quill and Staples Business Advantage, with programmatic display using 'first-party cookie data'; no outcome metric, attribution rule, lift offer or access route is documented, so the default is coded unknown. | 3 | vendor assertion | unresolved | medium | ||
| Among the 27 E2 rows, 4 document a controlled-lift or lift offer on their own pages (Best Buy Ads Sponsored Products; Best Buy Ads display, video, search and CTV media; Orange Apron Media; DG Media Network), Chewy Ads names 'incrementality measurement' without a design, Gulp describes lift reporting in the future tense, Lowe's Ipsos measurement rests on reporting only, and the rest document none on the pages screened; no row documents a request rule, minimum spend, sample size or significance rule on its own pages, the only eligibility statement being reported (Lowe's iROAS reporting 'to top spenders', Marketing Dive); no row documents assignment for any offer; and only DG Media Network lists lift metrics in its Reporting Dashboard, without saying they are the default. | 3 | synthesis | supported with qualification | medium | ||
| Albertsons Media Collective's Bob's Red Mill case reports, for a Sponsored Search campaign run August to December 2023, 'a 66% increase in return on ad spend (ROAS) and a 9% increase in click through rate (CTR)' and 'double digit year-over-year dollar growth (3x over overall category benchmark)', with no ROAS definition, window, baseline period or counterfactual. | 3 | vendor assertion | supported as stated | medium | ||
| Albertsons Media Collective's onsite lift offer, announced 16 April 2026 as 'the launch of onsite incrementality measurement' that 'isolates the causal impact of media from purchases that may have occurred regardless' (a vendor assertion), uses 'test and control frameworks to isolate the effect of media exposure' and, per its results page, 'advanced matching methodologies grounded in predictive purchase behavior to construct credible test and control groups'; no page opened states allocation, matching variables, balance metrics, sample rules or uncertainty; on whether reports include lift, the 23 March 2026 post says only that advertisers 'will see' matched reporting on onsite display 'when programming reaches the appropriate scale' (scale undefined), and the onsite offerings page ties incremental measurement to display ads, with no measurement statement for sponsored products or search. | 3 | vendor assertion | supported with qualification | medium | ||
| On the pages screened by 3 November 2026, none of Albertsons Media Collective's three rows documents a value for any of the eight attribution parameters (click and view windows, touch model, match basis, in-store inclusion rule, new-to-brand lookback, sales basis, deduplication rule, halo rule); the only parameter statements are 'flexible attribution windows via The Trade Desk' (offsite, 2024), 'on and offline sales' for the TTD closed loop, a 'deduplicated' multi-touch view (pilot) and store-level 'sales, orders and unit lift' for in-store. | 3 | synthesis | supported as stated | medium | ||
| Amazon's Reporting API v1 counts 'Halo' conversions, for 'highly relevant products that were not directly promoted in the campaign', inside base metrics, maps the Sponsored Products 'Other SKU' fields to halo, and lists multi-touch variants of purchases and sales whose 'Program eligibility' is 'N/A' on 3 November 2026, while base purchases and sales are eligible for Sponsored Products, Sponsored Brands, Sponsored Display, Sponsored TV, Amazon DSP and Full-Funnel. | 3 | vendor assertion | supported as stated | high | ||
| Since 1 January 2026, Amazon Sponsored Brands and display ads campaigns billed on viewable impressions, and Amazon DSP campaigns serving ads in the Store, report Purchases, Sales and ROAS under a 'shopping-signal enhanced last-touch attribution model' with 'a shorter attribution window' that Amazon does not quantify, keep 14-day 'all views' metrics as a parallel series, and leave click-based attribution, and so Sponsored Products, unchanged (US among 23 countries). | 3 | vendor assertion | supported with qualification | high | ||
| Amazon defines new-to-brand by a 12-month lookback (Sponsored Brands page: 'all within a 12-month look-back window'; long-term sales page: no conversion with the brand 'in the last 12 months'; version 3 columns: a shopper 'who had not purchased a product from the brand in the last 12 months'), but no new-to-brand or long-term-sales column lists a Sponsored Products report type on the version 3 columns page read on 3 November 2026, so a new-to-brand lookback is not documented for the Sponsored Products row on the screened record. | 3 | synthesis | supported as stated | medium | ||
| Amazon's reporting includes a modelled outcome announced on 15 October 2024 for US 'Sponsored Ads and Amazon DSP customers' (then in closed beta, with US open beta promised 'later this year'): long-term sales (LTS) and LTS ROAS are 'ad-attributed measures that estimate the incremental sales value your brand can expect to generate over the next 12 months'; the version 3 field longTermSales defines it as the 12-month potential value of new-to-brand engagements 'plus the 14d total sales of the campaign' and lists Sponsored Display, Sponsored Brands, Sponsored TV and DSP campaign reports, not Sponsored Products. | 3 | vendor assertion | supported with qualification | high | ||
| Instacart's Ads Manager documentation (last updated 17 September 2026) defines Sponsored Products 'Attributed sales' as 'How many dollars in sales resulted from purchases occurring within 14 days of a customer clicking on a sponsored product ad', new-to-brand sales against a 26-week lookback (available from 1 August 2022), ROAS only as comparing 'ad cost vs revenue', viewable impressions at 50% for one second, updates 'within 24 hours', CSV exports, and excludes Storefront and Storefront Pro from viewable impressions, clicks and CTR; it states no halo, view-through, in-store or lift metric. | 3 | vendor assertion | supported as stated | high | ||
| 84.51°'s Report Center page (last updated 17 August 2026) documents the attribution rules for onsite campaigns in the Kroger Ad Platform: a 14-day window ('Shoppers have 14 days to purchase an attributable product after a click on a Carousel or PLA ... or a view on a TOA'), 'the click is always prioritized over the view', 'the last interaction is prioritized for attribution', halo credit ('All Kroger Ad Platform products follow Halo attribution', for sales 'under the same Brand and Commodity selected in the Ad Group'), and KAP reporting that 'shows sales for the exposure on the day it occurs'; the operator's example shows that KAP reporting places the ad cost on the exposure date and the attributed sale on the purchase date. | 3 | vendor assertion | supported with qualification | medium | ||
| 84.51°'s own help pages state the attribution model three ways: the KPM Managed Service FAQs (last updated 20 October 2026) say '84.51° Prism reporting uses a "last-touch" or "last exposure to the ad" attribution model' with a 14-day window, the Report Center says managed-service owned-channel campaigns use a '"first-to-last" exposure attribution methodology', and for offsite campaigns that 'Unlike Onsite, there is NO attribution priority order' and 'Each view of the advertisement resets the 14-day rolling attribution window'. | 3 | vendor assertion | supported as stated | medium | ||
| Walmart's Sponsored Search developer documentation names no touch model, match basis, cancellation or return treatment or deduplication rule; its help centre does (Overview of attribution, article 000012578, read in a browser on 3 November 2026): since 1 June 2026 a unified, deduplicated last-touch model across Sponsored Search, Onsite Display, Walmart DSP and self-checkout store ads, with clicks the only eligible Sponsored Search touchpoint (before that date, last click not deduplicated with Display); online orders cancelled within 72 hours removed and returns not deducted (before 1 June 2026 Sponsored Search deducted neither); full credit to one campaign; halo of same-brand items (and, where applicable, the same Marketplace seller ID); first- and third-party sales combined. Match basis and the treatment of taxes and shipping remain unstated. | 3 | synthesis | supported with qualification | medium | ||
| Walmart Connect pages define new-to-brand as an 'ad-exposed customer who purchases a brand's product from Walmart site or app after not purchasing from the brand the previous 12 months' (lookback 'at or on Walmart for the 12 prior months'), state that 46% of 'sales revenue attributed to Walmart Connect ads in FY25 came from new-to-brand customers' on 'a 14-day attribution for Sponsored Search, Onsite display and Offsite media ads', and define ROAS as 'the ratio of total attributed sales revenue divided by ad spend'. | 3 | vendor assertion | supported as stated | high | ||
| Walmart's Onsite Display reporting API (definitions updated 28 October 2026) attributes sales, orders and units 'within 14 days of ad exposure' or 30 days, reports an 'attributionModel' of FAIR_SHARE or LAST_TOUCH, a configurable new-buyer lookback ('months6' to 'months36'), featured versus halo itemsets and in-store, pickup and delivery splits, and defines campaign lift metrics against a control group of 'households that were not exposed to the ad campaign and have similar digital & purchase propensity as the test group', with test and control group sizes and a 'Significance level of overall sales lift'. | 3 | vendor assertion | supported with qualification | high | ||
| Q1 (what kind of number): the 41 of 49 eligible cohort A service rows that name an outcome metric name 129 metric records, counted once per row and kind of metric; 89 are attributed with no counterfactual (attributed sales 32, attributed orders or units 20, ROAS 24, new-to-brand 13) and 6 are modelled; of the 34 lift and brand-lift metrics, 2 disclose random assignment, both for Instacart Sponsored Products and both documented only on pages from 2022 and 2023, and the other 32 leave assignment unknown or name lift with no comparison. Eight rows name no outcome metric on the pages read. | 3 | synthesis | supported as stated | medium | ||
| eBay's own documents describe the general-strategy trigger differently on 3 November 2026: the attribution page and the announcement say any buyer's purchase within 30 days of any click counts 'regardless of whether the buyer themselves clicked', while the Marketing API overview says a seller is charged 'when a buyer clicks on their promoted listed ad and purchases the promoted items within 30 days'; the attribution page itself warns that attribution 'is subject to change' and carries no last-updated date, the dated revisions found are the 13 January 2026 change and an ebay.de first-ad-slot change from 26 February 2025, and the help centre returned a verification wall. | 3, 4 | synthesis | supported as stated | medium | ||
| Etsy attributes orders and sales to Etsy Ads when someone interacts with an ad (a click, in Etsy's definition) and then buys any item in the shop within 30 days, charges per click, flags attributed orders with a megaphone icon, and defines ROAS as 'a measure of how profitable your ad is'. | 3, 4 | vendor assertion | supported as stated | high | ||
| 84.51°'s Glossary (last updated 22 May 2026) defines aROAS as 'The total sales of the targeted test group (for the products being measured) relative to the Total Campaign Cost' ('Total Sales / Total Spend') and iROAS as '(Total Test Group Sales - Total Control Group Sales) / Total Cost', the latter worded for 'households targeted for the coupon'; its managed-service page says aROAS 'doesn't factor in whether the ad truly affected customer behavior'. | 3, 4 | vendor assertion | supported as stated | medium | ||
| Walmart's Sponsored Search reporting API (definitions updated 16 July 2026) defines its attributed outcomes on click windows of 3, 14 and 30 days ('over 14 day click'), reports online sales separately ('This is for online sales only') from in-store attributed sales, orders and units on the same three click windows, splits advertised items from same-brand 'Other' items, counts new-to-brand orders 'over a one year look back window', and adds view-through metrics only for video campaigns, credited after an 'MRC defined view' of two seconds. | 3, 4 | vendor assertion | supported with qualification | high | ||
| Walmart Connect said on 4 April 2025 that advertisers can access its measurement through Partner Network measurement partners that 'complement our suite of first-party capabilities by integrating our various data sources': visualization partners (including Incremental), data connectors (Adverity, Improvado, Saras Daton, DataHawk), brand-lift partners (Kantar, Upwave, Nielsen, Lucid), 'Rest-of-Market partners, like NCSolutions and TransUnion' tracking 'incremental sales, including those made outside of Walmart', reach and frequency partners 'iSpot, Nielsen and Samba TV' and attribution partners TransUnion and Kochava; the data each receives and the terms are not stated. | 3, 4, 5, 6, 12, 13, 14 | vendor assertion | supported as stated | high | ||
| Amazon's 'Approved third-party providers' page (last updated 1 June 2026) documents third-party ad verification for Amazon DSP only ('available on Amazon DSP' with DoubleVerify, Integral Ad Science and Adloox, by server-to-server measurement) and lists 'Samba TV' with vendor type 'Measurement' among about sixty measurement vendors; it carries no verification or measurement-partner statement for Sponsored Products, Sponsored Brands or display ads beyond a link to the sponsored ads policies. | 3, 4, 5, 6, 12, 13, 14, appendix | vendor assertion | supported with qualification | high | ||
| Who holds the pen at Kroger Precision Marketing: KPM (powered by 84.51°) computes both the attributable and the incremental sales it reports, states that its incremental methodology 'adheres to IAB standards and best practices' (self-attested), and has its onsite outcome metrics (attributed sales, incremental sales, aROAS, iROAS) in MRC's pre-audit, which is a status rather than assurance; no restatement policy, change log, dispute procedure or sales-measurement separation was found on the KPM pages read (silence coded unknown). | 3, 4, 13 | synthesis | supported as stated | medium | ||
| Walmart Connect's changelogs record identity and attribution upgrades effective 1 June 2026 (new Sponsored Search in-store advertised, other, units and orders metrics; a Display attributionModel column). Its help article Overview of attribution (000012578) describes them: Sponsored Search moved from last click, not deduplicated with Display, to a unified, deduplicated last-touch model shared with Onsite Display, Walmart DSP and self-checkout store ads, with clicks still the only Sponsored Search touchpoint and clicks prioritised over views; Display moved from fair share across all delivered impressions to last touch over viewable impressions and clicks for campaigns starting 1 June 2026 or later, with fair share deprecated from 1 September 2026; online orders cancelled within 72 hours are now deducted and returns are not; history is not restated and Walmart advises against comparing post-June figures with earlier ones. The article states no change to the match basis or to the windows. | 3, 4, 13, 14 | vendor assertion | supported with qualification | medium | ||
| Orange Apron Media's measurement post (listed 22 April 2026) offers 'advanced studies such as incrementality, geo lift, and brand lift', clean-room environments with 'turnkey access to deeper insights', Quick-Start Reports in Orange Access, an Insights Console and a Supplier Analytics link to total assortment sales and inventory; no study states its assignment, counterfactual construction, eligibility, request route or presence in the default report. | 3, 5 | vendor assertion | supported with qualification | medium | ||
| Best Buy Ads documents two lift designs by placement class: 'Time-Based Regression Geographic Experimentation' for offsite (counterfactual geo; assignment unknown) and 'customer-level holdouts' for onsite (counterfactual holdout; assignment unknown); request route, eligibility, minimum spend, significance rule and whether lift enters the default report are not stated. | 3, 6 | vendor assertion | supported as stated | medium | ||
| Albertsons Media Collective's release of 6 January 2026 (re-opened 3 November 2026) measures in-store screen campaigns by comparing 'test stores exposed to ads' with 'a rigorously selected set of control stores with no media exposure', using 'nearly 60 variables for advanced store-level matching', so that 'the model delivers a statistically validated measure of true lift' in 'sales, orders and unit lift'; it does not say whether stores are allocated at random or how impressions are counted, says it is 'closely aligning our measurement roadmap with IAB current and future standards for in-store media', and describes closed-loop measurement as a future capability. | 3, 6 | vendor assertion | supported as stated | high | ||
| Walmart Connect Store Ads states that 'Store ads measurement includes both brand and sales lift' and that 'Initial proof-of-concept testing has shown that Store ads campaigns achieved up to 5% sales lift for Food suppliers on Deli screens and Bakery screens', footnoted to Deli and Bakery screen campaigns of 'food suppliers only, September 2024 – January 2025' with 'Sales lift validated at the 80% confidence level or above'; its audience footnote says 'Gross impressions are estimated projections based on historical and aggregated data' and cites 'ComScore, Dec 23, 2024 -April 24, 2025'. | 3, 6, 14 | vendor assertion | supported as stated | medium | ||
| On 2 September 2026 DoorDash announced that DoorDash Ads had earned the Alliance for Audited Media's incrementality certification, describing it as 'an independent review of DoorDash Ads' first-party incrementality measurement and reporting for Sponsored Products and Sponsored Brands' (US). | 3, 7, 8 | vendor assertion | supported as stated | high | ||
| DoorDash's 24 March 2026 release launched 'Sales Lift Measurement' for 'select advertisers via Managed Service for Sponsored Products campaigns', reporting 'over 30% average sales lift for CPG advertisers' (headline: 'Up to 30%') from '30 Sponsored Product Lift Measurement tests conducted in H1 2025', a 65% average lift for Snacks from 19 test campaigns, and Heritage Kellanova (now part of Mars Snacking) iROAS of '2.4X to 2.9X across three studies' in Q4 2025; the averages are footnoted as exposed customers versus 'a non-ad-exposed holdout group'; the release says the solution is 'based on Ghost Ads' and 'experiment-based' and 'builds on our previous incrementality measurement tool for Enterprise Restaurants'; assignment is not stated, and the footnote does not cover the Heritage Kellanova Q4 2025 studies, whose design is not stated. | 3, 8 | vendor assertion | supported with qualification | medium | ||
| DoorDash's ghost-ads resource (4 October 2025) contrasts the intent-to-treat randomized controlled trial, in which 'eligible user accounts are randomly assigned to either a treatment group ... or a control group', with ghost ads that 'identify equivalent populations' of exposed and would-have-been-exposed accounts and that 'now powers Sponsored Listings lift studies', claiming the change 'reduced experimentation dilution by 92%' and 'improved iROAS confidence intervals by 35%', with no base period, experiment count, minimum sample or significance rule stated. | 3, 8 | vendor assertion | supported with qualification | medium | ||
| A 23 March 2026 post by Albertsons Media Collective's Vice President of Media & Measurement says that 'Clean experimental isolation is not always possible', that before reporting lift it evaluates 'match quality' and may add 'time-series modeling', that it offers 'a randomized testing methodology upon request', and that matched reporting covers 'Pinterest, In-Store and Offsite Display and, now, onsite display capabilities when programming reaches the appropriate scale'. | 3, 8 | vendor assertion | supported as stated | medium | ||
| Amazon describes its standard attribution as last touch ('It credits a conversion to one ad click or view, usually the last ad encountered on the path to purchase') and announced at unBoxed 2024 (press release 15 October 2024: testing 'later this year, with a full rollout planned for 2025') a multi-touch attribution beta that 'divides credit for Amazon purchase conversions across Amazon Ads touchpoints in proportion to their value', offered as 'an additional way' to view results. The undated beta page says US advertisers 'will soon' have access, lists console and downloadable-report availability 'for all US advertisers using Amazon DSP, Sponsored Display, Sponsored Brands, or Sponsored TV' (Sponsored Products not named in that list) and names Sponsored Products only for Amazon Marketing Stream conversion datasets; on 3 November 2026 the v3 columns page carries no multi-touch column and the v1 multi-touch fields show 'Program eligibility: N/A', so whether the beta reaches Sponsored Products, and at what stage, is unconfirmed. | 3, 8 | vendor assertion | supported with qualification | medium | ||
| Amazon's multi-touch attribution paper (arXiv 2508.08209, 11 August 2025; eight authors listed with Amazon affiliations, including Brett R. Gordon and Florian Zettelmeyer, also of Northwestern University) records calibration experiments: randomised controlled trials in which 'one group of shoppers is eligible to see ads from the focal campaign ("treatment") while another group of shoppers is not ("holdout")', drawn from 'a database of hundreds of thousands of RCTs across our ad products'; it reports no sample sizes, holdout shares, periods, intervals or results. | 3, 8 | vendor assertion | supported with qualification | medium | ||
| Walmart Connect announced on 28 April 2025 a 'Search Incrementality solution' for Sponsored Search built on 'industry-leading, test-and-control methodologies' to evaluate 'incremental sales online', with plans 'to roll out more broadly to managed advertisers this summer'; its MRC post of 16 October 2025 still describes 'expanding ... such as incremental sales measurement for Sponsored Search'; no assignment, unit, eligibility, minimum spend, significance rule, availability notice or result was found on the pages read. | 3, 8 | vendor assertion | supported as stated | high | ||
| Instacart's engineering post of 30 June 2023 documents a randomised holdout assigned per user-advertiser pair ('hash of (user_id, advertiser_id)', by its 'Roulette' engine) with ghost ads simulating the holdout's would-be impressions, measuring total (organic plus attributed) sales of the advertised products as total incremental sales, lift percentage and incremental sales on ad spend, and limits 'continual ads incrementality measurement' to Sponsored Products; it states no significance rule, minimum sample or interval. | 3, 8, 13 | vendor assertion | supported as stated | medium | ||
| Lift at Instacart is a separate test, not the default report: a September 2022 post launched a 'pilot testing experience' measuring 'true total sales lift' with 'randomized test and holdout groups', the May 2023 Reckitt case says 'Lift testing services may not be available to all partners. Please consult with your account team', the measurement page captured on 28 September 2026 offers 'Sales lift testing' with 'gold-standard experimentation' and no stated design or format, and the 2026 Ads Manager analytics index lists no lift, incrementality or A/B-test page. | 3, 8, 13 | vendor assertion | supported as stated | medium | ||
| This review found no Amazon page documenting a sales-lift offer (randomised holdout, ghost ads or another counterfactual) delivered to advertisers for Sponsored Products campaigns, on the pages screened by 3 November 2026 (the list as written, plus Amazon's Brand Lift page and its 15 October 2024 press release) and in two searches; and Amazon's survey-based Brand Lift product, which compares 'an ad-exposed audience and an unexposed control group', states that sponsored ads and Display, Video, and Audio campaigns are eligible 'except for Sponsored Products'. The documented outcome measures for Sponsored Products are click attribution, a multi-touch beta of unconfirmed reach and buyer-run AMC aggregates. | 3, 11 | synthesis | supported with qualification | medium | ||
| Amazon documents that Sponsored Products conversion metrics (purchases, sales, units sold, ROAS) default to a 7-day look-back for sellers and a 14-day look-back for vendors after an ad click, 'consistent with the advertising console and downloadable reports', and that Sponsored Products, which does not measure view-through attribution, reports 0 for conversions from views: the documented default is click-based attribution with no counterfactual (Amazon API documentation read 3 November 2026; global). | 3, 11, 13 | vendor assertion | supported as stated | high | ||
| Amazon's version 3 reporting columns define Sponsored Products conversions by days after an ad click (1, 7, 14 and 30 days) and split them into same-SKU and other-SKU variants, e.g. 'sales7d' is 'Total value of sales occurring within 7 days of an ad click' while 'salesOtherSku7d' counts sales 'where the purchased SKU was different from the SKU advertised'; no column read states whether sales are gross or net of returns, cancellations, discounts, taxes or shipping. | 3, 11, 13 | vendor assertion | supported as stated | high | ||
| DoorDash Ads CPG reports attributed sales by default with no counterfactual: sales of relevant items from the first order within 14 days of a click (Sponsored Products: the clicked item; Sponsored Brands: any campaign item), at any DoorDash store, credited to the most recent ad interaction; ROAS is attributed sales over spend; Brand Halo Sales and 180-day and one-year new-to-brand sales are separate; view-through is not mentioned. | 3, 12 | vendor assertion | supported as stated | high | ||
| Chewy Ads lists its outcome metrics (New-to-Brand, Autoship, 'Lifetime Value ROAS' and 'incrementality') on two own pages read on 3 November 2026 without naming the default report, an attribution window, touch model, match basis, new-to-brand lookback or any lift design; the default outcome method is coded unknown, and the lift offer is coded 'named without a design'. | 3, 14 | vendor assertion | supported as stated | medium | ||
| Gulp Media Network documents two outcome constructs in the future tense: 'Exposed customer purchase behavior will measure the ROAS for the campaign based on measured sales in C-Shopper' and 'Control/exposed methodology will enable lift reporting' to identify net new purchases and calculate an iROAS, plus 'Enhanced campaign measurement' and a rewards-member research panel; the measured sales come from transactions collected at every 7-Eleven and Speedway store, linked to 7REWARDS and SpeedyRewards members, and the default between the two constructs, the window, the control construction and the assignment are not stated. | 3, 14 | vendor assertion | supported as stated | medium | ||
| Leap Media Group's solutions page (copyright 2022) offers 'Measurable Analytics, including ROAS' and lets advertisers 'monitor the audiences, impressions and overall effectiveness and efficiency' of their investment, against '1.5 million myPerks households' scanning loyalty cards 'with every purchase'; no ROAS definition, window, lift design or access route is documented, so the default is coded unknown. | 3, 14 | vendor assertion | supported as stated | medium | ||
| Across the 27 workstream E2 rows, 21 document a default outcome construct and 6 do not (Chewy Ads, Kohl's, Wayfair, Gopuff, Staples, Leap); of the 21 documented, 19 are attribution with no counterfactual (14 with at least one parameter, 5 by term only), one (DG Media Network) lists lift metrics in its default dashboard and one (Gulp) describes attribution and lift without naming a default. At network level, with the frozen representative services, 15 of 21 networks document a default and 13 of those are attribution-only; if term-only rows are treated as undocumented, 16 of 27 rows and 10 of 21 networks document a default, of which 14 and 8 are attribution-only. | 3, 14 | synthesis | supported with qualification | medium | ||
| On the record read, the documented default for the KAP self-serve onsite formats is attribution (14-day window, last interaction, click over view, halo; cost booked on the exposure date and the sale on the purchase date), and incremental measurement naming those formats is PrecisionView 360 (paid, opt-in, minimum US$400K across three or more channels) and the Incremental Sales and iROAS metrics in MRC's pre-audit scope. Managed-service campaigns in 84.51° Prism, which include homepage, department, seasonal and promotions Premium Placements, report Sales Uplift against tightly-matched control households by default. KAP's marketing page promises "sales lift and household penetration"; KAP's own definition of those metrics, and whether a self-serve advertiser receives them by default, are behind the KAP Learning Center sign-in. | 3, 14 | synthesis | supported with qualification | medium | ||
| Amazon's 1 January 2026 view-attribution change is a documented, dated reporting revision for Amazon Sponsored Brands, Amazon display ads (vCPM) and Amazon DSP in-Store campaigns, published on a page dated 1 January 2026 (the effective date), with the legacy 14-day 'all views' metrics kept available as a parallel series; the page says nothing about restating historical data. | 4 | vendor assertion | supported with qualification | high | ||
| Instacart's documentation records dated reporting revisions (linear attribution from 1 August 2022, affecting 'sales reporting data for dates August 1, 2022, and onward'; clicks-only credit from 1 April 2025) and maintains public changelogs for Ads Manager, the Data Hub and the Ads API; no other network screened in this stream exposed a changelog, and no restatement policy or dispute procedure was found on Instacart's screened pages. | 4 | vendor assertion | supported as stated | medium | ||
| Who holds the pen at Roundel: Roundel reports its own closed-loop and multi-touch attributed outcomes and runs the incrementality designs for eligible campaigns on Target data; its pages name no third-party measurement partner, no accreditation, no metric-definition standard, no restatement or correction policy and no separation between sales and measurement (silence coded unknown), and Target's privacy policy (updated 15 August 2026) never mentions Roundel. | 4 | synthesis | supported as stated | medium | ||
| Who holds the pen at Uber Eats sponsored listings: Uber computes the attributed metrics on its own glossary definitions and runs the platform experiments and the cross-channel incrementality studies (with Affinity Solutions supplying card data), and the reporting pages state 'All reporting API endpoints are subject to standard Uber Ads API rate limits' but no data-finality, restatement or correction policy, change log, dispute procedure or separation between sales and measurement (silence coded unknown). | 4 | synthesis | supported as stated | medium | ||
| Vrbo Sponsored Listings, announced as launched globally in Expedia Group's release of 1 September 2026 after a pilot, charges per booked night only when 'a guest who clicks on your ad actually books their stay' (minimum bid $5 per booked night), so the click-to-booking attribution is the billing event; the portal reports impressions, clicks and bookings, campaigns run on the Expedia Group Advertising platform, and no attribution window or cancellation rule is documented. The pilot results (49% bookings, 39% revenue, 30% booked nights) are averages on 'Expedia Group internal data, 2026' with no design. | 4 | vendor assertion | supported as stated | high | ||
| For four E2 rows the attributed outcome is also the invoice: eBay Promoted Listings General charges its ad fee on the attributed sale, Etsy Offsite Ads charges a percentage fee on attributed orders, Vrbo Sponsored Listings charges per booked night after a click, and Hy-Vee RedMedia bills on-site paid search per conversion; in each case the operator writes the attribution rule, reports the outcome and bills on it, and none of the four documents a dispute or restatement procedure for attributed events on the pages screened. | 4 | synthesis | supported with qualification | high | ||
| No E2 row documents an outcome-metric accreditation or certification on the pages screened; third parties appear as a named study author (Circana for Best Buy Ads), a reported incrementality partner (Ipsos for Lowe's), enablement platforms (Koddi for Kohl's; Criteo for UB Media and Walgreens sponsored search; Vantage, Pentaleap and Kevel for Orange Apron Media; Carrot Ads for Gopuff, announced), clean-room or partner environments (LiveRamp for Chewy, Best Buy and DGMN; Meta and Google for DGMN), approved verification vendors (CMX) and TAG anti-fraud and anti-malware logos (CMX); none of these is shown to validate an outcome number. | 4 | synthesis | supported as stated | medium | ||
| Lyft Ads documents outcome measurement only as partner-run studies listed by tier, with no default method, window, design or match basis: it offers to 'Tie ad exposure to in-store visits, offline sales lift, and online conversions' through Foursquare, NCS and Kochava, brand metrics with Kantar, Nielsen and IAS, and verification with DoubleVerify, IAS and Nielsen ONE Ads; its home page uses 'closed-loop reporting' for store traffic, not sales. | 4 | vendor assertion | unresolved | medium | ||
| Chase Media Solutions documents attributed sales from Chase's own card transactions as its default report: it promises to 'accurately attribute sales to campaigns', tracks 'online and in-store purchases', reports 'budget utilization, ROAS, impressions and sales' on a downloadable Performance Dashboard, and charges brands only 'when a customer views a campaign and makes a purchase'; no attribution window, touch rule, sales basis or deduplication is documented on the five Chase pages opened. | 4 | vendor assertion | supported as stated | high | ||
| Chase Media Solutions is itself the reporting party and states privacy commitments without naming a legal basis: it 'does not share customers' personally identifiable information', says 'All our customers are fully authenticated', and targets audiences 'based on up to one year of purchase history'; none of the pages opened names a consent basis, a GLBA or state-law notice specific to the offers business, a restatement policy or a separation between media sales and measurement. | 4 | vendor assertion | unresolved | medium | ||
| Cardlytics' default self-serve report is attributed, activation-based revenue: the Ads Manager dashboard shows 'budget spend, engagements, attributed sales, revenue and ROAS', and its glossary (updated 19 February 2026) defines revenue and ROAS 'post-engagement' as revenue 'from purchases by users who activated an offer' divided by ad spend, an engagement being a user 'who clicked on or activated an offer', with purchases split by channel into 'In-store, Online, or Unknown'; no counterfactual enters these metrics. | 4 | vendor assertion | supported as stated | high | ||
| Fetch's validation record is an assessment it engaged KPMG to perform: 'Fetch engaged KPMG to conduct an independent performance measurement assessment' of 'the processes and methods we use to measure partner campaign performance', and Fetch reports that 'KPMG was able to exactly replicate the outcomes Fetch reported to partners' and that its methodology 'is well founded'; Fetch's posts of 8 June 2023 state no attestation standard, period, campaign sample or fee, and the KPMG report is not provided. | 4 | vendor assertion | supported as stated | medium | ||
| Ibotta's validation evidence is Circana's, as published by Ibotta: a meta-study of '48 Ibotta campaigns' 'using a causal methodology' that compared 'matched buyer groups', reporting '+16.5% average lift in incremental sales', '+17% average increase in new household penetration' and an 'average iROAS of $5.05' for existing brand buyers (page of 19 June 2026); the page states no period, assignment, panel or commissioner, while the Danone case says Circana used 'loyalty-card panel data matched against campaign exposure'. | 4 | vendor assertion | supported as stated | medium | ||
| Klarna's Search & Compare terms (revised 23 July 2025, effective 1 September 2025) define a click-priced placement whose default report is delivery: Klarna will 'provide estimated live statistics for Klarna and its affiliates for Clicks and advertising impressions', supplies conversion-tracking code only where payment depends on purchases on the partner's site, and sells extra data as 'Analytics Premium'. | 4 | vendor assertion | supported with qualification | high | ||
| Upside's default and billing measure is incremental and built from matched non-users: for each user who claims an offer it finds '10 to 100 non-Upside users with similar purchasing histories' and treats the spend difference as incremental; its contract methodology (July 2023) builds the control from 'control payment identifiers' with similar spend, subtracts a pre-period difference (difference-in-differences), measures per user and merchant network through 7 days after the user's latest Upside transaction, caps the result at Upside-attributed sales and at zero when negative, and computes its fee as the agreed profit-share percentage of incremental gross profit minus promotional costs (zero where incremental gross profit does not cover the user's cash back, with Upside funding the shortfall); the pages do not state assignment, and the test group is defined as shoppers who claim an offer, so assignment is non-random by construction (inference from the design). | 4 | vendor assertion | supported with qualification | high | ||
| On the measurement pages of Roundel and Walmart Connect read on 3 November 2026, the words 'IAB', 'MRC', 'Media Rating Council', 'standard', 'glossary' and 'definition' do not appear; both describe 'closed-loop measurement' produced by the network itself (Roundel: 'As a media company that's part of Target', footnotes '2026 Target Internal Data'; Walmart Connect: 'Our closed-loop measurement tracks sales across our site, app and in our stores.' via 'our self-serve Walmart Connect Ad Center'), and Roundel offers incrementality 'for eligible campaigns' using 'synthetic control, matched market and randomized controlled trials'. | 4 | vendor assertion | supported as stated | high | ||
| Amazon versions its reporting surfaces with dated deprecations: version 2 Sponsored Products reporting endpoints were to be deprecated on 30 March 2023, version 3 reporting and Amazon Marketing Stream version 1 are 'To be shutoff after June 30, 2027', and the successor Reporting API v1 renames conversion metrics (dropping interval suffixes and standardising 'Halo') with the statement that 'These naming changes do not affect measurement logic'. | 4 | vendor assertion | supported as stated | high | ||
| For Amazon DSP, Amazon's Reporting API v1 documents a reporting revision: conversions move from the conversion date to the 'traffic date', base metrics now include promoted and halo conversions and 'may return higher totals than in Version 3', and DSP campaigns with flexible look-back windows report 'the advertiser-selected click and view windows', while campaign types other than Sponsored Products use a standard 14-day look-back. | 4 | vendor assertion | supported as stated | high | ||
| On 31 August 2026 the FTC and 22 states sued Amazon (Case 2:26-cv-03097, W.D. Wash.) alleging that 'Since 2019' it 'secretly and systematically overcharged its approximately 1.2 million advertising customers' through undisclosed 'soft reserve' prices in its Sponsored Products, Sponsored Brands and Sponsored Display auctions, that in 2024 Sponsored Products advertisers paid 'their own winning bid close to 80% of the time', and that Amazon used advertisers' ROAS, which the complaint calls 'the leading measure of the cost-efficiency of advertisements', as 'the main signal to tune our pricing controls'; all of this is allegation. | 4 | direct record | supported as stated | high | ||
| The Kroger Co.'s fiscal 2025 Form 10-K (filed 31 March 2026) calls Kroger Precision Marketing 'our retail media business' and places 84.51° LLC, 'Kroger's data analytics subsidiary', among the alternative profit businesses for which its executive Yael Cosset is responsible; an 84.51° boilerplate of 16 September 2022 calls Kroger Precision Marketing 84.51°'s 'retail media advertising solution'. | 4 | direct record | supported as stated | high | ||
| The legacy Direct Connect reporting page (84.51° Prism, last updated 8 May 2026, marked for deprecation on 1 June 2026 and still served on 3 November 2026) defines offsite DSP-campaign metrics with 'a 14 day lookback window for both in-store and online', aROAS on 'the targeted test group', household penetration 'within the attribution window', New, Existing and Lapsed households by 52- and 26-week lookbacks, a modelled 'Rest of Market' uplift 'including non-Kroger stores', viewable impressions at '50% ... for at least one second', and states that iROAS reporting requires spending 'at least $150K in Media'. | 4 | vendor assertion | supported as stated | medium | ||
| The KPM Direct Connect reporting policy (84.51° Prism, last updated 8 May 2026, scoped to 'Campaigns run through Direct Connect only' and marked for deprecation on 1 June 2026) sets a correction and restatement rule: advertisers 'must notify KPM if data is not flowing' within five business days, otherwise 'a makegood or restatement will NOT be submitted for approval'; 'Immaterial changes to such reports may be made at any time' while substantive changes 'will be discussed with Client and modified where appropriate'; reports stay accessible 'for up to 13 months' and 'are the confidential Information of KPM'. | 4 | vendor assertion | supported with qualification | medium | ||
| 84.51° keeps a dated public change log for its Prism platform: the release notes list 33 entries from January to September 2026, including 'New PrecisionView 360 and PrecisionView Lite reporting' (February 2026), an 'Updated Glossary' (January 2026) and 'Direct Connect Campaign topics deprecated' (June 2026: 'All Direct Connect and related topics in the 84.51° Prism Learning Center have been removed'), and its roadmap records a completed Q1 2025 change in which 'Sales to Cost metrics' were 'retired and replaced with iROAS' and aROAS for Best Customer Communications; no entry concerns a restatement, a data correction or a change to an attribution method. | 4 | vendor assertion | supported as stated | medium | ||
| The developer documentation and the help centre (readable in a browser) document dated change logs, a documented reporting revision with notice to buyers (the 1 June 2026 attribution change, with a statement that history will not be restated and advice not to compare across the change), retroactive updating of attributed metrics within the window, a dated methodology description for Sponsored Search delivery metrics (last updated 9 September 2026; no version history seen), and support routes (billing dispute, support case). No metric-specific dispute procedure or general correction policy was found; the help-centre screen remains incomplete (Billing dispute, Open a case, and the Walmart Connect Advertising Amended IAB Terms were not read). | 4 | synthesis | supported with qualification | medium | ||
| The Instacart Ads API terms (last updated 18 October 2021; required minimum functionality 4 October 2022) license agencies to build reporting that must 'Show ROAS Metric and associated calculation', 'Show Attributed Sales', 'Show Attributed Quantities', 'Explain IC attribution windows' and 'Make advertiser aware of any delays in reporting', and require that 'Agency must provide Instacart a list of metrics displayed to Advertisers and calculation methodology thereof'. | 4, 5 | vendor assertion | supported as stated | high | ||
| Marketing Dive reported on 13 October 2026 that Lowe's Media Network is partnering with Ipsos 'to provide independent incrementality measurement', 'including bespoke marketing mix modeling and agile attribution, for brand partners', and that the network claims to be among the first retailers to offer incremental return-on-ad-spend reporting; it quoted the network's VP and GM: 'I don't want an advertiser grading their own work'. No Lowe's own source opened states the partnership, so it is coded reported, with assignment, counterfactual and unit unknown. | 4, 5 | vendor assertion | supported with qualification | low | ||
| Rokt documents data access for the advertiser and discloses that its counts move: automatic exports of referral (click) data 'for your own reporting purposes or to integrate with your CRM', scheduled secure file transfer or 'real-time updates via API', manual export of transaction data and a Query API 'to pull performance data at the account, campaign, and creative levels'; referral-time conversion counts 'may continue to accumulate until the attribution window is reached', while conversion-time counts 'do not change over time'. | 4, 5 | vendor assertion | unresolved | high | ||
| Albertsons Media Collective's standard terms (Version 3.0 of the internet-advertising template, undated) make its ad server 'the measurement used for invoicing' and the 'Controlling Measurement', allow only 'Media Company tags' on its own properties, restrict advertisers' use of Site Data to performing the IO and 'campaign reporting and measurement' with deletion at campaign end, oblige it to cure 'an incomplete or inaccurate report' within five business days, and leave one clause as an unfilled placeholder. | 4, 5 | vendor assertion | supported with qualification | high | ||
| Instacart's 10 June 2025 release (read through a PR Newswire reprint) says agency partners such as Omnicom's Flywheel 'can access closed-loop measurement from Instacart across all of their campaigns' run in The Trade Desk across 'CTV, audio, video, and display', describing 'Instacart's deterministic purchase-based data' without naming any identity mechanism, window or counterfactual. | 4, 5, 6, 12, 13, 14, appendix | vendor assertion | supported as stated | medium | ||
| Circana and DoorDash announced on 2 September 2026 'a partnership to bring independent, third-party validation to DoorDash's measurement capabilities', 'Now available to select advertisers for Sponsored Products and Sponsored Brands'; DoorDash's release attributes '$0.59 in incremental in-store sales per household' for a major beverage company to 'a household-level lift study', Circana's release presents the same $0.59 as 'Early measurement results' in a paragraph that opens 'New findings from DoorDash measurement', and the footnoted Circana analysis covers 'an eight-week campaign on DoorDash, between 11/10/25 and 1/4/2026 in the U.S.'; no design, panel, match basis or payer is stated. | 4, 5, 8, 9, 12 | vendor assertion | supported with qualification | medium | ||
| MMN's terms let the advertiser use 'Click/View Data' 'solely for Advertiser's own internal purposes', treat it as confidential, forbid selling it, combining it with another advertiser's data or transferring it to any ad network, ad exchange or data broker, and state that 'MMN owns any data generated from Users' interactions with ads'; they contain no third-party tag, clean-room or API clause, offer a make-good only where product identifiers are rejected through MMN's error, and are 'subject to updates at any time in MMN's sole discretion' with no version number. Sponsored Products can also be bought and reported through the Amazon Ads console and APIs under the Amazon Retail Ad Service pilot. | 4, 5, 10 | vendor assertion | supported as stated | high | ||
| Cardlytics' US master advertiser terms for Cost per Served Sale make its own data "the definitive data used to calculate the fees" (the client also pays fees above the IO budget); bar disclosure of Insights Portal data "to any consultant, subcontractor or third party without Cardlytics' written consent"; limit any transaction-level or individual offer-activation information to "internal offer verification or internal evaluation" and require its deletion within 45 days; treat campaign-performance reports as confidential information shareable with the client's contractors or agents who need to know under equivalent confidentiality; and let Insights Portal report types change with bank-partner obligations. An insertion order prevails where it conflicts. The full page text contains no billing-dispute window, audit right or third-party measurement clause; its only dispute clause sets Georgia law and Fulton County venue. | 4, 5, 10 | vendor assertion | supported with qualification | medium | ||
| Etsy attributes an order to Offsite Ads when a buyer clicks an Etsy-run off-site ad and purchases from the seller's shop within 30 days, and the attribution is the billing event: a 15% fee on attributed orders, 12% for shops with $10,000 or more in sales over any 365 days, capped at $100 per order, with enrolment mandatory for the lifetime of a shop once it crosses that threshold; partners include search engines and social platforms that 'may change over time'. | 4, 6 | vendor assertion | supported with qualification | high | ||
| Kroger Precision Marketing's own pages (read in a browser on 3 November 2026) say its incremental measurement uses 'many IAB-recommended methods as well as test and control groups' and that its 'methodology adheres to IAB standards and best practices', while its FAQ states KPM is 'proudly certified by the Interactive Advertising Bureau (IAB) and Merchant Risk Council (MRC)', naming the wrong body for the MRC acronym; the Media Rating Council's own list shows 'Kroger (Pre-audit in process as of 6/2026)', not an accreditation. | 4, 7 | vendor assertion | supported with qualification | medium | ||
| The IAB/MRC Retail Media Measurement Guidelines (January 2024) apply to 'retail media organizations or other parties measuring this activity' (p. 10), so they accept the network itself as the measuring party, and they rely on audit, validation and disclosure rather than on separating measurement from sales: a retail media measurement organisation 'is encouraged to conduct frequent (at least annual) data quality audits (internal audits as well as external audits performed by independent parties)', with 'Periodic audits by MRC or other independent third parties' (p. 14); 'If outcomes are reported for retail media organization activity and used for pricing or optimization considerations, they are encouraged to be subject to third-party audit and validation.' and the document 'recommends retail media measurement be compliant with MRC requirements and ultimately, accredited' (p. 25); 'Known errors should be disclosed to measurement users as soon as is possible through client notification and quantification of impact or data restatement where applicable.' (p. 15); no passage addresses separation between a network's sales function and its measurement function. | 4, 7 | direct record | supported with qualification | high | ||
| The Ovative paper states that The Collective 'cannot always rely on RCTs to measure campaign incrementality at scale' and 'often uses matching methods to construct control groups', and that an RMN not using an RCT is 'likely using a matched control and less frequently a synthetic control group'; these are the study's characterisation of practice during the 42 campaigns and of omni-channel RMNs generally, and are recorded with the study, not as Albertsons Media Collective's documented default. | 4, 8 | vendor assertion | supported as stated | high | ||
| Cardlytics' only named third-party measurement is Nielsen Sales Lift Measurement, which its FAQ describes as 'an independent, third-party process' that is 'based on purchase data seen through the Cardlytics platform' and uses 'Nielsen's own methodology' for sales, purchases and basket size; who pays, the control design and Nielsen's access terms are not stated. | 4, 8 | vendor assertion | supported as stated | medium | ||
| This review found no independent validation of any Instacart outcome number in the sources screened by 3 November 2026: the lift results are Instacart's own (Reckitt, the 2023 average, Applegate, Challenge Dairy), the Circana work comprises buyer-overlap descriptions (August 2023, February 2024) and Circana-provided in-store lift figures for two unnamed sponsored-product advertisers against a 'Circana control group' in a study Instacart says it 'collaborated' on (funding, control construction and assignment not stated), MRC accreditation covers delivery metrics only, and AAM's incrementality certification list names DoorDash alone. | 4, 8 | inference | supported with qualification | medium | ||
| Intentwise documents its Amazon hourly feed (Amazon Marketing Stream) as push-only data from Amazon, enabled on request with 'no historical data', summarised hourly by campaign, keyword, match type, ASIN and placement, populating in about 24 hours, 'based on the 14-day attribution window', with negative values explained as 'corrections from Amazon's click and conversion validation process'. | 4, 9 | vendor assertion | supported as stated | high | ||
| Who holds the pen at DoorDash Ads (CPG): DoorDash computes attributed and incremental results on its own metric definitions (help-centre key metrics), had its incrementality 'metric calculations, and customer disclosures' reviewed by AAM, and its ghost-ads resource says 'DoorDash Ads is working toward MRC accreditation'; no restatement or correction policy, change log, dispute procedure or sales-measurement separation was found on the pages screened (silence coded unknown); the only timing rule found is that Event Date 'is based on UTC time'. | 4, 12 | synthesis | supported as stated | medium | ||
| Cardlytics documents offer-level timing rather than an attribution window: its API page cites 'various redemption windows ranging from 30-45 days' and says purchases 'can have up to 7-14 days delay for transaction data completeness', the FAQ says performance data are 'updated hourly after an initial 4-day delay', the Reporting Hub says 'Performance data updates daily', and reports reach back two years. | 4, 12, 14 | vendor assertion | supported as stated | high | ||
| Macy's Media Network documents 'closed-loop reporting' that 'tracks online and in-store conversion activity, as well as new buyer acquisition rates', shows ROAS labels of 8x (Bulova), 2x (Prada) and 15x (Nespresso) with new-customer rates of 89%, 85% and 83% and no footnotes, and its Self-Serve Ad Terms say MMN tracks delivery 'through its own ad server', that reporting 'will be final and not subject to challenge unless otherwise required by applicable law', and that the Sponsored Ad Service is a CPC program; no window, touch model, match basis or lookback is stated. | 4, 13 | vendor assertion | supported as stated | high | ||
| Amazon documents a restatement schedule for sponsored ads data: initial clicks within 12 hours with changes 'up to three days' after the click from traffic validation, initial conversions within 24 hours restated '1, 7, and 28 days after the conversion event' (so a 14-day window can change up to 42 days after the report date), while its reporting overview says conversion data 'is subject to change for up to 60 days' and invalid-traffic detection runs 'for up to 30 days after the fact'; the FAQ adds that conversions are reported on the date of the ad interaction, so restatement covers every date in the attribution window, and the overview links click validation to charging ('Amazon validates every click to make sure advertisers are not charged unnecessarily') and says Amazon Marketing Stream v2 backfills up to 70 days 'as attribution validation completes'. | 4, 13 | vendor assertion | supported with qualification | high | ||
| For Amazon DSP, AMC access with 'DVA+ ... advanced settings' is by request to an Ad Tech Account Executive while partners on the Amazon Ads Partner Network can use AMC on advertisers' behalf and agency partners 'offer professional services on query development and custom analytics in AMC'; the DSP product page lists AMC, attribution, brand lift and sales lift only as navigation links. | 5 | vendor assertion | supported with qualification | medium | ||
| Amazon's 'Approved third-party providers' page (last updated 1 June 2026) documents third-party verification for Amazon DSP only: DoubleVerify, Integral Ad Science and Adloox via S2S, custom-audience campaigns limited to DSP-hosted creatives, 'Certain products and placements do not support third-party ad serving', third-party impression tracking on the Amazon Mobile Shopping site and app (by 1x1 pixel only, with JavaScript tags refused on the app), a 'Self-Service Third-Party Measurement Studies' section (Amazon DSP only), and 'Samba TV' listed as a 'Measurement' vendor. | 5 | vendor assertion | supported with qualification | high | ||
| This review found no documentation of third-party verification tags or of event-level exposure export for Amazon Sponsored Products on the pages screened by 3 November 2026; buyer-run outcome analysis is documented through AMC (F09). | 5 | synthesis | supported with qualification | medium | ||
| Instacart's Data Hub is a Snowflake clean room in which partners install and view clean rooms with SQL commands and run pre-built Instacart templates (including Ads Taxonomy, New to Brand and Path to Purchase), with 'partners ... restricted on what they can run' and no guaranteed access; its launch release (January 2026) says advertisers can 'measure campaign impact with flexible attribution approaches' and that it was piloted with select agencies and CPG brands. No page states an output threshold, export rule or destination, names a Sponsored Products exposure dataset, or documents free-form buyer queries. H3 axis (b) for Instacart Sponsored Products: not documented at service scope. | 5 | vendor assertion | supported with qualification | medium | ||
| Instacart's documentation cites an 'MRC-accredited measurement methodology for gross impressions, viewable impressions, and clicks' and MRC lists Instacart (GI060126) for delivery metrics; Instacart's own DoubleVerify and Integral Ad Science integration post (20 October 2023 per the brief) and IAS's release were not read, and DoubleVerify's own release of 12 October 2023 scopes its Instacart integration to viewability and invalid-traffic verification of 'Instacart video and display ads', with the integration still to be built 'over the coming months', and does not name sponsored product ads; so verification tags are not documented for sponsored product ads, and DV viewability and IVT verification for display and video is documented as announced (2023), with no completion date on the pages read. | 5 | direct record | supported with qualification | medium | ||
| Roundel's identity and access disclosure: '100M+ Target Circle members powering Roundel targeting and measurement' (2026 Target internal data), measurement 'grounded in insights from first-party data across in-store, in-app and online shopping behavior', and 'secure data collaboration environments' named without output rules, thresholds, destinations or buyer query rights; the solutions page documents self-service audience activation 'with your DSP' and login-based campaign reporting, and no API. | 5 | vendor assertion | supported as stated | medium | ||
| This review found no documentation of event-level exposure access or buyer-run outcome queries for Kroger Precision Marketing onsite on the KPM pages read; the 84.51° Collaborative Cloud appears only in trade reporting (19 November 2021) as a hosted environment for CPG data scientists over transaction-level data for 60 million de-identified households, with no mention of KPM or ad exposures; 8451.com refuses the fetch tool. | 5 | synthesis | supported with qualification | medium | ||
| Walmart Connect's accreditation post (16 October 2025) names onsite partnerships 'with third-party measurement leaders' 'such as DoubleVerify and Integral Ad Science' beside MRC accreditation of 'first party click and impression metrics for Sponsored Search'; on the pages screened, no Walmart page documents event-level exposure data for a buyer or an independent measurer, or buyer-run outcome queries on Sponsored Search exposures (H3: axes (a) and (b) not documented on the readable pages, and coded 'source inaccessible' because the US help centre and the Developer Portal were not read). | 5 | vendor assertion | supported as stated | medium | ||
| Sam's Club Connect's measurement page offers a 'self-service dashboard' that puts advertising performance and 'point-of-sale results together' and 'Flexible data exports' ('Data Your Way'), and says 'Independent partners validate campaign performance' without naming them; no clean room, buyer query, API, event-level export or verification vendor is documented on the Sam's Club Connect pages screened (H3-type axes, recorded for the access record: (a) not documented; (b) not documented). | 5 | vendor assertion | supported as stated | medium | ||
| This review found no documentation of event-level exposure data, buyer-run outcome queries, data exports, an API or third-party verification for Albertsons Media Collective onsite on the pages screened by 3 November 2026 (home, about, difference, onsite, offsite and results pages, the 6 January and 16 April 2026 releases, the privacy policy); a 'Client Hub' is linked without any description (H3: axes (a) and (b) not documented in the screened record; screen incomplete, with the Client Hub and standard terms unread). | 5 | synthesis | supported with qualification | medium | ||
| DoorDash Ads CPG documents self-serve and managed buying, Brand Center exports, a direct API connection and approved API partners (Pacvue, Flywheel, Skai, CommerceIQ), marketing reports covering up to a year, and Offsite Ads on search, social and programmatic partners with 'closed-loop attribution'; no clean room, event-level export or verification vendor is documented. | 5 | vendor assertion | unresolved | high | ||
| Chewy Ads names a LiveRamp-powered clean room ('Chewy Marketing Cloud') 'to inform campaign optimization and performance' without stating who writes and runs queries, output granularity, thresholds or destinations; no event-level export, reporting API or third-party verification tag policy appears on the pages read. | 5 | vendor assertion | supported as stated | medium | ||
| Best Buy Ads documents My Ads as a self-service platform with 'measurement access to a handful of ad products' and a LiveRamp clean room through which buyers 'work with us to activate' audiences; a 2025 recap adds clean-room audience collaboration with partners and audiences sold through The Trade Desk. Buyer-run outcome queries, output rules, event-level export, a reporting API and on-site verification tags are not documented on the five Best Buy Ads pages read. | 5 | vendor assertion | supported as stated | medium | ||
| CMX's ads policy approves third-party tags from CM360 and Flashtalking (click and impression), Innovid, IAS and DoubleVerify (all 1x1), states 'CMX does not accept ad blocking tags', offers Reddit verification through server-to-server DV and IAS integrations, and makes any other tag subject to approval; the same policy says the advertiser 'agrees not to use any data obtained from CVS for measurement or analytics purposes' without prior written consent, so buyer-run outcome analysis is documented only as consent-gated, and no clean room, event-level export or reporting API appears on the seven CMX pages read. | 5 | vendor assertion | supported with qualification | high | ||
| eBay documents a seller-side Advertising dashboard and a Marketing API whose Promoted Listings reporting service creates 'five different types of reports' retrieved with getReport (not runnable in the Sandbox); per-sale 'Attributed sale' rows appear in the 'Sale type' reporting column. No exposure-level (impression or click) log, clean room, verification-tag policy or buyer-run outcome analysis is documented for any eBay row on the pages read. | 5 | vendor assertion | supported as stated | high | ||
| Cardlytics documents exports and two data feeds: CSV export and scheduled email reports from the Reporting Hub, a REST performance API returning offer-level impressions, reach, activations, engagements, purchases and revenue aggregated across flights, and an hourly Redemption Feed in which 'Each row in the feed represents a unique, anonymized transaction where a Cardlytics offer has been redeemed', with identifiers 'further obfuscated to ensure no cross-partner PII linkage'; the feed carries redemptions only, and no exposure-level export or buyer-run analysis environment is documented. | 5 | vendor assertion | unresolved | high | ||
| Numerator's media page names retailers only as purchase-data coverage ('Sephora, Home Depot, Walmart, Amazon, Aldi, CVS, Kroger, & McDonalds'), not as networks measured, and documents three route families for media clients: sales lift studies, 'survey-based studies, and data feeds for use in measurement models', with the SecureMeasure case adding a clean-room match; no retail media network, API, dashboard for lift results, or network count appears on the four Numerator pages opened. | 5 | vendor assertion | supported as stated | medium | ||
| Amazon documents several routes by which a Sponsored Products buyer receives outcome data: the Ads Console, the Amazon Ads API (Reporting API v1 and version 3, to be shut off after 30 June 2027), Amazon Marketing Stream ('Complete advertising datasets delivered to your S3 bucket as Parquet files, with updates as frequent as every 15 minutes'), Marketing Mix Modeling reports, Amazon Marketing Cloud by self-service, and partners on the Amazon Ads Partner Network. | 5 | vendor assertion | supported as stated | high | ||
| Buyer-side analysis of Instacart Sponsored Products outcomes runs through operator-defined outputs: Ads Manager dashboards with CSV exports and a separate Impressions & Clicks export, an Ads API licensed to agencies whose tools must 'Show ROAS Metric and associated calculation' and 'Explain IC attribution windows', and the Snowflake-based Data Hub clean room where 'queries initiated by a user are allowed' but access 'not guaranteed' and no output threshold or export rule is stated; the ads terms make Instacart Data and Reporting Information confidential and contain no clause on third-party measurement. | 5 | vendor assertion | supported as stated | medium | ||
| 84.51°'s public help documentation for the Kroger Ad Platform and its API is behind a sign-in: the Kroger Ad Platform Learning Center redirects to an Okta login, and the 84.51 API Catalog shows only 'Sign in' and 'Not a member yet? Contact [email protected] to get access.'; this review therefore found no public documentation of a KAP reporting API, event-level export or buyer-run query environment for onsite outcomes on the pages read by 3 November 2026. | 5 | synthesis | supported as stated | medium | ||
| 84.51°'s Collaborative Cloud is documented, in a Vestcom release of 16 September 2022, as 'a platform built for data scientists, providing access to the shopping behavior of Kroger's 60MM households with transaction-level granularity' in a 'hosted environment that has been built to the individual client's specifications', through which Vestcom, 'the first in-store media provider to gain direct access', reports the incremental sales of its own shelfAdz media using 'rigorous test and control measurement'; no page read documents Collaborative Cloud access for buyers of KPM's onsite or offsite services. | 5 | vendor assertion | supported with qualification | medium | ||
| Walmart Connect's accreditation post (16 October 2025) adds that the accreditation 'is in addition to Walmart Connect’s rich onsite partnerships with third-party measurement leaders such as DoubleVerify and Integral Ad Science', without stating which metrics, placements or rows those partnerships cover or whether their tags run on Sponsored Search; Walmart's help centre (Data use and privacy requirements, last updated 23 June 2026) scopes the DoubleVerify and IAS tags to Onsite Display, as viewability trackers approved for monitoring only (blocking not permitted), on desktop and mobile web (no in-app viewability); no verification-tag route is documented for Sponsored Search. | 5 | vendor assertion | supported with qualification | high | ||
| Sponsored Search reaches buyers through four documented route families: the self-serve Ad Center (dashboards that 'Compare performance across different timeframes and attribution windows' and let users 'Download custom views for offline analysis'), the Sponsored Search APIs for 'Full-Service' and 'Reporting Only' API partners, agency and technology partners in the Partner Network, and a managed service ('managed advertisers'); since 8 October 2025 API snapshot reports can be retrieved for 62 days (previously 90), 'aligning with Walmart Ad Center'. | 5 | vendor assertion | supported as stated | high | ||
| Instacart's Data Hub (announced 6 January 2026, in pilot with select agencies and CPG brands) lets marketers join their own first-party data with Instacart's purchase signals in a clean room, build audiences for off-platform activation and "Measure campaign impact with flexible attribution approaches"; the release does not say that off-site media exposure is joined to Instacart purchases or who runs the queries. | 5, 6 | vendor assertion | supported as stated | medium | ||
| Amazon documents that Retail Ad Service retailers receive, in a dedicated Retailer console, reports of ad revenue, ad-serving funnel metrics, average cost-per-click, click-through rate, attributed sales (units and sale amount) and advertiser ROAS by advertiser, product category and keyword, with ad measurement performed inside AWS Clean Rooms producing aggregated, anonymised reports for the retailer and its advertisers; no attribution window, touch model, match basis or control design is documented. | 5, 9 | vendor assertion | supported as stated | medium | ||
| Outcome-feed provenance (dimension 3) is documented unevenly across the 47 product-by-network retail media feed rows recorded (ChatGPT rows excluded): 5 rows document origin plus at least two of granularity, constraints and limitations (Perpetua x Walmart Connect; Intentwise Optimize x Walmart Connect; Intentwise Optimize x Amazon Ads; Xnurta x Criteo; Criteo Commerce Max x Criteo's own retailer network), of which 2 document all four (Intentwise Optimize x Amazon Ads; Criteo Commerce Max x Criteo's own retailer network); 6 rows document origin plus one further field; 5 document origin only; 31 are named integrations with no provenance field documented in the sources opened. | 5, 9 | synthesis | supported with qualification | medium | ||
| Commerce Max documents the limits of its outcome feed: metrics 'are updated with a delay, typically between 8 and 24 hours', collection is 'close to real-time' for most retailers but slower for some, dashboard figures 'can change and reflect new events, even if your campaign has stopped' (late in-window sales raise them, invalid-traffic removal lowers clicks), offline (in-store) sales are 'imported once every few days' and take '3 days (around 74 hours)' to appear after ingestion, and 'not all retailers provide offline sales data'. | 5, 9 | vendor assertion | supported with qualification | high | ||
| Albertsons Media Collective launched on 7 January 2025 an API 'that will enable advertisers to integrate Albertsons Cos. campaign performance data into their own measurement models for analysis', delivered to a destination of the advertiser's choice and 'now available for all clients', plus a TransUnion partnership that supplies campaign data to TruAudience marketing mix modelling; the release does not state the API's fields, granularity or whether sales outcomes are included. | 5, 10 | vendor assertion | supported as stated | medium | ||
| On the pages read, buyers of Sponsored Search receive Walmart-defined aggregate reports (Ad Center dashboards and downloads; API snapshot report types such as 'attributedPurchases', which returns purchased and advertised item IDs and an attribution type), and no clean room, buyer-written query, event-level exposure export or independent replication is documented for Sponsored Search on walmartconnect.com, the Sponsored Search developer portal or Walmart's measurement pages read on 3 November 2026; the help centre was a JavaScript shell to every route tried. | 5, 10 | synthesis | supported with qualification | medium | ||
| Amazon documents that advertisers using sponsored ads can access Amazon Marketing Cloud (a clean room built on AWS Clean Rooms) by self-service, at no cost to eligible advertisers, and run their own SQL there over event-level Sponsored Products impression and click records, joinable to conversion events that Amazon has attributed to those traffic events (14-day window), receiving only aggregated, anonymous outputs. | 5, 10, 12 | vendor assertion | supported with qualification | high | ||
| Amazon Marketing Cloud documents buyer-run SQL queries, self-service and at no cost for sponsored ads advertisers, over a table with one record per Sponsored Products impression and per click, joinable to conversion tables that pair traffic and conversion events, with only 'aggregated, anonymous outputs' leaving the clean room under per-column aggregation thresholds whose numeric values are not published. H3 axis (b) is documented at the exact service scope (Amazon Sponsored Products): this is an established counterexample, not a candidate, so H3 is contradicted at step 1 of its decision order whatever the other searches show. Axis (a) (event-level exposure data available to the buyer or an independent measurer) is not documented: queries run over event-level records but no event-level export is documented. | 5, 13 | vendor assertion | supported with qualification | medium | ||
| H4b record for Kroger Precision Marketing offsite programmatic: exposures arise in partner DSPs (The Trade Desk, Yahoo DSP, DV360 via 'Direct Connect' self-service in KAP), transactions are Kroger loyalty purchases, and for DV360 the join is described as 'CAPI-based online and in-store sales measurement', while the FAQ says KPM 'uses hashed identifiers ... and employs privacy-preserving environments (clean rooms) where appropriate'; the exact identity match for each DSP is not described. | 6 | vendor assertion | supported as stated | medium | ||
| eBay Promoted Offsite promotes listings 'in external search channels' (Google is the channel named) on CPC only and reports Direct and Halo sales on the same 30-day click rule as priority campaigns; the documented join is a click on the off-site ad followed by a purchase on eBay, so no identity match outside eBay's own surface is documented for this off-site row. | 6 | vendor assertion | supported with qualification | high | ||
| Walmart announced its partnership with The Trade Desk to build a demand-side platform on 28 January 2021 (planned 'in time for the 2021 holiday season') and on 25 August 2021 said 'Walmart DSP will be available for select suppliers by the end of October'; the 2021 texts describe the DSP as 'Closing the measurement loop across digital and in-store sales' by 'leveraging Walmart's purchase data to measure both online and in-store' and name no identity match mechanism, window or control. | 6 | vendor assertion | supported as stated | high | ||
| Disney Advertising's release of 8 May 2024 says Walmart Connect's measurement on Disney's addressable streaming inventory (Hulu and Disney+) is 'enabled by a sync to Disney's Audience Graph via clean room technology', letting advertisers 'measure the impact of their campaigns through closed-loop attribution' using 'Walmart's proprietary omnichannel purchase data', bought 'through a private marketplace programmatically', as a 'proof of concept' launching in Q2 2024 'with select advertisers'. | 6 | vendor assertion | supported as stated | high | ||
| The Home Depot's release of 22 April 2026 announces 'a new beta integration with Pinterest Media Network Connect' through which 'brands not sold at The Home Depot' can run Pinterest campaigns on 'more than 20 first-party audiences', calls it 'the third environment where non-endemic advertisers can activate The Home Depot audiences, following integrations with The Trade Desk and Yahoo', and names no attribution, clean-room or closed-loop mechanism; Orange Apron Media's measurement page separately promises closed-loop measurement 'to both online and in-store sales and conversions' and a clean-room offer for 'publishers and other partners'. | 6 | vendor assertion | supported with qualification | medium | ||
| Instacart's release of 10 June 2025 (PR Newswire, read through a reprint because the issuer host geo-redirects) says agency partners such as Omnicom's Flywheel 'can access closed-loop measurement from Instacart across all of their campaigns' and that brands 'can access real-time sales measurement signals from Instacart directly in The Trade Desk platform', for audiences used 'across channels like CTV, audio, video, and display'; it describes 'Instacart's deterministic purchase-based data' but names no identity mechanism (no UID2, clean room or hashed identifier), and mentions a 'beta' only in Danone's quote. | 6 | vendor assertion | supported with qualification | medium | ||
| Amazon Marketing Cloud's page describes 'a secure, privacy-safe, and cloud-based clean room solution' joining 'Amazon Ads signals as well as their own inputs' and 'onboarded third-party providers', with 'hundreds of fields ... including ad-attributed impressions, clicks, and conversions', returning 'only ... aggregated, anonymous outputs' under 'built-in aggregation thresholds' (no number stated), to 'Quantify advertising impact at online and offline Amazon stores, advertiser-owned sites, and beyond', 'at no cost to eligible advertisers'; Amazon's DSP product page itself carries AMC, attribution, sales lift and closed loop only as navigation links. | 6 | vendor assertion | unresolved | high | ||
| Albertsons Media Collective's off-site offering runs 'Offsite Display through Google's DV360 and Collective TV' and a self-service route with The Trade Desk ('helping advertisers understand the connection between their CTV campaigns across the open internet through TTD and customer sales'); its 2026 multi-touch attribution release, 'Developed in collaboration with LiveRamp', 'connects impression-level exposure data across Albertsons Media Collective's owned and operated channels' and offsite partners (Meta; DV360, Google Ad Manager, Perion) 'with Albertsons transaction data in a neutral, privacy-conscious environment', 'isolating the incremental contribution of each touchpoint to a sale' with PepsiCo as 'early pilot partner' and no method, assignment or availability stage stated. | 6 | vendor assertion | supported with qualification | medium | ||
| Roundel documents off-site measurement for Programmatic by Roundel ('Receive closed-loop programmatic reporting directly in demand-side platforms like The Trade Desk'; 'key sales metrics directly via The Trade Desk or Reports') and CTV by Roundel ('closed-loop reporting with product sales numbers, along with standard video metrics'; 'new purchaser insights data for products sold at Target (availability based on investment level)'), on 'First-party data from 100M+ Target Circle members', and its measurement page says only that 'We use privacy-conscious methodologies and secure data collaboration environments'; no partner, window or control is named for either rail. | 6 | vendor assertion | supported with qualification | medium | ||
| Dollar General's release of 7 June 2022 says DG Media Network delivers 'closed-loop 1:1 measurement, insights and reporting that identifies both attribution and incrementality' and names 'Live Ramp' for 'data collaboration and advanced data clean room capabilities' and Goodway Group as partners; its 13 April 2026 release says the network 'integrates point of sales (POS) data' into 'its closed-loop reporting'; DG Media Network is on The Trade Desk's 17 June 2026 partner list; no page opened states the join mechanism beyond the clean-room partnership or any control design. | 6 | vendor assertion | unresolved | medium | ||
| Walgreens Advertising Group's offsite page (footer 2024) promises 'SKU-level targeting and omnichannel, closed loop reporting features' with no mechanism, partner, window or control; the closed-loop sentence follows the Trade Desk and OpenX activation sentence (so it is tied to the off-site offer), CTV runs 'with Infillion', and 'wagID' is a targeting construct built from loyalty sales data, not a documented measurement join; the group is on The Trade Desk's 17 June 2026 partner list; Lowe's Media Network's releases of 13 October 2021 and 8 August 2024 promise 'In-depth measurement and closed-loop reporting' for the whole network without naming a channel, mechanism or measurement partner, so Lowe's off-site measurement is documented only as a pooled network-level statement. | 6 | vendor assertion | supported with qualification | medium | ||
| As tallied by stream H before its three open linkage leads were read on 3 November 2026: of the 13 cohort A rows that tie outcome measurement to an off-site channel, 2 (15%) document in their own pages that the off-site exposure-to-transaction join runs outside the operator's logged-in surface: Walmart Connect Offsite Display (Disney's Audience Graph via clean room, 2024 proof of concept; LiveRamp's data clean room for Meta, 2026) and Albertsons Media Collective offsite (impression-level exposure from offsite media, Meta, DV360, Google Ad Manager and Perion joined with Albertsons transactions in a LiveRamp-powered neutral environment; 1 of 13, 8%, if that environment is not accepted as a clean room). Counting network-level clean-room or data-collaboration statements as well (Kroger, Roundel, Chewy, Best Buy) gives 6 of 13 (46%); over the broad denominator of 22, 2 (9%) strict and 9 (41%) lenient. Under the hypothesis's own rule the verdict is unresolved, not contradicted: the linkage searches for Roundel, CVS and Instacart are incomplete (partner and operator clean-room announcements surfaced as leads and were not opened), and Kroger's pages refuse the fetch tool. After those leads were read, and after the coding rule was applied as written on 3 November 2026, the paper's count is 3 of 13 (RTL-HYP-H4b). | 6 | synthesis | supported with qualification | medium | ||
| eBay documents that a Promoted Offsite campaign earns a 'Direct sale' credit 'when a buyer purchases the same item' from the clicked ad within 'the most recent 30 days' and a halo credit when the buyer 'then purchases a different item from the same seller within 30 days of that click', with ad fees 'charged on a cost per click basis': an off-site outcome join that runs through a click to eBay's own marketplace, with no partner identity match described. | 6 | vendor assertion | supported as stated | medium | ||
| Best Buy Ads documents 'closed-loop, SKU-level reporting inclusive of online and in-store sales, in near real-time' for formats that include offsite display and offsite video, and an off-site incrementality design, 'Time-Based Regression Geographic Experimentation', 'splitting media investment geographically across regions that behave similarly', while on-site testing uses 'customer-level holdouts'; no identity mechanism is described for the off-site reporting. | 6 | vendor assertion | supported with qualification | medium | ||
| Chewy Ads lists 'Closed loop measurement' as a bullet under its Offsite Search & Social product (Google PLA, Meta and TikTok), reports an average 'LTV ROAS' footnoted to 'all 2026 on and offsite campaigns', and describes 'Chewy Marketing Cloud: Clean Room solutions powered by LiveRamp to inform campaign optimization and performance' at network level, without tying the clean room to an off-site rail. | 6 | vendor assertion | supported as stated | medium | ||
| DICK'S Media's offsite page (social, search, display, CTV and online video 'across the open web, social, and streaming platforms') carries the same measurement sentence as its onsite page, 'we track every touchpoint with closed-loop measurement to show your media's true impact.', with 'our first-party loyalty data' as the data claim and no window, partner or join mechanism. | 6 | vendor assertion | supported as stated | medium | ||
| Six more cohort A rows that sell off-site media document outcome measurement only as a network-level statement not tied to an off-site channel: Gulp Media Network ('Exposed customer purchase behavior will measure the ROAS for the campaign based on measured sales in C-Shopper.'), Hy-Vee RedMedia ('effective offline and online sales attribution'), Macy's Media Network ('Our closed-loop reporting tracks online and in-store conversion activity'), Petco RMN (a dashboard 'including ROAS, Sales, CTR, ROI and more'), UB Media ('Our closed-loop reporting across channels measures campaign-level ROAS') and Lowe's Media Network ('In-depth measurement and closed-loop reporting'); none names an identity mechanism. | 6 | synthesis | unresolved | medium | ||
| The IAB and IAB Europe in-store standard (final, December 2024) asks retail media networks to report 'ad play and gross impressions' now and to disclose 'The formula used', defines 'Ad impression = audience x ad play', names opportunity to see as the future 'best proxy available for a viewable ad impression in the In-Store environment', asks that 'The calculation used to determine the ad impression' and how store traffic was gathered be disclosed, points in-store measurement to the 'MRC Digital Place-Based Audience Measurement Standards where applicable', and says sales lift 'should exclude any stores where the media is not activated'. | 6 | direct record | supported as stated | high | ||
| STRATACACHE's news item of 28 January 2026 says Albertsons 'has officially debuted its in-store digital display network, powered by STRATACACHE technology', 'reaching approximately 800 stores across the Albertsons footprint by 2026', under a bullet labelled 'Closed-Loop Measurement' that promises brands 'proof of play' and direct sales attribution; STRATACACHE's in-store retail media page offers only 'Accurately measure campaign performance to optimize marketing spend.' and no impression method. | 6 | vendor assertion | supported as stated | medium | ||
| Grocery TV's pages and its 2026 media kit describe 'the largest in-store retail media network' with outcome claims ('Measure brand lift and sales impact tied to actual purchases with an average of 14% sales lift and 27% brand awareness across brands', while the media kit gives an 'Average sales lift' of '+14%' and 'Average brand awareness lift +22%'; 'Understand the impact on sales, ROAS, and iROAS both in Grocery TV stores and in other retailers'; 'an average 49% increase in incremental reach'; case figures of 19% sales lift for a personal care brand and 34% brand awareness lift for a FinTech app, none with a design, base, period or measurer), but none of the four documents opened states how impressions are counted, which test design produced the lifts, or who measured them. | 6 | vendor assertion | supported with qualification | medium | ||
| Stingray Advertising's release of 20 June 2024 reports a '20.3% average sales lift' for CPG brands across 100 matched panel tests 'Using matched panel analysis with data from Nielsen Scantrack', 'comparing sales data from stores where the ads were played to similar stores where no ads were played', and says 'This method relies on random assignment and statistical methods to account for variables such as price'; it gives 'over 21,100 stores' in the US and 'more than 30,000 major retail locations' in North America without reconciling them, and labels the network retail media. | 6 | vendor assertion | supported as stated | medium | ||
| The in-store audio services of three cohort A networks document no audio-specific impression or outcome method on the pages opened: Dollar General announced on 13 April 2026 'the planned rollout of an enhanced, AI-enabled in-store audio network across approximately 6,000 DG stores in 48 states' under a network-level closed-loop statement; Lowe's lists 'Audio advertising in Lowe's stores nationwide' under network-level 'In-depth measurement and closed-loop reporting'; and Walgreens' only closed-loop sentences sit under coupons, not audio; Gulp Media Network claims 'an average overall sales lifts of 5–9%' for Gulp Radio with no design, base, period or control, says lift is measured on 'first-party data' and documents ad-play logging ('Qsic's system timestamps each ad'; ad plays, not impressions). | 6 | synthesis | supported with qualification | medium | ||
| CVS Media Exchange's Social page, under the heading 'Attributed sales on Pinterest', offers 'the attributable sales data you need to optimize your campaigns' and says 'CMX's innovative partnership with Pinterest and LiveRamp utilizes clean room technology to create a data-safe targeted approach'; the page ties the clean room to targeting and does not say the sales attribution is computed there, nor describe the join keys or who sees what. | 6 | vendor assertion | supported with qualification | medium | ||
| Instacart's off-platform partnerships page (capture of 1 March 2026) offers "closed-loop measurement" with NBCUniversal, Pinterest, Roku and TikTok and "Instacart first-party audiences and measurement" inside The Trade Desk, but states no identity match, clean room or exposure-to-purchase join for any partner. | 6 | vendor assertion | supported as stated | high | ||
| Instacart's release of 7 October 2025 says it is integrating its purchase data into TikTok Ads Manager so that select CPG advertisers can measure campaign performance there with Instacart conversion data ("closed-loop measurement on purchases"); it does not say how Instacart purchases are matched to TikTok exposure (no identifier, clean room or match process is named). | 6 | vendor assertion | unresolved | medium | ||
| Albertsons Media Collective sells near-store digital out-of-home on '120,000 digital screens near stores' with 'incremental measurement that proves performance', and its Primo Water case (April to May 2025, powered by Perion) reports 'a 5.5% incremental sales lift' with no design, control, sample or interval stated. | 6 | vendor assertion | supported with qualification | medium | ||
| 84.51°'s managed-service documentation lists the channels an account team runs (Brand Shop, Custom Influencer, Display Ad, Email Module, Meta Ad, Online Video Ad, Pinterest Ad, Premium Placement Ad, Programmatic Audio, Programmatic CTV, Push Notification, Single Subject Email, Snapchat Ad, Targeted Digital Coupon), says the campaigns 'benefit from an integrated KPM ecosystem connecting digital and in-store touchpoints', applies 'a rest-of-market multiplier', reads results after an attribution window of 'approximately two (2) weeks', and defines a Display Ad as 'embedded within a non-Kroger Owned and Operated Page'. | 6 | vendor assertion | supported as stated | medium | ||
| Walmart Connect's off-site rails beyond social, as dated by their own sources: a Walmart-Roku shoppable pilot announced 16 June 2022 and run 22 November 2022 to 10 February 2023; Disney's 8 May 2024 release of a proof of concept measuring Hulu and Disney+ campaigns through 'a sync to Disney's Audience Graph via clean room technology'; the VIZIO NewFronts release of 23 March 2026 ('closed-loop sales attribution'); VIZIO inventory through Yahoo DSP and Magnite in 'a closed proof of concept phase' (28 May 2026); and Spotify and live linear TV private-marketplace deals in Walmart DSP (6 October 2026); none of these pages states how an off-site exposure is matched to a Walmart purchase except the two clean-room statements. | 6 | synthesis | unresolved | high | ||
| The Walmart DSP was announced with The Trade Desk on 28 January 2021 and said on 25 August 2021 to be 'available for select suppliers by the end of October' 2021, with 'Closing the measurement loop across digital and in-store sales' as its measurement claim; today Walmart sells it as 'Powered by The Trade Desk' for 'the open web, social media, CTV, audio, gaming and more', to endemic and non-endemic advertisers, with no DSP-specific window, match method or lift design on the pages read. | 6 | vendor assertion | supported with qualification | high | ||
| Albertsons Media Collective measures in-store screen campaigns with a 'matched market framework' that compares 'test stores exposed to ads against a rigorously selected set of control stores with no media exposure' matched on 'nearly 60 variables'; its Mondelēz case (6 January 2026) reports in-store screens 'across 116 Albertsons banner stores', '$2.41 incremental return on as spend' and '14% sales lift in-store' for its 'first in-store Matched Market campaign', with no store allocation rule, control-store count, period or interval, and calls closed-loop measurement a future capability. | 6, 8, 13 | vendor assertion | supported as stated | medium | ||
| This review found no Samba TV partnership with a retail media network on samba.com (newsroom listing of six 2026 items read; one web search on 3 November 2026) or on any network page opened; Samba's documented purchase-outcome partnerships are with Catalina (release of 12 August 2020: 'a 1:1 deterministic match between TV exposure and in-store purchase' on 'more than five million frequent shopper cards', for Georgia-Pacific) and Affinity Solutions (16 October 2023: 'deterministic purchase data capturing over 140 million credit and debit cards' for 'cross-platform measurement of return on ad spend (ROAS)'), neither of which describes measurement of a retail media network's media: the Catalina arrangement measures a CPG brand's own TV and digital advertising against frequent-shopper purchases, and the Affinity integration measures against card transactions. | 6, 9, 14 | vendor assertion | supported with qualification | medium | ||
| On 19 October 2026 the FTC announced a complaint and a proposed consent order, out for 30 days of public comment, under which Publicis Groupe would complete its US$2.2 billion acquisition of LiveRamp subject to five years of 'neutral and nondiscriminatory access' to LiveRamp's products and firewalls against Publicis's access to LiveRamp customer data, with an independent auditor 'every six months for the first two years after the date that the proposed transaction closes'; the release treats the acquisition as not yet completed. | 6, 10 | direct record | unresolved | high | ||
| Walmart Connect said on 3 April 2026 that it is 'evolving from our existing identity framework to Walmart's more advanced, proprietary identity graph that connects in-store transactions, online activity and media engagement', that it is introducing 'individual-level targeting for Onsite Display', that the approach 'will extend to Walmart DSP and Social' over time and into CTV 'through our integration with VIZIO', and that it is bringing in 'a new unified attribution model across Sponsored Search, Onsite Display and Walmart DSP'; no match method, matchable share or effective date is stated. | 6, 10, 13 | vendor assertion | supported with qualification | medium | ||
| Walmart Connect's store ads page says 'Gross impressions are estimated projections based on historical and aggregated data.', 'Store ads measurement includes both brand and sales lift.' and 'Sales lift validated at the 80% confidence level or above.', citing a Deli and Bakery screen proof-of-concept that 'achieved up to 5% sales lift for Food suppliers'; it names no test design, control stores or impression formula, and Walmart's 2021 release counted 'nearly 170,000 digital screens across 4,500+ stores'. | 6, 13 | vendor assertion | supported with qualification | medium | ||
| On MRC's digital accredited-services list as read on 3 November 2026, six entries concern register services or their operators (Amazon Ads Sponsored Products, Amazon DSP, Criteo, Epsilon, Instacart, Walmart Connect Sponsored Search), and every accredited metric in them is a delivery metric (impressions, clicks, CTR, viewable impressions, invalid-traffic filtration) except the entry 'Epsilon', which lists 'Correlated Outcomes (Online and Offline)' for D, MW and MA. | 7 | direct record | supported with qualification | high | ||
| MRC's letter of 1 October 2025 continues the accreditation of a service named only 'Epsilon' for 'Display and Video Impressions and Correlated Outcomes (online and offline) within Desktop, Mobile Web and Mobile In-App environments, reported Gross and Net of General Invalid Traffic (GIVT)', under the Minimum Standards, the IVT 2.0 General Traffic provisions, MRC's Data Integration Guidelines, the IAB display and video guidelines and the MRC/IAB/MMA mobile guidelines; the letter names no product, no retailer, no retail environment and does not cite the IAB/MRC Retail Media Measurement Guidelines. | 7 | direct record | supported as stated | high | ||
| MRC's letter of 15 March 2024 grants Criteo's Retail Media Platform (RMP) initial accreditation for 'Display Impressions and Clicks reported Gross and Net of General Invalid Traffic (GIVT) within Desktop, Mobile Web and Mobile In-App', 'specifically for On-site Sponsored Ads, Commerce Display Ads and Standard Banner Ads', with accredited metrics Impressions, Clicks, Invalid Impressions and Invalid Clicks, under the Minimum Standards, IVT 2.0 GIVT provisions, the IAB Display Impression and Click Measurement Guidelines and the mobile guidelines; the IAB/MRC Retail Media Measurement Guidelines are not cited. | 7 | direct record | supported as stated | high | ||
| MRC's letter of 1 June 2026 continues accreditation of 'Instacart's Ads Manager Platform' for display and video impressions, clicks and CTR (gross and net of GIVT) and viewable impressions and viewability metrics (net of GIVT) 'across Instacart Ads and Carrot Ads within Desktop, Mobile Web and Mobile In-App', 'specifically for Sponsored Products Ads (display), Display Ads, Shoppable Display Ads and Shoppable Video Ads placements served on the Instacart Marketplace as well as on retailer properties through Storefront and Storefront Pro integrations and Connect API integrations'; MRC's Q4 2025 update records the Carrot Ads grant in October 2025. | 7 | direct record | supported as stated | high | ||
| MRC's letter of 29 May 2026 consolidates the January 2026 decisions (initial grants for Amazon Retail video BTR impressions and click-throughs and display and video viewable impressions; IMDb display viewable impressions; APS and 3PX display and video viewable impressions) and the May 2026 grant for Alexa connected devices, and continues display BTR impressions and click-throughs for Retail, IMDb, APS and 3PX (D, MW, MA), Fire TV display in CTV, and APS/3PX video. | 7 | direct record | supported with qualification | high | ||
| None of the six MRC accreditation letters in force on 3 November 2026 for register services (Epsilon 1 October 2025, Criteo 15 March 2024, Walmart Connect 26 September 2025, Instacart 1 June 2026, Amazon Sponsored Products 8 July 2026, Amazon DSP 29 May 2026) cites the IAB/MRC Retail Media Measurement Guidelines among the standards applied; each cites the MRC Minimum Standards, the IVT addendum and the IAB display, click, video, viewable-impression or mobile guidelines. | 7 | direct record | supported as stated | high | ||
| MRC's five quarterly accreditation updates from Q3 2025 to Q3 2026 record these retail-relevant actions: September 2025 Walmart grant (Sponsored Search/'Sponsored Products service') and Epsilon continuation; October 2025 Instacart grant for Carrot Ads and continuation for the marketplace; November 2025 IAS third-party reporting of Amazon DSP granted (Amazon O&O Retail and IMDb plus APD and 3PX; Fire TV and Twitch excluded); January 2026 Amazon DSP grants within Retail, IMDb, APS and 3PX; May 2026 Amazon DSP grant for Alexa connected devices and Instacart continuation; June 2026 Amazon Sponsored Products continuation; and no retail media action in July to September 2026. | 7 | direct record | supported as stated | high | ||
| Between the 13 September 2023 draft (open for comment until 13 October 2023) and the January 2024 final, the attribution passage changed from 'user-level attribution (also sometimes referred to as "closed loop")' under a heading 'Extrapolation' to 'event-level attribution (also sometimes referred to as closed-loop attribution)' under 'Modeled attribution for unknown data', and the incrementality passage widened from 'Experimental design forms the foundation of any incrementality test' with 'marketers should use RCTs' to 'Experimental design and non-experimental approaches form the foundation' with 'marketers are recommended to use RCTs'; the final adds, at the head of its 'General Measurement Principles' (p. 10), the scope sentence 'Retail Media, for purposes of this document, refers to marketing to consumers near their point of purchase or choice between brands or products and includes online and/or in-store advertising', which the draft does not carry; the final has 60 pages against the draft's 56. | 7 | direct record | supported as stated | high | ||
| The IAB and IAB Europe In-Store Retail Media Definitions and Measurement Standards (final, dated Dec. 03, 2024 on IAB's page; no version number) were finalised 'following a public comment period during September, October and November 2024'; they define five store zones, say that today RMNs 'should report on ad play and gross impressions' while 'opportunity to see is the best proxy available for a viewable ad impression in the In-Store environment', give an OTS example built on eight variables (p. 16), recommend sales reported 30 days pre-exposure and 30 days post-exposure with Sales Lift and Brand Sales Lift metrics (p. 17), define 'Sales Variance' and 'Brand Variance' (p. 18) and an 'Incremental Sales/Brand Lift' section (pp. 19-21) whose test-versus-control method says consumers are 'ideally' sorted randomly and requires group sizes and statistical significance to be reported, and state on p. 4, in a paragraph about the MRC Digital OOH Measurement Guidelines, that 'The MRC Guidelines have been used as the basis for the development of these standards', while IAB's web page says 'these standards build on the IAB/MRC Retail Media Measurement Guidelines'. | 7 | direct record | supported with qualification | high | ||
| IAB Europe's commerce media measurement standards moved from the 'Retail Media Measurement Standards' (V1, posted 25 April 2024 after 'a public comment period between February and March 2024') to V2 released on 22 January 2026 after a public comment period the standards text dates 'September to November 2025' and the comment release of 9 October 2025 closes on 'Friday, 14th November 2025'; V2 adds a formal incrementality definition (with 'approved methodologies' in the words of the 22 January 2026 release; the standard says 'one of the methods specified in the Guidelines'), gross and net sales definitions, quick-commerce metrics, five new-to-brand timeframes and a 30-day default lookback with mandatory flexible windows; its transition rule reads 'Between January and July 2026, companies may choose to be certified under either Version 1 (V1) or Version 2 (V2). After July 2026, certification will only be available under V2.'; and it promises travel and finance editions 'in 2026'. | 7 | direct record | supported with qualification | high | ||
| IAB Europe's Retail Media Certification Programme certifies retailers and ad tech companies after an independent audit against the standards by one of three auditors (AAM, ABC, CESP), in which organisations 'must demonstrate compliance with the mandatory requirements from the compliance table' and the programme allows 'compliance to be achieved during the audit itself', with a badge valid for two years 'from the date issued on the Certificate', fees of EUR 7,500 for RMNs and EUR 7,500 or EUR 15,000 for ad tech companies (below or above EUR 50m annual media revenue) plus a separately paid audit, a requirement of membership of a national IAB in Europe, and a scope that can be 'their whole retail media offering or specific ad products', published with the certificate; IAB Europe and the MRC have coordinated so that an MRC letter to accredited European RMN measurement providers 'can, in turn, be provided to IAB Europe to demonstrate independent validation of compliance with certain IAB Europe Retail Media certification requirements'. | 7 | direct record | supported with qualification | medium | ||
| AAM launched its Incrementality Certification on 9 June 2026 'to review and certify commerce media networks and intermediaries' incrementality methodologies' in three areas, 'measurement methodologies and data sources, reporting rigor and transparency, and disclosure practices', aligned with 'IAB and IAB Europe's Guidelines for Incremental Measurement in Commerce Media released in November 2025' and 'the Association of National Advertisers' upcoming guidelines for retail media'; certified organisations may use 'the AAM certification seal' and are listed on AAM's Assurance List. | 7 | direct record | unresolved | high | ||
| AAM's compliance guide presents the certification as 'independent validation that the methodology is valid and aligns with industry expectations', recognises the IAB/IAB Europe method families (experiments, model-based counterfactuals, econometric models, hybrid or proxy approaches), requires a defined counterfactual and 'a structured intervention (test vs. control)', traceable inputs and 'statistically meaningful signals', documentation of 'any reliance on modeled, inferred, or third-party data', disclosure of 'methodology, key assumptions, limitations', an annual review, and an AAM 'review and gap analysis'; it does not state a randomisation or holdout requirement, names only the IAB and IAB Europe guidelines, and does not say what the certification does not attest. | 7 | direct record | unresolved | medium | ||
| The ANA document AAM names as 'ANA Retail Media Network Measurement Standards' is, on ANA's own page of 18 August 2026, a research report titled 'Retail Media Measurement Standardization' that 'highlights key learnings and establishes an initial framework to help drive outcome comparability', gated behind a free log-in, with no sample size, fielding dates or authors on the open page; AAM's post of 2 September 2026 describes it as based on 'a survey of Association of National Advertisers (ANA) members' and recommending standardised baseline metrics (impressions, viewability, clicks, invalid traffic) and 'expanding the role of third-party assurance'. | 7 | direct record | supported with qualification | medium | ||
| ANA's release of 23 July 2024 for 'Retail Media Networks: Optimism Tempered with Caution' (survey 'fielded in November 2023', following 'an initial survey in Summer 2022') reports that '63% say retailers have a lot of or complete influence on a marketer's decision to use various RMNs', that '55% of marketers see the lack of standardization across platforms as the greatest challenge', 'followed by attribution to sales at 48%', and states no sample size; the respondents are 'ANA client-side marketers'. | 7 | direct record | supported as stated | high | ||
| Between the September 2023 draft and the January 2024 final the guidelines' normative wording softened: counted case-insensitively in the extracted text, 'must' falls from 49 to 35 occurrences, 'should' from 126 to 86, and 'is recommended' or 'are recommended' rises from 1 to 45 (49 when occurrences broken across a line are counted); for example the draft's 'Retail media organizations must use robust measurement methodologies and technologies' becomes 'Retail media measurement is recommended to use robust measurement methodologies and technologies'. | 7 | direct record | supported with qualification | high | ||
| ANA's release of 30 January 2023 for 'Retail Media Networks: A Forced Marriage or Perfect Partnership?' rests on an online survey of ANA members conducted 'Over the summer of 2022' in which 'A total of 138 members participated in the survey, 80 of whom utilize RMNs', plus 'nine qualitative follow-up interviews', and reports that 'an overwhelming 88 percent believe they are somewhat or heavily influenced by retailers' and that 'a lack of measurement standardization and transparency are preventing advertisers from deriving full value'. | 7 | direct record | supported as stated | high | ||
| AAM also offers a 'Retail Media Certification' (described in its post of 29 October 2025) that 'provides third-party, independent assurance that demonstrates applicable measurement guidelines' are applied, assessed against the IAB/MRC Retail Media Measurement Guidelines (US) or IAB Europe's standards, with certified companies listed on AAM's Assurance List; on 3 November 2026 no Assurance List filter among its sixteen filter labels was labelled retail media and a request for 'Retail Media Certification' returned no company; whether AAM lists Retail Media Certification holders under 'AAM Platform Certification' was not checked. | 7 | direct record | supported with qualification | medium | ||
| MRC's under-review entry 'Kroger (Pre-audit in process as of 6/2026)' covers 'Targeted Onsite Ads, Product Listing Ads, Onsite Video, Promoted Product Carousels' with 'Impressions, Clicks, Viewable Impressions, Attributed sales, Incremental Sales, aROAS, iROAS' in desktop, mobile web and mobile in-app; it is the only screened entry for a cohort A service that puts attributed and incremental sales and their ROAS forms into an MRC process, and it is a pre-audit in process, not an accreditation; no MRC entry names Kroger's offsite programmatic service. | 7 | direct record | unresolved | high | ||
| Carrot Ads, Instacart's platform for other retailers, is inside the same MRC letter of 1 June 2026 ('across Instacart Ads and Carrot Ads') for placements 'on retailer properties through Storefront and Storefront Pro integrations and Connect API integrations'; MRC's Q4 2025 update records the initial grant for Carrot Ads in October 2025; delivery metrics only. | 7 | direct record | supported as stated | high | ||
| Among Amazon's other rows, MRC accredits Amazon DSP delivery metrics by supply source (letter of 29 May 2026: Amazon Retail display and video impressions, click-throughs and viewable impressions; Fire TV display in CTV; IMDb, APS, 3PX and Alexa connected devices with their own scopes) and separately accredited Integral Ad Science's server-to-server third-party reporting of Amazon DSP in November 2025 (Fire TV and Twitch excluded); Sponsored Brands, display ads and Amazon Retail Ad Service are not on MRC's digital accredited list read on 3 November 2026. | 7 | direct record | supported as stated | high | ||
| MRC's letter of 26 September 2025 grants initial accreditation to 'Walmart Connect's first-party measurement and reporting of Sponsored Search' for display, video and search clicks, CTR, impressions and invalid metrics from Sponsored Product, Sponsored Brand and Sponsored Video placement types, with GIVT filtration, in desktop, mobile web and mobile in-app; MRC's list read on 3 November 2026 shows 'Rendered Ad Impressions: Display', 'Rendered Ad Impressions: Video', 'Clicks: Search' and 'Click-through Rate (CTR)'; no outcome metric is accredited and the letter does not cite the retail media guidelines. | 7 | direct record | supported with qualification | high | ||
| H2 screen result, decision order: step 1 (contradicted) is not met, because no letter in force on 3 November 2026 is verified as accrediting an outcome metric of an eligible cohort A, B or C service at an exact service, metric, environment and date (Epsilon's 'Correlated Outcomes' is the only outcome-type accredited metric, and Epsilon's own record ties it to Epsilon PeopleCloud Digital Media Solutions, formerly Conversant, which is not a register row; Kroger, Kevel and Schwarz carry outcome-type metrics only as under-review statuses, not letters); the review stops at step 2 and H2 is UNRESOLVED on this record, because the logged screen is dated 3 November 2026 rather than the 20 November 2026 cutoff, the accredited list was read by keyword rather than enumerated in full, and 8 of the 14 under-review entries were not named; the Epsilon applicability check is mapped, on the vendor's record, to a non-register service. Step 3 (supported) is reached only if the cutoff-date screen enumerates every digital entry and finds no other candidate. This was workstream A's reading before stream X enumerated MRC's lists in full on 3 November 2026; on the complete screen step one still finds no counterexample, and H2 stays unresolved at step two because the Epsilon mapping is an inference, not a resolved check (RTL-HYP-H2). | 7, 9 | synthesis | supported with qualification | medium | ||
| IAB Europe's compliance table (Update May 2026, v2.1) tests attribution reporting on pp. 5-8: sale definitions ('When the CMN doesn't specify which sale they are reporting on, it is assumed to be a gross sale'), the 30-day default lookback, same-SKU and halo attribution ('Same Brand, Same Category'), disclosure of sales extrapolation, and new-to-brand categorisation; its sales-uplift row (3.3) requires a 'Statistically significant A/B Test (not before and after test)' and the control 'Review methodologies for determining Sales Uplift. Review and test results.'; its incrementality row (4.1) says 'A CMN doesn't have to provide incrementality measurement to be certified but if they do they have to use one of the methods specified in the IAB Europe Incrementality Measurement Guidelines'; and its campaign audit (6.1) says 'Auditors will select 10 campaigns at random from the past 6 months and ensure that those campaigns comply with the applicable Guidelines 1. Media Metrics, 2. Attribution, and 3. Insights'. | 7, 9 | direct record | supported with qualification | high | ||
| IAB Europe's certified list carries three certificates, all audited by ABC: Albert Heijn (Retail Media Network, 16 September 2025; 'On-site display ads (static only)' and 'On-site sponsored products', with the footnote 'SIVT filtration and viewability for sponsored products were out of scope for this audit'), Nectar 360 (Retail Media Network, 15 June 2026; 'Display and video ad products for off-site properties' including Meta Instagram and Facebook, Google DV 360 and YouTube) and Moloco (Technology Provider, 7 October 2026; on-site standard display and sponsored product ads, web, mobile and in-app); the page states no standards version, country or validity period for any entry, and none of the three is a US register row (Moloco Commerce Media is a cohort C register row, so its certificate is recorded against that row with the scope unmapped to a US deployment). | 7, 9 | direct record | supported as stated | high | ||
| No Albertsons Media Collective service appears on MRC's accredited or under-review digital lists (41 and 14 entries enumerated on 3 November 2026), on AAM's Incrementality Certification list or on IAB Europe's certified list; Albertsons Media Collective is listed among the company participants of the IAB/MRC Retail Media Measurement Guidelines (January 2024), which is participation, not assurance, and its in-store release states that it is 'remaining compliant with and closely aligning our measurement roadmap with IAB current and future standards for in-store media', a self-asserted compliance claim with no attestation, not only an alignment statement. | 7, 9 | direct record | supported with qualification | high | ||
| Criteo's own descriptions of its MRC accreditation reach beyond the letter's scope: the FY2025 10-K says 'Our MRC-accredited technology provides our clients with the unique ability to measure against product sales at the product SKU level', and the Commerce Yield methodology note says MRC 'certifies measurement systems meeting the standards set by the IAB/MRC Guidelines', while the letter accredits display impressions and clicks (and their invalid counts) only and cites the MRC minimum standards, the IVT addendum and the IAB display, click and mobile guidelines, not the retail media guidelines. | 7, 9 | synthesis | supported as stated | high | ||
| MRC's letter of 1 June 2026 continues accreditation of 'Instacart's Ads Manager Platform' for display and video impressions, clicks and CTR (gross and net of GIVT) and viewable impressions and viewability (net of GIVT) 'across Instacart Ads and Carrot Ads' in desktop, mobile web and mobile in-app, for Sponsored Products Ads (display), Display, Shoppable Display and Shoppable Video placements on the Instacart Marketplace and on retailer properties through Storefront, Storefront Pro and Connect API integrations; MRC's Q4 2025 update records the Carrot Ads grant in October 2025; no outcome metric is accredited and the retail media guidelines are not cited. | 7, 9 | direct record | supported as stated | high | ||
| On 3 November 2026, MRC's services-under-review list for digital held 14 entries; five concern retail or commerce media: "DoorDash Ads (Pre-audit reviewed 1/2026, Audit not begun)" (impressions, clicks and CTR for Sponsored Products and Sponsored Brands), "Kevel (Pre-audit reviewed 9/2026)" (including "Clicks, Conversions and incrementality"; the entry names no retail environment, and Kevel is counted as retail media by the paper's register, not by MRC), "Kroger (Pre-audit in process as of 6/2026)" (including "Attributed sales, Incremental Sales, aROAS, iROAS"), "Schwarz Media Onsite Advertising (Audit reviewed 9/2026)" (including "shoppers reached, units purchased", for display carousel ads in the Lidl and Kaufland apps, not a US service) and "Walmart Connect Onsite Display (Audit in process as of 11/2025)" (delivery and viewability metrics); a listing under review is a status, not an accreditation. | 7, 9, 10, 12, 13 | direct record | supported with qualification | high | ||
| On AAM's Assurance List filtered to 'Incrementality Certification', the session's read of 3 November 2026 found one entry, DoorDash, with no date shown; AAM's programme page lists no certified companies itself and links to that list. | 7, 9, 12 | direct record | supported as stated | medium | ||
| MRC's letter of 15 March 2024 grants initial accreditation to 'Criteo’s Retail Media Platform (RMP)' for display impressions and clicks reported gross and net of general invalid traffic in desktop, mobile web and mobile in-app, 'specifically for On-site Sponsored Ads, Commerce Display Ads and Standard Banner Ads', with accredited metrics Impressions, Clicks, Invalid Impressions and Invalid Clicks under the MRC Minimum Standards, the GIVT provisions of the IVT Addendum 2.0, the IAB display impression and click guidelines and the mobile web and in-app guidelines; no outcome metric is accredited, the IAB/MRC Retail Media Measurement Guidelines are not cited, and Criteo's own Commerce Yield methodology note (11 August 2026) states that off-site inventory is outside the scope. | 7, 9, 12 | direct record | supported as stated | high | ||
| Instacart's own pages describe its MRC scope more broadly than MRC's letter: the methodology page (12 August 2026) says 'All of these impression metrics are accredited by the Media Rating Council' on a page whose platform list includes Caper and Rosie; Instacart's Caper Carts page (12 August 2026) says eligible shoppable display campaigns extend automatically to Caper Carts and 'All Ads Manager reporting displays aggregated data for all surfaces, including Caper Carts', with a Caper impression counted at '1 second on the screen'; and the 6 November 2025 release promises 'Verified Ad Metrics to More Than 240 Ecommerce Sites' in the U.S. and Canada while its footnote offers Carrot Ads to retailers using 'Storefront Pro, Connect, or their own e-commerce storefronts'. MRC's letter names only placements on the Instacart Marketplace and on retailer properties through Storefront, Storefront Pro and Connect API integrations; it names neither Caper nor Rosie nor retailers' own storefronts, and whether 'Connect API integrations' include Carrot Ads API partners is not stated. Inference: Ads Manager display totals that pool Caper impressions are not shown by the letter to be accredited as reported. | 7, 9, 13 | synthesis | supported with qualification | medium | ||
| MRC's letter of 8 July 2026 continues accreditation of Amazon Advertising's first-party measurement and reporting of Sponsored Products for 'Display and Search Clicks inclusive of Sophisticated Invalid Traffic (SIVT) filtration for Amazon placements', applying 'to total reporting' without environment segmentation (statistics: Gross Clicks; Clicks Total Net of SIVT); impressions and every outcome metric (attributed sales, ROAS) are outside the accredited scope, and the letter's standards do not include the IAB/MRC Retail Media Measurement Guidelines. | 7, 11 | direct record | supported as stated | high | ||
| DoorDash Ads is listed by MRC as 'DoorDash Ads (Pre-audit reviewed 1/2026, Audit not begun)' for 'Measurement and reporting of Impressions, Clicks and CTR - Desktop, Mobile Web and Mobile In-App', with 'Ad formats included in-scope: Sponsored Products and Sponsored Brands'; the same formats hold AAM's incrementality certification (RTL-A-F30), so the row carries two instruments at different stages for different metrics. | 7, 12 | direct record | supported with qualification | high | ||
| MRC's letter of 8 July 2026 continues accreditation of Amazon Advertising's 'first-party measurement and reporting of Sponsored Products' for 'Display and Search Clicks inclusive of Sophisticated Invalid Traffic (SIVT) filtration for Amazon placements', which 'applies to total reporting and does not include any environment (Desktop, Mobile Web, In-app) segmentations'; the accredited statistics are Gross Clicks and Clicks (Total Net of SIVT); a TAG Certified Against Fraud addendum of the same date records domain-list and data-center IP filtering; the retail media guidelines are not cited. | 7, 13 | direct record | supported as stated | high | ||
| IAB Europe's commerce media standards carry two version labels for one document: the 22 January 2026 release and the PDF body say 'V2' ('updated as of January 2026'), while the PDF cover reads 'IAB EUROPE COMMERCE (INCL. RETAIL) MEASUREMENT STANDARDS V2.1' and 'May 2026' (file created 19 May 2026, file name V2.1-May-2026), the knowledge-hub heading says 'V2.1' under a URL slug ending 'v2', and the compliance table's header reads 'Update May 2026' and 'Compliance Standards v2.1' while its PDF metadata title reads 'Compliance Table for Retailers (update Jan 2026).docx' and its file name says 'for Retailers' although its title is 'for CMNs (including RMNs, CMNs & Ad Tech)'. | 7, 14 | direct record | supported as stated | high | ||
| In the same 42-campaign analysis, each lever moved the average aggregate iROAS materially: pre-filtering to past brand purchasers cut it from $2.27 to $0.22 (category filter $2.20) while shrinking the test group to 17% of its unfiltered size; clustering gave $1.80 against $0.73 for PSM 1:many and $0.84 for PSM 1:1, with average standardised mean differences of 0.158 versus 0.013 and 0.012; dropping past brand sales as a PSM feature moved the average from $1.23 to -$0.14 (clustering stayed at $1.81); and observed-sales estimates averaged $1.56 against $0.97 under BSTS, a difference the authors put at '~90%', which is the mean of campaign/variation-level percentage differences, (iROAS - BSTS iROAS)/BSTS iROAS, over clustering and PSM 1:many variations only, not the gap between the two printed averages ($1.56 vs $0.97 is about 61%). | 8 | commissioned measurement | supported with qualification | high | ||
| Instacart's engineering post (30 June 2023) documents a user-level randomised holdout with ghost ads for Sponsored Products only: users are 'randomly split' by a hash of (user_id, advertiser_id) into an 'ads' group and a 'holdout' group before the auction, the outcome is total (organic plus attributed) sales of the advertised products, and the metrics are total incremental sales, incremental sales lift percentage and incremental sales on ad spend; the post states no significance rule, minimum sample or interval and describes the system's measurement, not what an advertiser receives by default. | 8 | vendor assertion | supported with qualification | high | ||
| The JMR 2017 version could not be opened (Semantic Scholar gives the bibliographic record only); a working-paper version of 18 February 2016 (NBER conference copy) was opened: it describes ghost ads as identifying 'the control-group counterparts of the exposed consumers in a randomized experiment', implemented on Google's display network, and demonstrated on an online retailer's display retargeting campaign; the design is also described by Instacart, the IAB/MRC guidelines and Gordon, Moakler and Zettelmeyer, as quoted. | 8 | synthesis | supported with qualification | low | ||
| Gordon, Zettelmeyer, Bhargava and Chapsky (working paper of 12 April 2018; published in Marketing Science 2019) compared 15 US randomised Facebook advertising experiments from January to September 2015 ('500 million user-experiment observations and 1.6 billion ad impressions'; 2 million to 140 million users each) with observational estimates and found that 'in half of our studies, the estimated percentage increase in purchase outcomes is off by a factor of three across all methods', that observational methods 'generally ... overestimate ad effectiveness relative to the RCT', and that six of the 14 checkout-outcome RCTs were not significant at 5%. | 8 | independent measurement | supported as stated | high | ||
| Blake, Nosko and Tadelis (NBER WP 20171, May 2014; eBay) found that stopping brand-keyword paid search cost almost no traffic because natural search substituted, and that suspending all non-brand paid search in 68 US DMAs for 60 days (65 matched control DMAs; 142 control DMAs in total) showed 'The entire regime of paid search adds only 0.66 percent to sales', with an experimental ROI of -63% (95% CI [-124%, -3%]) against OLS estimates of 'over 4,100%' and 'over 1,400%'; the test DMAs were drawn from a random candidate pool and then split by a matching algorithm, so 'the test group is not a purely random sample'. | 8 | independent measurement | supported with qualification | high | ||
| Lewis and Rao (QJE 130(4), 2015, pp. 1941-1973) analysed 'Twenty-five large field experiments' with major US retailers and brokerages, 'most reaching millions of customers' and 'collectively representing $2.8 million in digital advertising expenditure', and argue (per the article page, read in part) that individual sales are so volatile relative to per-person ad cost that the median confidence interval on return on investment is more than 100 percentage points wide and informative experiments can require more than ten million person-weeks. | 8 | independent measurement | supported with qualification | medium | ||
| Lewis and Reiley (QME 12, 2014, pp. 235-266) ran 'A randomized experiment with 1.6 million customers' of an anonymous major retailer on Yahoo!, finding that the advertising 'profitably increases purchases by 5%' (point estimate 5.0%, standard error 2.3%), that '93% of the increase occurs in brick-and-mortar stores', that '78% of the increase derives from consumers who never click the ads', and that a cross-sectional observational estimator 'exhibits a large negative bias three times the estimated experimental effect'; Yahoo! 'provided financial and data assistance and guaranteed academic independence'. | 8 | commissioned measurement | supported with qualification | high | ||
| Sahni and Zhang (QME 22(1), 2024, pp. 63-114) report 'a large-scale randomized field experiment in which a search engine varied the prominence of search ads for 3.3 million US users', finding that users shown fewer ads 'decrease their overall search engine usage' and that traffic to the newest decile of websites fell by 10%; it is a search-engine experiment on ad prominence and user welfare, not a marketplace sponsored-listing lift study. | 8 | independent measurement | supported as stated | medium | ||
| Gordon, Moakler and Zettelmeyer (arXiv 2201.07055 v2, 4 October 2022; Marketing Science 2023) analysed 663 treatment-control pairs from 563 US Facebook experiments run November 2019 to March 2020 ('approximately 7.9 billion user-experiment observations and over 38 billion ad impressions'; median experiment 30 days and about 7.3 million users) and found that double/debiased machine learning and stratified propensity-score matching, with over 5,000 user-level features, could not recover the experimental effects: median RCT lifts of 29%, 18% and 5% by funnel against 83%, 58%, 24% (DML) and 173%, 176%, 64% (SPSM). | 8 | independent measurement | supported with qualification | high | ||
| Einav, Leibtag and Nevo (NBER WP 14436, October 2008; QME 8(2), June 2010, pp. 207-239) validated Nielsen Homescan by matching 'thousands of individual transactions that were recorded by both the retailer (at the store) and the Nielsen Homescan panelist (at home)' at 'a large grocery chain', reporting unreported trips and recording errors in trip, product, price and quantity, with price errors 'more prevalent' and partly 'due to the way Nielsen constructs the price data'; the work was 'funded by a cooperative agreement between the USDA ERS and Northwestern University'. | 8 | independent measurement | supported as stated | medium | ||
| Abhishek, Jerath and Sharma (Information Systems Research, 2025, DOI 10.1287/isre.2022.0188) report 'a large-scale field experiment at Flipkart, a leading online marketplace in India' on sponsored listings in the most salient result positions, finding category-dependent effects (sponsored listings performed worse than displaced organic listings in electronics and better in clothing and footwear) and that 'increasing the fraction of sponsored listings (by about 10% points) does not affect clicks and conversions in any product category'; it is a platform-side study of listing mix, not an advertiser-level lift measurement. | 8 | independent measurement | supported with qualification | medium | ||
| This review found no public validation of a receipt or loyalty panel (Numerator, Circana, NCS) against point-of-sale data with a published design in the one search run and the pages screened by 3 November 2026; the only peer-reviewed panel-versus-retailer validation opened is Einav, Leibtag and Nevo on Nielsen Homescan (RTL-B-F15), and the Circana pages surfaced were marketing posts arguing that receipt panels need point-of-sale integration, not validation studies. | 8 | inference | supported with qualification | low | ||
| DoorDash's release of 2 September 2026 reports that a household-level lift study found a major beverage company's DoorDash media "drove $0.59 in incremental in-store sales per household", "three times the beverage category benchmark" (Circana's release of the same day: "three times the Circana Beverage Benchmark"). Neither release states who ran the study (Circana's release introduces it as "Early measurement results" within "New findings from DoorDash measurement"), its period, control construction, assignment or panel size; the eight-week US campaign of 10 November 2025 to 4 January 2026 is footnoted only to Circana's buyer-overlap analysis (76/13/10). | 8 | vendor assertion | supported with qualification | medium | ||
| Ipsos MMA's 'DoorDash-funded study' (post dated 13 October 2026) built marketing mix models for six quick-service restaurant advertisers, tested five representations of DoorDash advertising ('campaign orders, campaign sales, campaign spend, impressions, and clicks') and three prior settings, and reported that 'Campaign Orders performed best across all six advertisers'; DoorDash's release adds that 'an average of 42% of DoorDash's modeled effect carried into the following week', while Ipsos MMA describes the following-week effect as 'a small tail into the following week or two'. | 8 | commissioned measurement | supported with qualification | medium | ||
| Roundel's undated measurement page offers 'incrementality measurement' using 'methodologies like synthetic control, matched market and randomized controlled trials', footnoted as 'available for eligible campaigns', beside 'closed-loop measurement' with a 'multi-touch attribution model', and reports two cases without design, period or spend: a pet care brand's '280%' 'lift in CTR' and '10%' 'lift in ROAS', and Spin Master's '88%' 'increase in ROAS' and '40%' 'new-to-brand guests'. | 8 | vendor assertion | supported with qualification | medium | ||
| Fetch's post of 8 June 2023 says the company 'enlisted KPMG' to review its methodology and that KPMG found Fetch provides 'a uniquely clear view of campaign impact' through randomized controlled trials in which Fetch 'randomly assigns users' to treatment or control, 'withholds those offers from the control group' and uses 'two types of holdouts: partner level and offer level'; the KPMG report, its standard, period and payer are not on the page. | 8 | vendor assertion | supported with qualification | medium | ||
| Chase Media Solutions' LensDirect case (published 7 July 2025) reports a 90-day acquisition campaign from 15 September 2024 with '$878K revenue', '$509K in incremental sales', '10:1 ROAS', '$223 average order value', '175MM impressions' and '81MM unique impressions' on a $25-cash-back offer with a $130 minimum purchase, with no control, holdout or lift method stated. | 8 | vendor assertion | supported as stated | medium | ||
| PayPal's release of 6 January 2026 reports that 'Compared to the weeks prior to the campaign, Ulta saw a 20% increase in transaction spend via PayPal during the campaign' (footnote: 'PayPal Internal Data, Jun-Jul 2025'), a before/during comparison with no control group, beside a Lucid/Cint brand-favorability result '136% above the Lucid benchmark' and a 'Transaction Graph Measurement' suite offering 'campaign reporting, insights, attribution, and incrementality' with named validation partners. | 8 | vendor assertion | supported as stated | high | ||
| 7-Eleven's release of 12 November 2024 states that 'Products promoted through Gulp Radio 7-Eleven® locations have seen an average overall sales lifts of 5–9%' and that Gulp Media Network 'is a closed loop solution' able 'to track ROAS and iROAS', with no control, exposed group, test design, campaign count or period given for the lift figure. | 8 | vendor assertion | supported as stated | medium | ||
| Best Buy's post of 28 September 2026 announced 'a new multi-touch attribution solution' spanning 'onsite and offsite media, sponsored products, paid search, CTV, social, influencers, affiliates and in-store experiences' and 'expanded incrementality testing capabilities', and reported a Circana study in which 'Best Buy Ads outperformed the national media ROAS average by 17%', 'represented just 17% of media spend analyzed' while generating '35% of measured media contribution to sales at Best Buy', 'nearly 16% of media-driven sales at competing retailers' and '9.9% of the measured brand equity', with no study type, sample, period or commissioning party stated. | 8 | vendor assertion | supported as stated | medium | ||
| ANA's survey released 30 January 2023 was fielded 'Over the summer of 2022' among ANA members, with 'A total of 138 members participated in the survey, 80 of whom utilize RMNs', followed by 'nine qualitative follow-up interviews'; 88 percent 'believe they are somewhat or heavily influenced by retailers to buy advertising on their RMNs', and the base of that item (all 138 or the 80 RMN users) is not stated on either open ANA page. | 8 | direct record | supported with qualification | high | ||
| In the ANA 2023 survey (n = 138 members, summer 2022), 73 percent 'expect to be spending somewhat or significantly more on RMNs in the future', 56 percent 'are currently working with five or more different RMNs', and spending was described as 'shifted to RMNs from existing budgets for shopper marketing, brand marketing, and trade spending'; the words 'joint business', 'JBP' and 'pressure' do not appear on the ANA release; WARC's summary of the same report records a 'feel pressured by retailers to support RMNS' item (85%; see RTL-I-F01). | 8 | direct record | supported with qualification | high | ||
| ANA's second survey, released 23 July 2024 (report page 24 July 2024), was 'a quantitative survey of ANA client-side marketers' 'fielded in November 2023'; 63% 'say retailers have a lot of or complete influence on a marketer's decision to use various RMNs' and 'Nearly two-thirds of respondents still feel RMNs are a "have to buy" versus "want to buy"'; the sample size and the items' bases are not on the open pages. | 8 | direct record | supported as stated | medium | ||
| The Microsoft Advertising and Digital Shelf Institute joint-business-plan study is described by Microsoft (8 August 2024) as 'interviewing 18 senior-level executives across retailers and brands' and by DSI's report page as 'interviews 14 leading brands and retailers'; neither page gives interview dates, a sampling frame, a brand/retailer split or the method, the report is behind a DSI membership form, and Microsoft's sentence that improving the JBP process 'can increase ROI by 10%' carries no source. | 8 | vendor assertion | supported with qualification | high | ||
| Digiday (republished by Modern Retail, 25 March 2025) reported, on anonymous agency and CPG sources only, that some advertisers were leaving retailer joint-business-plan negotiations over 'the transparency in negotiations, otherwise known as the joint business planning (JBP) process', with one agency executive quoted: 'I've had two clients walk away from JBPs this year with Walmart because they felt that they got so shook down.' | 8 | synthesis | supported as stated | medium | ||
| Digiday's Marketing Briefing of 1 April 2025 reported, on five named agency sources, that brands were seeking flexibility in 2025 retail media joint business plans: Boris Litvinov (president, Left Off Madison) said 'If I was a brand, I'd be cautious about signing anything this year unless there's extreme outs and all this stuff.', Sammy Rubin (VP of integrated media, Wpromote) called flexibility a 'pressure release valve' against tariffs, and Hillary Kupferberg (VP of performance marketing, Exverus Media) described the ideal as 'no penalty for not reaching spend'. | 8 | synthesis | supported as stated | medium | ||
| This review found no survey of US advertisers with a stated sampling frame, field dates and n that measures retailer influence or joint-business-plan-linked retail media spend for 2025 or 2026, in the one web search run and the pages opened by 3 November 2026; the only opened surveys with influence items are ANA 2023 and ANA 2024 (member samples). | 8 | inference | supported with qualification | low | ||
| Albertsons Media Collective's 16 April 2026 release reports that an S. Martinelli & Co. multi-placement onsite campaign 'delivered a $7.45 iROAS, drove 65% new-to-brand buyers, and generated a 33% lift in sales', while citing its own research that 'iROAS can vary by 6.5x and flip results in 83% of campaigns based on methodology alone'; the case states no design, period, spend, sample or interval, and the release drops the study's qualifiers. | 8 | vendor assertion | supported as stated | medium | ||
| The Ovative Group and Albertsons Media Collective study 'Retail Media iROAS Demystified' (March 2026), by two Albertsons Media Collective staff, two Ovative staff and Eric T. Anderson and Brett R. Gordon of Northwestern Kellogg, with no funding statement, analysed 42 onsite display campaigns on Albertsons websites without a randomised benchmark and reports that within campaigns the highest-to-lowest iROAS gap 'averaged 6.5x (median 2.5x)' and that 83% of campaigns 'flipped from positive to negative depending on methodology' (p. 22), worded 'could flip' on p. 12 and in the executive summary, from Exhibit 12's no-filter variations. | 8 | vendor assertion | supported as stated | high | ||
| The attribution number Amazon's paper describes is a model prediction, not a per-campaign lift estimate: campaign-level 'Causal Calibration Models' fit RCT conversions from last-touch and model-driven attributions and 'help extend the RCT ad effects to other ad campaigns with similar features, even those that were not run as RCTs', with touchpoint credits normalised to shares and no interval or accuracy figure reported. | 8 | vendor assertion | supported as stated | medium | ||
| This review found no validation of Cardlytics' outcome numbers (no comparison of its attributed or test-versus-control results with an experiment, panel, point-of-sale data or replication) in the filings and pages opened by 3 November 2026; the record holds only accuracy claims such as 'our measurements are not probabilistic or based on models' and 'There is no need to estimate performance'. | 8 | synthesis | supported with qualification | medium | ||
| The Circana analyses Instacart cites (shopper analyses of August 2023 and February 2024) are buyer-overlap descriptions, not lift estimates: '75% of people who purchased the beverage and snack brands on Instacart did not buy the same brands in-store', '90%' for pasta sauce brands, and a '12% higher in-store purchase amount' per household versus in-store-only purchasers; Instacart presents them as 'additional third-party measurement and validation'. | 8 | vendor assertion | supported as stated | low | ||
| KPM's post of 19 July 2024 reports that 'a sales attribution model that broadly measured total households led to over-attributing sales to campaign performance by over $2.8MM' relative to $98K, which KPM calls 'the true amount', after restricting to 'households that were exposed to viewable ads with a custom date range'; the example compares two attribution filters on one unnamed campaign and contains no counterfactual. | 8 | vendor assertion | supported as stated | medium | ||
| This review found no independent or out-of-sample validation of Kroger Precision Marketing's attributed or incremental outcome numbers for either row in the sources screened by 3 November 2026; the nearest items are KPM's own accuracy statements ('allow us to accurately measure sales impacts'; 'Our methodology adheres to IAB standards and best practices'), the 10-K's 'best-in-class performance measurement', and a 2022 release in which a partner, Vestcom, reports an average '10-20% incremental sales lift' for its own shelf-edge media. | 8 | synthesis | supported as stated | medium | ||
| In a Walmart Connect article of 30 October 2026, QBR Media analysed one anonymised food brand that cut Search spend from an average $6,434 to $350 a day for one week in August 2026: average daily digital GMV fell 35.1% (from $35,883 to $23,293) while in-store GMV rose 0.8%, and QBR 'estimated $2.15 in incremental total GMV for every $1 of Search spend', using day-of-week-matched week-over-week counterfactuals with a prior-year seasonality index, six baseline specifications (within a $0.36 range), log-log regressions with Newey–West errors (n=30), a 5,000-draw block bootstrap and in-store GMV as a placebo outcome; the figures are 'GMV, not net revenue, and are unaudited'. | 8 | vendor assertion | supported as stated | medium | ||
| A Walmart Connect post of 8 April 2026 states that "Among Walmart Connect campaigns that qualified for a sales lift study" it saw "2.5x stronger incremental ROAS and 1.8x higher sales lift than social media industry benchmarks", footnoted to a "Walmart Connects sales lift benchmarks analysis performed by Circana, September 2024–September 2025" limited to Food, Consumables, Beverages or Alcohol campaigns, with "Industry benchmarks ... provided by Circana" and the statement "Circana is an independent third party and not affiliated with Walmart Connect". Circana performed the aggregate analysis on both sides of the comparison; the page does not say who ran the underlying per-campaign sales lift studies, whether those campaigns were social campaigns, or who paid. | 8, 9 | vendor assertion | supported with qualification | medium | ||
| Best Buy reported on 28 September 2026 that 'Circana conducted a first-of-its-kind study examining the impact of Best Buy Ads', finding that Best Buy Ads 'outperformed the national media ROAS average by 17%' and generated '35% of measured media contribution to sales at Best Buy' on 17% of spend; no design, sample, period or commissioning party is stated, and the contribution language points to a modelled (mix-type) study rather than a lift test. | 8, 9 | vendor assertion | supported as stated | medium | ||
| Criteo's Commerce Max help centre (the buyer-side product's help centre, which also hosts the 'Retailer Support' section that Criteo's SSP API page links to) documents the sales attribution Criteo applies to onsite and offsite campaigns as read on 3 November 2026: post-click lookback options of 7, 14 or 30 days; post-view options of none, 1, 7, 14 or 30 days, never longer than the post-click window; windows set per campaign for all its line items; clicks take precedence over impressions and the interaction closest to the purchase wins; one event credited per sale, other interactions recorded as assisted sales; sales deduplicated between offsite and onsite; product-match options at the same product, category, brand or seller level; no default window is stated; whether a Commerce Yield retailer configures or constrains these options is undocumented here because the retailer-side settings page returned 403. | 9 | vendor assertion | supported as stated | low | ||
| Criteo's Commerce Max help centre names a deterministic identity basis for attribution (the retailer's CRM ID for authenticated users, a hashed email, a first-party retailer visitor cookie and a Criteo cookie), states that IP addresses and probabilistic models are never used for user-level attribution, and documents in-store matching for some retail partners through email collection and the retailer's CRM ID; no match rate or coverage figure is documented. | 9 | vendor assertion | supported as stated | low | ||
| Criteo's own release of 28 March 2024 scopes its first MRC accreditation to display rendered impression and click metrics for Onsite Sponsored Products and Onsite Display Ads across desktop, mobile web and mobile in-app environments on Commerce Max and Commerce Yield, with a review of GIVT detection; the release does not name sales, ROAS or attribution as accredited. | 9 | vendor assertion | supported with qualification | high | ||
| Criteo's Supply Side Analytics (SSP) page documents that SSP revenue reports can be downloaded through the Retail Media API with fully customisable report attribution windows and time zones, and the Commerce Grid product page says media owners have direct access to its user interface and reporting tools; the pages opened document no attribution method, incrementality design, named US retailer or US availability for Commerce Grid. | 9 | vendor assertion | unresolved | medium | ||
| Epsilon's US retail media pages (the retailer page and the 'For advertisers' page, read 3 November 2026) assert exposure-to-sales measurement online and in-store on a persistent identity basis and document no attribution parameter, control design, access route, per-class measurement, accreditation, named retailer, US availability statement or product-specific privacy control. | 9 | vendor assertion | unresolved | medium | ||
| Epsilon's own release of 11 May 2021 names the MRC-accredited Correlated Outcomes service as Epsilon PeopleCloud Digital Media Solutions (DMS, formerly Conversant), defines a correlated outcome as an event connected to an underlying ad served across display and video on desktop, mobile web and mobile in-app, and contains no causality, lift or test-and-control sentence and no mention of CitrusAd or retail media; a search of epsilon.com on 3 November 2026 found no page tying the Epsilon Retail Media product to an MRC accreditation. | 9 | vendor assertion | supported as stated | medium | ||
| Epsilon's release of 18 August 2022 says Gopuff Ads is the first retail media network to roll out the off-site capabilities of CitrusAd, powered by Epsilon, the former name of the Epsilon Retail Media platform, giving one dated US retailer deployment; its date falls before 2024. | 9 | vendor assertion | supported as stated | low | ||
| Amazon Retail Ad Service is documented by Amazon as a beta available to multi-brand retailers in the US with an e-commerce site or app; Amazon's announcement of 9 January 2025 names iHerb, Oriental Trading Company and Weee! as beta retailers and says the service is launching with Tilly's. | 9 | vendor assertion | supported as stated | high | ||
| Kevel documents attribution as an opt-in part of its Retail Media feature set requiring Catalog: last-touch only (clicks checked before views), lookback windows of 1, 7, 14 or 30 days set per flight and ad with post-click and post-view windows and match types adjustable independently, match types Same Product, Same Category and Brand, Same Brand and Same Merchant (marketplaces), one ad credited per purchase, GMV = quantity × price, purchase events accepted up to 15 days after the window closes, conversions dated to the engagement day; no default window, no multi-touch model and no in-store matching are documented. | 9 | vendor assertion | supported as stated | high | ||
| Kevel documents its route families as APIs (Management, Decision and Reporting APIs), Data Shipping log files, Catalog data shared as NDJSON through an S3 bucket, a self-serve interface, programmatic demand (Kevel Relay and Kevel Exchange) and first- or second-price auctions; the Retail Media Guide mentions no holdout, incrementality or A/B testing and names no clean room, DSP or SSP, and the attribution pages describe last-touch attribution only. | 9 | vendor assertion | unresolved | high | ||
| Koddi's developer documentation (updated 5 January 2026) sets last-touch attribution as the non-configurable default with alternative models available on request, a configurable look-back window of any number of days under 30 (longer on request), attribution types Direct, Exposed, Halo Direct and Halo Exposed with configurable halo grouping, a hashed loyalty ID in the winning-ads call as the match key, offline visits and purchases ingested server-to-server with CRM or loyalty matchback (location-data partners optional), configurable deduplication rules, and reporting metrics Revenue, ROAS, Transactions and Units with Exposed and Halo variants; no default window value is stated. | 9 | vendor assertion | supported as stated | high | ||
| Koddi's product pages document Koddi Ads for on-site, off-site and in-store ads, an SSP serving commerce media formats programmatically to buyers' DSPs, direct DSP integrations, a customisable UI and APIs, an API status page and a privacy centre, and assert 'incrementality testing' without any design, sample or significance rule; no named US retailer with a date was found (Tractor Supply appears only in an undated blog title; Kohl's page returned 403 to U). | 9 | vendor assertion | unresolved | medium | ||
| Topsort's documentation lets marketplaces configure attribution per ad format with a last-click or last-impression model and a 1 to 30 day window (an example table shows Sponsored Listings last-click 7 days, Sponsored Brands last-click 14 days, banner and video last-impression 30 days, labelled an example configuration), gives direct attribution priority over halo, credits each conversion to one interaction, links clicks and purchases through the user ID, and accepts purchase, click and impression events with an idempotency id and an optional channel of onsite, offsite or instore; no default model or window, deduplication mechanics, in-store matching or ROAS definition is documented on the opened pages. | 9 | vendor assertion | supported as stated | high | ||
| Topsort's documentation landing page documents an auction-based ad server API (v2 paths), an Analytics.js SDK, 'Ad Intelligence' tools for forecasting, retrieval and attribution and a DSP for agencies and brands across multiple retailers; the landing page names no holdout, A/B or incrementality design, reporting or export documentation, changelog, versioning policy or privacy content, and the events and attribution pages opened describe event reporting and attribution only. | 9 | vendor assertion | unresolved | medium | ||
| Pentaleap documents ad serving for sponsored products, unified ranking of sponsored and organic products, a Campaign UI for Sponsored Products, Display and Video, and demand connections to programmatic partners, Amazon and Google; its only test design is a randomised A/B test of ad revenue lift from adding its ranking layer (vendor-claimed lifts of 80–140 percent), which is a platform yield test, not an advertiser outcome control; no attribution parameter, reporting, API, measurement method or privacy control is documented on the three pages opened. | 9 | vendor assertion | unresolved | medium | ||
| Pentaleap's release of 22 January 2026 says it has been selected by Macy's, Inc. to power the Macy's Media Network with its ad server and mediation technology, with no launch or go-live date, adding Macy's (selected, not yet production-evidenced) to the dated US retailer record of The Home Depot (3 October 2024) and Staples (29 July 2026). | 9 | vendor assertion | supported with qualification | high | ||
| Moloco Commerce Media's attribution methodology (updated 17 April 2026) is last-touch and click-based across Sponsored Product, Sponsored Brand and Sponsored Display, with the window configured once at platform level from options of 1, 7 or 14 days (an older solution-components page, updated 7 February 2025, says '7 or 14 days for your campaigns'), look-back direction crediting the click date, direct (same item) and indirect (same ad account) attribution, a consistent user identifier across ad requests and events, deduplication of identical events and one credit per conversion, returns and cancellations not accounted for, and shipping or VAT inclusion depending on the platform; metrics are shown on the MCM Portal and through a Report API; no view-through attribution, default window, in-store conversion or ROAS formula is documented. | 9 | vendor assertion | supported as stated | high | ||
| Vantage documents an orchestration and workflow layer for onsite, offsite and in-store media with self-serve advertiser access, reporting as 'one source of truth for performance, spend, and outcomes', automated billing and SOC 2 certification, and shows integrations only as logos (Criteo and Kevel for onsite, Google Ads, Meta, Pinterest and Reddit for offsite, Salesforce, Metarouter, BigQuery, Snowflake, Stripe); no attribution parameter, control design, API, export, clean room or named retailer is documented on its platform or home page. | 9 | vendor assertion | unresolved | medium | ||
| This review found no Mirakl-owned documentation of Mirakl Ads attribution windows, control designs, reporting access routes or a named US retailer with a date: workstream U's product-page read recorded no measurement, attribution, reporting or ROAS sentences, and one WebSearch on 3 November 2026 returned no Mirakl documentation page on attribution or reporting. | 9 | synthesis | unresolved | low | ||
| NIQ Activate's product page asserts that it 'closes the loop between exposure and purchase' through AI-powered measurement and attribution and that it increases ROAS, with no attribution parameter, control design, access route, identity basis, accreditation, US retailer or privacy control documented; the terms incrementality and reporting do not appear. | 9 | vendor assertion | unresolved | medium | ||
| Osmos documents an incrementality product for onsite sponsored display campaigns (Phase 1, early access) that holds out eligible users inside the campaign ('Eligible users assigned exposed or control'; 'a small holdout does not' see the ad). The page says the control group 'matches the exposed profile' and gives a configurable 95/5 default split and a minimum of eight weeks. Lift is reported at product view, add-to-cart and purchase, with per-campaign incremental conversions, revenue and ROAS and exports. The assignment mechanism (random or not), sample thresholds, significance rule, identity or transaction basis and who runs the analysis are not stated. Its home page also lists geo-holdouts, audience holdouts and synthetic control. | 9 | vendor assertion | supported with qualification | medium | ||
| Osmos documents route families of a retailer-branded white-label DSP, retailer seats on Google, Meta and TikTok, programmatic and pDOOH demand through Osmos Demand, a brand self-serve orchestration layer, closed-loop attribution on offsite campaigns, in-store impression verification against the play and a Developer Hub; it displays SOC 2, GDPR, CCPA, DPDP and ISO 27001 badges and a DPA link, claims retailers in 11 countries, names no US retailer in text (Wakefern Food Corp appears as an undated logo) and makes no US availability statement. | 9 | vendor assertion | unresolved | medium | ||
| Rippl documents itself as a consortium of regional grocers and convenience stores that delivers off-site retail media to national brands against deterministic purchase data ('over 140 million deterministic purchasers', 'derived from point-of-sale transactions') made available in syndicated audience marketplaces and buyers' DSPs (LiveRamp, OpenX and PubMatic are quoted or named), with one undated Schnucks case claiming 16 percent incremental category sales and an undated Danone case citing ROAS above 3.5; its 'incremental' claim on the advertisers page concerns audiences, not measured outcomes; no attribution parameter, holdout or test-and-control design, reporting route, API, clean room or dated retailer deployment is documented on the three pages opened. | 9 | vendor assertion | unresolved | medium | ||
| STRATACACHE documents in-store retail media network technology (screens, content, 20-plus years of deployments) that uses sales, store and loyalty data where permitted, with a dated US deployment at Albertsons Media Collective (news item of 28 January 2026 promising 'proof of play' and 'direct sales attribution' under a 'Closed-Loop Measurement' heading) and a deployment for Iceland and The Food Warehouse (22 January 2026; UK grocers by common knowledge, the listing states no country); its solutions page documents no attribution method, opportunity-to-see method, API, programmatic connection or privacy control beyond a form consent line. | 9 | vendor assertion | supported as stated | medium | ||
| Zitcha's platform page documents Zitcha Core (ad serving, channels, inventory, brand self-serve, billing, an 'API-first' ad server), Margin Manager as a Snowflake-native recommendation engine 'live with a handful of retailers' (its home page says 'in beta'), offsite connectors to Meta, Google, TikTok, Snapchat and Pinterest, access to 'the world's largest independent DSP', Pentaleap's unified ranking, and measurement claims of SKU-level attribution, conversions split into new-to-brand and repeat, loyalty-data matching of exposure to in-store purchases and 'the sales your media actually caused', with no attribution window, touch model, holdout design, method or named retailer documented; regions are listed as US, APAC and EMEA. | 9 | vendor assertion | unresolved | medium | ||
| ADvendio documents a Salesforce-native order-management, inventory, billing and self-serve portal platform with real-time ROAS visibility and performance data for suppliers, and documents no attribution method, control design, API reporting or dated US retailer; its named retailers (Albertsons, Ahold Delhaize, Ulta Beauty, Shoprite, Loblaw Media) are undated. | 9 | vendor assertion | unresolved | medium | ||
| Streaem's developer documentation (pages updated 22 May 2025) documents retailer-set post-click and post-view lookback windows (no day values stated for post-click; a one-day post-view window is advised), a conversion-type attribute on add-to-cart and purchase events, and a requirement that the retailer identify whether a user clicked or only saw an ad; no attribution model, product-match rule, identity method, in-store handling, deduplication rule or reporting output is documented, and the product's owner and US status remain unverified because vml.com/streaem refuses the fetch tool. | 9 | vendor assertion | unresolved | low | ||
| Kevel's GMV and ROAS page (updated 1 August 2026) defines ROAS as gross merchandise value divided by advertising spend, the spend being Kevel's 'Revenue' metric for the flight, expressed as a percentage; GMV is a value the customer appends to event-tracking URLs and can be tracked on impressions, clicks, conversions or custom events at the customer's choice; GMV appears in data-shipping logs, the Real-Time Reporting API and reports, and ROAS in real-time and scheduled reporting and the Kevel UI; returns, cancellations, taxes, shipping, currency and deduplication are not addressed. | 9 | vendor assertion | supported as stated | high | ||
| STRATACACHE's Data & Analytics page documents in-store measurement as ad-impression and video-play tracking at each digital screen, sensor-based behavioural data, real-time people-counting sensors claimed to be 99 percent accurate (no method), dwell time and dashboards, offered for in-store retail media networks; it documents no opportunity-to-see method, audience measurement, sales attribution, POS or loyalty data source, test-and-control design, API or anonymisation method beyond 'privacy-friendly'. | 9 | vendor assertion | unresolved | medium | ||
| CommerceIQ's public pages opened by this stream (the Retail Media Management page and its 15 August 2025 Prime Day blog) assert 'incrementality' and 'iROAS' optimisation with no holdout, test-control, geo or clean-room design described; the blog says the AllyAI engine 'factors in over 50 variables' and recalibrates intra-day. | 9 | vendor assertion | supported as stated | medium | ||
| Commerce Max publishes dated monthly release notes for 2026 (January to October), including 'Changes in Attribution Latency for Commerce Max Offsite Ads' (February 2026), 'Change History Tool' (March 2026), 'Cost Transparency Report' (July 2026), 'AI Shopping Assistant Page Type' and 'In-Platform Agent' (June 2026) and 'More Reliable Reporting' (October 2026). | 9 | vendor assertion | supported with qualification | high | ||
| Criteo's only documented incrementality framework opened by this stream (blog of 22 July 2025: 'user-split or geo-split methodology', 'audience-level holdouts or regional comparisons', a 'Real-time Incrementality' iROAS metric 'currently in pilot with select clients', with Haus named as an external partner) is described for the Commerce Growth offering; the Commerce Max page and the help-centre categories screened carry no incrementality or lift article. | 9 | vendor assertion | supported as stated | medium | ||
| Flywheel's retail media software page asserts causal measurement as an on/off comparison ('Measure causality by comparing sales when your ads are displayed versus when they're not') and 'custom attribution in a secure environment', with no assignment, holdout, window or method stated. | 9 | vendor assertion | supported as stated | medium | ||
| For Walmart in-store attribution, Intentwise documents that 'Walmart maps these interactions to subsequent online and in-store purchases', that 'Walmart is experimenting with different attribution windows—3 days, 14 days, and 30 days', that 'At Intentwise, we utilize a 14-day attribution window for our calculations', and that in-store and omnichannel metrics are automatic for 1P suppliers; it defines In-Store Attributed, Advertised and Other (same-brand) Sales, Omnichannel ROAS and a revised ROAS of 'Total Attributed Sales / Spend'. | 9 | vendor assertion | supported as stated | high | ||
| Intentwise's help centre documents four network feeds for Ad Optimizer: Walmart Connect through 'the Walmart Connect Ads APIs' (RTL-G1-S33), Instacart through an API user ('You need to add Intentwise as an API user', RTL-G1-S42), Amazon through the push-only Amazon Marketing Stream hourly feed and an Amazon DSP connection article (RTL-G1-S47, S26), and Criteo through a consent login covering twenty listed US retailers (RTL-G1-S41, S32); it also offers an MCP 'AI Gateway', Amazon Marketing Cloud ('Intentwise Explore') and BI exports; no incrementality or experiment article appears among the roughly 90 Ad Optimizer titles screened. | 9 | synthesis | supported as stated | medium | ||
| Pacvue's own documentation (pacvue.com, its iROAS page, its newsroom and help.pacvue.com) could not be read by this stream: pacvue.com refuses the fetch tool, the Internet Archive route is blocked for the tool, and help.pacvue.com failed on an expired TLS certificate; every Pacvue capability cell therefore rests on the Italian-locale about page read by the session's browser (RTL-U-S305) or stays n/e pending the browser items listed. | 9 | inference | supported as stated | high | ||
| Perpetua documents the provenance and limits of its Walmart outcome feed (help article of 13 May 2026): data is pulled from 'Walmart's API', 'the most current data in Perpetua will typically lag Walmart by 1–2 days', the platform 'won't show data from Walmart for the last two days', attributed sales 'are not final at the time of the click' because Walmart applies attribution windows ('commonly 14-day and 30-day') and 'Conversions may be added retroactively', and the Walmart Connect Ad Center 'updates faster than API-connected platforms'. | 9 | vendor assertion | supported as stated | high | ||
| Skai's retail media pages document network data integrations by name and kind (product-level data from 'Amazon, Walmart, and Criteo'; Amazon's 'hourly Stream data'; Instacart campaign, ad group, keyword and ad data), claim management 'across 100+ retail media networks' (a press-release line on the same page says 'more than 200'), and offer 'Custom Metrics' that 'Create metrics not provided by publishers' and a normalisation that will 'bridge the gap between differing Instacart, Walmart & Amazon attribution & auction models' without stating the method. | 9 | synthesis | supported as stated | medium | ||
| Xnurta documents three network integrations: Criteo through an API (RTL-U-S209: 'connects to Criteo via API'; 'Offsite campaigns are not currently supported via the Criteo API.'; attribution windows editable in bulk with no values listed), Walmart Connect through an API (RTL-G1-F33), and Amazon (Sponsored Products, Sponsored Brands, Sponsored Display, a managed DSP service and AMC) connected 'in just a few clicks' with no API statement (RTL-G1-S46); its help centre is a JavaScript shell to the fetch tool. | 9 | synthesis | supported as stated | medium | ||
| Commerce Max defines its outcome metrics on its 'Demand-Side Metrics' page: Attributed Sales is 'A sale that was a direct result of your marketing campaign (in accordance with your attribution rules)', 'A sale can only be attributed to one line item across all campaigns', ROAS is attributed sales over spend and 'can vary based on your attribution settings', and new-to-brand covers shoppers making 'their first purchase of the advertised brand at a specific retailer within the previous 180 days', a classification 'separate from the campaign click- or view-attribution window'. | 9 | vendor assertion | supported as stated | high | ||
| Perpetua connects Instacart Ads through API access granted as an agency user, after which 'Perpetua can create OAuth applications' and 'create, update, and view performance metrics via the API and through Ads Manager'; the article (dated 1 May 2024, footer 'last updated December 2023') states 'Instacart ads are currently only supported in the US and Canada' and says nothing about attribution or sync frequency. | 9 | vendor assertion | supported as stated | medium | ||
| Stackline's attribution page presents a Multi-retailer module that 'links verified ad exposures to real purchase receipts' using 'deterministic, first-party data' 'across every major retailer—both online and in-store', and an Incrementality analytics module (linked to an 'Ad Manager' button) that 'deciphers the net new customer impact of every impression' from search, click, media-intelligence and 'five years of verified shopper purchase histories'; no window, holdout, control group or method is stated, and case figures ('48%' of Amazon Ad attributed sales off Amazon; OLLY's '47.7% lift in ROAS') carry no design. | 9 | vendor assertion | supported as stated | medium | ||
| Attain's independence and ownership record from its own pages: 'No media inventory. No incentives to inflate results. Just objective measurement aligned with your business outcomes.' (measurement page); 'Attain began as Klover.' (2019 timeline), rebranded as Attain in 2022, launched the Frisbee rewards app in 2023, 'With offices in Chicago, New York, San Francisco and Denver,', 'Backed by' logos Core Innovation Capital, Mercato Partners, lightbank and Starting Line (no amounts), leadership Brian Mandelbaum (CEO & Co-founder), Dominic Bennett (CTO & Co-founder) and Ben Kartzman (President & COO); no legal entity name, headquarters or ownership percentage appears on any of the six Attain pages opened, and the terms page URL guessed returned 404. | 9 | vendor assertion | supported with qualification | medium | ||
| Attain's privacy record splits by surface: the B2B privacy policy ('Last Updated / Effective Date: April 16, 2026') names the CPRA ('If you are a California resident, the CPRA allows you to request that Attain no longer “sells” or “shares” your personal information.'), says 'Attain may “sell” your Personal Information as defined by each applicable state law', states 'Attain is SOC 2 Type II certified.' and 'We may create aggregated, de-identified or anonymized data from the Personal Information we collect', while the consumer privacy index says 'At Attain, we operate three consumer-facing applications.' (Klover, Frisbee, Merryfield) and points to each app's own policy on a separate domain, none opened; so the panel's consent basis for ad measurement is not documented in this stream's record. | 9 | vendor assertion | supported as stated | medium | ||
| Incremental states on its home page that it 'operates completely independently from retailers, media sellers, and bidding platforms.' while its about page says 'We're a trusted platform for advertisers, media sellers, and agencies.' and both carry a 'Walmart Connect partner directory' badge; its how-it-works page shows partner logos for TikTok, Google, Walmart, Amazon, Shopify, Target, Ulta and Meta; scale is stated as '$53B' 'in sales tracked annually', '550,000+' 'unique products', '500+' 'of the world's largest brands' and 'Every major agency holding company and 7 of the 10 largest CPGs.'; no legal entity, headquarters, ownership, investor or US-availability statement appears on any of the five Incremental pages opened. | 9 | vendor assertion | supported as stated | medium | ||
| Descriptive profile, Haus: three designs with different assignment states on Haus's own pages: GeoLift, where 'Haus uses random stratified sampling to assign test and control groups across a country or region' (random assignment at geography level), Fixed Geo Tests, 'ideal for campaigns where you select the test regions' with 'advanced modeling techniques to create a robust Haus Holdout for comparison' (client-selected test regions, modelled counterfactual), and Time Tests, which evaluate a period 'against a high-precision daily forecast' (pre/post versus forecast); channels named include 'select retail media platforms' and 'DTC, Amazon/marketplaces, and offline/retail'; a US DMA map appears in image alt text and no confidence-interval, minimum-spend or duration text appears. | 9 | vendor assertion | supported as stated | medium | ||
| Descriptive profile, LiftLab: an Agile MMM that 'models retail media as an explicit channel alongside national media and DTC' plus an Incrementality Testing Suite that 'provides controlled reads' 'Where retail media can be tested cleanly' (blog, 10 September 2026); the testing engine page states 'both Stratified Random Sampling and Synthetic Controls are supported' for geo selection, an audit trail 'showing how test and control markets were assigned', results 'as ranges rather than single lift numbers' with 'Confidence intervals are explicit, not hidden.', a contamination check ('Spillover contamination is detected and corrected before results are calculated'), and feedback 'directly back into your AMM as calibration inputs'; the company is 'LiftLab Analytics, Inc.' at '1111 Broadway, 5th floor, Oakland, CA 94607', 'SOC 2 Compliant' with 'an ISO 27001:2013 Certificate'; no retailer sales feed, panel, Amazon, Walmart or Instacart is named on the two product pages opened. | 9 | vendor assertion | supported with qualification | medium | ||
| On 23 June 2026 Albertsons Media Collective announced an integration with Criteo through which 'eligible sponsored products can appear within AI-powered conversational search product carousels' in Albertsons' app, describing it as building on its 'ongoing partnership with Criteo'; Criteo called Albertsons Cos. its 'first retailer to bring Sponsored Products into an AI-powered shopping assistant'. | 9 | vendor assertion | supported with qualification | high | ||
| Albertsons Media Collective announced on 16 June 2025 an in-store digital display network built with STRATACACHE, whose pilot 'will launch this summer in select Albertsons Cos. stores in two key regions', promising 'proof of play, direct sales attribution and sales lift'; STRATACACHE's item of 28 January 2026 says the network debuted and would reach 'approximately 800 stores across the Albertsons footprint by 2026', while Albertsons' 6 January 2026 release plans 'approximately 800 additional stores' in 2026 across 10 divisions and cites 'more than 50 unique advertising partners'. | 9 | vendor assertion | supported as stated | medium | ||
| Circana's Media Lift page (undated, footer '© 2026 Circana'; read 3 November 2026) describes Household Lift as 'the most precise, household-level A/B testing in the industry, connecting the people exposed to your media with the products they buy in-store', for addressable media 'at national or brand campaign level', with the bullet 'Enact controlled household tests'; the page states no allocation rule, control construction, minimum sample, significance rule, balance check or uncertainty, so assignment is coded unknown and the counterfactual as a control group asserted without a stated construction. | 9 | vendor assertion | supported as stated | high | ||
| The same page describes Market Lift as delivering 'sales uplift for your campaign test executions, projected sales impact, and ROAS across your total campaign geography' with 'a thorough matching process and proven methodology', for 'Non-addressable or in-market A/B testing for shopper campaigns and regional retailer-specific campaigns' on point-of-sale data ('Conduct controlled market or store tests'); allocation, matching variables, the projection method and uncertainty are not stated, so assignment is unknown and the counterfactual is coded matched market or store. | 9 | vendor assertion | supported as stated | high | ||
| Circana's Household Lift case study (undated) says Chomps ran 'digital-to-shelf offers across the Ibotta Performance Network' 'using Circana Household Lift for independent measurement of incremental impact' and reports '4.5X' 'Sales lift Vs. Category Benchmarks', '15%' 'Increased Spend Among Exposed HH' and '9X' 'HH Penetration vs. Category Benchmarks', with no period, sample, control construction, geography or payer. | 9 | vendor assertion | supported as stated | medium | ||
| In the screened record Commerce Yield documents attribution only: no control design (holdout, ghost ads, PSA, geo or matched control) is documented for retailers' advertiser campaigns; none of the 209 English titles on the Commerce Yield help centre's home page concerns incrementality, lift, holdout, experiment or validation; and the two Criteo incrementality texts opened describe user-split or geo-split tests for Commerce Growth (22 July 2025) and ghost bidding 'optimized for the open web' (18 September 2024), neither mentioning retail media in its body. | 9 | synthesis | supported with qualification | medium | ||
| Criteo's own dated statements name seven US retailers for its retailer-side retail media business in 2024-2025: Albertsons Media Collective (9 January 2024; Commerce Yield 'will allow for Albertsons to continue to monetize its online assets'), Dollar General (31 May 2024; 'supports Dollar General's onsite sponsored ad offerings'), Belk (8 October 2024; 'Belk is harnessing Criteo's retailer monetization platform, Commerce Yield') and, in the Q2 2025 results release of 30 July 2025, BJ's Wholesale Club, Weis Markets, Winn-Dixie and Harveys Supermarkets as 'new retailers' added (no product named); Criteo's FY2025 10-K counts 'approximately 235 retailers' worldwide without names. | 9 | vendor assertion | supported with qualification | medium | ||
| Commerce Yield's methodology note (published 13 July 2026, updated 11 August 2026) documents a correction route for delivery data: when a new type of invalid traffic is found Criteo runs 'a backward-looking analysis' and notifies affected advertisers if previously reported data was materially affected, and where issues affect spend or data 'Impacted advertisers are notified and data can be reprocessed to correct billing and reporting metrics' (a sentence in a section on delivery-server logging that speaks of issues affecting 'advertiser spend or data' without excluding attributed metrics; whether it covers attributed sales is not stated either way); with Criteo's API versioning policy and the retroactive application of attribution-setting changes, that is the governance the screened record shows, while no dispute procedure and no methodology version history were found. | 9 | vendor assertion | supported with qualification | medium | ||
| Carrot Ads' developer documentation defines five retailer-facing metrics (Revenue, Clicks, Impressions, CTR, Average CPC net value), requires externally fulfilled orders to be ingested 'within three days of order creation' to count, and documents no attribution window, touch model, ROAS or matching rule; its changelog (18 March 2024 to 10 October 2026, 'All changes are non-breaking changes.') records on 10 October 2025 'Added clarification on how ad impressions are calculated for both sponsored products and display banners.' and on 1 May 2025 a correction of required fields. | 9 | vendor assertion | supported as stated | high | ||
| By 11 November 2025 Instacart's own releases and blog named six US networks as using Carrot Ads: Schnucks (live on Schnucks.com and its app since 2021 per Instacart's blog of 19 August 2025, which reports its retail media revenue up '7x'), Hy-Vee, Sprouts, Thrive Market and Uber Eats ('use this solution', 6 November 2025; Hy-Vee's March 2025 release had said 'later this year') and Associated Food Stores ('is also leveraging Carrot Ads', 11 November 2025), whose Davis Food & Drug said its 'monthly retail media revenue grew by 2-3x'. Instacart counted 'more than 220 retailer banners' (March and August 2025), 'more than 240' partners (September and November 2025) and 'more than 310 partners' (September 2026). Gopuff (22 September 2026: 'will be utilizing Carrot Ads') and Vroom Delivery (11 September 2025: 'will have access') were announced as future. | 9 | vendor assertion | supported with qualification | medium | ||
| Swiftly's retail media page (timestamped 2 November 2026) asserts IAB-standard closed-loop attribution tied to in-store sales, closed-loop reporting of 'real sales lift' and an analytics dashboard showing impressions, engagement and in-store sales lift, and carries a PepsiCo quote of 'more than $100 iROAS' (a dollar-denominated iROAS, as written) and over 39 percent incremental sales lift with no method; named retailers are Homeland, Better Health Market and The Save Mart Companies without deployment dates. The page carries a publication timestamp of 2 November 2026, not deployment dates. Two of the three retailer items describe enhancing 'digital presence' (Homeland) and expanding 'digital circular reach' with Audience Optimizer (Better Health Market), not documented retail media campaigns. | 9, 12 | vendor assertion | supported with qualification | medium | ||
| Amazon Ads' privacy notice page is a country index whose United States entry links to the amazon.com customer privacy notice; that US notice and Amazon's Interest-Based Ads notice both returned HTTP 503 to the fetch tool on 3 November 2026, so the product-specific privacy posture of the four Amazon rows is coded 'source inaccessible' in this stream. | 10 | synthesis | unresolved | high | ||
| The 'Privacy Notice - Sam's Member Access Platform' (last updated 23 February 2026) governs the advertising platform's own site visitors: it states 'Sam's Member Access Platform does not sell personal information', offers California requests and the Global Privacy Control, shares 'aggregated information that is not individually identifiable', and says nothing about members' purchase data, sensitive categories or clean rooms, so the product-specific consent basis for using member purchase data in advertising is not documented on it. | 10 | vendor assertion | unresolved | high | ||
| Target's company-wide privacy policy (updated 15 August 2026) states that Target 'uses your personal information for targeted advertising and "sells"' it (Colorado section), lists categories 'collected through Target Circle for targeted advertising', treats Global Privacy Control signals as opt-outs of sale, obtains consent for sensitive personal information 'If required', and never mentions Roundel, a clean room or hashed matching, so Roundel's product-specific privacy posture rests on a generic policy. | 10 | vendor assertion | unresolved | high | ||
| Walmart Connect's own privacy policy (last updated 17 March 2026) covers business users of the Walmart Connect site ('Walmart Connect is a business-to-business service') and says it 'does not address other information collected and processed by Walmart on other Walmart sites'; the shopper-facing basis for using purchase data in Walmart Connect advertising is therefore Walmart's company-wide customer notice (RTL-E1-F31), and no product-specific consent, sensitive-category or partner-matching notice for Walmart Connect was found on the pages screened. | 10 | vendor assertion | unresolved | high | ||
| The FTC's monopolisation complaint against Amazon, announced 26 September 2023 and filed in the U.S. District Court for the Western District of Washington as Civil Action No. 2:23-cv-01495 with 17 states after a 3-0 Commission vote, alleges that Amazon degraded the shopping experience by 'replacing relevant, organic search results with paid advertisements' and 'deliberately increasing junk ads that worsen search quality', and by 'Biasing Amazon's search results to preference Amazon's own products'; these are allegations ('The complaint alleges'), not findings. | 10 | direct record | supported as stated | high | ||
| On 31 August 2026 the FTC and 22 states filed a second case, FTC et al. v. Amazon.com, Inc., Case 2:26-cv-03097 (W.D. Wash.), alleging under Section 5(a) of the FTC Act and state laws that 'Since 2019, Amazon.com, Inc. ("Amazon") has secretly and systematically overcharged its approximately 1.2 million advertising customers by manipulating the "auctions" that it uses to set the price of ads on its platform' for 'Sponsored Products, Sponsored Brands and Sponsored Display', via undisclosed 'soft reserve' prices described internally as a 'proxy 2nd price that we calculate'; the complaint alleges the scheme 'has likely illegally extracted over 20 billion dollars' and that the 2024 surcharge increase 'also resulted in Amazon charging Sponsored Products advertisers their own winning bid close to 80% of the time' (¶76, read with ¶75); the Commission vote was 2-0 and all of this is allegation. | 10 | direct record | supported with qualification | high | ||
| On 23 July 2024 the FTC issued Section 6(b) orders to eight companies (Mastercard, Revionics, Bloomreach, JPMorgan Chase, Task Software, PROS, Accenture and McKinsey & Co.) seeking information on surveillance pricing, by a 5-0 vote, under a statute the release describes as authorising studies 'that do not have a specific law enforcement purpose'. | 10 | direct record | supported as stated | high | ||
| The FTC's staff perspective of 17 January 2025 ('FTC Surveillance Pricing 6(b) Study: Research Summaries — A Staff Perspective', January 2025) presents staff's 'initial findings' and says it is 'not an exhaustive declaration of the findings'; it was released on a 3-2 vote (Commissioners Ferguson and Holyoak dissenting), rests on documents from six respondents (Mastercard, Revionics, Bloomreach, PROS, Accenture and McKinsey & Co.), 'only includes hypothetical examples of surveillance pricing' (release), and 'does not reflect or suggest any assessment as to whether anyone has engaged in illegal conduct' (PDF); the study was 'still ongoing' and a request for information took comments until 17 April 2025. | 10 | direct record | supported with qualification | high | ||
| The FTC's 2024 6(b) order (Matter No. P246202) defined the solutions studied by reference to 'purchase history' among other variables. The staff document gives an illustrative example of a retailer targeting shoppers who joined its loyalty program in-store with web-only deals (fn 33), and separately reports, citing redacted respondent documents, that targeted customers 'can include those who are coupon-inclined, who have perceived customer "loyalty," or who demonstrate a "willingness to pay"' (fn 39); neither names a company or a retail media network. | 10 | direct record | supported with qualification | medium | ||
| On 1 February 2023 the FTC announced its first Health Breach Notification Rule action, against GoodRx: a US$1.5 million civil penalty and a proposed order (referred to the Department of Justice by a 4-0 vote, subject to court approval) over disclosure of users' prescription and health data to advertising platforms the release names as Facebook, Google and Criteo, and to Branch and Twilio. | 10 | direct record | supported as stated | high | ||
| Colorado's Privacy Act took effect 1 July 2023; the Attorney General's page states that controllers must obtain affirmative consent before collecting and processing sensitive data, must accept universal opt-out mechanisms from 1 July 2024, and that using bona fide loyalty-program data for targeted advertising or sale is a secondary use requiring consent. | 10 | direct record | unresolved | medium | ||
| Texas's Data Privacy and Security Act (HB 4, effective 1 July 2024) bars processing sensitive data, which includes personal data revealing a 'mental or physical health diagnosis', without the consumer's consent, requires a controller that sells sensitive data to post 'NOTICE: We may sell your sensitive personal data.', and defines targeted advertising to exclude ads based on activity within the controller's own sites or apps and processing 'solely for measuring or reporting advertising performance, reach, or frequency'. | 10 | direct record | supported as stated | high | ||
| Minnesota's Consumer Data Privacy Act (Laws 2024, ch. 121, art. 5, codified at Minn. Stat. §§ 325M.10 to 325M.21; effective 31 July 2025) bars processing sensitive data without consent (§ 325M.16, subd. 2(d)), bars targeted advertising or sale of personal data of consumers known to be 13 to 16 without consent (subd. 2(f)), and bars small businesses from selling sensitive data without prior consent (§ 325M.17(a)). | 10 | direct record | supported with qualification | high | ||
| Under the HIPAA Privacy Rule (45 CFR 164.508, 2024 edition) a covered entity, such as a pharmacy, must obtain an individual's authorization for any use or disclosure of protected health information for marketing (except face-to-face communications and promotional gifts of nominal value), the authorization must state when the marketing involves financial remuneration from a third party, and any disclosure that is a sale of protected health information requires an authorization stating that remuneration will result. | 10 | direct record | supported as stated | high | ||
| California's current data-broker registration file (registry.csv, 625 records, page last updated 30 October 2026) lists Catalina Marketing Corporation (ID 1067), 'Bridg, a division of Cardlytics, Inc.' (ID 1119), Circana, LLC (ID 1270), Criteo Corp. (ID 1416), Affinity Solutions Inc. (ID 1453), PaeDae, Inc. (Infillion; ID 1431), Fetch Rewards, LLC (ID 1373), Koddi, Inc (ID 1481), LiveRamp Holdings, Inc. (ID 1089), Epsilon Data Management, LLC (ID 1026), The Nielsen Company, LLC and eXelate, Inc. (Nielsen Marketing Cloud) and VIZIO Services, LLC (ID 1758); Cardlytics, Inc. itself, Attain, Numerator, Kroger, 84.51°, Albertsons, Walmart, Amazon, Instacart, Target, Roundel, JPMorgan Chase, Ibotta and NIQ do not appear by name or DBA in any of the four files. | 10 | direct record | supported as stated | medium | ||
| Amazon (Privacy Notice last updated 30 June 2026; Interest-Based Ads notice) states that it uses personal information to display interest-based ads, does not use information that personally identifies a customer for them, does not share 'specific shopping actions like purchases, product views, or searches with ad companies' but may share an advertising identifier derived by 'irreversible cryptography' from an email address, and the Amazon Marketing Cloud page states that AMC 'only accepts pseudonymized information', that advertisers 'can only access aggregated, anonymous outputs' and that it has 'built-in aggregation thresholds'; none of these pages is specific to Sponsored Products, Sponsored Brands, display ads or Amazon DSP, and none names an independent privacy certification (Amazon's health-data and state disclosures are RTL-J-F26). | 10 | vendor assertion | unresolved | medium | ||
| Walmart's Privacy Notice (updated 20 August 2026) lists 'Performing and providing advertising and marketing services, including targeted advertising' among its uses, says Walmart processes sensitive personal information with consent 'Where required by law', honours the Global Privacy Control to opt out of sale/sharing, states that disclosures to business partners 'may be deemed selling under certain state privacy laws', and does not mention Walmart Connect, retail media, clean rooms or a Washington consumer-health-data section; pharmacies are covered by separate notices. | 10 | vendor assertion | supported with qualification | medium | ||
| Target's Privacy Policy (updated 15 August 2026) states that Target participates in interest-based advertising under the DAA principles, treats Global Privacy Control signals as opt-outs of sale and of targeted / cross-context behavioural advertising and lists categories of personal information collected through Target Circle for targeted advertising; its California section says 'We do not collect or process sensitive personal information (as defined by California law) for the purpose of inferring characteristics about individuals.'; its Health Privacy Policy (30 June 2025) says health data, including 'purchase of health-related items', is used for marketing only with consent and 'will not be collected or shared' for cross-site advertising 'unless your consent is obtained'; neither document names Roundel in its body, while Roundel's own measurement page says 'We use privacy-conscious methodologies and secure data collaboration environments'. | 10 | vendor assertion | supported with qualification | medium | ||
| Instacart's Privacy Policy (updated 15 January 2026) discloses personal information to 'marketing and advertising partners' so Instacart can 'provide advertising services to other companies (e.g., to assist in serving and measuring our or their advertisements)', receives loyalty-status data from retailers, lists 'Advertising and marketing partners' among parties to whom categories were sold or shared for targeted advertising, treats prescription-delivery data as HIPAA-protected, limits use of sensitive personal information to necessary purposes, and honours Global Privacy Control; it does not mention clean rooms, hashing or 'Instacart Ads'. | 10 | vendor assertion | supported as stated | medium | ||
| Kroger Precision Marketing's own FAQ states that it 'uses hashed identifiers, adheres to platform rules and regional regulations, and employs privacy-preserving environments (clean rooms) where appropriate' and that it is 'certified by the Interactive Advertising Bureau (IAB) and Merchant Risk Council (MRC)'; Kroger's privacy policy and privacy centre could not be opened by the fetch tool. | 10 | vendor assertion | unresolved | low | ||
| Criteo's privacy policy (effective 17 July 2026) states that Criteo 'does not process sensitive personal information', prohibits advertisers and publishers from sharing special categories of data and advertising that targets children under the consent age, discloses that in the past 12 months it has 'disclosed or sold' categories 'for the commercial purpose of providing cross-context behavioral advertising' with a 'Do Not Sell or Share My Personal Information' link, and declares that Criteo Corp. is a registered data broker under Texas Business and Commerce Code chapter 509 and in California; the policy body says nothing about retail media or purchase data from retailers and names no independent certification. | 10 | vendor assertion | supported with qualification | medium | ||
| Circana's global privacy policy (English effective-date line '23 October 2025') covers its 'mobile application panelists', lists among third-party sources 'the manufacturers or sellers of the products you purchase or use' and 'companies that combine purchase information' and data on 'your exposure to advertising', says that for its media products combined data is keyed by 'a unique ID' that 'does not directly identify consumers', lists 'health and fitness data' and email contents read 'to identify purchases that may be eligible for loyalty and rewards' as sensitive personal information collected, lists 'Data brokers' among recipient categories, and states that it may sell or share with partners that use its media products categories 'including certain information considered sensitive personal information under the CCPA'; Circana, LLC is a registered California data broker. Its registry entry states that NC Ventures, LLC (NCSolutions) merged into Circana, LLC on 30 September 2025. | 10 | vendor assertion | supported as stated | medium | ||
| On 19 August 2026 the FTC (vote 2-0) published for comment a proposed enforcement policy statement on personalized pricing (comments due 18 September 2026, extended on 3 September to 25 September 2026) stating that, where consumers reasonably expect prices not to vary with their personal data, businesses 'should clearly and conspicuously disclose not just that the price is personalized, but also the basis for that personalization and the types of data on which the personalization is based', that failing to do so is likely unfair or deceptive under Section 5, and that a shopper misled into believing a personalized price 'is a discount based on their purchase history with that retailer' may be deceived; this review found no final version by 3 November 2026. | 10 | direct record | supported with qualification | high | ||
| Amazon's Consumer Health Data Privacy Disclosure and Additional State-Specific Privacy Disclosures (both last updated 30 June 2026) state that consumer health data under Washington's and Nevada's laws is used to provide requested services (no advertising use is listed) and is not sold, that Amazon uses sensitive personal information only as the California Privacy Rights Act permits, that under-18s are opted out of cross-context behavioural ads by default, that GPC opts a browser out, that Amazon sold no personal information in the prior twelve months, and that the controller entities include 'Amazon Advertising LLC'; Amazon reports about 349.5 million cross-context behavioural advertising opt-out requests in 2025, mostly from browsers with GPC enabled. | 10 | vendor assertion | unresolved | medium | ||
| Albertsons Media Collective's own privacy policy (last updated 12 January 2026; Albertsons Companies, Inc.) covers its website and offline collection and 'does not address other information collected and processed by Albertsons.com on other Albertsons sites'; it says Albertsons does not 'sell' personal data under the CCPA/CPRA but may 'share' it for cross-context behavioural advertising subject to opt-out, discloses customer records to marketing and advertising agencies, offers an opt-out of sale, sharing, targeted advertising and profiling, and does not mention sensitive personal information, consumer health data, pharmacy, loyalty programmes, Global Privacy Control or any certification. | 10 | vendor assertion | supported as stated | medium | ||
| On 28 October 2026 the FTC ordered nine companies, including Instacart, DoorDash, Uber, Lyft, Target and Walmart, to provide information under Section 6(b) on the use of personal data to set prices and compensation; the release says the agency "learned nothing about actual pricing decisions from the 2024 study". This is a study inquiry, not an allegation or finding. | 10 | direct record | supported as stated | high | ||
| The consent basis for using Albertsons shopper purchase and loyalty data in its media is not documented in the record opened: the Media Collective's site policy (last updated 12 January 2026) excludes data 'collected and processed by Albertsons.com on other Albertsons sites', the shopper-facing albertsons.com policy refused automated access, and the in-store network sells 'deli and pharmacy placements' with no stated handling of health or pharmacy data. | 10 | synthesis | supported as stated | medium | ||
| Cardlytics states that banks supply de-identified data and that it receives no PII and has no 'insight into transactions in sensitive categories such as healthcare'; US consumers activate offers in their bank's app and opt out through the bank, its UK notice rests processing on 'the consent that has been given by the customers to our banking partners', and the 10-K names the Gramm-Leach-Bliley Act, the CCPA and California's Delete Act among the regimes it is subject to. | 10 | vendor assertion | supported with qualification | medium | ||
| Commerce Yield's help centre documents the identity basis of attribution as three signals the retailer passes through the API (Customer ID and Retailer Visitor ID required, Hashed Email recommended) plus a Criteo ID set automatically by Criteo's tag; all are evaluated in parallel and the best match is kept in the order Customer ID, Hashed Email, Retailer Visitor ID, Criteo ID; Commerce Yield's About Attribution article states that user matching is purely deterministic, with no IP-address or probabilistic matching; missing signals undercount attributed sales; no match rate or coverage report is published. | 10 | vendor assertion | supported with qualification | high | ||
| Criteo's FY2025 10-K records three legal matters touching personal or health data (none tied in the record to Commerce Yield, Commerce Max or Commerce Grid): the CNIL's June 2023 GDPR decision reducing the sanction to €40 million, paid in Q3 2023 and under appeal, which 'relates to past matters and does not include any obligation for Criteo to change its current practices'; Doe v. GoodRx Holdings (N.D. Cal.), where Criteo agreed in Q3 2025 to settle for US$7.0 million (GoodRx to indemnify US$5.5 million) and the court denied approval in January 2026; and a putative class action filed 9 July 2025 against Criteo, CVS and others in the Central District of California over sensitive health and personal information, dismissed as to Criteo on 27 October 2025, refiled on 7 November 2025, with a stipulation to dismiss filed on 30 January 2026 and mediation under way. | 10 | direct record | supported as stated | high | ||
| Instacart's privacy record for its advertising is company-level only: the Privacy Policy (updated 15 January 2026) discloses personal information to 'marketing and advertising partners' to 'assist in serving and measuring our or their advertisements', treats prescription-delivery data under HIPAA and honours Global Privacy Control, but names no 'Instacart Ads' control; the ads terms carry no data-protection clause; Maplebear is not on California's data-broker registry by name; and its FTC matters are a December 2025 complaint and January 2026 stipulated order over Instacart+ consent and fees and a 28 October 2026 Section 6(b) pricing inquiry, neither about advertising data. | 10 | synthesis | supported with qualification | medium | ||
| KPM's only product-level privacy statement found is its FAQ answer that it 'uses hashed identifiers, adheres to platform rules and regional regulations, and employs privacy‑preserving environments (clean rooms) where appropriate'; 84.51°'s help pages link Kroger's company privacy policy as their 'Privacy Policy' (not readable to the fetch tool); no consent basis for advertising use of loyalty purchase data and no rule for sensitive categories (pharmacy, health, alcohol) was found on the KPM or 84.51° pages read. | 10 | synthesis | supported with qualification | medium | ||
| Amazon's Sponsored Products prompts on the Amazon store (US) are explicitly generally available: announced on the page of 10 March 2026, introduced 'in November 2025 in open beta' (evidenced start) and moved 'from open beta to general availability in the U.S. on March 25, 2026' (planned start stated on 10 March 2026; general availability date 25 March 2026). | 11 | vendor assertion | supported with qualification | high | ||
| Third-party attribution for Sponsored Products prompts is unknown from Amazon's pages: neither the 10 March 2026 prompts page nor the 11 June 2026 agentic-shopping page names an advertiser-side attribution partner for the format (this review found none on those two pages by 3 November 2026). | 11 | vendor assertion | supported as stated | medium | ||
| Causal outcome measurement for Sponsored Products prompts is unknown from Amazon's pages: the 10 March 2026 page lists attributed metrics only, and the 11 June 2026 page claims 'closed-loop measurement that tracks impressions all the way to conversions' without any holdout, control, incrementality or lift wording (the fetch tool reported none of those words on the page). | 11 | vendor assertion | supported as stated | medium | ||
| Ad-quality verification (viewability, brand safety or suitability by a named third party) for Sponsored Products prompts is unknown from Amazon's pages: neither the 10 March 2026 nor the 11 June 2026 page mentions viewability, brand safety or a verification vendor (the fetch tool reported 'brand safety' and 'viewability' absent from the 11 June page). | 11 | vendor assertion | supported as stated | medium | ||
| Amazon's Sponsored Brands prompts on the Amazon store (US) are explicitly generally available: announced on the page of 10 March 2026, introduced 'in November 2025 in open beta' (evidenced start) and moved 'from open beta to general availability in the U.S. on March 25, 2026' (planned start stated on 10 March 2026; general availability date 25 March 2026). | 11 | vendor assertion | supported with qualification | high | ||
| Causal outcome measurement for Sponsored Brands prompts is unknown from Amazon's pages: the 10 March 2026 page lists attributed metrics only, and the 11 June 2026 page claims 'closed-loop measurement that tracks impressions all the way to conversions' without any holdout, control, incrementality or lift wording (the fetch tool reported none of those words on the page). | 11 | vendor assertion | supported as stated | medium | ||
| Sponsored Products and Sponsored Brands prompts on Alexa for Shopping are available with commercial stage unstated: Amazon's page of 11 June 2026 says advertisers 'today' can take part 'through sponsored ads and Sponsored Products and Sponsored Brands prompts on Alexa for Shopping' and that the experiences 'are available today', with no 'beta', 'general availability' or 'rolling out' wording on the page (evidenced availability 11 June 2026; no separate announcement or planned-start date stated). | 11 | vendor assertion | supported with qualification | medium | ||
| Operator reporting for the Alexa for Shopping prompts is available with stage unstated: the 11 June 2026 page says advertisers 'benefit from these today for our agentic offerings as they do for all our solutions' and describes 'closed-loop measurement that tracks impressions all the way to conversions'; the page itemises no metric list for the Alexa surface (the itemised Prompts report sits on the 10 March 2026 page). | 11 | vendor assertion | supported as stated | medium | ||
| Third-party attribution for the Alexa for Shopping prompts is unknown: the 11 June 2026 page names no attribution partner and the fetch tool reported 'third-party' absent from it. | 11 | vendor assertion | supported as stated | medium | ||
| Ad-quality verification by a named third party for the Alexa for Shopping prompts is unknown: the 11 June 2026 page mentions no viewability, brand-safety or verification vendor (the fetch tool reported 'brand safety' and 'viewability' absent). | 11 | vendor assertion | supported as stated | medium | ||
| ChatGPT's standard ads at the bottom of answers (US, Free and Go) are available with commercial stage unstated: announced 16 January 2026 with a planned start 'in the coming weeks'; evidenced start 9 February 2026 ('Today, we're beginning to test ads in ChatGPT in the U.S.'); OpenAI later calls that date the 'launch' ('In less than 200 days after launch', 31 August 2026; 'Since we launched ads just over eight months ago', 21 October 2026) and says ads are 'available in over 40 countries' (31 August) and in '89 countries worldwide' (21 October), but no post read says 'generally available' for the format. | 11 | vendor assertion | supported with qualification | medium | ||
| Operator reporting for ChatGPT's standard ads is available with stage unstated: advertisers 'only receive aggregate information about how their ads perform such as number of views or clicks' (9 February 2026); OpenAI 'recently launched Conversions API and pixel-based measurement' and advertisers 'receive aggregated performance insights' (5 May 2026); 'Pixel and Conversions API have become important foundations for measurement and optimization' (31 August 2026). | 11 | vendor assertion | supported as stated | medium | ||
| Third-party attribution for ChatGPT ads is available with stage unstated (platform-level statement applied to the standard format, the format in market on that date): on 5 October 2026 OpenAI said 'We now support leading attribution partners across the web and app space, enabling Conversions API, reporting, and click attribution solutions with companies including AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge, and Tenjin', with no stage label. | 11 | vendor assertion | supported with qualification | medium | ||
| Causal outcome measurement for ChatGPT ads is at pilot or test stage (platform-level statement applied to the standard format): OpenAI says 'our work on incrementality is still in its early stages' and that it is 'partnering with Haus, Measured, and WorkMagic to explore geo-based experiments to help advertisers understand the causal impact of ChatGPT Ads' (5 October 2026). | 11 | vendor assertion | supported with qualification | medium | ||
| Ad-quality verification by a named third party for ChatGPT ads is at pilot stage (platform-level statement applied to the standard format): OpenAI is 'developing brand suitability evaluation pilots with DoubleVerify (DV) and Integral Ad Science (IAS)' that 'allows independent partners to assess how our brand safety standards are applied in a controlled testing environment, without accessing private user conversations' (5 October 2026); no viewability verification is mentioned. | 11 | vendor assertion | supported with qualification | medium | ||
| ChatGPT's visual ad format is announced, not started: on 5 October 2026 OpenAI said it will 'test this new ad format during image generation in ChatGPT for users on the Free and Go plans' and 'Testing will begin later this month in the US with an initial group of advertisers' (announcement 5 October 2026; planned start October 2026; no evidenced start on the 21 October 2026 post or any other OpenAI post read by 3 November 2026). | 11 | vendor assertion | supported as stated | high | ||
| ChatGPT's Sponsored Agents are a test actually started: on 16 September 2026 OpenAI said 'Sponsored Agents are now being tested with select advertisers in the United States' (announcement and evidenced start both 16 September 2026; the operator describes the format as letting a shopper work out 'whether a product or service is the right fit'). | 11 | vendor assertion | supported as stated | high | ||
| Operator reporting for Sponsored Agents is unknown: the 16 September 2026 post describes no reporting for the format (its 'tracking performance' sentence concerns Shopify merchants' ChatGPT ad campaigns generally). | 11 | vendor assertion | supported as stated | medium | ||
| Causal outcome measurement for Sponsored Agents is unknown: no format-specific statement on the posts read. | 11 | vendor assertion | supported as stated | medium | ||
| Ad-quality verification for Sponsored Agents is unknown: no format-specific statement on the posts read (the post says the sponsored conversation 'is distinct from ChatGPT's independent answers', an operator design statement, not third-party verification). | 11 | vendor assertion | supported as stated | medium | ||
| Amazon Ads' page of 10 September 2026 announces a partnership with OpenAI and says 'Select advertisers in the U.S. are currently testing ChatGPT Ads as part of a pilot' (pilot actually started, US). The page does not name the Amazon DSP as the buying route or mention first-party data in its body, and its shopping-use wording ('explore options, compare choices, and make decisions') is platform-level, with a travel advertiser (Delta Vacations) as the only named participant; the record enters N with the same platform-level caveat as R04. | 11 | vendor assertion | supported with qualification | high | ||
| Operator reporting for Amazon-sold ads in ChatGPT is unknown: Amazon's 10 September 2026 page describes no reporting, metrics, attribution or conversions for the placement (the fetch tool reported those topics absent), and no OpenAI page on the partnership was found in one search on openai.com. | 11 | vendor assertion | supported as stated | medium | ||
| Ad-quality verification for Amazon-sold ads in ChatGPT is unknown: not mentioned on Amazon's page. | 11 | vendor assertion | supported as stated | medium | ||
| Ads in Google's AI Overviews and AI Mode are available with commercial stage unstated (US): ads in AI Overviews were 'now available for mobile users in the US' on 3 October 2024 and 'expanding to desktop in the U.S.' on 21 May 2025, when Google was 'also starting to test ads in AI Mode' (evidenced test start 21 May 2025); the Google Ads help page says 'Ads within AI Overviews and AI Mode are available in English in the United States and select English-speaking markets' and 'Text ads and Shopping ads can appear within AI Overviews and AI Mode'; on 28 October 2026 Google was 'scaling ads in AI Mode' to ten more countries and, in the US copy, 'To allow ads in follow-up queries in AI Mode, we are launching multiturn ads in the U.S.'; no page read says 'generally available'. | 11 | vendor assertion | supported with qualification | medium | ||
| Third-party attribution for ads in AI Overviews and AI Mode is unknown: none of the Google pages read (21 May 2025 post, help page, 28 October 2026 post) names an attribution partner for the formats. | 11 | vendor assertion | supported as stated | medium | ||
| Causal outcome measurement is unknown. The 28 October 2026 CTR statement is an internal engagement comparison footnoted on the US copy to 'Google Internal Data, US, August 20, 2026- September 3, 2026', with no sample or design, and is not outcome reporting. | 11 | vendor assertion | supported with qualification | medium | ||
| Ad-quality verification by a named third party for ads in AI Overviews and AI Mode is unknown: the help page and the two blog posts read mention no viewability, brand-safety or verification vendor (the tool reported them absent). | 11 | vendor assertion | supported as stated | medium | ||
| Google's Direct Offers in AI Mode are a pilot actually started: on 11 January 2026 Google wrote 'we're continuing to test ads in AI Mode and we're now introducing Direct Offers', 'This new Google Ads pilot allows advertisers to present exclusive offers for shoppers who are ready to buy', 'We are initially focusing on discounts for the pilot' and 'We're already collaborating with brands like Petco, e.l.f. Cosmetics, Samsonite, Rugs USA and Shopify merchants' (announcement and evidenced start 11 January 2026); on 20 May 2026 Google said UCP-integrated brands will 'Soon' let people 'buy on the spot' from Direct Offers. | 11 | vendor assertion | supported as stated | high | ||
| Operator reporting for Direct Offers is unknown: neither the 11 January 2026 nor the 20 May 2026 Google post describes how offers are reported to advertisers (the tool reported no reporting, measurement or attribution statements on either). | 11 | vendor assertion | supported as stated | medium | ||
| Kroger Precision Marketing's Shopping Assistant PLAs are available with commercial stage unstated: the KPM post of 3 August 2026 says 'The AI Shopping Assistant is now available across the Kroger Family of Companies websites and apps', that 'sponsored placements from active Search & Browse PLA campaigns may appear within Shopping Assistant results and are clearly identified as sponsored content' and that they are 'automatically included when running a Search & Browse PLA ad group' (evidenced availability 3 August 2026; no 'beta' or 'general availability' word on the page). | 11 | vendor assertion | supported as stated | high | ||
| Operator reporting for Shopping Assistant PLAs is available with stage unstated: 'Advertisers may use KPM's report builder to analyze Shopping Assistant PLA performance separately from other Search & Browse inventory' (3 August 2026); the post lists no metrics for the placement. | 11 | vendor assertion | supported as stated | high | ||
| Causal outcome measurement for Shopping Assistant PLAs is unknown at format level: the 3 August 2026 post says nothing about incrementality for the placement; KPM's network-level FAQ statement that its 'incrementality and closed-loop attribution methods compare exposed and unexposed groups' is not format-specific and is recorded as context only. | 11 | vendor assertion | supported as stated | medium | ||
| Ad-quality verification by a named third party for Shopping Assistant PLAs is unknown: the post mentions none; KPM's TAG Platinum certification (home page) is an operator-level certification for brand safety, fraud and malware, not a verification of this placement's delivered impressions. | 11 | vendor assertion | supported as stated | medium | ||
| Microsoft's posts imply shopping ads serve in Copilot (a quick-back comparison for 'shopping ads' on global first-party data, February to May 2025) and describe Offer Highlights in ads on product detail pages in Copilot as 'now available for retail use cases in English-speaking markets' while also inviting advertisers to 'live pilots' for Offer Highlights; no Microsoft page read names shopping ads in Copilot as a product with a stage or documents US availability for it. | 11 | vendor assertion | supported with qualification | low | ||
| Operator reporting specific to shopping ads in Copilot is unknown: the 21 April 2026 post cites 'search term and asset reporting' for AI Max and the 6 August 2025 post gives aggregate first-party comparisons, but neither describes Copilot-placement reporting an advertiser receives for shopping ads; two searches found no Microsoft help page on it. | 11 | vendor assertion | supported as stated | medium | ||
| Causal outcome measurement for shopping ads in Copilot is unknown: the figures on the posts read are internal comparisons (Copilot versus traditional search; Brand Agent-assisted versus unassisted sessions) with no counterfactual design disclosed. | 11 | vendor assertion | supported as stated | medium | ||
| Perplexity carries no H6 record and its status is unsettled from a Perplexity source: its three hub posts (12 and 18 November 2024; 9 July 2025) refused the fetch tool (403 on the two attempted, the third not attempted after two refusals), so the 'These cards aren't sponsored' statement of 18 November 2024 was not read by this stream; the October 2025 pause of new advertisers rests on TipRanks' report of conference remarks (10 October 2025: 'currently not adding advertisers'; the sponsored follow-up question format 'is still active') and is recorded as reported. | 11 | synthesis | supported as stated | low | ||
| OpenAI's product results in ChatGPT are stated to be unsponsored and Instant Checkout enablement is one listed ranking factor among several: on 29 September 2025 OpenAI wrote 'Product results are organic and unsponsored, ranked purely on relevance to the user' and 'Instant Checkout items are not preferred in product results. When ranking multiple merchants that sell the same product, ChatGPT considers factors like availability, price, quality, whether a merchant is the primary seller, and whether Instant Checkout is enabled, to optimize the user experience' (a ranking factor in ChatGPT's implementation, not an ACP specification rule); on 24 March 2026 OpenAI said it was 'allowing merchants to use their own checkout experiences while we focus our efforts on product discovery', with no ads or measurement on that page. | 11 | vendor assertion | supported as stated | high | ||
| ACP is pinned at spec version 2026-04-17 (the README's 'latest stable' folder), on commit 7fdd78df677a94dce04c770644b0fbbb1401272b (HEAD of main on 3 November 2026; committer date 2026-07-18T04:51:51Z); the repository has no GitHub releases ('There aren't any releases here.') and no tags, and its README states the protocol's task in its own words: 'an interaction model and open standard for connecting buyers, their AI agents, and businesses to complete purchases seamlessly', 'maintained by OpenAI and Stripe and is currently in beta'. | 11 | direct record | supported as stated | high | ||
| Scoped use case for ACP (representing a sponsored placement inside a shopping surface and reporting its outcome), per pinned release 2026-04-17: schema representation none in the ACP repository (the README, spec folders and RFC list carry checkout, delegated payment, discount, capability negotiation, discovery and feed; no 'sponsored', 'ads', 'advertising', 'attribution' or 'affiliate' term appears in the README or the release listing); OpenAI's separate product-feed specification carries an ads-eligibility flag ('is_ads_eligible', 'Conditionally required for Ads processing'; 'ads_metadata', 'Optional for Ads'; 'Ads and checkout have separate requirements.'), which marks catalog items for OpenAI's Ads processing and represents neither a sponsored placement nor an outcome report; implemented connector, interoperability test, production use and demonstrated transaction for the scoped use case: n/a (nothing to connect). | 11 | direct record | supported as stated | medium | ||
| ACP's own task (checkout) has operator-documented production use and operator-asserted transactions, no independent volume: OpenAI's 29 September 2025 post says 'U.S. ChatGPT Plus, Pro, and Free users can now buy directly from U.S. Etsy sellers right in chat' 'beginning with Instant Checkout, powered by the Agentic Commerce Protocol, built with Stripe', and its 8 December 2025 post says 'Instacart is the first app to offer a checkout experience directly within ChatGPT, powered by the Agentic Commerce Protocol'; no transaction count or independent demonstration was found by this stream. | 11 | vendor assertion | supported with qualification | medium | ||
| The Universal Commerce Protocol (Google-announced, 11 January 2026) is pinned at release v2026-08-25 (published 2026-08-25T13:53:03Z, marked Latest; earlier tags v2026-04-08 published 9 April 2026, v2026-01-23 and v2026-01-11 both published 23 January 2026), and its README states its task as 'An open standard enabling interoperability between various commerce entities to facilitate seamless commerce integrations' with capabilities 'Checkout', 'Identity Linking', 'Order' and 'Payment Token Exchange'; Google's announcement words are 'UCP establishes a common language for agents and systems to operate together' and 'UCP was co-developed with industry leaders including Shopify, Etsy, Wayfair, Target and Walmart'. | 11 | direct record | supported with qualification | high | ||
| Scoped use case for the Universal Commerce Protocol at v2026-08-25: schema representation partial, limited to an attribution context field: the release body says the Attribution feature 'Adds support for platform-driven referral and ad attribution context' (and v2026-04-08's changelog names an 'attribution field for platform referral context'); no release mentions 'sponsored', 'ads', 'advertising', 'retail media', 'promotion' or 'measurement' as standalone terms and the README contains none of 'attribution', 'ads', 'sponsored', 'advertising', 'retail', 'promotion', 'measurement' or 'reporting' (its extension examples are 'Discounts, Fulfillment'); implemented connector, interoperability test and demonstrated transaction for the scoped use case: none found; production use: Google says UCP 'will soon power a new checkout feature' (11 January 2026) and checkout features will be tried 'soon' with Nike, Sephora, Target, Ulta Beauty, Walmart and Wayfair (20 May 2026), and Microsoft states 'support for UCP-ready feeds in Microsoft Merchant Center' is 'Generally available today for everyone in the U.S.' (21 April 2026), a feed-format use by a named third party, not a sponsored placement or a transaction. | 11 | direct record | supported with qualification | medium | ||
| AP2 (Agent Payments Protocol, google-agentic-commerce) is pinned at v0.2.0 (published 2026-04-28T12:40:54Z, 'This is the second release of AP2. It focuses on providing Human Not Present flows.') after v0.1.0 (published 2025-09-16T12:32:33Z, 'Create Agent Payments Protocol (AP2)'); its README describes the repository as 'code samples and demos of the Agent Payments Protocol' under the banner 'Building a Secure and Interoperable Future for AI-Driven Payments', and none of 'sponsored', 'ads', 'advertising', 'retail', 'merchant', 'attribution', 'measurement' or 'UCP' appears in the README; for the scoped sponsored-placement use case every state is n/a (payments mandates only). | 11 | direct record | supported as stated | high | ||
| AdCP is pinned at release v3.2.8 (published 2026-10-28T20:56:12Z, marked Latest) with v3.3.0-beta.3 (published 2026-10-31T08:35:47Z, pre-release) and v3.1.31 (published 2026-10-30T10:31:18Z) beside it, while the documentation introduction page states 'AdCP 3.2 is generally available (release `3.2.1`, wire pin `"3.2"`)', lagging the release tag; the introduction describes AdCP as 'an open agentic advertising standard' whose goal is to become 'the enterprise coordination protocol for a trillion-dollar global advertising economy'. | 11 | direct record | supported as stated | high | ||
| AAMP 3.0 (IAB Tech Lab) is a draft in public comment: the release of 22 September 2026 says the OpenProposal specification is in 'Public Comment until October 22, 2026' and, on the same page, 'we are extending public comment to Tuesday, November 10, 2026'; the OpenProposal repository's release is v3.0-draft-1 ('OpenProposal 3.0 — draft 1 (public comment)', published 2026-09-24T15:14:53Z) and its README still shows the comment period closing 2026-10-22 ('November 10' does not appear), states 'This is a draft. Nothing here is a ratified IAB Tech Lab standard.' and defines the task as 'OpenProposal is the core AAMP 3.0 object: the descriptive specification for what is being bought and sold'; IAB Tech Lab's words for the task are 'introducing a standardized way for buyers and sellers to move from an RFP through proposal evaluation' and OpenProposal 'defines how sellers can represent advertising products and respond to buyer briefs in a standardized, agent-readable format', with 'Performance can flow back through existing reporting standards.' | 11 | direct record | supported as stated | high | ||
| Scoped use case for AAMP 3.0 (OpenProposal v3.0-draft-1): schema representation none specific (the press release and the README contain none of 'retail', 'commerce', 'sponsored', 'shopping' or 'attribution'; GitHub code search of the OpenProposal repository returned 0 hits for 'retail' and 0 for 'commerce'); outcome reporting is delegated ('Performance can flow back through existing reporting standards'); implemented connector, interoperability test, production use and demonstrated transaction: none found (a draft under comment); retail-specific vocabulary: absent; generic representability: a seller's 'advertising products' in a proposal could describe a sponsored placement in generic terms, which this stream did not test and which a keyword absence does not refute. | 11 | direct record | supported with qualification | medium | ||
| AAMP 2.3 is the released version: IAB Tech Lab 'announced the release of AAMP 2.3' on 30 July 2026, describing the framework's task as 'enabling organizations to deploy model-agnostic agentic stacks that connect to major inventory sources' and the release as 'enterprise-ready deployment, integrated privacy diligence, stronger pricing integrity', with a 'Google Ad Manager reporting integration' and 'Support for Meta buying'; none of 'retail', 'commerce', 'sponsored', 'shopping' or 'attribution' appears on the release page, and the AAMP standards page (last updated 27 October 2026) gives no release dates and lists the component repositories (OpenProposal, agentic-direct, buyer-agent, seller-agent, AAMP). | 11 | direct record | supported as stated | high | ||
| Sponsored Products prompts, introduced 'in November 2025 in open beta', moved to general availability in the US on 25 March 2026, with Sponsored Products and Sponsored Brands campaigns 'automatically enrolled', CPC billing from general availability, and a Prompts report listing 'sales, ACOS, ROAS, and 7-day orders and units' in the console and API; prompts appear in shopping results and product detail pages and 'may open a dialog in Rufus' (Amazon's assistant was renamed Alexa for Shopping after this page). | 11 | vendor assertion | supported as stated | high | ||
| Amazon's page of 11 June 2026 says 'Existing sponsored ads campaigns are automatically eligible to serve in Alexa for Shopping' and promises 'closed-loop measurement that tracks impressions all the way to conversions' for its agentic offerings, naming no holdout, control, lift or incrementality design; 'closed-loop' is Amazon's term for exposure-to-conversion tracking and is coded as attribution with no counterfactual. | 11 | vendor assertion | supported as stated | high | ||
| Amazon Ads announced on 10 September 2026 a partnership with OpenAI 'to support advertising in ChatGPT', with 'Select advertisers in the U.S.' testing ChatGPT Ads 'as part of a pilot' (Delta Vacations names itself 'among the first brands testing'); the page describes no reporting, attribution or measurement for the placement and associates it with Amazon DSP (a quotation from an Amazon DSP director and a closing Amazon DSP link) without stating the buying route. | 11 | vendor assertion | supported with qualification | high | ||
| Operator reporting for Sponsored Brands prompts is generally available in the operator's words: the page of 10 March 2026 lists a Prompts report with 'impressions, clicks, click-through rate, cost per click, spend, sales, ACOS, ROAS, and 7-day orders and units' and 'performance metrics and pause controls accessible through the Ads Console and via API'. | 11, 13 | vendor assertion | supported with qualification | high | ||
| Operator reporting for ads in AI Overviews and AI Mode is available with stage unstated and aggregate: the Google Ads help page says 'Performance metrics for ads in AI Overviews and AI Mode are included in aggregate' in campaign reports ('metrics such as impressions, clicks, conversions, and cost') and that triggering queries 'are included in aggregate' in the search terms report; the page does not say advertisers can segment AI Overviews or AI Mode performance separately. | 11, 13 | vendor assertion | supported as stated | medium | ||
| AdCP's schema represents a sponsored placement inside a shopping surface (state: schema representation, yes, at release tag v3.2.8, identical on main): the media channel taxonomy defines `retail_media` as 'Retail media networks and commerce marketplaces'; the canonical format `sponsored_placement` is titled 'Canonical Format: Sponsored Placement (retail-media catalog-driven)' and is 'Marked experimental at 3.1 GA'; the sponsored-placement-type enum carries three values, 'sponsored_search' ('e.g., Amazon Sponsored Products on search, Walmart Connect Sponsored Search'), 'sponsored_display' ('e.g., Criteo Sponsored Display, Amazon Sponsored Display on-site') and 'sponsored_native' ('e.g., CitrusAd native product cards, Pinterest catalog-driven collections'), described as 'seller-declared discovery metadata, not independent verification of inventory quality or delivery context'; the commerce-media page's worked example is a fictional retailer's product `shopgrid_sponsored_products`, named 'Sponsored Products - Search & Browse' (section heading 'Sponsored product listings'; `format_kind: sponsored_placement`, CPC, `platform_managed: true`); no Amazon-named example exists on the page. | 11, 14 | direct record | supported with qualification | high | ||
| AdCP's commerce-media page documents an outcome-report basis it calls closed-loop attribution: 'Commerce media's defining advantage is deterministic purchase attribution. Retailers match ad exposure to transactions using loyalty card data, login state, and point-of-sale records'; products declare `reporting_capabilities.available_metrics` ('conversions', 'conversion_value', 'roas', 'cost_per_acquisition', 'units_sold', 'new_to_brand_rate', 'incremental_sales_lift') and 'When `platform_managed` is `true`, the retailer provides always-on measurement'; the page's comparison table sets the 'Measurement owner' for commerce media as 'Retailer platform (`platform_managed`)' against 'Third-party (IAS, DV)' for traditional media, and notes that 'Different retailers use different attribution windows (e.g., 14-day click vs. 7-day click)'; `outcome_measurement` is marked deprecated. | 11, 14 | direct record | supported as stated | high | ||
| AdCP's retail media seller specialism `sales-retail-media` is at status preview at release tag v3.2.8, identical on the main branch (specialism version '0.1.0'; summary 'Retail media network seller', then 'onsite sponsored search, sponsored display, and offsite extensions tied to SKU-level catalog feeds, keyword targeting, and deterministic purchase attribution'; enum description 'PREVIEW', then 'Retail media network seller ... Agents claiming this specialism today pass the media-buy baseline plus retail-media-specific phases (catalog sync, keyword buy, item availability)'), and its storyboard is a fixture ('The fixture buyer (Summit Foods, via Pinnacle Agency) pushes a grocery catalog and buys sponsored search + sponsored display on ShopGrid's (Daniel Park's) retail platform'), i.e. a conformance test kit, not a demonstrated interoperability test between named third parties. | 11, 14 | direct record | supported with qualification | high | ||
| AdCP's own documentation states a promotion gate: the sponsored-placement canonical remains experimental at v3.2.8 under a gate of format_schema evidence from at least two retail media networks; the adapter-contracts page says the canonical 'stays `experimental: true` past 3.1 GA until at least two retail-media networks ship `format_schema` evidence', and its readiness matrix keeps three of the four contract families ('Amazon SP', 'Criteo SP, CitrusAd SP', 'Pinterest Collection, Snap Collection') at 'Keep experimental until evidence lands' and the generative-per-SKU family at 'Vendor-specific only' and 'Keep experimental'; the page does not itself say that no network has shipped the evidence; this stream found no documented production use, implemented connector by a named third party, or demonstrated transaction for a sponsored placement under AdCP in the pages opened (states: implemented connector none found; interoperability test none beyond the author organisation's test kit; production use none found; demonstrated transaction none found). | 11, 14 | direct record | supported with qualification | medium | ||
| Fetch's privacy policy (effective 22 January 2026, revised 13 October 2026) treats participation as consent to disclosure to brand partners ('you are consenting to Fetch disclosing your information with our brand partners'), lists advertising measurement as a purpose, excludes phone number, precise geolocation, consumer health data and other sensitive information from brand-partner sharing, requires consent for e-receipt collection, and offers a 'Do Not Sell or Share My Personal Information' opt-out for identifiers it sells or shares. | 12 | vendor assertion | unresolved | high | ||
| Moloco Commerce Media: dimension-8 scale figure 1 distinct named US retailer networks with dated production evidence for Wayfair: 'analysis period: Q4 2024' (customer provided data) (Wayfair (S16: 'Double-digit growth in display ad revenue as a % of GMV (A2G) over the past 12 months.' Source: 'Customer provided data; analysis period: Q4 2024.'). Named, undated: Costco (logo). Non-US: Olive Young, SSG, Yogiyo, W Concept, ManoMano, Damaecan; perimeter network; press release or product page). | 12 | vendor assertion | unresolved | medium | ||
| Perpetua (optimisation): status live; operating entity Flywheel Digital Limited and/or its affiliates ('Flywheel is the controller of your personal data', privacy policy last updated 06 August 2025, RTL-U-S202); owner Omnicom Group Inc. (via Flywheel; Omnicom named in the privacy contact address, RTL-U-S202; Flywheel acquisition closed 2 January 2024, RTL-U-S200); US availability yes: US office '36 Maplewood Ave, Portsmouth, NH 03801' on the offer page (RTL-U-S201); named clients (Crocs, Hero Cosmetics, Cartograph, Mi. | 12 | vendor assertion | unresolved | medium | ||
| Instacart documents at least five activation and reporting route families for its advertising: the self-serve Ads Manager, the Instacart Ads API licensed to agencies (with documented connections by Perpetua, Intentwise and Skai), The Trade Desk for off-site audiences and 'real-time sales measurement signals' (10 June 2025; Instacart Ads listed among The Trade Desk's retail partners on 17 June 2026), the Data Hub clean room, and managed insights for 'qualifying CPG brands'. | 12 | vendor assertion | supported as stated | medium | ||
| Walmart discloses its advertising business only as a global, unaudited earnings-release figure: 'grew approximately 28% to reach $3.4 billion' (fiscal 2024, ended 31 January 2024), 'grew 27% to reach $4.4 billion' (fiscal 2025) and 'grew 46% to nearly $6.4 billion, including VIZIO' (fiscal 2026), each footnoted as 'recorded in either net sales or as a reduction to cost of sales, depending on the nature of the advertising arrangement', while Walmart Connect in the U.S. appears only as fourth-quarter growth rates (22%, 24%, 41%). | 12 | direct record | supported as stated | high | ||
| Amazon.com, Inc.'s audited 'Advertising services' net sales were US$46,906 million (2023), US$56,214 million (2024) and US$68,635 million (2025), a global line that the 10-K's footnote defines as 'sales of advertising services to sellers, vendors, publishers, authors, and others, through programs such as sponsored ads, display, and video advertising'. | abstract, 1, 2 | direct record | supported as stated | high | ||
| The IAB/PwC Internet Advertising Revenue Report for full-year 2025 (April 2026) reports US 'Commerce media advertising revenue' of US$37.5 billion (2022), US$43.7 billion (2023, +16.3%), US$53.7 billion (2024, +23.0%) and US$63.4 billion (2025, +18.0%): full-year figures that PwC compiles from a survey of companies, publicly disclosed information and its own estimates for non-participating companies (mix not disclosed), with no on-site, off-site or in-store split. | abstract, 1, 2, 6 | commissioned measurement | supported with qualification | high | ||
| The FY2025 IAB/PwC report defines the surveyed line narrowly ('Commerce Media Network ads that allow a retailer to promote/ sponsor their own white-label products and/or the products they sell from their vendors directly on their own eCommerce site or on other sites', p. 32) while its narrative says commerce media 'now encompasses media powered by commerce data across on-site, off-site, and in-store environments, as well as new, growing entrants beyond retail' (p. 15); no sentence reconciles the two, and a full-text screen of the 43-page PDF found no on-site, off-site or in-store revenue split. | abstract, 1, 2, 6 | direct record | supported as stated | high | ||
| DoorDash, Inc. agreed on 28 October 2026 to acquire Maplebear Inc. (Instacart) at US$65.00 a share, an equity value of approximately US$17 billion (announced 29 October 2026), with closing 'expected' at 'the end of 2027 or the first half of 2028', subject to Instacart stockholder approval, HSR expiry and approvals including in Canada and Israel; termination fees are US$650 million (Instacart) and US$1.25 billion (DoorDash), and until closing the two 'will continue to operate as separate and independent companies'. | abstract, 1, 2, 9, 12 | direct record | supported as stated | high | ||
| H1a (attribution by default) is supported: of the 49 eligible cohort A service rows, 37 document a default outcome method and 36 of them (97.3%) report attribution with no counterfactual, above the two-thirds threshold; the other is Albertsons Media Collective in-store's matched-store lift, with assignment unknown. Counting the five 'closed loop' term-only rows as attribution gives 41 of 42. At network level all 20 documented defaults of the 31 networks are attribution (24 of 24 in the lenient reading), and the disclosure gate passes in every reading. | abstract, 1, 3, 13, 14, appendix | synthesis | supported as stated | high | ||
| H1b (controlled lift as a conditional offer) is descriptive and carries no verdict. Of the 49 eligible cohort A rows, 22 document a controlled-lift offer, 2 name one without a product, 1 is reported only by trade press, 23 have none found and 1 is unknown. Of the 22 documented offers, 15 disclose a design, 21 leave assignment unknown and one documents random assignment (Instacart Sponsored Products, on pages of 2022 and 2023 only); 2 must be requested, 1 need not be and 19 do not say; 2 state an eligibility rule or minimum; and lift appears in the default report for 1 (Albertsons Media Collective in-store), is listed in standard reporting for 4 and is conditional for 4. | abstract, 1, 3, 13, appendix | synthesis | supported as stated | medium | ||
| Amazon's Advertising Agreement (posted 23 July 2026; 'formerly known as the Sponsored Ads Agreement') bills on Amazon's own counts ('Fees are based solely on our measurements'), requires fee claims within 60 days through the console's 'Contact Us' feature, disclaims guarantees about 'THE ACCURACY OF AD SERVICES DATA, TARGETING, REPORTING', limits the buyer's use of Ad Services Data to managing campaigns through the Ad Services, evaluating performance 'only for its internal purposes' and passing Performance Data to the advertiser, allows disclosure to contractors bound by equivalent confidentiality, and states 'All Ad Services Data is Amazon’s exclusive property.' | abstract, 1, 4, 5, 10, 13 | vendor assertion | supported with qualification | high | ||
| In every cohort B row whose outcome reporting is documented, the party that produces the default number is the operator that sold the media (or, for Rokt, the network operator), on its own metric definitions; no row cites the IAB/MRC retail media guidelines' definitions or IAB Europe's, and none documents a separation between the media-sales and measurement functions; named third parties appear only as optional studies (Nielsen for Cardlytics, Circana for Ibotta, Lyft's and PayPal's partner lists) or as assessments the operator engaged, commissioned or publishes (KPMG and Nielsen for Fetch, Giant Eagle for Upside). | abstract, 1, 4, 13 | synthesis | supported with qualification | medium | ||
| Circana sells NCS Sales Effect on its NCS Measurement page, listed under an "NCSolutions Portfolio" beside NCS Targeting, NCS Optimization and NCS Enablement, as a separately branded study service with its own "Speak to an Expert" contact; "No matter which Sales Effect you choose" indicates more than one Sales Effect variant. The page states no price, contract or product code, so separate purchasability is implied by the offer page, not documented in terms. It measures "incremental sales lift and ROAS" for national CPG campaigns or campaigns "in major and consolidated markets", reports how "creative, audience, frequency and media" affect outcomes, offers cross-media reads "across TV, digital" and, for retail media networks, sales "beyond your network"; its only method words are "supercharged by machine learning" and its data basis is "Verified Purchase Data" (source not named); results are delivered by a CPG consulting team as a managed service or self-service "in a dashboard or clean room". No design, panel, window or sample is stated. | abstract, 1, 5, 9, 14, appendix | vendor assertion | supported with qualification | high | ||
| On 3 November 2026, MRC's list of accredited digital services named 41 services, and the only outcome-type metric on it, "Correlated Outcomes (Online and Offline)", was held by a single service named "Epsilon" (desktop, mobile web, mobile in-app); every other register service on the list (Amazon Ads Sponsored Products, Amazon DSP, Criteo, Instacart, Walmart Connect Sponsored Search) holds delivery metrics only. | abstract, 1, 7, 9 | direct record | supported as stated | high | ||
| H2 (no MRC outcome accreditation) is unresolved under its registered decision order. The complete screen of MRC's lists on 3 November 2026 finds no letter in force verified as accrediting an outcome metric of any of the 89 eligible cohort A, B and C services, so step one finds no counterexample, and the seven eligible services with completed MRC accreditation hold it for delivery and invalid-traffic metrics. The list's one outcome-type metric, 'Correlated Outcomes (Online and Offline)', belongs to a service named only 'Epsilon', which is mapped to Epsilon PeopleCloud Digital Media Solutions, not a register row, only by inference from Epsilon's release of 2021; an unresolved mapping blocks support at step two. Accepting the inference, H2 would be supported. | abstract, 1, 7, 9, 13, 14, appendix | synthesis | supported as stated | medium | ||
| Between the FY2024 report (April 2025) and the FY2025 report (April 2026) IAB/PwC renamed the series: the FY2024 PDF, under a section already headed 'Commerce media', charts 'Retail media network advertising revenue, full-year results (2022 – 2024)' at US$37.5, 43.7 and 53.7 billion (16.3%, 23.0%) with the survey term 'Retail Media Network ads (Retail Media / Shopper)' (p. 27); the FY2025 PDF charts 'Commerce media advertising revenue' with identical 2022-2024 values and the survey term 'Commerce Media Network ads' (p. 32), whose definition sentence is unchanged apart from its opening words, the parenthetical '(Retail Media / Shopper)' being dropped; neither report explains the change. | abstract, 2 | direct record | supported with qualification | high | ||
| Circana agreed on 26 August 2024 to acquire NCSolutions (NCS), 'the joint venture between Nielsen and Catalina', and Nielsen's Marketing Mix Modeling business, expecting to close 'during the fourth quarter of 2024'; its post dated 2 June 2025 announces 'the successful completion of its acquisition of NCSolutions', says NCS 'will join Circana’s Media team' and that Circana 'continues to work towards closing' the Nielsen Marketing Mix Modeling purchase. | abstract, 2 | vendor assertion | supported as stated | high | ||
| In LiveRamp's release of 23 October 2025, which says retail media networks can connect Meta campaign results with their own first-party sales data in the LiveRamp Clean Room to see how off-property sites 'drive sales, orders, and ROAS', Roundel's VP of Product Management says 'Clean room measurement of Meta ads with LiveRamp allows us to combine Target's first-party guest insights with Meta's platform ads data in a privacy-safe way', and Albertsons Media Collective's VP of Product Innovation says that with 'LiveRamp's expanded Meta insights' it connected 'audience-targeted Meta campaigns directly to item level sales' through 'privacy-preserving solutions' (his quote does not name the clean room); LiveRamp says incrementality, halo analysis and new-buyer insights are 'Coming soon'. | abstract, 2, 5, 6, 10, 13, 14 | vendor assertion | supported with qualification | medium | ||
| Publicis Groupe announced on 17 May 2026 an agreement to acquire LiveRamp for a total enterprise value of US$2.2 billion, expected to close before the end of 2026, and announced on 2 November 2026 "the closing of its acquisition of LiveRamp". | abstract, 2, 6, 10, 14 | direct record | unresolved | high | ||
| H4a (off-site share rising) is unresolved: no eligible series exists, because none of the six candidate series screened discloses a US on-site, off-site and in-store split on a constant definition across 2023 to 2026. The IAB and PwC commerce media revenue series for 2022 to 2025, the one read in full, has no placement split; the eMarketer, WPP Media, MAGNA and dentsu series were gated or returned errors, and vendor series were not opened for a split. | abstract, 2, 6, 13, appendix | synthesis | supported as stated | medium | ||
| Instacart documents dated reporting revisions for Sponsored Products and a public changelog: impressions switched from served to viewable impressions 'Starting January 18, 2022', applied retroactively to reporting from 1 January 2022 (earlier data 'remain unaltered'), with in-product notices where periods span the change and with clicks, CPC, attributed sales and ROAS stated to be unaffected; linear attribution from 1 August 2022; clicks-only credit from 1 April 2025; an Ads Manager changelog with nine dated entries since 25 March 2026. No restatement policy for metrics, no dispute procedure for reported outcomes and no methodology version history were found; the only correction rules are contractual: invoice disputes within 60 days, and 'Instacart shall not adjust invoices retroactively if Instacart, in its sole discretion, finds an inaccuracy in the billing amount of +/- 1% of the initial invoiced amount' (a de minimis rule, not a bar on all retroactive adjustment). | abstract, 4 | vendor assertion | supported with qualification | high | ||
| The Ovative Group and Albertsons Media Collective paper (March 2026) is a retrospective sensitivity analysis of matching choices on 42 of The Collective's own onsite display campaigns on Albertsons website properties, with no randomised benchmark. Its four levers can yield 54 iROAS values per campaign, but its headline statistics come from Exhibit 12, which uses only the no-filter variations (PSM 1:many or clustering, with and without brand sales as a feature): within campaigns the gap between highest and lowest iROAS 'averaged 6.5x (median 2.5x)', and '83% of campaigns flipped from positive to negative depending on methodology' (full report, p. 22), worded 'could flip' on p. 12 and in the executive summary (p. 4); the report does not say how a highest-to-lowest ratio was computed for campaigns whose iROAS changed sign. | abstract, 4, 8, 13 | commissioned measurement | supported with qualification | high | ||
| Across the 27 E2 rows, the bounded governance screen found no restatement or correction policy, no documented separation between the sales organisation and the measurement function, and no methodology version history; the only dispute term is Macy's Media Network's clause that its ad-server reporting is final and not subject to challenge, the only dated metric revision is eBay's general-attribution change of 13 January 2026, two networks claim IAB alignment without naming the document (CMX, DG Media Network), and the only reported third-party incrementality partner (Ipsos for Lowe's) rests on trade press. | abstract, 4, 13 | synthesis | supported with qualification | medium | ||
| For Amazon Sponsored Products, Amazon documents buyer-controlled outcome analysis in Amazon Marketing Cloud (H3 axis (b) documented): sponsored-ads advertisers 'can access AMC via self-service', run SQL-style queries over a table with '1 record per sponsored ads impression event and 1 record per click event', and over conversion tables in which 'Each conversion event in the table is attributed to its corresponding traffic event' within a 14-day window, receiving 'only ... aggregated, anonymous outputs'; per-column aggregation thresholds are labelled NONE, LOW, MEDIUM, HIGH or VERY_HIGH with no numeric value; VERY_HIGH fields (user_id, household_id, ip_address, event_id) cannot appear in a final SELECT, and no event-level export is documented (H3 axis (a) not documented). | abstract, 5 | vendor assertion | supported with qualification | high | ||
| Albertsons Media Collective announced on 17 September 2026 'Retail Media Incrementality-Based Multi-Touch Attribution', which it calls a launch, built with LiveRamp and 'powered by its Cross-Media Intelligence infrastructure', that 'connects impression-level exposure data across Albertsons Media Collective’s owned and operated channels, offsite media and participating partners including Meta, third-party inventory with Display & Video 360, Google Ad Manager and Perion, with Albertsons transaction data in a neutral, privacy-conscious environment' (the word 'neutral' comes from the operator and from LiveRamp, which sells the environment), with PepsiCo 'an early pilot partner'; no method, assignment or window is stated and general availability is not stated (the results page uses the future tense, 'will give you'). | abstract, 5, 6, 13 | vendor assertion | supported with qualification | high | ||
| Walmart Connect's article of 8 April 2026 states that 'Closed-loop measurement via LiveRamp's data clean room is now available for Meta', that 'Closed-loop measurement is also available for TikTok and Pinterest' (no mechanism named for those two), that it measures 'both online and in-store sales at Walmart', that self-serve social access is available for Meta, with TikTok self-service 'planned to rollout this half' (the page sentence is a run-on without a full stop after 'Meta and TikTok'), and that 'Add-to-Cart is currently available in beta for managed-serve clients in Meta, TikTok, and Pinterest'; of its result claims, the 2.5x incremental ROAS and 1.8x sales lift come from a sales-lift benchmark analysis performed by Circana (September 2024 to September 2025, campaigns that qualified for a sales lift study in Food, Consumables, Beverages or Alcohol only, benchmarks supplied by Circana), with no design, sample size or interval disclosed; the 48% higher ROAS is Walmart first-party FY26 data on a 14-day attribution window; and the 42% of ad-attributed sales is Walmart Connect data for 22 April to 31 May 2025 from one beta campaign. | abstract, 5, 6, 13, 14 | vendor assertion | supported with qualification | high | ||
| H3 (no documented independent loop on-site) is contradicted at the first step of its decision order: Amazon Sponsored Products, one of the six pre-registered services, documents buyer-controlled outcome analysis, since sponsored-ads advertisers can run their own SQL queries in Amazon Marketing Cloud over one record per Sponsored Products impression and per click, paired with attributed conversions, and receive only aggregated, anonymous outputs. None of the six documents event-level exposure data available to the buyer or to an independent measurer in the record screened. | abstract, 5, 13, appendix | synthesis | supported as stated | high | ||
| CVS Media Exchange's announcement (Woonsocket, R.I. dateline; reprinted by Mass Market Retailers on 29 February 2024 and quoted by Beet.TV on 1 March 2024 as 'CVS Media Exchange's announcement') says ExtraCare data are pseudonymised before entering a LiveRamp-provided clean room in which Pinterest's integration with LiveRamp reveals 'the overlapping population in ExtraCare data and Pinterest's platform data', and that 'with this technology' CMX can measure attributed sales related to campaigns and share them with CPG suppliers in the CMX Performance Dashboard, so that partners can 'track attributed Pinterest sales for the first time'; CMX was one of a few retail media networks in Pinterest's beta clean room partnership. The announcement does not say which Pinterest data (impressions, clicks or other) enter the clean room or which keys are matched. | abstract, 6, 10, 13 | vendor assertion | supported with qualification | medium | ||
| 84.51°'s roadmap (last updated 9 September 2026) marks 'Meta Advanced Analytics', which 'Enables KPM ... to measure the impact of media on Meta in a secure clean room environment', completed in Q1 2025, 'Direct Connect Meta Activations in Kroger Ad Platform' completed in Q3 2025 and 'More Direct Connect Activations in the Kroger Ad Platform' completed in Q2 2026, and lists 'Audience A/B testing' ('Split any audience segment in half for clean A/B testing') as in development for Q4 2026. | abstract, 6, 13, 14 | vendor assertion | supported as stated | medium | ||
| H4b (the off-site loop is an identity match) is contradicted under its registered rule: of the 13 cohort A services that tie outcome measurement to an off-site channel, 3 (23.1%) document that the exposure-to-purchase join runs outside the operator's logged-in surface (Walmart Connect Offsite Display, Albertsons Media Collective offsite and CVS Media Exchange), at or below the one-third threshold, with every linkage search complete. Kroger Precision Marketing offsite programmatic and Roundel name a clean room for Meta campaigns but no purchase outcome, so neither is credited. Every stricter reading gives 1 or 2 of 13; the loosest gives 8 of 13, which would leave it unresolved. | abstract, 6, 13, 14, appendix | synthesis | supported as stated | medium | ||
| IAB Europe's certificate for Moloco, created 7 October 2026, certifies 'TECHNOLOGY PLATFORM CAPABILITY TO SUPPORT MEASUREMENT AND REPORTING FOR ONSITE STANDARD DISPLAY AND SPONSORED PRODUCT ADS (WEB, MOBILE AND IN-APP)', states that it 'COVERS THE CAPABILITY OF THE MOLOCO COMMERCE MEDIA PLATFORM EXCLUSIVELY', and is 'VALID TO: 01/10/2028'; it names no standards version, metric list, retailer deployment or geography. | abstract, 7, 9, 12 | direct record | supported as stated | high | ||
| Epsilon maps its MRC 'Correlated Outcomes' accreditation to Epsilon PeopleCloud Digital Media Solutions ('DMS, formerly known as Conversant'), not to Epsilon Retail Media: its release of 11 May 2021 calls DMS 'the first offering to be accredited by the Media Rating Council (MRC) for correlated outcomes' 'across display and video advertising on desktop, mobile web and mobile in-app' and 'the only offering in the integrated suite to be accredited by the MRC', and names no retail media product; the 1 October 2025 letter continues the same scope (display and video impressions and correlated outcomes in desktop, mobile web and mobile in-app) for a service it names only 'Epsilon'; Epsilon's US retail media page makes no accreditation or 'correlated outcomes' claim. This review therefore maps the accredited service to an Epsilon service that is not a register row, with medium confidence because MRC's own letters never name the product. | abstract, 7, 9, 13, 14 | vendor assertion | supported with qualification | medium | ||
| AAM's post of 10 September 2026 states that 'AAM evaluated incrementality measurement and reporting for DoorDash Ads' Sponsored Products and Sponsored Brands' 'including its methodology, data governance, causal testing framework, metric calculations and customer disclosures', and that the certification 'provides independent verification of incrementality measurement practices including methodologies and data sources, reporting and transparency and disclosure practices' against 'industry guidelines, including the IAB and IAB Europe's Guidelines for Incremental Measurement in Commerce Media'; the post cites the ANA's Retail Media Measurement Standardization report only for a survey finding and for recommending assurance (AAM's programme page, not the DoorDash post, names the ANA document among the guidance the certification aligns with); the post names no expiry, renewal or excluded format, and does not mention MRC. | abstract, 7, 12, 13 | direct record | supported with qualification | high | ||
| MRC's services-under-review list (read in full on 3 November 2026) carries 'Kroger (Pre-audit in process as of 6/2026)' for 'Targeted Onsite Ads, Product Listing Ads, Onsite Video, Promoted Product Carousels' with 'Impressions, Clicks, Viewable Impressions, Attributed sales, Incremental Sales, aROAS, iROAS - Desktop, Mobile Web, Mobile In-App', and MRC's accredited digital list of 41 services has no Kroger entry; KPM's FAQ nevertheless says it is 'certified by the Interactive Advertising Bureau (IAB) and Merchant Risk Council (MRC)', and its home page cites TAG Platinum certification for brand safety, fraud and malware. | abstract, 7, 13 | direct record | supported as stated | high | ||
| This review found no peer-reviewed study published 2024-2026 that measures advertiser-level incremental sales lift from a US retail media network's advertising with a published design, in three targeted searches (B.json queries 2, 3 and 8), one verifier search and the pages screened by 3 November 2026 (SSRN, Google Scholar and INFORMS were not searchable with the tools used); the only 2024-2026 peer-reviewed field experiment on marketplace sponsored listings found is the Flipkart study (India; platform-side); the retail-media-specific effectiveness evidence opened is operator-authored or operator-co-authored (Ovative with Albertsons Media Collective; Amazon; Instacart). | abstract, 8 | inference | supported with qualification | medium | ||
| Instacart's pages report more lift figures than the two quoted: Reckitt's '16.5% incremental sales lift' (sponsored product ads, Q3-Q4 2022, 'two randomly selected groups'); 'more than a 15% incremental sales lift – and sometimes twice that' on average (internal tests, year ending 31 December 2023); Kellogg's 'double-digit total sales lift' (2022); Applegate 'a 35% sales lift' and Challenge Dairy 'an incremental sales lift of over 60%' (17 December 2024 post; format, period and design not stated); and Circana-provided in-store sales lifts of 21% and 15% for two unnamed CPGs 'among consumers exposed to its sponsored product ads' (exposed users 1 April to 16 June 2024 'vs. Circana control group'; control construction and assignment not stated). None gives sample size, interval, significance rule or spend; every figure is published by Instacart, which sells the media. | abstract, 8 | vendor assertion | supported with qualification | medium | ||
| This review found no Circana document, among the 15 circana.com pages read (plus two that returned 404) and four searches by 3 November 2026, that describes Media Lift's allocation, minimum sample, significance rule, uncertainty, balance checks, metric definitions, method versioning, correction or dispute procedure, buyer data access or export, or any validation of Media Lift's own numbers against an experiment, a replication or point-of-sale truth. | abstract, 8, 9 | inference | supported as stated | medium | ||
| Across the eleven cohort D rows, this stream found incrementality testing documented by the platform itself for none: Flywheel asserts an ads-on versus ads-off sales comparison, Stackline's attribution page an 'Incrementality analytics' module that 'deciphers the net new customer impact of every impression', CommerceIQ 'iROAS' optimisation, Criteo a user-split or geo-split framework that belongs to Commerce Growth and is in pilot, and Pacvue an Incrementality Console that could not be read; no row documents assignment, a counterfactual construction with method, minimum samples or a significance rule. | abstract, 9 | synthesis | supported as stated | medium | ||
| Numerator's media-industry page, addressed to media companies, offers 'sales lift studies that analyze exposed vs non-exposed consumers' to 'help prove the effectiveness of your advertisers’ campaign investments'; the only method document opened, the SecureMeasure client success story (PDF created 30 June 2023), is a CRM-marketing case in which first-party CRM data were matched to the panel 'using clean room technology' and 'six continuous months of consumer purchase behavior were analyzed to measure the pre and post impact of marketing touchpoints'; how the unexposed comparison for media sales-lift studies is constructed is not documented in the opened record, so assignment is coded unknown, and because the studies are marketed to media sellers the commissioning party of a given study may be the seller (funding unknown per study). | abstract, 9 | vendor assertion | supported with qualification | medium | ||
| Catalina's HUB360 pages document exposure-to-purchase attribution as the measurement basis ('Our Catalina Pixel and direct integration in the store means we can attribute your media to an actual purchase.'; 'We use actual shopper transactions to fuel our sales impact measurement and anonymously connect exposures to a purchase in real time.') and list 'Campaign Snapshot', 'Incrementality Report' and 'Multi-Touch Attribution' as HUB360 Measurement features without describing any of them; no attribution window, touch model, match basis, test-and-control design, holdout or significance rule appears on the three product pages opened, and the one case cited ('using Catalina's Multi-Touch Attribution to A/B test CTV audiences and ad creative', '4.3% Sales Lift') states no design. | abstract, 9 | vendor assertion | supported as stated | medium | ||
| Nielsen's Sales Lift page documents the standard design in one sentence, 'Our privacy-first approach uses industry-standard control/exposed methodology and retailer-verified purchase data.', sourced 'Whether through third-party integrations or proprietary datasets', with channels including 'digital, linear, CTV, retail media and more' and delivery through 'always-on dashboards, AI integrations and intuitive reporting'; the page states no allocation rule, control construction, attribution window, significance rule, panel or feed size, data partner or country, and uses both 'incremental sales' and 'the sales attributed to marketing' for its output. | abstract, 9 | vendor assertion | supported as stated | medium | ||
| Attain's measurement page offers attribution and incrementality as separate products: 'Sales Attribution' ('Did exposed consumers actually buy? Tie ad exposure to real transactions across retailers and channels') with no counterfactual, and 'Incrementality', described as 'Measure true causal lift using advanced holdout methodologies—and quantify impact through ROAS and iROAS'; exposure is ingested by tags, uploaded logs or CAPI, a pre-measurement exposure-to-panel match check is asserted, and results 'update daily'; the page does not say how the holdout is formed or by whom, so for the Incrementality product assignment is coded unknown and the counterfactual as holdout (construction undisclosed), and the Sales Attribution product is coded attributed with no counterfactual. | abstract, 9 | vendor assertion | supported with qualification | medium | ||
| Incremental documents its Measurement & Reporting method at principle level only: 'Multiple causal techniques link investment to outcome, not the click that came last.'; 'Econometric techniques combined with designed and synthetic experiments separate media impact from other factors by retailer and channel'; 'An ensemble causal model decomposes sales into what caused them.'; 'Daily retraining keeps the model in sync with the business.'; its FAQ question 'Can I use Incremental to run A/B tests?' is answered (per the fetch tool's summary) that Incremental does not run A/B tests itself but integrates results from tests on retail platforms such as Amazon and Walmart to validate and calibrate its models; no baseline definition, counterfactual construction, uncertainty statement or iROI formula appears on the five Incremental pages opened, so the default is coded as a model-based (synthetic) counterfactual with no assignment. | abstract, 9 | vendor assertion | supported with qualification | medium | ||
| Moloco documents, in an undated vendor blog, a randomised ghost-bidding holdout built into its ad decisioning: ads are withheld from a randomised subset of users whose ad would otherwise have won, all purchases during the experiment are counted in both groups, and the method is offered to advertisers meeting unstated minimum traffic thresholds; reported incremental ROAS ranged from 253 to 1,609 percent across unnamed advertisers and platforms with no period, sample size, duration or significance rule. | abstract, 9, 12 | vendor assertion | supported as stated | medium | ||
| Commerce Yield's own help centre documents per-campaign attribution settings with defaults: post-click lookback of 7, 14 or 30 days (default 30) and post-view of none, 1, 7, 14 or 30 days (default 1), post-view never longer than post-click, and a product match per window (default post-click 'Same Product + Same Category', post-view 'Same Product', with options widening to same brand for brand accounts or same seller for seller accounts); settings can be changed after launch, take up to two hours to reach reporting and are 'retroactively applied to the full extent of the campaign'. | abstract, 9, 12 | vendor assertion | supported as stated | high | ||
| California: the CCPA (text as effective 1 January 2026) gives consumers a right to opt out of the sale or sharing of personal information (§ 1798.120), defines 'share' as disclosure 'for cross-context behavioral advertising, whether or not for monetary or other valuable consideration', gives a right to limit the use of sensitive personal information (§ 1798.121), which includes 'Personal information collected and analyzed concerning a consumer's health', and the regulations effective 1 January 2026 require businesses to treat a compliant opt-out preference signal as a valid opt-out of sale/sharing (§ 7025) and to offer a 'Limit the Use of My Sensitive Personal Information' route (§ 7027). | abstract, 10 | direct record | supported as stated | high | ||
| Washington's My Health My Data Act (RCW 19.373; the duties in RCW 19.373.020 to .070 apply from 31 March 2024, small businesses from 30 June 2024) defines consumer health data to include 'Use or purchase of prescribed medication' and health data inferred from non-health information 'by any means, including algorithms or machine learning'; permits collecting it only with consent for a specified purpose or to the extent necessary to provide a product or service the consumer requested, and sharing it only with a separate consent or to that same extent; makes selling it unlawful 'without first obtaining valid authorization'; bars geofences within 2,000 feet of in-person health-care providers used to track consumers seeking care, collect their health data, or send them notifications, messages or advertisements; does not accept acceptance of general terms of use as consent; and makes violations unfair or deceptive acts under the Consumer Protection Act. | abstract, 10 | direct record | supported with qualification | high | ||
| Maryland's Online Data Privacy Act (2024 ch. 455, SB 541; effective 1 October 2025) provides that a controller may not collect, process or share sensitive data (which includes consumer health data) except where strictly necessary to provide or maintain a specific product or service the consumer requested (the enrolled text strikes 'and unless the controller obtains the consumer's consent'), may not sell sensitive data, must limit collection to what is 'reasonably necessary and proportionate', may not process personal data for targeted advertising, or sell it, where it knew or should have known the consumer is under 18, and had to honour opt-out preference signals by 1 October 2025; Section 2 of the chapter bars applying § 14-4612 as enacted to processing activities before 1 April 2026. | abstract, 10 | direct record | supported with qualification | medium | ||
| This review found no FTC or state attorney-general order, settlement or court finding, and no breach notice, concerning a register network's use of shopper data for its retail media service (Amazon, Walmart, Instacart, Kroger, Target, Albertsons, Criteo, Chase, Cardlytics, Circana, Numerator) in the searches and pages logged on 3 November 2026; the official matters opened that touch a register company are inquiries (the 2024 6(b) orders, JPMorgan Chase; the 2026 6(b) orders, Instacart, Target, Walmart, DoorDash, Uber, Lyft and United), Criteo as a named recipient in GoodRx (2023), and Instacart's December 2025 settlement over fees and subscriptions, which does not concern data. | abstract, 10 | synthesis | supported with qualification | low | ||
| On 28 October 2026 the FTC (vote 2-0; Chairman Ferguson and Commissioner Meador) issued Section 6(b) orders to Delta, United, JetBlue, Lyft, Uber, DoorDash, Instacart, Target and Walmart (the release names no legal entities and groups them as 'airlines, ride-sharing companies and food-delivery firms') to study whether pricing 'varies from one person to the next based on factors such as participation in loyalty programs or analysis of personal data by superintelligence algorithms', and personalised driver compensation. The template order carries FTC Matter No. P246202, the number on the 2024 surveillance-pricing order, covers 1 January 2023 onward, defines personal information to include 'purchases, leases, subscriptions, browsing history, reactions to advertising' and asks for the exact wording of each disclosure to consumers or workers. The release says 'The agency learned nothing about actual pricing decisions from the 2024 study', which it says 'examined only middlemen consultancies rather than the businesses that actually set prices'. | abstract, 10 | direct record | supported with qualification | medium | ||
| Operator reporting for Sponsored Products prompts is generally available in the operator's words ('prompts and reporting are now generally available', 10 March 2026), with a Prompts report listing impressions, clicks, CTR, CPC, spend, sales, ACOS, ROAS and 7-day orders and units, in the Ads Console and via API. | abstract, 11, 13 | vendor assertion | supported with qualification | high | ||
| H6a (operator-reported inside the agent) is supported: of the eleven frozen sponsored-format records on agent surfaces, the two explicitly generally available, Amazon's Sponsored Products prompts and Sponsored Brands prompts on the Amazon store, both have operator reporting generally available, through a Prompts report of impressions, clicks, spend, sales, ACOS, ROAS and 7-day orders and units produced by Amazon. | abstract, 11, 13, appendix | synthesis | supported as stated | medium | ||
| H6b (third-party attribution mostly none or pilot) is unresolved: none of the eleven records documents explicit general availability of third-party attribution, so it is not contradicted, but none documents the axis as none or pilot either. ChatGPT's standard ads name attribution partners with no stage stated and the other ten records are unknown, so (G + U) / N is 11 of 11, above one half. | abstract, 11, 13, appendix | synthesis | supported as stated | medium | ||
| H6c (causal outcome measurement mostly none or pilot) is unresolved: none of the eleven records documents explicit general availability of causal outcome measurement, and only ChatGPT's standard ads document a pilot (geo-based experiments with Haus, Measured and WorkMagic, which OpenAI calls early); the other ten records are unknown, so (G + U) / N is 10 of 11, above one half. | abstract, 11, 13, appendix | synthesis | supported as stated | medium | ||
| H6d (ad-quality verification mostly none or pilot) is unresolved: none of the eleven records documents explicit general availability of ad-quality verification by a named third party, and only ChatGPT's standard ads document a pilot (brand-suitability evaluation pilots with DoubleVerify and Integral Ad Science); the other ten records are unknown, so (G + U) / N is 10 of 11, above one half. | abstract, 11, 13, appendix | synthesis | supported as stated | medium |
The claim ledger lists the 572 claims marked in the text. The print edition leaves it out; it is listed in full in the web edition and in data/claim-ledger.csv, with each claim’s evidence class, review outcome and confidence.
The corrections
The ledger in full.
Corrections ledger
Every substantive change made after the verification and challenge passes, including withdrawn claims. Already reflected in the chapters; reproduced so the difference between first research and final text can be audited.
Machine-readable copy: data/corrections-ledger.csv.
| ID | Original claim or score | Challenge | Evidence checked | Verdict | Change made | Reason | Affected |
|---|---|---|---|---|---|---|---|
| C001 | On MRC's digital accredited-services list as read on 3 November 2026, six entries concern register services or their operators (Amazon Ads Sponsored Products, Amazon DSP, Criteo, Epsilon, Instacart, Walmart Connect Sponsored Search), and every accredited metric in them is a delivery metric (impressions, clicks, CTR, viewable impressions, invalid-traffic filtration) except the entry 'Epsilon', which lists 'Correlated Outcomes (Online and Offline)' for D, MW and MA. | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | Six letters opened and extracted locally (pymupdf). Every accredited statistic in five of them is a delivery metric (impressions, clicks, CTR, viewable impressions and viewability metrics, invalid-traffic metrics); Epsilon's letter (1 October 2025) alone adds an outcome-type metric: 'Display and Video Impressions and Correlated Outcomes (online and offline) within Desktop, Mobile Web and Mobile In-App environments, reported Gross and Net of General Invalid Traffic (GIVT)'. MRC's Q4 2025 update (22 January 2026, p. 2) also records, for November 2025: 'IAS Third-Party Reporting of Amazon DSP ... | supported with qualification | Keep the metric-class statement (verified from the letters). Add: 'The list strings quoted (entry names, "Clicks: Search": Total only, the Criteo and Amazon DSP footnotes) rest on the 3 November keyword read and are confirmed only by the full enumeration. If the digital list carries a separate entry for IAS third-party reporting of Amazon DSP (granted November 2025 per MRC Q4 2025 update), entries concerning a register service number seven, the seventh being a third-party reporting accreditation of Amazon DSP traffic for delivery metrics only.' | MRC's two list pages were not opened by this verifier: per the assignment, the main session is enumerating both pages in full in a browser on 3 November 2026, which settles the list strings. Consistency check done here: none of MRC's five quarterly updates (Q3 2025 to Q3 2026, all opened) records any accreditation decision (grant, continuation, denial) for Kroger, Kevel, DoorDash, Schwarz or Walma | RTL-A-F01 (evidence/verify/A.json) |
| C002 | Epsilon maps its MRC 'Correlated Outcomes' accreditation to Epsilon PeopleCloud Digital Media Solutions ('DMS, formerly known as Conversant'), not to Epsilon Retail Media: its release of 11 May 2021 calls DMS 'the first offering to be accredited by the Media Rating Council (MRC) for correlated outcomes' 'across display and video advertising on desktop, mobile web and mobile in-app' and 'the only offering in the integrated suite to be accredited by the MRC', and names no retail media product; the 1 October 2025 letter continues the same scope (display and video impressions and correlated outcom | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | Epsilon release dated 'May 11, 2021': 'Epsilon today announced that Epsilon PeopleCloud Digital Media Solutions (DMS, formerly known as Conversant) is the first offering to be accredited by the Media Rating Council (MRC) for correlated outcomes—a metric defined as an event (i.e. purchase, action, etc.) connected to an underlying ad served—across display and video advertising on desktop, mobile web and mobile in-app.' 'The rigorous accreditation process evaluated the correlated outcome, tracked ads, and rendered impression metrics available through Epsilon PeopleCloud DMS along with internal pr | supported with qualification | Mapping verdict: HOLDS as the best-supported reading; it is not wrong, and it stays formally unresolved on MRC's own record because the 2025 letter and the list name only 'Epsilon'. Add to the support the MRC CEO's quoted sentence naming 'Epsilon PeopleCloud DMS' (inside Epsilon's release, so still not an MRC record) and the scope-continuity inference (2021: correlated outcomes, tracked ads, rendered impressions, display and video, desktop/mobile web/in-app; 2025: display and video impressions and correlated outcomes, same three environments). Add, labelled inference: each of the other five le | The mapping is an inference from a vendor release plus continuity of scope; code the mapping row as inference (evidence of the 2021 release itself stays vendor_assertion). | RTL-A-F04 (evidence/verify/A.json) |
| C003 | MRC's letter of 29 May 2026 continues and extends accreditation of Amazon Advertising's DSP by supply source: for the Amazon Retail supply source it continues display BTR impressions and click-throughs (D, MW, MA) and newly grants video BTR impressions and click-throughs and display and video viewable impressions (D, MW, MA); Fire TV is accredited for display impressions and click-throughs in CTV; APS, 3PX, IMDb and Alexa connected devices carry their own scopes; the retail media guidelines are not cited. | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | Letter of May 29, 2026 continues (1) display BTR impressions and click-throughs for Amazon Retail, IMDb, APS and 3PX (D, MW, MA), (2) Fire TV display BTR impressions and click-throughs in CTV, (3) APS and 3PX video BTR impressions and click-throughs; and grants initial accreditation for (1) Amazon Retail video BTR impressions and click-throughs and display and video viewable impressions, (2) IMDb display viewable impressions, (3) APS and 3PX display and video viewable impressions, (4) 'Amazon Alexa connected devices for Display BTR Impressions and Click-throughs ... and Viewable Impressions'. | supported with qualification | Restate: 'MRC's letter of 29 May 2026 consolidates the January 2026 decisions (initial grants for Amazon Retail video BTR impressions and click-throughs and display and video viewable impressions; IMDb display viewable impressions; APS and 3PX display and video viewable impressions) and the May 2026 grant for Alexa connected devices, and continues display BTR impressions and click-throughs for Retail, IMDb, APS and 3PX (D, MW, MA), Fire TV display in CTV, and APS/3PX video.' The date of the Retail grants is January 2026, not May 2026. | RTL-A-F09 (evidence/verify/A.json) | |
| C004 | H2 screen result, decision order: step 1 (contradicted) is not met, because no letter in force on 3 November 2026 is verified as accrediting an outcome metric of an eligible cohort A, B or C service at an exact service, metric, environment and date (Epsilon's 'Correlated Outcomes' is the only outcome-type accredited metric, and Epsilon's own record ties it to Epsilon PeopleCloud Digital Media Solutions, formerly Conversant, which is not a register row; Kroger, Kevel and Schwarz carry outcome-type metrics only as under-review statuses, not letters); the review stops at step 2 and H2 is UNRESOLV | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | The decision logic follows from the letters: the only accredited outcome-type metric among register-relevant letters is Epsilon's 'Correlated Outcomes (online and offline)'; its mapping to Epsilon PeopleCloud DMS (formerly Conversant) is supported by Epsilon's 2021 release, including MRC's CEO quoted naming 'Epsilon PeopleCloud DMS'. The quarterly updates record no accreditation decision for Kroger, Kevel or Schwarz. | supported with qualification | Keep UNRESOLVED at step 2. Add the MRC CEO quote to the Epsilon applicability note. The statuses of Kroger, Kevel and Schwarz rest on the list read (not opened by this verifier; main session enumerating). | Synthesis label is right; the Epsilon applicability judgement inside it is an inference and should say so. | RTL-A-F11 (evidence/verify/A.json) |
| C005 | Between the September 2023 draft and the January 2024 final the guidelines' normative wording softened: counted case-insensitively in the extracted text, 'must' falls from 49 to 35 occurrences, 'should' from 126 to 86, and 'is recommended' or 'are recommended' rises from 1 to 45; for example the draft's 'Retail media organizations must use robust measurement methodologies and technologies' becomes 'Retail media measurement is recommended to use robust measurement methodologies and technologies'. | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | Re-run with pymupdf, case-insensitive: \bmust\b 49 (draft) to 35 (final); \bshould\b 126 to 86; \b(is|are) recommended\b with a single space 1 to 45; with any whitespace between the words (matches split across a line) 1 to 49. 'require' 22 to 34 reproduces only as require+requires+required (draft 1+9+12, final 7+11+16), excluding 'requirement(s)' (39 to 47). Example sentences as quoted (draft p. 9, final p. 10). | supported with qualification | Restate: "'is recommended' or 'are recommended' rises from 1 to 45 (49 when occurrences broken across a line are counted)"; in the derivation, define the 'require' count as require, requires and required, excluding requirement(s). | RTL-A-F37 (evidence/verify/A.json) | |
| C006 | The IAB and IAB Europe In-Store Retail Media Definitions and Measurement Standards (final, dated Dec. 03, 2024 on IAB's page; no version number) were finalised 'following a public comment period during September, October and November 2024'; they define five store zones, say that today RMNs 'should report on ad play and gross impressions' while 'opportunity to see is the best proxy available for a viewable ad impression in the In-Store environment', give an OTS example built on eight variables (p. 16), recommend sales reported 30 days pre-exposure and 30 days post-exposure with Sales Lift and B | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | V1 PDF (35 pp., created 2 December 2024): p. 3 'This is the final version of the standards following a public comment period during September, October and November 2024'; p. 4 the MRC DOOH paragraph ending 'The MRC Guidelines have been used as the basis for the development of these standards.'; p. 7 'IAB has defined five zones'; p. 10 'Today RMNs should report on ad play and gross impressions. ... In the future for digital activations opportunity to see is the best proxy available for a viewable ad impression in the In-Store environment.'; p. 16 OTS example with eight variables, 'Final Estimat | supported with qualification | Add to the claim: V1 also defines 'Sales Variance' and 'Brand Variance' (p. 18) and an 'Incremental Sales/Brand Lift' section (pp. 19-21) whose test-versus-control method says consumers are 'ideally' sorted randomly and requires group sizes and statistical significance to be reported. Rewrite the limitation: p. 17 states no method for 'Sales Lift' beyond the reporting windows and excluding non-activated stores; calling it a pre/post comparison is an inference, and V1's controlled methods sit in the separate 'Incremental Sales/Brand Lift' section. | IAB's web page (S21, 'Dec. 03, 2024', 'build on the IAB/MRC Retail Media Measurement Guidelines') was not re-opened. | RTL-A-F18 (evidence/verify/A.json) |
| C007 | IAB Europe's 'In-Store Retail Media Definitions and Measurement Standards, Version 2' is a 'Draft For Industry Review - September 2026', released for comment on 17 September 2026 with feedback requested 'by 23rd October'; per its own change table it keeps the five store zones (adding zone-by-zone measurement guidance), re-anchors the in-store impression on footfall ('visual contact with the ad') instead of ad play, replaces Version 1's eight-variable OTS example with a 'simplified six-metric core calculation' (average daily store footfall x stores x days x placements x aisle penetration x camp | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | V2 draft p. 4 change table: '... All other definitions carry over from Version 1 unchanged.' Rows: zones unchanged with guidance added; impression 'defined as visual contact with the ad'; OTS 'Had an example based on 8 variables' -> 'Simplified six-metric core calculation based on daily footfall (OTS 36,000,000)'; aisle penetration 'four measurement methods'; reporting duration 30 -> 28 days; 'Incrementality (Section 6) | Standalone sub-section | Removed. Incrementality continues to be addressed under Incremental Sales/Brand Lift'. The table lists no change to Sales Lift. V2 p. 18 'Sales Varia | contradicted | Replace the Sales Lift sentence with: "Version 2 drops Version 1's 'Sales Lift and Brand Sales Lift' metrics (V1 p. 17) and Version 1's 'Incremental Sales/Brand Lift' section (V1 pp. 19-21: test-versus-control, multivariate and one-to-one methods, with consumers 'ideally' sorted into groups at random and group sizes and statistical significance to be reported); it keeps Version 1's 'Sales Variance' (before-and-after testing, to be used 'only when Incremental Sales studies are not possible') and 'Brand Variance' definitions unchanged, and its only incrementality text is the line 'Incrementality | Consequential for the paper's in-store narrative: the proposed V2 removes the only in-store text that described a controlled lift design and asked for group sizes and significance, while the before-and-after 'Sales Variance' method survives. The ExchangeWire release copy (S34) was not re-opened. | RTL-A-F19 (evidence/verify/A.json) |
| C008 | IAB Europe's commerce media measurement standards moved from the 'Retail Media Measurement Standards' (V1, posted 25 April 2024 after 'a public comment period between February and March 2024') to V2 released on 22 January 2026 after a public comment period the standards text dates 'September to November 2025' and the comment release of 9 October 2025 closes on 'Friday, 14th November 2025'; V2 adds a formal incrementality definition with 'approved methodologies', gross and net sales definitions, quick-commerce metrics, five new-to-brand timeframes and a 30-day default lookback with mandatory fl | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | V2.1 PDF p. 3 and p. 4 as quoted; p. 8 lookback as quoted; p. 9 five NTB/NTC timeframes; p. 10 'A CMN doesn't have to provide incrementality measurement to be certified, but if they do, they have to use one of the methods specified in the Guidelines.' Release (dated 22 January 2026; dateline 'Brussels, Belgium, 22nd January, 2026'): 'A formal definition of incrementality, with approved methodologies specified'; 'Updated guidance on new-to-brand and new-to-category timeframes, now expanded to five generic timeframes'; 'reinforces that a 30-day lookback window is the default for reporting, while | supported with qualification | Attribute "'approved methodologies'" to the 22 January 2026 release, not to the standard (the standard says 'one of the methods specified in the Guidelines'). In the limitation, the 'six-month grace period ... until the end of July 2026' wording is in the release itself as well as on the hub page. | V1 date and comment period (S50) and the comment release (S28) were not re-opened. | RTL-A-F23 (evidence/verify/A.json) |
| C009 | IAB Europe's compliance table (Update May 2026, v2.1) tests attribution reporting on pp. 5-8: sale definitions ('When the CMN doesn't specify which sale they are reporting on, it is assumed to be a gross sale'), the 30-day default lookback, same-SKU and halo attribution ('Same Brand, Same Category'), disclosure of sales extrapolation, and new-to-brand categorisation; its sales-uplift row (3.3) requires a 'Statistically significant A/B Test (not before and after test)' and the control 'Review methodologies for determining Sales Uplift. Review and test results.'; its incrementality row (4.1) say | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | Text as quoted on pp. 2, 5, 6, 7, 8. The legend on p. 1 reads 'REQUIREMENTS THAT MUST BE COMPLIED WITH' (green), 'REQUIREMENTS THAT ARE OPTIONAL' (yellow), 'ADDITIONAL INFORMATION FOR COMPLIANCE'. Rendering the pages and sampling the guideline cells: green (must) = 1.1 Ad Impression, 1.2 Click-Through, 1.3 Viewability ('required if video ad products are in scope'), 1.4 Video Completion Rate, 1.5 GIVT, 2.1 Sale Definitions, 2.1 ROAS, 2.2 Lookback Windows, 2.3 Same SKU and Halo, 2.4 Sales Extrapolation, 3.1 New to Brand, 3.3 Sales Uplift, 6.1 Media Audit; yellow (optional) = 1.7 SIVT, 2.1.1 mark | supported with qualification | Replace the limitation 'which rows are mandatory could not be read' with the reading above, and add to the claim: 'Sales Uplift (3.3) is a mandatory row for any CMN that reports the metric ("Statistically significant A/B Test (not before and after test)"), while incrementality (4.1) and SIVT filtration (1.7) are optional rows.' Keep the caution that the A/B wording establishes neither random assignment nor how groups are formed. | Colours read from the PDF render (pymupdf pixel sampling of each guideline cell, confirmed visually at higher resolution for pp. 5 and 7). | RTL-A-F25 (evidence/verify/A.json) |
| C010 | IAB Europe's Retail Media Certification Programme certifies retailers and ad tech companies after an independent audit against the standards by one of three auditors (AAM, ABC, CESP), with a badge valid for two years, fees of EUR 7,500 for RMNs and EUR 7,500 or EUR 15,000 for ad tech companies (below or above EUR 50m annual media revenue) plus a separately paid audit, a requirement of membership of a national IAB in Europe, and a scope that can be 'their whole retail media offering or specific ad products', published with the certificate. | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | Programme page: 'IAB Europe has selected three non-profit auditors (AAM, ABC, and CESP)'; 'the IAB Europe Retail Media Certification badge is valid for two years'; 'Certification is valid for two years from the date issued on the Certificate'; fees 'The certification fee for RMN businesses is 7,500 Euros.' 'ad tech companies generating more than 50m+ Euros of media revenue per year will be 15,000 Euros, and those with less than 50m Euros of media revenue will be 7,500 Euros'; 'the independent audit will need to be paid for'; 'Any company wishing to be certified must be a member of a National I | supported with qualification | Cite the scope sentence to the programme page (S31) rather than the summary-only post (S33), and raise confidence to high. Add the mandatory-requirements rule, 'compliance to be achieved during the audit itself', validity 'from the date issued on the Certificate', and the MRC coordination paragraph (relevant to Schwarz Media's MRC audit and to how an MRC letter can count toward IAB Europe certification). | No '90%' rule appears on the page, as the file says. | RTL-A-F26 (evidence/verify/A.json) |
| C011 | AAM's post of 10 September 2026 states that 'AAM evaluated incrementality measurement and reporting for DoorDash Ads' Sponsored Products and Sponsored Brands' 'including its methodology, data governance, causal testing framework, metric calculations and customer disclosures', against 'IAB and IAB Europe's Guidelines for Incremental Measurement in Commerce Media' and 'The ANA's Retail Media Measurement Standardization report', and that the certification 'provides independent verification of incrementality measurement practices including methodologies and data sources, reporting and transparency | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | Post (datePublished 2026-09-10): 'During the certification, AAM evaluated incrementality measurement and reporting for DoorDash Ads’ Sponsored Products and Sponsored Brands including its methodology, data governance, causal testing framework, metric calculations and customer disclosures.' 'AAM’s Incrementality Certification provides independent verification of incrementality measurement practices including methodologies and data sources, reporting and transparency and disclosure practices against industry guidelines, including the IAB and IAB Europe’s Guidelines for Incremental Measurement in | supported with qualification | Contradicted element, correct it: the post does not say DoorDash was evaluated against the ANA report. Restate: "... against 'industry guidelines, including the IAB and IAB Europe's Guidelines for Incremental Measurement in Commerce Media'; the post cites the ANA's Retail Media Measurement Standardization report only for a survey finding and for recommending assurance (AAM's programme page, not the DoorDash post, names the ANA document among the guidance the certification aligns with)." Apply the same fix to the assurance_instruments AAM row ('standard_applied'). | Standard-applied field misattributed (survey context coded as an audit standard). | RTL-A-F30 (evidence/verify/A.json) |
| C012 | The ANA document AAM names as 'ANA Retail Media Network Measurement Standards' is, on ANA's own page of 18 August 2026, a research report titled 'Retail Media Measurement Standardization' that 'highlights key learnings and establishes an initial framework to help drive outcome comparability', gated behind a free log-in, with no sample size, fielding dates or authors on the open page; AAM's post of 2 September 2026 describes it as based on 'a survey of Association of National Advertisers (ANA) members' and recommending standardised baseline metrics (impressions, viewability, clicks, invalid tra | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | AAM programme page link text 'Association of National Advertisers (ANA) Retail Media Network Measurement Standards' points to https://www.ana.net/miccontent/show/id/rr-2026-08-retail-media-measurement; AAM's DoorDash post names the same URL 'ANA's Retail Media Measurement Standardization report'; AAM's assurance brief calls it 'a recent report on retail media by the Association of National Advertisers'. | supported with qualification | The one-document reading is confirmed by the shared link target. Replace the limitation sentence 'The AAM assurance brief of 14 October 2026 ... was not located' with the located source: 'AAM Assurance Brief: Increasing Confidence in Incrementality Measurement', blog.auditedmedia.com/newsviews/assurance-brief-incrementality (datePublished 2026-10-14). | ANA's report page (S39) and AAM's 2 September post (S38) were not re-opened, so the gating, missing sample and 'survey of ANA members' wording are not re-verified. | RTL-A-F32 (evidence/verify/A.json) |
| C013 | AAM also offers a 'Retail Media Certification' (described in its post of 29 October 2025) that 'provides third-party, independent assurance that demonstrates applicable measurement guidelines' are applied, assessed against the IAB/MRC Retail Media Measurement Guidelines (US) or IAB Europe's standards, with certified companies listed on AAM's Assurance List; on 3 November 2026 the Assurance List offered no 'Retail Media' certification filter among its sixteen filter labels and a request for 'Retail Media Certification' returned no company, so this review found no listed holder of AAM's Retail M | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | Post dated 'October 29, 2025': 'AAM’s Retail Media Certification provides third-party, independent assurance that demonstrates applicable measurement guidelines are implemented correctly and consistently across networks, products or platforms.' Step 1 links 'IAB U.S.' to iab.com/guidelines/final-iab-mrc-retail-media-measurement-guidelines-available-for-download/ and 'IAB Europe' to the V1 knowledge-hub page. Step 4: 'Certified companies receive a listing on AAM’s Assurance List'. AAM's programme page: 'AAM also provides Retail Media Certifications to ensure retailers’ systems and ad tech compa | supported with qualification | Bound the absence: 'no Assurance List filter is labelled retail media; whether AAM lists Retail Media Certification holders under "AAM Platform Certification" was not checked'. Add the programme-page sentence as a second, issuer-level source for the guideline basis. | The direct request for a 'Retail Media Certification' filter (S53) was not repeated. | RTL-A-F41 (evidence/verify/A.json) |
| C014 | Kroger Precision Marketing's home page states that it 'has earned the Trustworthy Accountability Group (TAG) Platinum certification, recognizing excellence in brand safety, ad fraud prevention, and malware protection', a company-level vendor claim that this review could not check against TAG's registry; separately, MRC's Amazon Sponsored Products letter of 8 July 2026 carries a 'Trustworthy Accountability Group Certified Against Fraud Addendum' stating that IVT 2.0 compliance 'is inclusive of certain Trustworthy Accountability Group (TAG) Certified Against Fraud requirements' (domain-list and | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | Amazon Sponsored Products letter p. 3 addendum as quoted (domain-list and data-center IP filtering). | supported with qualification | No change to the MRC half. The KPM half remains a vendor claim not checked against TAG's registry (8451.com and KPM pages refuse the fetch tool; not re-tried). | AAM's Assurance List has 'TAG Certified Against Fraud', 'TAG Brand Safety Certification' and 'TAG Certified Against Malvertising' filters; AAM audits TAG programmes, so a browser read of those filters (or tagtoday.net) could settle KPM's level. | RTL-A-F42 (evidence/verify/A.json) |
| C015 | DoorDash Ads is listed by MRC as 'DoorDash Ads (Pre-audit reviewed 1/2026, Audit not begun)' for 'Measurement and reporting of Impressions, Clicks and CTR - Desktop, Mobile Web and Mobile In-App', with 'Ad formats included in-scope: Sponsored Products and Sponsored Brands'; the same formats hold AAM's incrementality certification (RTL-A-F30), so the row carries two instruments at different stages for different metrics. | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | AAM part verified: the DoorDash post and AAM's 14 October 2026 assurance brief both name 'Sponsored Products and Sponsored Brands'. MRC under-review wording not opened by this verifier; no DoorDash accreditation decision appears in the five quarterly updates. | supported with qualification | Keep; MRC strings to be confirmed by the enumeration. | MRC's two list pages were not opened by this verifier: per the assignment, the main session is enumerating both pages in full in a browser on 3 November 2026, which settles the list strings. Consistency check done here: none of MRC's five quarterly updates (Q3 2025 to Q3 2026, all opened) records any accreditation decision (grant, continuation, denial) for Kroger, Kevel, DoorDash, Schwarz or Walma | RTL-A-F48 (evidence/verify/A.json) |
| C016 | data_tables.assurance_instruments (17 rows) | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | The seven MRC letter and IAS rows match the letters and the Q4 2025 update (service names, metrics, environments, standards). The IAB Europe rows match the certified list and the Moloco certificate (auditor ABC on both). The AAM row's 'standard_applied' does not match the DoorDash post (see RTL-A-F30). | supported with qualification | (1) AAM row, standard_applied: "industry guidelines, including the IAB and IAB Europe Guidelines for Incremental Measurement in Commerce Media (AAM post, 10 September 2026); AAM's programme page also names the ANA Retail Media Network Measurement Standards among the guidance the programme aligns with". (2) Amazon DSP row, stage: "continuing accreditation plus initial grants decided January 2026 (Retail, IMDb, APS, 3PX) and May 2026 (Alexa connected devices), consolidated in the letter of 29 May 2026". (3) MRC under-review rows: unverified here (list not opened). | AAM standard_applied cell | TABLE:assurance_instruments (evidence/verify/A.json) |
| C017 | data_tables.standards_timeline (38 rows) | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | Dates checked against the documents opened all match (draft comment close 'Until October 13, 2023' on the draft cover; final PDF created 26 January 2024; V1 in-store comment period 'September, October and November 2024'; V2 release 22 January 2026; transition 'until the end of July 2026'; letters' dates; AAM 29 October 2025 and 10 September 2026; Moloco 7 October 2026; MRC Q3 2026 16 October 2026). | supported with qualification | (1) Order: the row dated 2025-10-29 (AAM) sits before 2025-10-01 (MRC Epsilon) and 2025-10-09; re-sort. (2) Add rows: 2026-01 MRC Amazon DSP initial grants (Retail video and viewability; IMDb, APS, 3PX viewability; source MRC Q1 2026 update, 17 April 2026); 2026-05 MRC Amazon DSP grant for Alexa connected devices (Q2 2026 update); 2026-10-14 AAM Assurance Brief on incrementality. (3) The 2026-09-17 in-store row should note that the V2 draft deletes V1's 'Incremental Sales/Brand Lift' and 'Sales Lift and Brand Sales Lift' text (RTL-A-F19). (4) The 2026-10-23 close: the drafts say 'by 23rd Octob | Rows resting on S19, S23, S25, S28, S36, S39, S49, S50 and S34 were not re-opened. | TABLE:standards_timeline (evidence/verify/A.json) |
| C018 | data_tables.dimension10_assurance_record (89 rows) | Refute-mode verifier for evidence/research/A.json re-opened the cited source. | MRC-accredited, AAM and IAB Europe cells match the letters, the AAM incrementality filter (DoorDash only) and the three-entry IAB Europe list. The AAM 'not listed' cells rest on the absence of a retail media filter, but the list has an 'AAM Platform Certification' filter that was not checked. | supported with qualification | Bound the AAM 'not listed' note: 'no retail media filter exists; AAM Platform Certification holders not screened'. Under-review MRC cells await the enumeration. Kroger TAG cells stay vendor-claimed. | TABLE:dimension10_assurance_record (evidence/verify/A.json) | |
| C019 | The Ovative and Albertsons Media Collective paper (March 2026) is a retrospective sensitivity analysis of matching choices on 42 of The Collective's own onsite display campaigns on Albertsons website properties, with no randomised benchmark, which reports that 54 methodology variants per campaign produced highest-to-lowest iROAS gaps that 'averaged 6.5x (median 2.5x)' and that 83% of campaigns 'flipped from positive to negative depending on methodology' (full report, p. 22), hedged as 'could flip' in the executive summary (p. 4) and on p. 12 of the full report. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Full report p. 22 (Analysis Summary), verbatim: 'By varying matching methodologies, our analysis produced 54 distinct iROAS outcomes across 42 campaigns. Within individual campaigns, the gap between the highest and lowest iROAS averaged 6.5x (median 2.5x), and 83% of campaigns flipped from positive to negative depending on methodology (Exhibit 12).' Exhibit 12 is titled 'Distribution of Campaign Level % Difference Between Highest and Lowest iROAS Generated' and its note reads: 'Exhibit Note: All variations include no filtering. Variations include PSM (1:many) or Clustering with and without Bra | supported with qualification | The Ovative Group and Albertsons Media Collective paper (March 2026) is a retrospective sensitivity analysis of matching choices on 42 of The Collective's own onsite display campaigns on Albertsons website properties, with no randomised benchmark. Its four levers can yield 54 iROAS values per campaign, but its headline statistics come from Exhibit 12, which uses only the no-filter variations (PSM 1:many or clustering, with and without brand sales as a feature): within campaigns the gap between highest and lowest iROAS 'averaged 6.5x (median 2.5x)', and '83% of campaigns flipped from positive t | Evidence class commissioned_measurement presumes a commissioning relationship the paper does not document: there is no funding, compensation or competing-interest statement anywhere in the 34 pages; it is a party study co-authored by the operator's own staff and its consultant, funding unknown. Brett R. Gordon's bio (p. 33) names only Kellogg, while the Amazon paper of 11 August 2025 lists him wit | RTL-B-F01 (evidence/verify/B.json) |
| C020 | In the same 42-campaign analysis, each lever moved the average aggregate iROAS materially: pre-filtering to past brand purchasers cut it from $2.27 to $0.22 (category filter $2.20) while shrinking the test group to 17% of its unfiltered size; clustering gave $1.80 against $0.73 for PSM 1:many and $0.84 for PSM 1:1, with average standardised mean differences of 0.158 versus 0.013 and 0.012; dropping past brand sales as a PSM feature moved the average from $1.23 to -$0.14 (clustering stayed at $1.81); and observed-sales estimates averaged $1.56 against $0.97 under BSTS, a difference the authors | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Every figure is printed in the report: p. 16 'iROAS worsened substantially ($2.27 to $0.22) while a category sales filter remained more stable ($2.27 to $2.20)'; Exhibit 6 test-group index 17 / 84 / 100 and control 16 / 81 / 100, SMD 0.101 / 0.080 / 0.075; p. 18 '(Clustering: $1.80 vs. PSM 1:many: $0.73)'; Exhibit 7 SMD 0.158 / 0.013 / 0.012 and iROAS $1.80 / $0.73 / $0.84; p. 20 'When brand sales are not used for PSM (1:many), the average iROAS changes from $1.23 to -$0.14. When Clustering, omitting brand sales results in an average iROAS that holds steady at $1.81 (Exhibit 9).'; p. 21 'the a | supported with qualification | Append: the '~90%' is the mean of campaign/variation-level percentage differences, (iROAS - BSTS iROAS)/BSTS iROAS, over clustering and PSM 1:many variations only; it is not the gap between the two printed averages ($1.56 vs $0.97 is about 61%). | The Exhibit 9 column header says only 'Propensity Score Matching'; the 1:many specification comes from the p. 20 text. In the ovative_sensitivity table the two revenue-method rows carry '126 campaign-variations' each; the exhibit note's 'two methods' most naturally refers to the two matching methods (the reason given is 'Only considering PSM 1:Many'), so the 126 attribution is ambiguous and should | RTL-B-F02 (evidence/verify/B.json) |
| C021 | Amazon's paper summarises Gordon, Moakler and Zettelmeyer (2023) as 'a 2023 study of almost 2,000 representative campaigns at Meta' with observational errors 'ranging from 488% to 948%', whereas the arXiv v2 text of that study (4 October 2022) reports 663 treatment-control pairs from 563 experiments and median lifts of 29%, 18% and 5% (RCT) against 83%, 58%, 24% (DML) and 173%, 176%, 64% (SPSM); the published Marketing Science version was not opened, so the discrepancy is recorded, not resolved; Amazon's quotation that the IAB/MRC guidelines 'highly recommends using RCTs' matches the guideline | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Amazon (p. 3): 'For example, a 2023 study of almost 2,000 representative campaigns at Meta showed that sophisticated models using observational data led to errors in the estimated ad effects ranging from 488% to 948%, depending on the model and outcome type (Gordon et al., 2023).' Amazon's reference list maps 'Gordon et al., 2023' to 'Close Enough? A Large-Scale Exploration of Non-Experimental Approaches to Advertising Measurement. Marketing Science, 42(4):768–793. Working paper version available at https://arxiv.org/abs/2201.07055', and PIE separately to 'Gordon, B. R., Moakler, R., and Zette | contradicted | Amazon's paper (11 August 2025) attributes to Gordon, Moakler and Zettelmeyer (2023) 'a 2023 study of almost 2,000 representative campaigns at Meta' with errors 'ranging from 488% to 948%'. The error range matches the study's arXiv v2 exactly: 488% and 948% are the smallest and largest median absolute percentage errors in its Table 5 (DML and SPSM, mid-funnel, computed only for experiments with a positive RCT lift), a qualifier Amazon's sentence omits. The sample count does not match any opened text: arXiv v2 reports 563 experiments (663 treatment-control pairs), an earlier version 850 experim | The counterevidence_and_tensions entry ('median lifts that imply different relative errors') and the limitation ('the 488% to 948% figures could be relative errors computed on a different basis') should be withdrawn: the figures are printed in the arXiv v2 that B.json read. Only the sample-size discrepancy stands. | RTL-B-F06 (evidence/verify/B.json) |
| C022 | Instacart's engineering post (30 June 2023) documents a user-level randomised holdout with ghost ads for Sponsored Products only: users are 'randomly split' by a hash of (user_id, advertiser_id) into an 'ads' group and a 'holdout' group before the auction, the outcome is total (organic plus attributed) sales of the advertised products, and the metrics are total incremental sales, incremental sales lift percentage and incremental sales on ad spend; the post states no significance rule, minimum sample or interval and describes the system's measurement, not what an advertiser receives by default. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Checked against the session's verbatim browser-pane notes only (browser-notes-U.md, 'Instacart engineering post'; tech.instacart.com refuses the fetch tool, and the live page was not re-read here). All seven passages quoted in the finding appear verbatim in those notes, with author Jason Kim and date 30 June 2023, including 'Currently, continual ads incrementality measurement is limited to Instacart's Sponsored Products offering.' and 'Users are randomized based on hash of (user_id, advertiser_id)'. The notes also record 'We can measure ad effectiveness, often termed ad incrementality, using R | supported with qualification | Replace 'user-level randomised holdout' with 'randomised holdout assigned per user-advertiser pair (hash of user_id and advertiser_id), so one user can be in the holdout for one advertiser and in the ads group for another'. | measurement_record.assignment_unit 'user' should read 'user within advertiser (user_id x advertiser_id hash)'. | RTL-B-F07 (evidence/verify/B.json) |
| C023 | The Ghost Ads paper (Johnson, Lewis and Nubbemeyer, Journal of Marketing Research, vol. 54, 2017, pp. 867-884) could not be opened by any fetch route on 3 November 2026; the design it names is described by three opened sources: Instacart ('simulating which holdout impression would have won the auction'), the IAB/MRC guidelines ('ghost ad or traditional A/B tests can usually be configured to apply the same targeting and bidding logic to determine the populations that are measured for treatment and control'), and Gordon, Moakler and Zettelmeyer ('Facebook implements the same counterfactual as Go | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Semantic Scholar record (re-fetched): JMR volume 54, pages '867 - 884', 2017, authors Garrett A. Johnson, Randall A. Lewis, Elmar I. Nubbemeyer; abstract elided by the publisher. A working-paper version is open at conference.nber.org: 'Ghost Ads: Improving the Economics of Measuring Ad Effectiveness', 18 February 2016; abstract: 'We develop a methodology we call 'Ghost Ads,' which facilitates this comparison by identifying the control-group counterparts of the exposed consumers in a randomized experiment.' ... 'our implementation records more than 100 million predicted ghost ads per day. We de | supported with qualification | The JMR 2017 version could not be opened (Semantic Scholar gives the bibliographic record only); a working-paper version of 18 February 2016 (NBER conference copy) was opened: it describes ghost ads as identifying 'the control-group counterparts of the exposed consumers in a randomized experiment', implemented on Google's display network, and demonstrated on an online retailer's display retargeting campaign; the design is also described by Instacart, the IAB/MRC guidelines and Gordon, Moakler and Zettelmeyer, as quoted. | Source RTL-B-S07 and REG-RTL-006 say 'two authors were Google employees per secondary descriptions': the opened working paper lists one Google affiliation (Nubbemeyer) and Lewis at Netflix as of February 2016; it carries thanks to 'many Google employees and advertisers' and no funding statement. | RTL-B-F08 (evidence/verify/B.json) |
| C024 | Blake, Nosko and Tadelis (NBER WP 20171, May 2014; eBay) found that stopping brand-keyword paid search cost almost no traffic because natural search substituted, and that suspending all non-brand paid search in 68 US DMAs for 60 days (65 matched control DMAs; 142 control DMAs in total) showed 'The entire regime of paid search adds only 0.66 percent to sales', with an experimental ROI of -63% (95% CI [-124%, -3%]) against OLS estimates of 'over 4,100%' and 'over 1,400%'; the test DMAs were drawn from a random candidate pool and then split by a matching algorithm, so 'the test group is not a pur | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Numbers verbatim: p. 14 'The entire regime of paid search adds only 0.66 percent to sales.'; p. 7 'a ROI of over 4, 100% without time and geographic controls, and a ROI of over 1, 400% with such controls ... a ROI of −63%, with a 95% confidence interval of [−124%, −3%]'; p. 13 'the 68 test DMAs where advertising ceased, 65 matched control DMAs, and the remaining 77 control DMAs (yielding a total of 142 control DMAs)'. Body text p. 13: 'A purely random subsample of DMAs were chosen as candidates for the test. Next, candidate DMAs were divided into test and control DMAs using an algorithm that m | supported with qualification | Replace the last clause with: 'the 68 test DMAs were assigned by restricted randomisation: 133 of 210 DMAs were chosen as candidates purely at random, random 68/65 splits were drawn and those with very low historical correlation discarded, and a coin flip decided which group was switched off; the authors call the result not a purely random sample because of that balance screen, and the other 77 DMAs serve as additional controls.' | Assignment is miscoded. The measurement_record ('non-random (random candidate pool of DMAs, matched allocation to test and control)'), the study_design, the studies-table design cell and the counterevidence_and_tensions entry rest on the body-text label 'matched'; the paper's own appendix discloses random allocation with re-randomisation for balance and a coin flip. Code assignment: random (restri | RTL-B-F10 (evidence/verify/B.json) |
| C025 | Lewis and Rao (QJE 130(4), 2015, pp. 1941-1973) analysed 'Twenty-five large field experiments' with major US retailers and brokerages, 'most reaching millions of customers' and 'collectively representing $2.8 million in digital advertising expenditure', and argue (per the article page, read in part) that individual sales are so volatile relative to per-person ad cost that the median confidence interval on return on investment is more than 100 percentage points wide and informative experiments can require more than ten million person-weeks. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | The OUP page again rendered the abstract only as paraphrase (WebFetch) and returned 403 to curl. The verbatim abstract was read from a journal-index record (SUFE) and from the typeset article PDF mirrored at gwern.net (stamped 'Downloaded from qje.oxfordjournals.org'): 'Twenty-five large field experiments with major U.S. retailers and brokerages, most reaching millions of customers and collectively representing $2.8 million in digital advertising expenditure, reveal that measuring the returns to advertising is difficult. The median confidence interval on return on investment is over 100 percen | supported with qualification | Replace the paraphrase with the verbatim statements: 'The median confidence interval on return on investment is over 100 percentage points wide' and 'informative advertising experiments can easily require more than 10 million person-weeks'; the 25 RCTs were with retailers (19) and financial service firms (6), each a random holdout of targeted users from online display ads. | Evidence class independent_measurement is not supported: the experiments ran on Yahoo!'s network (the media seller) while the authors were Yahoo! employees ('our former employer'), many are 'taken from past work from Yahoo! Labs', and no funding statement appears. Code it consistently with Lewis and Reiley (RTL-B-F12): platform-employee research, funding not stated, independence unknown. The study | RTL-B-F11 (evidence/verify/B.json) |
| C026 | Lewis and Reiley (QME 12, 2014, pp. 235-266) ran 'A randomized experiment with 1.6 million customers' of an anonymous major retailer on Yahoo!, finding that the advertising 'profitably increases purchases by 5%' (point estimate 5.0%, standard error 2.3%), that '93% of the increase occurs in brick-and-mortar stores', that '78% of the increase derives from consumers who never click the ads', and that a cross-sectional observational estimator 'exhibits a large negative bias three times the estimated experimental effect'; Yahoo! 'provided financial and data assistance and guaranteed academic indep | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Springer page (read with curl and a cookie jar): abstract 'A randomized experiment with 1.6 million customers measures positive causal effects of online advertising for a major retailer. The advertising profitably increases purchases by 5%. 93% of the increase occurs in brick-and-mortar stores; 78% of the increase derives from consumers who never click the ads. ... the latter exhibits a large negative bias three times the estimated experimental effect.' Acknowledgments: 'Yahoo! Inc. provided financial and data assistance and guaranteed academic independence prior to our analysis so that the re | supported with qualification | Replace '(point estimate 5.0%, standard error 2.3%)' with '(the paper's percentage estimate E[Δ] = 5.0%, standard error 2.3%, is a log difference-in-differences between exposed and unexposed customers, which the authors place midway between the R$0.083 experimental and R$0.102 DID level estimates)'. | Sales are reported in 'R$' units on the readable page, not shown as US$; the measurement_record's 'in US$' should read 'in R$ as reported'. Eligibility should record that the retailer chose the customer subset by an undisclosed rule. Note 19 ('75,000 observations with nonzero purchases') was not re-checked. | RTL-B-F12 (evidence/verify/B.json) |
| C027 | Gordon, Moakler and Zettelmeyer (arXiv 2201.07055 v2, 4 October 2022; Marketing Science 2023) analysed 663 treatment-control pairs from 563 US Facebook experiments run November 2019 to March 2020 ('approximately 7.9 billion user-experiment observations and over 38 billion ad impressions'; median experiment 30 days and about 7.3 million users) and found that double/debiased machine learning and stratified propensity-score matching, with over 5,000 user-level features, could not recover the experimental effects: median RCT lifts of 29%, 18% and 5% by funnel against 83%, 58%, 24% (DML) and 173%, | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | All numbers verbatim (abstract; PDF pp. 2-3 and 11). Selection (PDF p. 11): 'The experiments we analyze are a random subset from the set of experiments started between November 1, 2019, and March 1, 2020, and had at least one million users in the test group. For each experiment, we selected all outcomes with at least 5,000 conversions in the test group.' Footnote 15: 'An earlier version of this paper, dated February 17, 2021, used 850 experiments that were selected using the same selection criteria.' PDF p. 44, 'Funding and Competing Interests': 'To be allowed to access the data required for t | supported with qualification | Claim stands; correct the record: funding and interests 'all three authors were Facebook/Meta employees for this paper (Gordon and Zettelmeyer part-time, as a condition of data access; Moakler full-time with Meta stock)'; eligibility 'a random subset of experiments started 1 November 2019 to 1 March 2020 with at least one million test-group users; outcomes with at least 5,000 test-group conversions'. | Contradicted labels: evidence_class independent_measurement, the study_design and measurement_record ('one Meta employee co-author; Facebook data; no funding statement'), source RTL-B-S13 ('no funding statement in the front matter beyond thanks'), the studies-table sponsor cell, and the counterevidence_and_tensions line saying 'the Facebook studies carry an explicit no-compensation statement'. The | RTL-B-F14 (evidence/verify/B.json) |
| C028 | Abhishek, Jerath and Sharma (Information Systems Research, 2025, DOI 10.1287/isre.2022.0188) report 'a large-scale field experiment at Flipkart, a leading online marketplace in India' on sponsored listings in the most salient result positions, finding category-dependent effects (sponsored listings performed worse than displaced organic listings in electronics and better in clothing and footwear) and that 'increasing the fraction of sponsored listings (by about 10% points) does not affect clicks and conversions in any product category'; it is a platform-side study of listing mix, not an adverti | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | SUFE index record (re-fetched with curl): 'In this paper, taking the perspective of the marketplace, we obtain insights into the impact of sponsored listings being shown at the most salient positions in the list of results. To do so, we analyze data from a large-scale field experiment at Flipkart, a leading online marketplace in India.' ... 'we find that increasing the fraction of sponsored listings (by about 10% points) does not affect clicks and conversions in any product category.' Institutions listed (not mapped to authors): 'University of California System; University of California Irvine | supported with qualification | Claim stands; correct the record: assignment 'unknown (the abstract says field experiment; randomisation and unit not disclosed in the opened record)'. | The measurement_record and study_design code assignment 'random (a field experiment ...)' from the label 'field experiment', which the brief forbids; and independent_measurement is assumed with funding unknown and Flipkart supplying the data. A search result (Columbia write-up, not opened; snippet only) describes an eight-day experiment with about 1.2 million app users in three groups; treat as a | RTL-B-F16 (evidence/verify/B.json) |
| C029 | This review found no peer-reviewed study published 2024-2026 that measures advertiser-level incremental sales lift from a US retail media network's advertising with a published design, in the ten searches and the pages screened by 3 November 2026; the only 2024-2026 peer-reviewed field experiment on marketplace sponsored listings found is the Flipkart study (RTL-B-F16), and the retail-media-specific effectiveness evidence opened is operator-authored (Ovative and Albertsons; Amazon; Instacart). | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Not a source claim. B.json's search log has ten WebSearch entries, but only three bear on this question (queries 2, 3 and 8); the others are bibliographic look-ups (Ovative, Einav, Sahni and Zhang, the Flipkart record, IAB/MRC, Close Enough) or the panel question. One verifier search on 3 November 2026 ('randomized field experiment sponsored product ads incremental sales e-commerce retailer' with Marketing Science, JMR or Management Science, 2024 or 2025) surfaced no peer-reviewed US retail-media lift study either; it surfaced Instacart's post, the Flipkart study's Columbia pages, and a trade- | supported with qualification | This review found no peer-reviewed study published 2024-2026 that measures advertiser-level incremental sales lift from a US retail media network's advertising with a published design, in three targeted searches (B.json queries 2, 3 and 8), one verifier search and the pages screened by 3 November 2026 (SSRN, Google Scholar and INFORMS were not searchable with the tools used); the only 2024-2026 peer-reviewed field experiment on marketplace sponsored listings found is the Flipkart study (India; platform-side); the retail-media-specific effectiveness evidence opened is operator-authored or opera | Bounded absence claim; correctly coded as inference. | RTL-B-F17 (evidence/verify/B.json) |
| C030 | This review found no public validation of a receipt or loyalty panel (Numerator, Circana, NCS) against point-of-sale data with a published design in the one search run and the pages screened by 3 November 2026; the only peer-reviewed panel-versus-retailer validation opened is Einav, Leibtag and Nevo on Nielsen Homescan (RTL-B-F15), and the Circana pages surfaced were marketing posts arguing that receipt panels need point-of-sale integration, not validation studies. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | USDA ERS Technical Bulletin 1942 (April 2016; Muth, Sweitzer, Brown, Capogrossi, Karns, Levin, Okrent, Siegel, Zhen), which B.json's search 4 surfaced as 'USDA receipt-capture work' but did not open, documents IRI's (now Circana's) Consumer Network household panel and InfoScan store data and 'compared certain components of the datasets with Government or commercial data sources'. It reports that 'total U.S. expenditures reported by households in the weighted IRI data are less than those in other nationally representative datasets' and that 'Certain households are less likely to report purchase | supported with qualification | Add after 'in the one search run and the pages screened': '; the nearest public independent assessment found is USDA ERS Technical Bulletin 1942 (April 2016), which compares the IRI (now Circana) Consumer Network household panel with government benchmarks, finding lower weighted expenditures than nationally representative datasets and inconsistent reporting by some household groups, but does not validate the panel against point-of-sale records'. | Low confidence remains right; the bound should name the vendor methodology pages not opened (numerator.com, circana.com, NIQ/NCS), as B.json's limitation already does. | RTL-B-F18 (evidence/verify/B.json) |
| C031 | RTL-B data_tables[studies] row 1: Ovative Group with Albertsons Media Collective (2026): 42 campaigns; gap averaged 6.5x (median 2.5x); 83% 'flipped' / 'could flip'; sponsor Ovative Group and Albertsons Companies (party study). | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | As RTL-B-F01: the 6.5x, 2.5x and 83% come from Exhibit 12's no-filter variations; all other cells (brand filter $2.27 to $0.22, PSM 1:many $0.73 vs clustering $1.80, BSTS $0.97 vs observed $1.56) match the exhibits. | supported with qualification | Effect cell: prefix the headline statistics with 'no-filter variations (Exhibit 12):'. | Sponsor cell 'party study' is right; 'commissioned' should not be implied (no funding statement). | RTL-B-DT1-R01 (evidence/verify/B.json) |
| C032 | RTL-B data_tables[studies] row 4: Instacart engineering post (2023): user-level randomised holdout with ghost ads, assigned before the auction by hash of (user_id, advertiser_id); Sponsored Products only. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Matches the session's browser-pane notes verbatim; live page not re-read. | supported with qualification | Design cell: 'randomised holdout per user-advertiser pair (hash of user_id, advertiser_id) with ghost ads'. | Checked against notes only (needs_browser). | RTL-B-DT1-R04 (evidence/verify/B.json) |
| C033 | RTL-B data_tables[studies] row 5: Johnson, Lewis and Nubbemeyer, Ghost Ads (2017): primary not read; 'not read' in n, outcome, effect, uncertainty. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | A February 2016 working-paper version is open (NBER conference copy): randomised experiment with ghost ads identifying control-group counterparts; implemented on Google's display network; application to an online retailer's display retargeting campaign; abstract 'lifted website visits by 17% and purchases by 11%'. | supported with qualification | Fill from the working paper, labelled as such: setting 'Google display network; an online sports and outdoors retailer's retargeting campaign'; effect 'visits +17%, purchases +11% (WP abstract; WP intro gives visitors +26.6%, transactions +12.0%, sales +10.8%)'; sponsor 'unknown; authors at Rochester, Netflix and Google (Feb 2016)'. | 'Google-affiliated authors' should read 'one Google-affiliated author at the working-paper date'. | RTL-B-DT1-R05 (evidence/verify/B.json) |
| C034 | RTL-B data_tables[studies] row 7: Blake, Nosko and Tadelis (2014 WP / 2015): 'DMA-level switch-off; random candidate pool then matched split (non-random allocation)'; sponsor 'eBay employed two authors'. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Appendix 'Randomization Procedure Detail': 133 of 210 DMAs chosen 'purely at random'; 'groups of 68 were drawn at random' with low-correlation draws discarded; 'Which group was actually turned off was decided by a coin flip.' Blake's affiliation is eBay Research Labs and the title page says Tadelis and Nosko were employed by eBay Research Labs. | contradicted | Design cell: 'DMA-level switch-off; restricted randomisation (random candidates, re-randomised 68/65 split, coin flip); difference-in-differences; 2SLS'. Sponsor cell: 'all three authors at eBay Research Labs (advertiser, not media seller)'. Tag: 'RCT (geo, re-randomised)'. | Allocation is random with a balance screen, not non-random. | RTL-B-DT1-R07 (evidence/verify/B.json) |
| C035 | RTL-B data_tables[studies] row 8: Lewis and Rao (2015): 25 randomised experiments; effects given as paraphrase; sponsor unknown; tag 'independent; power; partial read'. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Verbatim abstract now read: 'The median confidence interval on return on investment is over 100 percentage points wide'; 'informative advertising experiments can easily require more than 10 million person-weeks'; 19 retailers and 6 financial service firms; experiments on Yahoo! by authors who call Yahoo! 'our former employer'. | supported with qualification | Replace paraphrase with the verbatim statements; n cell '25 RCTs (19 retailers, 6 financial service firms)'; sponsor 'authors were Yahoo! employees when the experiments ran (Yahoo! sold the media); funding not stated'; tag: drop 'independent'. | Not independent by provenance; overlaps Lewis and Reiley's data. | RTL-B-DT1-R08 (evidence/verify/B.json) |
| C036 | RTL-B data_tables[studies] row 9: Lewis and Reiley (2014): 1.6 million customers; +5.0% purchases; SE 2.3%; 93% in stores; 78% non-clickers; Yahoo! funded with independence guaranteed. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | As RTL-B-F12: the 5.0% (2.3%) is the log difference-in-differences estimate E[Δ], placed 'midway between the R$0.083 experimental and the R$0.102 DID estimates'. | supported with qualification | Uncertainty cell: 'SE 2.3% on the log-DID percentage estimate'; units 'R$ as reported'. | 75,000 nonzero purchasers not re-checked. | RTL-B-DT1-R09 (evidence/verify/B.json) |
| C037 | RTL-B data_tables[studies] row 10: Sahni and Zhang (2024): randomised search-ad prominence experiment, 3.3 million US users; tag 'independent; search; abstract only'. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Abstract verbatim as under RTL-B-F13; no funding information on the record. | supported with qualification | Tag: replace 'independent' with 'independence unknown (funding not stated; unnamed search engine ran the experiment)'. | Funding unknown is not independence. | RTL-B-DT1-R10 (evidence/verify/B.json) |
| C038 | RTL-B data_tables[studies] row 11: Gordon, Moakler and Zettelmeyer (2022 arXiv / 2023): 563 experiments (663 pairs); sponsor 'one Meta employee co-author; Facebook data; no funding statement'; tag 'independent-academic with platform co-author'. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | 'Funding and Competing Interests: To be allowed to access the data required for this paper, Gordon and Zettelmeyer were part-time employees of Facebook with the title of Academic Researchers, employed for three hours per week. Moakler is an employee of Meta Platforms, Inc. and owns stock in the company.' | contradicted | Sponsor cell: 'all three authors Facebook/Meta employees for this paper (two part-time, as a condition of data access); Facebook data'. Tag: 'platform-employed authors; RCT vs observational'. | Not independent. | RTL-B-DT1-R11 (evidence/verify/B.json) |
| C039 | RTL-B data_tables[studies] row 13: Abhishek, Jerath and Sharma (2025): large-scale field experiment varying sponsored-listing share at Flipkart; tag 'independent; marketplace; abstract only'. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Abstract verbatim as under RTL-B-F16; randomisation and funding not stated in the opened record. | supported with qualification | Tag: replace 'independent' with 'independence unknown (funding not stated; Flipkart data)'; design: 'field experiment (allocation not disclosed in the abstract)'. | Assignment coded from a label. | RTL-B-DT1-R13 (evidence/verify/B.json) |
| C040 | RTL-B data_tables[ovative_sensitivity], all 16 rows: average aggregate iROAS, sample indices and SMD by lever, and the headline 6.5x / 2.5x / 83% / 54. | Refute-mode verifier for evidence/research/B.json re-opened the cited source. | Every value matches Exhibits 6, 7, 9, 10 and 12 and p. 22 (2.27 / 2.20 / 0.22; 100/100, 84/81, 17/16; SMD 0.075 / 0.080 / 0.101 / 0.158 / 0.013 / 0.012; new-to-brand 41% / 38% / 0%; unique matched controls 100 / 22 / 29; 1.80 / 0.73 / 0.84; 1.23 / -0.14 / 1.81 / 1.81; 1.56 / 0.97; 6.5x, 2.5x, 83%; 54 per campaign per executive summary p. 3). | supported with qualification | Revenue-method rows: '252 campaign-variations (126 clustering, 126 PSM 1:many), each estimated both ways' rather than '126 campaign-variations' each; the ~90% note: 'mean of campaign/variation-level % differences, not the gap between the two averages'; the 54 row: 'per campaign (executive summary p. 3); not reproducible from the option list on p. 14'. | RTL-B-DT2 (evidence/verify/B.json) | |
| C041 | A Circana 'household-level lift study' reported in the joint DoorDash-Circana releases of 2 September 2026 found a major beverage company's DoorDash media drove '$0.59 in incremental in-store sales per household', 'three times the Circana Beverage Benchmark', over an eight-week US campaign between 10 November 2025 and 4 January 2026, with no control-group construction, assignment or panel size disclosed. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | Circana release (S02), one paragraph: 'New findings from DoorDash measurement illustrate the platform's incremental value. An overlap analysis of buyers of measured brands showed that 76% purchased in store only, 13% purchased on DoorDash only, and 10% purchased through both channels.* Among the buyers who purchased on DoorDash, 56% bought exclusively on DoorDash. These results provide proof that DoorDash drives incremental demand by reaching shoppers whose brands are not otherwise reaching in store. Early measurement results underscore the value of that incremental reach. A major beverage com | supported with qualification | Restate: 'DoorDash's release of 2 September 2026 reports that a household-level lift study found a major beverage company's DoorDash media "drove $0.59 in incremental in-store sales per household", "three times the beverage category benchmark" (Circana's release of the same day: "three times the Circana Beverage Benchmark"). Neither release states who ran the study (Circana's release introduces it as "Early measurement results" within "New findings from DoorDash measurement"), its period, control construction, assignment or panel size; the eight-week US campaign of 10 November 2025 to 4 Januar | Reporting party and period over-attributed: the footnote asterisk sits on the overlap sentence only, and Circana's own release does not say Circana computed the $0.59. Evidence class vendor_assertion and assignment 'unknown' are correct; the 'Circana' attribution of the lift study is an inference and must be marked as such (this matters for H3, which reads the partnership as independent household- | RTL-C-F02 (evidence/verify/C.json) |
| C042 | Ipsos MMA's 'DoorDash-funded study' (13 October 2026) built MMMs for six QSR advertisers, tested five representations of DoorDash advertising and three prior settings, reported 'Campaign Orders performed best across all six advertisers'; DoorDash's release adds 'an average of 42% of DoorDash's modeled effect carried into the following week'. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | Ipsos MMA post (author Joe Lasala; datePublished 2026-10-13): 'We tested five ways to represent DoorDash advertising in the model: campaign orders, campaign sales, campaign spend, impressions, and clicks. Campaign Orders performed best across all six advertisers, which follows from how the platform works.'; 'Across advertisers, most of DoorDash's impact landed in the week the ads ran, with a small tail into the following week or two.'; 'We tested three versions: loose assumptions, neutral ones, and assumptions anchored in the results of controlled DoorDash experiments.'; 'And of course, the te | supported with qualification | Name all five inputs (the page lists them: 'campaign orders, campaign sales, campaign spend, impressions, and clicks'); replace 'no model outputs ... are published' (open_questions, reliability_notes) with 'no model outputs are on the open post; a white paper sits behind a contact form (not read)'. Add the tension: Ipsos MMA describes the following-week effect as 'a small tail into the following week or two', while DoorDash's release gives 'an average of 42% of DoorDash's modeled effect carried into the following week'. | Evidence class: commissioned_measurement is right for the findings on Ipsos MMA's own page (funder named by the measurer: DoorDash). The 42% appears only in DoorDash's release, so that number (and the vendor_evidence row carrying it) rests on the funder's account: code the 42% as vendor_assertion (DoorDash's report of a DoorDash-funded Ipsos MMA study) unless the white paper is read and confirms i | RTL-C-F03 (evidence/verify/C.json) |
| C043 | DoorDash's 24 March 2026 release launched Sales Lift Measurement for 'select advertisers via Managed Service for Sponsored Products campaigns', reporting 'over 30% average sales lift for CPG advertisers' (headline: 'Up to 30%') from 30 tests in H1 2025, 65% for Snacks from 19 test campaigns, and Kellanova iROAS '2.4X to 2.9X across three studies' in Q4 2025, measured against a 'non-ad-exposed holdout', assignment not stated. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | Title, schema headline and H1: 'DoorDash Ads Launches Sales Lift Measurement, Driving Up to 30% Incrementality Lift in Campaigns on Average for Brands'. Body: 'Ghost Ads methodology is widely recognized as the industry-leading standard for incrementality measurement. This new solution, based on Ghost Ads, equips brands with rigorous, experiment-based measurement designed to quantify true incremental sales and return on ad spend, moving beyond traditional attribution metrics.'; 'Sales lift studies on Sponsored Products show that DoorDash is a powerful driver of incremental growth for brands, de | supported with qualification | Headline/body conflict confirmed as the file states ('Up to 30% Incrementality Lift in Campaigns on Average' vs 'over 30% average sales lift'). Restate the design clause: 'the averages are footnoted as exposed customers versus "a non-ad-exposed holdout group"; the release says the solution is "based on Ghost Ads" and "experiment-based" and "builds on our previous incrementality measurement tool for Enterprise Restaurants"; assignment is not stated, and the footnote does not cover the Heritage Kellanova Q4 2025 studies, whose design is not stated.' Name the advertiser as 'Heritage Kellanova (no | Assignment 'unknown' is correct (a ghost-ads holdout is not thereby randomised on this page). measurement_record.counterfactual should read 'holdout ("non-ad-exposed holdout group"); solution "based on Ghost Ads"'. The vendor_evidence rows for Kellanova/Cheez-It wrongly say 'holdout (per footnote)'. | RTL-C-F04 (evidence/verify/C.json) |
| C044 | DoorDash's ghost-ads resource (4 October 2025) contrasts the ITT RCT ('eligible user accounts are randomly assigned') with ghost ads that 'identify equivalent populations', which 'now powers Sponsored Listings lift studies', claiming 92% less dilution and 35% better iROAS intervals, with no base period, experiment count, minimum sample or significance rule. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | Page shows 'Oct 4, 2025' (publishDate 2025-10-04; dateModified 2026-10-20). 'Prior to ghost ads, the industry standard for measuring lift was the intent-to-treat (ITT) randomized controlled trial (RCT). In these experiments, once an auction begins, eligible user accounts are randomly assigned to either a treatment group, which may see the ad, or a control group, which does not.'; 'In 2017, Google and Amazon pioneered the ghost ads technique, which allows us to accurately identify equivalent populations of user accounts in the control and treatment groups of a lift experiment'; 'For example, an | supported with qualification | Add that the 92% and 35% are attributed to 'our recent studies' (count and dates not stated) and that the page documents automated balance checks ('standard statistical tests run automatically to detect experimentation bias at the auction stage'; the 80/20 vs 80.3/19.7 example) but no significance rule for the lift estimate. | measurement_record.assignment is filled from the ITT predecessor's sentence. For the ghost-ads studies the page gives a treatment/control split (80/20) run by an 'experimentation platform' but does not restate allocation. Code assignment 'unknown (random assignment stated for the ITT design that ghost ads replaced; the ghost-ads section describes an 80/20 treatment/control split without stating al | RTL-C-F05 (evidence/verify/C.json) |
| C045 | Walmart Connect's undated measurement page states 'Our closed-loop measurement tracks sales across our site, app and in our stores' and ties 'ad exposure - across Walmart's digital properties ... to sales on Walmart's site and in stores', offering 'attribution, incrementality, brand impact, and campaign insights' through the self-serve Ad Center, with no window, lift, test, control or holdout wording. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | 'Our closed-loop measurement, powered by our first-party data, ties ad exposure - across Walmart's digital properties and offsite channels - to sales on Walmart's site and in stores. When a customer sees your ad and makes a related purchase at Walmart, we recognize and report it so you can track the results of your onsite and offsite campaigns in near real time.' Footnotes visible: '2. Walmart annual report papers, March 14, 2025.' '3. Walmart first-party data, February 2025.' | supported with qualification | Restore the words the ellipsis drops: 'across Walmart's digital properties and offsite channels'. Add 'When a customer sees your ad and makes a related purchase at Walmart, we recognize and report it ... in near real time' ('related purchase' undefined: halo scope unknown). | Absence wording (no window, lift, test, control, holdout) confirmed for this page. | RTL-C-F07 (evidence/verify/C.json) |
| C046 | Roundel's undated measurement page offers incrementality via 'synthetic control, matched market and randomized controlled trials', footnoted 'available for eligible campaigns', beside closed-loop measurement and a multi-touch attribution model, and reports two cases without design, period or spend: pet care '280%' lift in CTR and '10%' lift in ROAS; Spin Master '88%' increase in ROAS and '40%' new-to-brand guests. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | Measurement page footnote 3 in full: 'Incrementality measurement is available for eligible campaigns. Based on your campaign objectives, investment level, and measurement needs, your Roundel team can recommend the right approach for your business.' Each tile has a 'View case study' link. Pet tile -> roundel.com/case-studies/pets/: 'A robust search campaign and smart optimizations through Criteo Retail Media'; 'Build a 14-day click + 14-day view attribution reporting model'; 'April - May 2021: Campaigns with all products achieved 342% ROAS'; 'June 2021: Multiple product-specific campaigns impro | supported with qualification | Replace '(criteria unstated)' with 'eligibility set case by case on "campaign objectives, investment level, and measurement needs"'. For the cases: the pet tile's 280%/10% are not found on the case page it links to, which describes a 13-week search campaign via Criteo Retail Media in April-June 2021 reported on a 14-day click + 14-day view attribution model (ROAS 342% to 669%; 437% overall; $700k sales); record the tile figures as unsourced on the linked page. Spin Master: a Precision Plus pilot running Google Performance Max via Search Ads 360, Target internal data 2025, with '98 % increase i | Coding of the case ROAS figures as attributed is confirmed by the pet case's stated 14-day click + 14-day view window (priority 4 passed). Assignment for the offer correctly left unknown despite the RCT label (priority 3 passed). | RTL-C-F08 (evidence/verify/C.json) |
| C047 | Kroger Precision Marketing's post of 19 July 2024 (browser pane) describes a '96% loyalty card capture rate', 'test and control groups' matched 'across variables, such as pre-campaign spend, media interaction, loyalty and geography/division', and an attribution example over-attributed 'by over $2.8MM' against '$98K'. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | Notes (KPM blog, 19 July 2024): 'To determine the incremental impact of advertising, we use many IAB-recommended methods as well as test and control groups to create accurate comparisons across variables, such as pre-campaign spend, media interaction, loyalty and geography/division'; 'a sales attribution model that broadly measured total households led to over-attributing sales to campaign performance by over $2.8MM. By sifting the data to only include households that were exposed to viewable ads with a custom date range, the true amount was revealed at $98K.' FAQ: 'Our incrementality and clos | supported with qualification | Drop 'matched': the post says test and control groups are used 'to create accurate comparisons across variables'; it does not say the groups are matched. measurement_record.counterfactual: 'test and control groups (construction not stated; "comparisons across variables")'. In counterevidence_and_tensions, restate the example as 'over-attributed by over $2.8MM relative to a filtered figure of $98K that Kroger calls "the true amount"' (the $2.8MM is the excess, not the broad total). | Assignment 'unknown' correct. The $98K is a filtered attributed figure that the seller calls 'true'; the file correctly does not treat it as a counterfactual. | RTL-C-F09 (evidence/verify/C.json) |
| C048 | Fetch's post of 8 June 2023 says Fetch 'enlisted KPMG' and that KPMG found 'a uniquely clear view of campaign impact' through RCTs in which Fetch 'randomly assigns users', withholds offers from control and uses partner-level and offer-level holdouts; the KPMG report, standard, period and payer are not on the page. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | Dated June 8, 2023. 'As part of that process, the company enlisted KPMG, a global leader in data insights and professional services, to provide an independent third-party audit of its capabilities.'; 'Digging into the full methodology and evaluating the outcomes of Fetch offers and campaigns, KPMG was able to exactly replicate the results that Fetch delivers for partner brands.'; 'Because Fetch has billions of receipts from millions of users, the platform can ensure that the randomized control and test groups exhibit the same behavior - similar purchase patterns, brand spend and category spend | supported with qualification | Add: Fetch calls the engagement 'an independent third-party audit' and says KPMG 'was able to exactly replicate the results that Fetch delivers for partner brands' (no report, standard or period given). Fix measurement_record: the incremental-return divisor is 'the dollar amount that's been invoiced to the partner' (not cut off); balance claim: 'the randomized control and test groups exhibit the same behavior ... as verified by transaction data' (no statistics shown). | vendor_assertion is right: 'independent' is Fetch's word for a review it commissioned ('enlisted'); payer unstated, so independence is unknown. Assignment 'random' is disclosed by the design's own description ('randomly assigns users'), so the field is properly filled. | RTL-C-F11 (evidence/verify/C.json) |
| C049 | Chase's undated Solo Stove case: '$300K+ in incremental sales (46% incrementality for new customers)', '$750K+', '94% new customer transactions', 2,098 new customers, '4.7MM impressions', Q4 2023, Chase Sapphire cardholders, no control or holdout. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | 'Published Apr 3, 2024' (datePublished 2024-04-03). Figures verbatim. Also: 'the team crafted a campaign offer construct with a minimum spending requirement to ensure customers would make higher-ticket purchases'. | supported with qualification | Replace 'undated' with 'published 3 April 2024' in the claim and in RTL-C-S16.pub_date; add the minimum-spend offer construct (amount not stated). | Numbers all confirmed; only the date is wrong. | RTL-C-F14 (evidence/verify/C.json) |
| C050 | Albertsons' 6 January 2026 release: in-store incrementality 'comparing sales performance in test stores exposed to ads against a rigorously selected set of control stores', 'nearly 60 variables', 'a statistically validated measure of true lift'; Mondelez Sargento Cheese Bakes 'significant sales lift' without value, allocation, dates or interval. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | 'Unlike traditional attribution models that assign credit to purchases that may have occurred anyway, Albertsons Media Collective's matched market framework measures the true incremental impact of in-store media ... comparing sales performance in test stores exposed to ads against a rigorously selected set of control stores with no media exposure. By estimating the counterfactual, what would have happened without advertising, the model delivers a statistically validated measure of true lift.'; 'Mondelez integrated in-store digital screens with high-engagement onsite and offsite media'; 'After | supported with qualification | Add the release's own design term 'matched market framework' (used four times) to source_term, and 'control stores with no media exposure'. Add that the Sargento campaign combined in-store screens with onsite and offsite media, so the 'significant sales lift' is credited to 'the addition of in-store screens' inside an omnichannel campaign. | Assignment 'unknown' correct: a matched-market label does not establish allocation (priority 3 passed). | RTL-C-F19 (evidence/verify/C.json) |
| C051 | Summary: 'The "incremental" figures (DoorDash's $0.59 per household and "over 30% average sales lift", Albertsons' $7.45 iROAS, Citi's "up to 5x", Chase's $509K and $300K+) carry no stated counterfactual construction, interval or sample.' | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | DoorDash's footnote: '* Incremental sales are measured by comparing a brand's sales among ad-exposed customers versus a non-ad-exposed holdout group. Data sourced from DoorDash's 30 Sponsored Product Lift Measurement tests conducted in H1 2025'; body: 'This new solution, based on Ghost Ads'. | supported with qualification | Take DoorDash's 30% out of that list, or restate: 'DoorDash's "over 30% average sales lift" names a non-ad-exposed holdout group, 30 tests and a ghost-ads basis but no assignment or interval; the others carry no stated counterfactual construction, interval or sample.' Also correct the summary's DoorDash design screen ('holdout only in the footnote') by adding 'based on Ghost Ads' in the body. | SUMMARY (evidence/verify/C.json) | |
| C052 | counterevidence_and_tensions: 'Kroger Precision Marketing's own example shows an attribution filter choice moving one campaign's attributed sales from over $2.8MM to $98K'. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | 'led to over-attributing sales to campaign performance by over $2.8MM ... the true amount was revealed at $98K.' | supported with qualification | Restate: 'an attribution filter choice over-attributed one campaign's sales by over $2.8MM relative to the $98K Kroger calls "the true amount" after restricting to viewable-exposed households in a custom date range'. | TENSION-Kroger (evidence/verify/C.json) | |
| C053 | DoorDash with Circana: $0.59 incremental in-store sales per household; 3x Circana Beverage Benchmark; period 2025-11-10 to 2026-01-04; design "household-level lift study (Circana)". | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | The eight-week footnote is attached to the overlap analysis only; Circana's release calls the $0.59 'Early measurement results' within 'New findings from DoorDash measurement'. | supported with qualification | Set period to 'not stated (the eight-week footnote attaches to the overlap analysis)'; design to "'household-level lift study' (DoorDash's term; party that ran it not stated)"; S02's wording is 'incremental sales per household', S01's 'incremental in-store sales per household'. | See F02. | TABLE:vendor_evidence row 2-3 (evidence/verify/C.json) |
| C054 | 42% of modelled effect carried into the following week; sponsor DoorDash; evidence_class commissioned_measurement; source RTL-C-S01. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | 42% appears only in DoorDash's release; Ipsos MMA's post: 'most of DoorDash's impact landed in the week the ads ran, with a small tail into the following week or two'. | supported with qualification | Code evidence_class vendor_assertion (DoorDash's account of a DoorDash-funded Ipsos MMA study) unless the gated white paper confirms the figure; add the Ipsos 'small tail' wording as a tension. | evidence_class: the number rests only on the funder's release. | TABLE:vendor_evidence row 6 (evidence/verify/C.json) |
| C055 | Heritage Kellanova iROAS 2.4x-2.9x (three studies) and Cheez-It nearly 46% lift, Q4 2025, control 'holdout (per footnote)'. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | Values and Q4 2025 verbatim. The footnote marker '*' is on the 30%/65% sentence; the Kellanova paragraph carries no footnote. | supported with qualification | Set control to 'not stated (the holdout footnote covers the 30% and 65% averages; the solution is "based on Ghost Ads")'; name the advertiser 'Heritage Kellanova (now part of Mars Snacking)'. | TABLE:vendor_evidence row 9-10 (evidence/verify/C.json) | |
| C056 | Ghost ads: 92% less experimentation dilution; 35% tighter iROAS confidence intervals; control "random treatment/control split of user accounts with ghost ads"; period not stated. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | 'From our recent studies, we have found that ghost ads have reduced experimentation dilution by 92% and improved iROAS confidence intervals by 35%.' Random assignment is stated for the ITT predecessor; the ghost-ads example gives an 80/20 split without stating allocation. | supported with qualification | Control: 'treatment/control split of user accounts with ghost ads (allocation not restated for ghost-ads studies)'; sample/period: "'recent studies' (count and dates not stated)". | See F05 label issue. | TABLE:vendor_evidence row 11-12 (evidence/verify/C.json) |
| C057 | Roundel pet care case: 280% lift in CTR; 10% lift in ROAS; design and period not stated. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | Tile figures verbatim on the measurement page. The case page the tile links to (premium pet food brand, Criteo Retail Media, April-June 2021, 13 weeks, '14-day click + 14-day view attribution reporting model') reports ROAS 342% to 669%, '437% overall ROAS' and '$700k in sales'; neither 280% nor a 10% ROAS lift appears there. | supported with qualification | Add a note: 'tile figures not found on the linked case page'; if the rows stay, add the case page's attribution window (14-day click + 14-day view) and campaign period (2021) as the linked case's parameters, flagged as not reconciled with the tile. | TABLE:vendor_evidence row 14-15 (evidence/verify/C.json) | |
| C058 | Roundel Spin Master: 88% increase in ROAS; 40% new-to-brand guests; period not stated. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | Case page: Precision Plus pilot with Google, 'Performance Max campaign via Search Ads 360'; '98 % increase in CTR', '22 % decrease in CPC', '88 % increase in ROAS' with source 'Target internal data (2025)'; 'converted over 40% of new-to-brand Target guests'. | supported with qualification | Period: '2025 (Target internal data, per the case page)'; design: 'Precision Plus pilot (Google Performance Max via Search Ads 360); ROAS attributed, window not stated'; note the case page's wording of the 40%; add rows for the 98% CTR increase and 22% CPC decrease if every number is to be rowed. | TABLE:vendor_evidence row 16-17 (evidence/verify/C.json) | |
| C059 | Roundel incrementality offer, sample "'eligible campaigns'". | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | Footnote 3: 'Based on your campaign objectives, investment level, and measurement needs, your Roundel team can recommend the right approach for your business.' | supported with qualification | Sample/eligibility: "'eligible campaigns', chosen on 'campaign objectives, investment level, and measurement needs'". | TABLE:vendor_evidence row 18 (evidence/verify/C.json) | |
| C060 | Kroger: over-attributed sales (all households) vs viewable-exposed households, over $2.8MM vs $98K. | Refute-mode verifier for evidence/research/C.json re-opened the cited source. | 'over-attributing sales ... by over $2.8MM ... the true amount was revealed at $98K.' | supported with qualification | Metric: 'over-attribution (excess) under an all-households model vs the filtered figure Kroger calls "the true amount" (USD)'. | Checked against browser-pane notes. | TABLE:vendor_evidence row 19 (evidence/verify/C.json) |
| C061 | The IAB/PwC Internet Advertising Revenue Report for full-year 2025 (April 2026) reports US 'Commerce media advertising revenue' of US$37.5 billion (2022), US$43.7 billion (2023, +16.3%), US$53.7 billion (2024, +23.0%) and US$63.4 billion (2025, +18.0%), full-year survey actuals with no on-site, off-site or in-store split. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | p. 27: 'Commerce media advertising revenue, full-year results (2022-2025) ($ billions) Source: IAB / PwC Internet Ad Revenue Report, FY 2025. 2022 2023 2024 2025 $37.5 $43.7 $53.7 $63.4 16.3% 23.0% 18.0%'. p. 35 (Survey scope and methodology), PwC 'Acquires supplemental data through the use of publicly disclosed information;' and 'Identifies non-participating companies and applies a conservative revenue estimate based on available public sources;'. | supported with qualification | The IAB/PwC Internet Advertising Revenue Report for full-year 2025 (April 2026) reports US 'Commerce media advertising revenue' of US$37.5 billion (2022), US$43.7 billion (2023, +16.3%), US$53.7 billion (2024, +23.0%) and US$63.4 billion (2025, +18.0%): full-year figures that PwC compiles from a survey of companies, publicly disclosed information and its own estimates for non-participating companies (mix not disclosed), with no on-site, off-site or in-store split. Edit the measurement_record: observed_vs_modelled = 'survey responses plus PwC estimates for non-participants'; data_provenance = ' | The measurement_record codes the series as an 'observed accounting amount' from 'company-reported revenues'; the report's own method (p. 35) includes estimates for non-participants, so part of each value is modelled. | RTL-D-F01 (evidence/verify/D.json) |
| C062 | Between the FY2024 report (April 2025) and the FY2025 report (April 2026) IAB/PwC relabelled the same series: the FY2024 PDF charts 'Retail media network advertising revenue, full-year results (2022 - 2024)' at US$37.5, 43.7 and 53.7 billion with the survey definition 'Retail Media Network ads (Retail Media / Shopper)', while the FY2025 PDF charts 'Commerce media advertising revenue' with identical 2022-2024 values and a survey definition whose only change is 'Commerce Media Network ads' in place of 'Retail Media Network ads'; neither report explains the change of label. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | FY2024 p. 22, section heading 'Commerce media', chart 'Retail media network advertising revenue, full-year results (2022 – 2024) ($ billions)' with '16.3% 23.0% 2022 2023 2024 $37.5 $43.7 $53.7'. FY2024 'Survey definitions' begins on p. 26; the entry sits on p. 27: 'Retail Media Network ads (Retail Media / Shopper)' – 'Retail Media Network ads that allow a retailer to promote/sponsor their own white-label products and/ or the products they sell from their vendors directly on their own eCommerce site or on other sites.' FY2025 p. 32: term 'Commerce Media Network ads' with the same sentence begi | supported with qualification | Between the FY2024 report (April 2025) and the FY2025 report (April 2026) IAB/PwC renamed the series: the FY2024 PDF, under a section already headed 'Commerce media', charts 'Retail media network advertising revenue, full-year results (2022 – 2024)' at US$37.5, 43.7 and 53.7 billion (16.3%, 23.0%) with the survey term 'Retail Media Network ads (Retail Media / Shopper)' (p. 27); the FY2025 PDF charts 'Commerce media advertising revenue' with identical 2022-2024 values and the survey term 'Commerce Media Network ads' (p. 32), whose definition sentence is unchanged apart from its opening words, t | confidence_reason says both PDFs were 'downloaded from the publisher'; the FY2025 copy is served by IAB Hong Kong (iabhongkong.com), not by iab.com or PwC. The text is identical in substance to the publisher's report as far as can be told, but the provenance should read 'IAB Hong Kong-hosted copy'. | RTL-D-F03 (evidence/verify/D.json) |
| C063 | The FY2025 IAB/PwC report gives no 2026 figure for commerce media; its only forward statement is qualitative and attributed to a cited third party: 'Commerce media ad spend is expected to grow 30% faster than the overall advertising market, driven by expanding off-site and in-store opportunities.' | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | p. 15: 'Commerce media ad spend is expected to grow 30% faster than the overall advertising market, driven by expanding off-site and in-store opportunities.32' Endnote 32 (p. 42): 'Interactive Advertising Bureau. (2026). 2026 Outlook Study: A Snapshot of U.S. Ad Spend, Opportunities, and Strategies for Growth. www.iab.com'. Other 2026 statements in the report concern video/CTV ('Our research also projects an increase in revenues for both digital video (OLV) and in connected TV/ over-the-top video (CTV/OTT) in 2026', p. 21), social (p. 26) and the creator economy (p. 14); none gives a commerce | supported with qualification | The FY2025 IAB/PwC report gives no 2026 figure for commerce media; its only forward statement on it is qualitative, 'Commerce media ad spend is expected to grow 30% faster than the overall advertising market, driven by expanding off-site and in-store opportunities' (p. 15), sourced in endnote 32 to the IAB's own '2026 Outlook Study: A Snapshot of U.S. Ad Spend, Opportunities, and Strategies for Growth' (a buyer-side outlook survey by the report's commissioning publisher, not a third party). | 'attributed to a cited third party' is wrong: the cited source is IAB, the trade body that commissions the report. | RTL-D-F04 (evidence/verify/D.json) |
| C064 | Walmart's advertising figure exists only in its earnings releases as a global approximation: 'Global advertising business grew approximately 28% to reach $3.4 billion' (fiscal 2024, release of 20 February 2024), 'grew 27% to reach $4.4 billion' (fiscal 2025, 20 February 2025) and 'grew 46% to nearly $6.4 billion, including VIZIO' (fiscal 2026, 19 February 2026), each with the same footnote, while Walmart Connect in the US is given only as a growth rate (22%, 24% and 41% for the three fourth quarters). | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | FY24 release (dateline 'Feb 20, 2024'): 'Global advertising business2 grew approximately 28% to reach $3.4 billion'; Q4: 'grew approximately 33%, including 22% for Walmart Connect in the U.S.'; footnote 2 'Our global advertising business is recorded in either net sales or as a reduction to cost of sales' 'depending on the nature of the advertising arrangement.' FY25 release ('February 20, 2025'): 'Global advertising business3 grew 27% to reach $4.4 billion'; Q4 'grew 29%, including 24% for Walmart Connect in the U.S.'; footnote 3 same wording. FY26 release ('February 19, 2026'): 'Global advert | supported with qualification | Among Walmart's EDGAR filings the advertising amount is given in the earnings-release exhibits (8-K exhibit 99.1): 'grew approximately 28% to reach $3.4 billion' (fiscal 2024, 20 February 2024), 'grew 27% to reach $4.4 billion' (fiscal 2025, 20 February 2025) and 'grew 46% to nearly $6.4 billion, including VIZIO' (fiscal 2026, 19 February 2026), each footnoted 'recorded in either net sales or as a reduction to cost of sales, depending on the nature of the advertising arrangement'; Walmart Connect in the US is given only as fourth-quarter growth (22%, 24%, 41%). The phrase 'global advertising b | Calling all three amounts 'approximations' is the source's wording for two of them only; the denominators table's FY2025 unit 'approximate' should read 'rounded ($4.4 billion)'. | RTL-D-F07 (evidence/verify/D.json) |
| C065 | Maplebear Inc. (Instacart) reports audited 'Advertising and other' revenue of US$871 million (2023), US$958 million (2024) and US$1,065 million (2025); its accounting policy describes the line as advertising services that 'include Sponsored Product ads, display ads, coupons, and a variety of other online advertising services' plus software subscription services and other immaterial revenue, recorded gross or net depending on control. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | R12: 'The advertising services include Sponsored Product ads, display ads, coupons, and a variety of other online advertising' 'services.'; 'For advertising arrangements that involve third parties, the Company records advertising revenue on a gross or net basis' 'based on whether it acts as a principal or agent in the transaction, which is assessed on a contract by contract basis.'; 'The Company also offers software subscription services to certain partners that enhance the omnichannel shopping experience' 'and generates an amount of other revenue from software subscriptions and other goods an | supported with qualification | Maplebear Inc. (Instacart) reports audited 'Advertising and other' revenue of US$871 million (2023), US$958 million (2024) and US$1,065 million (2025), company-wide; its accounting policy says the advertising services 'include Sponsored Product ads, display ads, coupons, and a variety of other online advertising services', that third-party arrangements are recorded gross or net 'based on whether it acts as a principal or agent in the transaction, which is assessed on a contract by contract basis', and that other revenue from software subscriptions and other goods and services is 'not material' | scope_geo 'US' is not established: the line is company-wide and the 10-K mentions Canada (full-text search hit); no geographic revenue split was read. | RTL-D-F08 (evidence/verify/D.json) |
| C066 | Uber discloses no advertising revenue line: its Q4 2025 earnings release (4 February 2026) contains no sentence on advertising, and the only figures are vendor press-page run-rate statements ('surpassed $1.5 billion in annual revenue run rate (Q1'25)'; 'US$900m revenue run rate in Q4 2023'), which are run rates, not annual revenue, and span Uber Eats and rides advertising. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | Release (dateline 'SAN FRANCISCO – February 4, 2026'): no sentence with 'advertising', 'ads', 'Uber Advertising' or 'ad revenue'. Full-text search of Uber's February 2026 filings for 'advertising' returns one hit, the FY2025 10-K (uber-20251231.htm, filed 2026-02-13), and not the release. Press page: 'Since then, Uber Advertising has surpassed $1.5 billion in annual revenue run rate (Q1'25)' (June 17, 2025 item); 'reaching a US$900m revenue run rate in Q4 2023' (March 22, 2024 item); 'exceeding $500 million annual revenue run rate' (February 9, 2022 item). | supported with qualification | Uber discloses no advertising revenue amount in the documents opened: its Q4 2025 earnings release (4 February 2026) contains no sentence on advertising (confirmed by EDGAR full-text search), its fiscal 2025 10-K mentions advertising but was not read, and the figures found are run-rate statements on Uber's advertising press page ('exceeding $500 million annual revenue run rate', February 2022; 'US$900m revenue run rate in Q4 2023'; 'surpassed $1.5 billion in annual revenue run rate (Q1'25)'), which annualise a quarter and span Uber Eats and rides advertising. | 'the only figures' is an unbounded absence claim; bound it to the release and the press page, and name the unread 10-K. | RTL-D-F14 (evidence/verify/D.json) |
| C067 | DoorDash discloses no advertising revenue line in the documents opened: its Q4 2025 shareholder letter (8-K exhibit 99.2, 18 February 2026) contains no sentence mentioning advertising, 'Ads', ad revenue or an ad run rate, and no DoorDash filing or letter opened by workstreams D or U carries an advertising amount. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | Letter: none of 'advertising', 'ads', 'DoorDash Ads', 'ad revenue' or 'run rate' appears (fetch read); full-text search for 'advertising' and for 'advertisers' over DoorDash's 10-25 February 2026 filings does not return the letter either, but returns the Q4 2025 press release (EX-99.1) and the FY2025 10-K (dash-20251231.htm, filed 2026-02-18). The press release (fetch rendering, fragments): '...increased merchant value by launching new services such as reservations for restaurants and Smart Campaigns for ads'; 'The Y/Y increase in GAAP sales and marketing expense in Q4 2025 was driven primaril | supported with qualification | DoorDash discloses no advertising revenue amount in the documents opened: its Q4 2025 shareholder letter (8-K exhibit 99.2, 18 February 2026) does not mention advertising, and the Q4 2025 press release (exhibit 99.1) mentions 'advertising revenue' only as a driver of growth in accounts receivable, with no amount; the fiscal 2025 10-K (filed 18 February 2026) mentions advertising but was not read. | Edit the denominators table's DoorDash row ('no advertising amount disclosed in the documents opened') to name the press release and the unread 10-K. | RTL-D-F15 (evidence/verify/D.json) |
| C068 | eMarketer's US retail media series is gated (403 to every tool; forecasts behind PRO+); the only eMarketer forecast sentences read, in two free articles opened in the browser pane, are 'Financial media network ad spending will exceed $3 billion by 2028, per our forecast', growing 'at an explosive 77.3% compounded annually between 2023 and 2028 versus just 21.2% for the similarly sized travel media network segment' (23 September 2026), and that Amazon's ad business is expected 'to more than double from 2024 through 2028' (10 September 2026), each on an undisclosed definition. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | Notes: 'Financial media network ad spending will exceed $3 billion by 2028, per our forecast'; 'grow at an explosive 77.3% compounded annually between 2023 and 2028—versus just 21.2% for the similarly sized travel media network segment.' (23 September 2026, Katherine Smith); Amazon's ad business 'we expect to more than double from 2024 through 2028' (10 September 2026, Rachel Wolff and Marisa Jones). | supported with qualification | eMarketer's US retail media series page returned HTTP 403 to the fetch tool (one attempt) and its forecast tables were not read; the only eMarketer forecast sentences read, in two free articles opened in the browser pane, are ... (sentences as in the claim, with the em dash restored before 'versus'). | '403 to every tool' overstates the record: the free emarketer.com articles opened in the browser pane; only the series page is recorded as 403 (RTL-U-S247). 'Forecasts behind PRO+' is not in the notes or the U record and was not checked. | RTL-D-F16 (evidence/verify/D.json) |
| C069 | Kantar's H1 2026 interim results show Numerator divisional gross revenue of US$332.2 million for the six months to 30 June 2026 against US$333.4 million (down 0.4%), 'with 9.3% growth in Numerator North America offset by a decline in Numerator International', after a 1 January 2026 reorganisation that combined the former Worldpanel and Numerator cash-generating units into 'the Numerator CGU'. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | p. 3 table: 'Numerator 332.2 333.4 (1.2) (0.4%)'. p. 4: 'Numerator revenues decreased by 0.4%, with 9.3% growth in Numerator North America offset by a decline in Numerator International.' p. 3: 'From 1 January 2026, the Insights and Profiles divisions, which had been reported as separate categories prior to that date, were combined into one division named Kantar.' Notes: 'During the period, the Group revised its CGU structure following an internal reorganisation that resulted in four CGUs being consolidated into two larger, operationally integrated businesses. The reorganisation was completed | supported with qualification | Kantar's H1 2026 interim results (reporting entity Kantar Global Holdings S.à r.l.; ultimate controlling party Bain Capital Europe Fund V, SCSp) show Numerator divisional Gross Revenue of US$332.2 million for the six months to 30 June 2026 against a reclassified US$333.4 million (down 0.4%), 'with 9.3% growth in Numerator North America offset by a decline in Numerator International'; during the period the group combined the former Worldpanel and Numerator cash-generating units into 'the Numerator CGU' (the 1 January 2026 date in the report applies to the combination of the Insights and Profile | The 1 January 2026 date is attached to the wrong reorganisation; the Worldpanel-Numerator CGU combination is dated only 'During the period'. The same correction applies to the limitation ('includes the former Worldpanel business from 1 January 2026') and to the consolidation_events row of 2026-01-01. | RTL-D-F19 (evidence/verify/D.json) |
| C070 | Circana's formation is settled from its own posts only in part: IRI and The NPD Group completed their merger on 1 August 2022, while the rebrand post's body could not be read, so the Circana rename date stays reported (6 March 2023 on a circana.com snippet; 7 March 2023 on a Business Wire stamp), and the NCSolutions purchase is settled only at agreement (26 August 2024), its completion dated 2 June 2025 solely by a page title. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | Merger post: dateline 'CHICAGO and PORT WASHINGTON, NY, Aug. 1, 2022'; 'have completed their merger'; 'A unified name and brand for the combined company will be announced at a later date.'; 'H&F, a premier global private equity firm, is the majority investor in the combined company.' Hellman & Friedman's post of the rebrand release: headline 'IRI and NPD Rebrand as Circana, the Leading Advisor on the Complexity of Consumer Behavior', date 'March 7th, 2023', dateline 'Chicago, IL', first sentence 'Following their merger in August 2022, Information Resources, Inc. (“IRI”) and The NPD Group (“NPD | supported with qualification | IRI and The NPD Group completed their merger on 1 August 2022 (Circana post); the Circana name was announced in a release dated 7 March 2023 on the website of Hellman & Friedman, the combined company's majority investor (a party, not the issuer), matching the Business Wire URL stamp, while Circana's own post remains unread and a circana.com snippet's 6 March 2023 is unexplained; the NCSolutions purchase is settled only at agreement (Nielsen release, 26 August 2024), its completion on 2 June 2025 resting on a page title and snippet. | The 7 March 2023 date can move from 'reported' to 'party release (H&F)'; the issuer's own dateline still needs a browser. | RTL-D-F20 (evidence/verify/D.json) |
| C071 | Infillion's two 2026 acquisitions are announced, not shown closed: Catalina ('today announced the acquisition of Catalina', 23 February 2026, no closing or price wording) and Foursquare ('Infillion has announced the acquisition of Foursquare', 18 September 2026, with 'Several members of Foursquare's leadership team will join Infillion after the transaction closes' and no price or expected-close date); the Infillion blog listing carried no completion post for either by 3 November 2026. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | Catalina release: headline 'Infillion Acquires Catalina, One of the World's Largest Sources of Deterministic Purchase Data'; dateline 'NEW YORK, NY and ST. PETERSBURG, FL, February 23, 2026'; 'today announced the acquisition of Catalina, the pioneer in deterministic purchase intelligence'; no closing date, 'expected to close', completion wording, price or terms. Foursquare post: headline 'Infillion to Acquire Foursquare, Closing the Loop Between Advertising and Real-World Results'; dateline 'NEW YORK, NY, September 18, 2026'; 'Several members of Foursquare's leadership team will join Infillion | supported with qualification | Infillion's two 2026 acquisitions carry different wording and neither states a closing date: the Catalina release of 23 February 2026 is headed 'Infillion Acquires Catalina' and says 'today announced the acquisition of Catalina' (present-tense acquisition wording, no closing date, conditions or price), and Infillion's own boilerplate of 18 September 2026 lists Catalina among the technologies it was 'formed by unifying'; the Foursquare release of 18 September 2026 is headed 'Infillion to Acquire Foursquare' and says leadership 'will join Infillion after the transaction closes', so that deal was | Classifying Catalina as 'announced (closing unstated)' alongside Foursquare's 'pending close' blurs a real difference in the issuers' wording; Catalina should read 'acquisition announced, closing date unstated', not 'not shown closed'. | RTL-D-F21 (evidence/verify/D.json) |
| C072 | NIQ Global Intelligence plc priced its IPO on 22 July 2025 (50,000,000 ordinary shares at US$21.00; trading 'expected to begin' on the NYSE on 23 July 2025 and closing 'expected' on 24 July 2025) and filed an 8-K with date of report 24 July 2025, but no opened source states the first trading day or the closing itself, so those dates stay expected, not settled. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | Pricing release ('CHICAGO (July 22, 2025)'): 'expected to begin trading on the New York Stock Exchange on July 23, 2025,'; 'The offering is expected to close on July 24, 2025, subject to customary closing conditions.' 10-Q, Note 1, 'Initial Public Offering': 'On July 24, 2025, NIQ Global Intelligence plc completed its IPO,' 'in which NIQ Global Intelligence plc sold 50,000,000 ordinary shares' 'at the initial public offering price of $21.00 per share.' 'NIQ Global Intelligence plc received aggregate net proceeds of $985.1 million'. Full-text search finds 'began trading' in the FY2025 10-K (fil | supported with qualification | NIQ Global Intelligence plc priced its IPO on 22 July 2025 (50,000,000 ordinary shares at US$21.00; pricing release) and completed it on 24 July 2025 with aggregate net proceeds of US$985.1 million (10-Q for the quarter ended 30 June 2025, Note 1); the first NYSE trading day (expected 23 July 2025) is stated only as expected in the documents read, though the FY2025 10-K contains the phrase 'began trading'. | Add the 10-Q as a source and a 2025-07-24 'completion' row to consolidation_events; the first-trade date stays 'expected' until the 10-K's Item 5 sentence is read. | RTL-D-F22 (evidence/verify/D.json) |
| C073 | Citi launched Citi Commerce Media on 23 September 2026 (issuer release), but its acquisition of Kard is recorded only from reporting ('powered in part by Kard, the rewards platform Citi acquired in August', eMarketer, 23 September 2026): no Citi release on the Kard agreement, its closing or its conditions was opened. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | Citi release, 'NEW YORK', 'August 13, 2026', headline 'Citi Expands Customer Engagement and Commerce Media Capabilities with Addition of Kard': Citi's U.S. Consumer Cards business 'has entered into an agreement to acquire Kard Financial, Inc. (Kard)'; 'Terms of the transaction were not disclosed and are not material to Citi's financial results.'; 'The transaction is subject to satisfaction of customary closing conditions.'; 'Until the transaction closes, Citi and Kard will continue to operate as independent organizations.' No expected closing date. The launch release of 23 September 2026 that | contradicted | Citi launched Citi Commerce Media on 23 September 2026 (issuer release). Citi's U.S. Consumer Cards business announced on 13 August 2026 that it 'has entered into an agreement to acquire Kard Financial, Inc.', on undisclosed terms 'not material to Citi's financial results' and 'subject to satisfaction of customary closing conditions', with the two companies operating independently until closing; the 23 September launch release still describes the deal as 'Citi's agreement to acquire Kard'. No completion record was found by 3 November 2026, so the event is an agreement (pending), and eMarketer' | Event type and status are wrong (completion versus agreement), the evidence class 'synthesis' with confidence 'low' should become direct issuer record with high confidence for the agreement, and the claim that no Citi source on the Kard agreement was opened is contradicted by the source D itself cites (RTL-U-S314). | RTL-D-F25 (evidence/verify/D.json) |
| C074 | The remaining Q3 events are classified in the consolidation_events table from the sources workstream U opened: WBA take-private completed 28 August 2025; Kroger-Albertsons enjoined 10 December 2024 and terminated by both parties' notices of 10 and 11 December 2024; Walmart-VIZIO agreement 19 February 2024, announcement 20 February 2024, completion 3 December 2024 for net cash consideration of US$1.9 billion; Omnicom closed Flywheel Digital 2 January 2024; Neptune completed Quotient 5 September 2023; Assembly took the Pacvue name 27 March 2023; Kenshoo became Skai 8 June 2021; Lowe's One Roof M | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | WBA 8-K: 'Date of Report (Date of earliest event reported): August 28, 2025'; 'On August 28, 2025, pursuant to the Merger Agreement, Merger Sub merged with and into the Company'; '$11.45 in cash, without interest thereon' plus 'one divested asset proceed right'; 'ceased trading on the Nasdaq effective prior to the opening of trading on August 28, 2025.' Kroger 8-K (report date December 11, 2024): 'On December 11, 2024, The Kroger Co. ... delivered a notice'; 'the December 10, 2024 decision of United States District Court for the District of Oregon' ('3:24-cv-00347-AN') 'issued a preliminary in | supported with qualification | Add to the Kroger-Albertsons entry that Kroger's 8-K states Albertsons' 10 December letter 'is not an effective termination' (the terminating notice is disputed between the parties), and mark the Kroger quote 'today terminated its merger agreement with Albertsons' as exhibit 99.1 wording not present in the 8-K body read. | Checked: WBA, Kroger, Albertsons, VIZIO agreement and completion, the US$1.9 billion phrase (full-text hit in the fiscal 2025 10-K), Omnicom, Neptune, Sam's Club Connect, and the three browser-pane items against the session notes. Not re-opened: Skai (8 June 2021), Lowe's (13 October 2021; 8 August 2024), MARRIOTT MEDIA (12 June 2025), Kinective Media (7 June 2024, brief record). | RTL-D-F26 (evidence/verify/D.json) |
| C075 | Filing and market-series denominators: every row carries perimeter, audit status and claim status; rows never added or spliced. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | Every numeric value in the table matches its source: Amazon 46,906 / 56,214 / 68,635 and 19,637 / 17,703 / 56,689 / 47,318; Walmart 3.4 / 4.4 / 6.4 billion and 22% / 24% / 41%; Maplebear 871 / 958 / 1,065; Criteo 209.007 / 258.303 / 263.872 and 203.46 / 253.846 / 259.684; Cardlytics 309.204 / 278.298 / 233.273 with the segment and UK splits; Expedia 483 / 639 / 758; eBay 544 / 517; Uber run rates 900 / 1,500; IAB 37.5 / 43.7 / 53.7 / 63.4; eMarketer 3,000 floor and 77.3% / 21.2%; Kantar 332.2 / 333.4. | supported with qualification | Row edits: (1) Maplebear rows geo 'US' -> 'company-wide (not shown to be US-only)'. (2) Walmart FY2025 unit "approximate: '$4.4 billion'" -> "rounded: '$4.4 billion'"; only FY2024 ('approximately') and FY2026 ('nearly') are worded as approximations. (3) eBay 'includes' cells -> 'not defined in the release (Promoted Listings named only in forward-looking statements)'. (4) DoorDash row 'unstated' -> 'no advertising amount in the Q4 2025 letter or press release (the release names advertising revenue only as a receivables driver); FY2025 10-K not read'. (5) IAB rows 'audit_status' -> 'survey of co | IAB rows' measurement treatment (see RTL-D-F01): partly estimated, not purely reported. | table:denominators (evidence/verify/D.json) |
| C076 | Q3 timeline classifying every event since 2021 by type, status and evidence basis. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | Rows checked against issuer records and filings: 2022-08-01 IRI-NPD completion; 2023-09-05 Neptune-Quotient; 2024-01-02 Omnicom-Flywheel; 2024-02-19 VIZIO agreement; 2024-02-20 Walmart announcement (Q4 FY24 release: 'Walmart agrees to buy VIZIO HOLDING CORP. to further accelerate Walmart Connect in the U.S.'); 2024-12-03 VIZIO completion; 2024-12-10 injunction and Albertsons notice; 2024-12-11 Kroger notice; 2025-07-22 NIQ pricing; 2025-07-24 NIQ 8-K; 2025-08-28 WBA; 2026-01-07 Edge (browser notes); 2026-01-23 / 01-26 / 03-24 Cardlytics-PAR; 2026-02-23 Infillion-Catalina; 2026-06-22 Sam's Club | supported with qualification | Row edits: (1) CONTRADICTED ROW: '2026-08 (reported), completion, Citigroup Inc.; Kard, completed (reported)' -> '2026-08-13, agreement, Citigroup Inc. (U.S. Consumer Cards business); Kard Financial, Inc., pending close ("subject to satisfaction of customary closing conditions"; terms undisclosed), issuer release'. (2) '2023-03-06/07 rename' -> '2023-03-07, rename announcement, release dated March 7th, 2023 on Hellman & Friedman's site (majority investor; a party) and Business Wire URL stamp 20230307; Circana's own post unread'. (3) Add '2025-07-24, completion (IPO closed), NIQ Global Intellig | Citi-Kard row: wrong event type and status (completion versus agreement) and wrong evidence basis (reporting, when the issuer release is available and was already cited by U). | table:consolidation_events (evidence/verify/D.json) |
| C077 | Market-research series screened for Q4 and H4a: publisher, definition, vintage, geography, CTV and in-store treatment, split disclosed, access. | Refute-mode verifier for evidence/research/D.json re-opened the cited source. | IAB rows: definitions, vintages and 'no split' confirmed (full-text screens of both extracts); FY2024 definition sits on p. 27 (section 'Survey definitions' begins on p. 26). eMarketer row: sentences confirmed in the browser notes. WPP Media, MAGNA, dentsu, Advertiser Perceptions and Coresight rows are access or search records, not re-tested. | supported with qualification | Edit the FY2024 IAB row: 'survey definition p. 26' -> 'p. 27'. Add to both IAB rows: method includes PwC estimates for non-participating companies (FY2025 p. 35; FY2024 same wording). H4a 'no eligible series' stands: neither IAB report, nor any other series read, discloses a US on-site, off-site and in-store split. | table:market_series (evidence/verify/D.json) | |
| C078 | Amazon Sponsored Products' documented default outcome reporting in the US is click-based attribution of sales, orders and units to the ad click with no counterfactual: Amazon's 1 January 2026 change left 'Click-based attribution' and 'all other campaign types' unchanged, and the Sponsored Products prompts report (US general availability 25 March 2026) lists 'sales, ACOS, ROAS, and 7-day orders and units'. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Update page (dated 'January 01, 2026'): 'Conversion reporting for all other campaign types and inventory types remains unchanged.' 'Click-based attribution remains unchanged.' Sponsored Products is not in the scope list. Prompts page (session read): 'The Prompts report includes prompt text, associated ad, impressions, clicks, click-through rate, cost per click, spend, sales, ACOS, ROAS, and 7-day orders and units.' No opened page states the Sponsored Products touch model or window. | supported with qualification | Amazon Sponsored Products' documented default outcome reporting is attribution of sales, orders and units with no counterfactual: Sponsored Products sits outside the 1 January 2026 view-attribution change ('Conversion reporting for all other campaign types and inventory types remains unchanged'), and the Sponsored Products prompts report (US general availability 25 March 2026) lists 'sales, ACOS, ROAS, and 7-day orders and units'. The touch model (click-only) and the console window are not stated on the pages opened; 'click-based' is an inference from CPC billing. | 'Click-based attribution' is quoted as if it named the Sponsored Products model; on the page it is a general statement that click attribution is unchanged. Code touch model 'click (inference)'. | RTL-E1-F01 (evidence/verify/E1.json) |
| C079 | Since 1 January 2026 Amazon Sponsored Brands campaigns billed on viewable impressions (vCPM), including branded keyword campaigns, report Purchases, Sales and ROAS under a 'shopping-signal enhanced last-touch attribution model' with 'a shorter attribution window' that Amazon does not quantify, while 'all views' metrics within a 14-day window remain available through Unified Reporting interfaces and APIs; click-based attribution is unchanged (US among 21 countries). | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | 'We are introducing a shopping-signal enhanced last-touch attribution model on January 1, 2026' ... 'applies a shorter attribution window to reflect shopper behavior in the store' (length not given); scope: 'Sponsored Brands, Sponsored Display campaigns billed on a viewable impression (vCPM) basis' (including branded keyword campaigns) and 'Amazon DSP campaigns that serve ads in the Store'; 'all views' metrics 'within a 14-day window'. Availability lists six regional lines: 'North America: United States, Canada, Mexico' / 'South America: Brazil' / 'Europe: United Kingdom, Germany, Spain, Italy | supported with qualification | Replace '(US among 21 countries)' with '(the United States among 23 countries named on the page as fetched on 3 November 2026)'. Record the scope phrase verbatim and flag that it is grammatically ambiguous whether 'billed on a viewable impression (vCPM) basis' qualifies Sponsored Brands as well as Sponsored Display; the vCPM-only reading for Sponsored Brands is one reading, not a stated rule. | The session's browser note says 'the US and 20 other countries'; the live page lists 23. Re-check the list at the 20 November freeze. No statement on restating historical data. | RTL-E1-F02 (evidence/verify/E1.json) |
| C080 | Amazon display ads (formerly Sponsored Display) billed on a vCPM basis moved to the same shopping-signal enhanced last-touch view model on 1 January 2026, with 14-day 'all views' metrics retained and click attribution unchanged; the display ads product page itself states no window, touch model or counterfactual. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Scope list: 'Sponsored Brands, Sponsored Display campaigns billed on a viewable impression (vCPM) basis' ... 'Amazon DSP campaigns that serve ads in the Store'; 'Click-based attribution remains unchanged.' | supported with qualification | As stated, adding that the page names 'Sponsored Display' (the former name) and does not itself say 'display ads'; the alias rests on RTL-U-S003. | Same scope ambiguity as F02. | RTL-E1-F03 (evidence/verify/E1.json) |
| C081 | Amazon's 1 January 2026 view-attribution change is a documented, dated reporting revision disclosed to buyers in advance for Amazon Sponsored Brands, Amazon display ads (vCPM) and Amazon DSP in-Store campaigns, with the legacy 14-day 'all views' metrics kept available as a parallel series. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Page date 'January 01, 2026'; 'We are introducing a shopping-signal enhanced last-touch attribution model on January 1, 2026'; 'all views' metrics 'within a 14-day window' kept; no statement on whether historical data is restated. | supported with qualification | Amazon's 1 January 2026 view-attribution change is a documented, dated reporting revision for Amazon Sponsored Brands, Amazon display ads (vCPM) and Amazon DSP in-Store campaigns, published on a page dated 1 January 2026 (the effective date), with the legacy 14-day 'all views' metrics kept available as a parallel series; the page says nothing about restating historical data. Delete 'disclosed to buyers in advance': no earlier notice was opened. | RTL-E1-F05 (evidence/verify/E1.json) | |
| C082 | Amazon Marketing Cloud documents buyer-run queries for sponsored ads advertisers: self-service access at no cost via UI and API, queries over 'event-level and user-level signals' including 'ad-attributed impressions, clicks, and conversions', and only 'aggregated, anonymous outputs' under aggregation thresholds whose values are not stated (H3 axis (b) documented for Amazon Sponsored Products; axis (a) not documented). | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Product page (live, 3 November 2026): 'Through AMC, advertisers can query event-level and user-level signals across sources' ... 'with built-in aggregation thresholds ensuring user privacy.'; 'including ad-attributed impressions, clicks, and conversions.'; 'You can only access aggregated, anonymous outputs from AMC.'; 'Amazon Marketing Cloud is available at no cost to eligible advertisers via web-based UI and API.'; 'those using sponsored ads can access AMC via self-service.'; 'Built on AWS Clean Rooms'. Developer docs (session browser read): 'Tables in AMC can be queried like SQL tables.'; co | supported with qualification | Amazon Marketing Cloud documents buyer-run SQL queries, self-service and at no cost for sponsored ads advertisers, over a table with one record per Sponsored Products impression and per click, joinable to conversion tables that pair traffic and conversion events, with only 'aggregated, anonymous outputs' leaving the clean room under per-column aggregation thresholds whose numeric values are not published. H3 axis (b) is documented at the exact service scope (Amazon Sponsored Products): this is an established counterexample, not a candidate, so H3 is contradicted at step 1 of its decision order | The finding says axis (a) 'not documented' while F12 and access_record row 1 code it 'source inaccessible'; reconcile to 'not documented in the screened record (AMC docs read in the browser; Ads API reporting docs RTL-E1-S34 unread)'. | RTL-E1-F09 (evidence/verify/E1.json) |
| C083 | For Amazon DSP, AMC access with 'DVA+ ... advanced settings' is by request to an Ad Tech Account Executive while partners on the Amazon Ads Partner Network can use AMC on advertisers' behalf and agency partners 'offer professional services on query development and custom analytics in AMC'; the DSP product page lists AMC, attribution, brand lift and sales lift only as navigation links. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | AMC page: 'Advertisers looking to leverage DVA+ with advanced settings (within Display, Video, and Audio)' 'can reach out to their Ad Tech Account Executive to request access'; 'Advertisers working with partners registered on the Amazon Ads Partner Network can also utilize AMC via their partners.'; 'Agency partners offer professional services on query development and custom analytics in AMC'. Approved-providers page, section 'Self-Service Third-Party Measurement Studies': 'Amazon Ads offers a self-service workflow for third-party measurement studies within Amazon DSP.' 'Advertisers and agencie | supported with qualification | Add: Amazon's approved-providers page (last updated 1 June 2026) documents a self-service workflow in which Amazon DSP advertisers configure, fund and read third-party measurement studies from '50+ vendor products' inside Amazon DSP, available to managed-service and self-service advertisers; study types (brand lift, sales lift) and designs are not named. The H1b flags for Amazon DSP become: a third-party study route documented (request not required: self-service; funded by the advertiser or vendor; design unknown). | This also belongs in access_record row 4 (Amazon DSP). | RTL-E1-F10 (evidence/verify/E1.json) |
| C084 | Amazon's 'Approved third-party providers' page (last updated 1 June 2026) documents third-party verification for Amazon DSP only: DoubleVerify, Integral Ad Science and Adloox via S2S, custom-audience campaigns limited to DSP-hosted creatives, 'Certain products and placements do not support third-party ad serving', third-party impression tracking on the Amazon Mobile Shopping site and app, and 'Samba TV' listed as a 'Measurement' vendor. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | 'Last updated: June 1, 2026'; 'Ad verification via third-party vendors such as DoubleVerify, Integral Ad Science, and Adloox is available on Amazon DSP' 'via server-to-server (S2S) integrations.'; 'Advertisers can enable third-party ad verification on campaigns targeting custom audiences only' 'via Amazon DSP hosted creatives.'; 'This feature is not supported on Third Party Ad Server creatives.'; 'Certain products and placements do not support third-party ad serving'; 'Third party impression tracking is available for the Amazon Mobile Shopping website and app.' 'Please note that we cannot acce | supported with qualification | As stated, adding the 'Self-Service Third-Party Measurement Studies' section (Amazon DSP only; see F10) and that third-party impression tracking on Amazon Mobile Shopping is by 1x1 pixel only, with JavaScript tags refused on the app. | Samba TV appears as one 'Measurement' vendor in a list of about 55; the listing records approval to serve or measure on Amazon DSP, not an integration with any sponsored ads product (shared rule 6: a partner name alone earns nothing). Keep the author's interest mark wherever Samba TV is named. | RTL-E1-F11 (evidence/verify/E1.json) |
| C085 | This review found no documentation of third-party verification tags, event-level exposure export or buyer-run outcome analysis specific to Amazon Sponsored Products on-site other than AMC (RTL-E1-F09) on the Amazon pages screened by 3 November 2026 (Sponsored Products product page, the approved third-party providers page, the AMC page, the prompts page, the 1 January 2026 notice); the Amazon Ads API reporting docs and AMC developer docs were JavaScript shells to the fetch tool. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Approved-providers page: no Sponsored Products or Sponsored Brands verification statement ('sponsored ads' only in policy links). AMC docs (session browser read): one record per Sponsored Products impression and click, queryable in SQL, aggregated outputs only. | supported with qualification | This review found no documentation of third-party verification tags or of event-level exposure export for Amazon Sponsored Products on the pages screened by 3 November 2026; buyer-run outcome analysis is documented through AMC (F09). Code H3 axis (a) 'not documented in the screened record' (AMC docs read in the browser; Ads API reporting docs unread) rather than 'source inaccessible', and axis (b) 'documented'. | Internal inconsistency with F09 on axis (a). | RTL-E1-F12 (evidence/verify/E1.json) |
| C086 | MRC's accredited-services list (read 3 November 2026) carries Amazon Ads Sponsored Products for 'Clicks: Display', 'Clicks: Search' and 'Sophisticated Invalid Traffic Detection/Filtration', each 'Total only', and Amazon DSP for viewable impressions, rendered ad impressions and clicks across display, video and CTV environments with the footnote 'Not all metrics and environments are accredited for every supply source'; no outcome metric is accredited for either row. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Logged screen: Amazon Ads Sponsored Products (GI070826): 'Clicks: Display' / 'Clicks: Search' / 'Sophisticated Invalid Traffic Detection/Filtration', each 'Total only'. Amazon DSP (GI052926; listed as 'Amazon DSP Retail and IMDb, Amazon Publisher Services (APS), Fire TV and Third-party Exchanges (3PX), Alexa Connected Devices'): viewable impressions and viewability display and video (D, MW, MA); rendered ad impressions display (D, MW, MA, CTV) and video (D, MW, MA); clicks display (D, MW, MA, CTV) and video; footnote 'Not all metrics and environments are accredited for every supply source.' | supported with qualification | As stated, but narrow 'across display, video and CTV environments': CTV appears only for display rendered impressions and display clicks; video metrics are desktop, mobile web and mobile in-app; and the DSP entry is scoped by supply source (Amazon DSP Retail and IMDb, APS, Fire TV, 3PX, Alexa Connected Devices). | RTL-E1-F18 (evidence/verify/E1.json) | |
| C087 | Walmart Connect Offsite Display (managed offsite display, video, CTV and social) shares the pooled offsite measurement statement ('connect real-time impressions, clicks, and conversions to verified Walmart transactions online') and, for social, 'Closed-loop measurement via LiveRamp's data clean room is now available for Meta' and 'also available for TikTok and Pinterest' (8 April 2026), with the social statistic 'aligned to a 14-day attribution window'. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | 'Closed-loop measurement via LiveRamp's data clean room is now available for Meta.' 'Closed-loop measurement is also available for TikTok and Pinterest.' Footnote 2 supports '48% higher ROAS and 20% higher click-through rate' for advertisers running Walmart Connect Social alongside Onsite Display versus display-only, source 'Walmart first-party data, FY26 (Feb. 1, 2025–Jan. 31, 2026)', 'aligned to a 14-day attribution window', including only display campaigns with at least 14 days of overlap with social campaigns. | supported with qualification | Keep the LiveRamp sentences; replace 'with the social statistic aligned to a 14-day attribution window' by 'and one footnoted FY26 comparison (Social alongside Onsite Display versus display-only) is aligned to a 14-day attribution window'. The 14-day figure is the analysis window of that comparison, not a documented reporting window for social closed-loop measurement; attribution_parameters row 11 should read window 'unknown (one FY26 analysis aligned to 14 days)'. | A footnote's analysis window was coded as a product parameter. | RTL-E1-F26 (evidence/verify/E1.json) |
| C088 | H4b record for Walmart Connect Offsite Display (social): the join for Meta, TikTok and Pinterest campaigns runs through 'LiveRamp's data clean room', an identity match outside Walmart's logged-in surface, measuring online and in-store Walmart sales; the lift benchmarks cited rest on a Circana analysis (September 2024 to September 2025) of campaigns that qualified for a sales-lift study. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | 'Walmart Connects sales lift benchmarks analysis performed by Circana, September 2024–September 2025.'; the benchmark covers campaigns that qualified for a sales lift study in Food, Consumables, Beverages or Alcohol only; 'Circana is an independent third party and not affiliated with Walmart Connect.' | supported with qualification | As stated, adding that the 2.5x and 1.8x figures cover only campaigns that qualified for a sales-lift study in Food, Consumables, Beverages or Alcohol, and that Walmart describes Circana as 'an independent third party and not affiliated with Walmart Connect'. | E1's limitation calls the Circana benchmark 'vendor-commissioned'; the page states neither commissioning nor funding. Code funding 'unknown' and record Walmart's independence statement as the vendor's assertion, not as established independence. | RTL-E1-F27 (evidence/verify/E1.json) |
| C089 | Walmart Connect Store Ads documents a lift-type default with a confidence rule: 'Store ads measurement includes both brand and sales lift', 'Sales lift validated at the 80% confidence level or above', the Deli and Bakery screen proof-of-concept 'achieved up to 5% sales lift for Food suppliers', and 'Gross impressions are estimated projections'; assignment and counterfactual construction are not stated. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | 'Store ads measurement includes both brand and sales lift.' sits in the 'Measure performance' section. 'Sales lift validated at the 80% confidence level or above.' is a sentence inside footnote 4, attached to the claim of 'up to 5% sales lift for Food suppliers on Deli screens and Bakery screens'; footnote 4 reads 'Walmart first-party data, In-store Deli and Bakery screen campaigns, food suppliers only, September 2024 – January 2025', 'based on initial proof-of-concept testing', then the 80% sentence and 'Past results do not guarantee future performance.' 'Gross impressions are estimated proje | contradicted | Walmart Connect Store Ads states that 'Store ads measurement includes both brand and sales lift' (lift is the described store-ads measurement; assignment, counterfactual, eligibility and any significance rule for that measurement are not stated). Its Deli and Bakery screen proof-of-concept 'achieved up to 5% sales lift for Food suppliers' (food suppliers only, September 2024 to January 2025), a figure footnoted as 'validated at the 80% confidence level or above'. The 80% threshold qualifies that one proof-of-concept statistic and is not a documented significance rule for store-ads lift reports | A footnote on one vendor statistic was coded as a standing significance rule and its test population ('food suppliers only') as eligibility; this changes an H1b flag and a summary sentence. | RTL-E1-F28 (evidence/verify/E1.json) |
| C090 | Sam's Club Connect documents attribution as its default ('Our membership model links ad exposure to sales across every touchpoint') in a self-service dashboard and a separate 'Cross-channel incrementality' offer (Omni-Impact) whose '36%' average sales lift cites a 'Circana FY26 report (1,440 campaigns)'; validating partners are unnamed and no assignment, counterfactual, window or match basis is stated. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | 'Our membership model links ad exposure to sales across every touchpoint, delivering clearer, more accurate measurement.'; 'Independent partners validate campaign performance, giving you greater confidence in your results.'; headline stats '54%¹' new buyers and '36%¹' incremental sales, footnote 'Circana FY26 report (1,440 campaigns)'; Omni-Impact bullets 'Cross-channel incrementality', 'Sales lift, including sponsored ads', 'Media efficiency and marginal iROAS'; dashboard bullets 'Attributed sales', 'ROAS', 'New buyers', 'Conversion'. | supported with qualification | As stated, adding that the measurement page labels the 36% 'incremental sales' (E1 quotes the home page's 'Average sales lift'), that the self-service dashboard lists 'Attributed sales' and 'ROAS' (supporting the attributed default), and that Omni-Impact names 'Sales lift, including sponsored ads' and 'marginal iROAS'. | The 36% is a Circana-footnoted vendor average over 1,440 campaigns; design unstated, so it is not a lift result with a known counterfactual. | RTL-E1-F35 (evidence/verify/E1.json) |
| C091 | Instacart offers 'Sales lift testing' using 'gold-standard experimentation', and its 30 June 2023 engineering post documents the design as user-level random assignment ('hash of (user_id, advertiser_id)') to ads and holdout groups with ghost ads, limited to Sponsored Products; neither page says whether a test must be requested, who qualifies, or whether lift appears in default reports. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Engineering post: 'users are randomly split into an "ads" group ... and "holdout" group'; 'Users are randomized based on hash of (user_id, advertiser_id)'; 'continual ads incrementality measurement is limited to Instacart's Sponsored Products offering.' Pilot page ('Last Updated: Sep 2, 2022'): 'Brand partners working with Instacart Ads now have access to a pilot testing experience where they can measure true total sales lift, leveraging industry-leading methodology using randomized test and holdout groups'. Reckitt page (5 May 2023): 'comparing the sales of two randomly selected groups'; '**L | supported with qualification | Instacart offers sales-lift testing for Sponsored Products, introduced as a 'pilot testing experience' using 'randomized test and holdout groups' (page last updated 2 September 2022) and documented in its 30 June 2023 engineering post as user-level random assignment ('hash of (user_id, advertiser_id)') with ghost ads; Instacart states that 'Lift testing services may not be available to all partners. Please consult with your account team' (5 May 2023). H1b flags: offer documented; request via the account team; eligibility 'not available to all partners' (criteria unstated); lift outside the def | 'Neither page says whether a test must be requested, who qualifies' is true of the two pages cited but not of Instacart's record: the Reckitt page answers both in part. default_method row 12 should not code request and eligibility as plain unknown. | RTL-E1-F40 (evidence/verify/E1.json) |
| C092 | Instacart's Data Hub is a Snowflake clean room in which partners run SQL analytics on combined datasets ('queries initiated by a user are allowed') with Instacart templates (including Ads Taxonomy, New to Brand, Path to Purchase) and no guaranteed access; the pages state no output threshold, export rule or destination and name no Sponsored Products exposure dataset (H3 axis (b): platform level, not service scope). | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Data Hub (last updated 20 July 2026): 'The Instacart Data Hub is a clean room environment built on Snowflake'; 'Install, configure, and view available clean rooms through SQL commands.'; 'View provider and consumer datasets available within your clean room through SQL queries.'; alt text 'Direct access is not allowed, but queries initiated by a user are allowed.'; 'Run pre-built Insights templates for' Ads Taxonomy, Overlap Analysis, Lifetime Value, New to Brand, Purchase Frequency, Path to Purchase, Spend-Based Segmentation; 'Access to Instacart Data Hub functionalities, capabilities, and dat | supported with qualification | Instacart's Data Hub is a Snowflake clean room in which partners install and view clean rooms with SQL commands and run pre-built Instacart templates (including Ads Taxonomy, New to Brand and Path to Purchase), with 'partners ... restricted on what they can run' and no guaranteed access; its launch release (January 2026) says advertisers can 'measure campaign impact with flexible attribution approaches' and that it was piloted with select agencies and CPG brands. No page states an output threshold, export rule or destination, names a Sponsored Products exposure dataset, or documents free-form | 'partners run SQL analytics on combined datasets' overstates the documented query control: the docs show SQL setup commands and pre-built templates, not partner-authored analysis. | RTL-E1-F42 (evidence/verify/E1.json) |
| C093 | Instacart's documentation cites an 'MRC-accredited measurement methodology for gross impressions, viewable impressions, and clicks' and MRC lists Instacart (GI060126) for delivery metrics; Instacart's own DoubleVerify and Integral Ad Science integration post (20 October 2023 per the brief) could not be opened by any route, so verification tags for Instacart's on-site formats are coded 'source inaccessible'. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Instacart docs: 'Learn about Instacart's MRC-accredited measurement methodology for gross impressions, viewable impressions, and clicks.' DoubleVerify's release ('NEW YORK – October 12, 2023'): DV will 'verify that Instacart video and display ads are viewable by a human being and are safe from Fraud/Invalid Traffic'; 'will be working on the technical integration for viewability and IVT verification solutions over the coming months'; IAS, brand safety and sponsored product ads are not mentioned. | supported with qualification | As stated for MRC. For verification: DoubleVerify's own release of 12 October 2023 scopes its Instacart integration to viewability and invalid-traffic verification of 'Instacart video and display ads', with the integration still to be built 'over the coming months'; sponsored product ads are not named. Code Instacart Sponsored Products verification tags 'not documented for sponsored product ads (DV release); Instacart's own post and IAS's release unread', and the display and video row 'DV viewability and IVT documented as announced (2023); completion date not documented on pages read'. | DV is an interested party; its release documents the announced scope, not completion. | RTL-E1-F43 (evidence/verify/E1.json) |
| C094 | Roundel documents a multi-touch attribution default ('Using a multi-touch attribution model, we distribute credit across the customer touchpoints leading to a purchase') with metrics such as 'New guest ROAS', and a separate incrementality offer using synthetic control, matched market and randomized controlled trials that 'is available for eligible campaigns'; design per campaign, assignment, eligibility rule, minimum and significance rule are not stated. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Measurement page: 'Using a multi-touch attribution model, we distribute credit across the customer touchpoints leading to a purchase'; 'We use methodologies like synthetic control, matched market and randomized controlled trials'; 'Incrementality measurement is available for eligible campaigns.' 'Based on your campaign objectives, investment level, and measurement needs,' 'your Roundel team can recommend the right approach for your business.'; metric lists: Incrementality 'Incremental sales', 'iROAS', 'Lift'; Full-funnel 'ROAS', 'Attributed sales', 'Site behavior', 'Influenced sales / engageme | supported with qualification | As stated, adding: the footnote ties eligibility to 'campaign objectives, investment level, and measurement needs' with the Roundel team recommending the approach (eligibility criteria named but thresholds unstated; route via the Roundel team), and the Matched Market Model is described as 'a geo holdout approach' (counterfactual construction for that design: geo holdout / matched market; assignment unknown). | RTL-E1-F46 (evidence/verify/E1.json) | |
| C095 | Kroger Precision Marketing's documented onsite method is an exposed-versus-unexposed household comparison that KPM calls both 'incrementality' and 'closed-loop attribution', reported in the Kroger Ad Platform as 'sales lift and household penetration', with test and control groups compared across variables such as pre-campaign spend, loyalty and geography (assignment not disclosed). | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | FAQ: 'KPM measures performance by matching media exposure to real household purchases across in‑store and ecommerce transactions. Our incrementality and closed-loop attribution methods compare exposed and unexposed groups to isolate true impact.' KAP page: 'Track meaningful metrics such as sales lift and household penetration—not just clicks.' MRC under review: Kroger services 'Targeted Onsite Ads, Product Listing Ads, Onsite Video, Promoted Product Carousels', metrics 'Impressions, Clicks, Viewable Impressions, Attributed sales, Incremental Sales, aROAS, iROAS'. | supported with qualification | Kroger Precision Marketing documents both attributed and incremental onsite outcomes: its network-wide FAQ names 'incrementality and closed-loop attribution methods' that 'compare exposed and unexposed groups', the Kroger Ad Platform reports 'sales lift and household penetration', and MRC's pre-audit scope for the onsite formats lists 'Attributed sales, Incremental Sales, aROAS, iROAS'. Which of the two is the default report is not stated; code default_method row 15 'both reported (default not stated)', assignment unknown, counterfactual matched/unexposed comparison, and run H1a with this row | Coding the onsite default as 'incremental' rests on a network-wide FAQ answer and a platform page; the MRC scope shows attributed sales and aROAS are also reported for the same formats. | RTL-E1-F49 (evidence/verify/E1.json) |
| C096 | Kroger Precision Marketing offsite programmatic documents an 'in-flight attributable sales' default (since 2023) and, for qualified self-service campaigns through The Trade Desk, post-campaign incremental sales against 'a control audience that mirrors the exposed audience', included in the standard wrap report and subject to a minimum campaign spend (launched December 2024; post of 14 January 2025); assignment is not disclosed. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | 'Incremental sales measurement is now available for qualified campaigns placed through The Trade Desk. This new capability, launched in December, represents the causal impact of media spend.'; 'our approach uses a control audience that mirrors the exposed audience'; 'In addition to the in-flight attributable sales measurement we've offered since 2023'; 'Available for self-service advertisers using The Trade Desk DSP'; 'Incremental sales measurement will be included in the standard campaign wrap report'; 'Requires a minimum campaign spend.'; 'Beta testing began in March'. | supported with qualification | As stated, adding that beta testing 'began in March' (2024, by inference from the post date) and that 'launched in December' is December 2024 by inference. In default_method row 16, code request_required 'not stated (included in the standard wrap report for qualified campaigns; no request step described)' rather than 'no'. | 'Request not required' is an inference from inclusion in the wrap report. | RTL-E1-F51 (evidence/verify/E1.json) |
| C097 | MRC's under-review list reads 'Kroger (Pre-audit in process as of 6/2026)' for 'Targeted Onsite Ads, Product Listing Ads, Onsite Video, Promoted Product Carousels' with metrics 'Impressions, Clicks, Viewable Impressions, Attributed sales, Incremental Sales, aROAS, iROAS - Desktop, Mobile Web, Mobile In-App', the only retail media outcome-metric candidate in MRC's process; KPM's FAQ claims certification by 'the Interactive Advertising Bureau (IAB) and Merchant Risk Council (MRC)' (a different MRC) and its home page cites TAG Platinum certification. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | MRC: Kroger 'Pre-audit in process as of 6/2026', 'Targeted Onsite Ads, Product Listing Ads, Onsite Video, Promoted Product Carousels', 'Impressions, Clicks, Viewable Impressions, Attributed sales, Incremental Sales, aROAS, iROAS', 'Desktop, Mobile Web, Mobile In-App'. Also under review: Kevel 'Pre-audit reviewed 9/2026' with 'Clicks, Conversions and incrementality'; Schwarz Media Onsite Advertising 'Audit reviewed 9/2026' with 'shoppers reached, units purchased' ('Display banner ads served in carousel ad format for Lidl & Kaufland apps'). | supported with qualification | Replace 'the only retail media outcome-metric candidate in MRC's process' with 'the only outcome-metric candidate in MRC's process among the E1 networks; MRC's under-review list also carries Kevel (Pre-audit reviewed 9/2026: 'Clicks, Conversions and incrementality') and Schwarz Media Onsite Advertising (Audit reviewed 9/2026: 'shoppers reached, units purchased', Lidl and Kaufland apps, not a US service)'. The FAQ and TAG sentences hold as stated. | An unbounded 'only' contradicted by the same MRC page. | RTL-E1-F53 (evidence/verify/E1.json) |
| C098 | This review found no documentation of event-level exposure access or buyer-run outcome queries for Kroger Precision Marketing onsite on the KPM pages read; the 84.51° Collaborative Cloud appears only in trade reporting (19 November 2021) as a hosted environment for CPG data scientists over transaction-level data for 60 million de-identified households, with no mention of KPM or ad exposures; 8451.com refuses the fetch tool. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | KPM pages read in the browser document no event-level exposure access or buyer query; FAQ mentions clean rooms 'where appropriate'. 8451.com (84.51°, which runs KPM measurement and the Collaborative Cloud) refuses the fetch tool. | supported with qualification | As stated for the KPM pages, but code H3 axes (a) and (b) for KPM onsite 'not documented on the KPM pages read; screen incomplete (8451.com inaccessible)', so the service enters H3 as an incomplete search, consistent with the Walmart rows, not as a completed 'not documented'. | A required source (the measurement operator's own site) was inaccessible; the decision rule treats that as incomplete. | RTL-E1-F54 (evidence/verify/E1.json) |
| C099 | Albertsons Media Collective onsite documents two constructions without saying which one standard reporting delivers: an 'incrementality-based multi-touch attribution solution' built with LiveRamp and onsite 'test and control frameworks' (16 April 2026) with groups built by matching on predicted purchase behaviour (H1a: default unknown; H1b: offer documented, assignment unknown, counterfactual matched audience). | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Release ('April 16, 2026'): 'The methodology uses test and control frameworks to isolate the effect of media exposure.'; names 'onsite display media' only; no group construction, reporting route, eligibility or significance; cites the Ovative/Kellogg study ('found iROAS can vary by 6.5x and flip results in 83% of campaigns based on methodology alone'). Results page (undated, network-wide): 'advanced matching methodologies grounded in predictive purchase behavior' to build 'credible test and control groups'; 'Our first-to-market, grocery-exclusive incrementality-based multi-touch attribution so | supported with qualification | Albertsons Media Collective onsite documents an onsite display incrementality measurement using 'test and control frameworks' (16 April 2026; group construction not stated in the release) beside a LiveRamp-built 'incrementality-based multi-touch attribution solution'; its network-wide results page says its studies build test and control groups with 'advanced matching methodologies grounded in predictive purchase behavior' and reports 70 incrementality studies in 2025, without tying that method to the onsite offer. H1a: default unknown; H1b: offer documented (onsite display), request and eligib | The matching detail comes from the results page, not the onsite release; the finding attaches it to the 16 April 2026 framework. | RTL-E1-F63 (evidence/verify/E1.json) |
| C100 | Albertsons Media Collective in-store measures in-store campaigns with matched test and control stores: its 6 January 2026 release compares stores with ad exposure against 'a rigorously selected set of control stores with no media exposure' and reports 'incremental sales, orders, and unit lift'; allocation of stores (random or not), eligibility and whether every in-store campaign receives the study are not stated, and closed-loop measurement is described as a future capability. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Dateline 'BOISE, Idaho--(BUSINESS WIRE)--' with no visible date; 'comparing sales performance in test stores exposed to ads against a rigorously selected set of control stores with no media exposure'; 'leverages nearly 60 variables for advanced store-level matching to reduce bias and noise'; 'a statistically validated measure of true lift'; metrics rendered by the extractor as 'sales, orders and unit lift'; 'We look forward to advancing the measurement capabilities we offer in the future with closed-loop measurement.'; 'more than 50 unique advertising partners'; no allocation rule, campaign co | supported with qualification | As stated, adding the release's claim of 'a statistically validated measure of true lift' with no threshold, interval or test named (so no significance rule is documented), and noting that the metric phrase could only be confirmed as 'sales, orders and unit lift' on the live page. | Date 6 January 2026 rests on the Business Wire ID and the newsroom copy, as E1 says. | RTL-E1-F64 (evidence/verify/E1.json) |
| C101 | This review found no documentation of event-level exposure data, buyer-run outcome queries, data exports, an API or third-party verification for Albertsons Media Collective onsite on the pages screened by 3 November 2026 (home, about, difference, onsite, offsite and results pages, the 6 January and 16 April 2026 releases, the privacy policy); a 'Client Hub' is linked without any description (H3: axis (a) not documented; axis (b) not documented in the screened record). | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Home page lists 'Client Hub' (hub.albertsonsmediacollective.com) under Advertising Resources with no description; no clean room, API or data-access statement; results page names LiveRamp only. | supported with qualification | As stated, but code the H3 axes for Albertsons onsite 'not documented in the screened record; screen incomplete (Client Hub and standard terms unread)', so the row counts as an incomplete search under the H3 decision rule. | RTL-E1-F66 (evidence/verify/E1.json) | |
| C102 | Uber's US merchant page bases its '20%' 'Increase in sales and orders with Sponsored listings' claim on a platform-level experiment in which 'We remove 5% of winning ads at random and replace them with, if available, the next ad', with results 'not guaranteed'; this is an average-effect experiment across advertisers, not a per-campaign lift report. | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Footnote 2 to '20%' 'Increase in sales and orders with Sponsored listings': 'Uber internal data science, November 2025.'; an ongoing global holdout analysis: 'We remove 5% of winning ads at random and replace them with, if available, the next ad.', compared with the 95% where the ad is shown; 'These averages are not a promise of any specific outcome, and your outcome is not guaranteed.' | supported with qualification | Uber's US merchant page bases its '20%' claim on an ongoing global holdout analysis (Uber internal data science, November 2025) that removes '5% of winning ads at random' and compares them with the 95% shown; a platform-wide, global average-effect experiment, not a US-specific or per-campaign lift report. | The experiment is global, not US-scoped; keep geography 'global (shown on a US page)'. | RTL-E1-F78 (evidence/verify/E1.json) |
| C103 | Data-table cells as coded in E1 (default_method row 4 (Amazon DSP)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | 'Amazon Ads offers a self-service workflow for third-party measurement studies within Amazon DSP.' ... 'configure and fund studies,' 'and access results all within Amazon DSP.' | supported with qualification | lift_offer_documented: 'third-party measurement studies, self-service in Amazon DSP (types and designs unnamed)'; request_required: 'no (self-service)'; eligibility: 'managed-service and self-service advertisers and agencies; 18 countries'. | TABLE default_method row 4 (Amazon DSP) (evidence/verify/E1.json) | |
| C104 | Data-table cells as coded in E1 (default_method row 9 (Walmart Connect Store Ads)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | The 80% sentence and 'food suppliers only' are footnote 4 to the Deli and Bakery proof-of-concept statistic. | contradicted | eligibility: 'unknown'; min_spend_or_sample: 'not stated (80% confidence qualifies one proof-of-concept statistic)'; default_target_contrast 'incremental' may stand on 'Store ads measurement includes both brand and sales lift'. Update the table note and the summary. | See F28. | TABLE default_method row 9 (Walmart Connect Store Ads) (evidence/verify/E1.json) |
| C105 | Data-table cells as coded in E1 (default_method row 12 (Instacart Sponsored Products)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | 'Lift testing services may not be available to all partners. Please consult with your account team' (5 May 2023); pilot with 'randomized test and holdout groups' (2 September 2022). | supported with qualification | request_required: 'yes (via account team)'; eligibility: 'not available to all partners (criteria unstated)'; lift_offer_documented: 'yes (pilot 2022; random user holdout with ghost ads per 2023 post)'. | TABLE default_method row 12 (Instacart Sponsored Products) (evidence/verify/E1.json) | |
| C106 | Data-table cells as coded in E1 (default_method row 14 (Roundel)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | 'Based on your campaign objectives, investment level, and measurement needs,' 'your Roundel team can recommend the right approach for your business.'; Matched Market Model 'using a geo holdout approach'. | supported with qualification | eligibility: ''eligible campaigns'; factors named: objectives, investment level, measurement needs (thresholds unstated)'; request_required: 'via Roundel team (recommendation)'; record the Matched Market Model's counterfactual as geo holdout / matched market, assignment unknown. | TABLE default_method row 14 (Roundel) (evidence/verify/E1.json) | |
| C107 | Data-table cells as coded in E1 (default_method row 15 (Kroger onsite)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | See F49. | supported with qualification | default_target_contrast: 'both attributed and incremental reported (default not stated)'; report H1a with this row as attributed and as lift. | TABLE default_method row 15 (Kroger onsite) (evidence/verify/E1.json) | |
| C108 | Data-table cells as coded in E1 (default_method row 16 (Kroger offsite)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | 'Incremental sales measurement will be included in the standard campaign wrap report'; 'Requires a minimum campaign spend.' | supported with qualification | request_required: 'not stated (wrap report for qualified campaigns)'. | Inference coded as a fact. | TABLE default_method row 16 (Kroger offsite) (evidence/verify/E1.json) |
| C109 | Data-table cells as coded in E1 (default_method row 17 (Albertsons onsite)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Matching 'grounded in predictive purchase behavior' is on the network-wide results page; the onsite release says only 'test and control frameworks'. | supported with qualification | default_counterfactual: 'matched (network-level description; not tied to the onsite offer)'. | TABLE default_method row 17 (Albertsons onsite) (evidence/verify/E1.json) | |
| C110 | Data-table cells as coded in E1 (default_method row 21 (DoorDash CPG)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | The release does not say whether lift results appear in standard reporting. | supported with qualification | lift_in_default_report: 'not stated (separate study for select managed-service advertisers; inference: not default)'. | Inference coded as 'no'. | TABLE default_method row 21 (DoorDash CPG) (evidence/verify/E1.json) |
| C111 | Data-table cells as coded in E1 (access_record row 1 (Amazon Sponsored Products)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | See F09. | supported with qualification | h3_axis_a: 'not documented (AMC: aggregated outputs only; Ads API reporting docs unread)'; h3_axis_b: 'documented (AMC sponsored_ads_traffic + conversion tables; self-service)'; output_rules: 'aggregated, anonymous; per-column thresholds LOW/NONE, numeric values unpublished'. The table note's 'candidate counterexample' becomes 'counterexample: H3 contradicted'. | TABLE access_record row 1 (Amazon Sponsored Products) (evidence/verify/E1.json) | |
| C112 | Data-table cells as coded in E1 (access_record row 4 (Amazon DSP)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | Self-service third-party measurement studies in Amazon DSP; 1x1 impression pixel only on Amazon Mobile Shopping. | supported with qualification | Add 'self-service third-party measurement studies (50+ vendor products)' and 'impression tracking by 1x1 pixel only; JavaScript tags refused on the app'. | TABLE access_record row 4 (Amazon DSP) (evidence/verify/E1.json) | |
| C113 | Data-table cells as coded in E1 (access_record rows 5-6 (Walmart Connect Sponsored Search, Onsite Display)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | DV and IAS named as 'onsite partnerships with third-party measurement leaders'; no metrics or placements stated; no clean room, API or export on the measurement page. | supported with qualification | verification_tags: 'DoubleVerify, IAS named as onsite partners; scope not stated'. Axes stay 'source inaccessible'. | Trade reporting (search lead, not opened) says Walmart Connect tested a clean room with an agency and several CPG clients; not documentation. | TABLE access_record rows 5-6 (Walmart Connect Sponsored Search, Onsite Display) (evidence/verify/E1.json) |
| C114 | Data-table cells as coded in E1 (access_record row 12 (Instacart Sponsored Products)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | DV: 'Instacart video and display ads' only; Data Hub: pre-built templates, restricted queries, pilot with select partners. | supported with qualification | verification_tags: 'not documented for sponsored product ads (DV release of 12 October 2023 names video and display ads); Instacart post and IAS release unread'; h3_axis_b: 'not documented at service scope; candidate (Data Hub release: 'Measure campaign impact with flexible attribution approaches')'. | TABLE access_record row 12 (Instacart Sponsored Products) (evidence/verify/E1.json) | |
| C115 | Data-table cells as coded in E1 (access_record rows 15 and 17 (Kroger onsite; Albertsons onsite)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | See F54 and F66. | supported with qualification | Code both axes 'not documented in the screened record; screen incomplete' (8451.com; Client Hub and terms). | Roundel (row 14): supported as stated; the LiveRamp customer story adds no buyer query. | TABLE access_record rows 15 and 17 (Kroger onsite; Albertsons onsite) (evidence/verify/E1.json) |
| C116 | Data-table cells as coded in E1 (attribution_parameters rows 1-2 (Amazon Sponsored Products, Sponsored Brands)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | No console window stated for Sponsored Products or Sponsored Brands CPC; AMC docs: 'the attribution window, which is 14d'; scope phrase for the vCPM change ambiguous. | supported with qualification | Row 1 touch_model 'click (inference)'; add an AMC note 'AMC conversion tables: 14d; SP by traffic time, SB by conversion time'. Row 2: mark the vCPM split as one reading of an ambiguous scope phrase. | TABLE attribution_parameters rows 1-2 (Amazon Sponsored Products, Sponsored Brands) (evidence/verify/E1.json) | |
| C117 | Data-table cells as coded in E1 (attribution_parameters row 11 (Walmart Connect Offsite Display, social)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | The 14-day window footnote supports a FY26 comparison of Social-plus-Onsite-Display advertisers with display-only campaigns. | supported with qualification | window_click_days: 'unknown (one FY26 analysis aligned to a 14-day window)'. | See F26. | TABLE attribution_parameters row 11 (Walmart Connect Offsite Display, social) (evidence/verify/E1.json) |
| C118 | Data-table cells as coded in E1 (attribution_parameters row 12 (Walmart Connect Store Ads)) | Refute-mode verifier for evidence/research/E1.json re-opened the cited source. | 'Gross impressions are estimated projections based on historical and aggregated data.' is footnote 3 to the reach claim (about 150M customers per week). | supported with qualification | touch_model: 'n/a (lift)'; move 'modelled impressions' to a reach note. | TABLE attribution_parameters row 12 (Walmart Connect Store Ads) (evidence/verify/E1.json) | |
| C119 | Orange Apron Media's measurement post (listed 22 April 2026) offers 'advanced studies such as incrementality, geo lift, and brand lift', clean-room environments with 'turnkey access to deeper insights', Quick-Start Reports in Orange Access, an Insights Console and a Supplier Analytics link to total assortment sales and inventory; no study states its assignment, counterfactual construction, eligibility, request route or presence in the default report. | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | 'Advertisers can also tap into advanced studies such as incrementality, geo lift, and brand lift'; '"Clean Room" environments provide scalable, turnkey access to deeper insights without adding complexity'; 'Within Orange Access, Quick-Start Reports continue to help brands quickly identify what's driving performance'; 'Insights Console gives suppliers a centralized place to monitor key performance indicators alongside competitors'; 'Through a collaboration between Orange Apron Media and The Home Depot's Supplier Analytics program,' (tying promoted products to 'total assortment sales, online per | supported with qualification | Replace the two fetch-tool renderings in the support with the verbatim fragments: 'Insights Console gives suppliers a centralized place to monitor key performance indicators alongside competitors' and 'Through a collaboration between Orange Apron Media and The Home Depot's Supplier Analytics program,' ... 'total assortment sales, online performance, and inventory levels'. Add that the studies are offered as 'advanced measurement solutions built around specific business questions' (wording that points to bespoke studies; request route still unstated). | RTL-E2-F13 (evidence/verify/E2.json) | |
| C120 | The 22 April 2026 Pinterest item resolves to an Orange Apron Media post (dated 22 April 2026 by its blog listing) announcing a beta with Pinterest Media Network Connect, 'set to launch later this summer', that lets advertisers outside home improvement use The Home Depot's audiences on Pinterest, self-managed or with OAM; the post documents no measurement, attribution, clean room or in-store match for those campaigns, and workstream U found no Home Depot investor-relations release of that date. | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Byline 'By Stephenie Cattonar, Senior Director, Strategy, Product & Analytics'; no date, no press-release label and no Home Depot dateline on the post; listing shows 'Orange Apron Media Expands Pinterest Partnership to Non-Endemic Advertisers. April 22, 2026'. Body: 'a new beta partnership with Pinterest Media Network Connect'; 'set to launch later this summer'; 'Advertisers outside of the home improvement category can now leverage' 'Orange Apron Media's first-party data for Pinterest campaigns'; 'Bridges the gap between Pinterest discovery and The Home Depot's verified shopper behavior'; 'a c | supported with qualification | Restate: 'The 22 April 2026 Pinterest item is an Orange Apron Media blog post bylined to Stephenie Cattonar (date from the blog listing only; no Home Depot release or dateline), announcing a beta with Pinterest Media Network Connect "set to launch later this summer" while also saying advertisers outside home improvement "can now leverage" OAM's first-party data on Pinterest; it documents no measurement method, attribution, clean room or in-store match for those campaigns, and asserts only "a clear line of sight to performance" and earlier "meaningful results, including a significant lift in br | The brand-ad-recall lift is a vendor result assertion with no design; it must not be coded as a lift offer or an outcome measurement for the beta. | RTL-E2-F14 (evidence/verify/E2.json) |
| C121 | Marketing Dive reported on 13 October 2026 that Lowe's Media Network is partnering with Ipsos 'to provide independent incrementality measurement', 'including bespoke marketing mix modeling and agile attribution, for brand partners', and that the network claims to be among the first retailers to offer incremental return-on-ad-spend reporting; it quoted the network's VP and GM: 'I don't want an advertiser grading their own work'. No Lowe's own source opened states the partnership, so it is coded reported, with assignment, counterfactual and unit unknown. | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Published Oct. 13, 2026, Chris Kelly, Senior Reporter. 'Lowe's Media Network, which was first launched in 2021, is also partnering with consulting firm Ipsos to provide independent incrementality measurement' 'including bespoke marketing mix modeling and agile attribution, for brand partners'; 'The media network claims to be among the first from a retailer to provide incremental return on ad spend reporting to top spenders.'; 'The partnership with Ipsos speaks to advertiser demands for accredited third-party measurement'; 'I don't want an advertiser grading their own work'; 'per details shared | supported with qualification | Add the eligibility the article reports: the network 'claims to be among the first from a retailer to provide incremental return on ad spend reporting to top spenders' (reported; no threshold given). In default_method, replace 'unknown (reported paraphrase mentions top spenders; unverified)' with 'reported: iROAS reporting "to top spenders" (Marketing Dive, 13 October 2026); threshold not stated'. | 'Independent' is the article's and Lowe's term; Ipsos's commissioning and funding are not stated, so independence is unknown. 'accredited third-party measurement' is the reporter's framing of advertiser demand and attests no accreditation of Ipsos's work. MMM is modelled, not a controlled design. | RTL-E2-F17 (evidence/verify/E2.json) |
| C122 | CMX's ads policy approves third-party tags from CM360 and Flashtalking (click and impression), Innovid, IAS and DoubleVerify (all 1x1), states 'CMX does not accept ad blocking tags', offers Reddit verification through server-to-server DV and IAS integrations, and makes any other tag subject to approval; the same policy says the advertiser 'agrees not to use any data obtained from CVS for measurement or analytics purposes' without prior written consent, so buyer-run outcome analysis is documented only as consent-gated, and no clean room, event-level export or reporting API appears on the seven | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Approved: 'CM360 - Click' 1x1; 'CM360 - Impression' 1x1; 'Flashtalking - Click' 1x1; 'Flashtalking - Impression' 1x1; 'Innovid' 1x1; 'IAS' 1x1; 'Double Verify' 1x1; 'Any tag that is not outlined above is subject to approval by the CMX Ad Policy team.'; 'CMX does not accept ad blocking tags'; '†Reddit offers direct S2S integrations with advertisers with existing DV & IAS relationships (No tags accepted)'; 'Advertiser agrees not to use any data obtained from CVS for measurement or analytics purposes' 'without prior written consent from CVS Media Exchange.'; dashboard: 'Inform your optimization s | supported with qualification | Replace 'offers Reddit verification through server-to-server DV and IAS integrations' with: 'notes that "Reddit offers direct S2S integrations with advertisers with existing DV & IAS relationships (No tags accepted)"' (the integration is Reddit's, not a CMX service). Replace the fetch-tool rendering of the Reddit line in the support with that verbatim footnote. | 'seven CMX pages read' absence claim is bounded to the stream's pages; this check re-opened three. | RTL-E2-F20 (evidence/verify/E2.json) |
| C123 | CMX's activation routes are a CPC sponsored-product offering with a 'CMX Private Marketplace', programmatic self-service through an advertiser's existing seat at The Trade Desk (eligibility determined by CMX), Pinterest and Reddit off-site partners and managed service; CorIQ, described as a proprietary AI media platform with 'Closed-loop measurement tied to verified outcomes', is credited with a '13% average ROAS improvement' on 'CVS Internal Data 2025'. The governance screen found no metric versioning, restatement policy, dated revision, dispute procedure or sales-measurement separation; the | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | 'With Cost Per Click, only pay for action taken'; 'Need quick campaign activation? Try the new CMX Private Marketplace for direct access to our Sponsored Product offering.'; policy: 'With respect to billing utilizing measurement data from the vendors outlined above,' ... 'all billing terms and conditions shall be governed by individual advertiser terms and conditions.'; 'We periodically update our ad policy.'; 'Advertiser represents and warrants that the Restricted Data Processing (RDP) feature will at all times be enabled.'; no effective date. | supported with qualification | Replace the two fetch-tool renderings in the RTL-E2-S17 support with the verbatim fragments above ('With respect to billing utilizing measurement data from the vendors outlined above,' ... 'all billing terms and conditions shall be governed by individual advertiser terms and conditions.' and 'We periodically update our ad policy.'), and correct RTL-E2-S17's reliability note accordingly. | The Trade Desk self-service page (RTL-E2-S20) and the CorIQ post (RTL-E2-S21) were not re-opened; the 13% CorIQ figure is not re-verified here. | RTL-E2-F21 (evidence/verify/E2.json) |
| C124 | DG Media Network's on-site and off-site solution pages both offer 'Closed-loop reporting and incrementality with our Reporting Dashboard', whose listed sales metrics are Sales Lift, iROAS, Transactions/Buyer Lift, $/Transaction Lift, Conversion Rate and % New-to-Brand Shoppers; lift metrics therefore appear in the documented default dashboard, but no assignment, counterfactual construction, window, match basis or significance rule is stated, and the network's releases describe the same construct as 'closed-loop 1:1 measurement' identifying 'both attribution and incrementality' with point-of-sa | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Both pages: 'Closed-loop reporting and incrementality with our Reporting Dashboard' under 'Insights At Your Fingertips'. Sales metrics: 'Sales Lift', 'iROAS', 'Transactions/Buyer Lift', '$/Transaction Lift', 'Conversion Rate', '% New-to-Brand Shoppers'; engagement: Viewability, Impressions, CPC, CTRs, Site Level, Budget Pacing; the off-site page adds dimensions Brand/Product, Dates, Tactics, Audience, Device Type, Metro/State. [fetch-tool rendering] neither page says whether the lift metrics come with every campaign or only on request or for eligible campaigns, and neither describes a control | supported with qualification | Restate the middle clause: 'lift metrics are listed in "our Reporting Dashboard" on both solutions pages (marketing pages); neither page says every campaign receives them or that they are the default, and the dashboard's listed sales metrics contain no attributed-sales or ROAS metric, so the attribution half of the construct rests on the term "closed-loop reporting" and the releases'. Code lift_in_default_report as 'listed in the dashboard; default status not stated' and keep counterfactual and assignment unknown. | Coding the dashboard listing as 'lift in the default report' is an inference from a marketing page, not a documented default; label it inference. 'iROAS' and 'Sales Lift' are labels, not evidence of a counterfactual. | RTL-E2-F29 (evidence/verify/E2.json) |
| C125 | DG Media Network documents four off-site measurement rails: Meta campaigns with 'closed-loop, attributable store sales' using 'opted-in first-party data' and Meta's Advanced Analytics 'secure environment' (27 March 2023, after testing); Google's Commerce Media Suite bringing DG's 'first-party retail signals' into YouTube and Display & Video 360 with 'SKU-level conversion reporting' described as 'closed-loop attribution' to in-store performance (undated, © 2026); a Kevel and The Trade Desk solution announced 4 May 2026 with testing to start in June 2026 and availability expected in the third qu | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Meta (March 27, 2023): 'First to Market to Offer Closed-Loop Measurement of In-Store Sales.'; 'utilizing our opted-in first-party data,'; 'while measuring closed-loop, attributable store sales.'; 'Meta's Advanced Analytics enables SQL-based custom measurement and analysis in a secure environment.'; 'Using this measurement and analysis, DGMN is able to evaluate the business impact of Meta marketing investments to a particular brand, ... inclusive of quantifiable metrics such as return on ad spend'. [fetch-tool rendering] the release does not say how or where DG data and Meta exposure are joined | supported with qualification | Replace the last sentence with: 'Only the Meta rail names a partner measurement environment (Meta's Advanced Analytics, "SQL-based custom measurement and analysis in a secure environment"), and the release implies rather than states that the store-sales measurement runs there; the Google item names the data route (Commerce Media Suite) and the reporting surface (SKU-level conversion reporting in DV360) but not where the exposure-to-sales join is made; the Kevel and DoorDash items name no join method. No source states match keys, matchable share, modelled components or validation.' | Do not code an identity match outside DG's surface as documented on these sources: the Meta release documents a partner secure measurement environment (implied join location, keys unstated) and the Google item documents only a data-ingestion route and a reporting surface. | RTL-E2-F30 (evidence/verify/E2.json) |
| C126 | eBay Promoted Listings General attributes a sale when any buyer purchases the promoted item within 30 days of any click on its ad, 'regardless of whether the buyer themselves clicked on the ad', with the item promoted at click and at sale; the attributed sale is also the billing event (the ad rate in effect at the time of sale), the definition applies on 'all eBay sites' as of 13 January 2026, and such sales appear as 'Attributed sale' in the 'Sale type' reporting column. | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Attribution page: 'eBay will report an Attributed sale from a general campaign when a buyer purchases the same item' 'that was featured in the ad that was clicked on by any buyer in the most recent 30 days.'; 'The ad fee is charged when a buyer purchases the promoted item from a general ad that any buyer clicked on' 'in the most recent 30 days.'; 'The item must be promoted at the time of click and the time of sale.'; 'The ad fee is based on the ad rate in effect at the time of the sale.'; the page opens 'eBay currently reports sales and charges for ads differently based on the product and list | supported with qualification | Restate the date clause: 'the any-buyer definition took effect on ebay.com, ebay.ca and cafr.ebay.ca on 13 January 2026, having already applied on ebay.co.uk, ebay.fr, ebay.it, ebay.es, ebay.au and ebay.de (date of that earlier change not given on the announcement), so that from 13 January 2026 "all eBay sites" use it; the attribution page itself carries no date and still says eBay reports and charges "differently based on the product and listing site ID".' | The any-buyer rule is on the attribution page itself ('clicked on by any buyer'), not only in the announcement. No view or impression attribution is mentioned for any eBay product. | RTL-E2-F33 (evidence/verify/E2.json) |
| C127 | eBay Promoted Listings Priority reports a 'Direct' sale when a buyer purchases the same item from a clicked priority ad within 30 days and a 'Halo' sale when the buyer then purchases a different item from the same seller within 30 days of that click, charges per click ('The purchased item doesn't need to be promoted, and the seller won't pay any additional ad fees for this subsequent sale'), and from 13 January 2026 holds exclusive access to the first ad slot in search results on ebay.com, ebay.ca and cafr.ebay.ca, with an organic listing shown when no priority listing is relevant. | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | 'eBay will report a Direct sale from a priority campaign when a buyer purchases the same item' 'that was featured in the ad they originally clicked on in the most recent 30 days.'; 'Ad fees are charged on a cost per click basis.'; 'eBay will report a Halo item sale from a priority campaign when a buyer clicks on a listing' 'promoted with a priority strategy, then purchases a different item from the same seller within 30 days of that click.'; 'The purchased item doesn't need to be promoted, and the seller won't pay any additional ad fees for this subsequent sale,' 'as charges are still only cha | supported with qualification | Make the buyer condition explicit: a Direct sale requires the purchaser to buy 'the same item that was featured in the ad they originally clicked on' within 30 days (the clicker and the purchaser are the same buyer, unlike the general strategy's any-buyer rule); a Halo sale is a different item from the same seller bought by that clicking buyer within 30 days. | ebay.de received the same first-slot exclusivity from 26 February 2025 (hub page). | RTL-E2-F34 (evidence/verify/E2.json) |
| C128 | eBay Promoted Offsite promotes listings 'in external search channels' (Google is the channel named) on CPC only and reports Direct and Halo sales on the same 30-day click rule as priority campaigns; the documented join is a click on the off-site ad followed by a purchase on eBay, so no identity match outside eBay's own surface is documented for this off-site row. | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Attribution page: Direct sale 'when a buyer purchases the same item' 'that was featured in the ad they originally clicked on in the most recent 30 days.'; Halo 'when a buyer clicks on a listing' 'promoted using Promoted Offsite, then purchases a different item from the same seller within 30 days of that click.'; 'Ad fees are charged on a cost per click basis.' Hub: 'Highlight your eligible listings on popular external channels, like Google.'; 'shown to buyers who start their shopping journey off of eBay.' API: 'promote their eBay listings in external search channels'; 'Promoted Offsite campaig | supported with qualification | Replace the fetch-tool rendering for RTL-E2-S06 with the verbatim 'Highlight your eligible listings on popular external channels, like Google.' and add that Direct and Halo sales both require the purchaser to be the buyer who clicked ('the ad they originally clicked on'; 'that click'). | The absence of an identity-match statement is bounded to these three eBay pages; the help centre (verification wall) is unread. | RTL-E2-F36 (evidence/verify/E2.json) |
| C129 | Etsy attributes an order to Offsite Ads when a buyer clicks an Etsy-run off-site ad and purchases from the seller's shop within 30 days, and the attribution is the billing event: a 15% fee on attributed orders, 12% for shops with $10,000 or more in sales over any 365 days, capped at $100 per order, with enrolment mandatory for the lifetime of a shop once it crosses that threshold; partners include search engines and social platforms that 'may change over time'. | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | 'Orders are attributed to Offsite Ads when a buyer clicks on an ad and completes a purchase from your shop within 30 days.'; 'If the buyer's final click before purchasing is through an Etsy Ad instead, only the Etsy Ads fee will apply.'; 'Charged a 15% fee for orders attributed to an Offsite Ad'; 'Charged a discounted 12% fee for orders attributed to an Offsite Ad'; 'The Offsite Ads fee will never exceed $100 USD, regardless of the order total for an order attributed to an Offsite Ad.'; 'If your shop has made $10,000 USD or more in any consecutive 365-day period, participation in Offsite Ads i | supported with qualification | Add the documented precedence rule: 'If the buyer's final click before purchasing is through an Etsy Ad instead, only the Etsy Ads fee will apply' (a last-click rule between Etsy Ads and Offsite Ads for the fee; the page does not say whether reporting follows the same rule). Record it in attribution_parameters (touch_model and deduplication) for Etsy Offsite Ads and Etsy Ads. | The threshold is shop 'revenue within any consecutive 365-day period'; 'sales' in the finding is acceptable shorthand. | RTL-E2-F41 (evidence/verify/E2.json) |
| C130 | Hy-Vee RedMedia documents 'effective offline and online sales attribution', 'post-campaign performance readouts and real-time analytics' and a 'pay-per-conversion model' for on-site paid search, so on that placement the attributed conversion is also the billing event; no window, match basis, conversion definition or lift design is stated, and its home page carries one brand case described by the fetch tool as more than half incremental, with no design. | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Solutions page structure: under 'On-Site.' / 'Industry-leading paid search capabilities.': 'In fact, RedMedia offers ROAs well above the national average.' and 'Our radical pay-per-conversion model is shaking up the industry' (sentence continues); 'conversion' is not defined. Under 'Omnichannel Measurement.': 'Through post-campaign performance readouts and real-time analytics, we uncover data-heavy insights' (continuing to customer data and 'effective offline and online sales attribution'). Under 'Excellent Results.': 'Digitally engaged shoppers are 15% more likely to convert.' Home page: 'tra | supported with qualification | Restate the billing clause as inference: 'its On-Site section, headed "Industry-leading paid search capabilities", describes a "pay-per-conversion model" without defining a conversion or saying it is the attributed sale, so treating the attributed conversion as the billing event is an inference'. Replace the fetch-tool rendering with the verbatim 'More than half of sales generated through these campaigns are incremental to their brand.' and add the home page's 'unified attribution across channels'. | Billing-equals-attribution for Hy-Vee is inference, not documented; the incremental-share sentence is an asserted result with no design and no lift offer. | RTL-E2-F44 (evidence/verify/E2.json) |
| C131 | Across the 27 workstream E2 rows, 21 document a default outcome construct and 6 do not (Chewy Ads, Kohl's, Wayfair, Gopuff, Staples, Leap); of the 21 documented, 19 are attribution with no counterfactual (14 with at least one parameter, 5 by term only), one (DG Media Network) lists lift metrics in its default dashboard and one (Gulp) describes attribution and lift without naming a default. At network level, with the frozen representative services, 15 of 21 networks document a default and 13 of those are attribution-only; if term-only rows are treated as undocumented, 16 of 27 rows and 10 of 21 | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Arithmetic re-computed from h1a_coding: A = 14 rows, B = 5, C = 1, D = 1, E = 6 (27); documented 21, attribution 19 (A+B). Network level (21 representatives): documented 15, attribution 13; strict reading (B treated as undocumented): 16 rows / 14 attribution-only and 10 networks / 8 attribution-only. All figures match the finding. | supported with qualification | Add three caveats: (1) DG Media Network's code C rests on a dashboard metric list on marketing pages that does not say lift is the default (see F29); (2) the E/B boundary is applied inconsistently: Chewy Ads ('Closed loop measurement' bullet, 'LTV ROAS', NTB) is E while DICK'S Media ('closed-loop measurement') is B, and Leap ('Measurable Analytics, including ROAS') is E while Petco ('ROAS, Sales' dashboard) is B; recoding Chewy and Leap to B gives 23 documented rows with 21 attribution-only (91%) and 17 networks with 15 attribution-only; (3) under the strict reading the network-level documente | The verdict concerns documented defaults; under every coding variant above the attribution share among documented defaults stays above two-thirds, so the H1a direction does not change; only the strict network-level disclosure gate is at risk. | RTL-E2-F53 (evidence/verify/E2.json) |
| C132 | Among the 27 E2 rows, 4 document a controlled-lift or lift offer on their own pages (Best Buy Ads Sponsored Products; Best Buy Ads display, video, search and CTV media; Orange Apron Media; DG Media Network), Chewy Ads names 'incrementality measurement' without a design, Gulp describes lift reporting in the future tense, Lowe's Ipsos measurement rests on reporting only, and the rest document none on the pages screened; no row documents a request rule, eligibility threshold, minimum spend, sample size or significance rule, no row documents assignment for any offer, and only DG Media Network plac | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Best Buy: geo experimentation and 'customer-level holdouts', no request or eligibility rule. OAM: 'advanced studies such as incrementality, geo lift, and brand lift', 'built around specific business questions'. Gulp: 'Control/exposed methodology will enable lift reporting'. DGMN: lift metrics listed in 'our Reporting Dashboard', default status not stated. Lowe's (reported): iROAS reporting 'to top spenders'. | supported with qualification | Two edits: (1) 'no row documents a request rule, eligibility threshold ...' should read 'no row documents a request rule, minimum spend, sample size or significance rule on its own pages; the only eligibility statement is reported: Lowe's iROAS reporting "to top spenders" (Marketing Dive)'; (2) 'only DG Media Network places lift metrics in its default report' should read 'only DG Media Network lists lift metrics in its Reporting Dashboard, without saying they are the default'. | RTL-E2-F54 (evidence/verify/E2.json) | |
| C133 | Of the E2 rows with off-site placements, 9 document off-site outcome measurement in a class-specific sentence (Chewy Ads; Best Buy Ads display, video, search and CTV media; CVS Media Exchange; Walgreens Advertising Group; DG Media Network; eBay Promoted Offsite; Etsy Offsite Ads; DICK'S Media; UB Media); only DG Media Network names a partner environment in which the off-site join is made (Meta's Advanced Analytics; Google's Commerce Media Suite), without match keys, and eBay Promoted Offsite and Etsy Offsite Ads document a click-through join to the operator's own site rather than an identity m | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Meta release: 'Meta's Advanced Analytics enables SQL-based custom measurement and analysis in a secure environment.' and 'Using this measurement and analysis, DGMN is able to evaluate the business impact of Meta marketing investments'; no join location or keys stated. Google item: first-party retail signals brought 'into YouTube and Display & Video 360 (DV360) through Google's Commerce Media Suite' and 'SKU-level conversion reporting in DV360'; join location not stated. eBay and Etsy: click rules only, no cookie, account or identity wording. | supported with qualification | Restate the DGMN clause: 'only DG Media Network names a partner environment, and only for the Meta rail (Meta's Advanced Analytics, a "secure environment" in which the release implies the measurement runs, keys unstated); its Google rail names a data route and a reporting surface (Commerce Media Suite; DV360), not a join location'. With DGMN at most 'implied' and eBay and Etsy documenting click-through joins, at most 1 of 9 rows (11%) documents an identity match outside the operator's surface. | Under H4b's rule, several of the nine rows rest on one or two marketing pages (Chewy, WAG, UB Media, DICK'S); the brief makes H4b unresolved when 'any linkage search for a documented row is incomplete or its source inaccessible', so the session must state whether those searches count as complete before reading the low share as 'contradicted'. | RTL-E2-F55 (evidence/verify/E2.json) |
| C134 | For four E2 rows the attributed outcome is also the invoice: eBay Promoted Listings General charges its ad fee on the attributed sale, Etsy Offsite Ads charges a percentage fee on attributed orders, Vrbo Sponsored Listings charges per booked night after a click, and Hy-Vee RedMedia bills on-site paid search per conversion; in each case the operator writes the attribution rule, reports the outcome and bills on it, and none of the four documents a dispute or restatement procedure for attributed events on the pages screened. | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | eBay: 'The ad fee is charged when a buyer purchases the promoted item from a general ad that any buyer clicked on' 'in the most recent 30 days.' Etsy: 'Charged a 15% fee for orders attributed to an Offsite Ad'. Vrbo: 'you only pay a fee when a guest who clicks on your ad actually books their stay at your property.' Hy-Vee: 'Our radical pay-per-conversion model is shaking up the industry' under 'On-Site.' / 'Industry-leading paid search capabilities.'; conversion undefined. | supported with qualification | Keep eBay General, Etsy Offsite Ads and Vrbo as documented. Mark Hy-Vee as inference: 'Hy-Vee RedMedia describes a pay-per-conversion model in its on-site paid-search section without defining a conversion or tying it to its sales attribution'. | Hy-Vee's inclusion is inference, not documentation. | RTL-E2-F56 (evidence/verify/E2.json) |
| C135 | Across the 27 E2 rows, the bounded governance screen found no restatement or correction policy, no documented separation between the sales organisation and the measurement function, and no methodology version history; the only dispute term is Macy's Media Network's clause that its ad-server reporting is final and not subject to challenge, the only dated metric revision is eBay's general-attribution change of 13 January 2026, two networks claim IAB alignment without naming the document (CMX, DG Media Network), and the only reported third-party incrementality partner (Ipsos for Lowe's) rests on | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Macy's: 'will be final and not subject to challenge unless otherwise required by applicable law.' eBay announcement: from January 13, 2026 on ebay.com, ebay.ca, cafr.ebay.ca; 'is already applied to items listed on ebay.co.uk, ebay.fr, ebay.it, ebay.es, ebay.au and ebay.de'. CMX: 'align our methodology with the IAB's measurement guidelines' (document not named); 'We periodically update our ad policy.' (undated). Marketing Dive: Ipsos 'independent incrementality measurement'. | supported with qualification | Add after 'the only dated metric revision is eBay's general-attribution change of 13 January 2026': 'which the same announcement says had already been applied, on an undated earlier occasion, to six non-North American eBay sites'. | DGMN's 'IAB-aligned reporting' (home page) was not re-opened. | RTL-E2-F57 (evidence/verify/E2.json) |
| C136 | E2 networks publish audience and membership counts in incompatible units: CMX '90M+ addressable shoppers', UB Media '46 million+' Rewards members, Gulp 'over 100 million measurable loyalty members' on one page and '95 million loyalty members' in a 2024 release, DG Media Network 'more than 90 million unique customer profiles', Leap '1.5 million myPerks households' and Staples '38 million reachable shoppers'; none is a share of transactions matchable to ad exposure, and none states the matchable share. | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Leap: '1.5 million myPerks households' and 'a general reachable audience of 14+ million'. DGMN: 'DG's more than 90 million unique customer profiles across the Meta ecosystem'. | supported with qualification | Add Leap's 'general reachable audience of 14+ million' (a second, different unit on the same page) and qualify DGMN's 90 million as 'unique customer profiles across the Meta ecosystem' (2023 Meta release). Add both to audience_units. | The CMX, UB Media, Gulp and Staples figures were not re-opened. | RTL-E2-F59 (evidence/verify/E2.json) |
| C137 | match_basis 'unknown for direct; the clicking buyer for halo ("that click")' | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Direct: 'when a buyer purchases the same item' 'that was featured in the ad they originally clicked on in the most recent 30 days.' Halo: 'when a buyer clicks on a listing' ... 'then purchases a different item from the same seller within 30 days of that click.' | supported with qualification | match_basis: 'the clicking buyer, for both direct ("the ad they originally clicked on") and halo ("that click")'. Window 30 days click; halo yes; CPC billing confirmed. | TABLE attribution_parameters: eBay Promoted Listings Priority (evidence/verify/E2.json) | |
| C138 | window 30; match basis 'click-through to eBay (mechanism not stated)'; halo yes | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Direct: 'the ad they originally clicked on in the most recent 30 days'; Halo: 'within 30 days of that click'; CPC only; channel example 'like Google'. | supported with qualification | match_basis: 'the clicking buyer (direct and halo); how the off-site click is linked to the eBay purchase is not stated'. | TABLE attribution_parameters: eBay Promoted Offsite (evidence/verify/E2.json) | |
| C139 | Etsy Ads: 30 days, interaction defined as a click, any item in shop, touch model and dedup unknown. Etsy Offsite Ads: 30 days click, touch model unknown, dedup unknown, fee 15%/12%, cap US$100 | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Etsy Ads: 'Orders are attributed to ads when someone interacts with an ad and then buys any item in your shop within 30 days.'; views counted, 'You aren't charged for views.' Offsite: 'Orders are attributed to Offsite Ads when a buyer clicks on an ad and completes a purchase from your shop within 30 days.'; 'If the buyer's final click before purchasing is through an Etsy Ad instead, only the Etsy Ads fee will apply.' | supported with qualification | Etsy Offsite Ads touch_model / deduplication: 'last click decides which fee applies between Etsy Ads and Offsite Ads ("If the buyer's final click before purchasing is through an Etsy Ad instead, only the Etsy Ads fee will apply"); whether reporting follows the same rule is not stated'. Mirror the note on the Etsy Ads row. Windows, fees and cap confirmed. | TABLE attribution_parameters: Etsy Ads and Etsy Offsite Ads (evidence/verify/E2.json) | |
| C140 | on/off: touch model unknown ('beyond last-click attribution'), in-store yes, '% New-to-Brand Shoppers' reported; off-site: match basis 'partner environment named (Meta Advanced Analytics; Google Commerce Media Suite); keys not stated' | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Dashboard sales metrics: Sales Lift, iROAS, Transactions/Buyer Lift, $/Transaction Lift, Conversion Rate, % New-to-Brand Shoppers (no attributed-sales or ROAS metric listed). Meta: Advanced Analytics 'SQL-based custom measurement and analysis in a secure environment'. Google: signals brought in 'through Google's Commerce Media Suite'; 'SKU-level conversion reporting in DV360'. Kevel: 'beyond last-click attribution' describes the planned solution, not the current dashboard. | supported with qualification | Off-site match_basis: 'Meta rail: partner secure measurement environment named (Advanced Analytics), join location implied, keys not stated; Google rail: data route (Commerce Media Suite) and reporting surface (DV360) named, join location not stated'. On/off row touch_model: cite the Kevel phrase as describing the announced Kevel and The Trade Desk solution only. | TABLE attribution_parameters: DG Media Network (on-site and off-site; off-site Meta and YouTube/DV360) (evidence/verify/E2.json) | |
| C141 | network-wide; on-site paid search billed per conversion; in-store yes ('offline and online sales attribution'); conversion undefined | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Pay-per-conversion sentence sits under 'On-Site.' / 'Industry-leading paid search capabilities.'; 'effective offline and online sales attribution' under 'Omnichannel Measurement.'; home page 'unified attribution across channels'. | supported with qualification | Label the placement link as inferred from page layout ('on-site paid search, per the section heading; conversion undefined; not stated to be the attributed sale'). | TABLE attribution_parameters: Hy-Vee RedMedia (evidence/verify/E2.json) | |
| C142 | default_target_contrast 'attributed and incremental (both in the default Reporting Dashboard)'; lift_in_default_report 'yes (as listed)' | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | 'Closed-loop reporting and incrementality with our Reporting Dashboard'; lift metrics listed; neither page says the metrics are standard for every campaign or available only on request; no counterfactual described. | supported with qualification | default_target_contrast: 'incremental metrics listed in the Reporting Dashboard (counterfactual unknown) beside the term "closed-loop reporting"; default status not stated'; lift_in_default_report: 'listed in the dashboard; whether every campaign receives it is not stated (inference)'. | Inference from a marketing page, not a documented default. | TABLE default_method: DG Media Network (evidence/verify/E2.json) |
| C143 | eligibility 'unknown (reported paraphrase mentions top spenders; unverified)' | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | 'The media network claims to be among the first from a retailer to provide incremental return on ad spend reporting to top spenders.' | supported with qualification | eligibility: 'reported: iROAS reporting "to top spenders" (Marketing Dive, 13 October 2026); threshold not stated'. lift_offer_documented stays 'reported only'. | TABLE default_method: Lowe's Media Network (evidence/verify/E2.json) | |
| C144 | Hy-Vee lift: 'not documented (one case claim of incremental sales, fetch-tool paraphrase)'; OAM request_required unknown | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Hy-Vee home: 'More than half of sales generated through these campaigns are incremental to their brand.' OAM post: 'offers advanced measurement solutions built around specific business questions'. | supported with qualification | Replace the Hy-Vee paraphrase with the verbatim sentence. For OAM keep request_required unknown but add the 'built around specific business questions' wording as a pointer to bespoke studies. | TABLE default_method: Hy-Vee RedMedia and Orange Apron Media (evidence/verify/E2.json) | |
| C145 | Chewy E, Leap E, Petco B, DICK'S B, DGMN C, Gulp D | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Chewy: bullet 'Closed loop measurement'; 'LTV ROAS'; 'New-to-Brand metrics and incrementality measurement'. DICK'S: 'we track every touchpoint with closed-loop measurement'. Leap: 'Measurable Analytics, including ROAS'. Petco: dashboard 'including ROAS, Sales, CTR, ROI'. DGMN: dashboard lists lift metrics, default not stated. Gulp: both constructs in the future tense. | supported with qualification | Apply one rule to the E/B boundary: either code Chewy B (term 'Closed loop measurement' plus ROAS-type metric) or D (attribution term plus named incrementality) and Leap B (ROAS named), matching DICK'S and Petco; or code all four E. DGMN's C should carry the label 'dashboard listing, default not stated'. | Representative selection checked: eBay General over Priority and Expedia over Vrbo follow the alphabetical tie rule; counts re-computed and match F53. | TABLE h1a_coding: coding consistency (Chewy, Leap, DG Media Network, Gulp) (evidence/verify/E2.json) |
| C146 | 'Reddit via server-to-server DV and IAS' | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | '†Reddit offers direct S2S integrations with advertisers with existing DV & IAS relationships (No tags accepted)' | supported with qualification | verification_tags: '... Reddit: no tags accepted; Reddit offers direct S2S integrations to advertisers with existing DV and IAS relationships'. Export 'detailed, downloadable data tables' and the consent-gated use clause are confirmed. | TABLE access_record: CVS Media Exchange verification_tags (evidence/verify/E2.json) | |
| C147 | 'partly for Meta campaigns: SQL-based custom measurement and analysis in a secure environment (Meta Advanced Analytics); who writes the queries is not stated' | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | 'Meta's Advanced Analytics enables SQL-based custom measurement and analysis in a secure environment.'; 'Using this measurement and analysis, DGMN is able to evaluate the business impact of Meta marketing investments to a particular brand'. | supported with qualification | h3_axis_b: 'not documented for buyers; the release describes DGMN running the analysis ("DGMN is able to evaluate the business impact") in Meta's environment'. Axis b is buyer-controlled outcome analysis, and the source names DGMN, not the advertiser, as the analyst. | TABLE access_record: DG Media Network h3_axis_b (evidence/verify/E2.json) | |
| C148 | eBay General/Priority api_reporting 'yes: Marketing API reporting service'; eBay Offsite 'yes: Promoted Offsite campaigns covered by the Marketing API'; OAM clean room 'named; provider not named' | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | API page: 'five different types of reports' (types not named), getReport, ad_report_task; Promoted Offsite mentioned as CPC-only campaigns; Promoted Stores not mentioned. OAM measurement page: 'Sharing data with publishers and other partners in a neutral and privacy-enhanced environment'. | supported with qualification | eBay Promoted Offsite api_reporting: 'campaigns named on the Marketing API overview (CPC only); whether the five report types cover them is not stated'. OAM clean_room: add the measurement page's sentence beside the post's 'Clean Room environments'. | TABLE access_record: eBay rows and Orange Apron Media (evidence/verify/E2.json) | |
| C149 | identity_linkage 'partner environment named for Meta and Google; keys not stated'; identity_match_outside_operator_surface_documented 'partly (environment named, keys not stated)' | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Meta: secure measurement environment named; DGMN's evaluation 'using this measurement and analysis'; join not described. Google: 'brings Dollar General's first-party retail signals directly into YouTube and Display & Video 360 (DV360) through Google's Commerce Media Suite'; 'SKU-level conversion reporting in DV360'; join not described. Kevel: 'decisioned and measured together in a single, coherent framework'; no identity wording. Off-site page: 'Leverage DG's first-party data'; no identity, clean-room or partner-environment method. | supported with qualification | identity_linkage: 'Meta (2023): partner secure measurement environment named, join location implied, keys not stated; Google: data route and reporting surface named, join location not stated; Kevel and The Trade Desk, DoorDash: not stated'. identity_match_outside_operator_surface_documented: 'implied for the Meta rail only; not documented for the other rails'. Exposure source and transaction source as coded; modelled components and validation unknown. | Treating the Meta environment as a documented identity match would be inference; label it so if the session counts it. | TABLE offsite_loop_fields: DG Media Network (evidence/verify/E2.json) |
| C150 | offsite_rail 'Yahoo DSP; Google, Meta, Pinterest; Roku and others (reported)'; modelled 'marketing-mix modelling (reported, Ipsos)'; validation 'Ipsos (reported)' | Refute-mode verifier for evidence/research/E2.json re-opened the cited source. | Marketing Dive names expansion partners 'Roku, DoorDash, Reddit, TikTok and YouTube'; Yahoo DSP, Google, Meta and Pinterest are not named as Lowe's partners in the article (Meta, Google and Amazon appear only as platforms that judge campaigns on their own systems). Ipsos provides 'independent incrementality measurement' 'including bespoke marketing mix modeling and agile attribution'. | supported with qualification | Cite Yahoo DSP, Google, Meta and Pinterest to RTL-U-S053 only (not re-opened here) and add 'DoorDash, Reddit, TikTok and YouTube (reported, Marketing Dive)'. Validation: Ipsos is reported as the measurement provider, not as a validator of Lowe's numbers; code validation 'not documented' and keep Ipsos under modelled components and reporting party (reported). | TABLE offsite_loop_fields: Lowe's Media Network (evidence/verify/E2.json) | |
| C151 | PayPal Ads announced on 6 January 2026 'Transaction Graph Measurement', 'PayPal's first-party measurement suite built from the ground up', covering 'campaign reporting, insights, attribution, and incrementality' and 'immediately available for US-based programs', with outside measurement partners named (AppsFlyer, Cint, Experian, iSpot, Kantar, Kochava and LiveRamp, per workstream U's reading); the release documents no attribution window, design, assignment, eligibility or default status for the incrementality component. | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | Dateline "San Jose, Calif., Jan. 6, 2026 /PRNewswire/"; "PayPal's first-party measurement suite built from the ground up"; "utilizing the PayPal Transaction Graph and deterministic-driven identity for campaign reporting, insights, attribution, and incrementality" (fragment, sentence longer); "immediately available for US-based programs with availability in the United Kingdom and Germany to follow."; the Measurement Partnership Program is to "enable independent campaign validation with a certified list of third-party partners": "AppsFlyer, Cint, Experian, iSpot, Kantar, Kochava, LiveRamp"; Luci | supported with qualification | Add after the partner list: the release says the suite uses "deterministic-driven identity" and that the partner programme exists to "enable independent campaign validation with a certified list of third-party partners" (AppsFlyer, Cint, Experian, iSpot, Kantar, Kochava, LiveRamp, read directly by this verifier, not only per workstream U); it documents no window, design, assignment, eligibility or default status for the incrementality component, nor what each partner validates. | access_record (PayPal) 'roles not stated' should read 'programme purpose stated ("independent campaign validation"); per-partner roles not stated'. attribution_parameters (PayPal) identity: add the suite-level claim "deterministic-driven identity" (method undescribed). | RTL-E3-F09 (evidence/verify/E3.json) |
| C152 | PayPal's only campaign result on the pages opened is a during-campaign comparison with no stated control: 'Ulta saw a 20% increase in transaction spend via PayPal during the campaign' (release of 6 January 2026). | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | "Compared to the weeks prior to the campaign, Ulta saw a 20% increase in transaction spend via PayPal during the campaign." Footnote 1: "PayPal Internal Data, Jun-Jul 2025." | contradicted | PayPal's only campaign result on the pages opened is a pre/post comparison with no unexposed control: 'Compared to the weeks prior to the campaign, Ulta saw a 20% increase in transaction spend via PayPal during the campaign' (release of 6 January 2026; footnote: PayPal internal data, June to July 2025). The baseline is the pre-campaign weeks; no comparison with unexposed users is reported. | measurement_record: counterfactual 'none stated' should be 'pre/post (weeks prior to the campaign)'; target_contrast 'descriptive (pre/post)'; confidence_reason 'no baseline, period or design' is wrong on baseline and period. vendor_reported_outcome_figures (PayPal row): comparator 'not stated' -> 'weeks prior to the campaign'; design 'none' -> 'pre/post; no unexposed control'; period '2025 campai | RTL-E3-F12 (evidence/verify/E3.json) |
| C153 | Cardlytics' identity basis is anonymized card-account purchase data supplied by bank partners: its fiscal 2025 10-K says its partners are 'predominantly financial institutions ("FI partners")', that 'Today, our FI partners include JPMorgan Chase Bank, National Association ("Chase")', that marketing 'is targeted and measured based on actual purchase data at a customer and account level', and that it 'analyzed approximately $5.7 trillion in purchases across all categories and geographies, both online and in-store'; its website says 'we see over $4.1T in consumer spend', and no matchable share of | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | The full sentence (fetch-tool segments): "Today, our FI partners include JPMorgan Chase Bank, National Association ("Chase")," "Wells Fargo Bank, National Association ("Wells Fargo")," "and American Express Travel Related Services Company, Inc. ("American Express")," "as well as many other national and regional financial institutions, financial technology companies and virtual-only banks." Also "The partners for the Cardlytics platform are predominantly financial institutions"; "targeted and measured based on actual purchase data"; "approximately $5.7 trillion in purchases across all categorie | supported with qualification | Replace "Today, our FI partners include JPMorgan Chase Bank, National Association ('Chase')" with the full list: Chase, Wells Fargo and American Express Travel Related Services Company, "as well as many other national and regional financial institutions, financial technology companies and virtual-only banks"; delete the limitation that the other partners were not captured. | Consequence: American Express, like Chase, is both a Cardlytics FI partner and the seller of its own cohort B row (Amex Offers). Add it to the Chase tension in counterevidence_and_tensions; how offers in Amex's channels divide between Amex Offers and Cardlytics is not documented. | RTL-E3-F17 (evidence/verify/E3.json) |
| C154 | Fetch documents a randomised holdout as its offer-measurement method: 'Fetch randomly assigns users into either "treatment" or "control" groups', serves offers to the treatment group and 'withholds those offers from the control group', measures sales, units or trips from user-submitted receipts ('Fetch does not model results'), and divides incremental revenue by the amount invoiced to the partner; these design statements date from February and June 2023, and no 2026 page opened says every Special Offers campaign carries a holdout or that lift appears in a default report. | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | 3 February 2023 post: "Our methodology uses randomized holdouts—the most rigorous standard in the control vs. exposed methodology—"; "Bias is minimized through random assignment"; "Minimizing bias in treatment and control groups, achieved by creating statistically significant group sizes,"; "Fetch does not model results—rather, we use receipts submitted by our users to measure real impact."; ROAS "is calculated by dividing the observed revenue by the cost of the promotion." No window, eligibility or default rule. 8 June 2023 release: "Fetch randomly assigns users into either "treatment" or "co | supported with qualification | Replace the H1b flag "minimum sample or significance rule not stated ('small but representative' control groups)" with "no numeric minimum is published; the February 2023 post says bias is minimised by 'creating statistically significant group sizes' and the June 2023 post that partners 'can create small but representative control groups'". Add the support passage for the source term 'Fetch Verified Incremental' (RTL-E3-S17: "Fetch Verified Incremental is not a modeling technique, but rather an empirical analysis."). The 2026 bound stands: no 2026 page opened says what a Special Offers buyer r | assignment 'random (user)' is disclosed (KPMG release), so the coding is correct; the source_term 'Fetch Verified Incremental' had no quoted passage in E3. | RTL-E3-F20 (evidence/verify/E3.json) |
| C155 | MARRIOTT MEDIA claims closed-loop attribution without documenting it: its launch release (12 June 2025) promises 'campaign performance with closed loop attribution' and 'robust measurement capability across the marketing funnel', and its home page (browser pane, 3 November 2026) shows partner-case metrics (site visits, sign-ups, brand consideration and awareness, 'YoY Gatorade Volume growth vs Other Benchmark Products') but no window, transaction source for an advertiser's sales, match basis or control. | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | Release: "BETHESDA, Md., June 12, 2025 /PRNewswire/"; "will offer marketers campaign performance with closed loop attribution" (fragment of a longer sentence pairing messaging design and audience selection with "robust measurement capability across the marketing funnel"); "200+ targetable attributes across the nearly 237-million member Marriott Bonvoy program"; no transaction source, partner, holdout or control; a spokesperson quote in the release names "brand lift and awareness studies" alongside attributable conversions and advanced modelling. Home page (browser-pane notes): "We are Marriott | supported with qualification | Add: the launch release states the attribution in the future tense ('will offer marketers campaign performance with closed loop attribution') and, in a spokesperson quote, names 'brand lift and awareness studies' beside attributable conversions and modelling, so the outcome construct behind 'closed loop attribution' is undetermined. | measurement_record target_contrast 'attributed (source term ...)' imports a construct the release does not fix; code 'undetermined (source term: closed loop attribution; brand lift studies and modelling also named)'. Same for the default_method row. | RTL-E3-F32 (evidence/verify/E3.json) |
| C156 | American Express reports Amex Offers' scale as spend and enrolments, not advertising revenue: 'In 2025, the Amex Offers program drove $24 billion in spend worldwide (inclusive of $21 billion in the U.S.) at merchants who accept American Express', and 'In 2025, there were 464M enrollments in Amex Offers by U.S. Card Members'; an enrolment counts offer sign-ups, not people. | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | Browser-pane notes: "In 2025, the Amex Offers program drove $24 billion in spend worldwide (inclusive of $21 billion in the U.S.) at merchants who accept American Express."; "In 2025, there were 464M enrollments in Amex Offers by U.S. Card Members."; footnote 1 cites "Nilson Report #1,301, February 2026". | supported with qualification | Mark 'an enrolment counts offer sign-ups, not people' as inference from the wording ('enrollments in Amex Offers by U.S. Card Members'); the page does not define an enrolment. | Inference stated as fact in the claim sentence. | RTL-E3-F37 (evidence/verify/E3.json) |
| C157 | Klarna's Search & Compare terms (revised 23 July 2025, effective 1 September 2025) define a click-priced placement whose default report is delivery: Klarna will 'provide estimated live statistics for Klarna and its affiliates for Clicks and advertising impressions', supplies conversion-tracking code only where payment depends on purchases on the partner's site, and sells extra data as 'Analytics Premium'. | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | "Revised and republished: July 23, 2025"; "This will take effect from September 1, 2025"; "“CPC” means cost per Click."; Klarna will "provide estimated live statistics for Klarna and its affiliates for Clicks and advertising impressions"; "Klarna does not warrant or guarantee clicks, click rates or ability to convert the clicks into sales."; "If payment for any of the Services is based upon a User completing a purchase on the Partner Site" ... "Klarna shall provide Partner with code for such conversion tracking"; "Analytics Premium is the paid interface of Analytics Data with extended access t | supported with qualification | Replace the limitation 'no correction procedure was found in workstream U's reading' with: the terms make Klarna's tracking count binding ('in no event ... will data provided by Klarna constitute final billing numbers'; only invoices are final), deem an invoice valid unless the partner objects within eight days, disclaim liability for 'incorrect or erroneous data', and do not guarantee clicks or 'ability to convert the clicks into sales'; the live statistics are 'estimated' but the billing count is final. | The coding 'delivery by default' holds: the default price is CPC and the default statistics are clicks and impressions; purchase-based payment with Klarna's tracking code is an option. | RTL-E3-F40 (evidence/verify/E3.json) |
| C158 | Upside's default and billing measure is incremental and built from matched non-users: for each user who claims an offer it finds '10 to 100 non-Upside users with similar purchasing histories' and treats the spend difference as incremental; its contract methodology (July 2023) builds the control from 'control payment identifiers' with similar spend, subtracts a pre-period difference (difference-in-differences), measures per user and merchant network through 7 days after the user's latest Upside transaction, caps the result at Upside-attributed sales and at zero when negative, and computes the p | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | Methodology page: "We measure impact using a test-versus-control approach."; "Each customer who uses an Upside offer is compared to similar customers who didn't"; "Next, we find 10 to 100 non-Upside users with similar purchasing histories."; "The difference is deemed incremental and attributed to Upside."; "Retailers only pay for proven, incremental results — never for sales they would have received anyway."; "Because Upside uses a profit-sharing model, it only makes money when retailers do."; "All program costs are based on measured results—not projections or estimates." Contract methodology | supported with qualification | Restate the end of the claim: '... and computes its fee as the agreed profit-share percentage of incremental gross profit minus promotional costs (zero where incremental gross profit does not cover the user's cash back, with Upside funding the shortfall); the pages do not state assignment, and the test group is defined as shoppers who claim an offer, so assignment is non-random by construction (inference from the design).' | assignment 'non-random (shopper self-selects by claiming an offer)' is an inference from the design, not a disclosed allocation; label it so. The zero floor is written per user and network ("at a user level for each network"), so the upward-bias inference in the limitation is better grounded than 'not stated in the rendered extract' suggests, though still not stated outright. | RTL-E3-F46 (evidence/verify/E3.json) |
| C159 | Across the 20 cohort B rows, this stream codes the documented default outcome method (3 November 2026) as: attributed or offer-linked outcomes with no counterfactual for 6 rows (Chase Media Solutions; PayPal Ads; Cardlytics card-linked offers; Ibotta Performance Network offers; Amex Offers; Rokt Ads); attributed by source term only, with no parameters, for 2 (MARRIOTT MEDIA; Expedia Group Advertising for non-endemic brands); incremental by a documented design for 2 (Fetch Special Offers; Upside merchant promotions); attribution and incrementality both claimed without a design for 1 (Citi Comme | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | Arithmetic: 6 + 2 + 2 + 1 + 2 + 6 + 1 = 20 rows, matching the default_method table. Judgement calls re-examined against the pages: PayPal's two releases name "attribution, and incrementality" (suite, 6 January 2026) and say "Attribution is then closed back to verified PayPal transactions" (Curated Ads, 22 April 2026), with no window, touch model, metric definition, default report or billing unit; Citi's release names "closed-loop measurement" to "prove return on ad spending and quantify incremental impact" and cites "first-party transaction data", likewise with no parameters. Chase, by contras | supported with qualification | Apply one rule: a row is 'attributed (documented)' only when a default report, metric definition or billing unit is documented. Under it PayPal moves to the Citi category ('attribution and incrementality both claimed, no design'), giving: attributed or offer-linked 5 (Chase, Cardlytics, Ibotta IPN, Amex Offers, Rokt); both claimed, no design 2 (PayPal, Citi); closed loop named without parameters 2 (MARRIOTT MEDIA, Expedia; construct undetermined, not 'attributed'); incremental by documented design 2 (Fetch, Upside); delivery only 2; not documented 6; inaccessible 1. If E3 prefers to keep PayPa | Category label 'attributed by source term only' for MARRIOTT MEDIA and Expedia should be 'closed loop named without parameters (construct undetermined)'. | RTL-E3-F50 (evidence/verify/E3.json) |
| C160 | Eight of the 20 cohort B rows document some sales or conversion lift measurement (Chase Media Solutions, PayPal Ads, Cardlytics card-linked offers, Fetch Special Offers, Ibotta Performance Network offers, Upside merchant promotions, Citi Commerce Media, and Lyft Ads through named partners), but only Fetch discloses random assignment (2023 posts) and only Upside discloses how its comparison group is built (matched non-users, non-random); Cardlytics' framework is on request outside the self-serve interface, Ibotta's LiveLift is limited to eligible clients, and Chase, PayPal, Citi and Lyft name l | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | Fetch discloses random assignment ("Fetch randomly assigns users into either "treatment" or "control" groups"); Upside discloses a matched control in operational detail (10 to 100 non-users; 12-month pre-period spend; pre-period difference subtracted); Circana's Ibotta meta-study names "matched buyer groups" without construction detail; Cardlytics names "a rigorous test vs. control process" only; Chase, PayPal and Citi name lift only. | supported with qualification | Change 'Eight of the 20 cohort B rows document some sales or conversion lift measurement' to 'Eight of the 20 cohort B rows name or document sales or conversion lift measurement (five only name it)'; and 'only Upside discloses how its comparison group is built' to 'only Upside describes a matched comparison group's construction in operational detail (Circana's Ibotta meta-study says only "matched buyer groups"; Fetch's comparison group is its random holdout)'. | RTL-E3-F51 (evidence/verify/E3.json) | |
| C161 | The identity and transaction bases documented across cohort B fall into six kinds: card or bank-account transactions held by the operator or its bank partners (Chase, Cardlytics, Amex Offers, Citi by its own description, Upside through card digits and merchant loyalty identifiers); payment-account transactions (PayPal); receipts (Fetch, Ibotta's own app); publisher or retailer purchase feeds (Ibotta's third-party publishers); advertiser-supplied conversion events (Rokt, Klarna Search & Compare's tracking code); and loyalty or account audiences with no documented join to an outside brand's sale | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | PayPal's suite claims "deterministic-driven identity"; Cardlytics' FI partners include Chase, Wells Fargo and American Express; Citi's release cites "first-party transaction data". | supported with qualification | Add to the payment-account kind: PayPal claims 'deterministic-driven identity' for its measurement suite without describing how off-site exposures are joined. Add to the card kind: Cardlytics' bank partners include American Express and Chase, which also sell their own offers rows, so two kinds of row draw on the same card data. | 'None of the 20 rows discloses the share of an advertiser's transactions it can match' holds on the pages opened by this verifier. | RTL-E3-F52 (evidence/verify/E3.json) |
| C162 | No cohort B row's pages opened by this stream cite an MRC accreditation, an AAM certification or an IAB Europe certification; the third-party records found are an assessment Fetch engaged KPMG to perform (2023), Circana studies Ibotta publishes, a Nielsen marketing-mix model Fetch commissioned (2025), Nielsen Sales Lift as a Cardlytics option, and a retailer re-computation Upside publishes; on governance, no row documents a restatement or correction policy or a dispute procedure, and the partial items found are Rokt's disclosure that referral-time counts accumulate, Cardlytics' dated documenta | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | Klarna Search & Compare terms, section 10(b): "the Partner agrees that Klarna’s tracking count shall be applied"; "in no event ... will data provided by Klarna constitute final billing numbers" (only invoices are final; quote corrected 3 Nov 2026 by the cohort B cell verifier); an invoice not objected to "within eight (8) days from the date of the invoice, such invoice will be deemed valid and payable"; section 11(a): "Klarna is not liable for any incorrect or erroneous data". Upside: "Upside may update this methodology in its reasonable discretion." | contradicted | Restate the governance half: 'on governance, no row documents a restatement or correction policy; one row documents a billing-dispute rule, in the operator's favour (Klarna Search & Compare: Klarna's tracking count applies, its data are never "final billing numbers" (only invoices are), and an invoice is deemed valid unless the partner objects within eight days; Klarna disclaims liability for incorrect data); the other partial items are Rokt's disclosure that referral-time counts accumulate, Cardlytics' dated documentation and latency notes, and Upside's dated methodology with a unilateral rig | The cohort_b_row_screen governance cell for Klarna Search & Compare needs the same addition; a scoring cell for governance could move from n/e or 0-adjacent to a low positive value for that row. | RTL-E3-F53 (evidence/verify/E3.json) |
| C163 | In every cohort B row whose outcome reporting is documented, the party that produces the default number is the operator that sold the media (or, for Rokt, the network operator), on its own metric definitions; no row cites the IAB/MRC retail media guidelines' definitions or IAB Europe's, and none documents a separation between the media-sales and measurement functions; named third parties appear only as optional studies (Nielsen for Cardlytics, Circana for Ibotta, Lyft's and PayPal's partner lists) or as assessments the operator engaged, commissioned or publishes (KPMG and Nielsen for Fetch, Gi | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | Operator definitions confirmed (Cardlytics glossary ROAS; Rokt ROAS formula). PayPal's partner programme exists to "enable independent campaign validation with a certified list of third-party partners". Klarna's terms make the operator's tracking count apply, while its data are never final billing numbers (only invoices are) ("in no event ... will data provided by Klarna constitute final billing numbers" (only invoices are final; quote corrected 3 Nov 2026 by the cohort B cell verifier)). | supported with qualification | Add: PayPal names a certified third-party validation programme (purpose stated, per-partner roles not), and Klarna's terms make the operator's own count binding for billing; neither changes who produces the default number. | RTL-E3-F54 (evidence/verify/E3.json) | |
| C164 | Attribution parameters per cohort B row (19 rows). | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | Checked rows: Chase, PayPal (both), Cardlytics, Upside, Rokt, Klarna Search & Compare, Citi against the pages above; MARRIOTT MEDIA, Kinective, Amex and Zillow against the browser-pane notes. Cells match except as corrected. | supported with qualification | PayPal rows: add the suite-level claim "deterministic-driven identity" to match_basis (method undescribed). Upside row: control matched on spend "in the 12 months prior" to the user's first Upside transaction; fee = profit share x (incremental gross profit minus promotional costs). Rokt row sales_basis: add the currency default ("Rokt will assume conversion amounts are in the default currency for your account"). | TABLE:attribution_parameters (evidence/verify/E3.json) | |
| C165 | Default outcome method and lift-offer flags per row (20 rows). | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | See RTL-E3-F50, F20, F32, F34 and F46. | supported with qualification | PayPal default_target_contrast -> 'attribution and incrementality both claimed, no design' (or Citi recoded to match PayPal; one rule for both). MARRIOTT MEDIA and Expedia -> 'closed loop named without parameters (construct undetermined)'. Fetch min_spend_or_sample -> 'no numeric minimum; "statistically significant group sizes" (Feb 2023); "small but representative" control groups (Jun 2023)'. Upside default_assignment -> 'non-random by construction (inference: test group = offer claimers)'. | A method name never fills assignment: the Fetch 'random' is disclosed; the Upside 'non-random' is inferred from the design and should say so. | TABLE:default_method (evidence/verify/E3.json) |
| C166 | Buyer access per row (20 rows). | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | PayPal: the partner programme's stated purpose is "independent campaign validation with a certified list of third-party partners". Klarna Search & Compare: "in no event ... will data provided by Klarna constitute final billing numbers" (only invoices are final; quote corrected 3 Nov 2026 by the cohort B cell verifier). | supported with qualification | PayPal verification_tags: 'certified Measurement Partnership Program for "independent campaign validation" (AppsFlyer, Cint, Experian, iSpot, Kantar, Kochava, LiveRamp); per-partner roles not stated'. Klarna Search & Compare output_rules: add 'Klarna's tracking count is the final billing number'. | Cardlytics rows were not re-opened (S05, S06). | TABLE:access_record (evidence/verify/E3.json) |
| C167 | Per-row screen of remaining dimension inputs (20 rows). | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | Klarna terms sections 10(b) and 11(a); Cardlytics FI partners Chase, Wells Fargo, American Express; Citi release has no privacy commitment and names no route beyond its channels; Kard agreement pending (13 August 2026). | supported with qualification | Klarna Search & Compare governance_items: add 'billing-dispute rule: Klarna's count final; 8-day invoice objection window; no liability for incorrect data'. Cardlytics privacy or validation note: add that its FI partners include Chase and American Express (both separate cohort B rows). Citi: add an event note 'agreement to acquire Kard (13 August 2026; pending, closing conditions; independent operation until close)'. | TABLE:cohort_b_row_screen (evidence/verify/E3.json) | |
| C168 | Counts of cohort B rows by default method (7 categories, 20 rows). | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | Sum checked: 6 + 6 + 2 + 2 + 1 + 2 + 1 = 20. | supported with qualification | Under one rule for PayPal and Citi: attributed or offer-linked 5; both claimed, no design 2; closed loop named without parameters 2; incremental by documented design 2; delivery only 2; not documented 6; source inaccessible 1 (total 20). | Rename 'attributed by source term only (no parameters)' to 'closed loop named without parameters (construct undetermined)'. | TABLE:default_method_counts (evidence/verify/E3.json) |
| C169 | Vendor-published outcome figures (19 rows). | Refute-mode verifier for evidence/research/E3.json re-opened the cited source. | Fetch Nielsen iROAS (2.64, 2.50, 2.29, 2.14, 1.91), Circana (16.5%, 17%, 5.05, 48 campaigns), Uber (11%, 41%, 39%) and Citi ('up to 5x') confirmed. PayPal: "Compared to the weeks prior to the campaign, Ulta saw a 20% increase in transaction spend via PayPal during the campaign" (PayPal Internal Data, Jun-Jul 2025). | supported with qualification | PayPal Ulta row: comparator 'weeks prior to the campaign'; design_disclosed 'pre/post; no unexposed control'; period 'June-July 2025'. | Fetch case headlines (53%, 13.3%), Ibotta 8x and Danone 1.5 million units, Upside 0.4% and Klarna 25x not re-opened. | TABLE:vendor_reported_outcome_figures (evidence/verify/E3.json) |
| C170 | Criteo's own release of 28 March 2024 scopes its first MRC accreditation to display rendered impression and click metrics for Onsite Sponsored Products and Onsite Display Ads across desktop, mobile web and mobile in-app environments on Commerce Max and Commerce Yield, with a review of GIVT detection; the release does not name sales, ROAS or attribution as accredited. | Refute-mode verifier for evidence/research/F.json re-opened the cited source. | Release: "NEW YORK – March 28, 2024 –"; "Now accredited for impression and click metrics across desktop, mobile web, and mobile in-app environments"; formats "Onsite Sponsored Products and Onsite Display Ads" on "Commerce Max and Commerce Yield"; no sales, ROAS or attribution wording. MRC letter (15 March 2024, local pymupdf extraction): "The Media Rating Council (MRC) Board of Directors has voted in favor of granting initial accreditation of Criteo's Retail Media Platform (RMP) ("the Service") for Display Impressions and Clicks reported Gross and Net of General Invalid Traffic (GIVT) within D | supported with qualification | Keep the sentence about Criteo's release, then add: "MRC's own letter of 15 March 2024, which is the controlling record, grants initial accreditation of 'Criteo's Retail Media Platform (RMP)' for Display Impressions and Clicks, gross and net of GIVT (Impressions, Clicks, Invalid Impressions, Invalid Clicks), in desktop, mobile web and mobile in-app, 'specifically for On-site Sponsored Ads, Commerce Display Ads and Standard Banner Ads'. It names neither Commerce Max nor Commerce Yield, accredits no viewability or outcome metric, and does not list the IAB/MRC Retail Media Measurement Guidelines | The accreditation scope rests only on the vendor release. Under the shared rules the accreditor's letter settles accreditation, so add it as the primary source (direct_record, high) and keep the release as a vendor_assertion for the product mapping. For the Commerce Yield row's dimension 10, the instrument record should read: service 'Retail Media Platform (RMP)', mapped to Commerce Yield by vendo | RTL-F-F03 (evidence/verify/F.json) |
| C171 | Pentaleap's release of 22 January 2026 says it has been selected by Macy's, Inc. to power the Macy's Media Network with its ad server and mediation technology, with no launch or go-live date, adding Macy's (selected, not yet production-evidenced) to the dated US retailer record of The Home Depot (3 October 2024) and Staples (29 July 2026). | Refute-mode verifier for evidence/research/F.json re-opened the cited source. | Macy's release: "January 22, 2026"; Macy's, Inc. selected Pentaleap; subtitle "Pentaleap's ad server and mediation technology"; no launch, go-live or timing language; boilerplate lists "The Home Depot, Farmacia San Pablo, and Pague Menos". Pentaleap post ("July 29, 2026"): "With Home Depot, several pharmacy chains in Brazil and Mexico, and Staples on the platform,"; CVS and Bloomingdale's do not appear; Macy's only as a sidebar link. Home Depot ("October 03, 2024"): "To enhance Orange Access' ad server capabilities, Orange Apron Media collaborated with Pentaleap and Kevel." and "today announce | supported with qualification | Replace 'and Staples (29 July 2026)' with 'and Staples (29 July 2026, Pentaleap's own blog post, "on the platform", with no country or go-live date; Staples, Inc. in the US and Staples Canada are separately owned, so the US attribution is the research file's inference)'. | The Staples dating is a vendor blog post's date, not a deployment date, and the post states no geography. Keep the dimension-8 count at 2 only if the post date is accepted as dating and Staples is read as the US business; otherwise 1 (The Home Depot). | RTL-F-F24 (evidence/verify/F.json) |
| C172 | Osmos documents an incrementality product for onsite sponsored display campaigns (Phase 1) that builds a matched control group inside the campaign with a configurable 95/5 default split, a minimum of eight weeks, lift reported at product view, add-to-cart and purchase stages, per-campaign incremental conversions, revenue and ROAS in the console with exports; random assignment, sample thresholds, a significance rule, the identity or transaction basis and who runs the analysis are not stated; its home page also lists geo-holdouts, audience holdouts and synthetic control. | Refute-mode verifier for evidence/research/F.json re-opened the cited source. | "builds a control group inside the campaign"; "Most eligible users see the ad, a small holdout does not."; "Eligible users assigned exposed or control."; "Control group matches the exposed profile for a clean, like-for-like comparison."; "Eligibility logic identifies users who would have qualified for ad exposure."; "95/5 default split, configurable per campaign based on audience size and statistical needs."; "Minimum 8 weeks with sufficient audience reach."; "Product View, Add to Cart, and Purchase tracked for both exposed and control."; "Per-campaign incremental conversions, revenue, and ROA | supported with qualification | Osmos documents an incrementality product for onsite sponsored display campaigns (Phase 1, early access) that holds out eligible users inside the campaign ('Eligible users assigned exposed or control'; 'a small holdout does not' see the ad). The page says the control group 'matches the exposed profile' and gives a configurable 95/5 default split and a minimum of eight weeks. Lift is reported at product view, add-to-cart and purchase, with per-campaign incremental conversions, revenue and ROAS and exports. The assignment mechanism (random or not), sample thresholds, significance rule, identity | Control coding: code the counterfactual construction as a holdout (user-level, in-campaign, with eligibility logic) and the assignment as unknown. Do not code it as 'matched audience': 'matches the exposed profile' is the page's comparability claim, not a documented matching estimator. 'Matched control group' in this finding, in RTL-F-F38, in the summary and in the platform_functions table should | RTL-F-F30 (evidence/verify/F.json) |
| C173 | Swiftly's retail media page (timestamped 2 November 2026) asserts IAB-standard closed-loop attribution tied to in-store sales, closed-loop reporting of 'real sales lift' and an analytics dashboard showing impressions, engagement and in-store sales lift, and carries a PepsiCo quote of more than 100 iROAS and over 39 percent incremental sales lift with no method; named retailers are Homeland, Better Health Market and The Save Mart Companies without deployment dates. | Refute-mode verifier for evidence/research/F.json re-opened the cited source. | Published "Nov 2, 2026, 6:58 PM UTC" (no modified date). "Measure results using IAB-standard closed-loop attribution tied to real in-store sales."; "Prove performance with closed-loop reporting that ties every campaign to real sales lift."; dashboard shows "the in-store sales lift each campaign generates"; heading "Over $100 iROAS"; PepsiCo quote: "With Swiftly, we achieved more than $100 iROAS and over 39% incremental sales lift,". Retailers: "This case study highlights how Homeland partnered with Swiftly to enhance its digital presence."; "Better Health Market used Swiftly's Audience Optimiz | supported with qualification | Change 'a PepsiCo quote of more than 100 iROAS' to 'a PepsiCo quote of "more than $100 iROAS" (a dollar-denominated iROAS, as written)'. Append: 'The page carries a publication timestamp of 2 November 2026, not deployment dates. Two of the three retailer items describe enhancing "digital presence" (Homeland) and expanding "digital circular reach" with Audience Optimizer (Better Health Market), not documented retail media campaigns.' | For dimension 8, the three retailers should not be counted as 'dated production evidence'. The only date is the page's publication timestamp, and two of the three items do not document a retail media deployment. Strict reading: 0; at most 1 (The Save Mart Companies, quoted on 'brand partnerships') if a page timestamp is accepted as dating. The platform_functions table lists all three in named_us_r | RTL-F-F33 (evidence/verify/F.json) |
| C174 | Across the 20 cohort C rows, by 3 November 2026 this review found two platforms documenting a control design for clients' campaigns with parameters (Moloco: randomised user-level ghost-bidding holdout, vendor blog, no sample or significance rule; Osmos: matched control group with a 95/5 default split and an eight-week minimum, assignment unstated, onsite display Phase 1), three asserting incrementality or sales lift without a design (Koddi, Swiftly, Zitcha); six documenting attribution windows or models in configurable form (Criteo Commerce Yield, read on the buyer-side Commerce Max help centr | Refute-mode verifier for evidence/research/F.json re-opened the cited source. | Every attribution page re-opened (Criteo Commerce Max, Kevel FAQ and overview, Koddi, Topsort overview and Direct Attribution, Moloco 2026 and 2025) states no default window. The Criteo page is the Commerce Max (advertiser) help centre and never mentions Commerce Yield. Osmos describes an in-campaign holdout ('Eligible users assigned exposed or control'). Koddi's developer page says 'Clients can use simple last-touch models or more advanced methods like weighted attribution or incrementality testing.' The 6 + 1 + 1 + 12 = 20 row arithmetic holds. | supported with qualification | Restate the attribution clause as: 'five documenting attribution windows or models that the retailer or marketplace configures (Kevel; Koddi; Topsort; Moloco; Streaem), plus Criteo, whose buyer-set windows are documented only on the advertiser-facing Commerce Max help centre and reach the Commerce Yield row by inference'. Replace 'Osmos: matched control group with a 95/5 default split' with 'Osmos: an in-campaign user-level holdout whose control group it says "matches the exposed profile", with a 95/5 default split'. Add that Koddi's assertion of incrementality testing also appears in its deve | The count mixes documented and inferred rows. Criteo Commerce Yield is an inference row, so carry it in a separate category rather than inside the 'six'. | RTL-F-F38 (evidence/verify/F.json) |
| C175 | Per cohort C row: control designs, attribution parameters, defaults, access routes, identity basis, classes measured, route families and named US retailers with dated production evidence (20 rows). | Refute-mode verifier for evidence/research/F.json re-opened the cited source. | Dated US retailer statements as found: Kevel: Dollar General, 20 May 2025, Kevel release ('is leveraging Kevel's Retail Media Cloud™'; DG has 'over 20,500 stores across 48 states'), and The Home Depot, 3 October 2024 ('collaborated with Pentaleap and Kevel'). Vantage: The Home Depot, 3 October 2024 ('Orange Access, powered by Vantage'). Pentaleap: The Home Depot, 3 October 2024; Staples, 29 July 2026 ('on the platform', no country); Macy's selected 22 January 2026 (no go-live). Moloco: Wayfair as a logo, a CTO testimonial and a stat footnote 'Customer provided data; analysis period: Q4 2024' o | supported with qualification | (1) Criteo Commerce Yield: the attribution_configurable cell is already marked as inference; add the same marker to identity_basis, to the off-site deduplication in classes_measured and to the DSP analytics reports in access_routes (Demand Side Analytics page, demand-side, names the 'Retail Media Platform (RMP)' and not Commerce Yield). (2) Swiftly named_us_retailers_dated: replace with 'none with a deployment date; Homeland, Better Health Market and The Save Mart Companies named on a page published 2 November 2026; the Homeland and Better Health Market items concern digital presence and digit | Cross-row transfer: buyer-side Criteo pages (Commerce Max help centre; Demand Side Analytics) fill Commerce Yield cells. The row rule forbids transfer between service rows unless the cell is coded as inference and kept out of scoring. | data_tables.platform_functions (evidence/verify/F.json) |
| C176 | Attribution window options, touch model and who configures them, as documented per platform (8 rows). | Refute-mode verifier for evidence/research/F.json re-opened the cited source. | Values re-found: Criteo post-click 7/14/30 and post-view none/1/7/14/30, not above post-click, campaign-level (Commerce Max). Kevel 1/7/14/30, independent click and view, flights and ads. Topsort 1–30 per format, last-click or last-impression, example table 7/14/30. Moloco 1/7/14 platform-level (2026) and 7 or 14 (2025). Carrot Ads: three-day ingestion only. Amazon Retail Ad Service: none. No default on any page. Koddi: 'The look back window is also configurable and can be set to any numbers of days less than 30'; beyond 30 days 'contact your Koddi team'. The page does not say a view window eq | supported with qualification | Criteo row: prefix configured_by with 'advertiser (Demand Brand / Demand Seller) per campaign in Commerce Max'. The cell already says inference for this row; keep it out of any Commerce Yield score. Koddi row: post_view_options_days 'not stated separately (exposed attribution types exist)'; configured_by 'configurable, by whom not stated; beyond 30 days via the Koddi team'. | The Criteo row is buyer-side evidence placed in a retailer-side (cohort C) table and should carry evidence_class 'inference'. | data_tables.attribution_window_options (evidence/verify/F.json) |
| C177 | Criteo Commerce Max's own help centre documents the attribution rules of the outcome feed it reports ('About Attribution', undated, read 3 November 2026): separate post-click windows of 7, 14 or 30 days and post-view windows of none, 1, 7, 14 or 30 days, set per campaign, with post-view no longer than post-click; each sale credited to 'a single, most-relevant ad interaction' (clicks before impressions, then closer product matches, then recency); sales counted on the same product, category or brand (or seller); contributing interactions recorded as assisted sales; 'Sales are deduplicated betwee | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | About Attribution (undated): 'Post-click options: 7, 14, or 30 days'; 'Post-view options: None, 1, 7, 14, or 30 days'; 'The post-view window cannot be longer than the post-click window.' No default window. Credit goes to 'the most relevant ad event': 'Clicks take precedence over impressions.'; 'More specific product matches take precedence.'; then the interaction closest to the purchase date. Qualifying sales: same product, same category and same brand (Demand Brand) or same seller (Demand Seller). 'Sales are deduplicated between Offsite and Onsite.'; 'Advertisements and media that aren't part | supported with qualification | Keep the claim text. In the limitations: (1) replace 'eight of eight' with 'seven of eight with values; sales basis (gross or net, returns, cancellations, taxes, shipping, discounts, spend basis) is not documented on any of the four Commerce Max pages read (About Attribution, Demand-Side Metrics, Data Freshness, About Sales Channels)'; (2) mark in-store inclusion as partial (retailers that supply offline sales are not named, no matchable share, and no statement that the click/view windows apply to offline sales); (3) record that matching is 'deterministic' on identifiers that include a first-p | The identifier list is no longer only the tool's paraphrase: the four identifiers and their descriptions are on the page as quoted here. | RTL-G1-F04 (evidence/verify/G1.json) |
| C178 | Dimension 2 for this feed: window, touch model, match basis, in-store inclusion, sales basis and halo (same product, category or brand), deduplication, and new-to-brand lookback (180 days, RTL-G1-F37) all carry documented values or rules: eight of eight, defaults unstated. | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | Per parameter for the Commerce Max feed (Criteo's own retailer network): (1) window click/view: DOCUMENTED as buyer-set options (post-click 7/14/30 days; post-view none/1/7/14/30; post-view no longer than post-click; set at campaign level), no default; (2) touch model: DOCUMENTED (single most-relevant interaction; clicks over impressions, then more specific product match, then recency; other contributing interactions reported as assisted sales); (3) match basis: DOCUMENTED (deterministic only; Customer ID, hashed email, retailer first-party visitor cookie, Criteo third-party cookie; no IP or p | contradicted | Seven of eight parameters documented with values (in-store partial; sales basis undocumented; default windows unstated), on one rule set stated for onsite and offsite. Under the dimension-2 anchors this is below 5 (all eight with values per placement class) and above 3 (window and touch model only); the vendor-documentation cap is 4 either way, so the cell value may not move, but the reason text must not say 'eight of eight'. | The finding merges 'sales basis' and 'halo' into one item ('sales basis and halo (same product, category or brand)'). In the brief, sales basis (gross or net; returns, cancellations, discounts, taxes, shipping; spend basis) and halo are separate parameters; the same merge appears in the network_feeds row ('sales basis and halo: same product, category or brand (or seller)'). | RTL-G1-F04 (limitations: 'eight of eight', reused by F37/F38 and the network_feeds row for Criteo Commerce Max) (evidence/verify/G1.json) |
| C179 | Commerce Max publishes dated monthly release notes for 2026 (January to October), including 'Changes in Attribution Latency for Commerce Max Offsite Ads' (February 2026), 'Change History Tool' (March 2026), 'Cost Transparency Report' (July 2026), 'AI Shopping Assistant Page Type' and 'In-Platform Agent' (June 2026) and 'More Reliable Reporting' (October 2026). | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | Monthly entries January to October 2026 are present with the headlines quoted. Bodies now read: February 2026 'Changes in Attribution Latency for Commerce Max Offsite Ads': 'We’ve updated our attribution pipeline to better support offline (in‑store) sales data.'; 'Commerce Max Offsite ad attribution and reporting will now appear after 72 hours,' 'instead of the previous 9-hour window.'; 'This change affects reporting only and sales data used by our optimization algorithms remains unchanged.' October 2026 'More Reliable Reporting': 'Eight high-volume dimensions ... are now available in Custom R | supported with qualification | Replace 'headlines only; entry bodies not returned' with the bodies: the February 2026 entry is a documented, dated change to offsite attribution reporting latency (9 hours to 72 hours, reporting only); the October 2026 entry moves eight high-volume dimensions to Custom Reports and is not a restatement. Update the governance_screen row for Criteo Commerce Max accordingly (documented_reporting_revisions: y, with the February entry as the dated example). | RTL-G1-F05 (evidence/verify/G1.json) | |
| C180 | Flywheel claims 'One login, optimized campaigns across +100 global retailer platforms and marketplaces' with logos (Amazon Ads, Walmart, Target, Instacart, Kroger, Criteo, The Trade Desk) and 'scale ROAS with 20% more efficiency than native retailer platform ad consoles'; no API documentation, methodology note, spend-under-management figure, US statement or ChatGPT Ads partnership was found on the two Flywheel pages opened or by search 4. | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | 'One login, optimized campaigns across +100 global retailer platforms and marketplaces.'; 'scale ROAS with 20% more efficiency than native retailer platform ad consoles'. Logos shown: TikTok Shop, Amazon Ads, MAP, Instacart, Target, Criteo, Kroger, PromoteIQ, Shopee, DoorDash, Mercado Libre, The Trade Desk, Walmart. The page also says 'Goal-based bidding (powered by Perpetua)' and 'Proven Perpetua-powered automation'. No API, spend figure, ChatGPT or US statement; Omnicom appears only in a footer link URL. | supported with qualification | Add the remaining logos (DoorDash, Mercado Libre, Shopee, TikTok Shop, PromoteIQ, MAP) or say the seven named are a subset; add that the page states Flywheel's bidding is 'powered by Perpetua', so the Flywheel Commerce Cloud and Perpetua rows share an engine and evidence must not be counted twice. | Interest mark missing: the frozen conflict screen (private/interests-declaration.md) and the product register put an interest mark on Flywheel Commerce Cloud (Omnicom owns Flywheel; Hearts & Science, a past employer, is an Omnicom agency). No G1 finding or table row on Flywheel carries it. | RTL-G1-F11 (evidence/verify/G1.json) |
| C181 | Reported: the same PPC Land article attributes to Pacvue the statements that its platform 'connects advertisers to more than 100 channels including Amazon, Walmart, Target, and Instacart' and 'currently powers approximately 12% of total retail media ad spend globally' (a share with no denominator source; the dollar figure is on the about page, RTL-U-S305). | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | PPC Land: Pacvue 'connects advertisers to more than 100 channels including Amazon, Walmart, Target, and Instacart' and 'currently powers approximately 12% of total retail media ad spend globally', attributed 'according to the company', with no base, method or source; no dollar figure and no 'excluding China'. Italian about page (session notes): 'A partire dal 2025, Pacvue gestisce il 12% della spesa pubblicitaria totale nel settore retail media a livello mondiale, esclusa la Cina'; the dollar sentence ('gestisce oltre 12 miliardi di dollari di spesa pubblicitaria su oltre 100 reti globali di r | supported with qualification | Add that Pacvue's own (Italian-locale) page scopes the 12% share 'as of 2025' and 'excluding China'. Change 'Dimension 8 reuses RTL-U-F175 (over US$12 billion, as of 2025)' to 'over US$12 billion, period not stated on the page (the "as of 2025" phrase belongs to the separate 12% share sentence)'. | RTL-G1-F19 (evidence/verify/G1.json) | |
| C182 | Quartile announced on 22 September 2026 (PR Newswire) that it 'Becomes Technology Partner Supporting Ads in ChatGPT', giving advertisers 'access advertising in ChatGPT through Quartile's existing tools and workflows'; the release restates '5,300+ brands' and annual ad spend 'exceeding $2 billion'. | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | Headline as quoted; 'PR Newswire', 22 September 2026, no city dateline; 'access advertising in ChatGPT through Quartile's existing tools and workflows'. A 'Quick Intel' block reads 'Quartile supports 5,300+ brands and manages over $2 billion in annual ad spend across major channels'; the boilerplate says 'an annual ad spend exceeding $2 billion'. Nothing on measuring, reporting or attributing ChatGPT campaigns; OpenAI is not named. | supported with qualification | Attribute '5,300+ brands' to the copy's 'Quick Intel' summary (which may be the syndicator's, not the release's) unless the release body is read; the '$2 billion' is the release boilerplate's 'annual ad spend', described in the same copy as 'across major channels'. | RTL-G1-F24 (evidence/verify/G1.json) | |
| C183 | Quartile's homepage figure '$2B+' 'Annual Retail Ad Spend Managed' (RTL-U-S203, undated) is corroborated by the 22 September 2026 release's 'exceeding $2 billion' annual ad spend, which supplies a date for the dimension-8 value (period: annual, as of September 2026). | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | Homepage (tool summary; the tool declined verbatim quoting): 'Over $2B in annual retail ad spend managed', '$25B' managed sales, '5,300' customers in '32' countries, a '41%' average ROAS increase. Release copy: 'manages over $2 billion in annual ad spend across major channels' and boilerplate 'an annual ad spend exceeding $2 billion'. | supported with qualification | The release does not date the homepage figure: it states an all-channel annual ad spend ('across major channels', including Google, Meta, Microsoft and ChatGPT per the homepage channel list), while the homepage states 'Annual Retail Ad Spend Managed'. Keep the retail-labelled figure as undated (page undated) and the release figure as all-channel 'as of 22 September 2026'; both sit above US$2 billion, so the rung does not change. | 'Corroborated' overstates: both statements come from the vendor (one evidence origin), so the release is a restatement, not corroboration. | RTL-G1-F25 (evidence/verify/G1.json) |
| C184 | Stackline's 17 June 2024 release (Business Wire) describes 'Stackline Multi-Retailer Attribution' as combining 'proprietary audience technology and Amazon Marketing Cloud' to measure purchases 'across retailers and channels at the individual products level' for shoppers exposed to Amazon media, with a client case ('nearly $10 million' of Thursday Night Football ads; viewers 'contributed an additional $63 million in sales from other retail partners') analysed in an AMC clean room; no window, matching method, holdout or retailer list is stated. | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | On Stackline's attribution page the Multi-retailer module ('Developed in partnership with Amazon, this groundbreaking technology links verified ad exposures to real purchase receipts') links to 'Shopper Analytics' (/shopper-analytics) and a demo form, not to Ad Manager; only the Incrementality analytics module links to 'Ad Manager'. | supported with qualification | State that Stackline's own page attaches Multi-retailer attribution to Shopper Analytics, not Ad Manager. Under the no-transfer rule the 2024 Multi-Retailer Attribution evidence should not fill the Stackline Ad Manager x Amazon feed row (origin y, granularity y) unless a source ties it to Ad Manager. | The Business Wire copy (RTL-G1-S36) was not reopened; its quotes are not re-verified. | RTL-G1-F30 (evidence/verify/G1.json) |
| C185 | Outcome-feed provenance (dimension 3) is documented unevenly across the 47 product-by-network retail media feed rows recorded (ChatGPT rows excluded): 5 rows document origin plus at least two of granularity, constraints and limitations (Perpetua x Walmart Connect; Intentwise Optimize x Walmart Connect; Intentwise Optimize x Amazon Ads; Xnurta x Criteo; Criteo Commerce Max x Criteo's own retailer network), of which 2 document all four (Intentwise Optimize x Amazon Ads; Criteo Commerce Max x Criteo's own retailer network); 6 rows document origin plus one further field; 5 document origin only; 31 | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | The group counts recompute exactly from the network_feeds table (47 retail media rows: A=5, B=6, C=5, D=31; all four fields y: 2). Two inputs move: (a) the Stackline Ad Manager x Amazon row (origin y, granularity y; group B) rests on Multi-retailer attribution, which Stackline's page links to Shopper Analytics, not Ad Manager; (b) the Criteo Commerce Max constraints field cites an '8 to 24 hours' delay, but the February 2026 release note says offsite attribution and reporting 'will now appear after 72 hours, instead of the previous 9-hour window'. | supported with qualification | If the Stackline row is recoded to no documented provenance for Ad Manager, the counts become A=5, B=5, C=5, D=32. Keep Criteo in group A but restate its constraint as '8 to 24 hours (Data Freshness, undated); offsite attribution and reporting after 72 hours since February 2026 (release note)'. | RTL-G1-F36 (evidence/verify/G1.json) | |
| C186 | Commerce Max documents the limits of its outcome feed: metrics 'are updated with a delay, typically between 8 and 24 hours', collection is 'close to real-time' for most retailers but slower for some, dashboard figures 'can change and reflect new events, even if your campaign has stopped' (late in-window sales raise them, invalid-traffic removal lowers clicks), offline (in-store) sales are 'imported once every few days' and take '3 days (around 74 hours)' to appear after ingestion, and 'not all retailers provide offline sales data'. | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | Data Freshness (undated): 'are updated with a delay, typically between 8 and 24 hours, depending on the retailer and processing time.'; 'For most Retailers, the data collection is close to real-time.'; dashboards 'can change and reflect new events,' after a campaign stops, with late in-window sales and invalid-traffic removal as the reasons. About Sales Channels: 'Offline sales data is imported once every few days.'; 'It will take 3 days (around 74 hours) to make offline sales available in the UI after ingestion.'; 'not all retailers provide offline sales data.' What's New, February 2026: 'Com | supported with qualification | Add: 'For offsite ads, Commerce Max's February 2026 release note moves attribution and reporting to 72 hours (from 9 hours) to capture in-store sales, for reporting only; the undated Data Freshness page's 8-to-24-hour figure may predate that change.' Use the dated release note as the governance example of a documented reporting revision. | RTL-G1-F38 (evidence/verify/G1.json) | |
| C187 | Flywheel Digital is owned by Omnicom (G1 labels RTL-G1-S20 'Flywheel Digital (Omnicom)') and Perpetua operates under Flywheel Digital Limited (G1 labels Perpetua sources 'Perpetua (Flywheel Digital Limited)'). | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | Omnicom: 'NEW YORK, Jan. 2, 2024 /PRNewswire/ -- ... today announced it has closed the previously announced acquisition of Flywheel Digital', 'the digital commerce business of Ascential'; no price in the closing release (terms first announced 30 October 2023); Perpetua is not named. Perpetua privacy policy ('Last updated on 06 August 2025'): 'Flywheel Digital Limited and/or its affiliates'; 'Flywheel is the controller of your personal data'; contact 'Omnicom Group Inc. Attn: General Counsel 280 Park Avenue New York, NY 10017'. Flywheel retail media page: 'Goal-based bidding (powered by Perpetu | supported with qualification | Record the chain as inference from three sources: Omnicom closed Flywheel Digital on 2 January 2024 (issuer release, Perpetua not named); Perpetua's controller is Flywheel Digital Limited with an Omnicom legal contact (policy of 6 August 2025); Flywheel's page sells bidding 'powered by Perpetua'. Perpetua's own acquisition by Ascential (closing 15 April 2021 per a law-firm deal record seen only as a search result) is a lead, not verified here. | Interest marks missing throughout G1: the frozen conflict screen requires an interest mark on Flywheel Commerce Cloud and Perpetua (past employer Hearts & Science is an Omnicom agency), and the product register carries it ('interest' field), but no G1 finding, table row or summary line on either product carries the mark or flags them for the sensitivity run. | ownership: Flywheel Commerce Cloud and Perpetua under Omnicom (source org fields RTL-G1-S01, S12-S14, S20, S24, S25, S45, S51; product register rows) (evidence/verify/G1.json) |
| C188 | One row per cohort D product x network feed with integration route, outcome metrics, parameters disclosed and four provenance fields. | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | Every 'API' integration_type is backed by a page naming an API for that network (Perpetua x Walmart: 'Walmart's API'; Perpetua x Instacart: 'via the API'; Intentwise x Walmart: 'Walmart Connect Ads APIs'; Intentwise x Instacart: 'API user'; Xnurta x Criteo and x Walmart: 'via API'). No API is credited where the opened page names none (AMS feeds, Xnurta x Amazon, Skai, Stackline, Flywheel, Quartile all coded unknown). | supported with qualification | Edit these cells: (1) Criteo Commerce Max parameters_disclosed: split 'sales basis and halo'; sales basis = not disclosed; halo = same category/brand or seller scope; in-store = some retailers, unnamed, no window statement; add '(seven of eight; sales basis missing)'. Constraints: add 'offsite attribution and reporting after 72 hours since February 2026'. (2) Perpetua x Instacart: touch model is verbatim ('linear multi-touch or last-touch'; default linear multi-touch), window click-only, record dated 22 July 2022. (3) Xnurta x Amazon: 'attribution is modeled' context = AMC Hub outputs 'from yo | Pacvue rows match the session's notes (Amazon, Walmart, Target, Kroger, Instacart, Criteo, CitrusAd). Teikametrics, CommerceIQ and Quartile name rows were not reopened beyond the Quartile homepage summary. | table: network_feeds (evidence/verify/G1.json) |
| C189 | Dimension-8 candidates: Pacvue over US$12 billion; Stackline US$2.4 billion; Quartile US$2 billion+ (homepage) and 'exceeding $2 billion' (release); Intentwise US$1.5 billion+ (Amazon only); Criteo US$4 billion company-wide; none found for the rest. | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | Pacvue (Italian page, session notes): 'gestisce oltre 12 miliardi di dollari di spesa pubblicitaria su oltre 100 reti globali di retail media' (manages over US$12 billion of ad spend across over 100 global retail media networks); no period and no 'annual' in that sentence; 'A partire dal 2025' opens the separate share sentence, which also says 'esclusa la Cina'. Stackline: 'Optimized over $2.4 billion in client ad spend last year' on an undated page whose footer reads 'Copyright © Stackline' with no year. Quartile homepage: over $2B 'annual retail ad spend managed' (tool summary). Quartile rel | supported with qualification | (1) Pacvue period: change 'as of 2025 (page undated)' to 'period not stated (not stated as annual); the "as of 2025" phrase belongs to the 12% share sentence'; because the cohort D unit is 'ad spend under management (US$, stated period)', the scorer must decide whether a period-less figure qualifies, and the English wording stays unverified. (2) Stackline period: drop '(copyright 2026)'; the page shows no year, so 'last year' cannot be dated. (3) Quartile release row: perimeter 'all channels ("across major channels")', not retail-only, and not a date for the homepage figure. (4) Intentwise: re | No change of rung follows for Stackline, Quartile or Intentwise (all US$1 billion to under 3 billion, rung 3, capped at 3); Pacvue's eligibility, not its rung, is what the period question affects. | table: spend_under_management (evidence/verify/G1.json) |
| C190 | Networks with documented integration and of which API-documented: Intentwise 4/2, Perpetua 3/2, Xnurta 3/2, Skai 5/0, Teikametrics 3/0, Stackline 4/0, CommerceIQ 6/0, Flywheel 6/0, Quartile 4/0, Pacvue 7/0, Criteo Commerce Max 1/0. | Refute-mode verifier for evidence/research/G1.json re-opened the cited source. | API-documented networks check out on the vendors' own pages: Intentwise (Walmart Connect Ads APIs; Instacart API user), Perpetua (Walmart's API; Instacart via the API), Xnurta (Criteo via API; Walmart Connect via API). Intentwise's AMS and Criteo-markets pages, Xnurta's Amazon page, Skai's retail media page, Stackline's advertising and attribution pages and Flywheel's retail media page name no API. Flywheel's page shows 13 logos (TikTok Shop, Amazon Ads, MAP, Instacart, Target, Criteo, Kroger, PromoteIQ, Shopee, DoorDash, Mercado Libre, The Trade Desk, Walmart). | supported with qualification | Flywheel's 'networks_with_documented_integration' of 6 undercounts the logos (names only, so dimension 6 stays n/e either way). Skai's count of 5 includes Kroger from a network-side listing on KPM's page and names without a documented integration route; label it 'named feeds' rather than 'documented integration'. The API column (2, 2, 2, else 0) stands; on the dimension-6 anchor for cohort D (1 = one or two API-documented networks) Intentwise, Perpetua and Xnurta sit at 1. | Teikametrics (3/0) and CommerceIQ (6/0) were not reopened. | table: api_integration_counts (evidence/verify/G1.json) |
| C191 | Circana's Media Lift page, read by the session in a browser pane on 3 November 2026 and relayed to this stream, describes Household Lift as 'household-level A/B testing' and Market Lift as using a 'thorough matching process'; neither relayed passage states how households or markets are allocated, so assignment is coded unknown, the counterfactual as matched audience (households) for Household Lift and matched market or store for Market Lift, and the estimation method as unknown. | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | Read in full with curl and a browser user agent (server-rendered HTML; the fetch tool gets only the title). Household Lift paragraph: 'Our Household Lift™ service gives you the most precise, household-level A/B testing in the industry, connecting the people exposed to your media with the products they buy in-store.'; its bullets include 'Enact controlled household tests', 'Leverage your loyalty card data' and 'Harmonize data from TV, digital household-level exposure, and your CRM'. Market Lift paragraph (separate): it 'delivers sales uplift for your campaign test executions, projected sales im | supported with qualification | Circana's Media Lift page (circana.com/solutions/media-lift, read in full with curl and a browser user agent on 3 November 2026; the fetch tool sees only its title) describes Household Lift as 'household-level A/B testing' that connects 'the people exposed to your media with the products they buy in-store', with 'controlled household tests', and Market Lift as in-market A/B testing on point-of-sale data that reports 'projected sales impact, and ROAS' using 'a thorough matching process'; neither paragraph states how households, markets or stores are allocated, so assignment is coded unknown for | Counterfactual mis-coded: the measurement record fills Household Lift with 'matched audience or store (Household Lift, per 'matching')', but the matching sentence sits in the Market Lift paragraph; the page states no matching for Household Lift. Access mislabelled: RTL-G2-S01 is a directly readable page (curl with a browser user agent), not a relay, so access_limits should read 'full (curl, browse | RTL-G2-F01 (evidence/verify/G2.json) |
| C192 | Circana's US production use is documented by the joint Circana–DoorDash release of 2 September 2026: Circana supplied 'independent, third-party validation to DoorDash's measurement capabilities' on an 'eight-week campaign on DoorDash, between 11/10/25 and 1/4/2026 in the U.S.', reporting '$0.59 in incremental sales per household'; the release validates DoorDash's number, not Circana's own service, names no panel size, no assignment and no test-and-control wording, and does not say who paid. | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | 'today announced a partnership to bring independent, third-party validation to DoorDash's measurement capabilities.' 'Now available to select advertisers for Sponsored Products and Sponsored Brands'. The paragraph opens 'New findings from DoorDash measurement illustrate the platform's incremental value.' The asterisk follows the buyer-overlap shares ('...10% purchased through both channels.*'); the footnote reads '*Based on Circana analysis of buyers of measured brands during an eight-week campaign on DoorDash, between 11/10/25 and 1/4/2026 in the U.S.' The $0.59 comes later, unfootnoted: 'Ear | supported with qualification | Circana's US production use is documented by the joint Circana–DoorDash release of 2 September 2026, which announces 'a partnership to bring independent, third-party validation to DoorDash's measurement capabilities', 'Now available to select advertisers for Sponsored Products and Sponsored Brands'; the release's footnoted buyer-overlap analysis is 'Based on Circana analysis of buyers of measured brands during an eight-week campaign on DoorDash, between 11/10/25 and 1/4/2026 in the U.S.', while the '$0.59 in incremental sales per household' for 'A major beverage company' ('three times the Circ | Period mis-attached: the eight-week window is footnoted to the overlap analysis, not to the $0.59 figure; the $0.59's reporting party (Circana or DoorDash) is ambiguous on the page, so the measurement record's reporting_party should read 'unclear (joint release; paragraph framed as DoorDash measurement)'. | RTL-G2-F02 (evidence/verify/G2.json) |
| C193 | No panel or feed size, US household count or projection method for Circana Media Lift was found in the opened record by 3 November 2026: the product page served only its title to the fetch tool, the DoorDash release names no panel size, and the two case-study pages returned 403 and 404; Circana's household data set is referred to only by the service names ('Household-Level', 'Market/Store'). | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | The Media Lift page states no household count, panel size or projection method. Its scale statements: 'Tracking $5.8T in consumer spend, no one brings more clarity to your media decisions than Circana.'; 'Circana’s full-scale household-level data has 4x ACV versus the nearest competitor.'; 'Backed by one of the world’s largest household data sets'; 'Join more than 7,000 clients and partners worldwide'. It also says 'Circana’s Media Lift solution drives up to 9x greater ROI through our optimizations.' Market Lift reports 'projected sales impact' (method not stated). | supported with qualification | No US household count, panel size or projection method for Circana Media Lift appears on the Media Lift page (read in full on 3 November 2026) or on the DoorDash release; the page states scale only in other units ('Tracking $5.8T in consumer spend', no geography or period; household data with '4x ACV versus the nearest competitor'), and the Household Lift case study returned HTTP 403 (AU locale) and 404 (locale-free URL). | Evidence class 'inference' is right for the bounded absence, but its basis changes: the page is now read, not unreadable. | RTL-G2-F03 (evidence/verify/G2.json) |
| C194 | Numerator's media page documents 'sales lift studies that analyze exposed vs non-exposed consumers' on its US purchase panel, and the only method document opened, the SecureMeasure client success story (PDF dated 30 June 2023), describes a CRM-marketing measurement in which first-party touchpoints were matched to the panel 'using clean room technology' and 'six continuous months of consumer purchase behavior were analyzed to measure the pre and post impact of marketing touchpoints', reporting a 'lift in buy rate above the baseline' on a '12M Pre- vs. Post-Buy Rate Lift Index'; how the unexpose | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | The media-industry page is addressed to media companies ('Leading media companies use Numerator data to power their advertising strategies', logos NBCUniversal, Pinterest, BET): 'Numerator offers solutions that help prove the effectiveness of your advertisers’ campaign investments, including sales lift studies that analyze exposed vs non-exposed consumers, survey-based studies, and data feeds for use in measurement models.' SecureMeasure PDF (2 pages; creation date 30 June 2023): 'Leveraging Numerator's CRM measurement solution, SecureMeasure, the manufacturer securely matched their first-part | supported with qualification | Numerator's media-industry page, addressed to media companies, offers 'sales lift studies that analyze exposed vs non-exposed consumers' to 'help prove the effectiveness of your advertisers’ campaign investments'; the only method document opened, the SecureMeasure client success story (PDF created 30 June 2023), is a CRM-marketing case in which first-party CRM data were matched to the panel 'using clean room technology' and 'six continuous months of consumer purchase behavior were analyzed to measure the pre and post impact of marketing touchpoints'; how the unexposed comparison for media sale | Independence field incomplete: the sales-lift sentence sits on a page that sells the studies to media companies to prove their advertisers' results, so 'funding: the client' should read 'unknown; media sellers are an addressed buyer'. The measurement record's observed_vs_modelled cell ('projection to Total U.S. documented for the Total Commerce Panel') rests only on the confidential whitepaper; re | RTL-G2-F06 (evidence/verify/G2.json) |
| C195 | Numerator states its panel in mixed units across its own pages: 'sourced directly from 1 Million+ US households' (media page), '1M+ HOUSEHOLDS SHARING TRIPS' (media solutions page), 'over a million consumers' and '1M+' 'PANELISTS UPLOADING RECEIPTS' (measurement panel page), while its 10 April 2025 release says the Total Commerce Panel static 'expanded its Total Commerce Panel from 105,000 to 200,000 U.S. households' requiring 'two trips per month for 12 months' with 'data collected from ~90% of all U.S. zip codes'; the projection method is documented only in the 2023 whitepaper (marked confid | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | Public pages: 'sourced directly from 1 Million+ US households' (Data Feeds blurb on the media-industry page); '1M+ HOUSEHOLDS SHARING TRIPS' (media solutions); 'purchasing data from over a million consumers' and '1M+ PANELISTS UPLOADING RECEIPTS' (measurement panel). Release of 10 April 2025: 'has expanded its Total Commerce Panel from 105,000 to 200,000 U.S. households'; 'The panel requires two trips per month for 12 months'; under 'Census Alignment': 'Panelist alignment with 35+ U.S. attributes within 8 key demographics categories, with data collected from ~90% of all U.S. zip codes.'; 'The | supported with qualification | rest on the public 10 April 2025 release (RTL-G2-S22) and drop the whitepaper quotes. Restated: Numerator states its panel in mixed units across its own pages: 'sourced directly from 1 Million+ US households' (Data Feeds blurb on the media-industry page), '1M+ HOUSEHOLDS SHARING TRIPS' (media solutions page), 'over a million consumers' and '1M+' 'PANELISTS UPLOADING RECEIPTS' (measurement panel page), while its 10 April 2025 release says it 'has expanded its Total Commerce Panel from 105,000 to 200,000 U.S. households', that 'The panel requires two trips per month for 12 months', and, under 'C | The whitepaper is a client-only document by its own stamp; treat it like the 84.51° deck (do not quote). The finding's reliability note 'every page' is slightly wrong: the stamp is on pages 2 to 19, not on the cover (p. 1) or the back page (p. 20). | RTL-G2-F07 (evidence/verify/G2.json) |
| C196 | Numerator's consumer-data-privacy page states panelist rights and a rewards exchange ('You have the right to opt out easily at any time'; 'You have the right to share in the value of your data'; 'An equitable exchange, where your data provides you with meaningful rewards.') and its footer carries a 'US: Do Not Sell Or Share My Personal Information / Opt-Out of Targeted Advertising' link, but the page names no statute (CCPA/CPRA absent), no sensitive purchase categories, no data-broker registration and no privacy certification; recruitment through the 'Receipt Hog app' with demographic 'MicroSu | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | Privacy page: 'You have the right to opt out easily at any time'; 'You have the right to share in the value of your data'; 'An equitable exchange, where your data provides you with meaningful rewards.'; 'Crystal clear language around what data is being collected and how it is being used.'; a 'Download the Consumer Privacy Bill of Rights' link (not opened); footer 'US: Do Not Sell Or Share My Personal Information / Opt-Out of Targeted Advertising'; no CCPA or CPRA text. Release: 'Numerator blends zero-party data from over 1 million US households'; 'Over 86% of panelists share via permissioned d | supported with qualification | rest on the public 10 April 2025 release (RTL-G2-S22) and drop the whitepaper quotes. Restated: Numerator's consumer-data-privacy page states panelist rights and a rewards exchange ('You have the right to opt out easily at any time'; 'You have the right to share in the value of your data'; 'An equitable exchange, where your data provides you with meaningful rewards.') and its footer carries a 'US: Do Not Sell Or Share My Personal Information / Opt-Out of Targeted Advertising' link, but the page names no statute (CCPA/CPRA absent), no sensitive purchase categories, no data-broker registration a | Decision (a). Survives without the whitepaper except for the recruitment-app sentence. The downloadable 'Consumer Privacy Bill of Rights' is a public document the dimension-12 record should open. | RTL-G2-F09 (evidence/verify/G2.json) |
| C197 | Catalina's data basis is stated network-wide, not for HUB360: the 23 February 2026 release describes 'Catalina's U.S. deterministic data' as '130 million households, $600 billion in annual spending', 'processing 11 billion annual shopping trips across seventy major retail banners' and 'more than 400 million shopper IDs'; no projection method is claimed (a deterministic record, not a projected panel), and no matchable share, window or banner list appears on the measurement pages. | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | 'Catalina is the only deterministic purchase intelligence network of its kind — four decades, seventy retail banners, 130 million households, $600 billion in annual spending.'; 'processing 11 billion annual shopping trips across seventy major retail banners'; separately, 'Infillion will be leveraging Catalina's U.S. deterministic data in its platform'; 'Exclusive access to the largest deterministic data set—including more than 400 million shopper IDs'; also 'For the past several years, this data hasn't been available to the market at all. Infillion is bringing it back — exclusively through Inf | supported with qualification | Catalina's data basis is stated network-wide, not for HUB360: the 23 February 2026 release describes Catalina's network as 'four decades, seventy retail banners, 130 million households, $600 billion in annual spending' and as 'processing 11 billion annual shopping trips across seventy major retail banners', and gives 'more than 400 million shopper IDs'; the release does not label the 130 million households as U.S. (the 'U.S. deterministic data' phrase is in a separate sentence about what Infillion will use); no projection method is claimed, and no matchable share, window or banner list appears | The panel table's row label 'households in Catalina's US deterministic data' adds a geography the sentence does not carry. | RTL-G2-F13 (evidence/verify/G2.json) |
| C198 | Catalina's privacy notice and terms could not be read: the /legal page served only navigation and footer to the fetch tool, so the consent basis, sensitive-category handling, state-law handling and any certification for HUB360 data are n/e from this stream's record; the HUB360 pages carry only the footer links 'Cookies, Terms, Privacy, and Your Privacy Choices'. | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | HUB360 and in-flight footers: '© Catalina 2026. All rights reserved.'; links 'Terms & Conditions', 'Cookies', 'Terms', 'Privacy', 'Your Privacy Choices'. The /legal page itself was not re-opened here. | supported with qualification | Catalina's privacy notice and terms were not read: the /legal page served only navigation and footer to the fetch tool, so the consent basis, sensitive-category handling, state-law handling and any certification for HUB360 data are n/e from this stream's record; catalina.com serves server-rendered pages to curl with a browser user agent (four Catalina pages read that way on 3 November 2026), so the notice should be retried that way before dimension 12 is left at n/e. | The finding is an access statement; it holds for the fetch tool, but the 'needs a browser' conclusion is premature. | RTL-G2-F15 (evidence/verify/G2.json) |
| C199 | Buyer access to HUB360 outputs is documented as a portal and dashboards only ('HUB360's all-in-one portal allows you to benchmark your store against the competition and identify real, actionable insights at the UPC level.'; 'View KPIs within simple, graphic dashboards'); no API, export, clean-room query or third-party access route appears on the three product pages opened. | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | The portal and dashboard text sits under 'HUB360 RETAIL INSIGHTS' ('See The Big Box Picture'): 'HUB360’s all-in-one portal allows you to benchmark your store against the competition and identify real, actionable insights at the UPC level.'; feature 'View KPIs within simple, graphic dashboards'. The 'HUB360 MEASUREMENT' block names Campaign Snapshot, Incrementality Report and Multi-Touch Attribution and states no delivery route. The parent page says 'maximize your campaign’s ROAS with in-flight optimization and reporting.' | supported with qualification | No delivery route for HUB360 Measurement outputs is documented on the pages opened: the 'all-in-one portal' and 'simple, graphic dashboards' text belongs to the HUB360 Retail Insights module (a store-benchmarking tool), while the Measurement block and the measurement-and-optimization page mention only 'in-flight optimization and reporting'; no API, export, clean-room query or third-party access route appears on the three product pages opened. | Module mis-attributed: dashboards are evidenced for Retail Insights, not for the measurement product; dimension 3 for the measurement row is 'reporting asserted, route unstated'. | RTL-G2-F16 (evidence/verify/G2.json) |
| C200 | Attain's measurement page describes its standard design as 'Measure true causal lift using advanced holdout methodologies—and quantify impact through ROAS and iROAS.', with exposure ingested by tags, logs or CAPI ('Tag campaigns, upload exposure logs from any platform, or activate via CAPI—no pixels or client data required.'), a pre-measurement match check ('Before measurement begins, Attain validates a strong exposure-to-panel match to ensure results you can trust.') and daily results ('Results update daily, with signals available within 24-48 hours'); the page does not say how the holdout is | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | The page lists four separate measurement products: 'Brand Lift'; 'Sales Attribution' ('Did exposed consumers actually buy? Tie ad exposure to real transactions across retailers and channels, with results updated daily—not weeks later.'); 'New Buyer Lift'; and 'Incrementality' ('What sales wouldn’t have happened without your media? Measure true causal lift using advanced holdout methodologies—and quantify impact through ROAS and iROAS.'). 'Tag campaigns, upload exposure logs from any platform, or activate via CAPI—no pixels or client data required. Before measurement begins, Attain validates a | supported with qualification | Attain's measurement page offers attribution and incrementality as separate products: 'Sales Attribution' ('Did exposed consumers actually buy? Tie ad exposure to real transactions across retailers and channels') with no counterfactual, and 'Incrementality', described as 'Measure true causal lift using advanced holdout methodologies—and quantify impact through ROAS and iROAS'; exposure is ingested by tags, uploaded logs or CAPI, a pre-measurement exposure-to-panel match check is asserted, and results 'update daily'; the page does not say how the holdout is formed or by whom, so for the Increme | Design over-generalised: the holdout sentence describes the Incrementality product, not a 'standard design'; the support passage 'Tie ad exposure to real transactions...' belongs to Sales Attribution and must not support the holdout coding. Two measurement records are needed. | RTL-G2-F21 (evidence/verify/G2.json) |
| C201 | Attain's panel is stated in consumers, not households: '14M+' opted-in consumers, '10M+' receipts captured per month, '5+ years' historical lookback, '50+' linked retailers and '10,000+' merchants and brands on the measurement page, with purchases seen 'across Walmart, Amazon, Target and beyond—online and offline'; its data page says the panel is 'combined with Census benchmarks and advanced modeling, to represent shopping behaviors of the full U.S. adult population', which is the only projection statement found, and no matchable share or household count appears on any Attain page opened. | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | Measurement page stat block: '14M+ Opted-in consumers'; '$700B+ Cumulative annual spend'; '$1.5T+ Cumulative transactions in/out'; '10,000+ Merchants and brands across all verticals'; '5+ years Historical lookback across checking, debit and credit'; '10M+ Receipts captured per month'; '50+ Linked retailers'. 'See purchases everywhere—across Walmart, Amazon, Target and beyond—online and offline'. Our Data page: 'Our panel of 14M+ verified shoppers is unmatched. We use this inside view of real shopper behavior, combined with Census benchmarks and advanced modeling, to represent shopping behavior | supported with qualification | Add to the claim: the same stat block states '$700B+' 'Cumulative annual spend' and '$1.5T+' 'Cumulative transactions in/out' (US$, period and geography not stated), dollar units that are not comparable with the consumer counts. | Add both dollar figures to panel_and_feed_scale with their labels. | RTL-G2-F22 (evidence/verify/G2.json) |
| C202 | Attain's independence and ownership record from its own pages: 'No media inventory. No incentives to inflate results. Just objective measurement aligned with your business outcomes.' (measurement page); 'Attain began as Klover.' (2019 timeline), rebranded as Attain in 2022, launched the Frisbee rewards app in 2023, 'With offices in Chicago, New York, San Francisco and Denver,', 'Backed by' logos Core Innovation Capital, Mercato Partners, lightbank and Starting Line (no amounts), leadership Brian Mandelbaum (CEO & Co-founder), Dominic Bennett (CTO & Co-founder) and Ben Kartzman (President & COO | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | 'Independent, conflict-free' / 'No media inventory. No incentives to inflate results. Just objective measurement aligned with your business outcomes.' About page: 'Attain began as Klover'; 2022 'The company rebranded as Attain'; 2023 Frisbee launched; 2024 'The acquisition of Merryfield added new dimensions to our ecosystem'; 'We extend that data across the advertising ecosystem—helping brands, agencies, retailers and media platforms connect media investment to verified sales outcomes.'; above the logo wall (Amazon, McDonald’s, Coca-Cola, Walmart, Verizon, Uber, IPG, PMG, Publicis Groupe, WPP) | supported with qualification | Add: Attain's about page says it extends its data to 'brands, agencies, retailers and media platforms', records the 2024 'acquisition of Merryfield', and frames its logo wall as companies that 'adopted' or integrated Attain; so the 'No media inventory' statement sits beside stated service to retailers and media platforms (media sellers), as with Incremental, and the Amazon and Walmart logos are framed as adopters without a stated role. | Independence field: media-sale interest of Attain itself is none documented, but media sellers are among its stated customers, so the commissioning party of a given study can be a seller; funding stays unknown per study. | RTL-G2-F23 (evidence/verify/G2.json) |
| C203 | Incremental documents its Measurement & Reporting method at principle level only: 'Multiple causal techniques link investment to outcome, not the click that came last.'; 'Econometric techniques combined with designed and synthetic experiments separate media impact from other factors by retailer and channel'; 'An ensemble causal model decomposes sales into what caused them.'; 'Daily retraining keeps the model in sync with the business.'; its FAQ question 'Can I use Incremental to run A/B tests?' is answered (per the fetch tool's summary) that Incremental does not run A/B tests itself but integr | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | How-it-works: 'Multiple causal techniques link investment to outcome, not the click that came last.'; 'An ensemble causal model decomposes sales into what caused them.'; 'Econometric techniques combined with designed and synthetic experiments separate media impact from other factors by retailer and channel, down to the line item.'; 'Daily retraining keeps the model in sync with the business.' A diagram's alt text lists 'Econometrics' (holdout, calibration, validation), 'Active Experimentation with bar charts labeled geo-based and randomized control' and 'Passive Experimentation' ('Quasi Experi | supported with qualification | Replace the fetch-tool paraphrase with the verbatim FAQ answer ('While Incremental doesn't directly run A/B tests, it excels in integrating and interpreting the results from such tests conducted across different retail platforms like Amazon, Walmart, or other channels'; 'These types of A/B tests are used for both validation and calibration of the model’s predictions'), and add that a how-it-works diagram (alt text) shows 'Active Experimentation' labelled 'geo-based and randomized control' among the inputs; the reported default stays a model-based counterfactual with no assignment by Incrementa | The FAQ paraphrase was accurate. The home page's 'Brands running on iROI see 15 to 25% more incremental sales' is a second unsourced performance claim beside the '15%' tile. | RTL-G2-F26 (evidence/verify/G2.json) |
| C204 | Descriptive profile, LiftLab: an Agile MMM that 'models retail media as an explicit channel alongside national media and DTC' plus an Incrementality Testing Suite that 'provides controlled reads' 'Where retail media can be tested cleanly' (blog, 10 September 2026); the testing engine page states 'both Stratified Random Sampling and Synthetic Controls are supported' for geo selection, an audit trail 'showing how test and control markets were assigned', results 'as ranges rather than single lift numbers' with 'Confidence intervals are explicit, not hidden.', and feedback 'directly back into your | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | 'This incrementality testing platform isolates causal lift by running controlled geo experiments, interpreting results as ranges rather than single lift numbers, and feeding those findings directly back into your AMM as calibration inputs.'; 'both Stratified Random Sampling and Synthetic Controls are supported, with each deployed where it performs best'; 'LiftLab's geo selection, whether Stratified Random Sampling or Synthetic Controls, produces a full audit trail showing how test and control markets were assigned. Confidence intervals are explicit, not hidden. Spillover contamination is detec | supported with qualification | Recode assignment: random for the Stratified Random Sampling option; unknown for the Synthetic Controls option (a synthetic control is a counterfactual construction and the page does not say how treated markets are chosen under it); and add the documented contamination check ('Spillover contamination is detected and corrected before results are calculated'). | Assignment coded from a method label: the measurement record's 'non-random (synthetic-control option)' fills assignment from the name 'Synthetic Controls', which the control-coding rule forbids. | RTL-G2-F37 (evidence/verify/G2.json) |
| C205 | Decision (a): the Numerator Total Commerce Panel whitepaper (RTL-G2-S20), stamped client-only but publicly served, may support RTL-G2-F07, RTL-G2-F09 and the Numerator table cells. | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | The PDF is served publicly (HTTP 200, application/pdf, 2,231,873 bytes, 20 pages, creation date 16 March 2023) from numerator.com's uploads path and carries 'Confidential & Proprietary - Non-Public Distribution to Numerator Clients Only - All Other Disclosure Prohibited' on pages 2 to 19 (the cover and back page carry no stamp in the text layer). The public 10 April 2025 release gives the 105,000-to-200,000 expansion, the two-trips rule, 'Census Alignment' ('Panelist alignment with 35+ U.S. attributes within 8 key demographics categories') and 'In-Market Promo Dampening', but no projection or | supported with qualification | Mark RTL-G2-F07 and RTL-G2-F09 supported_with_qualification with the correction 'rest on the public 10 April 2025 release (RTL-G2-S22) and drop the whitepaper quotes'. Also: in RTL-G2-F06's measurement record replace 'projection to Total U.S. documented for the Total Commerce Panel' with 'no public projection method documented'; in measurement_services (Numerator row) replace the projection_method cell with ''Census Alignment' asserted (release of 10 April 2025); projection method not publicly documented' and drop ''Receipt Hog app' per 2023 whitepaper' from panel_or_feed; in panel_and_feed_sc | Claim, not finding: the source note's 'EVERY PAGE IS MARKED' should read 'pages 2 to 19 of 20 are marked'. | RTL-G2-S20-decision-a (evidence/verify/G2.json) |
| C206 | Decision (b): RTL-G2-F01's Circana Media Lift coding (assignment unknown; matched counterfactuals) rests on the session's browser-pane quotes 'Household Lift™ service gives you the most precise, household-level A/B testing in the industry' and 'Market Lift™ ... With a thorough matching process'. | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | Same passages as RTL-G2-F01: the relayed quotes are verbatim and complete sentences on the page; 'thorough matching process' is in the Market Lift paragraph; Household Lift adds 'Enact controlled household tests'; no allocation is stated for either service. | supported with qualification | Keep 'assignment unknown' for both services. Recode Household Lift's counterfactual from 'matched audience' to 'control group asserted, construction not stated'. Re-record RTL-G2-S01 as opened directly with curl and a browser user agent on 3 November 2026 (full text), not as a relayed browser read. | Counterfactual for Household Lift coded from the wrong paragraph. | RTL-G2-S01-decision-b (evidence/verify/G2.json) |
| C207 | [measurement_services] product=Circana Media Lift; cohort=measurement; control_design=Household Lift: 'household-level A/B testing'; Market Lift: 'thorough matching process' (relayed browser read); assignment=unknown; assignment_unit=household (Household Lift); market or store (Market Lift); counterfactual=matched audience or store; matched market; estimation=unknown; metrics=incremental sales; '$0.59 in incremental sales per household' (DoorDash case); panel_or_feed=Circana household purchase data; point-of-sale data; size=not found; us_coverage=US production use (DoorDash campaign 2025-11-10 | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | See RTL-G2-F01 to F03: Household Lift states no matching; Market Lift reports 'projected sales impact, and ROAS' with 'a thorough matching process'; scale stated as 'Tracking $5.8T in consumer spend' and household data with '4x ACV versus the nearest competitor'. | supported with qualification | counterfactual: 'Household Lift: control group asserted ('Enact controlled household tests'), construction not stated; Market Lift: matched market or store'; metrics: add 'ROAS and projected sales impact (Market Lift)'; size: ''Tracking $5.8T in consumer spend'; '4x ACV versus the nearest competitor'; no household count'; projection_method: ''projected sales impact' (Market Lift), method not stated'; source_id: add the direct read of RTL-G2-S01. | Counterfactual cell mis-coded for Household Lift. | RTL-G2-TABLE-measurement_services-Circana (evidence/verify/G2.json) |
| C208 | [measurement_services] product=Numerator sales lift; cohort=measurement; control_design='sales lift studies that analyze exposed vs non-exposed consumers'; SecureMeasure CRM case: pre/post buy-rate index; assignment=unknown; assignment_unit=household; counterfactual=unknown (unexposed comparison; construction not stated); pre/post (CRM case); estimation=unknown; metrics=sales lift; buy rate; panel_or_feed=receipt-based purchase panel ('Receipt Hog app' per 2023 whitepaper); size='1 Million+ US households' (media page); 200,000 static US households (release 2025-04-10); 'over a million consumer | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | Media page addresses media companies and sells sales lift studies to prove 'your advertisers’ campaign investments'; the projection method and 'Receipt Hog app' come only from the client-only whitepaper. | supported with qualification | projection_method: ''Census Alignment' asserted (release of 10 April 2025); projection method not publicly documented'; panel_or_feed: 'receipt-based purchase panel ('zero-party data from over 1 million US households')'; independence_fields: add 'studies marketed to media companies to prove their advertisers' campaigns; funding unknown per study'; drop RTL-G2-S20 from source_id. | Independence field omits that media sellers are an addressed buyer of the sales-lift studies. | RTL-G2-TABLE-measurement_services-Numerator (evidence/verify/G2.json) |
| C209 | [measurement_services] product=Nielsen Sales Lift; cohort=measurement; control_design='industry-standard control/exposed methodology and retailer-verified purchase data'; assignment=unknown; assignment_unit=unknown; counterfactual=unknown (control group; construction not stated); estimation=unknown; metrics='incremental sales'; 'sales attributed to marketing'; panel_or_feed=retailer purchase data via 'third-party integrations or proprietary datasets' (partners not named); size=not stated; us_coverage=not stated for Sales Lift; Predictive Sales Lift 'presently available in the U.S. only'; proje | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | Sales Lift page as coded; the Predictive Sales Lift release says 'Publishers, meanwhile, will benefit' and offers the feature to 'current Nielsen ONE Ads customers'. | supported with qualification | independence_fields: replace 'funding the advertiser' with 'funding unknown (advertisers, agencies or publishers may commission)'. | Funding filled by assumption. | RTL-G2-TABLE-measurement_services-Nielsen (evidence/verify/G2.json) |
| C210 | [measurement_services] product=Attain; cohort=measurement; control_design='advanced holdout methodologies'; exposure by tags, logs or CAPI; 'exposure-to-panel match' validated before measurement; assignment=unknown; assignment_unit=consumer (panel member); counterfactual=holdout (construction not disclosed); estimation=unknown; metrics='Total incremental sales lift'; ROAS; iROAS; panel_or_feed=opted-in consumer panel (receipts via email scanning and photo uploads; linked bank, debit and credit accounts) from Attain's apps Klover, Frisbee, Merryfield; size='14M+' opted-in consumers; '10M+' rece | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | Separate 'Sales Attribution', 'New Buyer Lift', 'Brand Lift' and 'Incrementality' products; about page: data extended to 'brands, agencies, retailers and media platforms'. | supported with qualification | control_design: 'Sales Attribution (exposed buyers, no counterfactual) and Incrementality ('advanced holdout methodologies'); exposure by tags, logs or CAPI; 'exposure-to-panel match' validated before measurement'; metrics: add 'new buyer lift; brand lift'; independence_fields: add 'serves retailers and media platforms; funding unknown per study'. | Attribution product folded into the holdout coding; independence field omits seller customers. | RTL-G2-TABLE-measurement_services-Attain (evidence/verify/G2.json) |
| C211 | [measurement_services] product=LiftLab; cohort=listed (descriptive profile); control_design=Agile MMM with retail media as explicit channel; geo experiments with stratified random sampling or synthetic controls; confidence intervals; assignment=random or non-random (option); assignment_unit=geography (market); counterfactual=geo holdout; synthetic control; estimation=ranges with confidence intervals (method not named); metrics=revenue; profit; lift; panel_or_feed=client spend and outcome data; size=not stated; us_coverage=US address (Oakland, CA); projection_method=n/a; validation=none found ( | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | As RTL-G2-F37. | supported with qualification | assignment: 'random (Stratified Random Sampling option); unknown (Synthetic Controls option)'; validation: add 'spillover contamination detected and corrected (vendor statement)'. | Assignment filled from the label 'Synthetic Controls'. | RTL-G2-TABLE-measurement_services-LiftLab (evidence/verify/G2.json) |
| C212 | [panel_and_feed_scale] product=Numerator sales lift; figure=panel households (floor); value=1000000; unit=US households; scope='sourced directly from 1 Million+ US households' (service page); source_id=RTL-U-S213; period=undated (© 2026); claim_status=actual || product=Numerator sales lift; figure=Total Commerce Panel static households; value=200000; unit=US households; scope='from 105,000 to 200,000 U.S. households' (consumer-insights static panel); source_id=RTL-G2-S22; period=2025-04-10; claim_status=actual || product=Numerator sales lift; figure=Total Commerce Panel static households (prio | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | 'sourced directly from 1 Million+ US households' is the Data Feeds blurb on the media-industry page (SecureMeasure also says 'consumer panel of 1M+ households'); 200,000 and 105,000 confirmed in the release; 150,000 and 500,000 come only from the client-only whitepaper. | supported with qualification | Drop the 150,000 ('static panel households (2023 design)') and 500,000 ('monthly active participating users') rows; change the 1,000,000 row's scope to 'Data Feeds blurb on the media-industry page'. | Decision (a). | RTL-G2-TABLE-panel_and_feed_scale-Numerator (evidence/verify/G2.json) |
| C213 | [panel_and_feed_scale] product=Catalina HUB360; figure=households in Catalina's US deterministic data; value=130000000; unit=households; scope=network-wide, not scoped to HUB360; source_id=RTL-U-S218; period=2026-02-23 release ('annual', undated); claim_status=actual || product=Catalina HUB360; figure=shopper IDs; value=400000000; unit=shopper IDs; scope='more than 400 million shopper IDs'; source_id=RTL-U-S218; period=2026-02-23; claim_status=actual || product=Catalina HUB360; figure=retail banners in the data; value=70; unit=banners; scope='seventy major retail banners'; source_id=RTL-U-S218 | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | 'four decades, seventy retail banners, 130 million households, $600 billion in annual spending'; 'more than 400 million shopper IDs'; 'seventy major retail banners'. | supported with qualification | Relabel the 130,000,000 row 'households in Catalina's network (geography not stated in the sentence)'; optionally add '$600 billion in annual spending' and '11 billion annual shopping trips' as their own units. | RTL-G2-TABLE-panel_and_feed_scale-Catalina (evidence/verify/G2.json) | |
| C214 | [panel_and_feed_scale] product=Attain; figure=opted-in consumers (floor); value=14000000; unit=consumers (not households); scope='14M+' opted-in consumers; source_id=RTL-U-S226; period=undated (© 2026); claim_status=actual || product=Attain; figure=receipts captured per month (floor); value=10000000; unit=receipts per month; scope='10M+'; source_id=RTL-U-S226; period=undated (© 2026); claim_status=actual || product=Attain; figure=linked retailers (floor); value=50; unit=retailers; scope='50+' linked retailers; source_id=RTL-U-S226; period=undated (© 2026); claim_status=actual | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | '14M+ Opted-in consumers', '10M+ Receipts captured per month', '50+ Linked retailers' confirmed; the same block also states '$700B+ Cumulative annual spend' and '$1.5T+ Cumulative transactions in/out'. | supported with qualification | Add two rows: '$700B+' cumulative annual spend (US$, floor) and '$1.5T+' cumulative transactions in/out (US$, floor), both undated. | RTL-G2-TABLE-panel_and_feed_scale-Attain (evidence/verify/G2.json) | |
| C215 | [panel_and_feed_scale] product=Circana Media Lift; figure=panel or household count; value=None; unit=households; scope=not found in the opened record; source_id=RTL-G2-S01; period=2026-11-03; claim_status=n/a | Refute-mode verifier for evidence/research/G2.json re-opened the cited source. | No household count on the page; 'Tracking $5.8T in consumer spend'; household data with '4x ACV versus the nearest competitor'. | supported with qualification | Keep the null household row; add a row '$5.8T' consumer spend tracked (US$, geography and period not stated) from RTL-G2-S01 read directly. | RTL-G2-TABLE-panel_and_feed_scale-Circana (evidence/verify/G2.json) | |
| C216 | Walmart Connect's current offsite media page pools its off-site channels ('VIZIO Home Screen ads, the DSP, CTV, and social media (Meta, TikTok, Pinterest)') under one measurement statement, 'connect real-time impressions, clicks, and conversions to verified Walmart transactions online', and its only published CTV result rests on 'a pre/post-exposure 14-day attribution' (Walmart first-party data, 1 February 2024 to 31 January 2025); neither that page, the network measurement page nor the integrated measurement solutions page (read 3 November 2026) states a match mechanism, a window for other ch | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Channel tabs read 'VIZIO', 'DSP', 'CTV', 'Social media'; the tab texts are 'VIZIO Home Screen ads are integrated into the VIZIO TV operating system ...', 'Powered by The Trade Desk, Walmart's demand-side platform ...', 'Our CTV solutions sit at the intersection of premium content, retail-powered audiences and performance insights ...', 'Extend your brand story across Meta, TikTok or Pinterest ...'. Measurement: 'Optimize campaigns with near real-time performance insights / Track the full impact of your campaigns through performance dashboards and reports that connect real-time impressions, cli | supported with qualification | Replace the quoted channel string with the page's own wording (tabs 'VIZIO', 'DSP', 'CTV', 'Social media'; 'Extend your brand story across Meta, TikTok or Pinterest'), and state the CTV result itself: a 28% average increase in online attributed sales among CTV audiences against the same audience before exposure, 133 offsite CTV campaigns, Walmart first-party data 1 February 2024 to 31 January 2025, pre/post-exposure 14-day attribution. The measurement-page and integrated-measurement-page parts were not re-opened by the verifier. | The string 'VIZIO Home Screen ads, the DSP, CTV, and social media (Meta, TikTok, Pinterest)' is workstream U's paraphrase presented in quotation marks; it is not page text. The 28% is an attributed figure on a pre/post comparison of the same audience (assignment none), not a lift. | RTL-H-F02 (evidence/verify/H.json) |
| C217 | The Walmart-Roku shoppable-ads pilot was announced jointly on 16 June 2022 (read through a reprint of the Business Wire release) ...; Walmart's 15 June 2023 post dates the pilot 'from November 2022 to February 2023' (footnotes '11/22/22 - 2/10/23'), reports 'at least three times higher clickthrough rates than average video campaigns' on 'Roku and Walmart 1st Party Data', says advertisers 'receive insights on effectiveness through Walmart Connect's omnichannel measurement reports', and announces an expansion 'to include their premium inventory within our CTV offering'; neither text describes ho | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | 'During the pilot, which ran from November 2022 to February 2023, these innovative shoppable ads demonstrated strong performance ...' / 'These shoppable ads received at least three times higher clickthrough rates than average video campaigns powered by the Walmart DSP.' (footnote 3: 'Roku and Walmart 1st Party Data, 11/22/22 - 2/10/23') / 'Advertisers also receive insights on effectiveness through Walmart Connect's omnichannel measurement reports.' / 'Now, we're expanding our Roku partnership to include their premium inventory within our CTV offering now available on the Walmart DSP.' | supported with qualification | Restate the two truncated passages in full: the clickthrough comparison is against 'average video campaigns powered by the Walmart DSP', and the 2023 expansion puts Roku premium inventory 'within our CTV offering now available on the Walmart DSP'. The Roku CTV inventory therefore names a buying route (Walmart DSP) and belongs to the Walmart DSP row in data table 'offsite_rails'; the shoppable pilot's buying route remains unstated. | The 2022 joint release (01net.it reprint; Business Wire refuses) was not re-opened, so the Roku Pay and checkout passages are unverified here; the 2023 post's 'seamless, on-screen checkout experience with a Roku remote' is consistent with them. No match mechanism is described in the 2023 post. | RTL-H-F03 (evidence/verify/H.json) |
| C218 | Walmart Connect's article of 8 April 2026 states that 'Closed-loop measurement via LiveRamp's data clean room is now available for Meta', that 'Closed-loop measurement is also available for TikTok and Pinterest' (no mechanism named for those two), that it measures 'both online and in-store sales at Walmart', that self-serve social campaigns in the Walmart Connect Ad Center start with Meta ('Self-serve access is currently available for Meta and TikTok' appears alongside 'expanding to other platforms, such as TikTok, later this year'), and that 'Add-to-Cart is currently available in beta for man | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Dated 'April 08, 2026'. 'Closed-loop measurement via LiveRamp's data clean room is now available for Meta. Closed-loop measurement is also available for TikTok and Pinterest.' / 'Self-serve access is currently available for Meta and TikTok self-service is planned to rollout this half. An API for tech partners is also planned for later this year.' / 'For the first time, self-serve advertisers can use Walmart Connect-powered social media ... — starting with Meta and expanding to other platforms, such as TikTok, later this year.' / General Mills quote: 'The new Walmart Connect integration with Li | supported with qualification | Restate the self-serve clause: self-serve social access is available for Meta, and TikTok self-service is 'planned to rollout this half' (the page sentence is a run-on without a full stop after 'Meta and TikTok'); there is no inconsistency. Restate the results clause: the 2.5x incremental ROAS and 1.8x sales lift come from a sales-lift benchmark analysis performed by Circana (September 2024 to September 2025, campaigns that qualified for a sales lift study in Food, Consumables, Beverages or Alcohol only, benchmarks supplied by Circana), with no design, sample size or interval disclosed; the 48 | The quoted fragment 'Self-serve access is currently available for Meta and TikTok' is a truncation that manufactured the tension recorded in counterevidence_and_tensions; delete that tension. The reporting party for the 2.5x and 1.8x figures is Circana (a named third party whose independence is the operator's description; funding unknown), not the operator alone. | RTL-H-F06 (evidence/verify/H.json) |
| C219 | Walmart's 15 June 2023 Partner Lab post describes four further off-site rails with 'closed-loop' wording and no mechanism: NBCUniversal live sports streaming (...; Danone 'more than 30% new-to-brand buyers in one CTV campaign'), TikTok in-feed ads ('This is the first time any retailer has successfully provided closed-loop measurement for TikTok campaigns'), TalkShopLive livestreams, and Roku; its 6 October 2026 article adds Spotify (...) and live linear television bought 'Through private marketplace deals available in Walmart DSP' with FreeWheel's technology ('55% incremental reach beyond line | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | NBCUniversal: 'We're partnering with NBCUniversal to test a first-to-market offering: closing the loop on retail media-powered ads within live sports streaming.' / 'we've combined NBCUniversal's premium inventory with our targeting and omnichannel, closed-loop measurement capabilities as part of our premium CTV bundle'; Danone's quote: 'Utilizing purchase-based targeting with Walmart Connect's premium CTV bundle ... we connected to more than 30% new-to-brand buyers in one CTV campaign'. TikTok: after the quote of Jen Bryce (Unilever), which closes with '... driving an omnichannel purchase expe | supported with qualification | Roku's 2023 expansion does name a buying route ('within our CTV offering now available on the Walmart DSP'), so 'name no buying route' is wrong for Roku and the Roku CTV rail moves to the Walmart DSP row. Attribute the TikTok 'first time any retailer' sentence as text in Walmart's post immediately after Unilever's quote and written in an advertiser's voice ('our campaigns', 'our videos'), not as a plain Walmart claim. The Danone 30% new-to-brand figure sits in a quote about the 'premium CTV bundle' that combines NBCUniversal premium inventory with Walmart targeting and measurement. | The 6 October 2026 article (RTL-H-S08) was not re-opened (budget); its Spotify and live-linear passages are unverified here. No passage in the 2023 post states an identity mechanism. | RTL-H-F07 (evidence/verify/H.json) |
| C220 | The Home Depot's release of 22 April 2026 announces 'a new beta integration with Pinterest Media Network Connect' through which 'brands not sold at The Home Depot' can run Pinterest campaigns on 'more than 20 first-party audiences', calls it 'the third environment where non-endemic advertisers can activate The Home Depot audiences, following integrations with The Trade Desk and Yahoo', and names no attribution, clean-room or closed-loop mechanism; Orange Apron Media's measurement page separately promises closed-loop measurement 'to both online and in-store sales and conversions' and a clean-ro | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Dated 'April 22, 2026'. 'Through a new beta integration with Pinterest Media Network Connect, Orange Apron Media is expanding how non-endemic advertisers reach inspiration led home improvement audiences.' / 'The partnership enables brands not sold at The Home Depot to run Pinterest campaigns powered by more than 20 first‑party audiences built on insights from billions of annual store and site visits and over a billion transactions.' / 'Pinterest Media Network Connect is the third environment where non‑endemic advertisers can activate The Home Depot audiences, following integrations with The Tr | supported with qualification | Release part verified as stated. The Orange Apron Media measurement-page passages (RTL-U-S049, RTL-U-S051) were not re-opened by the verifier. | The three named environments are sold for 'non-endemic advertisers' and 'brands not sold at The Home Depot', which is cohort B use under the brief's cohort rule; recording them as rails of the cohort A Orange Apron Media row needs that flag in data table 'offsite_rails' (the finding notes the missing cohort B row). They should not enter the H4b population as cohort A rails. | RTL-H-F09 (evidence/verify/H.json) |
| C221 | Instacart's release of 10 June 2025 (PR Newswire, read through a reprint because the issuer host geo-redirects) says agency partners such as Omnicom's Flywheel 'can access closed-loop measurement from Instacart across all of their campaigns' and that brands 'can access real-time sales measurement signals from Instacart directly in The Trade Desk platform', for audiences used 'across channels like CTV, audio, video, and display'; it describes 'Instacart's deterministic purchase-based data' but names no identity mechanism (no UID2, clean room or hashed identifier), and mentions a 'beta' only in | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Issuer copy (Internet Archive capture of 12 July 2026; 'Published Jun 10, 2025'; 'SAN FRANCISCO, June 10, 2025 –'): 'Partners like Omnicom's Flywheel, leveraging data from Instacart in their campaigns, can access closed-loop measurement from Instacart across all of their campaigns, including attributed sales and return on ad spend (ROAS), directly within The Trade Desk platform.' / 'CPG brands can also access real-time sales measurement signals from Instacart directly in The Trade Desk platform.' / Jeff Daniel (The Trade Desk): 'By integrating Instacart's deterministic purchase-based data and | supported with qualification | Add: the issuer copy was read (Internet Archive raw capture of company.instacart.com, 12 July 2026), so the release no longer rests on a reprint; 'including attributed sales and return on ad spend (ROAS)' is verbatim release text, not a fetch-tool summary; and Flywheel's quote names 'the Conversion API between Instacart and The Trade Desk' as the route it tested. A conversions API is a data route whose matching key and match location the release does not state, so under the fixed rule it earns no outside-match credit, but it must be recorded beside Kroger's 'CAPI-based' DV360 measurement and t | H4b five fields for 'Instacart display and video ads' (Trade Desk rail): exposure = The Trade Desk campaign delivery across CTV, audio, video and display; identity linkage = not documented (Conversion API named, key and join unstated); transaction source = Instacart purchase data ('deterministic purchase-based data'); modelled components = not stated; validation = none (Danone 'beta'). Linkage sea | RTL-H-F10 (evidence/verify/H.json) |
| C222 | Albertsons Media Collective's off-site offering runs 'Offsite Display through Google's DV360 and Collective TV' and a self-service route with The Trade Desk (...); its 2026 multi-touch attribution release, 'Developed in collaboration with LiveRamp', 'connects impression-level exposure data across Albertsons Media Collective's owned and operated channels' and offsite partners (Meta; DV360, Google Ad Manager, Perion) 'with Albertsons transaction data in a neutral, privacy-conscious environment', 'isolating the incremental contribution of each touchpoint to a sale' with PepsiCo as 'early pilot pa | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Release, verbatim: 'Developed in collaboration with LiveRamp under a first-to-market, grocery exclusive collaboration and powered by its Cross-Media Intelligence infrastructure, the solution connects impression-level exposure data across Albertsons Media Collective's owned and operated channels, offsite media and participating partners including Meta, third-party inventory with Display & Video 360, Google Ad Manager and Perion, with Albertsons transaction data in a neutral, privacy-conscious environment.' Subhead: 'Built in partnership with LiveRamp, the solution gives advertisers a single, de | supported with qualification | Replace the limitation that the off-site channel list 'comes from the fetch tool's summary' with the verbatim sentence above: the release itself names 'offsite media and participating partners including Meta, third-party inventory with Display & Video 360, Google Ad Manager and Perion'. Add that Albertsons' in-store release states incrementality was already brought 'onsite and offsite' (method undisclosed), which belongs in the validation field. Date: search leads (Shelby Report URL dated 18 September 2026; search summaries) put the release at 17 September 2026; not confirmed from an opened pa | The release never uses the words 'clean room' or 'identity'; coding the 'neutral, privacy-conscious environment' as a clean room is the stream's inference, which the fixed rule accepts only if a neutral third-party join environment counts as a clean room. | RTL-H-F14 (evidence/verify/H.json) |
| C223 | Roundel documents off-site measurement for Programmatic by Roundel ('Receive closed-loop programmatic reporting directly in demand-side platforms like The Trade Desk'; 'key sales metrics directly via The Trade Desk or Reports') and CTV by Roundel ('closed-loop reporting with product sales numbers, along with standard video metrics'; 'new purchaser insights data for products sold at Target (availability based on investment level)'), on 'First-party data from 100M+ Target Circle members', and its measurement page says only that 'We use privacy-conscious methodologies and secure data collaboratio | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Programmatic page: 'Receive standard media metrics (available in any DSP), in addition to key sales metrics directly via The Trade Desk or Reports (our self-service reporting platform).' / 'First-party data from 100M+ Target Circle™ members ...'; the sentence 'Receive closed-loop programmatic reporting directly in demand-side platforms like The Trade Desk.' sits in the page's footer text block. CTV page: 'Receive closed-loop reporting with product sales numbers, along with standard video metrics.' / 'Explore new purchaser insights data for products sold at Target (availability based on investm | supported with qualification | Delete 'says only': the measurement page also documents a multi-touch attribution model and incrementality 'for eligible campaigns' by synthetic control, matched market and randomised controlled trials, at network level and not tied to either off-site rail. The rest stands: no partner, window or control is named for the Programmatic or CTV rail, and the 'closed-loop programmatic reporting' sentence is in the page's footer text. | H4b under the fixed rule: 'secure data collaboration environments' is a generic network privacy statement, so no outside-match credit (L1 in the lenient count). Linkage search is incomplete: a LiveRamp release of 23 October 2025 (search lead, not opened) reportedly quotes Roundel's VP of product management on combining Target first-party insights with Meta ad data through clean-room measurement; i | RTL-H-F15 (evidence/verify/H.json) |
| C224 | Walgreens Advertising Group's offsite page (footer 2024) promises 'SKU-level targeting and omnichannel, closed loop reporting features' with no mechanism, partner, window or control, and the group is on The Trade Desk's 17 June 2026 partner list; Lowe's Media Network's releases ... promise 'In-depth measurement and closed-loop reporting' for the whole network ... | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | 'wagDSP drives efficiency and effectiveness by combining its proprietary wagID (historical and real-time, omnichannel sales data of 101M+ loyalty members) with powerful AI (key intent and contextual signals) across 15,000 publishers.' / 'wag audiences are available for immediate activation with The Trade Desk and OpenX. Plus, leverage SKU-level targeting and omnichannel, closed loop reporting features that help your teams drive optimal decision making at the line-level.' / 'Ads delivered on premium CTV ... with Infillion.' / 'Leverage wag's 1PD across Meta (Facebook & Instagram), Snap, and Pin | supported with qualification | Walgreens part verified as stated; add that the closed-loop sentence follows the Trade Desk and OpenX activation sentence (so it is tied to the off-site offer) and that CTV runs 'with Infillion'. 'wagID' is a targeting construct built from loyalty sales data, not a documented measurement join. The Lowe's part (RTL-U-S053) and the Trade Desk list (RTL-U-S188) were not re-opened. | H4b five fields for Walgreens Advertising Group: exposure = wagDSP across 15,000 publishers, The Trade Desk and OpenX activation, CTV via Infillion, Meta, Snap, Pinterest; identity linkage = not documented; transaction source = Walgreens omnichannel loyalty sales data; modelled components = not stated; validation = none. No outside-match credit. The page predates the 28 August 2025 take-private. | RTL-H-F17 (evidence/verify/H.json) |
| C225 | This review found no Samba TV partnership with a retail media network on samba.com (newsroom listing of six 2026 items read; one web search on 3 November 2026) or on any network page opened; Samba's documented purchase-outcome partnerships are with Catalina (release of 12 August 2020: 'a 1:1 deterministic match between TV exposure and in-store purchase' on 'more than five million frequent shopper cards', for Georgia-Pacific) and Affinity Solutions (16 October 2023: 'deterministic purchase data capturing over 140 million credit and debit cards' for 'cross-platform measurement of return on ad sp | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Catalina release: 'ST. PETERSBURG, FL, Aug. 12, 2020 –' / 'Fully anonymized, single-source Measurement Solutions with a 1:1 deterministic match between TV exposure and in-store purchase.' / 'ad exposures are measured via a national ACR TV panel covering four million households and a deterministic linkage to 25 million digital devices.' / 'Purchase data for 2.5 million TV households is linked to more than five million frequent shopper cards'; Catalina describes itself as 'the market leader in shopper intelligence and highly targeted in-store and digital media'. Affinity release: 'NEW YORK, Octo | supported with qualification | Restate the last clause as 'neither describes measurement of a retail media network's media: the Catalina arrangement measures a CPG brand's own TV and digital advertising against frequent-shopper purchases, and the Affinity integration measures against card transactions'; Catalina itself sells in-store and digital media, so 'neither of which is retail media' is too broad. | Interest mark present ('interest': 'Samba'); the Samba sources RTL-H-S28 to RTL-H-S30 do not repeat it in their notes, which is acceptable if the paper reads marks from findings. | RTL-H-F20 (evidence/verify/H.json) |
| C226 | H4b tally as of 3 November 2026, over the 28 cohort A rows whose register entry or opened pages name an off-site service: 13 rows document outcome measurement tied to an off-site channel, and 2 of those 13 (15%) document an identity match outside the operator's logged-in surface for a named off-site rail (...); counting network-level clean-room or match-key statements as well (Kroger, Roundel, Chewy) gives 5 of 13 (38%); widening the denominator to the 9 further rows with pooled network-level measurement statements gives 2 of 22 (9%) strict and 8 of 22 (36%) lenient. | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Best Buy's article (RTL-U-S047, cited by the stream for its off-site row) also says: 'A clearer picture with data clean rooms ... To make this possible, we've teamed up with LiveRamp, leaders in clean room technology.' / 'Through our clean room environment, you can work with us to activate defined, high-performing audience groups that blend with Best Buy data—or choose to “bring your own audience” for targeting within our network.' (a network-level clean-room statement, stated for audiences, not for the reporting join). The other passages behind the 2/13 hold verbatim (RTL-H-F04, RTL-H-F06, RT | supported with qualification | Restated claim: 'Of the 13 cohort A rows that tie outcome measurement to an off-site channel, 2 (15%) document in their own pages that the off-site exposure-to-transaction join runs outside the operator's logged-in surface: Walmart Connect Offsite Display (Disney's Audience Graph via clean room, 2024 proof of concept; LiveRamp's data clean room for Meta, 2026) and Albertsons Media Collective offsite (impression-level exposure from offsite media, Meta, DV360, Google Ad Manager and Perion joined with Albertsons transactions in a LiveRamp-powered neutral environment; 1 of 13, 8%, if that environm | The finding presents 15% as falling in the 'contradicted' branch 'subject to' conditions, but the hypothesis's registered rule makes any incomplete linkage search for a documented row an 'unresolved' outcome, and three rows' searches are incomplete on the verifier's evidence. | RTL-H-F21 (evidence/verify/H.json) |
| C227 | Best Buy Ads documents 'closed-loop, SKU-level reporting inclusive of online and in-store sales, in near real-time' for formats that include offsite display and offsite video, and an off-site incrementality design, 'Time-Based Regression Geographic Experimentation', 'splitting media investment geographically across regions that behave similarly', while on-site testing uses 'customer-level holdouts'; no identity mechanism is described for the off-site reporting. | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | 'With My Ads, our self-service platform, we offer measurement access to a handful of ad products. You can currently access closed-loop, SKU-level reporting inclusive of online and in-store sales, in near real-time, for offsite display, offsite video, paid search, onsite display and onsite video, plus sponsored products (online sales only).' / 'For offsite, we use Time-Based Regression Geographic Experimentation, splitting media investment geographically across regions that behave similarly.' / 'For onsite testing, we can be even more specific with customer-level holdouts' / 'we've teamed up wi | supported with qualification | Add that the same article names a LiveRamp clean room at network level, described for audience collaboration and targeting, not for the off-site reporting join; this counts in the lenient (network-level) H4b tally, which the stream omitted. The article is undated but refers to the 2025 Best Buy Ads Showcase. | H4b five fields: exposure = offsite display and offsite video delivery; identity linkage = not documented for the reporting (LiveRamp clean room named for audiences); transaction source = Best Buy online and in-store sales; modelled = not stated; validation = geo experiment with matched regions, allocation not stated. No outside-match credit under the fixed rule. | RTL-H-F25 (evidence/verify/H.json) |
| C228 | Walmart Connect's store ads page says 'Gross impressions are estimated projections based on historical and aggregated data.', 'Store ads measurement includes both brand and sales lift.' and 'Sales lift validated at the 80% confidence level or above.', citing a Deli and Bakery screen proof-of-concept that 'achieved up to 5% sales lift for Food suppliers'; it names no test design, control stores or impression formula, and Walmart's 2021 release counted 'nearly 170,000 digital screens across 4,500+ stores'. | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Footnote 3: 'ComScore, Dec 23, 2024 -April 24, 2025; Gross impressions are estimated projections based on historical and aggregated data. Actual campaign results will vary. Denotes a standard TV Wall campaign running 6/x for 2 weeks.' Footnote 4: 'Walmart first-party data, In-store Deli and Bakery screen campaigns, food suppliers only, September 2024 – January 2025. Walmart first-party, based on initial proof-of-concept testing. Sales lift validated at the 80% confidence level or above.' Body: 'Initial proof-of-concept testing has shown that Store ads campaigns achieved up to 5% sales lift for | supported with qualification | Add that the gross-impression projection is footnoted to ComScore data for 23 December 2024 to 24 April 2025 and to 'a standard TV Wall campaign running 6/x for 2 weeks' (a named data source and a reference campaign, still no formula or store-traffic method), and that the up-to-5% proof of concept covers Deli and Bakery screen campaigns, food suppliers only, September 2024 to January 2025, Walmart first-party data. | The design behind the 80% confidence threshold remains undisclosed; assignment and counterfactual stay unknown. | RTL-H-F31 (evidence/verify/H.json) |
| C229 | Grocery TV's pages and its 2026 media kit describe 'the largest in-store retail media network' with outcome claims ('Measure brand lift and sales impact tied to actual purchases'; 'Understand the impact on sales, ROAS, and iROAS both in Grocery TV stores and in other retailers'; an 'Average sales lift' of '+14%' in the media kit; 'an average 49% increase in incremental reach') but none of the four documents opened states how impressions are counted, which test design produced the lifts, or who measured them. | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Brands page: 'Measure brand lift and sales impact tied to actual purchases with an average of 14% sales lift and 27% brand awareness across brands.' / 'Brands that add in-store retail media to their campaigns, see an average 49% increase in incremental reach.' / 'Understand the impact on sales, ROAS, and iROAS both in Grocery TV stores and in other retailers.' / 'Partnered with 125+ retailers' / case tiles '34% Brand Awareness Lift for a FinTech App' and '19% Sales Lift for a Personal Care Brand'. Media kit p. 5: 'Average brand awareness lift +22%', 'Average sales lift +14%', 'Average ad recal | supported with qualification | Complete the brands-page quote: it states the 14% average sales lift itself and a 27% average brand-awareness figure, while the media kit gives +22% average brand awareness lift; add that inconsistency and the two case figures (19% sales lift, personal care; 34% brand awareness lift, FinTech app), none with a design, base, period or measurer. | Media kit read from the research agent's download (5,793,942 bytes), re-extracted. The home and about pages were not re-opened. | RTL-H-F34 (evidence/verify/H.json) |
| C230 | The in-store audio services of three cohort A networks document no audio-specific impression or outcome method on the pages opened: Dollar General ...; Lowe's ...; and Walgreens' only closed-loop sentences sit under coupons, not audio; Gulp Media Network claims 'an average overall sales lifts of 5–9%' for Gulp Radio with no design. | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | 7-Eleven, 'Irving, Texas (November 12, 2024)': 'Through Qsic's AI-powered creative audio development, 7-Eleven more efficiently generates localized audio advertisements that reach target audiences during peak times – all backed by first-party data to measure the impact and determine revenue lift. Products promoted through Gulp Radio 7-Eleven® locations have seen an average overall sales lifts of 5–9%.' / 'Qsic surpasses IAB standards by dynamically adjusting audio volume based on ambient noise levels, ensuring each ad is heard clearly. Additionally, Qsic's system timestamps each ad, providing | supported with qualification | Gulp part: add that the release documents ad-play logging ('Qsic's system timestamps each ad') and says lift is measured on 'first-party data'; the 5–9% still has no design, base, period or control, and timestamps are ad plays, not impressions. Dollar General, Lowe's and Walgreens parts were not re-verified (Dollar General's page is a JavaScript shell to curl; the Lowe's and Walgreens in-store pages were not re-opened). | The Gulp Radio store counts (4,000+ now; 5,000 by end-2024 and 12,000+ by end-2025, forecasts as of 12 November 2024) are missing from data table 'instore_figures'. | RTL-H-F37 (evidence/verify/H.json) |
| C231 | The in-store services opened label their screens and audio as retail media and none of their pages labels that inventory DOOH (Walmart Connect store ads, Albertsons Media Collective in-store, STRATACACHE, Grocery TV, Stingray), while IAB's DOOH guide files retail stores under 'Retail and Venue-Based DOOH' and the in-store standard points to MRC's digital place-based standards, so the same store screens carry both labels by convention; separately, Albertsons sells 'Albertsons-powered Digital Out-of-Home' on '120,000 digital screens near stores' on a page that never says 'retail media', and Roun | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Walmart, 28 January 2021 (RTL-H-S01, a source this stream cites): 'The retailer is now offering media activations on in-store TV walls and self-checkout screens with nearly 170,000 digital screens across 4,500+ stores. ... These advancements make Walmart Connect one of the largest in-store activations and digital out-of-home ad platforms in the country.' The store ads page names the product 'Store ads'; its only 'retail media' string is a navigation promo ('See how full-funnel retail media leads to real results'); it uses no DOOH label. Albertsons' in-store offerings page uses neither 'retail | supported with qualification | Restate: 'On the current pages screened, the in-store sellers call the inventory store ads or in-store (retail) media and do not use a DOOH label, but Walmart's own release of 28 January 2021 described its in-store TV walls and self-checkout screens as making Walmart Connect one of the largest in-store and digital out-of-home ad platforms, so the dual label is used by a seller as well as by IAB's DOOH guide.' Also note that Walmart's store ads page and Albertsons' in-store page do not themselves say 'retail media' in body text; the retail media label comes from the operators' releases and bran | The screen list in the finding's limitation is bounded and correct as far as it goes; the counter-example comes from a source outside that list but inside this stream. | RTL-H-F42 (evidence/verify/H.json) |
| C232 | M1 / L0: release names no identity mechanism (read via reprint). | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Five fields: exposure source = The Trade Desk campaign delivery 'across channels like CTV, audio, video, and display'; identity linkage = not documented, though Flywheel's quote names 'the Conversion API between Instacart and The Trade Desk' (no key or join stated); transaction source = 'Instacart's deterministic purchase-based data'; modelled components = not stated; validation = none ('beta' in Danone's quote). | supported with qualification | Change 'read via reprint' to 'issuer copy read via Internet Archive capture of 12 July 2026'; record the Conversion API mention as a sensitivity alongside Kroger's CAPI. | Outside-match credit: does not hold on this release. Linkage search incomplete: Instacart's Data Hub clean-room release (company.instacart.com/pressreleases/instacart-debuts-clean-room-offering, January 2026) and its off-platform solutions page surfaced as leads; Instacart hosts geo-redirect, so read them via archive captures. This row triggers the H4b 'incomplete linkage search' condition until t | TABLE h4b_row_tally: Instacart display and video ads (evidence/verify/H.json) |
| C233 | M1 / L2: LiveRamp-built 'neutral, privacy-conscious environment' joining impression-level exposure with Albertsons transactions (off-site channel list from the fetch summary). | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Five fields: exposure source = 'impression-level exposure data across Albertsons Media Collective's owned and operated channels, offsite media and participating partners including Meta, third-party inventory with Display & Video 360, Google Ad Manager and Perion'; DV360 Offsite Display and Collective TV; Trade Desk CTV; identity linkage = join 'with Albertsons transaction data in a neutral, privacy-conscious environment', 'powered by' LiveRamp's 'Cross-Media Intelligence infrastructure' (no 'clean room' or 'identity' wording); transaction source = 'Albertsons transaction data'; 'customer sales | supported with qualification | Replace '(off-site channel list from the fetch summary)' with 'off-site channel list verbatim in the release'. | Outside-match credit: holds narrowly (a neutral third-party join environment for off-site impressions and retailer transactions); fails if the rule is read to require the words 'clean room' or an explicit identity match. Sources accessible. The LiveRamp release of 23 October 2025 (lead) would make the Meta join explicit. | TABLE h4b_row_tally: Albertsons Media Collective offsite (evidence/verify/H.json) |
| C234 | M1 / L1: 'secure data collaboration environments' (measurement page, network-level). | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Five fields: exposure source = DSP delivery (The Trade Desk named; 'the DSP of your choice'), Bullseye Marketplace publishers, CTV; identity linkage = 'privacy-conscious methodologies and secure data collaboration environments' (network-level privacy statement); transaction source = Target sales ('key sales metrics', 'product sales numbers'); modelled components = 'multi-touch attribution model' (network-level); validation = incrementality 'available for eligible campaigns' using 'synthetic control, matched market and randomized controlled trials' (network-level; allocation per test unknown). | supported with qualification | Add the modelled-components and validation entries above to data table 'offsite_rails' for both Roundel rows (as network-level, not rail-specific). | Outside-match credit: does not hold on Roundel's own pages. Linkage search incomplete: the LiveRamp release of 23 October 2025 (lead) reportedly quotes Roundel's VP of product management on Meta measurement through the LiveRamp clean room; whether a partner release quoting the operator counts as 'its own documentation' is a session decision. | TABLE h4b_row_tally: Roundel (evidence/verify/H.json) |
| C235 | M1 / L0: geo experiment for off-site lift; reporting join not stated. | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | See RTL-H-F25: closed-loop SKU-level reporting for offsite display and offsite video; Time-Based Regression Geographic Experimentation offsite; a LiveRamp clean room named at network level for audiences. | supported with qualification | Recode L0 to L1 (network-level clean-room statement); the strict (L2) result is unchanged. | Outside-match credit: does not hold under the fixed rule; counts in the lenient tally. | TABLE h4b_row_tally: Best Buy Ads display, video, search and CTV media (evidence/verify/H.json) |
| C236 | M1 / L0: none stated ('attributed sales based on viewable impressions'). | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Five fields: exposure = Pinterest (viewable impressions); identity linkage = not documented on the dashboard page; transaction source = CVS transactions; modelled = not stated; validation = none. | supported with qualification | Code as L0 on the page opened, with the linkage search marked incomplete. | Outside-match credit: does not hold on this page; leads (Retail TouchPoints, Beet.TV March 2024, Drug Store News, Mass Market Retailers) report CMX joining the Pinterest–LiveRamp clean room to attribute off-site sales. This row triggers the H4b 'incomplete linkage search' condition until CMX's own announcement is read. | TABLE h4b_row_tally: CVS Media Exchange (evidence/verify/H.json) |
| C237 | primary 2/13 (15.4%); lenient 5/13 (38.5%); broad 2/22 (9.1%); broad lenient 8/22 (36.4%). | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Recount with Best Buy's network-level LiveRamp clean-room statement: primary 2/13 = 15.4% (1/13 = 7.7% if Albertsons' 'neutral, privacy-conscious environment' is not accepted as a clean room); lenient 6/13 = 46.2%; broad 2/22 = 9.1%; broad lenient 9/22 = 40.9%. | supported with qualification | Replace rows 2 and 4 with 6/13 (46.2%) and 9/22 (40.9%). Change every rule_branch to 'unresolved (incomplete linkage searches for Roundel, CVS and Instacart; Kroger browser-only)' until the leads are read; after that, 'contradicted' applies only if the strict count stays at four or fewer of 13. | Best case on the leads found: 5/13 (38.5%), between thresholds. Two-thirds (9/13) is out of reach on any lead found. | TABLE h4b_shares (evidence/verify/H.json) |
| C238 | One row per cohort A product x off-site rail with the five H4b fields. | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | See the per-finding entries for the passages. | supported with qualification | (1) Roku: split the row; Roku premium CTV inventory goes to Walmart DSP ('within our CTV offering now available on the Walmart DSP', 15 June 2023); the pilot's CTR comparator is 'average video campaigns powered by the Walmart DSP'. (2) Walmart Connect Offsite Display, Meta row: validation = 'Circana sales-lift benchmark analysis, September 2024 to September 2025, qualifying campaigns in four categories; design undisclosed' instead of 'none documented'. (3) Pooled Walmart offsite row: add the 28% CTV attributed-sales statistic (133 campaigns, pre/post-exposure 14-day attribution, February 2024 | Rows not re-checked (budget or refusal): VIZIO rows, Spotify, live linear, Kroger rows, DG, Amazon DSP, Lowe's, Gulp, Hy-Vee, Macy's, Petco, UB Media. | TABLE offsite_rails (evidence/verify/H.json) |
| C239 | In-store surfaces: impression and outcome methods and labels. | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Yardstick row verified (pp. 10–17). Walmart Store Ads: impression projection footnoted to 'ComScore, Dec 23, 2024 -April 24, 2025' and 'a standard TV Wall campaign running 6/x for 2 weeks'; outcome: brand and sales lift, 80% confidence threshold, PoC September 2024 to January 2025. Albertsons in-store: matched control stores, 'nearly 60 variables', the source's label 'matched market framework'. STRATACACHE: 'proof of play'. Grocery TV: brands page states '14% sales lift and 27% brand awareness' averages; media kit '+22%' brand awareness. Stingray: verified. Gulp Radio: 'Qsic's system timestamp | supported with qualification | Walmart Store Ads row: impression_method add 'projection footnoted to ComScore data (23 Dec 2024–24 Apr 2025) for a standard TV Wall campaign; formula and store-traffic method not disclosed'; labels: the 2021 Walmart release calls the screens a 'digital out-of-home ad platform'. Gulp row: impression_method 'ad plays timestamped by Qsic (no impression formula)'; outcome add 'first-party data to measure the impact'. Grocery TV row: add the 27% (brands page) versus 22% (media kit) brand-awareness inconsistency. Albertsons in-store row: add the release's own label 'matched market framework'. | Rows not re-checked: Walmart non-endemic, Vibenomics, DG audio (JavaScript shell), Lowe's, Walgreens in-store, AD Retail Media (refuses), Kroger in-store, Cooler Screens (403), Roundel DOOH. | TABLE instore_measurement (evidence/verify/H.json) |
| C240 | Numbers stated by in-store and near-store services with definitions and periods. | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | Verified as stated: 170,000 screens / 4,500+ stores (28 Jan 2021); 800 additional stores in 2026 (Albertsons, 10 divisions); 800 stores by 2026 (STRATACACHE); 120,000 near-store screens; 21,100 US stores, 30,000 North American locations and 20.3% average lift (Stingray, 20 June 2024; over 100 studies, 42 categories); 14% average sales lift (Grocery TV media kit and brands page); 49% incremental reach; 125+ retailers; 5–9% Gulp Radio lift (12 Nov 2024). Walmart's up-to-5% PoC is footnoted 'September 2024 – January 2025'. | supported with qualification | Walmart 5% row: period 'September 2024 to January 2025 (Deli and Bakery screen campaigns, food suppliers only, Walmart first-party data)' instead of 'undated'. Add: Walmart '4.6K+' US stores (annual report papers, 14 March 2025) and 'nearly one in five in-store Walmart shoppers' reminded by TV Wall ads (Walmart first-party data, February 2025; survey item, base not stated); Grocery TV 27% average brand awareness (brands page) beside the media kit's +22%, +21% ad recall, +13% consideration, and the 19% and 34% case figures; Gulp Radio 4,000+ stores now, 5,000 by end-2024 and 12,000+ by end-2025 | None of these figures is a measure of matchable transactions; units must stay apart. | TABLE instore_figures (evidence/verify/H.json) |
| C241 | '... 2 of those 13 (15%) document an outside identity match for a named rail, 5 of 13 (38%) if network-level clean-room statements count ... Instacart's release of 10 June 2025 puts 'closed-loop measurement' inside The Trade Desk without naming an identity mechanism.' | Refute-mode verifier for evidence/research/H.json re-opened the cited source. | See TABLE h4b_shares and RTL-H-F10. | supported with qualification | Change '5 of 13 (38%)' to '6 of 13 (46%)'; add that Instacart's release names a 'Conversion API between Instacart and The Trade Desk' without a key or join; state that H4b is unresolved under its rule until the Roundel, CVS and Instacart leads are read. | SUMMARY (evidence/verify/H.json) | |
| C242 | ANA's survey released 30 January 2023 was fielded 'Over the summer of 2022' among ANA members, with 'A total of 138 members participated in the survey, 80 of whom utilize RMNs', followed by 'nine qualitative follow-up interviews'; 88 percent 'believe they are somewhat or heavily influenced by retailers to buy advertising on their RMNs', and the base of that item (all 138 or the 80 RMN users) is not stated on either open ANA page. | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | ANA release (read 2026-11-03): "A total of 138 members participated in the survey, 80 of whom utilize RMNs."; "Over the summer of 2022, the ANA conducted an online survey of its members into their usage and management of RMNs."; "an overwhelming 88 percent believe they are somewhat or heavily influenced by retailers to buy advertising on their RMNs." (the captured text begins mid-sentence at "an overwhelming", so the clause before it, which could name the item's base, was not read). The interview sentence was returned only in paraphrase (nine qualitative interviews with senior marketers). WARC | supported with qualification | Keep the claim. Replace the fragment passage 'Over the summer of 2022.' with the full sentence quoted here (it adds 'online survey'). Flag the interview passage as not confirmed verbatim (only a paraphrase was captured). In limitations replace "'influenced' is the survey's word and is not 'pressured'" with: "The release reports 88% 'somewhat or heavily influenced by retailers to buy advertising on their RMNs'; WARC's summary of the same report (RTL-VI-S01) records that 85% of 'brands surveyed' 'feel pressured by retailers to support RMNS'. Whether these are one item reported two ways or two it | The paper must not say the ANA survey has no 'pressure' item: the absence holds only for ANA's open pages. | RTL-I-F01 (evidence/verify/I.json) |
| C243 | In the ANA 2023 survey (n = 138 members, summer 2022), 73 percent 'expect to be spending somewhat or significantly more on RMNs in the future', 56 percent 'are currently working with five or more different RMNs', and spending was described as 'shifted to RMNs from existing budgets for shopper marketing, brand marketing, and trade spending'; the words 'joint business', 'JBP' and 'pressure' do not appear on either open page. | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | Release (harvest 2026-11-03): "73 percent said they expect to be spending somewhat or significantly more on RMNs in the future." (sentence continues); "56 percent said they are currently working with five or more different RMNs."; "58 percent of respondents said they expect to be using more RMNs than they do now over the next two years."; "being shifted to RMNs from existing budgets for shopper marketing, brand marketing, and trade spending."; "Fifty-two percent believe RMNs will be viewed more positively" (31 percent today, rest paraphrased); the two-thirds and 12 percent goal items came back | supported with qualification | Restate the final clause as: "the words 'joint business', 'JBP' and 'pressure' do not appear on the ANA release; WARC's summary of the same report records a 'feel pressured by retailers to support RMNS' item (85%; see RTL-I-F01)". The report page was not re-screened for 'pressure' or 'JBP'. Update limitations: the 75/18 figures are now captured verbatim (quote above); the 18 percent is 'brand awareness', a different item from the release's 12 percent 'to invest for future brand growth'; the two-thirds versus 75 percent sales items remain unreconciled. | The support passage '56 percent ... and 58 percent of respondents said they expect to be using more RMNs.' joins two separate sentences with ', and'; split it into the two verbatim sentences. The two-thirds/12 percent passage is a fetch-tool paraphrase, not verbatim. | RTL-I-F03 (evidence/verify/I.json) |
| C244 | ANA's retail media hub page lists four reports, 'Retail Media Networks: A Forced Marriage or Perfect Partnership?' (2023-01), 'Retail Media Networks: Optimism Tempered with Caution' (2024-07), 'Retail Media Internal Management Guidance' (2025-09) and 'Retail Media Measurement Standardization' (2026-08), and describes a 'Retail Media Working Group, composed of ANA brand members who bring senior-level expertise in retail media and related functions'. | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | Hub page (opened 2026-11-03) lists four reports: "Retail Media Measurement Standardization" (rr-2026-08), "Retail Media Internal Management Guidance" (rr-2025-09), "Retail Media Networks: Optimism Tempered with Caution" (rr-2024-07), "Retail Media Networks: A Forced Marriage or Perfect Partnership?" (rr-2023-01); no date is shown beside any. The Retail Media Working Group paragraph was returned only in paraphrase (ANA brand members with senior-level retail media expertise; three key issues: measurement standards, attribution and internal structures). Verbatim on the hub: heading "Advancing Ret | supported with qualification | Keep the four titles and ids. Mark the Working Group passage as not confirmed verbatim (the fetch tool paraphrased it) and the two description passages as not re-captured. Replace the limitation's snippet-based authorship note with: "The 2025 report page (26 September 2025, gated) names no author, survey or sample; the search-snippet description of it as developed with Working Group member brands is not on the open page." | RTL-I-F05 (evidence/verify/I.json) | |
| C245 | ANA's 'Retail Media Measurement Standardization' report page (18 August 2026) is listed under Research Reports, says the report 'establishes an initial framework to help drive outcome comparability', is gated to 'All Logged-In Visitors', and names no author, committee, partner, survey, sample or standards body on the open page; the AAM page cited in the brief links it as 'Association of National Advertisers (ANA) Retail Media Network Measurement Standards', a name the ANA page does not use. | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | Report page (opened 2026-11-03): title "Retail Media Measurement Standardization"; date August 18, 2026; category "Research Reports"; "All Logged-In Visitors"; "This report on measurement standardization highlights key learnings and establishes an initial framework to help drive outcome comparability."; "Retail media continues to increase, with Emarketer projecting a 17 percent compound annual growth rate through 2028, reaching $90 billion."; no author, committee, working group, partner, sample or method named. ANA's hub page heads the section "Advancing Retail Media Standards" and lists "esta | supported with qualification | Keep the claim. Soften the counterevidence: AAM's 'Measurement Standards' link text is not foreign to ANA's own framing (the hub's heading and the Working Group's issue list use 'standards'); the report page itself says 'initial framework'. The tension is between a framework document and a standards label, not between ANA and AAM vocabularies. | RTL-I-F06 (evidence/verify/I.json) | |
| C246 | The Microsoft Advertising and Digital Shelf Institute joint-business-plan study is described by Microsoft (8 August 2024) as 'interviewing 18 senior-level executives across retailers and brands' and by DSI's report page as 'interviews 14 leading brands and retailers'; neither page gives interview dates, a sampling frame, a brand/retailer split or the method, the report is behind a DSI membership form, and Microsoft's sentence that improving the JBP process 'can increase ROI by 10%' carries no source. | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | Microsoft post (harvest): author "Nate Pinkston" / "HEAD OF GROWTH, MICROSOFT RETAIL MEDIA"; date August 08, 2024; "interviewing 18 senior-level executives across retailers and brands"; "can increase ROI by 10%, making it worth everyone's time to improve"; no source or footnote for the 10% sentence; no interview dates, sampling frame or split. DSI page (harvest): "The report interviews 14 leading brands and retailers to understand how each approaches the JBP process."; "By submitting this form, you are requesting to become a member of the Digital Shelf Institute."; no date beyond the footer co | supported with qualification | Replace 'the report is behind a DSI membership form' with: 'the DSI page carries a form whose submission is "requesting to become a member of the Digital Shelf Institute"; the page does not state that the download requires it'. | RTL-I-S08 funding_or_interest says 'DSI is a Salsify-affiliated institute per public knowledge, not stated on the page': an unsourced assertion; remove it or cite an opened source. | RTL-I-F07 (evidence/verify/I.json) |
| C247 | Digiday's 'WTF are JBPs?' (6 March 2025) concerns joint business plans between agency holding companies and media platforms ('an increasingly common tactic to remain on clients' media plans'), not retailer trade terms; its only figure is Jeff Green's (The Trade Desk) statement that JBPs 'grow 50% faster than the rest of our business', a vendor assertion with no method. | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | Digiday (opened 2026-11-03): Ronan Shields, March 6, 2025; JBPs "are an increasingly common tactic to remain on clients’ media plans"; "straight out of Big Tech’s playbook"; JBPs "were at the center of The Trade Desk CEO Jeff Green’s damage limitation exercise after his company’s recent earnings miss"; JBPs "grow 50% faster than the rest of our business"; "retail", "retailer", "Walmart" and "Amazon" do not appear in the article body; scope: agency, platform and ad-tech vendor deals (including brand-direct relationships), not retailer trade terms. | supported with qualification | Restate: "Digiday's 'WTF are JBPs?' (6 March 2025) concerns joint business plans between agencies or brands and media and ad-tech platforms ('an increasingly common tactic to remain on clients’ media plans'), not retailer trade terms (the body never mentions retail, Walmart or Amazon); its only figure is The Trade Desk CEO Jeff Green's statement, reported in the context of an earnings miss, that JBPs 'grow 50% faster than the rest of our business', a vendor assertion with no method." | The finding records trade reporting (class synthesis) that quotes one vendor assertion; code the 50% figure as vendor_assertion inside it rather than the whole finding. | RTL-I-F10 (evidence/verify/I.json) |
| C248 | This review found no survey of US advertisers with a stated sampling frame, field dates and n that measures retailer influence or joint-business-plan-linked retail media spend for 2025 or 2026, in the one web search run and the pages opened by 3 November 2026; the only opened surveys with influence items are ANA 2023 and ANA 2024 (member samples). | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | The verifier's own search on 2026-11-03 (2025 OR 2026 survey of US advertisers on retail media networks with 'joint business plan' OR 'JBP', respondents, pressure, spend) returned no US survey measuring JBP-linked spend or retailer influence; leads returned, none opened: TransUnion '2025 RMN report' landing page, a Skai annual survey (KPI expectations), a Forrester spend-intention figure (62%), IAB Australia's 2025 retail media survey (161 decision makers; not US), an Agile Brand Guide research page. WARC (opened) adds an 85% 'feel pressured' item, but from the same 2022 ANA fieldwork. | supported with qualification | Keep the bounded absence; add the verifier's search and its unopened leads to 'searches' and 'open_questions', and add the WARC-recorded 'pressured' item to the list of opened influence items (same ANA 2023 study, not a 2025-2026 survey). | RTL-I-F11 (evidence/verify/I.json) | |
| C249 | Omnicom Group Inc. completed its acquisition of The Interpublic Group of Companies, Inc. on 26 November 2025 ('on the Closing Date ... the Merger was consummated', with 'IPG continuing as the surviving corporation and a direct wholly owned subsidiary of Omnicom' at 0.344 Omnicom shares per IPG share), the 8-K naming 'IPG's Media, Data and Technology offerings, including IPG Mediabrands, Acxiom, KINESSO and Matterkind'; Omnicom had closed the Flywheel Digital acquisition on 2 January 2024. | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | 8-K (opened 2026-11-03): "Date of Report (Date of earliest event reported): November 26, 2025"; Items 1.01, 2.01, 2.03, 3.03, 5.02, 7.01, 9.01; Item 2.01: "...the Merger was consummated." and "...converted into the right to receive 0.344 shares of common stock..."; Introductory Note: "...Merger Sub merged with and into IPG (the "Merger")..." and "...with IPG continuing as the surviving corporation and a direct wholly owned subsidiary of Omnicom."; Item 5.02: "...including IPG Mediabrands, Acxiom, KINESSO and Matterkind."; "Flywheel" and "Hearts" do not appear. Omnicom Media's own site (Omnicom | supported with qualification | Correct the locators: 'Merger Sub merged with and into IPG' and 'with IPG continuing as the surviving corporation and a direct wholly owned subsidiary of Omnicom.' sit in the 8-K's Introductory Note, not Item 2.01. Replace the limitation's search-snippet sentence with the issuer record: Omnicom Media's release of 28 August 2026 (RTL-VI-S03) combines Hearts & Science and Mediahub into Hearts United; the Kinesso and Annalect retirement remains unverified (the release has no 'retire' sentence). Restate the interest mark as: "past employer's parent: Omnicom. Hearts & Science (the author's past emp | Hearts & Science's Omnicom ownership is now settled from an Omnicom-owned page; apply the same restated mark to RTL-I-F13 and to the Omnicom rows of trade_table and holding_company_practices. | RTL-I-F12 (evidence/verify/I.json) |
| C250 | WPP's own commerce page documents a commerce practice by awards and a vendor claim ('WPP earned the highest possible scores across 12 criteria in The Forrester Wave™: Commerce Services, Q1 2026 report') and carries no text on retail media buying tools, retailer partnerships, measurement independence, incrementality or the IAB/MRC standards; 'WPP Media' appears only in an article summary and 'GroupM' only in a staff biography. | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | wpp.com/en/commerce (opened 2026-11-03): "WPP earned the highest possible scores across 12 criteria in The Forrester Wave™: Commerce Services, Q1 2026 report…"; "Not just your marketing partner, your business partner."; "WPP Media's Deni Petrova and Unilever's Charlotte Murphy on building partnerships that drive full-funnel commerce"; GroupM only in a biography ("She joined WPP in 2019 as Global Head of Commerce at GroupM..."). Retail media text present: "Rethinking the agency-brand relationship in retail media"; "The Drum Commerce Media Awards 2026: Best Integration of Trade or Shopper Market | supported with qualification | Restate: "WPP's own commerce page documents a commerce practice by awards and a vendor claim ('WPP earned the highest possible scores across 12 criteria in The Forrester Wave™: Commerce Services, Q1 2026 report'); its retail media text is limited to an article title ('Rethinking the agency-brand relationship in retail media'), award titles (including 'Best Integration of Trade or Shopper Marketing and Retail Media for Nestlé') and a '2025 Amazon Ads Partner Awards' entry; it names no retail media buying tool, documents no retailer partnership terms, and carries no text on measurement, incremen | The 'trade or shopper marketing and retail media' award title is a weak pointer to retail media sold inside trade budgets; it is an award category, not evidence of JBP terms. | RTL-I-F15 (evidence/verify/I.json) |
| C251 | Kroger Precision Marketing's own pages (read in a browser on 3 November 2026) say its incremental measurement uses 'many IAB-recommended methods as well as test and control groups' and that its 'methodology adheres to IAB standards and best practices', while its FAQ states KPM is 'proudly certified by the Interactive Advertising Bureau (IAB) and Merchant Risk Council (MRC)', naming the wrong body for the MRC acronym; the Media Rating Council's own list shows 'Kroger (Pre-audit in process as of 6/2026)', not an accreditation. | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | Session browser notes (browser-notes-U.md, pane reads of 2026-11-03; the host refuses the fetch tool, so the verifier did not re-open the pages): FAQ: "Kroger Precision Marketing is proudly certified by the Interactive Advertising Bureau (IAB) and Merchant Risk Council (MRC)."; blog of 19 July 2024: "To determine the incremental impact of advertising, we use many IAB-recommended methods as well as test and control groups to create accurate comparisons across variables, such as pre-campaign spend, media interaction, loyalty and geography/division"; blog of 14 January 2025 ('New Incremental Sale | supported with qualification | Quotes match the browser notes exactly. Product mapping: the 'adheres to IAB standards' passage concerns KPM's programmatic (offsite, The Trade Desk) incremental measurement, so that part belongs to the register row 'Kroger Precision Marketing offsite programmatic'; the FAQ and the 2024 blog are network-wide. Either split the finding or set product to 'Kroger Precision Marketing offsite programmatic' for the 2025 passage and keep the FAQ as a network-level statement. | product field 'Kroger Precision Marketing onsite search and display' does not fit the 2025 passage (offsite programmatic). | RTL-I-F18 (evidence/verify/I.json) |
| C252 | The IAB/MRC Retail Media Measurement Guidelines (January 2024) place measurement with the retail media organisation and ask for third-party audit rather than organisational separation: a retail media measurement organisation 'is encouraged to conduct frequent (at least annual) data quality audits (internal audits as well as external audits performed by independent parties)', with 'Periodic audits by MRC or other independent third parties' (p. 14); 'If outcomes are reported for retail media organization activity and used for pricing or optimization considerations, they are encouraged to be subj | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | PDF re-downloaded and extracted by the verifier (485,528 bytes, 60 pages, created 2024-01-26; printed page numbers equal PDF page numbers). All eleven passages are verbatim at the stated pages: p. 14 ("at least annual" audits; "Periodic audits by MRC or other independent third parties ..."), p. 15 ("Known errors should be disclosed ..."), p. 16 ("Transparency to data users is a paramount goal ..."), p. 25 ("If outcomes are reported ... used for pricing or optimization considerations, they are encouraged to be subject to third-party audit and validation." and "... recommends retail media measur | supported with qualification | Restate the opening: "The IAB/MRC Retail Media Measurement Guidelines (January 2024) apply to 'retail media organizations or other parties measuring this activity' (p. 10), so they accept the network itself as the measuring party, and they rely on audit, validation and disclosure rather than on separating measurement from sales: ..." (the rest of the claim stands as written). | 'Place measurement with the retail media organisation' overstated the text: the guidelines are neutral about who measures, which is the point the paper needs. | RTL-I-F20 (evidence/verify/I.json) |
| C253 | IAB Europe's Commerce Media Measurement Standards V2 (body dated January 2026) assign responsibility to the seller side and route assurance through certification: 'The party providing these metrics to the brand is responsible for adhering to these standards, that can be the Commerce Media Network or an Ad Tech provider that has a direct relationship with the brand. The only way to guarantee adherence to these standards is through IAB Europe's Retail Media Certification.' (p. 4); a CMN 'will need to disclose' which delivery route it uses (CMN reporting, CMN APIs with brand analysis, or clean ro | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | PDF re-downloaded and extracted by the verifier (711,741 bytes, 14 pages, author 'Nicole Laleva', created 2026-05-19). Verbatim at the stated pages: p. 4 "The party providing these metrics to the brand is responsible for adhering to these standards, ... The only way to guarantee adherence to these standards is through IAB Europe’s Retail Media Certification." and the V1/V2 note; p. 11 "A CMN will need to disclose in which of the following platforms they provide data to brands and agencies."; p. 10 "A CMN doesn’t have to provide incrementality measurement to be certified, but if they do, they h | supported with qualification | Cite the document as the IAB Europe Commerce (incl. Retail) Measurement Standards file whose cover reads 'V2.1' and 'May 2026' while the body says 'V2' 'updated as of January 2026'; the passages quoted are unaffected, but the version label in the claim ('V2 (body dated January 2026)') should add 'cover: V2.1, May 2026'. | RTL-I-S26 version field 'V2 per body; cover not re-read' should read 'body V2 (January 2026); cover V2.1 (May 2026)'. | RTL-I-F21 (evidence/verify/I.json) |
| C254 | On 31 August 2026 the FTC and 22 states filed a second case, FTC et al. v. Amazon.com, Inc., Case 2:26-cv-03097 (W.D. Wash.), alleging under Section 5(a) of the FTC Act and state laws that 'Since 2019, Amazon.com, Inc. ("Amazon") has secretly and systematically overcharged its approximately 1.2 million advertising customers by manipulating the "auctions" that it uses to set the price of ads on its platform' for 'Sponsored Products, Sponsored Brands and Sponsored Display', via undisclosed 'soft reserve' prices described internally as a 'proxy 2nd price that we calculate'; the complaint alleges | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | Complaint (the verifier obtained the PDF through the fetch tool's saved binary and extracted it independently: 181 pages, created 2026-08-31; every passage below matches the stream's text): caption "COMPLAINT FOR PERMANENT INJUNCTION, MONETARY JUDGMENT, AND OTHER RELIEF"; footer "Case 2:26-cv-03097 Document 1 Filed 08/31/26 Page 1 of 181"; ¶1 (printed p. 1, PDF p. 5) "Since 2019, Amazon.com, Inc. (“Amazon”) has secretly and systematically overcharged its approximately 1.2 million advertising customers ...", "... three of its advertising products: Sponsored Products, Sponsored Brand | supported with qualification | Keep the claim with three edits: (1) 'in 2024 Sponsored Products advertisers were charged ...' should read 'the 2024 surcharge increase "also resulted in Amazon charging Sponsored Products advertisers their own winning bid close to 80% of the time" (¶76, read with ¶75)'; (2) locator for ¶76 is printed p. 24 (PDF p. 28), matching the printed-page convention used for ¶1; (3) limitations: replace 'the FTC case page for this matter ... was not opened' with the case page facts above (pending; last updated 31 August 2026; no post-filing event by 3 November 2026), and reconcile the dates with ¶11's n | product: the complaint covers Sponsored Products, Sponsored Brands and Sponsored Display; the 70%/50% figures concern 'Amazon Sponsored Brands' (a register row). Consider recording the Sponsored Brands figures under that row rather than under 'Amazon Sponsored Products'. | RTL-I-F24 (evidence/verify/I.json) |
| C255 | The Media Rating Council lists 'MRC Digital Advertising Auction Transparency Standards' as 'Last updated and issued Jan 2026' on its standards page, beside the 'Retail Media Measurement Guidelines, Issued Jan 2024'; the auction-transparency document (not opened here) is the accreditor's instrument closest to the auction-disclosure allegations in Case 2:26-cv-03097, and the MRC page carries no retail media accreditation category text. | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | MRC standards page (opened 2026-11-03): "Retail Media Measurement Guidelines" — "Issued: Jan 2024"; "MRC Digital Advertising Auction Transparency Standards" — "Last updated: Jan 2026" and "Issued: Jan 2026" (link MRC_Digital_Advertising_Auction_Transparency_Standards_FINAL_0.pdf); "Outcomes and Data Quality Standards" — "Issued: Sep 2022"; "Guidelines for Incremental Measurement in Commerce Media" not listed; accreditation categories listed are Television, Radio, Multimedia, Digital, Out-of-Home and Outside the U.S. (no retail media category). | supported with qualification | Quote the labels as two fields ('Last updated: Jan 2026'; 'Issued: Jan 2026') rather than the composite 'Last updated and issued Jan 2026'. Add that the same page lists MRC's 'Outcomes and Data Quality Standards' (Issued Sep 2022), the MRC instrument the retail media guidelines lean on for outcomes and the one most relevant to who audits outcome metrics. | RTL-I-F26 (evidence/verify/I.json) | |
| C256 | Source title: 'Complaint for Permanent Injunction, Monetary Judgment, Civil Penalty Judgment, and Other Relief, FTC et al. v. Amazon.com, Inc., Case 2:26-cv-03097 (W.D. Wash.)' | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | The caption on PDF page 1 reads "COMPLAINT FOR PERMANENT INJUNCTION, MONETARY JUDGMENT, AND OTHER RELIEF"; the FTC case page lists the file as "Complaint for Permanent Injunction, Monetary Judgment, and Other Relief". 'Civil Penalty Judgment' appears only in the PDF's metadata title field ('Complaint for Permanent Injunction, Monetary Judgment Civil Penalty Judgment, and Other Relief'). | contradicted | Title the source 'Complaint for Permanent Injunction, Monetary Judgment, and Other Relief' (as in the docket_events table, which is already correct). | The states do seek civil penalties under state law (¶13), so the metadata title is not wrong in substance, but it is not the document's title. | RTL-I-S15 (source title) (evidence/verify/I.json) |
| C257 | Source title and version: 'IAB Europe Commerce Media Measurement Standards V2 (body: updated as of January 2026; file created 19 May 2026)'; version 'V2 per body; cover not re-read' | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | Cover (image, rendered by the verifier): "May 2026"; "IAB EUROPE COMMERCE (INCL. RETAIL) MEASUREMENT STANDARDS V2.1". Body p. 3: "updated as of January 2026 ... now the IAB Europe Commerce Media Measurement Standards V2". | supported with qualification | Version: 'body V2 (January 2026); cover V2.1 (May 2026)'. The brief's landmine on the V2/V2.1 drift is confirmed on this file. | RTL-I-S26 (source version) (evidence/verify/I.json) | |
| C258 | S08 funding note 'DSI is a Salsify-affiliated institute per public knowledge'; S02 reliability note 'Full report behind DOWNLOAD NOW (/getfile/34806) with a login form' | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | DSI page (harvest): no ownership or affiliation statement captured. ANA 2023 report page (opened 2026-11-03): a "Log in" panel appears ("Your company may already be a member"); a DOWNLOAD NOW button links to the full report; the page "does not state that a login is required to download". | supported with qualification | Remove the unsourced Salsify affiliation or cite an opened source. Restate S02's gate as 'a login panel and a DOWNLOAD NOW link; whether the download requires login is not stated on the page'. | Unsourced assertion in a source record ('per public knowledge'). | RTL-I-S08 / RTL-I-S02 (source notes) (evidence/verify/I.json) |
| C259 | Survey rows for ANA 2023, ANA 2024, Microsoft/DSI and ANA 2025-2026 documents | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | All ANA 2023 and 2024 cells match the release text (see RTL-I-F01, F03, F04). Report page: "Fully 75 percent of respondents indicated that sales were the "most important" goal of RMNs, while only 18 percent of advertisers said brand awareness was most important." WARC: Eighty-five percent of brands surveyed "feel pressured by retailers to support RMNS." ANA 2025 report page: dated September 26, 2025; gated; no survey, sample or author named. | supported with qualification | (1) ANA 2023 objective row: replace 'paraphrased by the fetch tool' with the verbatim report-page sentence, and note that the 18 percent is 'brand awareness' (a different item from the release's 12 percent 'to invest for future brand growth'). (2) Add a row: ANA 2023 (WARC summary of the report, RTL-VI-S01), item 'feel pressured by retailers to support RMNS', base 'brands surveyed' (not reconciled with 138 or 80), result 85%, method 'online survey' (31 questions per WARC), claim_status actual, period 2022. (3) 'Retail Media Internal Management Guidance' row: frame 'not stated on the open page | data_table: surveys (evidence/verify/I.json) | |
| C260 | Twelve rows of trade-term and governance sources | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | Rows match their sources (see the finding checks), except: the IAB/MRC guidelines apply to 'retail media organizations or other parties measuring this activity' (p. 10); the Flywheel closing row rests on an issuer investor-site release (RTL-U-S200), not a filing. | supported with qualification | IAB/MRC row: replace 'measurement sits with the "retail media measurement organization"' with 'applies to "retail media organizations or other parties measuring this activity" (p. 10)'. Flywheel closing row: type 'issuer release' (not 'filing'). ANA 2023 row: add the 85% 'feel pressured' item from WARC's summary. IAB Europe row: add 'cover V2.1, May 2026'. Restate the Omnicom interest mark as in RTL-I-F12. | Flywheel closing row typed 'filing' for an investor-site release. | data_table: trade_table (evidence/verify/I.json) |
| C261 | Five holding-company rows | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | Publicis, dentsu and IPG/KINESSO rows match. Omnicom: Omnicom Media's own release of 28 August 2026 records Hearts United, created from Hearts & Science and Mediahub, joining OMD, Initiative, PHD and UM ('Like its sibling Omnicom Media agencies OMD, Initiative, PHD and UM, Hearts United will leverage ...'). WPP: the page carries retail media article and award titles and a '2025 Amazon Ads Partner Awards' entry. | supported with qualification | Omnicom row: replace 'Omnicom agency list not reached (omc.com/agencies 404)' with 'Hearts & Science and Mediahub combined into Hearts United, 28 August 2026 (Omnicom Media release, RTL-VI-S03); Omnicom agency list itself not reached'. WPP row: replace 'no tool, retailer partnership or standards text on the page' with 'retail media appears only in an article title and award titles (incl. a 2025 Amazon Ads Partner Award); no tool, partnership terms or standards text'. Restate the interest note as in RTL-I-F12. | data_table: holding_company_practices (evidence/verify/I.json) | |
| C262 | Five network rows on IAB/MRC wording and MRC list status | Refute-mode verifier for evidence/research/I.json re-opened the cited source. | Roundel and Walmart Connect rows re-confirmed on the live pages (2026-11-03). Kroger and Instacart rows match the session browser notes word for word. MRC list cells rest on U.json (RTL-U-S021, S022), not re-opened here. | supported with qualification | Kroger row: note that the 'adheres to IAB standards' passage concerns KPM's offsite programmatic incremental measurement (register row 'Kroger Precision Marketing offsite programmatic'). Instacart row: cite the archive capture URL rather than the live URL recorded as an access error. | data_table: standard_metric_adoption (evidence/verify/I.json) | |
| C263 | The FTC's staff perspective of 17 January 2025 ('FTC Surveillance Pricing 6(b) Study: Research Summaries — A Staff Perspective', January 2025) is an interim document released on a 3-2 vote, based on documents from six respondents (Mastercard, Revionics, Bloomreach, PROS, Accenture and McKinsey & Co.), containing only hypothetical examples and stating that it 'does not reflect or suggest any assessment as to whether anyone has engaged in illegal conduct'; the study was 'still ongoing' and a request for information took comments until 17 April 2025. | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | Release (17 January 2025): 'The Commission voted 3-2 to allow staff to issue the report.'; '...and therefore the staff perspective only includes hypothetical examples of surveillance pricing.'; 'The FTC's study of the 6(b) documents is still ongoing.'; 'Comments are due by April 17.'; dissent by 'Commissioners Andrew Ferguson and Melissa Holyoak'; the release calls the content 'initial findings' and 'an initial analysis of documents provided by Mastercard, Accenture, PROS, Bloomreach, Revionics and McKinsey & Co.' PDF p. 1: 'These research summaries highlight FTC staff's initial findings from | supported with qualification | The FTC's staff perspective of 17 January 2025 ('FTC Surveillance Pricing 6(b) Study: Research Summaries — A Staff Perspective', January 2025) presents staff's 'initial findings' and says it is 'not an exhaustive declaration of the findings'; it was released on a 3-2 vote (Commissioners Ferguson and Holyoak dissenting), rests on documents from six respondents (Mastercard, Revionics, Bloomreach, PROS, Accenture and McKinsey & Co.), 'only includes hypothetical examples of surveillance pricing' (release), and 'does not reflect or suggest any assessment as to whether anyone has engaged in illegal | 'Interim' is the brief's and the research file's word, not the source's; keep it only as a paraphrase of 'initial findings' / 'not an exhaustive declaration'. | RTL-J-F02 (evidence/verify/J.json) |
| C264 | The FTC's 6(b) orders defined the solutions studied by reference to 'purchase history' among other variables, and the staff document's hypothetical examples include retailer loyalty-program lists and customers profiled by perceived 'loyalty' or 'willingness to pay' (examples, not findings about any named company). | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | fn 10 (p. 2) quotes the 'initial order' definitions of 'Targeted Pricing Solution' and 'User Segmentation Solution', both 'based on location, purchase history, ... browsing behavior, demographic information, or any other variables' (verbatim; the order PDF itself carries the same definitions). fn 33: 'For example, a retailer can compile a list of email addresses from shoppers who have signed up for its loyalty program in-store and then encourage those customers to shop online by targeting them with promotional deals exclusive to its website.' fn 39: 'Documents have indicated that customers tar | supported with qualification | The FTC's 2024 6(b) order (Matter No. P246202) defined the solutions studied by reference to 'purchase history' among other variables. The staff document gives an illustrative example of a retailer targeting shoppers who joined its loyalty program in-store with web-only deals (fn 33), and separately reports, citing redacted respondent documents, that targeted customers 'can include those who are coupon-inclined, who have perceived customer "loyalty," or who demonstrate a "willingness to pay"' (fn 39); neither names a company or a retail media network. | Footnote 39 is coded as a 'hypothetical example'; it is staff's anonymised report of what respondent documents indicated (a document-based observation), not a hypothetical. | RTL-J-F03 (evidence/verify/J.json) |
| C265 | Washington's My Health My Data Act (RCW 19.373, in force from 31 March 2024; small businesses from 30 June 2024) defines consumer health data to include 'Use or purchase of prescribed medication' and data derived by inference, requires consent to collect it and a separate consent to share it, makes it unlawful to sell it without a signed valid authorization, bans geofences around in-person health-care providers for advertising, and makes violations actionable under the state Consumer Protection Act. | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | RCW 19.373.010(8)(b)(iv) 'Use or purchase of prescribed medication;'; (8)(b)(xiii) data 'derived or extrapolated from nonhealth information (such as proxy, derivative, inferred, or emergent data by any means, including algorithms or machine learning)'; (6)(b)(i) consent may not be obtained by 'A consumer's acceptance of a general or broad terms of use agreement or a similar document...'. RCW 19.373.030(1)(a): '(i) With consent from the consumer for such collection for a specified purpose; or (ii) To the extent necessary to provide a product or service that the consumer to whom such consumer he | supported with qualification | Washington's My Health My Data Act (RCW 19.373; the duties in RCW 19.373.020 to .070 apply from 31 March 2024, small businesses from 30 June 2024) defines consumer health data to include 'Use or purchase of prescribed medication' and health data inferred from non-health information 'by any means, including algorithms or machine learning'; permits collecting it only with consent for a specified purpose or to the extent necessary to provide a product or service the consumer requested, and sharing it only with a separate consent or to that same extent; makes selling it unlawful 'without first obt | Material omission: the 'necessary to provide a requested product or service' limb in RCW 19.373.030(1)(a)(ii) and (b)(ii). For a pharmacy or grocer, processing a prescription purchase to fill the order needs no consent; consent bites on secondary uses such as advertising. The 'in force from 31 March 2024' date should not be attached to the geofence ban, which carries no date in the text read. | RTL-J-F06 (evidence/verify/J.json) |
| C266 | Minnesota's Consumer Data Privacy Act (2024 ch. 121, art. 5; effective 31 July 2025) bars processing sensitive data without consent, bars targeted advertising or sale of personal data of consumers known to be 13 to 16 without consent, and bars small businesses from selling sensitive data without prior consent. | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | Minn. Stat. § 325M.16 'RESPONSIBILITIES OF CONTROLLERS', subd. 2(d): 'a controller may not process sensitive data concerning a consumer without obtaining the consumer's consent'; subd. 2(f): 'A controller may not process the personal data of a consumer for purposes of targeted advertising' [or sell it without consent where the controller knows the consumer is 13 to 16]; history '2024 c 121 art 5 s 8'. § 325M.17 'REQUIREMENTS FOR SMALL BUSINESSES' (a): small businesses 'must not sell a consumer's sensitive data without the consumer's prior consent'; history '2024 c 121 art 5 s 9'; note citing a | supported with qualification | Minnesota's Consumer Data Privacy Act (Laws 2024, ch. 121, art. 5, codified at Minn. Stat. §§ 325M.10 to 325M.21; effective 31 July 2025) bars processing sensitive data without consent (§ 325M.16, subd. 2(d)), bars targeted advertising or sale of personal data of consumers known to be 13 to 16 without consent (subd. 2(f)), and bars small businesses from selling sensitive data without prior consent (§ 325M.17(a)). | Locators cite '[325O.07]', the bill's coding; the revisor codified the act as chapter 325M, which is why the 325O statute URLs returned 404. | RTL-J-F09 (evidence/verify/J.json) |
| C267 | Maryland's Online Data Privacy Act (2024 ch. 455, SB 541; effective 1 October 2025 as enrolled) provides that a controller may not collect, process or share sensitive data (which includes 'consumer health data') except where strictly necessary to provide or maintain a product or service the consumer requested, may not sell sensitive data, must limit collection to what is 'reasonably necessary and proportionate', may not process personal data of a consumer known to be under 18 for targeted advertising, and had to honour opt-out preference signals by 1 October 2025; the enrolled text strikes the | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | § 14-4607(A): item '(1) COLLECT PERSONAL DATA FOR THE SOLE PURPOSE OF CONTENT PERSONALIZATION OR MARKETING WITHOUT THE CONSENT...' is struck; new '(1) EXCEPT WHERE THE COLLECTION OR PROCESSING IS STRICTLY NECESSARY TO PROVIDE OR MAINTAIN A SPECIFIC PRODUCT OR SERVICE REQUESTED BY THE CONSUMER TO WHOM THE PERSONAL DATA PERTAINS [struck: AND UNLESS THE CONTROLLER OBTAINS THE CONSUMER'S CONSENT], COLLECT, PROCESS, OR SHARE SENSITIVE DATA CONCERNING A CONSUMER;' (the drawn strike line on that row runs from x 54.0 to 481.5 and covers 'CONSENT'); '(2) SELL SENSITIVE DATA;'; '(4) PROCESS THE PERSONAL | supported with qualification | Maryland's Online Data Privacy Act (2024 ch. 455, SB 541; effective 1 October 2025) provides that a controller may not collect, process or share sensitive data (which includes consumer health data) except where strictly necessary to provide or maintain a specific product or service the consumer requested (the enrolled text strikes 'and unless the controller obtains the consumer's consent'), may not sell sensitive data, must limit collection to what is 'reasonably necessary and proportionate', may not process personal data for targeted advertising, or sell it, where it knew or should have known | The struck-clause reading is confirmed independently. Three fixes: the minors' standard is 'knew or should have known', not 'known'; the sale of under-18 personal data is barred outright (its consent clause is struck too); and Section 2's 1 April 2026 processing date is omitted. Its literal scope is § 14-4612 as enacted, which in this enrolled text is the section opening 'NOTHING IN THIS SUBTITLE | RTL-J-F10 (evidence/verify/J.json) |
| C268 | Walmart's Privacy Notice (updated 20 August 2026) lists 'Performing and providing advertising and marketing services, including targeted advertising' among its uses, says Walmart processes sensitive personal information with consent 'Where required by law', honours the Global Privacy Control to opt out of sale/sharing, states that disclosures to business partners 'may be deemed selling under certain state privacy laws', and does not mention Walmart Connect, retail media, clean rooms or a Washington consumer-health-data section; pharmacies are covered by separate notices. | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | 'Updated: August 20, 2026'; 'Performing and providing advertising and marketing services, including targeted advertising.'; 'Where required by law, Walmart only processes this Sensitive Personal Information with your consent.'; 'You may use the Global Privacy Control (GPC) to opt out of sale/sharing of your Personal Information'; 'We may disclose personal information to third parties or process your personal information for various purposes' ... 'These kinds of disclosures may be deemed selling under certain state privacy laws.'; 'Walmart adheres to Digital Advertising Alliance (DAA) self-regu | supported with qualification | Add: the notice states that VIZIO OS (television) data is kept separate from Walmart account data 'unless and until you opt in to link and combine it', under a 'Consent to Combine VIZIO OS Data with Your Walmart Account Data', with identifiers encrypted or hashed; and it discloses data to advertising-technology vendors including 'identity management companies' and to measurement vendors 'that provide aggregate reporting'. | The VIZIO opt-in is the only documented consent statement for combining a data source with Walmart shopper accounts in any notice opened, and it bears on the consent-basis and partner-matching cells for the Walmart Connect rows that use VIZIO inventory or data. It is still company-level, not specific to Walmart Connect. | RTL-J-F14 (evidence/verify/J.json) |
| C269 | Target's Privacy Policy (updated 15 August 2026) states that Target participates in interest-based advertising under the DAA principles, treats Global Privacy Control signals as opt-outs of sale and lists categories of personal information collected through Target Circle for targeted advertising; its California section says 'We do not collect or process sensitive personal information (as defined by California law)' ... 'for the purpose of inferring characteristics about individuals' (two tool fragments); its Health Privacy Policy (30 June 2025) says health data, including 'purchase of health-r | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | Privacy policy: 'Last update: August 15, 2026'; the California sentence is continuous, with no words between the two fragments: 'We do not collect or process sensitive personal information (as defined by California law) for the purpose of inferring characteristics about individuals.'; 'If required, consent will be obtained before collecting sensitive personal information'; 'However, Target treats Global Privacy Control signals as opt-out of sale and opt out of targeted/cross-context behavioral advertising requests under applicable state laws (e.g., California Consumer Privacy Act).'; heading ' | supported with qualification | Join the two California fragments into the single sentence 'We do not collect or process sensitive personal information (as defined by California law) for the purpose of inferring characteristics about individuals.', and state that Target treats GPC signals as opt-outs of sale and of targeted / cross-context behavioural advertising (not of sale only). | The Roundel measurement-page sentence (RTL-U-S028) was not re-opened. Whether 'purchase history' sits under the Target Circle targeted-advertising heading in Section 6 should be quoted before it enters the posture table. | RTL-J-F15 (evidence/verify/J.json) |
| C270 | Criteo's privacy policy (effective 17 July 2026) states that Criteo 'does not process sensitive personal information', prohibits advertisers and publishers from sharing special categories of data and advertising that targets children under the consent age, discloses that in the past 12 months it has 'disclosed or sold' categories 'for the commercial purpose of providing cross-context behavioral advertising' with a 'Do Not Sell or Share My Personal Information' link, and declares that Criteo Corp. is a registered data broker under Texas Business and Commerce Code chapter 509 and in California; | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | 'Effective Date July 17, 2026'; under 'Right to limit use of sensitive personal information': 'Criteo does not process sensitive personal information.'; 'Criteo has adopted policies and guidelines strictly prohibiting advertisers and publishers to share with us' ... 'special categories of data as defined under the GDPR'; children sentence as quoted; 'In the past 12 months, we have disclosed or sold these categories of personal information' ... 'for the commercial purpose of providing cross-context behavioral advertising'; 'UNDER CHAPTER 509 OF THE TEXAS BUSINESS AND COMMERCE CODE.'; 'CRITEO CO | supported with qualification | Replace the final clause with: the policy body does not mention retail media or commerce media, but it does describe processing purchase events from advertisers' sites and apps and events at brick-and-mortar shops, matching purchases with ad clicks and displays, and receiving loyalty or membership card numbers from advertisers (whom it describes as 'like brands and retailers'); it names no independent certification. | The absence clause 'says nothing about ... purchase data from retailers' is contradicted by the policy body; the rest is verified. The privacy_posture partner-matching cell for Criteo should cite these passages instead of 'Product-specific: unknown' alone. Note the policy's own statutory slip ('Business and Professions Code § 1798.99.80'; the Delete Act sits in the Civil Code). | RTL-J-F18 (evidence/verify/J.json) |
| C271 | This review found no FTC or state attorney-general order, settlement or court finding, and no breach notice, concerning a register network's use of shopper data for its retail media service (Amazon, Walmart, Instacart, Kroger, Target, Albertsons, Criteo, Chase, Cardlytics, Circana, Numerator) in the searches and pages logged on 3 November 2026; the official matters opened that touch a register company are inquiries (the 2024 6(b) orders, JPMorgan Chase; the 2026 6(b) orders, Instacart, Target, Walmart, DoorDash, Uber, Lyft and United), Criteo as a named recipient in GoodRx (2023), and Instacar | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | NY AG, 'January 8, 2026': 'New York Attorney General Letitia James today sent a letter to online grocery shopping platform Instacart' ... 'demanding more information about its use of algorithmic pricing and price-setting experiments after a recent report revealed Instacart users were being charged significantly different prices for the same products.'; requests cover 'The automated tools used by Instacart and its partners to adjust prices and discounts, including how consumer data may be used for these price changes.'; 'New York's Algorithmic Pricing Disclosure Act took effect on November 10, | supported with qualification | Keep the bounded absence, and widen the list of official matters touching register companies: add the New York Attorney General's letter to Instacart of 8 January 2026 (an inquiry under the Algorithmic Pricing Disclosure Act about price experiments and use of consumer data in pricing; no finding) and California's surveillance-pricing investigative sweep of 27 January 2026 (letters to unnamed retail, grocery and hotel businesses; inquiry). Describe the Instacart settlement as concerning deceptive 'free delivery' and '100% satisfaction guarantee' claims, undisclosed service fees and Instacart+ c | Neither new matter concerns shopper data used for advertising, so the dimension-12 absence still holds as bounded; both are pricing inquiries like the 6(b) orders, and the Instacart one is a named, official, state-level inquiry the file does not record. Neither page reports an outcome. | RTL-J-F23 (evidence/verify/J.json) |
| C272 | On 28 October 2026 the FTC (vote 2-0) issued Section 6(b) orders to nine companies, rendered by the fetch tool as Delta, United, JetBlue, Lyft, Uber, DoorDash, Instacart, Target and Walmart, to study whether pricing 'varies from one person to the next based on factors such as participation in loyalty programs or analysis of personal data' and personalised driver compensation; the template order sits under the same FTC Matter No. P246202 as the 2024 surveillance-pricing orders, covers 1 January 2023 onward, defines personal information to include 'purchases, leases, subscriptions, browsing hist | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | Release, 'October 28, 2026': 'The orders were issued to Delta, United, JetBlue, Lyft, Uber, DoorDash, Instacart, Target and Walmart under Section 6(b) of the FTC Act.' (quoted verbatim by the tool on the second pass); 'nine major airlines, ride-sharing companies and food-delivery firms'; 'Each company must turn over information that will allow the agency to assess whether the pricing for a ticket, ride share, food delivery or other product or service varies from one person to the next based on factors such as participation in loyalty programs or analysis of personal data by superintelligence a | supported with qualification | On 28 October 2026 the FTC (vote 2-0; Chairman Ferguson and Commissioner Meador) issued Section 6(b) orders to Delta, United, JetBlue, Lyft, Uber, DoorDash, Instacart, Target and Walmart (the release names no legal entities and groups them as 'airlines, ride-sharing companies and food-delivery firms') to study whether pricing 'varies from one person to the next based on factors such as participation in loyalty programs or analysis of personal data by superintelligence algorithms', and personalised driver compensation. The template order carries FTC Matter No. P246202, the number on the 2024 su | The loyalty quotation is cut mid-sentence: the source continues 'by superintelligence algorithms' ('Super Intelligence' is a defined term in the order). Restore the words or mark the cut with an ellipsis. The recipient list is verbatim, not only the tool's rendering. The matter-number match now rests on the 2024 order itself. The 'middlemen consultancies' sentence belongs beside the 'learned nothi | RTL-J-F24 (evidence/verify/J.json) |
| C273 | On 19 August 2026 the FTC (vote 2-0) published for comment a proposed enforcement policy statement on personalized pricing (comments due 18 September 2026, extended on 3 September to 25 September 2026) stating that businesses 'should clearly and conspicuously disclose not just that the price is personalized, but also the basis for that personalization and the types of data on which the personalization is based', that failing to do so is likely unfair or deceptive under Section 5, and that a shopper misled into believing a personalized price 'is a discount based on their purchase history with t | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | PDF p. 1: 'Federal Trade Commission's Proposed Enforcement Policy Statement Regarding Personalized Pricing August 19, 2026'; 'Congress has not given the Commission the authority to prohibit personalized pricing in all circumstances, but the Commission intends to enforce the law aggressively ...'; 'Where consumers reasonably expect that prices for a product or service will not vary based on their personal data, businesses that engage in personalized pricing should clearly and conspicuously disclose not just that the price is personalized, but also the basis for that personalization and the type | supported with qualification | Insert the statement's condition: 'that, where consumers reasonably expect prices not to vary with their personal data, businesses ... should clearly and conspicuously disclose ...'. The rest stands. | The 'no final version by 3 November 2026' absence now rests on two ftc.gov-restricted searches (the researcher's and this one); the legal-library page title still reads 'Proposed'. | RTL-J-F25 (evidence/verify/J.json) |
| C274 | Instacart's only FTC matter opened in this stream is a complaint filed 18 December 2025 and a stipulated order for permanent injunction and monetary judgment dated 14 January 2026, settling allegations that it charged consumers 'for Instacart+ without their express informed consent' among 'numerous unlawful tactics that harmed shoppers' ($60 million in refunds per the release title); the FTC case page (last updated 14 January 2026) shows the status 'Pending' and mentions no personal data, advertising or pricing tests. | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | Case page: 'Last Updated: January 14, 2026'; Case Status 'Pending'; timeline file 'Stipulated Order for Permanent Injunction, Monetary Judgment, and Other Relief' grouped under January 14, 2026, with href /system/files/ftc_gov/pdf/Instacart-StipulatedOrder_0.pdf. That PDF is the court-entered order: caption 'FEDERAL TRADE COMMISSION, Plaintiff, v. MAPLEBEAR INC., a corporation, d/b/a INSTACART, Defendant.' (N.D. Cal., San Francisco Division); every page headed 'Case 4:25-cv-10783-JST Document 15 Filed 01/13/26'; p. 12 'IT IS SO ORDERED. Dated: January 13, 2026', signed by 'JON S. TIGAR, UNITED | contradicted | Instacart's FTC matter: the FTC sued Maplebear Inc. d/b/a Instacart on 18 December 2025 (N.D. Cal., No. 4:25-cv-10783-JST) over false 'free delivery' and '100% satisfaction guarantee' claims, undisclosed mandatory service fees and charging for Instacart+ without express informed consent, filing a stipulated order the Commission approved 2-0; Judge Jon S. Tigar entered the order on 13 January 2026 (Document 15), with a US$60 million monetary judgment and a ten-year term. The FTC posted it on 14 January 2026, and its case page still shows 'Pending'. The matter concerns no personal data, ad targe | Event status is coded pending; the order was entered by the court on 13 January 2026, so the settlement is final (a court-entered stipulated order), and the date is 13 January, not 14 January (the FTC's posting date). | RTL-J-F28 (evidence/verify/J.json) |
| C275 | Virginia's Consumer Data Protection Act (§ 59.1-578(A)(5)) and Connecticut's Data Privacy Act (§ 42-520(a)(4)) both bar processing sensitive data without the consumer's consent; Connecticut's sensitive data includes 'mental or physical health condition or diagnosis' and 'consumer health data', and § 42-526 bars selling consumer health data without first obtaining consent; both limit collection to what is 'adequate, relevant' and reasonably necessary. | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | Virginia § 59.1-578(A)(1) and (A)(5) verbatim as quoted; (D) disclosure of sale and targeted advertising and of the opt-out; history '2021, Sp. Sess. I, cc. 35, 36; 2024, cc. 840, 844; 2026, c. 820.' Connecticut: the page the finding used is the revision to 1 January 2025 and points to the 2026 Supplement for 2025-session amendments. The 2026 Supplement shows P.A. 25-113 'substantially revised' § 42-520 effective July 1, 2026: (a)(1)(A) 'Limit the collection of personal data to what is reasonably necessary and proportionate' ...; (a)(1)(D) 'not process sensitive data concerning a consumer unle | contradicted | Virginia's Consumer Data Protection Act (§ 59.1-578(A)(5)) bars processing sensitive data without the consumer's consent and limits collection to what is 'adequate, relevant, and reasonably necessary' (§ 59.1-578(A)(1); amended by 2026 c. 820, content not read). Connecticut's Data Privacy Act, as amended by P.A. 25-113 effective 1 July 2026, bars processing sensitive data unless it is reasonably necessary for the disclosed purpose and the consumer has consented (§ 42-520(a)(1)(D)), bars selling sensitive data without consent ((a)(1)(H)), bars targeted advertising and sale for consumers known ( | The Connecticut half presents superseded text as current law at the 3 November 2026 access date (and at the 20 November cutoff). The source note's remark that 'No 2025 public act appears in the history note returned for § 42-520' comes from a revision that predates the 2025 session, so it is not evidence that the section was unamended. | RTL-J-F29 (evidence/verify/J.json) |
| C276 | Data table 'rules' (17 rows: legal and regulatory instruments; 'requires', 'effective', 'kind_key'). | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | California, Washington (with the exceptions noted), Texas OPA, Minnesota (substance), Maryland (substance), HIPAA, GoodRx, both FTC 6(b) rows and the policy-statement row check out against the sources opened. Oregon DOJ FAQ: the OCPA 'takes effect on July 1, 2024'; 'Consent is required to collect, store, or otherwise process all categories of "sensitive data", as defined in the law'; sensitive data includes 'mental or physical health conditions or diagnoses'; 'As of January 1, 2026, it is unlawful to sell a consumer's precise geolocation (past or present location) data.'; 'As of January 1, 202 | supported with qualification | Row edits: Connecticut row: instrument 'Gen. Stat. §§ 42-515(39), 42-520(a)(1), 42-526 as amended by P.A. 25-113'; requires 'sensitive data processed only where reasonably necessary for the disclosed purpose and with consent; no sale of sensitive data without consent; no targeted advertising or sale for consumers known (or wilfully disregarded) to be 13-17; collection limited to what is reasonably necessary and proportionate; no sale of consumer health data without consent'; effective '2023-07-01; amended 2026-07-01'. Washington row: add 'or to the extent necessary to provide a requested produ | The table omits Oregon although the stream's prompt listed it and the Oregon DOJ FAQ is fetchable; and it omits the two state pricing laws that touch register operators directly (New York's disclosure act; Maryland's 2026 grocery and delivery personalised-pricing law, unverified). | table:rules (evidence/verify/J.json) |
| C277 | Data table 'privacy_posture' (11 deep-profile rows; dimension-12 inputs). | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | Coding check (priority 6): every row codes a silent product record as 'unknown' or 'Product-specific: unknown'; no cell scores an absence as 0 or as a failure; adverse events are recorded as events, and 6(b) orders are marked inquiries. Cell-level issues found against the sources opened: Walmart (consent and partner-matching cells) omits the VIZIO opt-in combine consent, the hashing of identifiers in that context, and disclosures to 'identity management companies' and aggregate-reporting measurement vendors; Instacart adverse events give the stipulated order as 'dated 14 January 2026' (entered | supported with qualification | Apply the cell edits listed; keep every 'unknown' as is. | One inference unlabelled (Amazon data-broker applicability); no silent record was coded as absent. | table:privacy_posture (evidence/verify/J.json) |
| C278 | Data table 'adverse_events' (9 rows). | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | JPMorgan, staff perspective, RFI, GoodRx, 2026 6(b) and policy-statement rows verified. Instacart row: the stipulated order was entered by Judge Tigar on 13 January 2026 (Document 15, N.D. Cal. No. 4:25-cv-10783-JST); the FTC posted it on 14 January 2026; the case page's 'Pending' is stale. | supported with qualification | Instacart row: status 'final: stipulated order entered 2026-01-13 (10-year term; US$60 million judgment)'; summary per RTL-J-F28's correction. Add rows: (1) Instacart (Maplebear), New York Attorney General letter of 2026-01-08, inquiry under the Algorithmic Pricing Disclosure Act into price experiments and use of consumer data in pricing, outcome not public on the page opened; (2) California AG surveillance-pricing investigative sweep, 2026-01-27, letters to unnamed retail, grocery and hotel businesses, inquiry. Neither is a shopper-data-for-advertising event. | Instacart event status coded pending; it is final. | table:adverse_events (evidence/verify/J.json) |
| C279 | Data table 'data_broker_registry' (17 rows). | Refute-mode verifier for evidence/research/J.json re-opened the cited source. | All registry.csv IDs, names, DBAs, flags and request counts in the table match (Catalina 140,020 / opt-out 10; Bridg 343,121 / 343,150; LiveRamp government-ID flag 'Yes', 132 deletions with 72 complied and 60 denied, 238 opt-outs all complied; Affinity 297 / 124 with 79 complied; Epsilon reproductive-health and foreign-actor flags 'Yes', 4,985 / 87,942; PaeDae precise geolocation 'Yes', 82 / 9; Fetch precise geolocation 'Yes', zeros; Criteo all 'No', zeros; Circana gender-identity flag 'Yes'). registry2025.csv contains 'NC Vnetures, LLC' (misspelt), DBA 'NCS', website 'https://www.ncsolutions. | supported with qualification | NC Ventures row: registry 'California (registry2024.csv as NC Ventures, LLC; registry2025.csv as "NC Vnetures, LLC" [sic], DBA NCS; older AG list as NC Ventures, LLC d/b/a NCS)'; listed 'yes in the 2024 and 2025 files; not in registry.csv (merged into Circana, LLC on 30 September 2025 per Circana's registry comment)'. | The NC Ventures row's 'not in ... registry2025.csv by that name' is literally true but wrong in substance: the 2025 file lists it under a misspelling, which an exact-name screen misses. | table:data_broker_registry (evidence/verify/J.json) |
| C280 | Amazon's Sponsored Products prompts on the Amazon store (US) are explicitly generally available (open beta November 2025; GA in the US on 25 March 2026). | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Page dated 'March 10, 2026', no updated date. Subtitle: 'Sponsored Products prompts and Sponsored Brands prompts and reporting are now generally available'. Body: 'are officially moving from open beta to general availability in the U.S. on March 25, 2026'; 'were introduced in November 2025 in open beta'; 'This feature is available to U.S. Amazon advertisers' (excluding authors and publishers). No pilot or select-advertiser wording. | supported with qualification | Keep the state 'explicitly generally available' but rest it on the operator's present-tense subtitle ('Sponsored Products prompts and Sponsored Brands prompts and reporting are now generally available', 10 March 2026) together with the body's GA date of 25 March 2026. The finding's own limitation reads the GA date as 'a stated plan at publication' and still codes GA; on that reading alone the frozen rule (a planned date passing is not evidence it happened) would not allow the promotion. The subtitle is what makes the GA explicit in the operator's words. | H6a rests on R01 and R02 being in A; with the subtitle the coding holds. | RTL-K-F01 (evidence/verify/K.json) |
| C281 | Operator reporting for Sponsored Products prompts is documented (available, stage unstated); no separate stage word is attached to the report. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Subtitle: 'Sponsored Products prompts and Sponsored Brands prompts and reporting are now generally available'. Body: 'Prompts reports are also available, which provide performance metrics at the individual prompt level'; 'The Prompts report includes prompt text, associated ad, impressions, clicks, click-through rate, cost per click' ... 'spend, sales, ACOS, ROAS, and 7-day orders and units'; 'Sellers and vendors can review and manage prompts directly in the Ads Console or via API.' | supported with qualification | Operator reporting for Sponsored Products prompts is generally available in the operator's words ('prompts and reporting are now generally available', 10 March 2026), with a Prompts report listing impressions, clicks, CTR, CPC, spend, sales, ACOS, ROAS and 7-day orders and units, in the Ads Console and via API. Delete 'no separate stage word is attached to the report': the subtitle attaches 'generally available' to reporting. | State miscoded: 'available with stage unstated' should be 'generally available' on the operator-reporting axis for R01. H6a's verdict is unchanged (either state satisfies the rule). | RTL-K-F02 (evidence/verify/K.json) |
| C282 | Amazon's Sponsored Brands prompts on the Amazon store (US) are explicitly generally available. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Same page and passages as RTL-K-F01; the subtitle names both formats: 'Sponsored Products prompts and Sponsored Brands prompts and reporting are now generally available'. | supported with qualification | As for RTL-K-F01: rest the GA state on the operator's subtitle plus the 25 March 2026 GA date, not on the planned date alone. | RTL-K-F06 (evidence/verify/K.json) | |
| C283 | Operator reporting for Sponsored Brands prompts is documented (available, stage unstated); no separate stage word is attached to the report. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Subtitle: '... prompts and reporting are now generally available'; Prompts report metrics as listed under RTL-K-F02. | supported with qualification | Operator reporting for Sponsored Brands prompts is generally available in the operator's words (10 March 2026); delete 'no separate stage word is attached to the report'. | State miscoded on R02 operator reporting: should be 'generally available'. H6a verdict unchanged. | RTL-K-F07 (evidence/verify/K.json) |
| C284 | Sponsored Products and Sponsored Brands prompts on Alexa for Shopping are available with commercial stage unstated (11 June 2026: 'available today', no beta or GA wording). | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Dated 'June 11, 2026', no author. 'Last month, Amazon brought together two powerful AI experiences' (Rufus and Alexa+ into Alexa for Shopping); 'Today, advertisers can be part of the conversation through' ... 'Sponsored Products and Sponsored Brands prompts on Alexa for Shopping.'; 'All these agentic and conversational experiences are available today'; 'It's not a future state; it's a current state.'; 'Prompts can appear on an Amazon product detail page. When clicked, they open a conversation with Alexa for Shopping.' None of 'beta', 'generally', 'pilot', 'test', 'rolling out' appears. | supported with qualification | Claim supported. Keep R03 to the frozen format (Sponsored Products and Sponsored Brands prompts on Alexa for Shopping); the agent_surfaces label 'plus sponsored ads' pools a second format ('Sponsored ads in Alexa for Shopping') into R03 and should be removed. | Record granularity: the table's R03 capability label adds 'sponsored ads', which the frozen population does not include in R03. | RTL-K-F11 (evidence/verify/K.json) |
| C285 | ChatGPT's standard ads at the bottom of answers (US, Free and Go) are available with commercial stage unstated: announced 16 January 2026; test started 9 February 2026; later called a 'launch'; no post says 'generally available'. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | All six quoted passages are verbatim in the browser notes. The 5 May 2026 post also says 'In the first phase of the pilot, advertisers could buy ChatGPT ads on a CPM (cost-per-mille impressions) basis' (pilot wording in May); later posts use 'after launch' (31 August), 'Since we launched ads' and 'extending availability to 89 countries worldwide' (21 October). | supported with qualification | Add the trail: the operator called the programme a pilot as late as 5 May 2026 ('In the first phase of the pilot ...'); the 'available, stage unstated' code rests on the later 'launch' and 'available' wording (31 August and 21 October 2026), not on the test start of 9 February. The state stays 'available with commercial stage unstated'. | OpenAI's developer page 'Product Feeds' (developers.openai.com/ads/product-feeds: 'Use product feeds to create ads from your catalog'; 'Use product-segmented Insights to see which products received impressions and clicks') documents a catalog-driven ad use at format level; it would strengthen the record's shopping-use caveat (see omissions). | RTL-K-F16 (evidence/verify/K.json) |
| C286 | Third-party attribution for ChatGPT ads is available with stage unstated (platform-level statement of 5 October 2026 applied to the standard format). | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | 'We now support leading attribution partners across the web and app space, enabling Conversions API, reporting, and click attribution solutions with companies including AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge, and Tenjin.' No stage word. (Verbatim in the notes.) | supported with qualification | Supported for R04. State the rule used for platform-level statements once: they are applied to R04 only, while R06 (an ad-click entry into Sponsored Agents) and R07 (which Amazon's page calls 'ChatGPT Ads') stay unknown. The choice changes no verdict (unknown and stage-unstated both count in U for H6b; applying the 5 October pilot statements to R06 and R07 would cut H6c and H6d's U from 10 to 8, ratio 0.73, still unresolved). | RTL-K-F18 (evidence/verify/K.json) | |
| C287 | Causal outcome measurement for ChatGPT ads is at pilot or test stage (incrementality 'still in its early stages'; geo experiments with Haus, Measured and WorkMagic). | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | 'While our work on incrementality is still in its early stages, we're partnering with Haus, Measured, and WorkMagic to explore geo-based experiments to help advertisers understand the causal impact of ChatGPT Ads.'; 'WorkMagic reported statistically significant lift for wellness brand Dose ...' | supported with qualification | Keep 'pilot or test', with the wording that the operator says it is partnering 'to explore' geo-based experiments; the one started test evidenced on the page is WorkMagic's partner-reported lift for Dose (no design, sample or period). | RTL-K-F19 (evidence/verify/K.json) | |
| C288 | Ad-quality verification by a named third party for ChatGPT ads is at pilot stage (DV and IAS brand suitability evaluation pilots). | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | 'We're also developing brand suitability evaluation pilots with DoubleVerify (DV) and Integral Ad Science (IAS)'; 'This approach allows independent partners to assess how our brand safety standards are applied in a controlled testing environment, without accessing private user conversations.' | supported with qualification | Add: the operator says it is 'developing' the pilots, so whether they had started by 5 October 2026 is not stated; the partner axis has no 'announced' state, and 'pilot or test' is the nearest code. No verdict effect (R04 is outside G and U on this axis either way). | DoubleVerify's release of 5 October 2026 (named in the brief) was not opened by the stream or the verifier. | RTL-K-F20 (evidence/verify/K.json) |
| C289 | Amazon-sold ads in ChatGPT are a pilot actually started, on Amazon's primary page of 10 September 2026; the record enters N. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Dated 'September 10, 2026'. 'Amazon Ads has announced a partnership with OpenAI to support advertising in ChatGPT.'; 'The new partnership enables advertisers to expand their Amazon Ads campaigns with a conversational ad experience on ChatGPT'; 'Select advertisers in the U.S. are currently testing ChatGPT Ads as part of a pilot.'; 'With access to ChatGPT Ads, advertisers can now extend their campaigns further'; '... as they explore options, compare choices, and make decisions on ChatGPT'. Buying route: 'Amazon Ads offers programmatic technology solutions'; 'trillions of proprietary browsing, sh | supported with qualification | Amazon Ads' page of 10 September 2026 announces a partnership with OpenAI and says 'Select advertisers in the U.S. are currently testing ChatGPT Ads as part of a pilot' (pilot actually started, US). The page does not name the Amazon DSP as the buying route or mention first-party data in its body, and its shopping-use wording ('explore options, compare choices, and make decisions') is platform-level, with a travel advertiser (Delta Vacations) as the only named participant; include R07 in N with the same platform-level caveat as R04. | The frozen R07 definition ('Amazon DSP-bought ads in ChatGPT using Amazon first-party data') is only partly matched by Amazon's own wording; the frozen entry rule (an Amazon or OpenAI primary source opened) is met. | RTL-K-F31 (evidence/verify/K.json) |
| C290 | Ads in Google's AI Overviews and AI Mode are available with commercial stage unstated (US); the 28 October 2026 US copy with the CTR footnote was not found. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | 3 October 2024 (Shashi Thakur): 'After several months of careful testing, we're excited to announce that Ads in AI Overviews are now available for mobile users in the US.' (heading 'Ads in AI Overviews: now live'). 21 May 2025: 'Starting today, Search and Shopping ads in AI Overviews will be expanding to desktop in the U.S.'; 'We're also starting to test ads in AI Mode ...'. Help page: 'Ads within AI Overviews and AI Mode are available in English in the United States and select English-speaking markets'; 'Text ads and Shopping ads can appear within AI Overviews and AI Mode.' US copy of 28 Octo | supported with qualification | Claim supported on the availability state. Replace the help-page quote with the full sentence ('... available in English in the United States and select English-speaking markets'), quote the multiturn sentence from the US copy ('in AI Mode', not 'on AI Mode' as on the Africa copy), and drop the statement that the US copy was not found: it is at https://blog.google/products/ads-commerce/ads-in-ai-mode-global/ and carries footnote 2, 'Google Internal Data, US, August 20, 2026- September 3, 2026'. | The needs_browser item for the US copy is resolved by the verifier. | RTL-K-F36 (evidence/verify/K.json) |
| C291 | Causal outcome measurement for ads in AI Overviews and AI Mode is unknown; the CTR statement has no sample, design or footnote on the copy read. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | US copy: 'In fact, click-through rates are already similar between sponsored and organic product list items within AI Mode.' with footnote 2: 'Google Internal Data, US, August 20, 2026- September 3, 2026'. No incrementality on the help page or the post. | supported with qualification | Causal outcome measurement is unknown. The 28 October 2026 CTR statement is an internal engagement comparison footnoted on the US copy to 'Google Internal Data, US, August 20, 2026- September 3, 2026', with no sample or design, and is not outcome reporting. | Matches the brief's landmine (scoped vendor assertion, US internal data, 20 August to 3 September 2026). | RTL-K-F39 (evidence/verify/K.json) |
| C292 | Microsoft documents shopping ads inside Copilot with commercial stage unstated (6 August 2025 quick-back footnote; 21 April 2026 Offer Highlights 'now available' alongside 'live pilots'). | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | 6 August 2025: 'Copilot's quick back rates are 59% lower for multimedia ads' ... '52% lower for responsive search ads, and 19% lower for shopping ads.' (Copilot ads versus search ads); footnote [10] defines quick back rate and carries 'Microsoft First Party Data', 'O&O' and 'Global, Feb 2025' (the verifier confirmed the footnote number and strings); the post uses no stage word and names no Copilot ad format beyond 'intuitive ad formats'. 21 April 2026 (Tim Frank): 'Offer Highlights are now available for retail use cases in English-speaking markets'; '... you'll start seeing them in ads that ap | supported with qualification | Microsoft's posts imply shopping ads serve in Copilot (a quick-back comparison for 'shopping ads' on global first-party data, February to May 2025) and describe Offer Highlights in ads on product detail pages in Copilot as 'now available for retail use cases in English-speaking markets' while also inviting advertisers to 'live pilots' for Offer Highlights; no Microsoft page read names shopping ads in Copilot as a product with a stage or documents US availability for it. Code availability 'available with commercial stage unstated' at low confidence, US availability inferred, and keep the verdic | Inference: the format's existence and US availability are read from a performance footnote and a feature post, not from a product page; mark as inference within the record. | RTL-K-F51 (evidence/verify/K.json) |
| C293 | ACP's 24 March 2026 'discovery' expansion is not a dated ACP release; the feed surface entered the spec in Version 2026-04-17 alongside a well-known discovery document. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Changelog 2026-04-17 line 43: '**Added** – an unreleased Feed API surface for feed metadata and product catalog management.'; line 65: '- promotions are intentionally deferred to a stacked follow-up PR'; line 198: '**Added** -- discovery document at `/.well-known/acp.json` for pre-session capability checks.' No 2026-03-24 changelog file. | supported with qualification | ACP's 24 March 2026 'discovery' expansion is not a dated ACP release: the changelog has no 2026-03-24 entry, and the 2026-04-17 version adds what its changelog calls 'an unreleased Feed API surface for feed metadata and product catalog management' (promotions deferred) plus a discovery document at /.well-known/acp.json. Do not say the feed 'entered the spec' without quoting the changelog's own 'unreleased' label. | RTL-K-F59 (evidence/verify/K.json) | |
| C294 | ACP's own task (checkout) has operator-documented production use and operator-asserted transactions, no independent volume. Sibling PRO-C067 confirmed for ACP. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Quotes verbatim in the notes. The 24 March 2026 post says 'the initial version of Instant Checkout did not offer the level of flexibility that we aspire to provide, so we're allowing merchants to use their own checkout experiences while we focus our efforts on product discovery.' | supported with qualification | Add the date range: ACP checkout production use is operator-documented from 29 September 2025 (Etsy sellers) and 8 December 2025 (Instacart app); on 24 March 2026 OpenAI said merchants may use their own checkout while it focuses on discovery, and no page read states which ACP checkout integrations remain live at the freeze. | The protocol_states ACP production_use cell needs the same date range. | RTL-K-F61 (evidence/verify/K.json) |
| C295 | UCP is pinned at release v2026-08-25 (earlier tags as listed); README task; Google's announcement words. Sibling protocols record (UCP v2026-08-25, commit f2206e68) 'rechecked on the release tag: confirmed'. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Releases: v2026-08-25 published 2026-08-25T13:53:03Z (releases/latest); v2026-04-08 2026-04-09T01:48:20Z; v2026-01-23 2026-01-23T22:48:51Z; v2026-01-11 2026-01-23T17:58:21Z. The v2026-08-25 tag is an annotated tag pointing to commit 87dd4b1fcff9105d68173120897237fa784c74ef (tagger date 2026-08-25T13:47:10Z). Commit f2206e680f64fe2cc73495acc3557569afd9097a is a main-branch commit of 2026-10-07T15:42:49Z ('docs(permalink): require explicit buyer action before cart merge (#789...'). The shopping paper's protocols.json pairs 'version: v2026-08-25' with 'commit: f2206e680f64...'; its claim ledger c | supported with qualification | The UCP release facts are supported. Replace the sibling-reuse sentence with: 'the shopping paper's protocols record (version v2026-08-25) is confirmed as the latest release; its commit f2206e68 is a main-branch commit of 7 October 2026, not the v2026-08-25 release commit (87dd4b1f, 25 August 2026), so the sibling record pairs a release version with a later main commit and should name which one it pins'. PRO-C064's qualification (checkout 'will soon power', not live at announcement) is supported by Google's post. | Sibling reuse mis-stated: 'rechecked on the release tag: confirmed' is not true of the commit. | RTL-K-F62 (evidence/verify/K.json) |
| C296 | UCP v2026-08-25: schema representation partial (an ad attribution context field); no release mentions 'sponsored', 'ads', 'advertising', 'retail media', 'promotion' or 'measurement'; README carries none of the listed terms; checkout 'soon'; Microsoft UCP-ready feeds GA in the US. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | v2026-08-25 body: '**📣 Attribution ([#391](...))**: Adds support for platform-driven referral and ad attribution context.' The same body also contains 'Profile Keys Promotion' (a heading), 'Refactored token binding to be vertical agnostic and reusable across Retail, Food, Lodging and future verticals' and 'Replaced `retail_location.json` with `common/types/location.json`'. 20 May 2026: 'Shoppers can try these select checkout features soon across retailers like Nike, Sephora, Target' ... 'Ulta Beauty, Walmart, Wayfair'; 'UCP-powered checkout will roll out across Canada and Australia in the comi | supported with qualification | Keep the states. Correct the term screen: the v2026-08-25 release body uses 'Retail' as a vertical name ('reusable across Retail, Food, Lodging and future verticals'), names a since-replaced 'retail_location.json' type and has a 'Profile Keys Promotion' heading; none of these is advertising or retail media vocabulary. Write 'no release uses sponsored, ads, advertising, retail media or measurement vocabulary' rather than claiming the words are absent. | Sibling PRO-C065 ('Universal Cart ... live with retailers'): Google's 20 May post introduces 'Our new Universal Cart' but says the named retailers' checkout features come 'soon'; the stream's 'confirmed with qualification' is right. | RTL-K-F63 (evidence/verify/K.json) |
| C297 | AdCP's schema represents a sponsored placement inside a shopping surface (schema representation yes); the commerce-media worked example is a fictional retailer's 'Sponsored product listings' product. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Enum and canonical files are byte-identical at tag v3.2.8 and on main. Enum description: '... This is seller-declared discovery metadata, not independent verification of inventory quality or delivery context.'; enumDescriptions as quoted. Canonical title 'Canonical Format: Sponsored Placement (retail-media catalog-driven)'; 'Marked experimental at 3.1 GA: the canonical covers 4 meaningfully different retail-media adapter contracts (Amazon SP, Criteo SP / CitrusAd SP, Pinterest Collection, generative-per-SKU). ... Promotion to non-experimental gated on the #4592 adapter-contract docs work.' Com | supported with qualification | State the schema files as read at release tag v3.2.8 (identical on main). Correct the example: the worked product is `shopgrid_sponsored_products`, named 'Sponsored Products - Search & Browse' (section heading 'Sponsored product listings'), for a fictional retailer; no Amazon-named example exists on the page. | Interest mark present. Representation only, per shared rule 12. | RTL-K-F66 (evidence/verify/K.json) |
| C298 | AdCP's sales-retail-media specialism is at status preview (version 0.1.0) on main; its storyboard is a fixture (test kit), not an interoperability test between named third parties. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | index.yaml identical at v3.2.8 and main: 'version: "0.1.0"', 'status: preview', the quoted summary and the Summit Foods / Pinnacle Agency / ShopGrid fixture narrative. specialism.json: the 'sales-retail-media' PREVIEW description is identical at v3.2.8 and main (main adds an unrelated 'sales-fixed-rate' entry and a longer broadcast TV description). | supported with qualification | Pin the specialism to release tag v3.2.8 (identical content), rather than 'main, not a release tag'. | Interest mark present. No author-built agent credited: correct under rule 12. | RTL-K-F68 (evidence/verify/K.json) |
| C299 | AdCP's own documentation states no retail media network had shipped the evidence to promote the canonical; the readiness matrix keeps every contract family at 'Keep experimental until evidence lands'; no connector, production use or transaction found. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Line 9: 'The canonical stays `experimental: true` past 3.1 GA until at least two retail-media networks ship `format_schema` evidence ...'. Matrix: 'Retail-media deterministic, buyer-uploaded | Ready | Keep experimental until evidence lands | Amazon `format_schema` evidence'; network-composed and collection-layout rows 'Keep experimental until evidence lands'; 'Generative-per-SKU | Vendor-specific only | Keep experimental | Generative-quality evidence; may not promote without per-vendor framing'. Line 71: '**Not a promotion gate.** ... gated on `format_schema` evidence across at least two adopt | supported with qualification | Three of the four contract families read 'Keep experimental until evidence lands'; the generative-per-SKU family reads 'Vendor-specific only' and 'Keep experimental'. The page states a promotion gate; it does not itself say that no network has shipped the evidence, so write 'the canonical remains experimental at v3.2.8 under a gate of format_schema evidence from at least two retail media networks'. | Interest mark present. The none-found states for connector, production use and transaction stay bounded to the pages listed; no state above schema representation is credited: correct. | RTL-K-F69 (evidence/verify/K.json) |
| C300 | Scoped use case for AAMP 3.0: schema representation none specific; the press release and README contain none of 'retail', 'commerce', 'sponsored', 'shopping' or 'attribution'; outcome reporting delegated; other states none found. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Press release: none of the five words. README: 'retail', 'commerce', 'sponsored' absent; 'Attribution' appears only in the licence name ('Creative Commons Attribution 3.0 License'); line 85: 'lands in the one its `transaction_mechanism` maps to (OpenDirect 2.1, Deals API, ...'. | supported with qualification | Add 'except in the Creative Commons Attribution licence name' to the README term screen. | The GitHub code-search counts (0 retail, 0 commerce in OpenProposal) were not re-run by the verifier. | RTL-K-F71 (evidence/verify/K.json) |
| C301 | Fifty-five rows (R01 to R11 x five axes) with frozen states and figure-ready status. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Availability states match the operator's words for all eleven records: R01, R02 explicitly GA (subtitle 'now generally available'); R03, R04, R08, R10 available with stage unstated; R11 available with stage unstated by inference; R06, R07, R09 pilot or test actually started; R05 announced (no start evidenced). | supported with qualification | (1) R01 and R02 operator-reporting rows: status_stated 'generally available' (Amazon subtitle 'prompts and reporting are now generally available'). (2) R01 and R02 third-party-attribution rows: the date 2026-03-10 is paired with source RTL-K-S15 (the 11 June 2026 page); cite both RTL-U-S312 and RTL-K-S15 or date the row 2026-06-11. (3) R03 capability label: remove '(plus sponsored ads)' to keep the frozen record. (4) R08 operator-reporting row: the date 2026-11-03 is the access date of an undated help page; mark it 'undated (accessed 2026-11-03)'. (5) R11 availability row: add 'inferred from a | No announced test was promoted on a passed planned month: R05 stays announced. No GA was given without the operator's words once the Amazon subtitle is cited. | TABLE:agent_surfaces (evidence/verify/K.json) |
| C302 | Five states per protocol for the scoped use case, per pinned release; AdCP rows carry the interest mark. | Refute-mode verifier for evidence/research/K.json re-opened the cited source. | Every pinned release and date was confirmed. No protocol is credited above schema representation for the scoped use case: ACP none (checkout task credited separately as operator-documented), UCP partial (ad attribution context field), AP2 n/a, AdCP schema representation only (connector, interoperability test, production use and transaction none found; author test kit not credited), AAMP 3.0 and 2.3 none specific. The AdCP row carries the interest mark. | supported with qualification | (1) UCP retail_vocabulary: replace "none in the README or release bodies ('retail', 'sponsored', 'ads' absent)" with 'no advertising or retail media vocabulary; the v2026-08-25 release body uses Retail only as a vertical name and a since-replaced retail_location type'. (2) UCP protocol label 'Google and Shopify': Google's post says UCP was 'co-developed with industry leaders including Shopify, Etsy, Wayfair, Target and Walmart'; label it 'announced by Google; co-developed with Shopify, Etsy, Wayfair, Target and Walmart' or cite the governance source for 'Google and Shopify'. (3) ACP production | Rule 12 respected throughout: no state above schema representation was credited without a documented connector, test, production use or transaction. | TABLE:protocol_states (evidence/verify/K.json) |
| C303 | On 3 November 2026, MRC's services-under-review list for digital held 14 entries; five concern retail or commerce media: "DoorDash Ads (Pre-audit reviewed 1/2026, Audit not begun)" (impressions, clicks and CTR), "Kevel (Pre-audit reviewed 9/2026)" (including "Clicks, Conversions and incrementality"), "Kroger (Pre-audit in process as of 6/2026)" (including "Attributed sales, Incremental Sales, aROAS, iROAS"), "Schwarz Media Onsite Advertising (Audit reviewed 9/2026)" (Lidl and Kaufland apps) and "Walmart Connect Onsite Display (Audit in process as of 11/2025)"; a listing under review is a statu | Refute-mode verifier for evidence/research/X.json re-opened the cited source. | 14 entries: Adelaide; DoorDash Ads; Escalated; Google YouTube Relay Log; IAS YouTube Brand Safety; Kevel; Kroger; Meta Third-Party Brand Suitability Pipeline; Ogury; Partnerize; Pixalate; Schwarz Media Onsite Advertising; Walmart Connect Onsite Display; XPLN. The five quoted status strings and the DoorDash, Kevel and Kroger passages are verbatim. Schwarz in full: 'Measurement and reporting of Display Impressions, Clicks and Viewability, shoppers reached, units purchased' / 'Applies to Display banner ads served in carousel ad format for Lidl & Kaufland apps'. Walmart: 'Display Impressions, View | supported with qualification | On 3 November 2026, MRC's services-under-review list for digital held 14 entries; five concern retail or commerce media: "DoorDash Ads (Pre-audit reviewed 1/2026, Audit not begun)" (impressions, clicks and CTR for Sponsored Products and Sponsored Brands), "Kevel (Pre-audit reviewed 9/2026)" (including "Clicks, Conversions and incrementality"; the entry names no retail environment, and Kevel is counted as retail media by the paper's register, not by MRC), "Kroger (Pre-audit in process as of 6/2026)" (including "Attributed sales, Incremental Sales, aROAS, iROAS"), "Schwarz Media Onsite Advertisi | scope_geo 'US' fits the accreditor, not every entry: the Schwarz entry is for Lidl and Kaufland apps. | RTL-X-F02 (evidence/verify/X.json) |
| C304 | Amazon documents that advertisers using sponsored ads can run their own SQL queries in Amazon Marketing Cloud (a clean room built on AWS Clean Rooms) over event-level Sponsored Products impression and click records paired with attributed conversion events, and receive only aggregated, anonymous outputs; access for sponsored-ads advertisers is self-service. | Refute-mode verifier for evidence/research/X.json re-opened the cited source. | Product page: 'Built on AWS Clean Rooms'; 'Amazon Marketing Cloud is available at no cost to eligible advertisers via web-based UI and API. ... those using sponsored ads can access AMC via self-service.'; 'You can only access aggregated, anonymous outputs from AMC.'; 'Through AMC, advertisers can query event-level and user-level signals across sources using custom metrics and attributes of their choice, with built-in aggregation thresholds ensuring user privacy.' Overview (rendered): the 'These tables both contain pairs of traffic and conversion events ... Each conversion event in the table is | supported with qualification | Keep the claim, adding 'eligible': "Amazon documents that advertisers using sponsored ads can access Amazon Marketing Cloud (a clean room built on AWS Clean Rooms) by self-service, at no cost to eligible advertisers, and run their own SQL there over event-level Sponsored Products impression and click records, joinable to conversion events that Amazon has attributed to those traffic events (14-day window), receiving only aggregated, anonymous outputs." Replace the limitation sentence "Aggregation thresholds are labelled LOW or NONE per column with no numeric values on the pages read" with: "The | measurement_record.identity_linkage 'deterministic account (pseudonymised)' is too narrow: per the sponsored_ads_traffic page, user_id may be an Amazon account ID, a device ID or a browser ID (user_id_type adUserId, sisDeviceId, adBrowserId) and is NULL where unresolved; code it as pseudonymous mixed identifiers (account, device, browser). Evidence class vendor_assertion is right (vendor documenta | RTL-X-F03 (evidence/verify/X.json) |
| C305 | Instacart documented a sales-lift test for Sponsored Products using randomised test and holdout groups as a pilot in September 2022, and in May 2023 said lift testing "may not be available to all partners" and is arranged through the account team; the brand results it publishes (Reckitt 16.5% in Q3-Q4 2022; Kellogg "double-digit") are reported by Instacart. | Refute-mode verifier for evidence/research/X.json re-opened the cited source. | 2022 capture: 'Last Updated: Sep 2, 2022'; 'Kellogg Company sees double-digit sales lift from Instacart Ads, as one of the first to pilot sales lift & A/B testing capabilities'; the 'Brand partners working with Instacart Ads now have access to a pilot testing experience ... randomized test and holdout groups to determine incremental sales impact.' passage is verbatim; Kellogg's 'randomized experiment designed to gain insight into the incremental sales lift of its Instacart Ads sponsored product campaigns. The latest study, completed in August this year, revealed a double-digit total sales lift | supported with qualification | Replace "and is arranged through the account team" with "and told partners to 'consult with your account team'"; the page does not say the test is arranged or requested through the account team. Date the 2022 page as "last updated 2 September 2022" (its only date stamp) rather than as a September 2022 publication. | measurement_record.assignment_unit ('user (user-advertiser pair per the 30 June 2023 methodology post)') and the exposure_capture and identity_linkage fields are not stated on either page cited here; the two pages say only 'randomized test and holdout groups' and 'two randomly selected groups'. Cite the 30 June 2023 engineering post as a source of this finding or mark those fields not stated on th | RTL-X-F06 (evidence/verify/X.json) |
| C306 | Instacart states that its advertisers "on average, see more than a 15% incremental sales lift" from internal tests in 2023, and presents two Circana shopper analyses (August 2023; February 2024) finding that 75% and 90% of the studied brands' Instacart buyers did not buy the same brands in-store in the analysis period; the Circana items are buyer-overlap analyses, not lift tests. | Refute-mode verifier for evidence/research/X.json re-opened the cited source. | Dated 'May 2, 2024'. 'Instacart Ads drive high incrementality for our advertisers, who on average, see more than a 15% incremental sales lift – and sometimes twice that.*' Footnote in full: '*Based on internal tests run across all brand partners using our Sponsored Product ads offering in the year ending December 31, 2023 and individual tests run for select brands or types of brands. There can be no assurance that the outcomes for these tests can be maintained or replicated due to a variety of factors, some of which may not be within Instacart's control or cannot be anticipated.' 'Circana and | supported with qualification | "Instacart states that its advertisers 'on average, see more than a 15% incremental sales lift', based on internal Sponsored Products tests in the year to 31 December 2023 and on individual tests for select brands (undated), and presents two Circana shopper analyses (August 2023; February 2024), each of one CPG company's brands, finding that 75% and 90% of those brands' Instacart buyers did not buy the same brands in-store in the analysis period; Instacart calls this 'incremental', but the Circana items are buyer-overlap analyses, not lift tests (an inference from their described design)." | The 'not lift tests' clause is the file's inference from the described design and should be labelled as inference. Source RTL-X-S10 funding note 'commissioned or co-run' should quote the page: 'Circana and Instacart collaborated'; funding not stated. | RTL-X-F07 (evidence/verify/X.json) |
| C307 | Albertsons Media Collective and Roundel each state on record, in LiveRamp's release of 23 October 2025, that they measure Meta campaigns by joining Meta's ad data with their own first-party sales or guest data inside the LiveRamp Clean Room; LiveRamp says incrementality, halo analysis and new-buyer insights are "Coming soon". | Refute-mode verifier for evidence/research/X.json re-opened the cited source. | Dated 'October 23, 2025', 'SAN FRANCISCO--(BUSINESS WIRE)--'. Body (LiveRamp's words): 'Retail media networks can now unlock new attribution insights from their Meta campaign investments through the LiveRamp Clean Room. By connecting Meta campaign results with their own first-party sales data, RMNs and their partners can see how off-property sites, including Meta, drive sales, orders, and ROAS across campaigns, brands, and products.' Also: 'many RMNs already leverage LiveRamp to securely send their sales data to Meta for in-flight optimization' and 'Coming soon, additional capabilities such as | supported with qualification | "In LiveRamp's release of 23 October 2025, which says retail media networks can connect Meta campaign results with their own first-party sales data in the LiveRamp Clean Room to see how off-property sites 'drive sales, orders, and ROAS', Roundel's VP of Product Management says 'Clean room measurement of Meta ads with LiveRamp allows us to combine Target's first-party guest insights with Meta's platform ads data in a privacy-safe way', and Albertsons Media Collective's VP of Product Innovation says that with 'LiveRamp's expanded Meta insights' it connected 'audience-targeted Meta campaigns dire | The 'product' field names only Roundel although the finding also carries Albertsons Media Collective. measurement_record.identity_linkage 'clean room match on pseudonymised identifiers' goes beyond the release, which names no identifier ('privacy-safe'; 'protected and anonymized'): recode to 'clean room match (keys and identifiers not stated)'. measurement_record.observed_vs_modelled 'observed (as | RTL-X-F08 (evidence/verify/X.json) |
| C308 | CVS Media Exchange documents attributed sales for Pinterest campaigns through a partnership with Pinterest and LiveRamp that "utilizes clean room technology"; the page does not describe the join keys or who sees what. | Refute-mode verifier for evidence/research/X.json re-opened the cited source. | Undated page ('© Copyright 1999 - 2026 CVS Health'). Section 'Attributed sales on Pinterest': 'Get in front of potential and existing customers on a visual brand discovery platform, along with the attributable sales data you need to optimize your campaigns. CMX’s innovative partnership with Pinterest and LiveRamp utilizes clean room technology to create a data-safe targeted approach.' Section 'Turn Conversations into Conversions on Reddit': 'By combining CMX’s first-party insights from 90M+ ExtraCare members with Reddit’s 110M+ global daily active unique visitors brands can connect with health | supported with qualification | "CVS Media Exchange's Social page, under the heading 'Attributed sales on Pinterest', offers 'the attributable sales data you need to optimize your campaigns' and says 'CMX's innovative partnership with Pinterest and LiveRamp utilizes clean room technology to create a data-safe targeted approach'; the page ties the clean room to targeting and does not say the sales attribution is computed there, nor describe the join keys or who sees what." | The finding has no measurement_record (F08 has one); if one is added, identity_linkage for the attribution join should read 'not stated (clean room named for targeting)'. Undated page coded scope_time 2026-11-03: correct as an access date, not a launch date. | RTL-X-F09 (evidence/verify/X.json) |
| C309 | CVS Media Exchange announced on 29 February 2024 (Woonsocket dateline, reprinted by Mass Market Retailers) that pseudonymised ExtraCare data and Pinterest platform data meet in a LiveRamp-provided clean room that reveals the overlapping population, so that CMX can measure attributed Pinterest sales and report them to suppliers in the CMX Performance Dashboard; Pinterest's programme was a beta at the time. | Refute-mode verifier for evidence/research/X.json re-opened the cited source. | Mass Market Retailers (the www URL 301-redirects to https://massmarketretailers.com/cvs-media-exchange-cmx-partners-with-pinterest/, HTTP 200; article:published_time 2024-02-29T16:17:57.000Z, displayed date 'February 29, 2024' and time '11:17 AM', byline 'MMR', modified 2024-07-16): 'WOONSOCKET, R.I. — CVS Media Exchange (CMX), a data-led, technology-driven media network that enables partners to reach and better serve CVS Pharmacy shoppers, is partnering with Pinterest and LiveRamp on data clean rooms to bring greater transparency to partners (CPG brand suppliers). Provided by LiveRamp, this c | supported with qualification | "CVS Media Exchange's announcement (Woonsocket, R.I. dateline; reprinted by Mass Market Retailers on 29 February 2024 and quoted by Beet.TV on 1 March 2024 as 'CVS Media Exchange's announcement') says ExtraCare data are pseudonymised before entering a LiveRamp-provided clean room in which Pinterest's integration with LiveRamp reveals 'the overlapping population in ExtraCare data and Pinterest's platform data', and that 'with this technology' CMX can measure attributed sales related to campaigns and share them with CPG suppliers in the CMX Performance Dashboard, so that partners can 'track attr | 29 February 2024 is the reprint's publication date; no CVS-hosted copy was found, so the CVS announcement date itself is not established beyond 'no later than 29 February 2024'. measurement_record.assignment 'none' and counterfactual 'none' are not stated by the announcement: recode 'unknown' and 'not stated' (allocation is unknown unless disclosed). exposure_capture 'Pinterest platform data' is t | RTL-X-F12 (evidence/verify/X.json) |
| C310 | On 23 June 2026 Albertsons Media Collective announced an integration with Criteo through which 'eligible sponsored products can appear within AI-powered conversational search product carousels' in Albertsons' app, building on its 'ongoing partnership with Criteo'; Criteo called Albertsons Cos. its 'first retailer to bring Sponsored Products into an AI-powered shopping assistant'. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | Page date 06/23/2026. 'today announced a new integration with Criteo (NASDAQ: CRTO) ... bringing sponsored product discovery into Albertsons Cos.’ AI-powered conversational search'; 'Through Criteo, eligible sponsored products can appear within AI-powered conversational search product carousels'; 'The collaboration builds on Albertsons Media Collective’s ongoing partnership with Criteo'; Criteo (Sherry Smith): 'With Albertsons Cos. as our first retailer to bring Sponsored Products into an AI-powered shopping assistant'. McCormick (Jenny Dickinson): 'Being part of an experience like Albertsons | supported with qualification | Keep the claim. In the limitations, replace "stage 'announced' with availability unstated" with: "availability: available, commercial stage unstated (the release and the onsite page use the present tense and a participating advertiser, McCormick, describes being part of the experience); operator reporting unknown (McCormick's 'clearer view of how those moments translate into consideration and sales' names no metric); partner axes unknown; no protocol named." | H6 availability state: under the brief's states, 'announced' is reserved for a planned test or launch the source has not shown to have started; this source shows it in use. | RTL-ALBERTSONS-F04 (evidence/verify/albertsons-media-collective.json) |
| C311 | The onsite lift offer, announced 16 April 2026 as 'the launch of onsite incrementality measurement', uses 'test and control frameworks to isolate the effect of media exposure' and 'advanced matching methodologies grounded in predictive purchase behavior to construct credible test and control groups'; no page opened states allocation, matching variables, balance metrics, sample rules, uncertainty or whether standard onsite reports include lift. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | 16 April 2026 release (page date April 16, 2026; albertsonscompanies.com): 'today announced the launch of onsite incrementality measurement, a new capability ... this solution provides clear insight into how onsite display media is driving new sales'; 'This new capability isolates the causal impact of media from purchases that may have occurred regardless'; 'The methodology uses test and control frameworks to isolate the effect of media exposure'. Results page and the 23 March post both carry 'advanced matching methodologies grounded in predictive purchase behavior to construct credible test a | supported with qualification | Restate the last clause: "... no page opened states allocation, matching variables, balance metrics, sample rules or uncertainty; on whether reports include lift, the 23 March 2026 post says only that advertisers 'will see' matched reporting on onsite display 'when programming reaches the appropriate scale' (scale undefined), and the onsite offerings page ties incremental measurement to display ads, with no measurement statement for sponsored products or search." Add the release's own causal wording ('isolates the causal impact of media from purchases that may have occurred regardless') as a v | Assignment correctly left unknown; nothing in the record codes the matched design as randomised. | RTL-ALBERTSONS-F06 (evidence/verify/albertsons-media-collective.json) |
| C312 | On 17 September 2026 the network announced an 'Incrementality-Based Multi-Touch Attribution' built with LiveRamp connecting 'impression-level exposure data across ... owned and operated channels, offsite media and participating partners including Meta, third-party inventory with Display & Video 360, Google Ad Manager and Perion, with Albertsons transaction data in a neutral, privacy-conscious environment', PepsiCo 'an early pilot partner', no method, assignment, window or availability stage stated; the results page uses the future tense. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | Page date 09/17/2026; feed 09/17/2026 13:57. 'today announced Retail Media Incrementality-Based Multi-Touch Attribution, a revolutionary measurement solution ... through a single, deduplicated view of performance.'; 'Developed in collaboration with LiveRamp under a first-to-market, grocery exclusive collaboration and powered by its Cross-Media Intelligence infrastructure, the solution connects impression-level exposure data across Albertsons Media Collective’s owned and operated channels, offsite media and participating partners including Meta, third-party inventory with Display & Video 360, G | supported with qualification | Restate: "... announced 'Retail Media Incrementality-Based Multi-Touch Attribution', which it calls a launch, built with LiveRamp and 'powered by its Cross-Media Intelligence infrastructure', ... with PepsiCo 'an early pilot partner'; no method, assignment or window is stated and general availability is not stated (the results page uses the future tense)." Note beside 'neutral' that the word comes from the operator and from LiveRamp, which sells the environment. | Availability: the source says 'launch' twice and 'early pilot partner' once, so code 'announced as a launch; one named pilot partner; general availability not stated' rather than plain 'pilot'. 'Neutral' is a vendor's adjective, not an independence finding (LiveRamp is a paid party; the X verifier recorded the newsroom banner 'LiveRamp is now part of Publicis Groupe!'). | RTL-ALBERTSONS-F11 (evidence/verify/albertsons-media-collective.json) |
| C313 | The standard terms (Version 3.0, undated) make the network's ad server 'the measurement used for invoicing' and the 'Controlling Measurement', allow only 'Media Company tags' on its own properties, restrict Site Data to performing the IO and 'campaign reporting and measurement' with deletion at campaign end, oblige a cure of 'an incomplete or inaccurate report' within five business days, and leave one clause as an unfilled placeholder. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | All six passages verbatim. Context: the Controlling Measurement sentence sits in 'XIII. THIRD PARTY AD SERVING AND TRACKING (Applicable if Third Party Ad Server is used)', after 'Media Company will track delivery through its ad server and, Agency may not use any Third Party Ad Server without Media Company’s prior written consent.' Ad Tags: 'Ads for placement on Network Properties may only be tagged by Agency/Advertiser with authorized Third-Party Ad Server tags approved by Media Company. Ads for placement on Media Company Properties may only be tagged with Media Company tags. Media Company or | supported with qualification | Add to the limitations: "These are default terms: 'In the event of any inconsistency between the terms of an IO and these Terms, the terms of the IO will prevail', so a negotiated IO can change the measurement, tagging and data clauses; the remedy for an uncured reporting failure is possible 'nonpayment for all activity for which data is incomplete or missing'." | The same precedence caveat belongs in the onsite buyer_analysis and the three governance cells, which quote these clauses as fixed. | RTL-ALBERTSONS-F13 (evidence/verify/albertsons-media-collective.json) |
| C314 | The June 2023 standardization framework asked networks to disclose 'the type of attribution window, the statistical model used to calculate incrementality, and the cash gross-up methodologies', but no page, release or filing screened discloses Albertsons' own windows, incrementality model or gross-up method for any of its three rows. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | White paper (one page; PDF created 2023-06-16, modified 2023-07-19): 'Definitions of the elements for which RMNs need to provide disclosure, for instance, are the type of attribution window, the statistical model used to calculate incrementality, and the cash gross-up methodologies.' Release (page date 06/22/2023; dateline CANNES): 'today is sharing a preliminary framework to standardize specifications, methodologies, terminology and disclosures across retail media networks'; 'At Cannes Lions, Albertsons Media Collective is hosting a critical discussion on the topic – Transforming Retail Media | supported with qualification | Keep the claim. In the limitations and the interest field, replace 'features Omnicom Media Group as co-host' / 'co-hosted the framework's launch panel' with: "Albertsons Media Collective hosted the Cannes panel 'alongside leaders from Omnicom Media Group, Pinterest and R3', and Omnicom Media Group is quoted in the release." | The interest mark is appropriate (Omnicom Media Group is quoted and on the panel), but its wording overstates the role: the release names Albertsons as host, not Omnicom as co-host. | RTL-ALBERTSONS-F14 (evidence/verify/albertsons-media-collective.json) |
| C315 | No Albertsons service appears on MRC's accredited or under-review digital lists (41 and 14 entries, 3 November 2026), on AAM's Incrementality Certification list or IAB Europe's certified list; Albertsons is listed among company participants of the IAB/MRC guidelines (January 2024), which is participation, not assurance, and its in-store release only states that it is 'closely aligning our measurement roadmap with IAB current and future standards for in-store media'. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | MRC pages (server-rendered, fetched 3 November 2026): no 'Albertsons' or 'Collective' string on either list; the under-review list has 14 entries (Adelaide, DoorDash Ads, Escalated, Google YouTube Relay Log, IAS YouTube Brand Safety, Kevel, Kroger, Meta Third-Party Brand Suitability Pipeline, Ogury, Partnerize, Pixalate, Schwarz Digits (Lidl & Kaufland Interact), Walmart Connect Onsite Display, XPLN). Criteo's accredited metrics are 'Clicks: Display' and 'Rendered Ad Impressions: Display', each with '* This accreditation applies specifically to On-site Sponsored Ads, Commerce Display Ads and S | supported with qualification | Quote the in-store sentence in full: it says 'remaining compliant with and closely aligning our measurement roadmap with IAB current and future standards for in-store media', a self-asserted compliance claim with no attestation, not only an alignment statement. In the onsite assurance cell, replace 'two of its staff among contributors (p. 5)' with 'one main contributor (Claire Wyatt, p. 5) and one other contributor (Vin Lay, p. 9)'. | Criteo's MRC scope (display clicks and rendered display impressions for Criteo's own On-site Sponsored Ads, Commerce Display Ads and Standard Banner Ads) is correctly not carried to the Albertsons onsite row: the record names no retailer and only delivery metrics. 'Albert Heijn' on IAB Europe's list is a Dutch Ahold Delhaize banner, not Albertsons. | RTL-ALBERTSONS-F15 (evidence/verify/albertsons-media-collective.json) |
| C316 | The network sells near-store DOOH on '120,000 digital screens near stores' with 'incremental measurement that proves performance', and its Primo Water case (April to May 2025, powered by Perion) reports 'a 5.5% incremental sales lift' with no design, control, sample or interval. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | Case page (date 09/24/2025): 'Boost awareness of Primo Water among Albertsons shoppers on the go, reaching them during their everyday routines across three priority divisions.'; 'Albertsons Media Collective activated a high-impact digital out-of-home campaign powered by Perion across a mix of screens from transit hubs to gyms and offices where Albertsons shoppers are most active.'; 'Mid-flight optimizations zeroed in on high-performing stores, enhancing efficiency and maximizing impact.'; 'resulting in a 5.5% incremental sales lift.'; '*Campaign duration: April–May 2025; performance improvemen | supported with qualification | Add to the claim or limitations: "The case places the screens 'from transit hubs to gyms and offices where Albertsons shoppers are most active' and does not say they were the '120,000 digital screens near stores' of the DOOH page; it also says 'Mid-flight optimizations zeroed in on high-performing stores', which, absent a stated design, means the 5.5% cannot be read as a pre-specified test-versus-control result." | 'Incremental' is the source's word; target contrast incremental (stated), counterfactual and assignment unknown: correctly coded. | RTL-ALBERTSONS-F19 (evidence/verify/albertsons-media-collective.json) |
| C317 | Default not documented; lift offer matched (assignment unknown), randomized testing 'upon request', onsite display reporting gated by 'appropriate scale'; Ovative statements stay with the study; Sponsored Search reports a ROAS change only; onsite incrementality dated four ways. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | See F06 to F08. Additional: the 23 March post says 'As an advertiser, you will see this level of disciplined reporting on ... onsite display capabilities when programming reaches the appropriate scale'; the onsite page ties incrementality to display ('delivering incremental return on ad spend with onsite display'); the Alby Awards release (11 December 2025) guarantees ROAS, not iROAS: 'Albertsons Media Collective will guarantee ROAS performance throughout Q1 in 2026' and 'will extend a custom Q1 2026 ROAS guarantee to each award recipient based on their Q1 2025 performance', calling it 'a powe | supported with qualification | Keep 'Default not documented' for the row, and add the reason in one sentence: "The only statement of what advertisers receive is conditional ('you will see' matched reporting on onsite display 'when programming reaches the appropriate scale', scale undefined) and covers display only; for sponsored products and search no reporting default is stated, and the network's one public performance guarantee (11 December 2025) is written on ROAS, an attributed metric, computed per advertiser." Coding: assignment unknown for the matched design (correct); random only for the on-request option, as labelle | Matched was coded unknown, not randomised, and not observational: correct under the brief's control-coding rule. If the coordinator codes H1a at row level, the conditional display-only statement should be recorded as a separate flag (lift in default report: conditional, display only, scale undefined) rather than filling the default. | CAPABILITY onsite default_method (Albertsons Media Collective onsite) (evidence/verify/albertsons-media-collective.json) |
| C318 | No qualifying advertising revenue figure: media advertising is an undisclosed share of 'Other' (US$1,305.6M FY2025, 'Consists primarily of wholesale sales to third parties, commissions, rental income, media advertising revenue ...') and partly booked as a reduction of cost of sales; 'Advertising costs ... $524.7 million' is an expense, never revenue. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | Filing text as under F01. scoring-rules.json scale_rules.perimeter_codes: 'broader parent proxy (a line that spans other named networks or non-advertising revenue: Walmart's global advertising business "including VIZIO"; Instacart's "advertising and other"; Cardlytics' company-wide revenue; a company-wide retail media segment that includes other businesses) caps at 3'; 'Undisclosed service revenue is never prorated from a broader line'; evidence_caps: 'No qualifying figure ... is n/e, never 0'; unit for retailer_network: 'annual advertising revenue (US$, stated fiscal period)'. | supported with qualification | n/e is defensible but departs from the literal perimeter text, so the cell reason must say why and the sensitivity must be logged. Suggested reason: "Not coded broader parent proxy: the line is a residual 'Other' line whose named principal components are non-advertising (wholesale, commissions, rental income), and it is not a superset of the network's advertising revenue, because media advertising 'classified as ... a reduction in Cost of sales' sits outside it; it is therefore neither a proxy nor a bound for advertising revenue, and no advertising revenue figure qualifies (n/e)." Sensitivity: | The expense line is correctly treated as an expense ('Advertising costs are included in Cost of sales and are expensed in the period the advertising occurs'), never as revenue. The perimeter examples all name advertising-led lines, which supports the profile's reading; the code's wording does not, which is why the departure must be explicit. | CAPABILITY scale (all three rows): no qualifying figure, n/e (evidence/verify/albertsons-media-collective.json) |
| C319 | No Albertsons entry on MRC accredited or under-review lists, AAM Incrementality Certification or IAB Europe certified list; IAB/MRC guidelines participant and 'two of its staff among contributors (p. 5)'; Criteo's MRC grant never transfers; no onsite alignment statement. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | See F15. | supported with qualification | Replace 'two of its staff among contributors (p. 5)' with 'one main contributor (Claire Wyatt, p. 5) and one other contributor (Vin Lay, p. 9)'. | Stage 'not begun or not public' is the right bound: absence from the lists says nothing about unannounced work. | CAPABILITY onsite assurance (evidence/verify/albertsons-media-collective.json) |
| C320 | Network API 'now available for all clients' (fields and granularity undocumented); TransUnion receives campaign data; multi-touch results operator-produced; Site Data restricted and destroyed at campaign end; ad server is the 'Controlling Measurement'; only Media Company tags on owned properties. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | See F12 and F13. | supported with qualification | Add: "Terms are defaults that yield to the IO ('the terms of the IO will prevail'); API access is described as 'near real time'." | No buyer-run query access or event-level export is documented; H3 axes (a) and (b) correctly not documented in the screened record. | CAPABILITY onsite buyer_analysis (evidence/verify/albertsons-media-collective.json) |
| C321 | No versioning, restatement policy or revision record; five-day cure clause and invoice corrections; ad server is the 'Controlling Measurement'; methodology claimed documented but not public; unfilled placeholder clause. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | Terms and post passages verbatim (see F07, F13). The post: 'Methodologies are documented and applied systematically at The Collective. Incrementality is not recalculated to fit a narrative or adjusted on a campaign-to-campaign basis without clear rationale.' | supported with qualification | Add the cure remedy ('Failure to cure may result in nonpayment for all activity for which data is incomplete or missing') and the IO-precedence clause to item (2)/(4). | CAPABILITY governance (all three rows) (evidence/verify/albertsons-media-collective.json) | |
| C322 | Multi-touch pilot joins impression-level exposure across owned channels, offsite media and partners (Meta, DV360, Google Ad Manager, Perion) with Albertsons transaction data in a LiveRamp environment; Collective TV uses PAIR with LiveRamp and TTD closed loop; keys and matchable share unstated; H coded L2. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | Channel list verbatim (F11). Collective TV: 'As an early adopter of PAIR, Albertsons Media Collective connects Albertsons’ informed shopper audiences with premium content, while leveraging LiveRamp's platform to enhance security and privacy for publisher and marketer data collaboration.' LiveRamp release of 23 October 2025 (as verified by the X verifier, RTL-X-S11): Albertsons Media Collective's VP of Product Innovation, Evan Hovorka: 'With LiveRamp’s expanded Meta insights, we have improved the visibility and accuracy of lower-funnel results. We did this by connecting audience-targeted Meta c | supported with qualification | Add RTL-X-S11 (LiveRamp, 23 October 2025) as the on-record operator statement for the Meta rail: Albertsons' VP of Product Innovation says it connected 'audience-targeted Meta campaigns directly to item level sales' using 'LiveRamp’s expanded Meta insights' (his quote does not name the clean room; LiveRamp's body text does). Record the join keys as not stated. | This firms up the H4b outside-match credit that H coded on the 2026 'neutral, privacy-conscious environment' sentence; it should be cited, not left to the home page's HTML comment. | CAPABILITY offsite identity_basis (H4b) (evidence/verify/albertsons-media-collective.json) |
| C323 | Store-level sales, no shopper match; '116 test stores in the Mondelēz campaign'; control-store count not stated; matching on 'nearly 60 variables'. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | 'The addition of in-store screens across 116 Albertsons banner stores contributed to a significant sales lift, as measured by The Collective’s Matched Market Incrementality (iROAS) methodology.' | supported with qualification | Replace '116 test stores' with '116 stores with in-store screens in the campaign (test and control store counts used in the analysis not stated)'; make the same change in the in-store iROAS measurement record ('reported campaign result (116 test stores)'). | CAPABILITY in-store identity_basis (evidence/verify/albertsons-media-collective.json) | |
| C324 | Dated events from 11 November 2021 to 13 October 2026. | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | Page date elements and feed entries match every event checked: 11/11/2021, 06/22/2023, 06/13/2024, 12/11/2024 (lawsuit headline), 07 January 2025 dateline (re-posted 02/24/2025), 06/16/2025, 01/06/2026, 02/12/2026 (OpenAI pilot headline), 03/23/2026, 04/16/2026, 2026-04-27 (10-K filing date), 06/17/2026 (P&G headline), 06/23/2026, 09/17/2026, 2026-10-13 (Q2 8-K). The 17 September 2026 release calls the multi-touch product a 'launch' with PepsiCo as 'an early pilot partner'. | supported with qualification | Change the 2026-09-17 event_type from 'pilot (announced)' to 'launch (announced; one named pilot partner, PepsiCo)'. Add two events: 2025-12-11, 'Alby Awards announced; ROAS guarantee for Q1 2026 extended to award recipients' (RTL-VALBERTSONS-S01); 2026-07-23, 'Q1 fiscal 2026 earnings release: no media or advertising sentence' (RTL-VALBERTSONS-S03). | The 2026-01-28 STRATACACHE and 2026-06-17 Trade Desk items were not re-opened. | TIMELINE (23 events) (evidence/verify/albertsons-media-collective.json) |
| C325 | 'A home-page block (inside an HTML comment, not rendered) describes a "custom ROAS guarantee" for 2026 award winners; it is not used as evidence.' | Refute-mode verifier for evidence/profiles/albertsons-media-collective.json re-opened the cited source. | Rendered release, page date 12/11/2025, listed in the operator's feed ('12/11/2025 09:00:00 | Albertsons Media Collective Doubles Down on Performance – Announcing Inaugural Alby Awards Honoring Top-Performing Advertising Partners'): 'The awards celebrate top-performing advertising partners who delivered the strongest Return on Ad Spend (ROAS) on the platform in 2025. In recognition of their proven performance, Albertsons Media Collective will guarantee ROAS performance throughout Q1 in 2026.'; 'Winners were selected through a clear, standardized evaluation framework that compared each brand’s | supported with qualification | Replace the sentence with: "On 11 December 2025 the network announced a custom Q1 2026 ROAS guarantee for each Alby Award recipient 'based on their Q1 2025 performance' (RTL-VALBERTSONS-S01): a contractual-guarantee use of an attributed metric; the terms of the guarantee (window, remedy, ROAS definition) are not published." Carry it to the paper's trade/guarantee table and to the onsite 'uses' attribute. | Use: contractual guarantee on ROAS (attributed, operator-defined), not on iROAS; never call the guaranteed ROAS a return. | HOW_IT_EARNS (ROAS guarantee sentence) (evidence/verify/albertsons-media-collective.json) |
| C326 | Amazon describes its standard attribution as last touch ('It credits a conversion to one ad click or view') and announced at unBoxed 2024 a multi-touch attribution beta that 'divides credit for Amazon purchase conversions across Amazon Ads touchpoints in proportion to their value' using 'insights from experiments, machine learning, and historical shopping signals', offered as 'an additional way' to view results; the page says US advertisers 'will soon' have access and the v1 multi-touch fields show 'Program eligibility: N/A' on 3 November 2026, so the beta's current stage for Sponsored Product | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | MTA page (en-us, browser pane; title 'Uncover ad value across the funnel with Multi-touch attribution (beta)'; 'Announced at unBoxed 2024'; no date): 'US advertisers will soon be able to access multi-touch attribution metrics in sponsored ads and Amazon DSP reporting'; 'Amazon Ads’ traditional, standard attribution is “last touch”: It credits a conversion to one ad click or view, usually the last ad encountered on the path to purchase.'; 'Multi-touch metrics are available alongside familiar last-touch metrics in Ad console, Amazon DSP campaign manager, sponsored ads downloadable reports (Campa | supported with qualification | Amazon describes its standard attribution as last touch ('It credits a conversion to one ad click or view, usually the last ad encountered on the path to purchase') and announced at unBoxed 2024 (press release 15 October 2024: testing 'later this year, with a full rollout planned for 2025') a multi-touch attribution beta that 'divides credit for Amazon purchase conversions across Amazon Ads touchpoints in proportion to their value', offered as 'an additional way' to view results. The undated beta page says US advertisers 'will soon' have access, lists console and downloadable-report availabili | A planned rollout date is an announcement, not evidence of general availability (shared rule 8). | RTL-AMAZON-F04 (evidence/verify/amazon-ads.json) |
| C327 | Amazon documents a restatement schedule for sponsored ads data: initial clicks within 12 hours with changes 'up to three days' after the click from traffic validation, initial conversions within 24 hours restated '1, 7, and 28 days after the conversion event' (so a 14-day window can change up to 42 days after the report date), while its reporting overview says conversion data 'is subject to change for up to 60 days' and invalid-traffic detection runs 'for up to 30 days after the fact'. | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | v3 FAQ: 'Initial impression and click data is available using the API within 12 hours. Note that small changes to impression and click data may happen up to three days after the initial click date due to the traffic validation process. ... Initial conversion data is available within 24 hours, but may change as part of the restatement process. Restatements of conversion data occur 1, 7, and 28 days after the conversion event. Because conversions are reported on the date of the ad interaction, restatement includes all dates in your attribution window. For example, if your attribution window is 1 | supported with qualification | Keep the claim; add to it or to its limitations: 'the FAQ adds that conversions are reported on the date of the ad interaction, so restatement covers every date in the attribution window; the overview links click validation to charging ("Amazon validates every click to make sure advertisers are not charged unnecessarily") and says Amazon Marketing Stream v2 backfills up to 70 days "as attribution validation completes"'. Replace the limitation 'whether billing follows restated clicks' with: 'the overview ties invalid-click validation to charging but does not say whether fees are adjusted after | Third horizon for the tension list: 42 days (FAQ example), 60 days (overview), 70-day Stream backfill. | RTL-AMAZON-F05 (evidence/verify/amazon-ads.json) |
| C328 | Amazon's Advertising Agreement (posted 23 July 2026; 'formerly known as the Sponsored Ads Agreement') bills on Amazon's own counts ('Fees are based solely on our measurements'), requires fee claims within 60 days through the console's 'Contact Us' feature, disclaims guarantees about 'THE ACCURACY OF AD SERVICES DATA, TARGETING, REPORTING', limits the buyer's use of Ad Services Data to managing campaigns through the Ad Services, evaluating performance 'only for its internal purposes' and passing Performance Data to the advertiser, allows disclosure to contractors bound by equivalent confidentia | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | Agreement (browser pane, lang en-US, 'Posted: July 23, 2026'). 4.A: 'Fees are based solely on our measurements for the Ad Services and the applicable billing metrics (e.g., impressions or clicks). Customer understands that third parties may generate impressions or clicks on Ads for improper purposes, and Customer accepts this risk.'; 'Customer waives all claims related to the Fees, unless made within 60 days after the date charged. To initiate a fee dispute, Customer may use the “Contact Us” feature in the Advertising Console.' Disclaimer B: 'EXCEPT AS EXPRESSLY SET FORTH IN THIS AGREEMENT, AN | supported with qualification | Keep every quotation (all verbatim) and add: 'Section 4.A also states "Customer understands that third parties may generate impressions or clicks on Ads for improper purposes, and Customer accepts this risk"; the Performance Data definition expressly includes "whether a User viewed or clicked on an Ad (i.e., User-level data)"; and for US customers any dispute relating to the Agreement or the Ad Services goes to binding arbitration.' In the limitations, change 'no procedure for challenging an outcome metric was found' to 'no measurement-specific challenge procedure was found; disputes other tha | The invalid-traffic risk clause and the user-level wording bear directly on the billing (Q2) and event-level access (H3 axis a) codings: the contract contemplates user-level Performance Data, while no route delivering it for Sponsored Products is documented. The page links the incorporated terms at https://advertising.amazon.com/terms/data-privacy and a change log at https://advertising.amazon.com | RTL-AMAZON-F08 (evidence/verify/amazon-ads.json) |
| C329 | For Amazon Sponsored Products, Amazon documents buyer-controlled outcome analysis in Amazon Marketing Cloud (H3 axis (b) documented): sponsored-ads advertisers 'can access AMC via self-service', run SQL-style queries over a table with '1 record per sponsored ads impression event and 1 record per click event', and over conversion tables in which 'Each conversion event in the table is attributed to its corresponding traffic event' within a 14-day window, receiving 'only ... aggregated, anonymous outputs'; per-column aggregation thresholds are shown only as labels (LOW or NONE) with no numeric va | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | sponsored_ads_traffic page (browser pane): table rows carry aggregation thresholds NONE, LOW, MEDIUM (event_dt, event_dt_utc), HIGH (household_size) and VERY_HIGH (event_id, household_id, ip_address, user_id); 'The field has a VERY_HIGH aggregation threshold, meaning it cannot be included in final SELECT statements but can be used within Common Table Expressions (CTEs) for aggregation purposes'; user_id: 'Pseudonymous identifier that connects user activity across different events. The field can contain ad user IDs (representing Amazon accounts), device IDs, or browser IDs. NULL values appear w | supported with qualification | Replace '(LOW or NONE)' in the claim with: 'per-column aggregation thresholds are labelled NONE, LOW, MEDIUM, HIGH or VERY_HIGH with no numeric value; VERY_HIGH fields (user_id, household_id, ip_address, event_id) cannot appear in a final SELECT'. Make the same correction in the X-S05 support locator, capabilities.buyer_analysis ('shown per column as LOW or NONE'), the rows_at_a_glance table ('thresholds LOW/NONE labels') and open question 3 ('Only LOW or NONE labels were shown'). | Confirms stream X's verifier (RTL-X-F03) on an independent read: the profile repeats the wrong 'LOW or NONE' wording in four places. The rest of the finding (self-service access, event-level SP traffic paired with attributed conversions, aggregated outputs, axis (a) not documented) is not disturbed. | RTL-AMAZON-F10 (evidence/verify/amazon-ads.json) |
| C330 | Since 1 January 2026, Amazon Sponsored Brands and display ads campaigns billed on viewable impressions, and Amazon DSP campaigns serving ads in the Store, report Purchases, Sales and ROAS under a 'shopping-signal enhanced last-touch attribution model' with 'a shorter attribution window' that Amazon does not quantify, keep 14-day 'all views' metrics as a parallel series, and leave click-based attribution, and so Sponsored Products, unchanged (US among 21 countries). | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | Page dated 'January 01, 2026'. All quoted passages verbatim. 'Where is the feature available? North America: United States, Canada, Mexico / South America: Brazil / Europe: United Kingdom, Germany, Spain, Italy, France, Belgium, Poland, Turkey, Netherlands, Sweden, Ireland. / Middle East: United Arab Emirates, Saudi Arabia, Egypt / Asia Pacific: Australia, India, Japan, Singapore. / Africa: South Africa' = 23 countries. Scope bullet: 'Sponsored Brands, Sponsored Display campaigns billed on a viewable impression (vCPM) basis, including branded keyword campaigns'. | supported with qualification | Change '(US among 21 countries)' to '(US among 23 countries)'; make the same change in product_capabilities 'Amazon display ads'.us_availability ('in the US among 21 countries'). | The workstream U browser note said 'US and 20 other countries'; the page lists 23. Whether the vCPM qualifier also attaches to Sponsored Brands is grammatically ambiguous on the page; the file's reading (vCPM for both) is consistent with 'Click-based attribution remains unchanged' and with the v1 note that Sponsored Brands CPC reports 0 view conversions. | RTL-AMAZON-F13 (evidence/verify/amazon-ads.json) |
| C331 | Amazon's standard reporting includes a modelled outcome for US 'Sponsored Ads and Amazon DSP customers': long-term sales (LTS) and LTS ROAS are 'ad-attributed measures that estimate the incremental sales value your brand can expect to generate over the next 12 months' (since 15 October 2024), which the version 3 column definition describes as the 12-month potential value of new-to-brand engagements 'plus the 14d total sales of the campaign'. | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | LTS page dated 'October 15, 2024', 'Announced at unBoxed 2024': 'Long term sales (LTS) and long term sales ROAS (LTS ROAS) are ad-attributed measures that estimate the incremental sales value your brand can expect to generate over the next 12 months, ... This is based on the historical 12-month return of a given customer engagement with your brand.'; 'Who can use it? Sponsored Ads and Amazon DSP customers in the US'. Amazon press release of the same date: 'Long-term sales insights are currently in closed beta and will be available to U.S. advertisers in open beta later this year.' v3 columns: | supported with qualification | Amazon's reporting includes a modelled outcome announced on 15 October 2024 for US 'Sponsored Ads and Amazon DSP customers' (then in closed beta, with US open beta promised 'later this year'): long-term sales (LTS) and LTS ROAS are 'ad-attributed measures that estimate the incremental sales value your brand can expect to generate over the next 12 months'; the version 3 field longTermSales defines it as the 12-month potential value of new-to-brand engagements 'plus the 14d total sales of the campaign' and lists Sponsored Display, Sponsored Brands, Sponsored TV and DSP campaign reports, not Spon | Timeline entry 2024-10-15 is coded 'launch'; per Amazon's release it is an announcement with the product in closed beta (open beta planned for later in 2024). | RTL-AMAZON-F17 (evidence/verify/amazon-ads.json) |
| C332 | Amazon's multi-touch attribution paper (arXiv 2508.08209, 11 August 2025; eight authors listed with Amazon affiliations, including Brett R. Gordon and Florian Zettelmeyer, also of Northwestern University) records calibration experiments: randomised controlled trials in which 'one group of shoppers is eligible to see ads from the focal campaign ("treatment") while another group of shoppers is not ("holdout")', drawn from 'a database of hundreds of thousands of RCTs across our ad products'; it reports no sample sizes, holdout shares, periods, intervals or results. | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | arXiv 2508.08209v1, dated 11 August 2025, author line: 'Randall Lewis1, Florian Zettelmeyer1,2,3, Brett R. Gordon1,2, Cristobal Garib1, Johannes Hermle4, Mike Perry4, Henrique Romero1, and German Schnaidt1' with '1Amazon 2Northwestern University 3NBER 4Work done at Amazon'. p. 3: 'In an advertising RCT, one group of shoppers is eligible to see ads from the focal campaign ("treatment") while another group of shoppers is not ("holdout").' p. 4: 'RCTs can produce results that are too imprecise (i.e., wide confidence intervals) ...'; p. 4 also cites the IAB/MRC guidelines as one that 'highly recom | supported with qualification | Replace 'eight authors listed with Amazon affiliations, including Brett R. Gordon and Florian Zettelmeyer, also of Northwestern University' with 'eight authors: six with an Amazon affiliation and two (Johannes Hermle, Mike Perry) marked "Work done at Amazon"; Florian Zettelmeyer (Amazon, Northwestern University, NBER) and Brett R. Gordon (Amazon, Northwestern University)'. | The 'no sample sizes, holdout shares, periods, intervals or results' statement holds: the only numbers are an illustrative group size and a toy example's weights. Calibration RCTs and the reported model are kept apart correctly (F18 vs F19). Amazon's 15 October 2024 release repeats the RCT claim ('hundreds of thousands of randomized control trial experiments – A/B tests') and is the dated public s | RTL-AMAZON-F18 (evidence/verify/amazon-ads.json) |
| C333 | Amazon's 'Approved third-party providers' page (last updated 1 June 2026) documents third-party ad verification for Amazon DSP only ('available on Amazon DSP' with DoubleVerify, Integral Ad Science and Adloox, by server-to-server measurement) and lists 'Samba TV' with vendor type 'Measurement' among about sixty measurement vendors; it carries no verification or measurement-partner statement for Sponsored Products, Sponsored Brands or display ads beyond a link to the sponsored ads policies. | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | Page 'Last updated: June 1, 2026'. Intro: 'Use of any third-party provider to serve and enhance ad content (including creative assets, pixels, and cookies) is restricted to certified technologies.' and 'The following table(s) is a comprehensive list of all currently certified technologies for ad serving, creative hosting, ad tagging, measurement, post bid ad verification, viewability, research studies, and similar services.' Table row 'Samba TV | Measurement' (58 rows typed Measurement). 'Ad verification via third-party vendors such as DoubleVerify, Integral Ad Science, and Adloox is available | supported with qualification | Keep the claim. Rewrite the limitation's scope sentence: 'Interest: Samba (the author's employer) is listed. The page presents the vendor table as "a comprehensive list of all currently certified technologies" without tying it to Amazon DSP or any ad product; the listing establishes certification, not a partnership, an integration or any campaign, and the page does not say whether Samba TV is among the "50+ vendor products" supported in the Amazon DSP self-service measurement-study workflow.' Add the self-service measurement-study section to the Amazon DSP row (buyer_analysis, routes). | The Samba interest mark is present in F23, the timeline and the DSP buyer_analysis cell, and is worded correctly; only the scope phrase 'approval to measure on Amazon DSP' over-reads the page. | RTL-AMAZON-F23 (evidence/verify/amazon-ads.json) |
| C334 | This review found no Amazon page documenting a controlled-lift offer (randomised holdout, ghost ads or another counterfactual) delivered to advertisers for Sponsored Products campaigns, on the pages screened by 3 November 2026 (Sponsored Products page, the 1 January 2026 attribution notice, the prompts page, the agentic-shopping page, the multi-touch attribution page, AMC's product and data-source pages, and the API reporting overview, columns, FAQ, v1 overview and Purchases family) and in two searches for such a page; the documented outcome measures are click attribution, a multi-touch beta a | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | Amazon Brand Lift page (en-us, browser pane, undated): 'Which ad products can Amazon Brand Lift measure? All ad campaigns run through Display, Video, and Audio campaigns and sponsored ads are eligible for Amazon Brand Lift, except for Sponsored Products.'; 'They are also available to sponsored ads advertisers in the United States.'; 'Amazon Brand Lift is free as long as campaigns meet spend and impression minimums.'; 'Amazon Brand Lift calculates results and generates a report that shows absolute lift, which is the difference between the ad-exposed group’s and control group’s response rate'; ' | supported with qualification | This review found no Amazon page documenting a sales-lift offer (randomised holdout, ghost ads or another counterfactual) delivered to advertisers for Sponsored Products campaigns, on the pages screened by 3 November 2026 (the list as written, plus Amazon's Brand Lift page and its 15 October 2024 press release) and in two searches; and Amazon's survey-based Brand Lift product, which compares 'an ad-exposed audience and an unexposed control group', states that sponsored ads and Display, Video, and Audio campaigns are eligible 'except for Sponsored Products'. The documented outcome measures for | Part of the absence is now a documented exclusion (a source documents absence for brand lift), which matters for the 0-versus-n/e choice: a 0 needs the scored criterion to match what the page excludes (survey brand lift, not sales lift); assignment for Brand Lift's control group is not disclosed ('a similar group of customers who haven’t seen an ad'). | RTL-AMAZON-F28 (evidence/verify/amazon-ads.json) |
| C335 | Amazon Ads announced on 10 September 2026 a partnership with OpenAI 'to support advertising in ChatGPT', with 'Select advertisers in the U.S.' testing ChatGPT Ads 'as part of a pilot'; the page describes no reporting, attribution or measurement for the placement and does not tie it to any scored Amazon row. | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | Page dated 'September 10, 2026': 'Amazon Ads has announced a partnership with OpenAI to support advertising in ChatGPT.'; 'Select advertisers in the U.S. are currently testing ChatGPT Ads as part of a pilot.'; a quotation from 'Chris Conetta, Director of Omnichannel Supply at Amazon DSP'; 'Katrin Koenig, President of Delta Vacations': 'We're proud to be among the first brands testing this new approach'; closing link 'Learn more about using Amazon DSP to grow your business'; related product 'Amazon DSP'. No reporting, attribution or measurement for the placement is described (boilerplate only: | supported with qualification | Amazon Ads announced on 10 September 2026 a partnership with OpenAI 'to support advertising in ChatGPT', with 'Select advertisers in the U.S.' testing ChatGPT Ads 'as part of a pilot' (Delta Vacations names itself 'among the first brands testing'); the page describes no reporting, attribution or measurement for the placement and associates it with Amazon DSP (a quotation from an Amazon DSP director and a closing Amazon DSP link) without stating the buying route. | The limitation 'The buying route (Amazon DSP) is ... not stated in the page body' stays true literally, but 'does not tie it to any scored Amazon row' over-states: the page points to Amazon DSP. | RTL-AMAZON-F31 (evidence/verify/amazon-ads.json) |
| C336 | Eight attribution parameters for Sponsored Products as coded in capabilities.attribution_disclosure | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | MTA page: 'It credits a conversion to one ad click or view, usually the last ad encountered on the path to purchase.' Purchases family: 'ROAS = sales / total cost'; 'ROAS (reconciled)' = 'sales / total cost reconciled', Program eligibility including Sponsored Products. v3 click variants on eight sp report types (not spGrossandInvalids). | supported with qualification | In (2) quote the full sentence including 'usually the last ad encountered on the path to purchase'. Add to (6) the spend basis: 'ROAS = sales / total cost', with a reconciled variant ('sales / total cost reconciled'; reconciliation undefined on the page). | No change to which parameters are documented with values (window, touch model, halo) or missing (match basis, in-store, new-to-brand, sales basis). | capabilities.attribution_disclosure (Amazon Sponsored Products) (evidence/verify/amazon-ads.json) |
| C337 | No page opened names the identity key behind the Sponsored Products click-to-purchase match | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | AMC sponsored_ads_traffic: user_id 'can contain ad user IDs (representing Amazon accounts), device IDs, or browser IDs. NULL values appear when Amazon Ads is unable to resolve an identifier for an event.'; household_id 'Pseudonymous household ID'. Agreement: Performance Data includes 'whether a User viewed or clicked on an Ad (i.e., User-level data)'. | supported with qualification | Keep 'basis asserted only' for the report match key, and add: 'In AMC, Amazon documents the identity fields on Sponsored Products traffic: a pseudonymous user_id that may be an Amazon account, device or browser ID (NULL when unresolved) and a pseudonymous household_id; this documents the clean room's identifiers, not the key used for the default click-to-purchase report.' | Stream X's verifier made the same point for X's measurement record (identity linkage is mixed, not account-only). | capabilities.identity_basis (Amazon Sponsored Products) (evidence/verify/amazon-ads.json) |
| C338 | Governance screen: versioning, restatement, revisions, dispute procedure, reproducibility | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | Agreement: binding arbitration for 'any dispute or claim relating in any way to this Agreement or the Ad Services' (US, CA, MX); page links 'Changes to the Amazon Advertising Agreement' (https://advertising.amazon.com/terms/changes). Overview: Stream v2 'automated backfill up to 70 days'; 'Amazon validates every click to make sure advertisers are not charged unnecessarily.' | supported with qualification | Under (4) add the general binding-arbitration clause beside the fee-claim window; under (2) add the 70-day Stream backfill and the click-validation-to-charging sentence; under (1) note that a dated change log exists for the Agreement (not read), separate from the API change log that did not load. | None of these additions creates a measurement-specific challenge procedure. | capabilities.governance (Amazon Sponsored Products) (evidence/verify/amazon-ads.json) |
| C339 | US production use: the FTC complaint alleges about 1.2 million US advertising customers across its sponsored ads | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | ¶4: 'approximately 1.2 million U.S. advertising customers, which include over 500,000 small and medium-size businesses' (not limited to sponsored ads in that paragraph); ¶1 says 'approximately 1.2 million advertising customers' and names three products. | supported with qualification | Change "'approximately 1.2 million U.S. advertising customers' across its sponsored ads" to "'approximately 1.2 million U.S. advertising customers' (¶4; ¶1 ties the overcharge to Sponsored Products, Sponsored Brands and Sponsored Display)". | Allegation; not a count of Sponsored Products advertisers. | capabilities.us_availability (Amazon Sponsored Products) (evidence/verify/amazon-ads.json) |
| C340 | No controlled-lift offer documented for Sponsored Brands or display ads | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | 'All ad campaigns run through Display, Video, and Audio campaigns and sponsored ads are eligible for Amazon Brand Lift, except for Sponsored Products.'; 'They are also available to sponsored ads advertisers in the United States.'; reports show 'absolute lift, which is the difference between the ad-exposed group’s and control group’s response rate'; free above 'spend and impression minimums'; results 'in as few as 10 business days'; also via the Amazon Ads API. | contradicted | Sponsored Brands default_method: replace 'no controlled-lift offer documented on the pages screened' with 'a survey-based brand-lift study is documented as an optional offer for US sponsored ads other than Sponsored Products (Amazon Brand Lift: ad-exposed versus unexposed control group, assignment not disclosed, outcome = survey response rates, free above spend and impression minimums); no sales-lift offer documented on the pages screened'. Apply the same addition to Amazon display ads (default_method and validation 'None found'). | Code Brand Lift's counterfactual as a control group with assignment unknown; outcome unit survey responses, not sales; it is an optional study, not the default method. | product_capabilities: Amazon Sponsored Brands and Amazon display ads, default_method / validation (controlled lift) (evidence/verify/amazon-ads.json) |
| C341 | Amazon DSP: 'brand lift' and 'sales lift' appear only as navigation links; no controlled-lift design documented | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | Brand Lift: 'All ad campaigns run through Display, Video, and Audio campaigns ... are eligible'; available 'to all Amazon Ads Agent advertisers in the United States, Canada, Germany, Spain, France, Mexico, Japan, and United Kingdom'. Approved providers page: 'Amazon Ads offers a self-service workflow for third-party measurement studies within Amazon DSP. Advertisers and agencies can discover which of 50+ vendor products are supported, configure and fund studies, and access results all within Amazon DSP.'; 'Managed within Amazon DSP: Studies are billed directly through Amazon.'; 'Nielsen Market | contradicted | Replace 'no controlled-lift design documented on the pages opened' with: 'Amazon Brand Lift (survey lift against an unexposed control group; assignment not disclosed) is documented for Display, Video, and Audio campaigns, and Amazon DSP documents a self-service workflow for third-party measurement studies ("50+ vendor products", billed through Amazon or managed with the vendor, 18 countries); no sales-lift design with method is documented on the pages opened.' Add the study workflow to the DSP routes and buyer_analysis cells. | Third-party studies billed through Amazon are commissioned by the advertiser via the operator's workflow; code reporting party and funding per study, not as independent by default. | product_capabilities: Amazon DSP default_method / buyer_analysis (controlled lift, third-party studies) (evidence/verify/amazon-ads.json) |
| C342 | Amazon Sponsored Products holds the only MRC accreditation of a sponsored-products service in the register | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | MRC accredited digital list (WebFetch, 3 November 2026): Criteo's footnote 'This accreditation applies specifically to On-site Sponsored Ads, Commerce Display Ads and Standard Banner Ads'; stream X's same-day read of the same page records 'Walmart Connect Sponsored Search' ('Rendered Ad Impressions: Display', 'Rendered Ad Impressions: Video', 'Clicks: Search', 'Click-through Rate (CTR)') and an entry 'Instacart'. The paper's product register carries 'Walmart Connect Sponsored Search' and 'Instacart Sponsored Products' as cohort A rows and Criteo Commerce rows in cohort C. | contradicted | Replace 'its Sponsored Products row holds the only MRC accreditation of a sponsored-products service in the register (clicks and invalid-traffic filtration, total only)' with 'its Sponsored Products row holds an MRC accreditation limited to clicks and SIVT filtration, total only (no impressions); other register services selling sponsored placements also appear on MRC's list with impression metrics (Walmart Connect Sponsored Search; Criteo, for "On-site Sponsored Ads"), and Instacart is listed (its letter's mapping to the Instacart Sponsored Products row was not checked here)'. | Not a numbered finding, but it is the profile's headline framing of its MRC record; 'only' does not survive the list. Walmart's and Instacart's letters were not opened in this pass; the Walmart Sponsored Search entry alone is enough to defeat 'only'. | why_profiled (MRC claim) (evidence/verify/amazon-ads.json) |
| C343 | 2024-10-15 launch: Long-term sales (LTS) and LTS ROAS metrics for US sponsored ads and Amazon DSP customers | Refute-mode verifier for evidence/profiles/amazon-ads.json re-opened the cited source. | LTS page dated 15 October 2024 headed 'What launched?'; Amazon's release of the same date: 'Long-term sales insights are currently in closed beta and will be available to U.S. advertisers in open beta later this year.' | supported with qualification | Recode event_type 'launch' as 'announcement (closed beta; US open beta planned later in 2024)'. Add a timeline row for 15 October 2024: Amazon announces multi-touch attribution testing 'later this year, with a full rollout planned for 2025' (press.aboutamazon.com). | timeline 2024-10-15 (long-term sales) (evidence/verify/amazon-ads.json) | |
| C344 | Two different exposure-to-purchase rules apply to the same service: Ads Manager's revenue and ROAS count purchases by users who clicked on or activated an offer, while Cost per Served Sale fees count all purchases from the marketer by served consumers regardless of engagement; neither carries a counterfactual, and the pages opened do not say which rule a CPS advertiser's dashboard ROAS uses. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | Glossary (updatedAt 2026-02-19): "Engagements | Number of users who clicked on or activated an offer."; "Purchases (post-engagement) | Number of purchases made by users who activated an offer."; "Revenue (post-engagement) | Revenue from purchases by users who activated an offer."; "ROAS (post-engagement) | Return on ad spend = Revenue of users who activated an offer divided by ad spend." The API reporting page lists Engagements ("clicked on or activated") and Activations ("activated") as separate fields and defines Purchases and Revenue on "users who activated an offer (post-engagement)". | supported with qualification | Replace 'Ads Manager's revenue and ROAS count purchases by users who clicked on or activated an offer' with 'Ads Manager's revenue and ROAS count purchases by users who activated an offer (the documentation labels them post-engagement, but a click without activation is counted only in Engagements)'. | This answers the brief's question: yes, the documented ROAS counts only purchases by users who activated an offer. Whether that revenue is limited to qualifying redemptions or covers any purchase from the marketer by an activator is not stated. | RTL-CARDLYTICS-F07 (evidence/verify/cardlytics.json) |
| C345 | Cardlytics offers incremental measurement only as a post-campaign Test vs. Control report outside the self-serve tool: its FAQ says it 'withholds a representative control group' to compare exposed and unexposed spending, the measurement page lists Test vs. Control, Long-Term Value and Share Shift as 'Commonly requested post-campaign reports', and the Ads Manager FAQ (5 October 2026) says the iROAS framework is not yet in Ads Manager; no allocation rule, control size, minimum spend, significance rule or interval is published. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | FAQ: "Using a best-in-class test vs. control methodology, Cardlytics withholds a representative control group to make apples-to-apples comparisons between the spending patterns of people who were exposed to an ad, and those who were not." and "Cardlytics' view of bank transactions provides a line of sight into consumer purchase behavior before and during the campaign." Measurement page: "Commonly requested post-campaign reports include:" Test vs. Control "Quantify sales lift and incremental return"; "We've set up a rigorous test vs. control process to answer three questions every marketer want | supported with qualification | Replace 'offers incremental measurement only as a post-campaign Test vs. Control report outside the self-serve tool' with 'offers its own incremental measurement as a post-campaign Test vs. Control report outside the self-serve tool (the Nielsen Sales Lift option is the other incremental route, RTL-CARDLYTICS-F09)'. 'Not yet in Ads Manager' is a fair paraphrase of 'planning to incorporate'; equating the 'iROAS testing framework' with the Test vs. Control report is an inference from the shared vocabulary ('incremental return on ad spend') and should be labelled so. | Assignment is correctly coded unknown: 'withholds a representative control group' names a holdout and says nothing about how members are allocated. The FAQ's pre-period line shows pre-campaign data exist, not that any balance or pre-trend check is run. | RTL-CARDLYTICS-F08 (evidence/verify/cardlytics.json) |
| C346 | For marketing-mix models and market-level tests Cardlytics says it 'typically' provides weekly impacted sales, online and in-store, by geography with the associated media spend, and the Ads Manager FAQ says it partners with 'many of the leading MMMs'; 'impacted sales' is not defined. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | FAQ: "For those using MMM to understand how our program impacts your topline sales, just put us in touch with your MMM consultant, and we'll deliver the data you need to validate our results. We typically provide weekly impacted sales (online and in-store) by geography, with the associated media spend." Ads Manager FAQ: "Yes, we partner with many of the leading MMMs and have processes in place to ensure the right data is passed for their models." | supported with qualification | Add: Cardlytics delivers the feed to the buyer's MMM consultant on introduction and presents it as the data 'you need to validate our results'. This bears on two other records: the CPS terms' consent requirement for disclosure to a consultant (F12) is met in practice by Cardlytics sending the data itself, and the validation cell (F21) should record a validation route offered, not a validation published. | 'Impacted sales' remains undefined on both pages. | RTL-CARDLYTICS-F11 (evidence/verify/cardlytics.json) |
| C347 | Cardlytics' US master advertiser terms for Cost per Served Sale make its own data 'the definitive data used to calculate the fees', restrict the client's use of campaign data to internal evaluation, bar disclosure to any consultant or third party without Cardlytics' written consent, require deletion within 45 days, and let report types change with bank-partner obligations; the text returned contains no dispute window, audit right or third-party measurement clause. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | Section 4 'Insights Portal' (applies 'In the event that Cardlytics issues Client with user credentials to Cardlytics' Insights Portal'): "certain features of the Insights Portal, including the types of reports and dashboards available thereon, are subject to change based on obligations imposed on Cardlytics by its financial institution partners. The Insights Portal and the data contained therein is intended for Client's internal use and as such, Client may not disclose such data or any derivatives thereof to any consultant, subcontractor or third party without Cardlytics' written consent." Sec | supported with qualification | Restate: 'Cardlytics' US master advertiser terms for Cost per Served Sale make its own data "the definitive data used to calculate the fees" (the client also pays fees above the IO budget); bar disclosure of Insights Portal data "to any consultant, subcontractor or third party without Cardlytics' written consent"; limit any transaction-level or individual offer-activation information to "internal offer verification or internal evaluation" and require its deletion within 45 days; treat campaign-performance reports as confidential information shareable with the client's contractors or agents who | The over-generalisation matters for the buyer-analysis cell: the 45-day deletion and use limit apply to row-level data, the consent bar to Insights Portal data, not to all campaign reporting. | RTL-CARDLYTICS-F12 (evidence/verify/cardlytics.json) |
| C348 | Cardlytics states that banks supply de-identified data and that it receives no PII and has no 'insight into transactions in sensitive categories such as healthcare'; US consumers activate offers in their bank's app and opt out through the bank, its UK notice rests processing on 'the consent that has been given by the customers to our banking partners', and the 10-K names the Gramm-Leach-Bliley Act, the CCPA and California's Delete Act among the regimes it is subject to. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | FAQ: "We do not have access to any personally identifiable information (PII) or insight into transactions in sensitive categories such as healthcare."; "Through our bank partnerships we see credit, debit, ACH, and bill pay transactions, including related purchases made via mobile wallets."; "If you are currently participating in and want to opt out of a Cardlytics card-linked marketing program, you can do so through your bank."; "No personally identifiable information — such as payment card information, names, addresses, phone numbers, email addresses, birth dates, and account numbers — ever l | supported with qualification | Add to counterevidence: (1) the Redemption Feed passes card type and the card's last four digits (if available) to advertiser partners, which sits uneasily with the FAQ's statement that no 'payment card information' ever leaves the bank (last four digits alone may not be PII; record the tension, not a breach); (2) Cardlytics' own UK insights piece reports pharmacy spend growth, so 'no insight into transactions in sensitive categories such as healthcare' does not exclude pharmacy as a merchant category (whether Cardlytics counts pharmacy as healthcare is not stated; UK, out of US scope). | Not re-opened here: the UK bank-customer notice (RTL-CARDLYTICS-S04), the 10-K sentence 'Our platforms are designed so that we do not receive or have access to any personal data from our FI partners' and the CCPA passage (verify/E3.json confirmed related 10-K privacy passages for RTL-E3-F18). | RTL-CARDLYTICS-F14 (evidence/verify/cardlytics.json) |
| C349 | Cardlytics replaced monthly active users with MQUs from 1 January 2025: it counted 224.2 million MQUs ('targetable customers that have made a transaction using their account with an FI Partner') in 2025 against 190.5 million in 2024, while ACPU fell from US$0.67 to US$0.50 'primarily as a result of lower billings'; none of its size figures measures the share of an advertiser's transactions it can match. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | 10-K: "As of January 1, 2025, we no longer report Cardlytics Monthly Active Users" ... "given we do not receive equivalent user data from our newer bank partners."; "We define MQUs as targetable customers that have made a transaction using their account with an FI Partner"; MQUs 224,159 / 190,482 (thousands); ACPU $0.50 / $0.67; "Cardlytics ACPU decreased by $0.17 during 2025 compared to 2024" ... "primarily as a result of lower billings and the ramp up of our newest FI Partner." Publisher documentation (28 August 2025) and the 24 March 2026 release boilerplate: "With visibility into approxima | supported with qualification | Complete the quote: ACPU fell 'primarily as a result of lower billings and the ramp up of our newest FI Partner', and MAUs were dropped 'given we do not receive equivalent user data from our newer bank partners'. Keep 'none of its size figures measures the share of an advertiser's transactions it can match', but add that Cardlytics asserts visibility into 'approximately half of all card-based transactions in the U.S. and a quarter in the U.K.' (publisher documentation, 28 August 2025; repeated in the release of 24 March 2026, after the Bank of America exit): a vendor-asserted, market-wide cove | The 'newest FI Partner' ramp is the 10-K's own context for the 2025 MQU rise and the ACPU fall. | RTL-CARDLYTICS-F16 (evidence/verify/cardlytics.json) |
| C350 | Cardlytics' finances are strained: accumulated net losses reached US$1,404.1 million by the end of 2025, US$172.5 million of 4.25% convertible notes fall due in 2029, cash was US$28.0 million at 30 June 2026, Nasdaq notified it on 3 June 2026 that its bid price had closed below US$1.00 for 30 business days (compliance date 30 November 2026), and a 1-for-10 reverse split took effect on 5 June 2026. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | 10-K: "Through December 31, 2025, we have incurred accumulated net losses of $1,404.1 million since inception"; "On April 1, 2024, we issued $172.5 million in principal amount of our 4.25% Convertible Senior Notes due in 2029". 8-K (date of report 3 June 2026), Item 3.01: notice received 3 June 2026 from Nasdaq's Listing Qualifications Staff; bid price closed below $1.00 "for the last 30 consecutive business days"; "the minimum closing bid price required by the continued listing requirements of Nasdaq Listing Rule 5550(a)(2)"; 180 calendar days, "until November 30, 2026 (the 'Compliance Date') | supported with qualification | Replace 'US$172.5 million of 4.25% convertible notes fall due in 2029' with 'it issued US$172.5 million of 4.25% convertible notes due 2029 (1 April 2024)': the passage establishes the amount issued, not the principal outstanding at the freeze, which was not checked. Strengthen the compliance limitation: no Cardlytics filing from 1 June to 3 November 2026 states that compliance was regained (EDGAR full-text search), and the Q2 2026 10-Q does not use the phrase 'minimum bid price'. | 'Strained' is the author's characterisation (inference) of filed facts; label it. | RTL-CARDLYTICS-F18 (evidence/verify/cardlytics.json) |
| C351 | This review found no validation of Cardlytics' outcome numbers (no comparison of its attributed or test-versus-control results with an experiment, panel, point-of-sale data or replication) in the filings and pages opened by 3 November 2026; the record holds only accuracy claims such as 'our measurements are not probabilistic or based on models' and 'There is no need to estimate performance'. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | 10-K: "our measurements are not probabilistic or based on models". Measurement page: "With insight into bank transaction data, we report our exact impact on sales. There is no need to estimate performance. These are real purchases, made by real people who saw ads in our platform." FAQ: "just put us in touch with your MMM consultant, and we'll deliver the data you need to validate our results." | supported with qualification | Add after the accuracy claims: 'Cardlytics offers buyers' MMM consultants data "to validate our results" (FAQ), a validation route, not a published validation.' The FAQ repeats the measurement page's wording and adds 'or rely on complex attributions or integrations'. | Absence bounded to the pages read; no case study was opened (the Dunkin' page was not re-tried). | RTL-CARDLYTICS-F21 (evidence/verify/cardlytics.json) |
| C352 | 8-K with Items 1.02 (termination of a material definitive agreement) and 8.01; not opened, so its subject is not established here. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | Date of report "April 22, 2025". Item 1.02: "On April 22, 2025, the Company received a written non-renewal notice (the "Notice") from Bank of America" for the General Services Agreement dated 7 July 2022 and related Statement of Work and Change Order; expiry "effective as of July 31, 2025"; 180-day continuation through 27 January 2026, extendable "by mutual agreement of the parties"; impact "will not be material during or after the Extension Period". Item 8.01: an amendment (16 April 2025) to the loan facility with Banc of California extending maturity to 15 April 2028, with "$60.0 million of | supported with qualification | Replace the entry with: '2025-04-22 (8-K filed 2025-04-28): Bank of America's written non-renewal notice for the agreements under which Cardlytics publishes offers to its customers (Item 1.02); separately, a loan-facility amendment with Banc of California (Item 8.01).' Merge it with the existing 2025-04-22 entry and cite the 8-K beside the 10-K. | TIMELINE-2025-04-28 (evidence/verify/cardlytics.json) | |
| C353 | 8-K with Items 2.05 (costs associated with exit or disposal activities) and 7.01; not opened. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | 8-K filed 2025-10-02, report date 2025-10-01, items 2.05 and 7.01 (0001666071-25-000135). | supported with qualification | Date the event 2025-10-01 (date of report) and the filing 2025-10-02. | Content still not opened. | TIMELINE-2025-10-02 (evidence/verify/cardlytics.json) |
| C354 | Default: attributed, no counterfactual. Ads Manager shows attributed sales, revenue and ROAS, with revenue and ROAS defined post-engagement as purchases by users who activated an offer divided by ad spend. The main pricing model bills on the same kind of match: CPS fees ... regardless of whether consumers engage. Controlled lift: a post-campaign Test vs. Control report ('withholds a representative control group') via the client success manager because the iROAS testing framework is not yet in Ads Manager; assignment unknown, counterfactual 'holdout' (source term); Nielsen Sales Lift option on | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | Every quoted passage matches the pages (see F06 to F09). 10-Q Note 5 revenue by pricing model, Q2 2026: served based 11,819, engagement based 23,872, other 1,191 (total 36,882). | supported with qualification | Replace 'The main pricing model bills on the same kind of match: CPS fees are ...' with 'Served-based pricing (Cost per Served Sale, 32.0% of Q2 2026 revenue; the engagement-based family, billed on purchases after engagement, was 64.7%) bills on a broader match: CPS fees are ...'. CPS is not the main pricing model in any period the 10-Q shows (served-based 30.7% of H1 2026, 40.5% of H1 2025). | Assignment unknown and counterfactual 'holdout (source term)' are coded correctly; no method label was promoted to random assignment. | CAP-default_method (evidence/verify/cardlytics.json) |
| C355 | Window: no attribution window stated for the Ads Manager metrics; activated offers valid for redemption 30-45 days; CPS billing counts purchases during the campaign period, no day value. Touch model: post-engagement (click or activation) for Ads Manager revenue and ROAS; served-before-purchase for CPS. Match basis account level; in-store included; NTB, deduplication, halo and returns missing; Cardlytics' fees exclude taxes. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | Redemption Feed (updatedAt 2026-02-13): "Lookback for Redemptions is 90 Days: This refers to the period during which Cardlytics considers a transaction eligible for an offer redemption. If an offer is activated, and a qualifying transaction occurs within 90 days of that activation, it will be considered a redemption and included in the Redemption Feed. Transactions occurring outside of this 90-day lookback window (e.g., a transaction from 91 days ago) will not be attributed as a redemption for active offers". Performance Reporting: "various redemption windows ranging from 30-45 days"; Purchase | supported with qualification | Window: add 'the Redemption Feed documents a 90-day lookback from activation for a transaction to count as a redemption'. Touch model: replace 'post-engagement (click or activation)' with 'activation (the source labels it post-engagement; clicks count only in Engagements)'. Revise the count line: a window in days is stated for redemptions (90-day lookback; 30-45-day offer redemption windows), not for the post-engagement Purchases, Revenue and ROAS metrics or for CPS billing; NTB, deduplication, halo and returns remain missing. Note that the CPS fee basis is a default an insertion order can ove | May change the dimension's parameter count if the anchors credit a stated window for any reported metric; redemptions are a reported quantity (redemption rates appear in the publisher documentation and the feed). | CAP-attribution_disclosure (evidence/verify/cardlytics.json) |
| C356 | Dashboard and Reporting Hub with CSV export and scheduled email reports, two-year retention; Performance Reporting API with offer-level aggregates; hourly Redemption Feed (one row per redeemed transaction, identifiers further obfuscated per partner); weekly impacted sales for MMMs; Nielsen option; no clean room, buyer-run query environment, exposure-level export or named third-party replication. Terms: report data for internal use, no disclosure to consultants or third parties without consent, internal evaluation only, deletion within 45 days. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | Reporting Hub: "Export all will directly export the report to your computer as a CSV file."; "Schedule will allow you to schedule future automated reports to be sent to your email."; "our two-year data retention window". Redemption Feed: "Each row in the feed represents a unique, anonymized transaction where a Cardlytics offer has been redeemed."; fields include "transaction ID, cardholder ID, merchant and offer identifiers, store information, merchant name, card type, and the last four digits of the card (if available), along with the transaction amount and timestamp"; "For each partner, thes | supported with qualification | Restate the terms sentence as in F12 (consent bar on Insights Portal data; internal-use limit and 45-day deletion on transaction-level or individual activation information; campaign-performance reports shareable with contractors or agents under confidentiality; IO prevails). Add that the redemption feed carries store information, card type and the card's last four digits where available. Add the FAQ's 'we'll deliver the data you need to validate our results' to the MMM item. | The terms also bar the client from using reporting to identify a participating customer or to 'identify, target, or analyze any individual or group as a customer of any given financial institution' (section 8), a further limit on buyer-side analysis. | CAP-buyer_analysis (evidence/verify/cardlytics.json) |
| C357 | De-identified FI-partner transaction data (credit, debit, ACH, bill pay), measured at customer and account level; in-store and online; matchable share not disclosed; MQUs, US$5.7 trillion purchases analysed and US$4.1 trillion spend seen are incompatible units; none states what share of a marketer's sales is paid with cards or accounts at partner banks. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | 10-K: "Our marketing is targeted and measured based on actual purchase data at a customer and account level."; "During 2025, the Cardlytics platform analyzed approximately $5.7 trillion in purchases"; "The data provided by the FI partner is anonymized so that it cannot be associated with any one individual." FAQ: "credit, debit, ACH, and bill pay transactions, including related purchases made via mobile wallets"; "With over 186M bank customers ... we've driven over $46B+ in measurable sales for marketers." Measurement page: "we see over $4.1T in consumer spend." Publisher documentation and the | supported with qualification | Add the vendor-asserted coverage share ('approximately half of all card-based transactions in the U.S. and a quarter in the U.K.'; undated measurement basis, repeated after the Bank of America exit) and the FAQ's '186M bank customers' and '$46B+ in measurable sales' to the list of incompatible size descriptors. The sentence 'None states what share of a marketer's sales is paid with cards or accounts at partner banks' stays true; the coverage claim is market-wide, not per marketer, and is a vendor assertion. | Evidence grade A is generous for a cell whose matching description rests partly on vendor pages; the filing passages carry it. | CAP-identity_basis (evidence/verify/cardlytics.json) |
| C358 | No validation record found; accuracy language only; Nielsen Sales Lift is an optional third-party report, not a published validation; the Dunkin' case study returned 404 and is not used. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | FAQ: "we'll deliver the data you need to validate our results" (to the buyer's MMM consultant). | supported with qualification | Add: 'Cardlytics offers buyers' MMM consultants weekly data "to validate our results"; no result of such a validation is published.' | Dunkin' 404 not re-tried. | CAP-validation (evidence/verify/cardlytics.json) |
| C359 | One served class: offers placed inside partners' digital channels (online banking and mobile apps); no off-site or in-store media sold in the record read; exposure capture = offer serves logged; engagement = click or activation; match = card or account transaction; parameters incomplete. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | Publisher page: "Cardlytics targets these specific customers with personalized offers via online banking and mobile apps."; publishers include "financial institutions and non-financial institutions (brands)". FAQ: "Can I use purchase data to target ads outside of the bank channel? No, our advertising solution is only available within our partner banks' brand-safe and fraud-free digital channels." API Overview: "the Cardlytics network, which includes a wide range of financial institutions and brand partners." Glossary: "Impressions | Total number of serves of an offer to all users." | supported with qualification | Cite the FAQ's explicit statement that the advertising is 'only available within our partner banks' ... digital channels' for the no-off-site point, and note that the publisher and API pages also name non-FI 'brand' publishers, so the served surfaces are partners' channels, mostly but not only banks'. | CAP-placement_coverage (evidence/verify/cardlytics.json) | |
| C360 | No change logs, restatement policy or dispute procedure; purchases up to 7-14 days delay and performance data updated hourly after a 4-day delay; CPS terms set 30-day payment and make Cardlytics' data definitive with no dispute window or audit right; Cardlytics determines targeting parameters in its sole discretion and may change reports and dashboards based on FI-partner obligations. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | "Most spend-related data (e.g. purchases) can have up to 7-14 days delay for transaction data completeness due to a lag time in data delivery from some publishers."; "Performance data will be updated hourly after an initial 4-day delay."; "Client shall pay such invoice within thirty (30) days from the invoice date."; "Unless otherwise stated in an Insertion Order, Cardlytics shall determine the targeting parameters and definitions set forth in each Insertion Order in its sole discretion."; the report-change clause sits in section 4 'Insights Portal'; section 13: disputes "will be governed by t | supported with qualification | Prefix the targeting-discretion clause with 'Unless otherwise stated in an Insertion Order'; say the report-change clause concerns the Insights Portal; record that the only dispute clause is governing law and venue, and that the client 'agrees to pay for the fees that exceed the budget stated in the Insertion Order'; add that an insertion order prevails over the terms. | Absences (restatement policy, audit right, billing-dispute window) are now verified on the full terms page, not only on the tool's excerpt. | CAP-governance (evidence/verify/cardlytics.json) |
| C361 | US consent via bank apps and opt-out through the bank; publisher documentation says customers must opt in; UK processing on consent given to banks; no PII; no insight into sensitive categories such as healthcare; GLBA, state laws, CCPA and Delete Act named; California registry: Cardlytics, Inc. absent, 'Bridg, a division of Cardlytics, Inc.' listed; no independent assurance; no adverse events in J's searches. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | See RTL-CARDLYTICS-F14 (FAQ and publisher passages match). FAQ: "No personally identifiable information — such as payment card information ... — ever leaves the bank." Redemption Feed fields include "card type, and the last four digits of the card (if available)". UK insights piece: pharmacy spend "growing 65–75% in every month of H1". Publisher page: "Segments can be based on demographics"; FAQ: "Can I target by audience demographics? No." | supported with qualification | Add the card-last-four field in the advertiser Redemption Feed and the pharmacy reporting as tensions with the FAQ's statements, and the publisher page's demographic segments as a tension with the FAQ's 'No' (all three recorded as vendor-document inconsistencies, not as breaches). | Not re-opened here: the UK notice (S04), the privacy policy (RTL-J-S24) and the California registry (RTL-J-S09). | CAP-privacy (evidence/verify/cardlytics.json) |
| C362 | Attributed; assignment none; estimation 'sum of purchases by users who clicked on or activated an offer'; exposure capture 'credit requires a click or activation'; touch model 'post-engagement (click or activation)'; window none stated; reporting party Cardlytics; access restrictions per CPS terms. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | Purchases, Revenue, AOV and ROAS (post-engagement) are each defined on "users who activated an offer"; Engagements are "users who clicked on or activated an offer". Source term on the measurement page: "Close the loop between campaign performance and actual marketing results". | supported with qualification | estimation_method: 'sum of purchases by users who activated an offer; ROAS = that revenue divided by ad spend'; exposure_capture: 'offer serve logged; credit requires activation'; touch_model: 'activation (source label post-engagement)'. access_restrictions: restate per F12 (the consent bar covers Insights Portal data; the 45-day deletion covers row-level information). | 'Close the loop' is kept as the source's term and not read as causal: correct. | MR-1 (Ads Manager revenue, purchases, AOV and ROAS) (evidence/verify/cardlytics.json) |
| C363 | Descriptive (no counterfactual described); counterfactual 'not stated'; estimation not described. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | Long-Term Value: "We look at the purchase patterns of people who converted from the original campaign to measure their continued loyalty to the marketer's brand – specifically, how many more times they purchase and how much they spend." Share Shift: "we help brands understand how their customers' behavior is changing in their category, and if they're picking up share from their competitive set", illustrated with 'SHARE OF SPEND: Before campaign / After campaign' for the advertiser and three competitors. | supported with qualification | Share Shift: counterfactual 'pre/post (before-campaign versus after-campaign share of spend, as illustrated)', target contrast descriptive. Long-Term Value: population 'people who converted from the original campaign', no comparison group described. | MR-5 (Long-Term Value and Share Shift) (evidence/verify/cardlytics.json) | |
| C364 | Target contrast unknown ('impacted sales' undefined); weekly sales by geography with media spend; analysis by the buyer or its MMM provider; CPS disclosure limit may apply. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | See RTL-CARDLYTICS-F11 and F12. | supported with qualification | access_restrictions: Cardlytics itself delivers the data to the buyer's MMM consultant on introduction ('just put us in touch with your MMM consultant, and we'll deliver the data'); the terms' consent bar covers Insights Portal data, so whether any restriction applies to this feed is not stated. Add the source's purpose statement: data 'to validate our results'. | MR-6 (weekly impacted sales for MMMs) (evidence/verify/cardlytics.json) | |
| C365 | The default report and the main pricing model are attributed with no counterfactual: post-activation ROAS in Ads Manager and CPS fees on all purchases by served consumers. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | Q2 2026 revenue by pricing model: served based 11,819 (32.0%); engagement based 23,872 (64.7%); other 1,191. | supported with qualification | Replace with: 'The default report and both pricing families are attributed with no counterfactual: post-activation ROAS in Ads Manager; engagement-based fees (64.7% of Q2 2026 revenue) on purchases after engagement; and Cost per Served Sale fees (32.0%) on all purchases by served consumers.' | The same misidentification appears in CAP-default_method; F07's counterevidence ('engagement-based models (most of revenue)') already has it right. | LIMITS-1 (evidence/verify/cardlytics.json) |
| C366 | No attribution window in days, new-to-brand lookback, deduplication rule or returns treatment is published. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | "Lookback for Redemptions is 90 Days" ... "If an offer is activated, and a qualifying transaction occurs within 90 days of that activation, it will be considered a redemption". | supported with qualification | Restate: 'No attribution window in days is published for the post-engagement Purchases, Revenue and ROAS metrics or for CPS billing (the Redemption Feed documents a 90-day lookback from activation for redemptions, and offers carry 30-45-day redemption windows); no new-to-brand lookback, deduplication rule or returns treatment is published.' | LIMITS-3 (evidence/verify/cardlytics.json) | |
| C367 | Buyers get dashboards, aggregates and a redemption-only feed under terms that bar third-party disclosure without consent and require deletion within 45 days. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | See RTL-CARDLYTICS-F12. | supported with qualification | Restate: '... under CPS terms that bar disclosure of Insights Portal data to third parties without consent and require deletion of transaction-level or individual activation data within 45 days (an insertion order can override).' | LIMITS-5 (evidence/verify/cardlytics.json) | |
| C368 | Three buyer routes are documented: an Ads Manager console, account-team managed campaigns and an Advertiser API with redemption feed and aggregate reporting. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | "It is designed for partners who have card-linked offers (CLO) and represent local merchants, such as restaurants." | supported with qualification | Restate the third route as 'an Advertiser API for demand partners with card-linked offers for local merchants, with a redemption feed and aggregate reporting'. | SWOT-strengths-3 (evidence/verify/cardlytics.json) | |
| C369 | Terms make Cardlytics' data definitive for fees, bar disclosure to third parties without consent and require deletion of campaign data within 45 days. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | See F12: the consent bar covers Insights Portal data; the 45-day deletion covers transaction-level or individual offer-activation information. | supported with qualification | Restate: 'Terms make Cardlytics' data definitive for fees, bar disclosure of Insights Portal data to third parties without consent and require deletion of row-level transaction or activation data within 45 days.' | SWOT-weaknesses-3 (evidence/verify/cardlytics.json) | |
| C370 | No MRC, AAM or IAB Europe instrument and no published validation cover any Cardlytics outcome metric. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | See F20 and F21. | supported with qualification | Bound the absence (shared hard rule 1): 'No MRC, AAM or IAB Europe instrument and no published validation of any Cardlytics outcome metric were found in the lists and pages screened by 3 November 2026.' | SWOT-weaknesses-4 (evidence/verify/cardlytics.json) | |
| C371 | Revenue fell 36.5% year on year in Q2 2026 after a US partner's exit; Bank of America's relationship ended on 16 February 2026. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | The 10-Q attributes the MQU decrease, not the revenue decrease, to 'an FI partner in the U.S. exiting'; in April 2025 the company said Bank of America's non-renewal "will not be material during or after the Extension Period to its overall financial results given the revenue contribution from Bank of America,". | supported with qualification | Restate: 'Revenue fell 36.5% year on year in Q2 2026, and MQUs fell 39.0 million after a US partner's exit; Bank of America's relationship ended on 16 February 2026.' Do not imply that the partner exit caused the revenue fall: the 10-Q text read gives no cause for it, and the company expected Bank of America's exit to be immaterial. | SWOT-threats-1 (evidence/verify/cardlytics.json) | |
| C372 | A Nasdaq bid-price notice, US$172.5 million of 2029 notes and Bridg indemnity costs strain the company's finances. | Refute-mode verifier for evidence/profiles/cardlytics.json re-opened the cited source. | See F17 and F18. | supported with qualification | Say 'US$172.5 million of 2029 notes issued in April 2024' (outstanding principal not checked) and keep 'strain' labelled as inference; the US$6.4 million settlement is partly offset if the insurance reimbursement the company is seeking is paid. | SWOT-threats-4 (evidence/verify/cardlytics.json) | |
| C373 | IRI-NPD merger completed 1 August 2022 with H&F 'the majority investor in the combined company' and Vestar and New Mountain Capital retaining 'significant investments'; H&F's portfolio page says H&F acquired NPD in 2021 and in 2022 merged NPD with IRI, 'renaming the combined company Circana'; no source names Vista; no filing or later ownership statement found. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | Merger post (header 'Jul 31, 2022', 'Updated: Sep 26, 2024'; dateline 'CHICAGO and PORT WASHINGTON, NY, Aug. 1, 2022'): "H&F, a premier global private equity firm, is the majority investor in the combined company. Vestar Capital Partners (“Vestar”) and New Mountain Capital (“NMC”), which were long-term investors in IRI, retain significant investments in the combined company." H&F page (dated 'December 15th, 2021'): "In 2021, H&F acquired The NPD Group, a leading provider of market research, and in 2022, merged NPD with Information Resources, Inc. (“IRI”), renaming the combined company Circana. | supported with qualification | Keep the 2022 structure as stated. Replace 'No filing or later ownership statement was found' with: 'No filing was found. A Vestar Capital Partners release of April 2024 announcing a US$1.2 billion continuation vehicle for its Circana stake surfaced in a verifier search but was not opened; until it is read, the current ownership split is unverified.' Do not use the H&F page to date the rename: it folds the renaming into the 2022 sentence, while the rename was announced on 7 March 2023 (RTL-CIRCANA-S10). | Vista Equity Partners appears on none of the pages opened. | RTL-CIRCANA-F04 (evidence/verify/circana.json) |
| C374 | Circana sells NCS Sales Effect on its NCS Measurement page as a separately branded service measuring 'incremental sales lift and ROAS for your national consumer packaged goods campaigns or for campaigns running in major and consolidated markets', for brands, publishers and retail media networks, delivered 'through a managed service model or a self-service model in a dashboard or clean room'; no design, panel, window or sample. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | Page listed in the site navigation under 'NCSolutions Portfolio' (NCS Enablement, NCS Optimization, NCS Targeting, NCS Measurement). "NCS Sales Effect studies not only help you understand what happened in your campaign, they help you see what’s next." "Sales Effect gives you a new level of detail that drives actionability and a precision that delivers the intelligence on how the creative, audience, frequency and media impact your campaign outcomes. Whether you’re launching a new product, trying to understand the full sales performance of your retail media investment or examining the value of a | supported with qualification | Restate: 'Circana sells NCS Sales Effect on its NCS Measurement page, listed under an "NCSolutions Portfolio" beside NCS Targeting, NCS Optimization and NCS Enablement, as a separately branded study service with its own "Speak to an Expert" contact; "No matter which Sales Effect you choose" indicates more than one Sales Effect variant. The page states no price, contract or product code, so separate purchasability is implied by the offer page, not documented in terms. It measures "incremental sales lift and ROAS" for national CPG campaigns or campaigns "in major and consolidated markets", repor | The 'Affordable and adaptable ... Our flexible plans are built around your business needs. Start Your Free Trial' block is a site template that appears verbatim on the Media Lift, NCS Measurement and Retail Media pages; it is not evidence of a product-specific trial, price or self-serve route for any of them. | RTL-CIRCANA-F08 (evidence/verify/circana.json) |
| C375 | Circana and DoorDash announced on 2 September 2026 'independent, third-party validation to DoorDash's measurement capabilities', 'Now available to select advertisers for Sponsored Products and Sponsored Brands'; DoorDash attributes '$0.59 in incremental in-store sales per household' to 'a household-level lift study'; Circana presents the same $0.59 as 'Early measurement results' after 'New findings from DoorDash measurement'; the footnoted Circana analysis covers an eight-week US campaign 11/10/25 to 1/4/2026; no design, panel, match basis or payer. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | Circana post ('Sep 2, 2026'): "New findings from DoorDash measurement illustrate the platform's incremental value. An overlap analysis of buyers of measured brands showed that 76% purchased in store only, 13% purchased on DoorDash only, and 10% purchased through both channels.* Among the buyers who purchased on DoorDash, 56% bought exclusively on DoorDash. These results provide proof that DoorDash drives incremental demand ... Early measurement results underscore the value of that incremental reach. A major beverage company saw $0.59 in incremental sales per household through DoorDash, three t | supported with qualification | Keep the first two clauses. Then: 'DoorDash's release attributes "$0.59 in incremental in-store sales per household" for a major beverage company to "a household-level lift study" and calls the comparator "the beverage category benchmark"; Circana's release gives "$0.59 in incremental sales per household through DoorDash, three times the Circana Beverage Benchmark" (no "in-store") as "Early measurement results" in a paragraph that opens "New findings from DoorDash measurement". The release's only footnote (an eight-week US campaign, 11/10/25 to 1/4/2026) is attached to the buyer-overlap shares | Circana's release calls a descriptive buyer-overlap analysis 'proof that DoorDash drives incremental demand'; an overlap share is not a counterfactual and should be coded descriptive if recorded. | RTL-CIRCANA-F11 (evidence/verify/circana.json) |
| C376 | Walmart Connect post of 8 April 2026: '2.5x stronger incremental ROAS and 1.8x higher sales lift than social media industry benchmarks', footnoted to 'Walmart Connects sales lift benchmarks analysis performed by Circana, September 2024–September 2025'; the footnote assigns the benchmark analysis to Circana, and the page does not say Circana measured the campaign itself. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | 'April 08, 2026': "Among Walmart Connect campaigns that qualified for a sales lift study, for example, we saw 2.5x stronger incremental ROAS and 1.8x higher sales lift than social media industry benchmarks. 1" Footnote 1: "Walmart Connects sales lift benchmarks analysis performed by Circana, September 2024–September 2025. Walmart Connect benchmark data includes campaigns that qualified for a sales lift data in either the Food, Consumables, Beverages, or Alcohol categories only. Industry benchmarks are provided by Circana. Circana is an independent third party and not affiliated with Walmart Co | supported with qualification | Restate: 'A Walmart Connect post of 8 April 2026 states that "Among Walmart Connect campaigns that qualified for a sales lift study" it saw "2.5x stronger incremental ROAS and 1.8x higher sales lift than social media industry benchmarks", footnoted to a "Walmart Connects sales lift benchmarks analysis performed by Circana, September 2024–September 2025" limited to Food, Consumables, Beverages or Alcohol campaigns, with "Industry benchmarks ... provided by Circana" and the statement "Circana is an independent third party and not affiliated with Walmart Connect". Circana performed the aggregate | The finding's measurement_record borrows a LiveRamp clean-room join (RTL-E1-F27) for exposure capture; on this page the LiveRamp clean room is tied to Meta closed-loop measurement, not to footnote 1's lift figures. Remove the transfer. 'Circana is an independent third party' is Walmart's assertion; independence of this analysis stays unknown because the payer is not stated. | RTL-CIRCANA-F16 (evidence/verify/circana.json) |
| C377 | Sam's Club Connect's home page footnotes its average sales lift and new-buyer statistics to a 'Circana FY26 report (1,440 campaigns)', while stating that unnamed 'Independent partners validate campaign performance'. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | Home page: "36% 1 Average sales lift", "54% 1 Average new buyers"; SOURCES list item 1: "Circana FY26 report (1,440 campaigns)". The phrase 'Independent partners validate' does not occur in the home page text served on 2026-11-03. | supported with qualification | Replace 'while stating that' with 'and its Measurement page (RTL-U-S020) states that'; the home page itself does not carry that sentence. | RTL-U-S020 (the Measurement page) was not re-opened within budget. | RTL-CIRCANA-F17 (evidence/verify/circana.json) |
| C378 | April 2016 USDA ERS bulletin documents IRI's Consumer Network (2008-12) as 'derived from over 120,000 households', IPF to Census targets, weights 'capped at a value of 20', weighted expenditures 'less than those in other nationally representative datasets', some households 'less likely to report purchases consistently'. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | PDF of 69 pages, created 2016-04-12, 'April 2016'. "The Consumer Network household scanner data are derived from over 120,000 households who report what food products they purchased". "The Consumer Network panel includes over 120,000 households, with 46 to 52 percent of the households providing sufficient purchase data to be included in the static panel used for analyses from 2008 to 2012." "However, total U.S. expenditures reported by households in the weighted IRI data are less than those in other nationally representative datasets." "IRI uses the Iterative Proportional Fitting (IPF) method | supported with qualification | Add after 'over 120,000 households': ', of which 46 to 52 percent provided enough purchase data to enter the static panel used for analyses in 2008-12'. Note that one of the three peer reviewers is ERS staff. | The static share matters wherever the 120,000 is set beside Complete Consumer's '200,000 static households' (capabilities.scale): the two counts are in different units. | RTL-CIRCANA-F21 (evidence/verify/circana.json) |
| C379 | No default named; Household Lift 'household-level A/B testing', 'Enact controlled household tests', assignment unknown, counterfactual a control group asserted with no construction; Market Lift matched market or store, assignment unknown; network-published Circana studies add only 'causal methodology', 'matched buyer groups' and 'household-level lift study'; labels do not fill assignment. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | Page passages as in F01 and F02. Media Lift page, 'For Digital Advertisers and Social Media Brand Managers': "These insights help you determine the sales value of exposed versus engaged users". Joint release (30 September 2025): "With the Household Lift solution, brands can now accurately compare the purchase behavior of consumers exposed to an Ibotta offer versus those who were not". | supported with qualification | Replace 'Circana studies published by networks add only ...' with: 'Circana studies published by networks add "causal methodology" and "matched buyer groups" (Ibotta meta-study) and "household-level lift study" (DoorDash); the Circana-Ibotta joint release of 30 September 2025 says Household Lift compares "consumers exposed to an Ibotta offer versus those who were not", and the Media Lift page offers social advertisers "the sales value of exposed versus engaged users". None states allocation.' For Household Lift, code the counterfactual as 'unexposed comparison group (construction not stated)' | Assignment correctly kept unknown; 'A/B testing', 'controlled household tests', 'thorough matching process' are treated as labels, as required. An exposed-versus-engaged contrast is self-selected and must not be read as a test. | CAP-default_method (evidence/verify/circana.json) |
| C380 | None of the eight parameters documented; 'The only scope statement is that Household Lift reads the products they buy in-store'; Market Lift ROAS and projected sales impact undefined; case metrics relative to undefined benchmarks; '$0.24 ROAS' without basis. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | Introduction: "With Circana’s Media Lift™ solution, you can quantify how media exposure drives consumer behavior — whether through in-person visits, online interactions, or brand engagement." For Manufacturers: "you can optimize campaigns to drive higher-quality website traffic and incremental sales. You can also prove the effectiveness of your email program". Household Lift: "...connecting the people exposed to your media with the products they buy in-store." | supported with qualification | Replace 'The only scope statement is that Household Lift reads "the products they buy in-store"' with: 'Scope statements are loose and inconsistent: Household Lift reads "the products they buy in-store", while the page introduction extends Media Lift to "in-person visits, online interactions, or brand engagement" and the manufacturer block to website traffic and email programmes; none defines a window, touch model, match basis or sales basis.' Add that the two DoorDash releases differ on the $0.59 outcome scope ('incremental in-store sales' at DoorDash; 'incremental sales ... through DoorDash' | The absence of the eight parameters holds. | CAP-attribution_disclosure (evidence/verify/circana.json) |
| C381 | Basis named, size not: 'one of the world’s largest household data sets', '4x ACV', loyalty card, TV, digital household-level exposure, CRM, POS; context: Complete Consumer 200,000 static households, Complete Audiences 96% of US households, Ibotta 'loyalty-card panel data matched against campaign exposure', USDA IRI panel over 120,000 households with IPF (2008-12). | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | All quoted passages match. USDA: "over 120,000 households, with 46 to 52 percent of the households providing sufficient purchase data to be included in the static panel used for analyses from 2008 to 2012." Complete Audiences: "Built from Circana’s industry leading receipt panel" and "loyalty card purchases, viewership, and eating behavior". | supported with qualification | Add the static share to the USDA context ('over 120,000 households, 46 to 52 percent of them in the static panel'), and note that Complete Audiences draws on the receipt panel as well as loyalty-card purchases. | Nothing in the cell is coded beyond the page; no count is transferred to Media Lift. | CAP-identity_basis (evidence/verify/circana.json) |
| C382 | No coverage list; dated Circana measurement on DoorDash (US campaign 2025-11-10 to 2026-01-04, 'household-level lift study'), Ibotta (Household Lift named; meta-study), Walmart Connect ('sales lift benchmarks analysis'), Sam's Club Connect (FY26 report, 1,440 campaigns), Best Buy Ads (likely not Media Lift); count: one network with Media Lift or Household Lift named (Ibotta), three more with Circana lift or benchmark work and the service unnamed. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | As in F11, F12, F15, F16, F17, F18. Joint release: "Circana’s Lift Studies for Ibotta campaigns will begin to roll out in late 2025 and will be available to all brands by early 2026." | supported with qualification | DoorDash: replace 'US campaign 2025-11-10 to 2026-01-04' with 'lift-study period not stated (the footnoted eight-week US campaign, 2025-11-10 to 2026-01-04, is the basis of the buyer-overlap shares)'. Ibotta: add 'Household Lift named for the Ibotta programme in the joint release of 2025-09-30, rollout late 2025, all brands by early 2026'. Walmart Connect: 'Circana-performed sales lift benchmarks analysis of qualifying Walmart Connect campaigns (Food, Consumables, Beverages or Alcohol), 2024-09 to 2025-09, against Circana social benchmarks'. The counts stand. | CAP-placement_coverage (evidence/verify/circana.json) | |
| C383 | Media Lift routes: managed service ('Our team will walk you through'; Mars 'tailored'); network integration (DoorDash to select advertisers; Household Lift on the Ibotta network); no API, self-serve, clean-room or DSP route for Media Lift. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | "Now available to select advertisers for Sponsored Products and Sponsored Brands" (DoorDash). "Circana and Ibotta ... today announced they plan to enter into a strategic partnership"; "Circana’s Lift Studies for Ibotta campaigns will begin to roll out in late 2025 and will be available to all brands by early 2026." | supported with qualification | Cite the joint release for the Ibotta network route (planned partnership announced 30 September 2025; rollout late 2025; all brands by early 2026) instead of resting it on the undated Chomps case alone. Mark 'managed service' as inferred: 'Our team will walk you through' is onboarding language, and only the Mars case ('tailored MediaLift solution') describes an engagement. | The absence of API, self-serve, clean-room and DSP routes for Media Lift holds on the pages read. | CAP-routes (evidence/verify/circana.json) |
| C384 | No qualifying household count for Media Lift; '$5.8T', '4x ACV' are not counts; Complete Consumer '200,000 static households' (2025-02-06); IRI Consumer Network 'over 120,000 households' (2008-12); Complete Audiences 96%; Retail Media '$5T'; Liquid Mix '750,000+ stores and 2,000+ retail partners'; 2023 '500,000+ stores in 20 countries'. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | Every figure matches its page. USDA: the 120,000 is the whole Consumer Network panel, of which '46 to 52 percent' were in the static panel in 2008-12. | supported with qualification | Write the predecessor figure as 'over 120,000 households in the panel, 46 to 52 percent of them static (2008-12)', so it is not read against Complete Consumer's 200,000 static households as like for like. | No count is transferred to Media Lift; the 'no qualifying figure' conclusion holds. | CAP-scale (evidence/verify/circana.json) |
| C385 | Hellman & Friedman's portfolio page records that 'In 2021, H&F acquired The NPD Group' (date 2021-12-15). | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | The page is headed 'News', dated 'December 15th, 2021', and its text, rewritten since, names Circana and the 2022 merger: "In 2021, H&F acquired The NPD Group". | supported with qualification | Date the event '2021 (month not stated)'; 15 December 2021 is the page's post date, not an acquisition or closing date. | The event_type already says the closing date is not stated; the date field should not carry the page date. | TIMELINE-2021-12-15 (evidence/verify/circana.json) |
| C386 | Private; H&F majority investor at the 1 August 2022 completion, Vestar and New Mountain retain significant investments; H&F page records 2021 NPD acquisition and 2022 merger; no filing or later ownership statement found; current split unverified; Vista not named. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | As under F04. | supported with qualification | As under F04: name the April 2024 Vestar continuation-vehicle release as an unopened later ownership statement. | Priority 2: H&F as 'the majority investor in the combined company' and Vestar and NMC retaining 'significant investments' are verbatim in the 1 August 2022 post; H&F's portfolio page confirms 2021 NPD and the 2022 merger; the 7 March 2023 rename is verbatim in H&F's copy; NCS completion 2 June 2025 and Nielsen MMM completion 21 August 2025 are verbatim in Circana's posts. | ENTITY-owner_or_parent (evidence/verify/circana.json) |
| C387 | HQ not stated in the sources opened; releases carry 'CHICAGO' datelines with 'NEW YORK' or 'PORT WASHINGTON, NY'. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | Privacy policy: "The Circana Group ... refers to Circana, LLC ( registered in Delaware, United States) as well as its subsidiaries and affiliates." Datelines as stated. | supported with qualification | Add: 'Circana, LLC is registered in Delaware (privacy policy, RTL-J-S27); headquarters address not stated.' | ENTITY-hq (evidence/verify/circana.json) | |
| C388 | DoorDash beverage campaign '$0.59 in incremental in-store sales per household', 'three times the Circana Beverage Benchmark'; design: 'household-level lift study'; eight-week US campaign 2025-11-10 to 2026-01-04. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | The footnote with the eight-week period is attached ('*') to the buyer-overlap shares in Circana's release; the $0.59 sentence carries no footnote and no period. The quoted pair splices two releases: 'in-store' is DoorDash's wording, 'Circana Beverage Benchmark' is Circana's; DoorDash says 'the beverage category benchmark'. | contradicted | Design field: 'household-level lift study (DoorDash's term); assignment, control and panel undisclosed; study period not stated (the release's eight-week US campaign, 2025-11-10 to 2026-01-04, is the basis of the buyer-overlap shares)'. Quote each release separately. Apply the same edit to the DoorDash row of data table circana_network_footprint (period column) and to rows_at_a_glance or circana_outcome_claims wherever the period is attached to the $0.59. | This is the most consequential error in the file: it dates the only US network lift figure with a period the sources give to a different analysis. | PERF-1 (evidence/verify/circana.json) |
| C389 | Walmart Connect social campaign: '2.5x stronger incremental ROAS and 1.8x higher sales lift than social media industry benchmarks', benchmarks by Circana; limits: 'Circana supplied the benchmark, not (on the page) the campaign measurement'. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | "Among Walmart Connect campaigns that qualified for a sales lift study"; footnote: "Walmart Connects sales lift benchmarks analysis performed by Circana ... Walmart Connect benchmark data includes campaigns that qualified for a sales lift data in either the Food, Consumables, Beverages, or Alcohol categories only. Industry benchmarks are provided by Circana." | contradicted | Claim: 'Walmart Connect campaigns that qualified for a sales lift study (Food, Consumables, Beverages or Alcohol), in aggregate: 2.5x incremental ROAS and 1.8x sales lift versus social media industry benchmarks'. Limits: 'Circana performed the analysis of the Walmart Connect campaign set and supplied the industry benchmarks (2024-09 to 2025-09); who ran the underlying lift studies, whether the campaigns were social, and who paid are not stated.' | Not a single campaign, and Circana's role on the page is wider than 'benchmark only'. | PERF-5 (evidence/verify/circana.json) |
| C390 | Offers retail media networks measurement of sales 'beyond your network' and claims store POS from '750,000+ stores and 2,000+ retail partners', a cross-retailer scope that no single network's own reporting claims. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | NCS: "Retail media networks, gauge the true reach of your retail media network campaign beyond your network for comprehensive sales analysis." Liquid Mix: "store-level POS data from 750,000+ stores and 2,000+ retail partners". | supported with qualification | Name the services: 'beyond your network' is NCS Sales Effect's; the 750,000+ stores is Liquid Mix's POS base. Replace 'that no single network's own reporting claims' with a bounded statement ('that none of the network pages read for this profile claims') or mark it as inference. | Unbounded absence claim (hard rule 1). | DIFF-3 (evidence/verify/circana.json) |
| C391 | Is the third party named in network releases as validator (DoorDash's 'independent, third-party validation'; Ibotta's 'independent measurement partners like Circana') [RTL-U-S212, RTL-CIRCANA-S08]. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | RTL-CIRCANA-S08 is Circana's own case page and does not contain 'independent measurement partners like Circana' (zero matches in the raw HTML); its wording is 'using Circana Household Lift for independent measurement of incremental impact'. In the record the quoted phrase belongs to RTL-E3-S21 (an Ibotta proof page, per evidence/research/E3.json). RTL-U-S212 is the joint release on circana.com, so 'independent, third-party validation' is Circana's and DoorDash's joint wording; DoorDash's own release says 'independent, household-level validation'. | contradicted | Cite RTL-E3-S21 for Ibotta's 'independent measurement partners like Circana' and drop RTL-CIRCANA-S08 from that clause (or quote S08's own wording as Circana's); describe RTL-U-S212 as a joint Circana-DoorDash release. | Mis-citation: a Circana-published page is presented as the network's word. | DIFF-4 (evidence/verify/circana.json) |
| C392 | Integration: Circana's pages already disagree on simple facts (benchmarks 'over 6,000' versus 'More than 10,000'; 'Media Lift' versus 'MediaLift'). | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | Media Lift page: "over 6,000 industry benchmarks"; Marketing Mix page: "More than 10,000 industry benchmarks"; Mars case: "MediaLift". | supported with qualification | Keep the benchmark discrepancy; replace the 'Media Lift' versus 'MediaLift' example (a spelling variant, not a factual disagreement) with the client-count drift ('more than 7,000 clients' in 2022, 'almost 7,000 brands and retailers' in 2023, 'more than 7,000 clients and partners' in 2026). | RISK-3 (evidence/verify/circana.json) | |
| C393 | Named as measurer or validator by DoorDash, Ibotta, Walmart Connect, Sam's Club Connect and Best Buy Ads on pages current in 2026. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | Each page names Circana (DoorDash partnership; Ibotta meta-study; Walmart footnote 'analysis performed by Circana'; Sam's footnote 'Circana FY26 report'; Best Buy 'Circana conducted a first-of-its-kind study'). | supported with qualification | 'Named as measurer, validator or benchmark analyst by ...' (Walmart Connect names Circana for a benchmark analysis). | SWOT-strengths-2 (evidence/verify/circana.json) | |
| C394 | Networks that sell the media publish its studies (DoorDash, Ibotta, Sam's Club Connect) without naming the payer or the design. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | No payer is named on any of the three. Design wording: DoorDash 'household-level lift study'; Ibotta 'causal methodology', 'matched buyer groups'; Sam's Club none. | supported with qualification | '... without naming the payer, and with design labels at most ("household-level lift study", "causal methodology", "matched buyer groups") and no allocation.' Circana itself co-published the DoorDash release. | SWOT-weaknesses-4 (evidence/verify/circana.json) | |
| C395 | Networks now buy outside validation of their own incrementality (DoorDash, 2 September 2026); a published Household Lift method would let buyers judge it. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | Neither release says who pays; they announce 'a partnership to bring independent, third-party validation to DoorDash's measurement capabilities', 'Now available to select advertisers'. | contradicted | 'Networks now offer outside validation of their own incrementality to selected advertisers (DoorDash, 2 September 2026; payer not stated); a published Household Lift method would let buyers judge it.' | 'Buy' asserts a payer that the profile itself codes as unknown (performance_evidence item 1, F11). | SWOT-opportunities-1 (evidence/verify/circana.json) |
| C396 | MRC's digital under-review list carries Kroger's and Kevel's 2026 pre-audits with incremental-sales metrics but no sales-lift measurer: room for a first audited lift method. | Refute-mode verifier for evidence/profiles/circana.json re-opened the cited source. | Kroger (Pre-audit in process as of 6/2026): "Impressions, Clicks, Viewable Impressions, Attributed sales, Incremental Sales, aROAS, iROAS". Kevel (Pre-audit reviewed 9/2026): "...Clicks, Conversions and incrementality". | supported with qualification | '... carries Kroger's pre-audit (Incremental Sales, iROAS) and Kevel's ("Conversions and incrementality") ...' Kevel's scope names incrementality, not incremental sales. | No third-party sales-lift measurer is on the list, as stated. | SWOT-opportunities-2 (evidence/verify/circana.json) |
| C397 | Criteo S.A. completed its redomiciliation from France to Luxembourg by 29 July 2026 and on 5 August 2026 entered a Merger Agreement to merge into Criteo Holdings, Inc. (Delaware) at 12:00:01 a.m. New York time on 1 January 2027 unless the boards agree otherwise, subject to shareholder approval, an effective Form S-4 and a US listing; pending on 3 November 2026. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Release of 29 July 2026 (criteo.com PDF; dateline 'NEW YORK, July 29, 2026'): completion of the transfer of legal domicile from France to Luxembourg via a cross-border conversion, 'effective July 29, 2026'; the ADS program was terminated (one ordinary share per ADS) and ordinary shares trade directly on Nasdaq; the listing is expected to move to the NYSE on completion of the U.S. Merger. 8-K of 5 August 2026, Item 1.01: Merger Agreement with 'Criteo Holdings, Inc., a Delaware corporation and wholly owned subsidiary of Lux Criteo'; Effective Time 'at 12:00:01 a.m., New York City time, on Januar | supported with qualification | Criteo S.A. completed its redomiciliation from France to Luxembourg by cross-border conversion effective 29 July 2026 (the ADS program was terminated and ordinary shares began trading directly on Nasdaq); on 5 August 2026 it entered a Merger Agreement to merge into its wholly owned Delaware subsidiary Criteo Holdings, Inc. at 12:00:01 a.m. New York time on 1 January 2027 unless the boards agree otherwise, subject to shareholder approval, an effective Form S-4 and a U.S. listing (Criteo expects to move to the NYSE); on 28-29 October 2026 the S-4 had not been declared effective and the vote had | The completion date rested on a release title (RTL-CRITEO-S09); the release body is now read (RTL-VCRITEO-S04). | RTL-CRITEO-F01 (evidence/verify/criteo.json) |
| C398 | Commerce Yield's help centre documents the identity basis as four signals the retailer passes through Criteo's API, matched in the order Customer ID, Hashed Email, Retailer Visitor ID, Criteo ID; missing signals undercount attributed sales; no match rate or coverage report; the identity graph's linking method is not stated. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Identity note (first and last published 2026-08-11): Customer ID and Retailer Visitor ID 'Required', Hashed Email 'Recommended', Criteo ID set by Criteo '(automatic)' and 'managed automatically by Criteo'; signals evaluated 'in parallel' and Criteo 'keeps the best match' in the order Customer ID, Hashed Email, Retailer Visitor ID, Criteo ID; Criteo ID 'dependent on third-party cookie availability'; 'Attributed sales are undercounted' without customerId; no match rate published. About Attribution (first published 2024-07-24, last 2025-12-17): identifiers are the retailer's CRM ID, the retailer' | supported with qualification | Commerce Yield's help centre documents the identity basis of attribution as three signals the retailer passes through the API (Customer ID and Retailer Visitor ID required, Hashed Email recommended) plus a Criteo ID set automatically by Criteo's tag; all are evaluated in parallel and the best match is kept in the order Customer ID, Hashed Email, Retailer Visitor ID, Criteo ID; Commerce Yield's About Attribution article states that user matching is purely deterministic, with no IP-address or probabilistic matching; missing signals undercount attributed sales; no match rate or coverage report is | The CY pages list the identifiers differently (identity note: Customer ID, Hashed Email, Retailer Visitor ID, Criteo ID; About Attribution: CRM ID, first-party cookie, Criteo cookie ID, Criteo X-device ID) and state the best-match rule differently (reliability order vs 'typically the most recent one'); record as a tension. The 'Identity method' tension in counterevidence_and_tensions is resolved f | RTL-CRITEO-F07 (evidence/verify/criteo.json) |
| C399 | In the screened record Commerce Yield documents attribution only: no control design is documented for retailers' advertiser campaigns; none of the 209 English help-centre titles concerns incrementality, lift, holdout, experiment or validation; Criteo's incrementality texts belong to Commerce Growth and open-web ghost bidding. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Help-centre home page (embedded state): no title or folder name contains 'incremental', 'lift', 'holdout', 'experiment' or 'validat' (verifier screen of 502 unique title and name strings, including Japanese and Korean versions and folder names; 'Understand Sales Attribution' is a folder holding 'About Attribution'). The CY product page describes no lift or test-and-control method. Offline Sales Reporting lists, among analytics-tab reports that exclude offline sales, '(Offsite activity + Campaign Summary + A/B Test Results)'. | supported with qualification | Add: Commerce Yield's analytics tab includes an 'A/B Test Results' report (named on the Offline Sales Reporting page) whose design (what is randomised, unit, outcome, whether a holdout) is not documented on any page opened; the absence of a documented control design for advertisers' campaigns stands, bounded to the pages opened, and the report should be read before the default_method cell is scored. | RTL-CRITEO-S19 and RTL-G1-S08 were not re-opened (budget). The 'A/B Test Results' report may be a creative or bid test rather than a lift test; it must not be scored as a control design until its page is read. | RTL-CRITEO-F08 (evidence/verify/criteo.json) |
| C400 | Criteo's own dated statements name seven US retailers for its retailer-side business in 2024-2025: Albertsons Media Collective, Dollar General and Belk (Commerce Yield named) and BJ's Wholesale Club, Weis Markets, Winn-Dixie and Harveys Supermarkets (Q2 2025 release, no product named); the 10-K counts approximately 235 retailers. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Albertsons ('NEW YORK, Jan. 9, 2024'): 'Criteo's retailer monetization suite, Commerce Yield, will allow for Albertsons to continue to monetize its online assets'; supports 'onsite sponsored ad offerings'; no go-live date; no geography stated for the deployment. Dollar General ('NEW YORK, May 31, 2024'): 'The partnership supports Dollar General's onsite sponsored ad offerings'; brands 'directly access DGMN's inventory within their private market via Commerce Yield' (the tool truncated the rest of the sentence); no go-live date. Belk ('NEW YORK, Oct. 8, 2024'): 'Belk is harnessing Criteo's reta | supported with qualification | Keep 7 names with 3 naming Commerce Yield, and add: the Albertsons and Dollar General releases are selection announcements with no go-live date and no stated deployment geography (US location rests on the companies' identity); Harveys was added 'via our digital commerce partner Mercatus in the U.S.' (an intermediated integration; wording now returned verbatim); the Q2 2025 release names no product, so BJ's, Weis, Winn-Dixie and Harveys are Retail Media additions, not documented Commerce Yield deployments. | The exact Dollar General phrase 'Commerce Yield, Criteo's retailer monetization platform' was not returned in full by the tool; Commerce Yield is named in the release. | RTL-CRITEO-F10 (evidence/verify/criteo.json) |
| C401 | Criteo's audited Retail Media segment revenue was US$209.0m (2023), US$258.3m (2024), US$263.9m (2025), recognised primarily net as agent, with segment TAC of US$5.5m, US$4.5m and US$4.2m; growth 'in particular in the U.S. market'; about 33% of Retail Media revenue from agencies. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | XBRL R60 ('$ in Thousands'): Retail Media Revenue 263,872 / 258,303 / 209,007; Traffic acquisition costs 4,188 / 4,457 / 5,547; Contribution ex-TAC 259,684 / 253,846 / 203,460 (2025 / 2024 / 2023). 10-K Item 1: 'approximately 33% of our Retail Media revenue comes from agencies.' Q4 2025 release, fiscal 2025: Retail Media revenue increased 2%, 'driven by continued strength in Retail Media onsite, new client integrations and growing network effects' (no U.S. wording). | supported with qualification | Figures, TAC and the agency share are verified. Mark the three MD&A quotations (net-basis recognition, the segment definition, 'in particular in the U.S. market') as not re-verified by the verifier: the fetch tool returns only the first 161,489 characters of the 10-K (Item 1 and part of Item 1A). Keep them pending a browser read. | See needs_browser. | RTL-CRITEO-F11 (evidence/verify/criteo.json) |
| C402 | Inference: the largest client was 5% of 2025 revenue (about US$97m of US$1,944.9m); if all Retail Media, about 37% of the segment; the Q2 2026 headwind equals about 34% of Q2 2025 Retail Media revenue. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | 10-K Item 1: 'our largest client represented 5% and 5% of our revenue, respectively, and our largest 10 clients represented 19% and 17%' (2025, 2024). R60 total revenue 1,944,901 (US$ thousand). Q2 2026 release: '$21 million headwind' stated against both the Retail Media revenue and Contribution ex-TAC declines; Q2 2025 Retail Media revenue 60,913. | supported with qualification | State the inference as a range, since 5% is rounded: about US$88-107 million (4.5%-5.5% of US$1,944.9 million), roughly 33%-41% of 2025 Retail Media revenue if it were all Retail Media. 21,000 / 60,913 = 34.5% is correct arithmetic, but the release does not say whether the US$21 million is measured in revenue or Contribution ex-TAC (both fell 21%; the two series differ by under US$1 million a quarter). | The goodwill-test quotation ('reduction in services from our largest customer') is in MD&A and was not re-read (tool truncation); R74 confirms Retail Media goodwill of 151,828 and no impairment row. | RTL-CRITEO-F13 (evidence/verify/criteo.json) |
| C403 | Commerce Yield's methodology note documents a correction route for delivery data (backward-looking IVT analysis, notification, reprocessing to correct billing and reporting metrics); with API versioning and retroactive attribution changes that is the governance shown; no dispute procedure or methodology version history found. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Methodology note (first published 2026-07-13, last 2026-08-11): 'If a new type of invalid traffic is discovered, Criteo conducts a backward-looking analysis'; under 'Data Logging and Reissuance': if Criteo 'detects issues that affect advertiser spend or data', impacted advertisers are notified and 'data can be reprocessed to correct billing and reporting metrics'; 'Filtering rules are reviewed quarterly'. About Attribution: settings changes make platform data and dashboards update retroactively. | supported with qualification | Qualify 'covers delivery and billing metrics': the reissuance sentence sits in a section on delivery-server logging but speaks of issues affecting 'advertiser spend or data' and of 'billing and reporting metrics' without excluding attributed metrics; whether it covers attributed sales is not stated either way. | RTL-F-S03 (API versioning) not re-opened. | RTL-CRITEO-F16 (evidence/verify/criteo.json) |
| C404 | Attribution only in the screened record: Commerce Yield's help centre documents attribution settings and identity matching and no control design for advertisers' campaigns; no help-centre title concerns incrementality, lift, holdout, experiment or validation; 'closed-loop reporting' has no counterfactual; Criteo's incrementality texts belong to other products. Grade C. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | No CY page opened (settings, identity, methodology, About Attribution, Demand-Side Metrics, Offline Sales Reporting, product page) describes a holdout, ghost-ad, PSA, geo or matched-control design. Offline Sales Reporting names an 'A/B Test Results' report in the CY analytics tab without describing it. | supported with qualification | Add the 'A/B Test Results' report as an unread lead and keep the cell's absence bounded to the pages opened; read that report's documentation before scoring dimension 1. | All sources in this cell are Commerce Yield pages or Criteo pages for other products used only to show non-transfer; no Commerce Max help page carries it. | CAP-CommerceYield-default_method (evidence/verify/criteo.json) |
| C405 | On Commerce Yield's own pages: window documented with defaults; touch model delegated to an unopened 'About Attribution' article; match basis documented; in-store asserted, 'Offline Sales Reporting' not opened; new-to-brand lookback not found; sales basis not found; deduplication not found; halo documented. Own-row count: 3 of 8 (window, match basis, halo). | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | About Attribution (help.cyield.criteo.com; first published 2024-07-24, last 2025-12-17): 'Sales deduplication means that a sales conversion is attributed once to a single advertising event'; prioritisation 'Clicks take precedence over impressions', 'Same SKU events take precedence over Same Category events', then the event 'closest to the time of purchase'; other events are 'assisted sales'; user matching 'purely deterministic'; 'Some retail partners also have the ability to correlate advertising spend to in-store purchases', matched by 'a combination of email address collection and the retail | contradicted | Rewrite the cell from Commerce Yield's own pages: (1) window: as stated (S22); (2) touch model: one 'most relevant' event per sale, clicks over impressions, same SKU over same category, then recency; other events recorded as assisted sales (About Attribution); (3) match basis: deterministic user matching on retailer CRM ID, first-party cookie, hashed email and Criteo IDs, no IP or probabilistic matching (About Attribution; S23); (4) in-store: retailer-supplied daily store-sales files matched on hashed email, customer ID or visitor ID and SKU under the campaign's window and product match, for o | These are retailer-side Commerce Yield pages (help.cyield.criteo.com), so they carry the Commerce Yield row directly; no transfer from Commerce Max is involved. Some wording mirrors Commerce Max's help centre (the CY Demand-Side Metrics page mentions viewing assisted sales 'in Commerce Max'), which is a note on provenance, not a row transfer. | CAP-CommerceYield-attribution_disclosure (evidence/verify/criteo.json) |
| C406 | Documented routes are operator-run reporting (in-platform, email, API Reporting Endpoints; Retail Media API demand-side analytics with 'attributed transaction logs' of undocumented contents); no buyer-run query environment, clean room or named third party with outcome-data access; Snowflake offline-sales app listed by title only. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Offline Sales Reporting: offline attribution results appear 'within the Commerce Yield Demand - Brand Activity Dashboard and Attributed Transaction log'; 'If shared by the Retailer, Brands will have access to the Offline Sales in the Analytics tab of their Commerce Max Demand Accounts'. About Attribution: 'Pulling the Attributed Transaction Log report on the demand-side platform will provide visibility on all transactions attributed to a campaign.' CY product page: 'API Reporting Endpoints', 'Automated Email Reports'. | supported with qualification | Cite Commerce Yield's own pages for the attributed transaction log (Offline Sales Reporting; About Attribution) instead of, or beside, RTL-F-S45; add that the retailer controls whether brands see offline sales in their demand accounts. The rest stands: no buyer-run environment or clean room found. | RTL-F-S45 is the Retail Media API's demand-side (DSP) analytics page; for the Commerce Yield row it is inference unless a CY page names the same report, which CY pages now do. | CAP-CommerceYield-buyer_analysis (evidence/verify/criteo.json) |
| C407 | Four signals passed by the retailer through Criteo's API (Customer ID required, Retailer Visitor ID required, Hashed Email recommended, Criteo ID via Criteo's identity graph), matched best-first; in-store via unopened articles; no coverage reporting; the identity graph's linking method is not stated. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Criteo ID is 'Set by' Criteo '(automatic)', not passed by the retailer; signals are evaluated in parallel and the best match kept by the stated order. About Attribution: user matching rules 'purely deterministic', no IP addresses or probabilistic models; identifiers include 'The Criteo X-device ID'. Offline Sales Reporting: in-store ingestion by SFTP file with at least one of customer_id, retailer_visitor_id or SHA-256 hashed email, sent in the file and in all ad calls; FAQ: '80-90% of transactions occur in stores' and under 15% of in-store transactions are cash that cannot be reconciled to an | supported with qualification | Three retailer-passed signals plus a Criteo-set ID; user matching declared deterministic (About Attribution); in-store channel documented (Offline Sales Reporting: daily files, three accepted identifiers, 2-day submission limit, about 74 hours to UI, onsite campaigns only). Coverage reporting: still none documented. Delete 'The identity graph's linking method is not stated' or restate as 'the graph's linking method is described only as deterministic'. | All sources are Commerce Yield pages. | CAP-CommerceYield-identity_basis (evidence/verify/criteo.json) |
| C408 | On-site (Sponsored Products CPC; display and video CPM) and off-site (open web); no in-store placements; on-site measurement documented (render-start impression, In-Unit Click, IVT rules, MRC on-site); off-site follows Criteo's standard methodology without MRC IVT filtering; one campaign-level attribution set for both classes. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Methodology note: impression counted when a placement or product 'is loaded and begins to render'; 'In-Unit Click' methodology; 'Impressions and clicks from offsite campaigns follow Criteo's standard measurement methodology, but MRC IVT filtering does not apply.' Offline Sales Reporting: offline attribution 'currently available for Onsite Display and Onsite Sponsored Product campaigns'; offsite reports exclude offline sales; offline attribution to offsite campaigns 'is not available today'. | supported with qualification | Add: in-store sales are attributed only to on-site display and sponsored-product campaigns; off-site campaigns get online-only attribution. The 'one rule set for both classes' holds for windows and product match, not for the outcome data used. | CAP-CommerceYield-placement_coverage (evidence/verify/criteo.json) | |
| C409 | Route families: API (Retail Media API, API Reporting Endpoints), self-serve console, owned DSP (Commerce Max), 60+ partner DSPs, agency/platform integrations (14 API buying partners, Google SA360), managed service; no clean-room route; none withdrawn. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | CY page: 'API Reporting Endpoints', '60+ third-party DSP partnerships', white-label private market, Professional Services. Albertsons: 'Both Commerce Max and Commerce Yield will be available as self-service solutions'. 10-K Item 1: 'the world's largest independent Retail Media API program, connecting with 14 leading buying partners'. Q3 2025: 'Criteo was named Google's first onsite Retail Media partner', campaigns through 'Google Search Ads 360'. | supported with qualification | Qualify route (5): the 14-partner API program and Search Ads 360 access are company-level demand routes into Criteo's retailer network (10-K; Q3 2025 release), not Commerce Yield documentation; they show demand reaching CY inventory, not a CY-specific integration. | RTL-F-S01 and RTL-F-S03 were not re-opened. | CAP-CommerceYield-routes (evidence/verify/criteo.json) |
| C410 | Named US retailers in Criteo's dated statements: Albertsons (9 Jan 2024), Dollar General (31 May 2024), Belk (8 Oct 2024) with Commerce Yield named; BJ's, Weis, Winn-Dixie, Harveys (30 Jul 2025, product not named): 7 names, 3 naming Commerce Yield; 'approximately 235 retailers' worldwide; continuing production undocumented. Grade D. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | As in RTL-CRITEO-F10. Albertsons and Dollar General: selection announcements, no go-live date. Belk: 'Belk Media Network launches'. Q2 2025: Harveys added 'via our digital commerce partner Mercatus in the U.S.'. CY page logos: Costco, FreshDirect, Albertsons; video titles Uber, CVS Media Exchange, Costco (undated). | supported with qualification | Keep 7 names and 3 naming Commerce Yield (the cohort C dimension-8 count). Add FreshDirect (undated logo on the CY page) to the 'not counted' list, and note that Harveys came through Mercatus, an intermediary. | Every name rests on Criteo's own releases (vendor); none is confirmed by the retailer, as the cell says. | CAP-CommerceYield-scale (evidence/verify/criteo.json) |
| C411 | MRC (15 March 2024): initial accreditation of the RMP for display impressions and clicks gross and net of GIVT, plus invalid counts, D/MW/MA, for on-site sponsored, commerce display and standard banner ads; mapped to Commerce Yield by Criteo; exclusions: off-site, onsite video, outcome metrics; MRC list 'Clicks: Display' and 'Rendered Ad Impressions: Display'; TAG claims company-wide. Grade A. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Letter and list as in RTL-CRITEO-F04; Criteo's release maps the accreditation to 'Commerce Max and Commerce Yield' for 'Onsite Sponsored Products and Onsite Display Ads'; methodology note addresses 'Retailers managing onsite inventory through Commerce Yield' and excludes offsite. The MRC list links only the 2024 initial letter. | supported with qualification | Add: GIVT only (no sophisticated invalid traffic accreditation); the only letter on MRC's public list is the 2024 initial accreditation, with no continuation letter linked as of 3 November 2026 (not evidence of a lapse; the list still carries Criteo); Criteo's 2024 release describes the accreditation as assurance about 'commerce outcomes', which the letter does not cover. | The cell rests on the accreditor's record plus Criteo's retailer-side mapping, as required; no Commerce Max page carries it. | CAP-CommerceYield-assurance (evidence/verify/criteo.json) |
| C412 | Bounded screen: API versioning with changelogs; 'What's New on Commerce Yield 2026' not opened; correction policy for delivery and billing data (S21); attribution changes retroactive (S22); no dated revision log; no dispute procedure; methodology note has publication metadata but no version history. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | As in RTL-CRITEO-F16. About Attribution adds that platform data and dashboards update retroactively after a settings change. Offline Sales Reporting: late submissions rejected, 'Delays require extensive recalculations'. | supported with qualification | Apply the RTL-CRITEO-F16 qualification on the scope of the reissuance sentence ('billing and reporting metrics', not limited to delivery in its wording). | RTL-F-S03 not re-opened. | CAP-CommerceYield-governance (evidence/verify/criteo.json) |
| C413 | Product-specific record: CY page commitments (consent, Co-Op opt-out, no resale); a Commerce Max help-centre FAQ on TCFv2 and GDPR (another row's page); company policy on sensitive data; company-level state-law routes; adverse events CNIL, GoodRx, CVS. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | CY page commitments verified (see RTL-CRITEO-F14). R28 legal matters verified (see RTL-CRITEO-F15). CY help centre lists its own 'Privacy FAQs' (guide id fqoNE5uQaG, update date 2025-10-31), not opened by the research or this check. | supported with qualification | Before scoring, replace the Commerce Max FAQ (RTL-CRITEO-S18) with Commerce Yield's own Privacy FAQs (help.cyield.criteo.com guide fqoNE5uQaG); until read, the TCFv2, retention and GDPR statements count for the Commerce Yield row only as inference. | RTL-CRITEO-S18 is a Commerce Max help-centre page; the cell flags it correctly, but it must not raise the Commerce Yield score. | CAP-CommerceYield-privacy (evidence/verify/criteo.json) |
| C414 | Self-service console; API (Retail Media API DSP analytics; 14-partner API program); agency and platform integrations (Google SA360; Microsoft demand integration announced for H2 2024; dentsu partnership title). | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Albertsons and Dollar General releases name Commerce Max as Criteo's 'self-service demand-side platform (DSP)'; 10-K: 14 buying partners; Q3 2025: Google Search Ads 360; Microsoft release: demand integration 'expected to roll out in the second half of 2024'. | supported with qualification | Record the Microsoft demand integration as announced (July 2024) with completion undocumented; it cannot count as a live route on the opened record. | RTL-F-S45, RTL-CRITEO-S16 and the dentsu title were not re-opened. | PCAP-CommerceMax-routes (evidence/verify/criteo.json) |
| C415 | US availability: Commerce Max self-service at Albertsons and Dollar General (2024), closed-loop reporting via Commerce Max at Belk (2024), Best Buy, Macy's and Shipt enrolled at GA (2023), a June 2026 report available to US & Canada users only. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Albertsons and Dollar General name Commerce Max as the self-service DSP; Belk: 'closed-loop reporting' 'via Criteo's demand-side platform, Commerce Max'. Dollar General adds an 'in-store sales integration within Commerce Max'. | supported with qualification | Verified for the three 2024 releases; the 2023 GA enrolment (RTL-CRITEO-S16) and the June 2026 release note (RTL-G1-S31) were not re-opened. | PCAP-CommerceMax-us_availability (evidence/verify/criteo.json) | |
| C416 | Commerce Yield attributed sales and ROAS record: windows post-click default 30 d, post-view default 1 d; product-match halo; touch_model 'unknown'; sales_basis 'unknown'; deduplication 'unknown'; identity-graph component unstated. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | About Attribution and Demand-Side Metrics document the touch model (one most relevant event; clicks over impressions; SKU over category; recency) and deduplication (one line item per sale; onsite and offsite deduplicated); matching declared 'purely deterministic'; ROAS = Attributed Sales / Spend on CY's own metrics page. | supported with qualification | Set touch_model to 'single most-relevant event: click > impression, same SKU > same category, then most recent; others recorded as assisted sales'; deduplication to 'one line item per sale across campaigns; onsite and offsite deduplicated'; observed_vs_modelled to 'observed deterministic matches (vendor statement)'; outcome_unit to cite CY's Demand-Side Metrics for ROAS instead of the Commerce Max definition. sales_basis stays 'unknown'. | MR-CommerceYield-1 (evidence/verify/criteo.json) | |
| C417 | Commerce Yield in-store (offline) attributed sales record: estimation_method, identity_linkage, observed_vs_modelled, access_restrictions, attribution_window, touch_model, sales_basis, deduplication and halo all 'unknown'; route listed but not opened. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Offline Sales Reporting: retailer uploads daily store-sales files by SFTP (identifiers, order ID, timestamp, SKU, unit price, quantity, store ID) within 2 days; Criteo matches user ID and SKU to prior ad events; attributed hourly; the campaign's product match and lookback windows apply ('Attribution rules apply equally to both offline and online orders'); onsite display and sponsored products only; results in the CY Brand Activity Dashboard and Attributed Transaction log, and in brands' Commerce Max demand accounts if the retailer shares them; about 74 hours to the UI. About Attribution: offli | contradicted | Fill the record: estimation_method 'deterministic user-ID and SKU match of retailer store-sales files to prior ad events within the campaign window'; identity_linkage 'hashed email, retailer customer ID or visitor ID'; observed_vs_modelled 'observed matches'; attribution_window and halo 'as set per campaign (same as online)'; touch_model and deduplication as in MR-CommerceYield-1; access_restrictions 'CY dashboards and Attributed Transaction log; brands see offline sales in Commerce Max only if the retailer shares'; scope 'onsite display and sponsored products only'. sales_basis stays 'unknown | The research marked these fields unknown because the article was not opened; the article is a Commerce Yield page and settles them. | MR-CommerceYield-2 (evidence/verify/criteo.json) |
| C418 | Headwind from scope changes with two Retail Media clients: 21,000 (US$ thousand, rounded); issuer's estimate of the quarter's revenue impact. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | 'Retail Media revenue decreased (21)%, or (22)% at constant currency, and Retail Media Contribution ex-TAC decreased (21)%, or (22)% at constant currency, reflecting a $21 million headwind from previously communicated scope changes with two specific Retail Media clients'; 'Excluding this impact, Contribution ex-TAC grew 20% in Q2 across the underlying client base.' | supported with qualification | Definition: 'Issuer's own estimate of the Q2 2026 year-over-year headwind attributed to the two clients' scope changes, stated against the declines in both Retail Media revenue and Contribution ex-TAC (measure not specified); clients unnamed; not an accounting line.' Add the issuer's ex-headwind Contribution ex-TAC growth of 20%. | Also affects supply_and_data_dependencies ('took about US$21 million from Q2 2026 segment revenue') and RTL-CRITEO-F13. | FIN-Headwind-Q2-2026 (evidence/verify/criteo.json) |
| C419 | Goodwill allocated to Retail Media 151,828 (US$ thousand) at 31 December 2025; no impairment; the test reflected 'the reduction in services from our largest customer'. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | R74 (Goodwill details): Retail Media goodwill 144,961 at the start of 2025, currency translation 6,867, 151,828 at the end of 2025; no impairment line for either year; 2 reportable segments in 2024 and 2025. | supported with qualification | Amount and absence of impairment verified; the quotation about the largest customer is in MD&A and was not re-read (tool truncation). The 2025 increase is currency translation only. | See needs_browser. | FIN-Goodwill-RetailMedia-2025 (evidence/verify/criteo.json) |
| C420 | Criteo and Google announce an onsite retail media integration; Q3 2025 release says Criteo was named Google's first onsite Retail Media partner via Search Ads 360. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Q3 2025 release: 'Criteo was named Google's first onsite Retail Media partner', campaigns through 'Google Search Ads 360'. | supported with qualification | The 10 September 2025 date rests on a release-list title (RTL-CRITEO-S09) not re-opened; the Q3 2025 release gives no date. | TL-2025-09-10 (evidence/verify/criteo.json) | |
| C421 | Redomiciliation from France to Luxembourg completed (release title; confirmed in the 5 August 2026 release). | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Release of 29 July 2026: conversion 'effective July 29, 2026'; ADS program terminated; ordinary shares trade directly on Nasdaq; listing to move to the NYSE on completion of the U.S. Merger. | supported with qualification | Source the event to the release body (RTL-VCRITEO-S04), not the title; add the ADS termination and the planned NYSE move. | TL-2026-07-29 (evidence/verify/criteo.json) | |
| C422 | Nasdaq Global Select Market, ticker CRTO (American Depositary Shares since 30 October 2013; ordinary shares per the 5 August 2026 8-K cover). | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | 8-K of 5 August 2026 cover: 'Ordinary Shares, nominal value €0.025 per share', 'CRTO', 'Nasdaq Global Select Market'. Release of 29 July 2026: ADS program terminated at the conversion; ordinary shares trade directly on Nasdaq from 29 July 2026; listing expected to move to the NYSE on completion of the U.S. Merger. | supported with qualification | Nasdaq Global Select Market, ticker CRTO: ADSs from 30 October 2013 until the ADS program was terminated on 29 July 2026, ordinary shares directly since then; Criteo expects to move the listing to the NYSE when the U.S. Merger completes. | The 2013 date was not re-verified (see TL-2013-10-30). | ENTITY-listing (evidence/verify/criteo.json) |
| C423 | Retailers pay a negotiated SSP fee and sometimes a technology fee, brands a negotiated DSP fee; revenue recognised primarily net as agent; about 33% of Retail Media revenue from agencies. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | 10-K Item 1: 'We typically charge retailers a negotiated SSP fee and sometimes a technology fee, while brands pay us a negotiated DSP fee.'; 'approximately 33% of our Retail Media revenue comes from agencies.' | supported with qualification | Fee and agency-share sentences verified; the net-basis sentence (MD&A) not re-read. | See needs_browser. | BUSINESS-how_it_earns-who_buys (evidence/verify/criteo.json) |
| C424 | MRC accreditation since 15 March 2024 for on-site display impressions and clicks, including invalid traffic, across desktop, mobile web and in-app. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Letter: 'Display Impressions and Clicks reported Gross and Net of General Invalid Traffic (GIVT) within Desktop, Mobile Web and Mobile In-App', 'specifically for On-site Sponsored Ads, Commerce Display Ads and Standard Banner Ads'; metrics include Invalid Impressions and Invalid Clicks. | supported with qualification | MRC accreditation since 15 March 2024 for on-site display impressions and clicks, gross and net of general invalid traffic (GIVT), across desktop, mobile web and in-app, for on-site sponsored, commerce display and standard banner ads. | 'including invalid traffic' should say general invalid traffic; sophisticated invalid traffic is not accredited. | SWOT-strengths-1 (evidence/verify/criteo.json) |
| C425 | No control design is documented for retailers' advertiser campaigns; 'closed-loop reporting' is attribution without a counterfactual. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | No CY page opened documents a control design; the CY page uses 'Closed-loop reporting' for attribution. Offline Sales Reporting names an 'A/B Test Results' report in the CY analytics tab, undocumented. | supported with qualification | No control design is documented, in the pages opened, for retailers' advertiser campaigns (an 'A/B Test Results' report is named but its design is not described); 'closed-loop reporting' is attribution without a counterfactual. | SWOT-weaknesses-1 (evidence/verify/criteo.json) | |
| C426 | No match rate, coverage report or sales basis is published, and changed attribution settings rewrite a campaign's whole history. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | No match rate or coverage report on the CY pages opened; Demand-Side Metrics defines Attributed Sales as 'Sum of all Sales' with no rule on returns, cancellations, taxes or shipping; settings changes 'will be retroactively applied to the full extent of the campaign' and 'all data and dashboards in the platform will update retroactively'. | supported with qualification | No match rate, coverage report or sales basis is published in the Commerce Yield pages opened, and changed attribution settings are applied retroactively to a campaign's whole reported history. | Bound the absence to the pages opened; the Offline Sales page speaks of 'match rate' only as advice on hashing. | SWOT-weaknesses-3 (evidence/verify/criteo.json) |
| C427 | Microsoft's planned preferred-partner role and Google's onsite integration could route new demand and retailers to the platform. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Microsoft (11 July 2024): intends to work with Criteo as preferred onsite partner, rollout 'expected' in H2 2024; Q3 2025: Criteo named Google's first onsite Retail Media partner via Search Ads 360. | supported with qualification | Microsoft's preferred-partner plan (announced July 2024, completion undocumented) and Google's Search Ads 360 integration could route new demand and retailers to the platform. | Labelled inference; the Microsoft plan is more than two years old with no documented completion. | SWOT-opportunities-2 (evidence/verify/criteo.json) |
| C428 | Its published correction route for delivery data could be extended to attributed sales, which no screened page covers. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Methodology note: issues affecting 'advertiser spend or data' lead to notification and reprocessing 'to correct billing and reporting metrics', in a section on delivery-server logging. | supported with qualification | Its published correction route, written for delivery and billing data, could be stated explicitly for attributed sales, which no screened page names. | The reissuance sentence does not exclude attributed metrics; the gap is that none of the pages names them. | SWOT-opportunities-3 (evidence/verify/criteo.json) |
| C429 | Large retailers may in-house retail media technology, as the 10-K warns; the 2025 scope changes show how fast segment revenue can fall. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | 10-K risk factors (as returned): 'Some large retailers, which could include our own clients, may develop retail media advertising technologies in-house' and may move demand to a direct sales model. Q2 2026: (21)% with a $21 million headwind from scope changes; the nature of the scope changes is not disclosed. | supported with qualification | Large retailers may in-house retail media technology, as the 10-K warns; the 2025 scope changes with two undisclosed clients, whose nature Criteo has not stated, show how fast segment revenue can fall. | Do not imply the scope changes were in-housing; the issuer calls them 'temporary'. | SWOT-threats-1 (evidence/verify/criteo.json) |
| C430 | A second domicile change, a merger into a Delaware company due 1 January 2027, still awaited shareholder approval on 28 October 2026. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | Effective Time 12:00:01 a.m. New York time on 1 January 2027 unless the boards agree otherwise; 8-K of 28 October 2026: shareholder meeting still to come, Form S-4 not yet declared effective. | supported with qualification | A second domicile change, a merger into a Delaware company scheduled for 1 January 2027 (unless the boards agree otherwise), still awaited an effective Form S-4 and shareholder approval on 28 October 2026. | The same period carries a reported, unconfirmed take-private approach by Vista Equity Partners and Quinti Capital (July 2026, trade press via search only); see omissions. | SWOT-threats-3 (evidence/verify/criteo.json) |
| C431 | Health-data litigation naming Criteo with CVS, a retail media operator, was unresolved at the 10-K date. | Refute-mode verifier for evidence/profiles/criteo.json re-opened the cited source. | R28: putative class action against the Company, CVS and others over 'sensitive health and personal information'; dismissed as to Criteo 27 October 2025; second amended complaint 7 November 2025; plaintiff 'filed a stipulation to dismiss the case' on 30 January 2026, parties in mediation. | supported with qualification | Health-data litigation naming Criteo with CVS was not finally resolved at the 10-K date: the plaintiff filed a stipulation to dismiss on 30 January 2026 and the parties were in mediation; the 10-K does not connect the case to retail media. | basis 'inference' is right for any link to retail media: CVS operates CVS Media Exchange, but the filing does not say the case concerns it. | SWOT-threats-4 (evidence/verify/criteo.json) |
| C432 | Maplebear's audited 'Advertising and other' revenue was US$871 million (2023), US$958 million (2024) and US$1,065 million (2025), company-wide; the 10-K policy says the advertising services 'include Sponsored Product ads, display ads, coupons, and a variety of other online advertising services', records third-party arrangements gross or net by contract, and calls software subscription and other revenue 'not material', so the line spans all three Instacart register rows plus coupons and other revenue (perimeter: broader parent proxy). | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | R13 disaggregation (US$ millions, 2023 / 2024 / 2025): Transaction $2,171 / $2,420 / $2,677; Advertising and other 871 / 958 / 1,065; Total revenue $3,042 / $3,378 / $3,742. R12: 'The advertising services include Sponsored Product ads, display ads, coupons, and a variety of other online advertising services.'; revenue recognised 'upon delivery of clicks, upon delivery of impressions, over the contract term on a fixed fee basis, or upon redemptions of coupons'; 'For advertising arrangements that involve third parties, the Company records advertising revenue on a gross or net basis based on whet | supported with qualification | Maplebear's audited 'Advertising and other' revenue was US$871 million (2023), US$958 million (2024) and US$1,065 million (2025), company-wide; the revenue note says the advertising services 'include Sponsored Product ads, display ads, coupons, and a variety of other online advertising services', records third-party arrangements gross or net by contract and calls software subscription and other revenue 'not material'. The note does not name Carrot Ads; that Carrot Ads revenue sits inside this line is an inference (Item 1 of the same 10-K calls Carrot Ads 'our enterprise ads offering'), so the | The financials rows (FY2023 definition) list 'Carrot Ads' among the line's contents as if the note said so; mark it '(inference; the note does not name Carrot Ads)'. | RTL-INSTACART-F02 (evidence/verify/instacart-ads.json) |
| C433 | Instacart's Sponsored Products reporting has used linear attribution since 1 August 2022 (even credit across ad interactions on the purchasing path; clicks only when clicks and views mix, from 1 April 2025; direct and halo sales reported separately), and a third-party help article of 22 July 2022 already described a 14-day click window with linear multi-touch as default and last-touch as an option, matching the 2026 Sponsored Products definition. | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | Linear page ('Last updated on Aug 12, 2026'): 'As of August 1, 2022, both Ads Manager and Instacart Ads API reporting views primarily use linear attribution.'; 'only affects sales reporting data for dates August 1, 2022, and onward'; '...each ad they interacted with on that purchasing path gets assigned an even split of that credit.'; April 1, 2025: 'if there's a mix of clicks and views on the purchasing path, we only assign the even credit split to clicks.'; Instacart 'historically used last touch attribution for sales reporting within each sponsored product and display ad format'; users can | supported with qualification | Instacart's Sponsored Products reporting has 'primarily' used linear attribution since 1 August 2022 (even credit across ad interactions on the purchasing path; clicks only when clicks and views mix, from 1 April 2025; an 'updated last touch model' remains viewable 'for comparison'; direct and halo sales reported separately); the linear page states no window, and the 14-day click window comes from the Sponsored Products page; a third-party help article of 22 July 2022 already described a 14-day window with linear multi-touch as default and last-touch as an option. | The Perpetua article (RTL-G1-S14) was not re-opened here; verify/G1.json (RTL-G1-F22) records its passages verbatim. Keep 'primarily' and the last-touch comparison view: the default is linear, not exclusively linear. | RTL-INSTACART-F05 (evidence/verify/instacart-ads.json) |
| C434 | Instacart reports two lift figures for Sponsored Products, both its own: Reckitt's '16.5% incremental sales lift' from a test in Q3-Q4 2022 described as 'comparing the sales of two randomly selected groups', and 'more than a 15% incremental sales lift – and sometimes twice that' on average from internal tests across brand partners in the year ending 31 December 2023; neither page gives sample size, interval, significance rule or the advertisers' spend. | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | Reckitt page: 'Reckitt's sales lift test proved a 16.5% incremental sales lift* from their sponsored product ads and a positive incremental return on ad spend (ROAS) result.'; 'comparing the sales of two randomly selected groups, one that is exposed to the ads being tested and a control group that is not.'; the difference 'represents the sales on Instacart that would only have occurred with running the ads'; 'When Reckitt ran their lift test in Q3-Q4 of 2022'. Instacart post of 17 December 2024 ('New Studies Show Instacart Ads Influencing In-Store Purchases'): 'Previous studies have shown how | supported with qualification | Instacart's pages report more lift figures than the two quoted: Reckitt's '16.5% incremental sales lift' (sponsored product ads, Q3-Q4 2022, 'two randomly selected groups'); 'more than a 15% incremental sales lift – and sometimes twice that' on average (internal tests, year ending 31 December 2023); Kellogg's 'double-digit total sales lift' (2022); Applegate 'a 35% sales lift' and Challenge Dairy 'an incremental sales lift of over 60%' (17 December 2024 post; format, period and design not stated); and Circana-provided in-store sales lifts of 21% and 15% for two unnamed CPGs 'among consumers ex | The Circana in-store figures are a party study (Instacart: 'we collaborated on additional studies'; 'Results provided by Circana'); funding not stated, so code commissioned/unknown, not independent; assignment unknown (a 'control group' is not thereby randomised). | RTL-INSTACART-F09 (evidence/verify/instacart-ads.json) |
| C435 | Instacart's own measurement pages describe its MRC scope more broadly than MRC's letter: the methodology page (12 August 2026) says 'All of these impression metrics are accredited by the Media Rating Council' beside a platform list that includes Caper and Rosie, and the 6 November 2025 release says the accreditation covers partner sites 'in both the U.S. and Canada', while the letter names only the Instacart Marketplace and retailer properties through Storefront, Storefront Pro and Connect API integrations. | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | Methodology page ('Last updated on Aug 12, 2026'): under 'Ad placements and ad types', 'Ads can appear on the following platforms—' lists 'Instacart website (desktop and mobile)', 'Instacart app (mobile only)', 'Storefront (desktop and mobile)', 'Storefront Pro (desktop and mobile)', 'Carrot Ads API Partners (desktop and mobile)', 'Caper', 'Rosie'; 'All of these impression metrics are accredited by the Media Rating Council.' follows the impression definitions (gross, gross video, sponsored product gross, viewable, viewable video, shoppable video). Caper Carts page ('Last updated on Aug 12, 202 | supported with qualification | Instacart's own pages describe its MRC scope more broadly than MRC's letter: the methodology page (12 August 2026) says 'All of these impression metrics are accredited by the Media Rating Council' on a page whose platform list includes Caper and Rosie; Instacart's Caper Carts page (12 August 2026) says eligible shoppable display campaigns extend automatically to Caper Carts and 'All Ads Manager reporting displays aggregated data for all surfaces, including Caper Carts', with a Caper impression counted at '1 second on the screen'; and the 6 November 2025 release promises 'Verified Ad Metrics to | Accreditation must not be extended to Caper, Rosie, retailers' own storefronts or 'Carrot Ads API Partners' unless MRC names them; the letter governs scope. | RTL-INSTACART-F12 (evidence/verify/instacart-ads.json) |
| C436 | Third-party verification on Instacart's on-site inventory is announced but not shown to be live: DoubleVerify's release of 12 October 2023 says DV and Instacart 'will be working on the technical integration' to verify that 'Instacart video and display ads' are viewable and free of invalid traffic 'over the coming months'; Instacart's own post (20 October 2023 per the brief) could not be opened, IAS's release was not opened, and Instacart's 2026 measurement methodology page names neither vendor. | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | DV release ('NEW YORK – October 12, 2023 –'): DV will 'verify that Instacart video and display ads are viewable by a human being and are safe from Fraud/Invalid Traffic'; 'DV and Instacart will be working on the technical integration for viewability and IVT verification solutions over the coming months'; sponsored products and brand safety not mentioned. DoubleVerify Q3 2024 results (8-K exhibit 99.1, 6 November 2024), under 'Third Quarter and Recent Business Highlights': 'Launched the first platform-wide invalid traffic detection and viewability verification solution on Instacart' (sentence c | contradicted | Third-party verification on Instacart's inventory is announced and reported as launched by the vendor, with format scope unstated: DoubleVerify's release of 12 October 2023 said DV and Instacart would build viewability and invalid-traffic verification for 'Instacart video and display ads' 'over the coming months', and DoubleVerify's Q3 2024 earnings release (8-K exhibit 99.1, 6 November 2024) lists 'Launched the first platform-wide invalid traffic detection and viewability verification solution on Instacart' (DV's assertion; formats, and whether sponsored product ads are covered, not stated). | Status field: change 'announced (2023), current status unknown' to 'launched per DoubleVerify (Q3 2024 release, 6 November 2024); format scope unstated; status after 2024 not documented on pages read'. DV is an interested party (it sells the verification); a vendor statement in its own furnished earnings release is still a vendor assertion about the product. | RTL-INSTACART-F13 (evidence/verify/instacart-ads.json) |
| C437 | Instacart documents dated reporting revisions for Sponsored Products and a public changelog: impressions switched from served to viewable impressions 'Starting January 18, 2022' with the change applied retroactively to reporting from 1 January 2022 (earlier data unaltered) and in-product notices where periods span the change; linear attribution from 1 August 2022 ('This change only affects sales reporting data for dates August 1, 2022, and onward'); clicks-only credit from 1 April 2025; an Ads Manager changelog with nine dated entries since 25 March 2026; but no restatement policy for metrics, | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | Impression page ('Last updated on Aug 12, 2026'): 'Starting January 18, 2022, Instacart Ads reports viewable impressions for all sponsored product ads'; 'This change retroactively affects reporting for dates January 1, 2022 onward.'; 'Historical campaign reporting on dates preceding January 1, 2022, for sponsored product ads remain unaltered and are based on served impressions'; notice that 'both new viewable impressions and old served impressions are present'; 'This change doesn't affect clicks, cost per click (CPC), attributed sales, and return on ad spend (ROAS).' (CTR, computed on viewable | supported with qualification | Instacart documents dated reporting revisions for Sponsored Products and a public changelog: impressions switched from served to viewable impressions 'Starting January 18, 2022', applied retroactively to reporting from 1 January 2022 (earlier data 'remain unaltered'), with in-product notices where periods span the change and with clicks, CPC, attributed sales and ROAS stated to be unaffected; linear attribution from 1 August 2022; clicks-only credit from 1 April 2025; an Ads Manager changelog with nine dated entries since 25 March 2026. No restatement policy for metrics, no dispute procedure f | The truncated quote 'Instacart shall not adjust invoices retroactively.' reverses the clause's scope; carry the full sentence in governance (both rows) and in the timeline/limits wherever it appears. | RTL-INSTACART-F14 (evidence/verify/instacart-ads.json) |
| C438 | By 11 November 2025 Instacart's own releases named five US networks as using Carrot Ads: Hy-Vee, Sprouts, Thrive Market and Uber Eats ('use this solution', 6 November 2025) and Associated Food Stores, whose Davis Food & Drug said its 'monthly retail media revenue grew by 2-3x' with Carrot Ads (11 November 2025); Instacart counted 'more than 240 ecommerce partners' in November 2025 and 'more than 310 partners' in September 2026, while Gopuff's deployment (22 September 2026) was announced as future and Vroom Delivery's release could not be read. | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | 6 November 2025: 'Partners like Hy-Vee, Sprouts, Thrive Market, and Uber Eats use this solution'; 'Carrot Ads powers white-label retail media for more than 240 ecommerce partners'. 11 November 2025: 'As a Retailer Owned Food Distributors & Associates (ROFDA) member, AFS is also leveraging Carrot Ads as their retail media network, providing in-store and online retail media revenue opportunities for owned and operated and member retailers using Storefront Pro and Caper Carts, Instacart's AI-powered smart cart.'; CJ Davis: 'our monthly retail media revenue grew by 2-3x with Instacart's retail med | supported with qualification | By 11 November 2025 Instacart's own releases and blog named six US networks as using Carrot Ads: Schnucks (live on Schnucks.com and its app since 2021 per Instacart's blog of 19 August 2025, which reports its retail media revenue up '7x'), Hy-Vee, Sprouts, Thrive Market and Uber Eats ('use this solution', 6 November 2025; Hy-Vee's March 2025 release had said 'later this year') and Associated Food Stores ('is also leveraging Carrot Ads', 11 November 2025), whose Davis Food & Drug said its 'monthly retail media revenue grew by 2-3x'. Instacart counted 'more than 220 retailer banners' (March and | Count rule: Schnucks has dated production evidence (vendor blog, 19 August 2025, describing live use since 2021) and should be counted: 6, not 5. Every count rests on vendor sources (cap 3). The 'will soon' quote misattributes Gopuff's timing word. | RTL-INSTACART-F19 (evidence/verify/instacart-ads.json) |
| C439 | Maplebear's board decided on 25 October 2026 to 'wind down Caper Cart operations', substantially by the end of Q2 2027, with pre-tax charges of about US$55 million to US$80 million; Caper carts are the in-store surface Instacart's Carrot Ads page offers ('personalized ads on Caper carts') and appear in the platform list of Instacart's measurement methodology page, so Carrot Ads' in-store class is being wound down (inference: the 8-K does not mention advertising). | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | 8-K Item 2.05 (signed 26 October 2026 by Emily Reuter, CFO): on October 25, 2026 the Board 'made the determination to wind down Caper Cart operations'; 'substantially completed by the end of the second quarter of 2027'; 'pre-tax charges of approximately $55 million to $80 million related to the Caper Wind-Down', primarily in Q4 2026 (contract terminations US$21-36m, Caper Cart asset impairments US$22-27m, employee termination US$12-17m; US$30-50m cash); no mention of advertising, ads, Carrot Ads or retail media. Caper Carts page (12 August 2026): 'Starting April 14, 2025, shoppable campaigns d | supported with qualification | Maplebear's board decided on 25 October 2026 to 'wind down Caper Cart operations', substantially by the end of Q2 2027, with pre-tax charges of about US$55 million to US$80 million. Caper carts carry two ad offers: Instacart display and video ads, whose eligible shoppable display campaigns have extended automatically to Caper Carts since 14 April 2025 through Ads Manager, and Carrot Ads' in-store surface ('personalized ads on Caper carts'; Associated Food Stores' members 'using Storefront Pro and Caper Carts'). Inference (the 8-K does not mention advertising): both rows' in-store placements ar | Event type correctly 'exit or disposal activity'. The inference is labelled, but it is attached to the wrong row only: the display and video row is the documented brand-facing seller of Caper Cart ads. | RTL-INSTACART-F20 (evidence/verify/instacart-ads.json) |
| C440 | Instacart's privacy record for its advertising is company-level only: the Privacy Policy (updated 15 January 2026) discloses personal information to 'marketing and advertising partners' to 'assist in serving and measuring our or their advertisements', treats prescription-delivery data under HIPAA and honours Global Privacy Control, but names no 'Instacart Ads' control; the ads terms carry no data-protection clause; Maplebear is not on California's data-broker registry by name; and its FTC matters are a December 2025 complaint and January 2026 stipulated order over Instacart+ consent and fees a | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | Terms: no data-protection or privacy clause (confirmed). For the FTC matter, the J verifier read the court-entered order: 'Case 4:25-cv-10783-JST Document 15 Filed 01/13/26'; 'IT IS SO ORDERED. Dated: January 13, 2026'; US$60,000,000 judgment; the 18 December 2025 release cites false 'free delivery' claims, undisclosed mandatory service fees, a false '100% satisfaction guarantee' and charging for paid memberships 'without their express informed consent'. | supported with qualification | ...its FTC matters are a complaint of 18 December 2025 (false 'free delivery' and '100% satisfaction guarantee' claims, undisclosed service fees and Instacart+ charges without express informed consent), settled by a stipulated order the court entered on 13 January 2026 (US$60 million judgment; posted by the FTC on 14 January 2026, case page still 'Pending'), and a 28 October 2026 Section 6(b) pricing inquiry; neither concerns advertising data. | Status: the order is court-entered (final), not pending; timeline rows 2025-12-18 and 2026-01-14 and the privacy cells should carry the J verifier's correction. | RTL-INSTACART-F21 (evidence/verify/instacart-ads.json) |
| C441 | The opened record documents Instacart Ads in the US and Canada, not the US alone: display metrics are reported 'in USD or CAD, as applicable', a 2024 tool article says 'Instacart ads are currently only supported in the US and Canada', the November 2025 accreditation release covers partner sites 'in both the U.S. and Canada', and the deal requires approvals 'including in Canada and Israel'; Maplebear's advertising line is company-wide. | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | Display page: 'in USD or CAD, as applicable'. 6 November 2025 release: environments 'in both the U.S. and Canada'. Merger release: approvals 'including in Canada and Israel'. Terms: fees payable 'in U.S. dollars for Services provided in the United States or in Canadian dollars for Services provided in Canada'. 10-K R13: revenue by bill-to location, United States US$3,600 million and International US$142 million of US$3,742 million in 2025 (2024: 3,247 and 131; 2023: 2,936 and 106); 'No individual international country represented 10% or more of the Company's total revenue.' | supported with qualification | The opened record documents Instacart Ads in the US and Canada, not the US alone (USD or CAD reporting; 'in both the U.S. and Canada'; terms priced for 'Services provided in the United States' or 'in Canada'); Maplebear's advertising line is company-wide, and the only geographic split in the 10-K is total revenue by bill-to location: US$3,600 million United States and US$142 million international in 2025 (96.2% US, derived), with no advertising split. | Resolves the profile's limitation 'the geographic note was not reached': R13 carries it. It does not give a US advertising figure. | RTL-INSTACART-F24 (evidence/verify/instacart-ads.json) |
| C442 | This review found no independent validation of any Instacart outcome number in the sources screened by 3 November 2026: the lift results are Instacart's own (Reckitt, the 2023 average), the Circana analyses are buyer-overlap descriptions with funding unstated, MRC accreditation covers delivery metrics only, and AAM's incrementality certification list names DoorDash alone. | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | 17 December 2024 post: Instacart 'collaborated' with Circana on studies of in-store effects; 'CPG A had a 21% in-store sales lift among consumers exposed to its sponsored product ads on Instacart; CPG B had a 15% lift.'; 'Data for exposed users from 4/1/24-6/16/24 measured for full Q2 '24 vs. Circana control group. Results provided by Circana.' MRC letter: delivery metrics only. | supported with qualification | This review found no independent validation of any Instacart outcome number in the sources screened by 3 November 2026: the lift results are Instacart's own (Reckitt, the 2023 average, Applegate, Challenge Dairy), the Circana work comprises buyer-overlap descriptions (August 2023, February 2024) and Circana-provided in-store lift figures for two unnamed sponsored-product advertisers against a 'Circana control group' in a study Instacart says it 'collaborated' on (funding, control construction and assignment not stated), MRC accreditation covers delivery metrics only, and AAM's incrementality c | The Circana in-store lift is a party study (collaboration with the operator; funding unstated), so it does not overturn the absence of independent validation; it does change the validation cell's description of what exists (a third-party-computed, operator-collaborated lift on sponsored products). | RTL-INSTACART-F26 (evidence/verify/instacart-ads.json) |
| C443 | Default report: attribution with no counterfactual. Sponsored Products 'Attributed sales' are purchases 'within 14 days of a customer clicking' (S08), credited by linear attribution since 1 August 2022, clicks only when clicks and views mix from 1 April 2025 (RTL-E1-S14). Controlled lift is a separate test: launched as a pilot on 2 September 2022 'using randomized test and holdout groups' (S01); documented in the 30 June 2023 method post as random assignment per user-advertiser pair ('hash of (user_id, advertiser_id)') with ghost ads, limited to Sponsored Products (RTL-U-S309); offered on requ | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | As in F04, F05, F07 and F08: 'within 14 days of a customer clicking on a sponsored product ad'; reporting views 'primarily use linear attribution' from 1 August 2022; clicks-only split from 1 April 2025; 2022 pilot 'randomized test and holdout groups'; 2023 'hash of (user_id, advertiser_id)' and Ghost Ads, 'limited to Instacart's Sponsored Products offering'; 2023 'may not be available to all partners. Please consult with your account team'; 28 September 2026 'Sales lift testing ... gold-standard experimentation' with no design; analytics index lists no lift page. | supported with qualification | Replace 'credited by linear attribution since 1 August 2022' with 'credited primarily by linear attribution since 1 August 2022 (a last-touch view remains available for comparison)'. The rest stands. | Default method stated correctly: attribution with no counterfactual; lift is separate, documented with a randomised design in 2022-2023 and as an undescribed 'Sales lift testing' offer in 2026. Evidence grade C is right. | CELL capabilities.default_method (Instacart Sponsored Products) (evidence/verify/instacart-ads.json) |
| C444 | Screen completed 2026-11-03 over S08-S13 and RTL-E1-S09, S10, S13, S14, S15, S32. (1) Metric versioning and change logs: yes, Ads Manager changelog with nine dated entries since 25 March 2026 (S11) and Ads API versioning with breaking and non-breaking definitions (RTL-E1-S09). (2) Restatement and correction policy: partial, the 2022 impression change was applied retroactively from 1 January 2022 with pre-2022 data unaltered and in-product notices (S10); no general policy found; 'Instacart shall not adjust invoices retroactively' (S13). (3) Documented reporting revisions: yes, served to viewabl | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | See F14: the full invoice sentence is 'Instacart shall not adjust invoices retroactively if Instacart, in its sole discretion, finds an inaccuracy in the billing amount of +/- 1% of the initial invoiced amount.'; the 2022 change 'doesn't affect clicks, cost per click (CPC), attributed sales, and return on ad spend (ROAS)'. | supported with qualification | In item (2), replace "'Instacart shall not adjust invoices retroactively' (S13)" with "'Instacart shall not adjust invoices retroactively if Instacart, in its sole discretion, finds an inaccuracy in the billing amount of +/- 1% of the initial invoiced amount' (S13), a de minimis rule", and add that the 2022 impression change left clicks, CPC, attributed sales and ROAS unaffected. | Nine changelog entries since 25 March 2026 confirmed. | CELL capabilities.governance (Instacart Sponsored Products) (evidence/verify/instacart-ads.json) |
| C445 | Classes served: on-site (display, shoppable display, shoppable video, brand pages; RTL-E1-S42) and off-site (products page lists social, open web, streaming video and CTV; The Trade Desk rail, provisional assignment, RTL-H-S40). Per-class measurement: on-site display documented with parameters (S09, S12); shoppable video exposure capture documented (50% of player for 2 continuous seconds, S12) but no outcome metric definitions; off-site measurement asserted ('closed-loop measurement') without parameters. | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | Caper Carts page (12 August 2026): 'Starting April 14, 2025, shoppable campaigns display on Caper Carts'; 'all eligible shoppable display campaigns will automatically extend delivery to Caper Carts'; 'only shoppable display campaigns with keyword targeting or with optimized performance are eligible.'; 'No, Caper-specific reporting isn't available.'; 'All Ads Manager reporting displays aggregated data for all surfaces, including Caper Carts.'; 'Impression: 1 second on the screen will count as a 1 impression.'; 'simplify campaign measurement across all touchpoints with unified attribution models | supported with qualification | Classes served: on-site (display, shoppable display, shoppable video, brand pages), in-store (eligible shoppable display campaigns extend automatically to Caper Carts from 14 April 2025; Caper-specific reporting 'isn't available' and Ads Manager reports aggregate all surfaces including Caper Carts; a Caper impression is '1 second on the screen'; Caper Cart operations being wound down by the end of Q2 2027) and off-site (The Trade Desk rail, provisional). Per-class measurement: on-site display documented with parameters; in-store not reported separately, pooled with on-site; shoppable video exp | The row's reporting pools two placement classes (on-site and in-store) that the paper's definitions keep apart; record it as the operator's pooling, not as the paper's. | CELL product_capabilities["Instacart display and video ads"].placement_coverage (evidence/verify/instacart-ads.json) |
| C446 | MRC accreditation, completed and continuing (letter 1 June 2026): display and video impressions, clicks, CTR, viewable impressions and viewability for 'Display Ads, Shoppable Display Ads and Shoppable Video Ads placements served on the Instacart Marketplace', D, MW, MA (RTL-A-S06). Delivery only; no outcome metric; off-site placements (The Trade Desk) are outside the letter. AAM and IAB Europe: not listed. | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | Letter: placements 'served on the Instacart Marketplace as well as on retailer properties through Storefront and Storefront Pro integrations and Connect API integrations'; Caper not named. Caper Carts page: 'All Ads Manager reporting displays aggregated data for all surfaces, including Caper Carts.' | supported with qualification | MRC accreditation, completed and continuing (letter 1 June 2026): display and video impressions, clicks, CTR, viewable impressions and viewability for 'Display Ads, Shoppable Display Ads and Shoppable Video Ads placements served on the Instacart Marketplace' (and retailer properties through Storefront, Storefront Pro and Connect API integrations), D, MW, MA. Delivery only; no outcome metric. Placements sold in this row but outside the letter: off-site (The Trade Desk) and in-store Caper Carts, whose impressions Ads Manager pools with on-site figures. AAM and IAB Europe: not listed. | Do not code this as delivery assurance 'in the environments sold': the row sells in-store (Caper) and off-site placements the letter does not name, so against the anchors it reads as assurance of a narrower delivery scope. | CELL product_capabilities["Instacart display and video ads"].assurance (evidence/verify/instacart-ads.json) |
| C447 | Distinct named US retailer networks stated as using Carrot Ads, with dates: five by 11 November 2025 (Hy-Vee, Sprouts, Thrive Market, Uber Eats per the 6 November 2025 release, S14; Associated Food Stores incl. Davis Food & Drug per the 11 November 2025 release, S15), all from vendor press releases (cap 3). Not counted: Gopuff (announced 22 September 2026, 'will soon', RTL-U-S093), Vroom Delivery (release unread, S18), Schnucks (case undated, RTL-U-S308). Partner counts in other units: 'more than 240 ecommerce partners' (November 2025, S14), 'more than 310 partners' (September 2026, RTL-U-S093 | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | See F19. | supported with qualification | Distinct named US retailer networks stated as using Carrot Ads, with dates: six by 11 November 2025 (Schnucks, live since 2021 per Instacart's blog of 19 August 2025; Hy-Vee, Sprouts, Thrive Market, Uber Eats per the 6 November 2025 release; Associated Food Stores incl. Davis Food & Drug per the 11 November 2025 release), all from vendor sources (cap 3). Not counted: Gopuff (announced 22 September 2026, 'will be utilizing'), Vroom Delivery (announced 11 September 2025, 'will have access'). Partner counts in other units: 'more than 220 retailer banners' (March and August 2025), 'more than 240' | Rung unchanged (5 to 9 = rung 4; capped at 3 by grade D). Update the data table's 'Total counted' row and the Schnucks row. | CELL product_capabilities["Carrot Ads"].scale (evidence/verify/instacart-ads.json) |
| C448 | MRC accreditation, completed: granted October 2025 for Instacart's measurement and reporting of Carrot Ads (display and video impressions, clicks, CTR, viewable impressions; desktop web, mobile web, mobile in-app) on retailer properties through Storefront, Storefront Pro and Connect API integrations (RTL-A-S10), continued in the letter of 1 June 2026 'across Instacart Ads and Carrot Ads' (RTL-A-S06). Delivery only; it does not attest what a retailer reports through other tools; no retailer is named in the letter. MRC's digital list entry 'Instacart' does not name Carrot Ads (RTL-U-S021). | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | Q4 2025 update: October grant for Carrot Ads 'specifically for Sponsored Product, Display, Shoppable Display and Shoppable Video Ads on retailer properties through Storefront and Storefront Pro integrations as well as the Connect API integrations.' 6 November 2025 release footnote: 'Carrot Ads is available to retailers using Instacart's Storefront Pro, Connect, or their own e-commerce storefronts'. Methodology page lists 'Carrot Ads API Partners (desktop and mobile)' separately from Storefront and Storefront Pro. | supported with qualification | Add: Carrot Ads placements on retailers' own e-commerce storefronts (Carrot Ads API partners) and on Caper carts are not named in MRC's letter or its Q4 2025 entry; whether 'Connect API integrations' covers Carrot Ads API partners is not stated (a 10 September 2026 Carrot Ads changelog topic on 'creating a Connect user account' suggests a link, not coverage). Code the accreditation as covering the named integrations only, a possibly narrower scope than Carrot Ads is sold in. | The release's headline ('Verified Ad Metrics to More Than 240 Ecommerce Sites') is the vendor's extension of the accreditation to every partner; the letter does not count or name sites. | CELL product_capabilities["Carrot Ads"].assurance (evidence/verify/instacart-ads.json) |
| C449 | Financials: FY2023-FY2025 Advertising and other 871 / 958 / 1,065; FY2025 total revenue 3,742; Q3 2026 Advertising and other 331 and total revenue 1,095; DoorDash equity value approx. US$17 billion; Caper Wind-Down pre-tax charges 55-80 (US$ million). | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | All values reproduced from R13, the Q3 2026 release, the merger release and the Item 2.05 8-K (charges 'primarily' in Q4 2026). | supported with qualification | FY2023 definition: mark 'Carrot Ads' as inference (the revenue note does not name it). | Q3 2025 total revenue US$939 million and the 17% growth check (1,095 / 939 = 1.166). Consider recording GTV US$10,528 million beside the 3.1% share. | FINANCIALS (all rows) (evidence/verify/instacart-ads.json) |
| C450 | Timeline entries 2022-01-18, 2022-08-01, 2022-09-02, 2023-05-05, 2023-06-30, 2023-10-12, 2025-04-01, 2025-10, 2025-11-06, 2025-11-11, 2025-12-18, 2026-01-14, 2026-06-01, 2026-09-22, 2026-10-25, 2026-10-28 (merger), 2026-10-29, 2027-12 to 2028-06. | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | Dates reproduced from each source; Gopuff page header 'Oct 9, 2026' versus body 'September 22, 2026'; Q3 release also 'has withdrawn previously issued long-term guidance'. | supported with qualification | (a) 2025-12-18 and 2026-01-14: per verify/J.json the court entered the stipulated order on 13 January 2026 (US$60 million; complaint also covers 'free delivery', service-fee and satisfaction-guarantee claims), posted by the FTC on 14 January 2026. (b) 2026-09-22: note the page header date of 9 October 2026. (c) 2026-10-29: add 'previously issued long-term guidance withdrawn'. (d) Add: 2024-11-06 DoubleVerify reports it 'Launched the first platform-wide invalid traffic detection and viewability verification solution on Instacart'; 2024-12-17 Instacart publishes Circana in-store lift figures for | TIMELINE (events checked) (evidence/verify/instacart-ads.json) | |
| C451 | Deep profile named in the brief: Instacart is the largest grocery marketplace network in the register, the only one whose operator published a randomised holdout design with ghost ads for its sponsored products, it licenses its ad stack to other retailers as Carrot Ads under the same MRC letter, and on 28 October 2026 DoorDash agreed to acquire it, a transaction expected to close at the end of 2027 or in the first half of 2028. | Refute-mode verifier for evidence/profiles/instacart-ads.json re-opened the cited source. | Method post confirms the randomised holdout with ghost ads for Sponsored Products; the letter covers Carrot Ads placements through Storefront, Storefront Pro and Connect API integrations only. | supported with qualification | Bound the uniqueness claim ('the only one in the register whose operator published ..., in the sources screened by the cutoff') and replace 'under the same MRC letter' with 'with Carrot Ads placements through Storefront, Storefront Pro and Connect API integrations under the same MRC letter'. | Unbounded absence claim across the register (hard rule 1); not checked against other operators by this verifier. | WHY_PROFILED (evidence/verify/instacart-ads.json) |
| C452 | Documented default reporting for the onsite row is attribution: KAP onsite reporting under a 14-day window, halo by brand and commodity, last interaction prioritised and click over view (84.51° Report Center); controlled lift naming the onsite formats is PrecisionView 360 (paid, opt-in, US$400K across three or more channels). Coding: default target contrast attributed, assignment none, counterfactual none; Workstream E1 (RTL-E1-F49) coded the default as an exposed-versus-unexposed comparison. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | Report Center (Prism, 'Last updated: 8/17/2026'), section 'Onsite Campaigns': 'The attribution window for Onsite Campaigns (for example, Product Listing Ad, Promoted Product Carousel, and Targeted Onsite Ad) is 14 days. Reporting within the Kroger Ad Platform shows sales for the exposure on the day it occurs.' ... 'So, if a shopper is exposed to a TOA days before purchasing an attributable product, reporting shows the ad cost on the date of exposure and the viewed sale on the date of purchase. For example, if a shopper is exposed to a TOA on December 27th, but purchases an attributable product | supported with qualification | Restate the observed field as: 'The onsite row has two documented reporting routes with different defaults. (a) KAP self-serve formats (Product Listing Ads, Targeted Onsite Ads, Promoted Product Carousels, including Video Carousels, and KAP Premium Placements such as HOME Carousels): the only public description of KAP onsite reporting, an 84.51° Prism Report Center section that names the Kroger Ad Platform, documents attribution: a 14-day window after a click (PLA, carousel) or a view (TOA), click over view, last interaction, halo (two stated scopes), cost shown on the exposure date and the sa | The Prism help centre is a different 84.51° product (Prism runs Incentives/BCC, managed-service and legacy Direct Connect campaigns), but its Report Center onsite section and its PrecisionView channel tables explicitly describe Kroger Ad Platform reporting, so it is usable vendor documentation for KAP onsite; it is not the KAP Learning Center and carries its own 'Last updated' date. The managed-se | capabilities.default_method (onsite row; priority 1) (evidence/verify/kroger-precision-marketing.json) |
| C453 | Kroger Precision Marketing's documented onsite method is an exposed-versus-unexposed household comparison that KPM calls both 'incrementality' and 'closed-loop attribution', reported in the Kroger Ad Platform as 'sales lift and household penetration' ... (assignment not disclosed). | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | FAQ and KAP-page passages are quoted correctly by E1. The format-specific rule for KAP onsite reporting is attribution: 'Shoppers have 14 days to purchase an attributable product after a click on a Carousel or PLA or a view on a TOA ... the click is always prioritized over the view ... the last interaction is prioritized for attribution' (Report Center). The exposed-versus-control comparison is documented for managed-service campaigns ('compared to tightly-matched control households') and for PrecisionView 360 and qualified Trade Desk campaigns. | contradicted | For the KAP self-serve onsite formats, code the documented default as attribution (14-day, last interaction, click over view, halo); record the network FAQ's exposed-versus-unexposed language as an unscoped vendor statement, and the matched-control Sales Uplift report as the managed-service route's default. | A network-level FAQ sentence and a marketing phrase ('sales lift') were coded as the default method of one product row. | RTL-E1-F49 (cross-check requested in the prompt; not a finding of this file) (evidence/verify/kroger-precision-marketing.json) |
| C454 | Kroger discloses no retail media revenue amount in the 10-K parts read (Items 1 and 7): third-party media revenue is one of four named components of 'Other sales', US$1,346 million in fiscal 2025 and US$1,177 million in fiscal 2024; the 2025 increase is 'primarily' due to 'other pharmacy services and third-party media revenue'. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | MD&A total sales table: 'Other sales(5)' 1,346 (2025) and 1,177 (2024); footnote (5): 'Other sales primarily relate to external sales at food production plants, other pharmacy services, third-party media revenue and data analytic services. The increase in 2025, compared to 2024, is primarily due to an increase in other pharmacy services and third-party media revenue.' Audited notes, Revenue Recognition, disaggregated revenues (XBRL R54): 'Other' US$1,129 million (period ended 31 January 2026), US$3,062 million (1 February 2025), US$4,229 million (3 February 2024); its footnote (R35): 'Consists | supported with qualification | Add: 'The audited revenue-recognition note carries a second, narrower residual category, "Other", of US$1,129 million in fiscal 2025 (US$3,062 million in fiscal 2024), which "consists primarily of sales related to third-party media revenue, data analytic services, specialty pharmacy and in-store health clinics"; its declines since fiscal 2023 are attributed to the disposal of Kroger Specialty Pharmacy, "partially offset by an increase in third-party media revenue". Neither line quantifies media revenue; the notes (revenue recognition, segment) were read and add no media amount.' Replace 'in th | Two filing lines with different perimeters exist; the MD&A line is a presentation table, the note category is an audited filing amount. | RTL-KROGER-F04 (priority 3) (evidence/verify/kroger-precision-marketing.json) |
| C455 | Report Center (last updated 17 August 2026) documents KAP onsite attribution: 14-day window, click over view, last interaction, halo by Brand and Commodity, and KAP reporting that 'shows sales for the exposure on the day it occurs'. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | All quoted fragments are verbatim. The sentence after 'shows sales for the exposure on the day it occurs' explains it: 'reporting shows the ad cost on the date of exposure and the viewed sale on the date of purchase', with the 27 December exposure and 29 December purchase example. 'You should run the end date in reporting as 14 days post-End Date (after the actual End Date) to capture all sales.' | supported with qualification | Append to the claim: 'the operator's example shows that KAP reporting places the ad cost on the exposure date and the attributed sale on the purchase date.' | The page is a Prism help page whose onsite section names the Kroger Ad Platform throughout; it is a valid public description of KAP onsite attribution rules. | RTL-KROGER-F08 (evidence/verify/kroger-precision-marketing.json) |
| C456 | PrecisionView 360: 'the premier omni-channel incremental sales measurement solution', 'A paid, cross-channel' study requiring 'a minimum investment of $400K across three (3) or more channels' and an opt-in, covering onsite channels (TOA, PLA, Promoted Product Carousels) as well as The Trade Desk and Yahoo, 'Includes richer methodology (marketing mix / precision Test vs. Control modeling)', delivered about five to nine weeks after the last exposure; assignment, sample and uncertainty not documented. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | Report Center: 'Your Cross-Channel Campaigns may qualify for PrecisionView 360 measurement and optimization, if your Campaign meets a minimum investment of $400K across three (3) or more channels.' 'You can opt into PrecisionView Lite or PrecisionView 360 measurement. To get started, contact your Account Management team'. PV360 channel table: Onsite (Email Module, Single Subject Email, Push Notification), Incentives (Targeted Digital Coupon), Offsite (Display Ad, Online Video Ad, Pinterest, Programmatic Audio, Programmatic CTV, Meta, Snapchat, Disney), Kroger Ad Platform (Targeted Onsite Ad, P | supported with qualification | Qualify: PV360 is offered to Cross-Channel Campaigns that 'may qualify' (eligibility and availability are discussed with the account representative, not guaranteed); its channels also include Prism managed onsite channels (email, push), Targeted Digital Coupons and managed offsite channels (display, online video, Pinterest, audio, CTV, Meta, Snapchat, Disney), and the KAP list includes Video Carousels. Whether TOA, PLA and carousels count as three 'channels' for a KAP-only campaign is not stated. | 'Test vs. Control modeling' is a label: assignment unknown, construction unstated, outputs modelled ('advanced modeling', 'marketing mix'). | RTL-KROGER-F10 (priority 2) (evidence/verify/kroger-precision-marketing.json) |
| C457 | On the record read, the documented default outcome report for KAP onsite search and display is attribution: KAP onsite reporting books attributable sales on the exposure day under a 14-day halo window; incremental measurement naming the onsite formats is PV360 and MRC's pre-audit metrics; no page read defines KAP's 'sales lift' or says a self-serve onsite advertiser receives it by default. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | See capabilities.default_method above: KAP shows 'the ad cost on the date of exposure and the viewed sale on the date of purchase'; the managed-service KPM Campaign Performance report leads with 'Sales Uplift' defined against 'tightly-matched control households'; the Glossary defines 'Uplift' (test versus a control 'not exposed') and 'Control Household / Group' ('intentionally excluded from a specific campaign or promotion during a test period'). | supported with qualification | Restate: 'On the record read, the documented default for the KAP self-serve onsite formats is attribution (14-day window, last interaction, click over view, halo; cost booked on the exposure date and the sale on the purchase date), and incremental measurement naming those formats is PrecisionView 360 (paid, opt-in, minimum US$400K across three or more channels) and the Incremental Sales and iROAS metrics in MRC's pre-audit scope. Managed-service campaigns in 84.51° Prism, which include homepage, department, seasonal and promotions Premium Placements, report Sales Uplift against tightly-matched | RTL-KROGER-F14 (priority 1) (evidence/verify/kroger-precision-marketing.json) | |
| C458 | 84.51° Collaborative Cloud, per a Vestcom release of 16 September 2022, gives data scientists transaction-level access to Kroger's 60MM households in a hosted environment, used by Vestcom for its own shelfAdz test-and-control measurement; no page read documents Collaborative Cloud access for buyers of KPM's onsite or offsite services. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | Vestcom release quotes are verbatim. KPM's own Collaborative Cloud page (undated; read by curl): 'How CPGs are using 84.51° Collaborative Cloud' ... 'Custom audiences: Activate tailored audiences across KPM’s onsite and offsite media solutions'; 'Built to accommodate data science teams writing code in Python, SQL, Scala and R, and applying modeling to transaction-level data.' The 2022 boilerplate: '84.51° Collaborative Cloud enables CPGs to unleash the potential of their data science teams'. | supported with qualification | Replace the last clause with: 'KPM's Collaborative Cloud page documents access for CPG data-science teams (KPM's buyer population) to transaction-level Kroger data and audience activation into KPM's onsite and offsite media; no page read documents KPM campaign exposure or outcome data inside it, or buyer-run measurement of KPM campaigns there.' | This also resolves the profile's needs_browser item for Collaborative Cloud (RTL-KROGER-S20, the 8451.com copy) through KPM's own page. | RTL-KROGER-F16 (evidence/verify/kroger-precision-marketing.json) |
| C459 | The KPM Direct Connect reporting policy (scoped to 'Campaigns run through Direct Connect only', deprecated 1 June 2026) requires advertisers to notify KPM 'if data is not flowing' within five business days, otherwise no makegood or restatement; immaterial changes 'at any time'; substantive changes 'discussed with Client'; reports 13 months; confidential. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | 'Advertisers must notify KPM if data is not flowing into the designated electronic portal after five (5) business days in order for KPM to assess and make appropriate remediations. If KPM is not timely notified, a makegood or restatement will NOT be submitted for approval.' Other quotes verbatim. 'KPM reserves the right to change our policies without notice.' | supported with qualification | Replace 'within five business days' with: 'when data has not flowed into the portal after five business days; the deadline for that notice is not stated ("If KPM is not timely notified ...")'. Apply the same fix to product_capabilities.offsite.governance ('restatement or makegood only if KPM is notified within five business days'). | Legacy Prism Direct Connect scope only; it is not current for KAP onsite or KAP Direct Connect, and the profile treats it so. | RTL-KROGER-F17 (priority 2, governance) (evidence/verify/kroger-precision-marketing.json) |
| C460 | Three route families for onsite: KAP self-service ('all-in-one self-service tool', 'second-price auction system', 'Now available to all advertisers'), Pacvue and Skai integrations, managed service ('Premium Placement Ad'); Shopping Assistant PLAs 'automatically included'. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | KAP page: 'The Kroger Ad Platform is your all-in-one self-service tool designed to boost your brand’s visibility on Kroger’s eCommerce sites and apps. Build impactful onsite search and display campaigns ...'; '... Keyword Bid Modifier within our second-price auction system'; 'Now available to all advertisers. Join the access list today!'; Pacvue and Skai under 'Platform Integrations'. Premium Placement page: 'For other Premium Placements (like Zone 1 and Zone 2 options), you'll use the Kroger Ad Platform'; managed Premium Placements need 'A six (6) week lead time' and 'run for one (1) week sta | supported with qualification | Add: 'Premium Placements split by route: homepage, department, seasonal and promotions banners are bought managed in 84.51° Prism; Zone 1 and Zone 2 placements and HOME Carousels are built in KAP. Onboarding to KAP is through an access list ("Join the access list today!").' | RTL-U-S296, S297 and S301 were not re-opened (403 from the host on 3 November 2026); their passages rest on workstream U's browser notes. | RTL-KROGER-F21 (evidence/verify/kroger-precision-marketing.json) |
| C461 | KPM's only product-level privacy statement is the FAQ answer (hashed identifiers, regional regulations, clean rooms 'where appropriate'); 84.51° links Kroger's privacy policy; no consent basis or sensitive-category rule found. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | FAQ verbatim as quoted, plus 'Audience freshness and retention windows help enforce privacy-by-design standards.' The KPM site footer links 'Privacy Policy' (https://www.kroger.com/i/privacy-policy/) and 'Privacy Center' (https://www.kroger.com/i/privacy-center), and shows 'Your Privacy Choices' and 'Do Not Sell' controls that are a OneTrust website-cookie settings button (ot-sdk-show-settings). The legacy Direct Connect page lists separate 'Endemic Intake Form', 'Non-Endemic Intake Form' and 'Alcoholic Beverage Intake Form'. | supported with qualification | Add the FAQ's retention sentence; note that the site's 'Do Not Sell' and 'Your Privacy Choices' are website cookie controls, not a product consent basis; note that a category-specific intake (alcoholic beverages) existed for legacy Direct Connect, without a published rule. | Kroger's privacy policy and Privacy Center remain unread (needs_browser). | RTL-KROGER-F25 (evidence/verify/kroger-precision-marketing.json) |
| C462 | Window 14 days, last interaction, click over view, halo with two stated scopes; match basis and in-store inclusion at network level; New Households 52 weeks platform-wide; sales basis and deduplication missing. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | As quoted in F08; Sub-Commodity rule: 'a shopper that clicks a Promoted Product Carousel or PLA or views a TOA, but purchases a product from a different Sub-Commodity (even if it is included in the Carousel, PLA, or TOA) will not be reflected in attribution'; KAP halo: 'same Brand and Commodity selected in the Ad Group'. PVLite: 'Audience and Ad Group-level data will NOT match data found in 84.51° Prism or the Kroger Ad Platform'. | supported with qualification | Add two documented items: KAP books the attributed sale on the purchase date and the cost on the exposure date; the operator states that PrecisionView Lite's cross-channel last-touch figures will not match KAP or Prism figures (a documented divergence between attribution regimes). | capabilities.attribution_disclosure (onsite) (evidence/verify/kroger-precision-marketing.json) | |
| C463 | Dashboard reporting only for onsite; no public API, event-level export or buyer-run query environment; Collaborative Cloud not documented for KPM buyers; offsite: legacy Report Center with 'Export data', 'Export all data'. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | Legacy Direct Connect export: 'Your final download includes one (1) page for each tab of your report, including visuals and filters, but not any raw data used to create the charts and tables'; chart-level 'Export Data' with 'the number of rows you export might be limited depending on the file type'. Managed-service report: chart-level 'Export Data' only. Collaborative Cloud: CPG data-science access and audience activation into KPM onsite and offsite media. | supported with qualification | Onsite: replace 'no page documents ... access for KPM's buyers' with 'Collaborative Cloud is documented for CPG data-science teams, with audience activation into KPM media; no page documents KPM exposure or outcome data in it'. Offsite: add that the legacy full export is a PowerPoint/PDF of report pages without raw data and chart exports are row-limited aggregates; neither is event-level. | A restriction on export is recorded as a restriction, not as proof of analytical dependence. | capabilities.buyer_analysis (onsite) and product_capabilities.offsite.buyer_analysis (evidence/verify/kroger-precision-marketing.json) |
| C464 | On-site class only (PLA, TOA, PPC, Premium Placements, Shopping Assistant PLAs); 'KAP reporting on the exposure day'; 'Onsite Video' in MRC scope but not on KPM's onsite page; Premium Placements not named in the onsite attribution rule. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | Report Center's KAP channel row reads 'Targeted Onsite Ad, Product Listing Ad, Promoted Product Carousels (and Video Carousels)'; KAP shows 'the ad cost on the date of exposure and the viewed sale on the date of purchase'; Premium Placements are split between Prism managed service and KAP (Zone 1/2, HOME Carousels). | supported with qualification | Replace 'KAP reporting on the exposure day' with 'KAP reporting that books cost on the exposure date and the sale on the purchase date'; note that KAP lists Video Carousels (a likely, not confirmed, match for MRC's 'Onsite Video'); note that managed Premium Placements are measured in the Prism KPM Campaign Performance report (Sales Uplift against matched controls) and that Premium Placements are outside MRC's pre-audit scope. | capabilities.placement_coverage (onsite) (evidence/verify/kroger-precision-marketing.json) | |
| C465 | No service- or network-scoped advertising revenue; 'Other sales' US$1,346M (FY2025) / US$1,177M (FY2024), perimeter broader parent proxy, 'mostly non-advertising components; media share undisclosed and never prorated'; US$1.5B alternative-profit operating profit context only; KPM profit +24% (Q2 2026); Investor Update gives no amount. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | Amounts and quotes confirmed (see F04 to F07). The audited note's 'Other' is US$1,129M (FY2025), 'primarily' third-party media revenue, data analytic services, specialty pharmacy and in-store health clinics. Neither filing line splits out media. | supported with qualification | (1) Delete 'mostly non-advertising components' from financials[0] and the scale cell: the filing states no split, so the share is undisclosed, not 'mostly' anything. (2) Add a financials row: fiscal 2025 'Other' (disaggregated revenues, audited note) US$1,129,000,000 (FY2024 US$3,062,000,000), definition quoted from R35, perimeter broader parent proxy. (3) Under scoring-rules.json's scale rules the literal mapping of either line is 'broader parent proxy' (a line spanning non-advertising revenue), which caps at 3; both amounts fall on the US$1–3 billion rung (4), so the cell would be 3 whicheve | Perimeter: not 'service' or 'network'; it should not be read as 'company-wide unallocated advertising line' either, because it is not an advertising line. The rule text fits broader parent proxy, but the paper should say that this Kroger line is mixed, with the media share undisclosed, unlike Instacart's 'advertising and other'. | capabilities.scale and financials (onsite and offsite; priority 3) (evidence/verify/kroger-precision-marketing.json) |
| C466 | Onsite: disclosure screen incomplete (KAP docs gated; KPM Specs and Policies refuses the fetch tool; ad-policies index 404); only platform-level change log, dated pages and a Direct Connect-only restatement policy found. Offsite: legacy Direct Connect restatement policy; screen for current KAP Direct Connect incomplete. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | dc-reporting is scoped 'for Campaigns run through Direct Connect only' and marked deprecated; resources/specs returned HTTP 403 to curl with a browser user agent on 3 November 2026. | supported with qualification | Offsite: fix the five-business-day paraphrase (see F17). Both rows: keep n/e for governance under the scoring rule ('an incomplete screen is n/e'). | The Direct Connect restatement policy is legacy (Prism-created Direct Connect campaigns) and is not evidence for KAP onsite governance. | capabilities.governance (onsite) and product_capabilities.offsite.governance (priority 2) (evidence/verify/kroger-precision-marketing.json) |
| C467 | estimation_method: 'last interaction within 14 days; click prioritised over view; sales booked on the exposure day'. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | 'reporting shows the ad cost on the date of exposure and the viewed sale on the date of purchase. For example, if a shopper is exposed to a TOA on December 27th, but purchases an attributable product on December 29th, the Kroger Ad Platform shows the ad cost on December 27th and the viewed sale on December 29th.' | contradicted | Replace 'sales booked on the exposure day' with 'ad cost booked on the exposure date and the attributed sale on the purchase date'. | Also add a measurement record for the managed-service default (see omissions). | measurement_records[0] (Attributed sales, KAP onsite) (evidence/verify/kroger-precision-marketing.json) |
| C468 | Glossary terms checked absent: 'Sales Uplift, control or test group, holdout, attribution window, halo, household penetration, rest of market, Sub-Commodity'. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | 'Control Household / Group A subset of a target audience that is intentionally excluded from a specific campaign or promotion during a test period. The Control Household or Group serves as a baseline or benchmark against which the effectiveness of the campaign is measured by comparing the results with the group that was exposed to the campaign (the test group).' 'Uplift The difference in responsiveness between a Test Group or Test Household (exposed to Campaign) and a Control Household / Group (not exposed).' 'A/B testing A 50/50 split run on a Campaign to test performance of an aspect of a Ca | contradicted | Move 'control or test group' and '(Sales) Uplift' to the present list, quoting both definitions. They are platform-level definitions in the Prism glossary; they say a control group is deliberately withheld from the target audience (a holdout-type construction) but not how it is allocated, so assignment stays unknown. | 'Sales Uplift' as a headword is absent from the Glossary but defined on the managed-service performance page. | sources RTL-KROGER-S08 reliability_notes (Glossary terms reported absent) (evidence/verify/kroger-precision-marketing.json) |
| C469 | why_profiled: KPM's MRC entry is 'a pre-audit in process since 6/2026'; timeline entries 2021 to 2026-Q4. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | MRC: 'Pre-audit in process as of 6/2026'. Roadmap and release-note dates as recorded. The legacy Direct Connect page of 5/8/2026 still calls DV360 'currently in Beta testing'; the roadmap marks Direct Connect in KAP for TTD, Yahoo, Disney and DV360 'Completed' in Q2 2026. | supported with qualification | Change 'since 6/2026' to 'as of 6/2026' in why_profiled and capabilities.us_availability, and the timeline label to 'status as of 6/2026 (start date not stated)'. Add timeline items: 2026-01-28 (Prism managed-service performance page documents Sales Uplift against tightly-matched control households, page date); 2026-05-08 (DV360 Direct Connect 'currently in Beta testing'); 2026-Q2 (roadmap: Direct Connect in KAP for TTD, Yahoo, Disney and DV360 completed). | Timeline years inferred by the profile (2024-03 beta; 2024-12 launch) are labelled as inferred and follow from the 14 January 2025 post. | why_profiled and timeline (evidence/verify/kroger-precision-marketing.json) |
| C470 | Default: in-flight attributable sales since 2023; rolling, view-reset 14-day window, no priority order; controlled lift for qualified Trade Desk campaigns (mirrored control, minimum spend unstated); legacy US$150K iROAS; PV360 covers TTD and Yahoo channels; assignment unknown. | Refute-mode verifier for evidence/profiles/kroger-precision-marketing.json re-opened the cited source. | As quoted in F12, F13, F28; PV360 channels include managed offsite Display Ad, Online Video Ad, Pinterest, Programmatic Audio, Programmatic CTV, Meta, Snapchat and Disney besides Direct Connect TTD and Yahoo. | supported with qualification | Replace 'PrecisionView 360 covers TTD and Yahoo channels' with 'PrecisionView 360 covers Direct Connect TTD and Yahoo and the managed offsite channels (display, online video, Pinterest, audio, CTV, Meta, Snapchat, Disney)'. Mark the US$150K rule as applying to legacy Prism Direct Connect campaigns only. Managed offsite campaigns (Display, CTV, audio and the rest) report through the Prism KPM Campaign Performance report, whose Performance Summary leads with Sales Uplift against tightly-matched control households; record that route's default separately. | product_capabilities.offsite.default_method (evidence/verify/kroger-precision-marketing.json) | |
| C471 | The Sponsored Search reporting API (definitions of 16 July 2026) defines attributed outcomes on 3-, 14- and 30-day click windows, reports online sales ('This is for online sales only') separately from in-store attributed sales, orders and units on the same windows, splits advertised from same-brand 'Other' items, counts new-to-brand orders over a one-year lookback, and adds view-through metrics only for video campaigns after an 'MRC defined view' of two seconds. Limitation as written: no default window is named and which window Ad Center shows by default is not stated. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | API page (updatedAt 2026-07-16; 137 element rows, 137 unique names): 'Total attributed sales revenue made from units sold over 14 day click. This is for online sales only.'; 'In-store attributed sales revenue based off the direct click of the advertised item, over 14 day click.'; ntbOrders14days 'over a one year look back window attributed over 14 days. This metric includes all placement types.'; viewThroughSales14days 'credited to a viewed ad (MRC defined view) within a 14-day window ... only for Advertised attributed sales, the attributed sales are measured after a 2-second view' (report typ | supported with qualification | Keep the claim; replace the limitation with: 'Walmart's help centre names the default: by default the standard attribution window is 14 days on a 14-day click basis, with 3- and 30-day click windows selectable (Advertiser reports, article 000008390). The default ROAS excludes in-store sales; a separate Omni ROAS adds digital and in-store attributed sales (Glossary, article 000008385).' Move the 14-day default from counterevidence to a documented parameter. | The 'attributedSales30days' omission of the word 'attributed' is confirmed as published. The help centre adds that where viewability cannot be measured it 'is inferred based on historical measurement data and features of the placement' (article 000012578), a modelled component of any view-based metric, including the Sponsored Videos view-through metrics. | RTL-WALMART-F07 (evidence/verify/walmart-connect.json) |
| C472 | Walmart's Sponsored Search developer documentation does not name a touch model, a match basis, a treatment of returns, cancellations, taxes or shipping, or a deduplication rule; its only attribution rule beyond the windows is the variant-click rule; spend basis is CPC times clicks. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | Developer pages: true as written ('As it works today whichever variant item is clicked on in the sponsored search will get the attribution.'; 'adSpend can be retrieved by the formula: CPC*clicks'). Walmart's help article 'Overview of attribution' (000012578; the target of the changelog's 'identity and attribution upgrades on June 1st' link) documents most of the missing parameters for this row: 'Walmart Connect offers a unified, deduplicated, last touch attribution model across Sponsored Search, Onsite Display, Walmart DSP and self-checkout store ad formats.'; 'For Sponsored Search, clicks are | supported with qualification | Restate as: 'Walmart's Sponsored Search developer documentation names no touch model, match basis, cancellation or return treatment or deduplication rule; its help centre does (Overview of attribution, article 000012578, read in a browser on 3 November 2026): since 1 June 2026 a unified, deduplicated last-touch model across Sponsored Search, Onsite Display, Walmart DSP and self-checkout store ads, with clicks the only eligible Sponsored Search touchpoint (before that date, last click not deduplicated with Display); online orders cancelled within 72 hours removed and returns not deducted (befor | Evidence class 'synthesis' with a bounded absence is acceptable for the developer pages, but the profile carries this absence into the Sponsored Search dimension-2 cell as row-level 'not documented'; that use is contradicted (see capabilities.attribution_disclosure). | RTL-WALMART-F08 (evidence/verify/walmart-connect.json) |
| C473 | The developer changelogs record 'identity and attribution upgrades on June 1st' 2026: new Sponsored Search in-store advertised, other, units and orders metrics on 3-, 14- and 30-day attribution, and a Display 'attributionModel' column returning 'fair_share' or 'last_touch'; the upgrades themselves are described only in a help-centre article that was not readable. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | Changelog quotes are accurate ('Walmart Connect introduced a number of identity and attribution upgrades on June 1st.'; in-store metrics list; 'The sum of “In-store Advertised Sales” and “In-store Other Sales” equals to “In-store Attributed Sales”. In-store Attributed Sales is not being expanded to additional report types.'; Display: 'Introducing attributionModel column to all report types. This column will return either fair_share or last_touch.'). The linked article is readable in a browser and describes the change. FAQ: 'Before June 1 ... Sponsored Search and Display were measured independe | supported with qualification | Restate as: 'Walmart Connect's changelogs record identity and attribution upgrades effective 1 June 2026 (new Sponsored Search in-store advertised, other, units and orders metrics; a Display attributionModel column). Its help article Overview of attribution (000012578) describes them: Sponsored Search moved from last click, not deduplicated with Display, to a unified, deduplicated last-touch model shared with Onsite Display, Walmart DSP and self-checkout store ads, with clicks still the only Sponsored Search touchpoint and clicks prioritised over views; Display moved from fair share across all | Confidence can move from medium to high for the description of the change (operator help centre, grade C, specific). | RTL-WALMART-F10 (evidence/verify/walmart-connect.json) |
| C474 | On 3 April 2026 Walmart Connect announced a move to a proprietary identity graph connecting in-store transactions, online activity and media engagement, individual-level targeting for Onsite Display, extension to DSP and Social over time and to CTV via VIZIO, and 'a new unified attribution model across Sponsored Search, Onsite Display and Walmart DSP'; no match method, matchable share or effective date is stated. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | Article dated 'April 03, 2026' (JSON-LD datePublished 2022-04-03; AEM modifyDate 2026-05-07): 'We are evolving from our existing identity framework to Walmart's more advanced, proprietary identity graph that connects in-store transactions, online activity and media engagement into a single, unified view across the shopper journey. In the future, this foundation will extend into connected TV environments through our integration with VIZIO.'; 'Display behavioral targeting is shifting to Walmart’s identity graph upgrade, and we’re introducing individual-level targeting for Onsite Display'; 'we’re | supported with qualification | Replace 'no ... effective date is stated' and the limitation 'the unified attribution model's rules are not described' with: 'The article states no match method, matchable share or date; Walmart's help centre (article 000012578) documents the unified model as deduplicated last touch, effective 1 June 2026, across Sponsored Search, Onsite Display, Walmart DSP and self-checkout store ads. The identity-graph and match-method elements remain undocumented.' The counterevidence sentence (separate online and in-store API metrics) is consistent with the help article: deduplication is across campaigns | The page-metadata inconsistency recorded by the profile (datePublished 2022) is confirmed. | RTL-WALMART-F11 (evidence/verify/walmart-connect.json) |
| C475 | MRC's letter of 26 September 2025 grants initial accreditation for Sponsored Search display, video and search clicks, CTR, impressions and invalid metrics (Sponsored Product, Brand and Video placements; GIVT; desktop, mobile web, mobile in-app); MRC's list shows Rendered Ad Impressions (Display, Video), Clicks: Search and CTR; no outcome metric is accredited and the letter does not cite the retail media guidelines. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | Letter dated 'September 26, 2025': 'granting initial accreditation to Walmart Connect's first-party measurement and reporting of Sponsored Search (`the Service') for Display, Video and Search Clicks, Click-Through Rate (CTR), Impressions and Invalid metrics originating from Sponsored Product, Sponsored Brand, and Sponsored Video ad placement types, inclusive of General Invalid Traffic (GIVT) filtration within Desktop, Mobile Web, and Mobile In-App environments.' Standards: MRC Minimum Standards, IVT Addendum v2.0 GIVT provisions, IAB Display Impression, Video Measurement and Click Measurement | supported with qualification | Add to the limitations: 'The letter's enclosure scopes the accredited metrics to those reported in the All Campaigns Dashboard and other Ad Center UI reports and in the IVT Report; the Sponsored Search API snapshot reports are not named, so the accreditation is not documented for API-delivered figures.' | The quarterly-update naming ('Sponsored Products service', RTL-A-S09) was not re-opened. | RTL-WALMART-F16 (evidence/verify/walmart-connect.json) |
| C476 | Walmart Connect's accreditation post (16 October 2025) says the accreditation is 'in addition to Walmart Connect’s rich onsite partnerships with third-party measurement leaders such as DoubleVerify and Integral Ad Science', without stating which metrics, placements or rows those partnerships cover or whether their tags run on Sponsored Search. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | Post: 'Our accreditation is in addition to Walmart Connect’s rich onsite partnerships with third-party measurement leaders such as DoubleVerify and Integral Ad Science.' (accurate). The help-centre ad policy 'Data use and privacy requirements' ('Last updated June 23, 2026') states the scope: under 'Approved onsite vendors', 'For Onsite Display: Tags for the below vendors are to be treated as viewability trackers regardless of format.' DoubleVerify and Integral AdScience: image or JavaScript 1x1 pixel, 'Approved for monitoring, blocking is not permitted', 'Viewability tracking is unavailable fo | supported with qualification | Append: 'Walmart's help centre (Data use and privacy requirements, last updated 23 June 2026) scopes the DoubleVerify and IAS tags to Onsite Display, as viewability trackers approved for monitoring only (blocking not permitted), on desktop and mobile web (no in-app viewability); no verification-tag route is documented for Sponsored Search.' The dimension-3 side field 'delivery-verification tags on-site' then reads: Sponsored Search, not documented; Onsite Display, DV and IAS monitoring tags documented (viewability, desktop and mobile web). | Answers verifier priority 5 on DV and IAS scope. | RTL-WALMART-F17 (evidence/verify/walmart-connect.json) |
| C477 | Buyers of Sponsored Search receive Walmart-defined aggregate reports, and no clean room, buyer-written query, event-level exposure export or independent replication is documented for Sponsored Search on walmartconnect.com, the developer portal or the measurement pages read; the help centre was a JavaScript shell to every route tried. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | The help centre renders in a browser and documents two further routes. Marketing Mix Modeling Data Portal (000001167): 'provides self-serve access to automated and standardized Walmart Connect MMM data across Sponsored Search, Onsite Display and Offsite media in Walmart Connect Ad Center'; 'Reports are standardized; dimensions and metrics cannot be customized in the request.'; 'The portal supports data requests for up to two years'; 'Data is set at the weekly level'; fields include campaign and ad group IDs and DMA; metrics 'Cost', 'Impressions', 'Clicks' ('Metrics are the same as in Walmart C | supported with qualification | Keep the bounded absence (no clean room, buyer-written query, event-level export or independent replication was found) and add: 'Walmart's help centre documents a self-serve MMM Data Portal giving advertisers standardised weekly, DMA-level exports of Sponsored Search cost, impressions and clicks for up to two years for their own marketing-mix modelling (media inputs only, no Walmart sales), and lists approved measurement partners connected by server-to-server integration activated by Walmart Connect (TransUnion MTA, Upwave brand lift, Crossix and IQVIA Rx sales lift, InMarket location lift, Ac | Dimension 3 needs a fresh look: the MMM portal is documented analysis on permitted exports with stated constraints, but the exports carry exposure and cost, not Walmart outcome data; the partner list is documented integration (not a name only) but without data terms. Whether that is rung 2, 3 or 4 is the scorer's call; rung 1 ('dashboard reports only') no longer describes the record. | RTL-WALMART-F20 (evidence/verify/walmart-connect.json) |
| C478 | The Onsite Display reporting API (28 October 2026) attributes sales, orders and units within 14 or 30 days of ad exposure, reports an attributionModel of FAIR_SHARE or LAST_TOUCH, a configurable new-buyer lookback, featured versus halo itemsets and in-store, pickup and delivery splits, and defines lift metrics against a control of unexposed households 'with similar digital & purchase propensity', with group sizes and a 'Significance level of overall sales lift'; counterfactual matched audience, assignment unknown. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | API (updatedAt 2026-10-28): 'Attribution window can be one of the following: days14, days30'; 'The attribution methodology used to calculate attributed metrics in the report. Values are FAIR_SHARE and LAST_TOUCH.'; buyerLookbackWindow 'months6' to 'months36'; pointOfSale 'Overall, In-store, Pickup, or Delivery'; controlGroupSize 'A sample of households that were not exposed to the ad campaign and have similar digital & purchase propensity as the test group.'; testGroupSize 'A sample of households that were exposed to the ad campaign'; overallSalesLift 'The percentage difference in dollar amoun | supported with qualification | Add: 'FAIR_SHARE is a legacy value: Display campaigns starting on or after 1 June 2026 use deduplicated last touch, fair share was fully deprecated on 1 September 2026, and since 1 June only viewable impressions (viewability inferred where it cannot be measured) and clicks are eligible touchpoints, clicks prioritised over views; before 1 June all delivered impressions were eligible.' Keep the control coding. | Control coding is correct: counterfactual 'matched audience' (propensity-similar unexposed households), assignment unknown (neither group-size definition discloses how exposure was allocated), unit household. The 'matched' label was not read as observational or randomised. No significance threshold, test, minimum sample or campaign eligibility rule is stated. | RTL-WALMART-F22 (evidence/verify/walmart-connect.json) |
| C479 | Walmart DSP announced 28 January 2021, 'available for select suppliers by the end of October' 2021; sold today as 'Powered by The Trade Desk' for open web, social, CTV, audio, gaming and more to endemic and non-endemic advertisers, with no DSP-specific window, match method or lift design on the pages read. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | Help article: 'Attribution is the process of identifying which media touchpoints a customer was exposed to prior to a conversion and assigning credit for the conversion to specific campaigns across Sponsored Search, Onsite Display and Walmart DSP.'; 'Walmart Connect provides 14-day or 30-day attribution window options for Display campaigns'; 'For Onsite Display and Walmart DSP, both clicks and viewable impressions are eligible touchpoints for attribution.'; 'The default ROAS for Onsite Display and Walmart DSP reported in Walmart Connect Ad Center reflects both online and in-store sales.'; Walm | supported with qualification | Replace 'with no DSP-specific window, match method or lift design on the pages read' with: 'Walmart's help centre documents the DSP's attribution: 14- or 30-day windows, clicks and viewable impressions as eligible touchpoints, deduplicated last touch shared with Sponsored Search and Onsite Display for campaigns starting 1 June 2026 or later (legacy fair share before; deprecated 1 September 2026), and a default Ad Center ROAS that includes online and in-store sales; spend in the MMM portal includes media and fees. No match method or lift design is documented for the DSP.' | The 2021 dates and the offsite-page quotes rest on workstream H and were not re-opened here. | RTL-WALMART-F27 (evidence/verify/walmart-connect.json) |
| C480 | The non-endemic page sells the DSP, Offsite Display audiences and in-store placements to automotive, financial services, insurance, QSR and travel brands, and its only results statement is 'Our data. Industry leading technology. Powerful results.'; inference: Walmart's closed loop measures purchases at Walmart, which such brands lack, and no page read says what outcome a non-endemic buyer receives. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | Help-centre ad policy, 'Approved onsite Measurement Partners': 'Acxiom Server-to-server integration Non-endemic attribution measurement'; 'Claritas Server-to-server integration Non-endemic attribution measurement'; 'Note: Server-to-server integrations must be activated by the Walmart Connect team. Reach out to your Walmart Connect representative prior to a campaign launch to ensure your campaign is measured by an approved partner.' | supported with qualification | Replace 'no page read says what outcome a non-endemic buyer receives' with: 'Walmart's own pages describe no outcome measurement for non-endemic buyers; its help centre lists Acxiom and Claritas as approved onsite measurement partners for non-endemic attribution measurement through server-to-server integrations that the Walmart Connect team activates, with parameters, data scope and terms not stated.' The non-endemic page itself was not re-opened. | The inference label stays appropriate for the remaining part (that Walmart's own closed loop cannot measure non-endemic outcomes). | RTL-WALMART-F30 (evidence/verify/walmart-connect.json) |
| C481 | A disclosure screen of the developer documentation found dated change logs, documented reporting revisions and timestamped definitions, but no restatement or correction policy, no metric dispute procedure and no methodology version history; the help centre could not be read, so the screen is incomplete (limitations add that Walmart's own pages do not state that conversions may be added retroactively). | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | Help article 000012578: 'Will historical data be updated to reflect the new last touch attribution methodology? No. Historical performance data will not be restated using the new last touch, deduplicated attribution methodology.'; 'we recommend using this update as a new baseline for evaluating campaign performance and avoiding direct comparisons with historical data prior to June 1, 2026'; 'Performance metrics continue updating through the end of the selected attribution window.'; 'metrics can continue updating within the selected attribution window to reflect conversions that occur after the | supported with qualification | Restate: 'The developer documentation and the help centre (readable in a browser) document dated change logs, a documented reporting revision with notice to buyers (the 1 June 2026 attribution change, with a statement that history will not be restated and advice not to compare across the change), retroactive updating of attributed metrics within the window, a dated methodology description for Sponsored Search delivery metrics (last updated 9 September 2026; no version history seen), and support routes (billing dispute, support case). No metric-specific dispute procedure or general correction p | The limitation that Walmart's pages do not state retroactive updating is contradicted by Walmart's own help article. Dimension 11 should stay n/e only until the remaining help-centre and terms pages are read, which is now feasible in a browser. | RTL-WALMART-F35 (evidence/verify/walmart-connect.json) |
| C482 | Documented with values: window, in-store inclusion, new-to-brand lookback, halo; partly: sales basis; not documented: touch model, match basis, deduplication. Touch model 'not named for Sponsored Search'; deduplication 'not stated'; returns, cancellations 'not stated'; no default window named. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | Touch model: 'Walmart Connect offers a unified, deduplicated, last touch attribution model across Sponsored Search, Onsite Display, Walmart DSP and self-checkout store ad formats.'; 'only clicks can receive credit for a Sponsored Search campaign'; before 1 June 2026, 'legacy last click attribution model that was not deduplicated with Display'. Deduplication: 'Attributed metrics will no longer be duplicated across channels.'; 'Full credit and the full value of the conversion is assigned to one campaign'. Sales basis: 'any online orders canceled within 72 hours of purchase are removed from repor | contradicted | Rewrite the cell: window click 3, 14 and 30 days, default 14-day click (Ad Center); view-through for Sponsored Videos only. Touch model: deduplicated last touch across Sponsored Search, Onsite Display, Walmart DSP and self-checkout store ads since 1 June 2026, clicks the only eligible Sponsored Search touchpoint, advertised items before halo and clicks before views (legacy last click, not deduplicated with Display, before 1 June 2026). Match basis: not stated. In-store: reported separately; default ROAS online only; Omni ROAS adds in-store. New-to-brand: one-year lookback. Sales basis: attribu | Evidence grade stays C (help centre is vendor documentation). The record the profile scored from moved from 'touch model and deduplication not documented' to seven of eight parameters documented; under the frozen dimension-2 anchors that is the difference between rung 2 ('one of window or touch model stated') and rung 4 ('six or seven of the eight documented with values'), cap C = 4. This is a ban | capabilities.attribution_disclosure (Walmart Connect Sponsored Search, dimension 2) (evidence/verify/walmart-connect.json) |
| C483 | MRC accreditation completed (initial grant, 26 September 2025) for Sponsored Search delivery metrics in desktop, mobile web and mobile in-app; standards listed; retail media guidelines not cited; viewability, SIVT and outcome metrics outside scope; no AAM or IAB Europe instrument. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | Letter and list as in F16; enclosure scopes the accredited metrics to the 'All Campaigns Dashboard' and other UI reports and the 'IVT Report'. DoubleVerify and IAS tags are approved for Onsite Display only, monitoring only. | supported with qualification | Add to the cell: 'Accredited metrics as enumerated for Ad Center UI reports and the IVT report; API-delivered reports not named in the enclosure. Third-party verification: DoubleVerify and IAS tags documented for Onsite Display only (monitoring, no blocking), none for Sponsored Search.' | Accreditation covers delivery metrics in the environments sold, which matches rung 4 as written; it attests nothing about attributed sales, in-store matching or new-to-brand, as the cell says. | capabilities.assurance (Walmart Connect Sponsored Search, dimension 10) (evidence/verify/walmart-connect.json) |
| C484 | Default: closed-loop attribution with no counterfactual; separate Search Incrementality offer undocumented in method; QBR single-brand analysis not a default. Identity basis asserted at network level; match method and matchable share not documented. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | WAMM: 'uses a proprietary, statistical approach to measure the incremental impact of Walmart Connect media on omnichannel sales'; 'Eligibility varies based on historical campaign data and investment.'; 'This is a directional analysis and provides proxy estimate only on incremental revenue based on media spend.'; metrics 'Incremental sales and units', 'Incremental ROAS'; 'Store tactics are not included'; 'All traceable orders from Walmart’s site and app and Store purchases are included in the report.' Attribution article: 'Because Walmart has both the media touchpoints and the retail data, we c | supported with qualification | Dimension 1 cell: add WAMM as a Walmart-run, modelled incrementality report on request for eligible advertisers (target contrast incremental, modelled; assignment none; counterfactual model-based, construction not documented; self-described 'directional' 'proxy estimate'). Dimension 4 cell: add that the June 2026 'identity and attribution upgrades' article describes attribution rules only, no identity or match method, and that WAMM counts 'traceable' orders, implying untraceable store purchases exist without a stated share. | Neither addition changes the documented default (attribution, no counterfactual) or the identity rung (basis asserted). WAMM's incremental ROAS is a model output, not a return, and must not be called iROAS from a controlled test. | capabilities.default_method and capabilities.identity_basis (Walmart Connect Sponsored Search, dimensions 1 and 4) (evidence/verify/walmart-connect.json) |
| C485 | Governance: bounded screen incomplete; no restatement or correction policy, no dispute procedure, no methodology version history; Walmart's pages do not state retroactive conversions; help centre unreadable, so n/e. Buyer analysis: Walmart-defined aggregate reports only; DV and IAS 'onsite partnerships' with scope not stated. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | See F35 (no-restatement statement, retroactive window updates, methodology description last updated 9 September 2026, billing-dispute and support-case routes), F20 (MMM Data Portal; server-to-server measurement partners) and F17 (DV and IAS tags scoped to Onsite Display). | supported with qualification | Apply the F35, F20 and F17 corrections to these two cells. Complete the dimension-11 screen in a browser (Billing dispute, Open a case with Advertiser Help, Walmart Connect Advertising Amended IAB Terms, Walmart Connect Product Terms) before scoring it. | capabilities.governance and buyer_analysis (Walmart Connect Sponsored Search, dimensions 11 and 3) (evidence/verify/walmart-connect.json) | |
| C486 | Onsite Display is separate in part because it has 'its own attribution model (fair-share or last-touch, 14/30 days of exposure)'; touch model FAIR_SHARE or LAST_TOUCH reported per row since 1 June 2026; returns treatment and deduplication not stated. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | 'Walmart Connect offers a unified, deduplicated, last touch attribution model across Sponsored Search, Onsite Display, Walmart DSP and self-checkout store ad formats.'; 'Any Onsite Display and Walmart DSP campaign starting June 1 or later will have deduplicated last touch attribution.'; 'Starting September 1, 2026, fair share attribution is completely deprecated'; 'Only viewable impressions* are eligible for Display attribution'; 'Cancellations occurring within 72 hours will be deducted from attributed sales reporting. Returns will not be deducted' (before June 1 Display removed cancellations | contradicted | In the note, replace 'its own attribution model (fair-share or last-touch, 14/30 days of exposure)' with 'its own reporting and lift elements, on the unified deduplicated last-touch model it shares with Sponsored Search and the DSP since 1 June 2026 (14- or 30-day windows over viewable impressions and clicks; fair share legacy, deprecated 1 September 2026)'. In the Onsite Display dimension-2 cell add: deduplication across channels and campaigns; cancellations within 72 hours deducted and returns not deducted since 1 June 2026 (cancellations and returns within six months removed before); viewab | The row separation itself still holds (separate purchase, API, lift elements and assurance stage); only the stated reason 'its own attribution model' is wrong. | product_capabilities_note and product_capabilities['Walmart Connect Onsite Display'].attribution_disclosure (evidence/verify/walmart-connect.json) |
| C487 | Attribution window documented only for pooled 14-day statistics; 'touch model, match basis, in-store inclusion (online only on the offsite page versus "online and in-store" in 2021), sales basis, deduplication and halo: not documented for the DSP'. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | As in F27: 14- or 30-day windows for Display campaigns; clicks and viewable impressions eligible for Onsite Display and Walmart DSP; deduplicated last touch for DSP campaigns starting 1 June 2026 or later; 'The default ROAS for Onsite Display and Walmart DSP reported in Walmart Connect Ad Center reflects both online and in-store sales.'; cancellation and return rules shared with Display; DSP spend in the MMM portal 'includes media spend and fees'. | contradicted | DSP dimension-2 cell: window 14 or 30 days (viewable impressions and clicks); touch model deduplicated last touch since 1 June 2026 (fair share before; deprecated 1 September 2026); in-store included in the default ROAS; sales basis 72-hour cancellations deducted, returns not; deduplication across Sponsored Search, Display and DSP; halo through measured item sets; match basis not stated; new-to-brand lookback for DSP not stated. Evidence grade C (help centre) for these parameters, not D. Dimension 1 unchanged (attribution; no DSP lift design documented). | The offsite page's 'verified Walmart transactions online' remains a pooled marketing statement; the help article is the more specific operator documentation and governs the in-store inclusion field. | product_capabilities['Walmart DSP'].default_method and .attribution_disclosure (evidence/verify/walmart-connect.json) |
| C488 | Store Ads: 'Not an attribution product on the page ... no window, match basis or sales basis'. Non-endemic: buyer analysis 'Not documented for this row'. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | 'Walmart Connect offers a unified, deduplicated, last touch attribution model across Sponsored Search, Onsite Display, Walmart DSP and self-checkout store ad formats.' Ad policy: Acxiom and Claritas 'Server-to-server integration' for 'Non-endemic attribution measurement'. | supported with qualification | Store Ads cell: 'The store-ads page describes lift, not attribution; Walmart's help centre places self-checkout store ad formats inside the unified, deduplicated last-touch attribution model (window and parameters for that format not stated); other in-store formats are not covered.' Non-endemic cell: 'Help centre lists Acxiom and Claritas as approved onsite measurement partners for non-endemic attribution via server-to-server integrations activated by Walmart Connect; data scope and terms not stated.' | product_capabilities['Walmart Connect Store Ads'].attribution_disclosure and ['Walmart Connect non-endemic advertising'].buyer_analysis (evidence/verify/walmart-connect.json) | |
| C489 | Timeline: 2026-06-01 'identity and attribution upgrades' with new in-store metrics and an attributionModel column, 'details in an unreadable help article'. Financials: FY26 'nearly $6.4 billion, including VIZIO'; Q2 FY27 growth 38% / 38% / 43% / 20%. | Refute-mode verifier for evidence/profiles/walmart-connect.json re-opened the cited source. | Financial rows for FY26 and Q2 FY27 match the releases. The help article gives the content of the 1 June change and a second dated step: 'Starting September 1, 2026, fair share attribution is completely deprecated'. | supported with qualification | Rewrite the 2026-06-01 event as a methodology change: 'Unified, deduplicated last-touch attribution across Sponsored Search, Onsite Display, Walmart DSP and self-checkout store ads; Display limited to viewable impressions and clicks; 72-hour cancellations deducted, returns not; no restatement of history.' Add '2026-09-01, retirement: fair share attribution fully deprecated for Onsite Display and Walmart DSP (campaigns still on it lose attribution after 31 August)'. Optionally add '2026-09-09, methodology document: Description of methodology for Sponsored Search last updated'. | FY24 and FY25 financial rows were not re-opened (stream D verifier). | timeline (2026-06-01 entry) and financials (evidence/verify/walmart-connect.json) |
| C490 | Sam's Club Connect / privacy = 1 (grade C) | Separate refute-mode cell verifier for scoring slice A1. | The only shopper-facing policy cited is Walmart's, whose scope section covers Walmart stores and sites and names Sam's Club as a separate corporate-family company; RTL-J-F14's 'applies identically ... to Sam's Club Connect' is therefore wrong, and the MAP notice concerns advertisers' site visits. The cell becomes n/e. | contradicted | Sam's Club Connect / privacy -> ne | Looked for a privacy record covering Sam's Club members' data in Sam's Club Connect advertising: none read. Walmart's Customer Privacy Notice (updated 20 August 2026) 'covers our Walmart retail stores in the United States and Puerto Rico, and our websites, mobile apps, and services where this Privacy Notice is posted' and treats Sam's Club as another company in the corporate family ('We may disclose your personal information with other retail and non-retail companies (e.g., Sam's Club and One Fi | data/scores.json or data/matrix-extra.json: Sam's Club Connect, privacy |
| C491 | Roundel / privacy = 1 (grade C) | Separate refute-mode cell verifier for scoring slice A1. | Value 1 -> 2, grade C -> D: the scorer read Roundel's sentence as naming no control, but the full sentence ends 'to measure campaign performance without exposing personally identifiable information (PII)', a product-specific control statement equivalent to Best Buy's, so the cell is not 'generic company policy only'. | supported with qualification | Roundel / privacy -> 2 (grade D) | Anchor 2: product-specific notice without named legal frameworks: Roundel's measurement page answers 'How do you ensure measurement remains privacy-safe?' with 'We use privacy-conscious methodologies and secure data collaboration environments to measure campaign performance without exposing personally identifiable information (PII).' [RTL-U-S028; full sentence per verify/H.json RTL-H-F15 and re-read on 3 November 2026], a control statement of the same kind as Best Buy's clean-room sentence (2). | data/scores.json or data/matrix-extra.json: Roundel, privacy |
| C492 | DG Media Network / placement_coverage = 1 (grade D) | Separate refute-mode cell verifier for scoring slice A1. | Value 1 -> 2: the anchor-1 reading rested on 'one dashboard (pooled)', which no page documents (the off-site dashboard adds a Tactics dimension); the on-site class has a documented reporting surface with incomplete parameters, the same position as CVS Media Exchange's 2. | supported with qualification | DG Media Network / placement_coverage -> 2 (grade D) | Anchor 2: the row serves on-site and in-app, off-site (Meta; YouTube and DV360; DoorDash) and in-store audio [RTL-E2-S32, RTL-E2-S33, RTL-E2-S50, RTL-U-S072]. The primary on-site class has its own documented reporting surface: the on-site page offers 'Closed-loop reporting and incrementality with our Reporting Dashboard' with engagement metrics (Viewability, Impressions, CPC, CTRs) and sales metrics (Conversion Rate, % New-to-Brand Shoppers, undesigned lift labels) [RTL-E2-S32], but no exposure- | data/scores.json or data/matrix-extra.json: DG Media Network, placement_coverage |
| C493 | CVS Media Exchange / us_availability = n/e | Cell verifier for scoring slice A1 scanned the n/e cells. | 3, grade D, date 2026-11, avail live: CMX's own In-Store Advertising page states '9K+ locations CVS is the largest retail pharmacy chain in America' and '85% of the U.S. population lives within 10 miles of a CVS, leading to over 5 million interactions daily with our customers' (https://www.cvsmediaexchange.com/solutions/instore): US availability documented on the row's own page (the scorer's convention); production use by a named US advertiser in the period not documented (the 14% and 13% ROAS f | structural | CVS Media Exchange / us_availability: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | CVS Media Exchange, us_availability |
| C494 | Roundel / validation = n/e | Cell verifier for scoring slice A1 scanned the n/e cells. | 1, grade D, date 2026-11, avail live: accuracy claims only: the measurement page says a pet care brand used Roundel's closed-loop measurement 'to connect media exposure to verified sales outcomes' and, under 'Designed for transparency', 'Clear methodologies and verified outcomes help brands see what's driving results' (https://roundel.com/measurement/, re-read 3 Nov 2026); no validation design. Coded as the scorer coded CVS ('Closed-loop measurement tied to verified outcomes' = 1). | structural | Roundel / validation: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Roundel, validation |
| C495 | Walgreens Advertising Group / validation = n/e | Cell verifier for scoring slice A1 scanned the n/e cells. | 1, grade D, date 2026-11, avail live: accuracy claim: 'Leverages wag's 1PD and closed-loop measurement for accuracy & efficiency.' (In-Store Solutions page, Coupons; re-read 3 Nov 2026 at https://www.walgreensadvertisinggroup.com/home/solution-detail.html); the scorer's n/e says no accuracy claim exists on the pages read, but the in-store page is one of them (RTL-U-S060). | structural | Walgreens Advertising Group / validation: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Walgreens Advertising Group, validation |
| C496 | walmart-connect / buyer_analysis = 4 (grade C) | Separate refute-mode cell verifier for scoring slice A2. | Ruling 1: the MMM Data Portal (re-read in the pane) exports only Cost, Impressions and Clicks ('Metrics are the same as in Walmart Connect Ad Center'), so it is not buyer-run analysis of the inventory's outcomes; the API snapshot aggregates give anchor 3. The value rests on the dated API pages, so the date is 2026-07-16; the MMM and tag records sit beside the cell. | contradicted | walmart-connect / buyer_analysis -> 3 (grade C) | Anchor 3: aggregates delivered through an API on the operator's queries only: Sponsored Search API snapshot reports for 'Full-Service' and 'Reporting Only' API partners, with Walmart-defined attributed sales, orders, units and new-to-brand elements and an attributedPurchases report [RTL-WALMART-S10, RTL-WALMART-S12, RTL-WALMART-S13; RTL-WALMART-F20 qualified]. Blocks 4: no buyer-run analysis of Walmart outcome data with output rules is documented. The Marketing Mix Modeling Data Portal gives sta | data/scores.json or data/matrix-extra.json: walmart-connect, buyer_analysis |
| C497 | walmart-connect / identity_basis = 3 (grade C) | Separate refute-mode cell verifier for scoring slice A2. | Ruling 2: the help-centre sentence (confirmed verbatim in the pane) asserts a customer-level link for online and in-store purchases but names no identifier, key or join; under the anchors that is 'the basis asserted only' (1), as the session coded Amazon Sponsored Products, Macy's, Orange Apron, Instacart, Albertsons onsite and Citi. Reported outcome channels cannot lift a rung whose basis is undocumented. | contradicted | walmart-connect / identity_basis -> 1 (grade C) | Anchor 1: the basis asserted only. The help centre says 'When a customer is served an ad and later makes a purchase (online or in-store) our closed-loop measurement connects that media exposure to the transaction' and 'Because Walmart has both the media touchpoints and the retail data, we can connect ad interactions, including clicks and views*, directly to sales', naming no identifier, key or join for either channel [RTL-VWALMART-S01]. The API reports online ('This is for online sales only') an | data/scores.json or data/matrix-extra.json: walmart-connect, identity_basis |
| C498 | Walmart Connect Onsite Display / buyer_analysis = 4 (grade C) | Separate refute-mode cell verifier for scoring slice A2. | Ruling 1: the MMM portal's 'Display (Onsite & Offsite)' export carries Cost, Impressions and Clicks only; the Display reporting API's operator-defined aggregates give anchor 3. Date: the API definitions (28 Oct 2026) the value rests on. | contradicted | Walmart Connect Onsite Display / buyer_analysis -> 3 (grade C) | Anchor 3: aggregates delivered through an API on the operator's queries only: the Display reporting API delivers Walmart-defined attributed sales, orders and units by featured and halo itemset and point of sale, and lift elements for campaign or campaign-group studies (definitions updated 28 October 2026) [RTL-WALMART-S15], and the Display Advertising API (26 February 2025) opened access to Partner Network platforms ('Beta partners Pacvue, Skai, and DataCaciques') [RTL-WALMART-S04]. Blocks 4: no | data/scores.json or data/matrix-extra.json: Walmart Connect Onsite Display, buyer_analysis |
| C499 | Walmart Connect Onsite Display / identity_basis = 3 (grade C) | Separate refute-mode cell verifier for scoring slice A2. | Ruling 2: the customer-level sentence names no key (see Sponsored Search); the API's 'households' define the lift study's test and control groups, not the basis of the attributed metrics (the session's Instacart coding treats a lift design's unit the same way). Basis asserted only: 1. Sensitivity: if the session counts the household unit as the documented basis, the cell is 3. | contradicted | Walmart Connect Onsite Display / identity_basis -> 1 (grade C) | Anchor 1: the basis asserted only. The help article that names Onsite Display says 'When a customer is served an ad and later makes a purchase (online or in-store) our closed-loop measurement connects that media exposure to the transaction', naming no identifier, key or join, and its viewable-impression touchpoints have no stated link to a purchaser [RTL-VWALMART-S01]. The Display API reports outcomes by point of sale and defines lift test and control groups as households [RTL-WALMART-S15]: the | data/scores.json or data/matrix-extra.json: Walmart Connect Onsite Display, identity_basis |
| C500 | Walmart Connect Onsite Display / privacy = 1 (grade C) | Separate refute-mode cell verifier for scoring slice A2. | Ruling 6: the ad policy (re-read in the pane) is a product-specific notice of data controls for the tags approved 'For Onsite Display' and DCM tags targeted to Walmart audiences, with no law named: anchor 2, the coding the session gives CVS Media Exchange's ads policy (same no-combining and RDP clauses), Macy's ad terms and Kroger's FAQ. No consent basis or sensitive-category rule, so not 3. | contradicted | Walmart Connect Onsite Display / privacy -> 2 (grade C) | Anchor 2: a product-specific notice without named legal frameworks. Walmart Connect's ad policy 'Data use and privacy requirements' (last updated 23 June 2026) sets data controls on the tags and vendors it approves for Onsite Display ('For Onsite Display: Tags for the below vendors are to be treated as viewability trackers'; DoubleVerify and IAS 'Approved for monitoring, blocking is not permitted'; Campaign Manager impression pixels): the advertiser 'shall not ingest or receive, collect, use, se | data/scores.json or data/matrix-extra.json: Walmart Connect Onsite Display, privacy |
| C501 | Walmart Connect Offsite Display / buyer_analysis = 4 (grade C) | Separate refute-mode cell verifier for scoring slice A2. | Ruling 1: the MMM export (media inputs only) does not give 4. The row's highest documented route is the LiveRamp clean-room closed-loop measurement for Meta: aggregates computed in a clean room on Walmart's queries, the rung-3 reading the session gave Amazon Retail Ad Service ('Ad measurement takes place within AWS Clean Rooms'); grade D caps at 3. Not 1 as the scorer's fallback had it. | contradicted | Walmart Connect Offsite Display / buyer_analysis -> 3 (grade D) | Anchor 3 (also the grade-D cap): aggregates computed in a clean room on the operator's queries only: 'Closed-loop measurement via LiveRamp's data clean room is now available for Meta', reporting 'both online and in-store sales at Walmart' to advertisers (General Mills: the integration 'gave us new transparency into how social contributed to Walmart sales outcomes'); the clean room is Walmart's, with no buyer query or output rules documented, and 'An API for tech partners is also planned' [RTL-H- | data/scores.json or data/matrix-extra.json: Walmart Connect Offsite Display, buyer_analysis |
| C502 | Walmart Connect Offsite Display / privacy = 1 (grade C) | Separate refute-mode cell verifier for scoring slice A2. | Ruling 6: the ad policy's 'Approved Offsite Vendors' and 'Connected TV' sections and its DCM clause ('targeted to Walmart audience onsite or offsite') make it a product-specific notice for off-site media without a named law: anchor 2. Not 3 (no consent basis or sensitive-category rule). | contradicted | Walmart Connect Offsite Display / privacy -> 2 (grade C) | Anchor 2: a product-specific notice without named legal frameworks. Walmart Connect's ad policy 'Data use and privacy requirements' (last updated 23 June 2026) sets data controls on the tags and vendors it approves for off-site media ('Approved Offsite Vendors': ad servers, DCO and high-impact vendors, DoubleVerify and IAS JavaScript tags 'Approved for monitoring, blocking is not permitted', 'Approved offsite Measurement Partners'; 'Connected TV' vendors): the advertiser 'shall not ingest or rec | data/scores.json or data/matrix-extra.json: Walmart Connect Offsite Display, privacy |
| C503 | Walmart Connect Store Ads / default_method = 1 (grade C) | Separate refute-mode cell verifier for scoring slice A2. | Ruling 4: dimension 1 scores the row's default for the formats it sells (session ruling 3: only the rung every product meets is evidenced). The store-ads page (re-read) names only 'closed-loop measurement and sales lift for store ads', a bare term and an undesigned label (n/e and nothing under ruling 8); the help article covers only 'self-checkout store ad formats'. Screens, TV Wall, audio, demos, events and sampling meet no documented rung, so the row is n/e. | contradicted | Walmart Connect Store Ads / default_method -> ne | Looked for the row's default outcome report across the formats sold (digital screens incl. Deli and Bakery, TV Wall, self-checkout, in-store audio and Walmart Radio, demos, events, sampling): the store-ads page names only 'closed-loop measurement and sales lift for store ads' and 'Store ads measurement includes both brand and sales lift', a bare term and an undesigned lift label that are n/e and earn nothing under session ruling 8 ('Sales lift validated at the 80% confidence level or above' is f | data/scores.json or data/matrix-extra.json: Walmart Connect Store Ads, default_method |
| C504 | Walmart DSP / buyer_analysis = 4 (grade C) | Separate refute-mode cell verifier for scoring slice A2. | Ruling 1: the MMM export ('Offsite Display (Walmart DSP)': cost, impressions, clicks) is media inputs only; no API, clean room or log-level export is documented for the DSP, and the ad policy's offsite measurement partners (pixel or Walmart-activated server-to-server) are not documented as accessing Walmart outcome data. Dashboard reports only: anchor 1. | contradicted | Walmart DSP / buyer_analysis -> 1 (grade C) | Anchor 1: dashboard reports only: the help centre reports the DSP's default ROAS in Walmart Connect Ad Center [RTL-VWALMART-S01] and the offsite page promises 'performance dashboards and reports' [RTL-U-S010]; no API, clean room, buyer query or log-level export is documented for the DSP. Not 2: the ad policy's 'Approved offsite Measurement Partners' connect by 1x1 pixel or by 'Server-to-server integration' that 'must be activated by the Walmart Connect team' (among them Attain 'Sales lift measur | data/scores.json or data/matrix-extra.json: Walmart DSP, buyer_analysis |
| C505 | Walmart DSP / privacy = 1 (grade C) | Separate refute-mode cell verifier for scoring slice A2. | Ruling 6: the ad policy's offsite sections (Dotdash Meredith 'offers Walmart DSP advertisers ...'; DV/IAS JavaScript tags; offsite measurement partners) and its DCM clause make it a product-specific notice for the DSP without a named law: anchor 2. | contradicted | Walmart DSP / privacy -> 2 (grade C) | Anchor 2: a product-specific notice without named legal frameworks. Walmart Connect's ad policy 'Data use and privacy requirements' (last updated 23 June 2026) sets data controls on the tags and vendors it approves for off-site buying ('Approved Offsite Vendors', including vendors that serve 'Walmart DSP advertisers', DoubleVerify and IAS JavaScript tags 'Approved for monitoring, blocking is not permitted', 'Approved offsite Measurement Partners'; 'Connected TV' vendors): the advertiser 'shall n | data/scores.json or data/matrix-extra.json: Walmart DSP, privacy |
| C506 | DoorDash Ads CPG Sponsored Products and Sponsored Brands / placement_coverage = 3 (grade C) | Separate refute-mode cell verifier for scoring slice A3. | Value 3 rested on counting Offsite Ads as a class of this row; DoorDash's overview lists Offsite Ads as its own product (with Banners, Spotlight and Promotions) and the register's purchased service is SP and SB only, so the row serves one class with documented measurement: 4 (cap C). Carrying another product's measurement gap into the row breaks the no-transfer rule. | supported with qualification | DoorDash Ads CPG Sponsored Products and Sponsored Brands / placement_coverage -> 4 (grade C) | Anchor 5 met on content, capped at 4 by grade C: the row (Sponsored Products and Sponsored Brands, the register's purchased service) serves one class, on-site DoorDash placements (Sponsored Products: collections, category pages, search results, DoubleDash; Sponsored Brands: store page, in-store search, category pages), and documents its measurement with exposure capture (impressions and clicks 'deduplicated within a five-minute window for a given customer and store'), match (orders by 'users who | data/scores.json or data/matrix-extra.json: DoorDash Ads CPG Sponsored Products and Sponsored Brands, placement_coverage |
| C507 | DoorDash Ads CPG Sponsored Products and Sponsored Brands / governance = 1 (grade C) | Separate refute-mode cell verifier for scoring slice A3. | The pages screened include no DoorDash Ads terms or IO terms; under session ruling 1 the bounded absence cannot be scored. | contradicted | DoorDash Ads CPG Sponsored Products and Sponsored Brands / governance -> ne | Session ruling 1 (frozen dimension-11 definition: 'Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'): E1's screen read help-centre and resource pages and releases only (overview, Brand Center dashboard and reporting, key metrics, product page, the 24 March and 2 September 2026 releases, the ghost-ads resource, the offer page) and found no change log, restatement policy, documented revision, dispute procedure or method | data/scores.json or data/matrix-extra.json: DoorDash Ads CPG Sponsored Products and Sponsored Brands, governance |
| C508 | DoorDash Ads restaurant Sponsored Listings / us_availability = 3 (grade C) | Separate refute-mode cell verifier for scoring slice A3. | The cell rests on locale only, which the slice's own convention (ruling 7) leaves n/e. | contradicted | DoorDash Ads restaurant Sponsored Listings / us_availability -> ne | No US statement on the row's own pages: the en-US merchant help article (re-read in the browser pane on 3 November 2026) and the US-locale offer page carry only locale indicia (en-US path, a toll-free support number), the ghost-ads resource is an en-us page with no geography, and the merchant learning page's '15.7K merchants' study (April 2023) names no country; under the dimension-9 convention (locale, currency, addresses and 'global' claims left n/e) US availability is not documented for resta | data/scores.json or data/matrix-extra.json: DoorDash Ads restaurant Sponsored Listings, us_availability |
| C509 | DoorDash Ads restaurant Sponsored Listings / governance = 1 (grade C) | Separate refute-mode cell verifier for scoring slice A3. | Two pages, no merchant advertising terms read: incomplete screen under ruling 1. | contradicted | DoorDash Ads restaurant Sponsored Listings / governance -> ne | Session ruling 1 (frozen dimension-11 definition: 'Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'): E1's screen read the en-US Sponsored Listings help article and the merchant learning page only and found no change log, restatement policy, documented revision, dispute procedure or methodology history [RTL-E1-S50, RTL-E1-S20]. No advertiser terms, IO terms or advertising agreement for this service were read, so the s | data/scores.json or data/matrix-extra.json: DoorDash Ads restaurant Sponsored Listings, governance |
| C510 | eBay Promoted Listings General / governance = 2 (grade C) | Separate refute-mode cell verifier for scoring slice A3. | The scorer itself records the screen as incomplete (help-centre wall) and no eBay advertising terms were read; ruling 1 makes an incomplete screen n/e even where items were found. | contradicted | eBay Promoted Listings General / governance -> ne | Session ruling 1 (frozen dimension-11 definition: 'Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'): E2's screen read the attribution page, the January 2026 seller announcement, the export-site hub and the Marketing API overview; the ebay.com help centre returned a verification wall [RTL-E2-S02]. It found one dated, disclosed revision (the any-buyer definition 'As of January 13, 2026') and 'eBay Advertising attributi | data/scores.json or data/matrix-extra.json: eBay Promoted Listings General, governance |
| C511 | Etsy Ads / identity_basis = 2 (grade C) | Separate refute-mode cell verifier for scoring slice A3. | The scorer read the buyer link as implied; the set-up article (RTL-E2-S04, cited in E2-F39, not re-opened by the E2 verifier) states it: 'it was that click that brought your buyer to your shop in the first place'. Same basis as Etsy Offsite Ads and eBay Priority (3): value 2 -> 3. | supported with qualification | Etsy Ads / identity_basis -> 3 (grade C) | Anchor 3: identity basis and matched channel documented, share not stated: orders are attributed 'when someone interacts with an ad and then buys any item in your shop within 30 days', and the set-up article ties the order to the clicking buyer ('the order is attributed to your Etsy Ads campaign since it was that click that brought your buyer to your shop in the first place'); the transaction source is the seller's Etsy shop orders, flagged on the Orders page with 'a small megaphone icon' [RTL-U | data/scores.json or data/matrix-extra.json: Etsy Ads, identity_basis |
| C512 | Etsy Ads / governance = 2 (grade C) | Separate refute-mode cell verifier for scoring slice A3. | Etsy's terms of use, seller policy and fees policy were not read: incomplete screen. | contradicted | Etsy Ads / governance -> ne | Session ruling 1 (frozen dimension-11 definition: 'Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'): E2's screen read four help articles (stats, set-up, performance, Offsite Ads) and found one disclosed reporting rule ('We filter out bot traffic again at the end of the day, which may lower your visit count') and no restatement policy, dispute procedure or methodology history [RTL-E2-S03, RTL-E2-S04, RTL-U-S076, RTL-U | data/scores.json or data/matrix-extra.json: Etsy Ads, governance |
| C513 | Etsy Offsite Ads / governance = 1 (grade C) | Separate refute-mode cell verifier for scoring slice A3. | No Etsy terms or fees policy read: incomplete screen. | contradicted | Etsy Offsite Ads / governance -> ne | Session ruling 1 (frozen dimension-11 definition: 'Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'): E2's screen read the Offsite Ads, stats, set-up and performance help articles and found no change log, restatement policy, revision, dispute procedure or methodology history; 'These partners may change over time' concerns placements [RTL-U-S077, RTL-E2-S03, RTL-E2-S04, RTL-U-S076; RTL-E2-F40]. No advertiser terms, IO | data/scores.json or data/matrix-extra.json: Etsy Offsite Ads, governance |
| C514 | Gulp Media Network / governance = 1 (grade D) | Separate refute-mode cell verifier for scoring slice A3. | Three marketing pages, no terms: incomplete screen. | contradicted | Gulp Media Network / governance -> ne | Session ruling 1 (frozen dimension-11 definition: 'Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'): E2's screen read the network home page, the non-endemic page and 7-Eleven's release of 12 November 2024 and found no versioning, restatement, revision, dispute or separation statement [RTL-E2-S39, RTL-E2-S40, RTL-U-S094; RTL-E2-F48]. No advertiser terms, IO terms or advertising agreement for this service were read, so | data/scores.json or data/matrix-extra.json: Gulp Media Network, governance |
| C515 | Staples Media Network / governance = 1 (grade D) | Separate refute-mode cell verifier for scoring slice A3. | One marketing page (the footer's general staples.com 'Terms & Conditions' were not read): incomplete screen. | contradicted | Staples Media Network / governance -> ne | Session ruling 1 (frozen dimension-11 definition: 'Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'): E2's screen read the single network page and found nothing [RTL-U-S090; RTL-E2-F51]. No advertiser terms, IO terms or advertising agreement for this service were read, so the screen is incomplete and neither a bounded absence (1) nor the items found can be scored. | data/scores.json or data/matrix-extra.json: Staples Media Network, governance |
| C516 | UB Media / governance = 1 (grade D) | Separate refute-mode cell verifier for scoring slice A3. | Three marketing pages, no terms: incomplete screen. | contradicted | UB Media / governance -> ne | Session ruling 1 (frozen dimension-11 definition: 'Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'): E2's screen read UB Media's home, outcomes and solutions pages and found no restatement, dispute, versioning or separation term [RTL-U-S066, RTL-E2-S30, RTL-E2-S53; RTL-E2-F27]. No advertiser terms, IO terms or advertising agreement for this service were read, so the screen is incomplete and neither a bounded absence | data/scores.json or data/matrix-extra.json: UB Media, governance |
| C517 | Gulp Media Network / default_method = n/e | Cell verifier for scoring slice A3 scanned the n/e cells. | 1 (grade D, live, 2026-11): under 'Tracking performance. Measuring success. Detailed reporting ...' the network's own page names an attributed metric in standard reporting, 'Attributable Sales: Exposed customer purchase behavior will measure the ROAS for the campaign based on measured sales in C-Shopper.', with no counterfactual; 'Incrementality: Control/exposed methodology will enable lift reporting ... and calculate an iROAS' names no construction or assignment and, under session rulings 8 and | structural | Gulp Media Network / default_method: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Gulp Media Network, default_method |
| C518 | Staples Media Network / routes = n/e | Cell verifier for scoring slice A3 scanned the n/e cells. | 1 (managed service only; grade D, live, 2026-11): the network page, re-opened 3 Nov 2026, says 'Staples Support: Never miss a beat with dedicated Account Managers and a team of media planners, AdOps support people, and data analysts.'; no self-serve, API or DSP route is stated. | structural | Staples Media Network / routes: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Staples Media Network, routes |
| C519 | UB Media / routes = n/e | Cell verifier for scoring slice A3 scanned the n/e cells. | 1 (managed service only; grade D, live, 2026-11): UB Media's home page, re-opened 3 Nov 2026: 'Partner with our experts: Our dedicated UB Media team provides custom solutions and actionable insights for brand partners. We also offer an extended team of industry-leading media buying and services partners to provide top digital media strategy, planning and buying resources across channels.' Criteo is named only as the serving platform for sponsored products. | structural | UB Media / routes: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | UB Media, routes |
| C520 | AD Retail Media / routes = n/e | Cell verifier for scoring slice A3 scanned the n/e cells. | 3 (one family, the Edge platform console for CPG partners; grade D, live, 2026-01-07): the release, re-read in the browser pane on 3 Nov 2026, is headed 'Now Available to CPG Partners, Edge Unifies Onsite Display, Sponsored Search and In-Store Digital Media' and says Edge consolidates 'audience insights, media planning, activation and optimization' and 'helps partners plan, launch and optimize campaigns more efficiently'. Edge is a child record of the row (its console), which makes it a route, n | structural | AD Retail Media / routes: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | AD Retail Media, routes |
| C521 | Wayfair advertising / routes = n/e | Cell verifier for scoring slice A3 scanned the n/e cells. | 3 (two families; grade D, live, 2026-11): the supplier page, re-opened 3 Nov 2026, offers 'Our digital advertising tools' / 'Get started with our online advertising tools' for the pay-per-click Sponsored Products and Sponsored Shops, and calls display 'This managed ad service intelligently places display ads for your products around Wayfair'. If the session does not accept 'online advertising tools' as a documented self-serve console, the managed display service alone gives 1. | structural | Wayfair advertising / routes: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Wayfair advertising, routes |
| C522 | Orange Apron Media / privacy = 1 (grade D) | Separate refute-mode cell verifier for scoring slice A4. | The 1 ('generic company policy only') is not supported as written: it rests on footer link labels whose policy text was never opened, and the record also holds a product-specific privacy statement on the measurement page ('privacy-enhanced environment ... all while adhering to privacy laws'), which is rung 2 (as Vantage and STRATACACHE in slice C, Chase and Kinective in slice B). | contradicted | Orange Apron Media / privacy -> 2 (grade D) | Anchor 2: a product-specific notice without a named legal framework: the measurement page's Clean Rooms section says 'Sharing data with publishers and other partners in a neutral and privacy-enhanced environment gives you an edge over your competitors. With an expanded view of your customers, you can uncover deeper insights and fine-tune your campaigns, all while adhering to privacy laws.' (no law named) [RTL-E2-S10, RTL-U-S049; read by the cell verifier 3 November 2026]. Blocks 3: no consent ba | data/scores.json or data/matrix-extra.json: Orange Apron Media, privacy |
| C523 | Chewy Ads / default_method = n/e | Cell verifier for scoring slice A4 scanned the n/e cells. | 1 (grade D, date 2026-11): Anchor 1: the row's own pages name attributed outcome metrics in its standard reporting and document no campaign counterfactual for the default: 'Chewy Ads provides insights like New-to-Brand, Autoship, Lifetime Value ROAS, and incrementality' (home-page FAQ), the proprietary 'LTV ROAS' 'connecting campaign performance to future Autoship purchases' (solutions page), '$7+ LTV Sales' 'For every $1 spent on Chewy Ads' footnoted 'Average LTV ROAS performance is based on al | structural | Chewy Ads / default_method: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Chewy Ads, default_method |
| C524 | Leap Media Group (Giant Eagle) / default_method = n/e | Cell verifier for scoring slice A4 scanned the n/e cells. | 1 (grade D, date 2026-11): Anchor 1: under 'Audience reach & impact measurement' the solutions page says advertisers 'can monitor the audiences, impressions and overall effectiveness and efficiency of your brand investment' with 'Measurable Analytics, including ROAS', and the home page offers 'a measurable (think ROAS and impressions)' opportunity; ROAS is an attributed outcome metric and no counterfactual or lift offer is documented (ruling 8 names Leap). Blocks 2: no counterfactual asserted. A | structural | Leap Media Group (Giant Eagle) / default_method: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Leap Media Group (Giant Eagle), default_method |
| C525 | Petco RMN / default_method = n/e | Cell verifier for scoring slice A4 scanned the n/e cells. | 1 (grade D, date 2026-11): Anchor 1: 'Transparent Reporting: Access to a custom reporting dashboard with real-time campaign analytics, including ROAS, Sales, CTR, ROI and more.', and for both the Self-Service and Managed Service routes 'Real time access to ROAS reporting'; no counterfactual or lift offer is documented, and in-store sales are a roadmap item ('enhanced measurement capabilities including ad influenced store sales'). Blocks 2: no counterfactual. Avail live. (Page footered 2024; date | structural | Petco RMN / default_method: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Petco RMN, default_method |
| C526 | Orange Apron Media / default_method = n/e | Cell verifier for scoring slice A4 scanned the n/e cells. | 1 (grade D, date 2026-11): Anchor 1: 'Our closed-loop measurement links your advertising and media spend directly to both online and in-store sales and conversions' (measurement page), reported in Orange Access dashboards 'featuring media diagnostics and key business metrics like single platform attribution and in-store sales for onsite' (release, 3 October 2024); the measurement post places 'advanced studies such as incrementality, geo lift, and brand lift' 'Beyond standard reporting', 'built a | structural | Orange Apron Media / default_method: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Orange Apron Media, default_method |
| C527 | Kohl's Media Network / us_availability = n/e | Cell verifier for scoring slice A4 scanned the n/e cells. | 3 (grade D, date 2026-11): Anchor 3: US availability documented: Kohl's release on the network, 'Kohl's Expands Advertising Opportunities with Kohl's Media Network' (21 July 2022), closes 'With more than 1,100 stores in 49 states and the online convenience of Kohls.com and the Kohl's App, Kohl's offers amazing national and exclusive brands at incredible savings for families nationwide', and the live offer page sells placements on Kohls.com and the app ('© 2026 Kohl's, Inc.'); this is the operato | structural | Kohl's Media Network / us_availability: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Kohl's Media Network, us_availability |
| C528 | Kohl's Media Network / attribution_disclosure = n/e | Cell verifier for scoring slice A4 scanned the n/e cells. | 1 (grade D, date 2022-07-21): Anchor 1: 'closed-loop reporting' asserted with no parameter: 'KMN has continued to expand with new enhancements including cross-targeting, industry-leading omnichannel data integration and closed-loop reporting' (release, 21 July 2022); the 2026 offer page names only 'real-time unified reporting' on the Koddi platform, and neither page states a window, touch model or match basis. Blocks 2: neither window nor touch model stated. Caveat for the session: the assertion | structural | Kohl's Media Network / attribution_disclosure: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Kohl's Media Network, attribution_disclosure |
| C529 | Leap Media Group (Giant Eagle) / routes = n/e | Cell verifier for scoring slice A4 scanned the n/e cells. | 1 (grade D, date 2026-11): Anchor 1: managed service only: 'We specialize in media buying, placing your message in the most impactful places. Let us put our data and intelligence to work for you.' and 'Our dedicated team and tools are here to ensure your message is reaching the right people, at the right time, at the right location', sold as 'Seasonally Relevant & New Item omnichannel package offerings'; no self-serve console, API, DSP, agency or clean-room route is documented. Blocks 3: no seco | structural | Leap Media Group (Giant Eagle) / routes: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Leap Media Group (Giant Eagle), routes |
| C530 | Amazon display ads / placement_coverage = 2 (grade C) | Separate refute-mode cell verifier for scoring slice A5. | The 2 rests on treating on-site as the primary class, which no source states: the display ads page body lists 'customers across Amazon properties including Twitch, Fire TV, and Echo Show' and 'premium placements across the open internet', names no primary placement, and mentions the Amazon store only in an image caption. Measurement is documented row-wide (14-day standard, click attribution) plus one class-specific statement for vCPM ads in the Store; the other classes carry none. That is the sa | contradicted | Amazon display ads / placement_coverage -> 1 (grade C) | Anchor 1: classes served with measurement pooled or undocumented. Display ads serve 'customers across Amazon properties including Twitch, Fire TV, and Echo Show' and 'premium placements across the open internet'; the page names no primary placement and states no measurement (the Amazon store appears only in an image caption) [RTL-U-S003]. Outcome rules are documented for the row as a whole (a 14-day standard look-back for campaign types other than Sponsored Products; click attribution) [RTL-AMAZ | data/scores.json or data/matrix-extra.json: Amazon display ads, placement_coverage |
| C531 | Albertsons Media Collective offsite / identity_basis = 3 (grade D) | Separate refute-mode cell verifier for scoring slice A6. | Rung 3 needs the identity basis documented; the record documents the transaction side ('Albertsons transaction data', 'item level sales'), the environment (LiveRamp) and the channel list, but not the identity on which exposures are matched (the X verifier's correction: join keys 'not stated'). The scorer's own onsite cell treats an unstated key as no documented basis, and the paper scores 'keys unstated' rows at 1 (Instacart display and video, DG Media Network's '1:1' closed loop, Chewy, Walmart | contradicted | Albertsons Media Collective offsite / identity_basis -> 1 (grade D) | Anchor 1: the basis is asserted only. The off-site record says what outcomes are joined to but not on what identity: the multi-touch product (17 September 2026; a launch with one pilot partner, PepsiCo) connects 'impression-level exposure data across Albertsons Media Collective's owned and operated channels, offsite media and participating partners including Meta, third-party inventory with Display & Video 360, Google Ad Manager and Perion, with Albertsons transaction data in a neutral, privacy- | data/scores.json or data/matrix-extra.json: Albertsons Media Collective offsite, identity_basis |
| C532 | uber-journey-ads / governance = 1 (grade D) | Separate refute-mode cell verifier for scoring slice B. | Scored 1 as a bounded absence on a screen that read no advertiser or insertion-order terms; under the frozen definition and session ruling 1 the screen is incomplete, so n/e (ruling Q1). | contradicted | uber-journey-ads / governance -> ne | Screen incomplete under the frozen definition of dimension 11 ('Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'; session ruling 1): E3 read the US landing and measurement pages, the Custom AU post and the press page to 100,000 of 107,924 characters; the Journey Ads page returned HTTP 404 and Uber Advertising's terms were not read [RTL-U-S043, RTL-U-S044, RTL-E3-S24, RTL-E3-S25, RTL-E3-S45]. E3 not flagging the screen | data/scores.json or data/matrix-extra.json: uber-journey-ads, governance |
| C533 | lyft-ads / governance = 1 (grade D) | Separate refute-mode cell verifier for scoring slice B. | Scored 1 as a bounded absence on a screen that read no advertiser or insertion-order terms; under the frozen definition and session ruling 1 the screen is incomplete, so n/e (ruling Q1). | contradicted | lyft-ads / governance -> ne | Screen incomplete under the frozen definition of dimension 11 ('Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'; session ruling 1): E3 read ads.lyft.com, its measurement page and the 2024 launch post; the Lyft Ads terms and privacy policy linked in the footer were not read [RTL-U-S102, RTL-U-S103, RTL-U-S104]. E3 not flagging the screen incomplete does not complete it, and a bounded absence (1) needs the completed sc | data/scores.json or data/matrix-extra.json: lyft-ads, governance |
| C534 | chase-media-solutions / governance = 1 (grade D) | Separate refute-mode cell verifier for scoring slice B. | Scored 1 as a bounded absence on a screen that read no advertiser or insertion-order terms; under the frozen definition and session ruling 1 the screen is incomplete, so n/e (ruling Q1). | contradicted | chase-media-solutions / governance -> ne | Screen incomplete under the frozen definition of dimension 11 ('Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'; session ruling 1): E3 read five Chase Media Solutions pages (its own limitation records the case-study pages as not opened) and no Chase Media Solutions advertiser or insertion-order terms [RTL-U-S105, RTL-U-S106, RTL-U-S107, RTL-E3-S08, RTL-E3-S09; RTL-E3-F06]. E3 not flagging the screen incomplete does n | data/scores.json or data/matrix-extra.json: chase-media-solutions, governance |
| C535 | fetch-special-offers / default_method = 3 (grade D) | Separate refute-mode cell verifier for scoring slice B. | Ruling Q3: the default the buyer receives at the cutoff decides; the randomised holdout is documented only for 2023 and no lower rung is documented for 2026, so n/e (not 3, and not a lower rung). | contradicted | fetch-special-offers / default_method -> ne | Default at the cutoff undocumented. No page dated at or near the cutoff states what a Special Offers buyer receives by default: the 2026 capabilities page names 'world-class incrementality and long-term value measurement' and 'Measure your impact - from sales lift to long-term value' with no design, metric or default and links the 8 June 2023 KPMG takeaways post [RTL-E3-S16, re-opened 3 Nov 2026]; the randomised user holdout ('Fetch randomly assigns users into either "treatment" or "control" gro | data/scores.json or data/matrix-extra.json: fetch-special-offers, default_method |
| C536 | fetch-special-offers / governance = 1 (grade C) | Separate refute-mode cell verifier for scoring slice B. | Scored 1 as a bounded absence on a screen that read no advertiser or insertion-order terms; under the frozen definition and session ruling 1 the screen is incomplete, so n/e (ruling Q1). | contradicted | fetch-special-offers / governance -> ne | Screen incomplete under the frozen definition of dimension 11 ('Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'; session ruling 1): E3 read the business home and capabilities pages, three 2023 posts, the 2025 MMM post and the consumer privacy policy; no brand-partner or insertion-order terms were read and the Snowflake post rendered only a listing [RTL-U-S117, RTL-U-S118, RTL-E3-S15 to RTL-E3-S20; RTL-E3-F24]. E3 not | data/scores.json or data/matrix-extra.json: fetch-special-offers, governance |
| C537 | fetch-special-offers / privacy = 3 (grade C) | Separate refute-mode cell verifier for scoring slice B. | Value raised 3 -> 4: the scorer's blocker ('the passages read name no legal framework for these controls') is contradicted by the policy's CCPA financial-incentive notice tied to sharing with brands and its Washington My Health My Data policy; grade C caps at 4. Date = the registry page (last updated 30 Oct 2026), the newest cited source. | supported with qualification | fetch-special-offers / privacy -> 4 (grade C) | Anchor 4: both checklist items are documented for the product with named legal frameworks. Consent basis: 'By utilizing the Fetch App, requesting Fetch Services or by providing your information to Fetch, you are consenting to Fetch disclosing your information with our brand partners', framed 'Under the CCPA' as a financial-incentive programme in which 'you direct us to use and share the personal information you provide ... with all our participating brands'; e-receipts collected or disclosed onl | data/scores.json or data/matrix-extra.json: fetch-special-offers, privacy |
| C538 | ibotta-performance-network-offers / governance = 1 (grade A) | Separate refute-mode cell verifier for scoring slice B. | Scored 1 as a bounded absence on a screen that read no advertiser or insertion-order terms; under the frozen definition and session ruling 1 the screen is incomplete, so n/e (ruling Q1). | contradicted | ibotta-performance-network-offers / governance -> ne | Screen incomplete under the frozen definition of dimension 11 ('Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'; session ruling 1): E3 read the IPN home, properties, proof, Danone and Circana pages and 10-K slices; no IPN client terms or insertion-order terms were read [RTL-U-S119, RTL-U-S120, RTL-U-S121, RTL-E3-S21, RTL-E3-S22, RTL-E3-S23]. E3 not flagging the screen incomplete does not complete it, and a bounded ab | data/scores.json or data/matrix-extra.json: ibotta-performance-network-offers, governance |
| C539 | ibotta-d2c-ad-products / governance = 1 (grade A) | Separate refute-mode cell verifier for scoring slice B. | Scored 1 as a bounded absence on a screen that read no advertiser or insertion-order terms; under the frozen definition and session ruling 1 the screen is incomplete, so n/e (ruling Q1). | contradicted | ibotta-d2c-ad-products / governance -> ne | Screen incomplete under the frozen definition of dimension 11 ('Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'; session ruling 1): E3 read the IPN home and properties pages and 10-K slices; no terms for the D2C ad products were read [RTL-U-S119, RTL-U-S120, RTL-U-S121]. E3 not flagging the screen incomplete does not complete it, and a bounded absence (1) needs the completed screen [RTL-E3-F53 limitation: 'complete o | data/scores.json or data/matrix-extra.json: ibotta-d2c-ad-products, governance |
| C540 | marriott-media / attribution_disclosure = 1 (grade D) | Separate refute-mode cell verifier for scoring slice B. | The scorer's anchor 1 rests on a future-tense promise; availability of that capability is 'announced', which is never scored. | contradicted | marriott-media / attribution_disclosure -> ne | No live attribution claim to place at anchor 1: the only 'closed loop attribution' statement is a future-tense promise in the 12 June 2025 launch release ('will offer marketers campaign performance with closed loop attribution'), beside 'brand lift and awareness studies' and modelling named in a spokesperson quote, so its construct is undetermined [RTL-U-S122; RTL-E3-F32 qualified]; the 2026 home page read in the browser pane carries partner-case metrics and no attribution statement [RTL-E3-S41] | data/scores.json or data/matrix-extra.json: marriott-media, attribution_disclosure |
| C541 | kinective-media-by-united-airlines / governance = 1 (grade D) | Separate refute-mode cell verifier for scoring slice B. | Scored 1 as a bounded absence on a screen that read no advertiser or insertion-order terms; under the frozen definition and session ruling 1 the screen is incomplete, so n/e (ruling Q1). | contradicted | kinective-media-by-united-airlines / governance -> ne | Screen incomplete under the frozen definition of dimension 11 ('Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'; session ruling 1): E3 read the June 2024 launch release and the home page (browser pane); no advertiser terms were read [RTL-U-S124, RTL-E3-S42]. E3 not flagging the screen incomplete does not complete it, and a bounded absence (1) needs the completed screen [RTL-E3-F53 limitation: 'complete only where the | data/scores.json or data/matrix-extra.json: kinective-media-by-united-airlines, governance |
| C542 | expedia-group-advertising-for-non-endemic-brands / governance = 1 (grade D) | Separate refute-mode cell verifier for scoring slice B. | Scored 1 as a bounded absence on a screen that read no advertiser or insertion-order terms; under the frozen definition and session ruling 1 the screen is incomplete, so n/e (ruling Q1). | contradicted | expedia-group-advertising-for-non-endemic-brands / governance -> ne | Screen incomplete under the frozen definition of dimension 11 ('Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'; session ruling 1): E3 read the non-endemic page, the off-site page and two releases; Expedia Group Advertising's advertiser terms were not read [RTL-U-S126, RTL-U-S127, RTL-U-S128, RTL-E3-S27]. E3 not flagging the screen incomplete does not complete it, and a bounded absence (1) needs the completed screen | data/scores.json or data/matrix-extra.json: expedia-group-advertising-for-non-endemic-brands, governance |
| C543 | klarna-ads-manager / buyer_analysis = 1 (grade D) | Separate refute-mode cell verifier for scoring slice B. | Ruling Q5: the row is not shown live at the cutoff, so none of its cells can carry a scorable availability label. | contradicted | klarna-ads-manager / buyer_analysis -> ne | Not scored: availability at the cutoff unconfirmed. The only first-party record is the 13 November 2023 release ('now available to Klarna's retail partners in the UK and Sweden', global rollout 'over the next few months') [RTL-U-S135]; no 2026 Klarna advertiser page was found (E3's two searches; the cell verifier's search of 3 November 2026 found only 2023 coverage) and both ad-policy URLs end at HTTP 404 (re-checked 3 November 2026) [RTL-E3-S46]; the register codes Ads Manager's US availability | data/scores.json or data/matrix-extra.json: klarna-ads-manager, buyer_analysis |
| C544 | klarna-ads-manager / placement_coverage = 1 (grade D) | Separate refute-mode cell verifier for scoring slice B. | Ruling Q5: the row is not shown live at the cutoff, so none of its cells can carry a scorable availability label. | contradicted | klarna-ads-manager / placement_coverage -> ne | Not scored: availability at the cutoff unconfirmed. The only first-party record is the 13 November 2023 release ('now available to Klarna's retail partners in the UK and Sweden', global rollout 'over the next few months') [RTL-U-S135]; no 2026 Klarna advertiser page was found (E3's two searches; the cell verifier's search of 3 November 2026 found only 2023 coverage) and both ad-policy URLs end at HTTP 404 (re-checked 3 November 2026) [RTL-E3-S46]; the register codes Ads Manager's US availability | data/scores.json or data/matrix-extra.json: klarna-ads-manager, placement_coverage |
| C545 | klarna-ads-manager / routes = 2 (grade D) | Separate refute-mode cell verifier for scoring slice B. | Ruling Q5: the row is not shown live at the cutoff, so none of its cells can carry a scorable availability label. | contradicted | klarna-ads-manager / routes -> ne | Not scored: availability at the cutoff unconfirmed. The only first-party record is the 13 November 2023 release ('now available to Klarna's retail partners in the UK and Sweden', global rollout 'over the next few months') [RTL-U-S135]; no 2026 Klarna advertiser page was found (E3's two searches; the cell verifier's search of 3 November 2026 found only 2023 coverage) and both ad-policy URLs end at HTTP 404 (re-checked 3 November 2026) [RTL-E3-S46]; the register codes Ads Manager's US availability | data/scores.json or data/matrix-extra.json: klarna-ads-manager, routes |
| C546 | klarna-ads-manager / us_availability = 1 (grade D) | Separate refute-mode cell verifier for scoring slice B. | Ruling Q5: the row is not shown live at the cutoff, so none of its cells can carry a scorable availability label. | contradicted | klarna-ads-manager / us_availability -> ne | Not scored: availability at the cutoff unconfirmed. The only first-party record is the 13 November 2023 release ('now available to Klarna's retail partners in the UK and Sweden', global rollout 'over the next few months') [RTL-U-S135]; no 2026 Klarna advertiser page was found (E3's two searches; the cell verifier's search of 3 November 2026 found only 2023 coverage) and both ad-policy URLs end at HTTP 404 (re-checked 3 November 2026) [RTL-E3-S46]; the register codes Ads Manager's US availability | data/scores.json or data/matrix-extra.json: klarna-ads-manager, us_availability |
| C547 | zillow-elevate-media-manager / governance = 1 (grade C) | Separate refute-mode cell verifier for scoring slice B. | Scored 1 as a bounded absence on a screen that read no advertiser or insertion-order terms; under the frozen definition and session ruling 1 the screen is incomplete, so n/e (ruling Q1). | contradicted | zillow-elevate-media-manager / governance -> ne | Screen incomplete under the frozen definition of dimension 11 ('Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'; session ruling 1): E3 read only the self-service FAQ (browser pane); zillow.com/brands returned HTTP 403 and the Media Manager terms were not read [RTL-E3-S43; RTL-E3-F41]. E3 not flagging the screen incomplete does not complete it, and a bounded absence (1) needs the completed screen [RTL-E3-F53 limitatio | data/scores.json or data/matrix-extra.json: zillow-elevate-media-manager, governance |
| C548 | rokt-ads / governance = 2 (grade C) | Separate refute-mode cell verifier for scoring slice B. | The positive item is real, but both the scorer's own conflict rule (no stream may record an unopened part of the row's documentation or terms) and session ruling 1 fail: E3 records two unopened Rokt sub-pages and no terms were read. | contradicted | rokt-ads / governance -> ne | Screen incomplete under the frozen definition of dimension 11 ('Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e'; session ruling 1): E3 read seven docs.rokt.com pages and the product page, but the 'Access referral data' and 'Measuring success and analyzing results' sub-pages were not opened (E3 open questions; RTL-E3-F44 limitation) and Rokt's advertiser terms and policies ('Rokt Policies' in the docs navigation) were | data/scores.json or data/matrix-extra.json: rokt-ads, governance |
| C549 | paypal-ads / default_method = n/e | Cell verifier for scoring slice B scanned the n/e cells. | 1 under session ruling 8: Curated Ads' standard measurement names an attributed outcome ('Attribution is then closed back to verified PayPal transactions'; 'Measurement is traced not to a click or a model, but to a confirmed PayPal purchase') with no campaign counterfactual; the suite's 'incrementality' has no design and neither raises nor blocks rung 1; the Ulta +20% is a pre/post case figure, not the default. No product is documented at a different rung, so ruling 3 does not bind. | structural | paypal-ads / default_method: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | paypal-ads, default_method |
| C550 | citi-commerce-media / default_method = n/e | Cell verifier for scoring slice B scanned the n/e cells. | 1 under session ruling 8: the platform's measurement names ROAS ('Uses closed-loop measurement to help advertisers evaluate campaign performance, prove return on ad spending and quantify incremental impact') with no documented counterfactual; 'quantify incremental impact' and 'up to 5x incremental return on ad spend' are undesigned labels that neither raise nor block rung 1 (compare Leap 'Measurable Analytics, including ROAS' = 1 in the ruling). | structural | citi-commerce-media / default_method: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | citi-commerce-media, default_method |
| C551 | amex-offers / routes = n/e | Cell verifier for scoring slice B scanned the n/e cells. | 1 (managed service only): 'We'll work with eligible merchants like you to create offers'; 'Step 1 We'll help you design an offer tailored to your business needs and set up custom targeting. Step 2 We'll promote your offer to Card Members'; 'Leave the busy work to us'; 'Amex Offers may not be available to all Merchants and Advertisers. Reach out to your Client Manager for more details.' | structural | amex-offers / routes: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | amex-offers, routes |
| C552 | klarna-search-compare-sponsored-placements / privacy = n/e | Cell verifier for scoring slice B scanned the n/e cells. | 2 (product-specific notice without a named framework for shopper data): Display Ads run 'as targeted ads for Users that have accepted the use of marketing cookies'; the terms' only data-protection clause (s.22) names the GDPR for the partner's contact persons, not shopper data; no consent basis for purchase or payment data and no sensitive-category rule. | structural | klarna-search-compare-sponsored-placements / privacy: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | klarna-search-compare-sponsored-placements, privacy |
| C553 | capital-one-ad-solutions / placement_coverage = n/e | Cell verifier for scoring slice B scanned the n/e cells. | 1: classes served are listed ('Search: Reach shoppers at the point of discovery with sponsored product placements across search results'; 'Feed'; 'Push'; 'Email'; 'Cardholder Offers: Reach 140M+ Capital One cardholders with exclusive offers delivered through the banking app') with no measurement documented per class. | structural | capital-one-ad-solutions / placement_coverage: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | capital-one-ad-solutions, placement_coverage |
| C554 | capital-one-ad-solutions / us_availability = n/e | Cell verifier for scoring slice B scanned the n/e cells. | 3: 'We reach 140M+ customers and ~40% of US households post-Discover acquisition' with 'Get Started' and 'Contact us' calls to action; named US campaigns not documented on the home page (case studies unread). | structural | capital-one-ad-solutions / us_availability: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | capital-one-ad-solutions, us_availability |
| C555 | criteo / attribution_disclosure = 2 (grade C) | Separate refute-mode cell verifier for scoring slice C. | Question 1: the Commerce Yield verifier's reading holds on the retailer-side pages themselves (no Commerce Max transfer): six of eight with values meets anchor 4; partial new-to-brand and missing sales basis are not counted (same standard as Kevel). Value 2 -> 4. | contradicted | criteo / attribution_disclosure -> 4 (grade C) | Anchor 4: six of eight parameters documented with values on Commerce Yield's own help centre. Window: post-click 7, 14 or 30 days (default 30), post-view none, 1, 7, 14 or 30 days (default 1), set per campaign [RTL-CRITEO-S22]. Touch model: one most relevant event per sale, clicks over impressions, same SKU over same category, then the event closest to purchase; other events recorded as assisted sales [RTL-VCRITEO-S01]. Match basis: deterministic user matching on the retailer's CRM ID, first-par | data/scores.json or data/matrix-extra.json: criteo, attribution_disclosure |
| C556 | criteo / buyer_analysis = 3 (grade D) | Separate refute-mode cell verifier for scoring slice C. | Value 3 stands; the route now rests on Commerce Yield's own help centre (grade C) as the Criteo verifier asked, not on the product page alone (D); the attributed transaction log is recorded as row-level portability beside the cell. | supported with qualification | criteo / buyer_analysis -> 3 (grade C) | Anchor 3: aggregates on the operator's queries. The Commerce Yield page documents in-platform reports, automated email reports and 'API Reporting Endpoints' [RTL-U-S147], and Commerce Yield's own help centre documents an Attributed Transaction log ('visibility on all transactions attributed to a campaign') and brands' access to offline sales in their Commerce Max demand accounts only 'If shared by the Retailer' [RTL-VCRITEO-S01, RTL-VCRITEO-S02]. No buyer-run query environment, clean room, outpu | data/scores.json or data/matrix-extra.json: criteo, buyer_analysis |
| C557 | criteo / routes = 5 (grade A) | Separate refute-mode cell verifier for scoring slice C. | Question 2: the 10-K passages exist (the verifier's needs_browser item is resolved from the full filing copy), but they describe the Retail Media segment and name no product; the Criteo verifier qualified the API program and SA360 as company-level demand routes. The cell rests on Commerce Yield's own documentation: 4, grade C. Value 5/A -> 4/C. | contradicted | criteo / routes -> 4 (grade C) | Anchor 5 in substance, capped at 4 by vendor documentation: Commerce Yield's own documentation shows three route families including an API: a self-serve console (the Commerce Yield help centre documents the platform's campaign, supply-side and private-market functions) [RTL-CRITEO-S20]; the Criteo Retail Media API with retailer-side guides (SSP revenue reports, Retailer Budgets, 'Retailers - Reach out to your Criteo representative'), shared authentication and versioning [RTL-F-S01, RTL-F-S03]; a | data/scores.json or data/matrix-extra.json: criteo, routes |
| C558 | criteo / scale = 3 (grade D) | Separate refute-mode cell verifier for scoring slice C. | The scored 3 counted two selection releases with no go-live date; the session coded Pentaleap's Macy's selection as not production, so Dollar General drops. Albertsons counts on its own dated 2026 release of ongoing Criteo-powered sponsored products. Two networks: 3 -> 2. | contradicted | criteo / scale -> 2 (grade D) | Anchor 2 (two networks): Belk ('Belk Media Network launches'; 'Belk is harnessing ... Commerce Yield', 8 October 2024) [RTL-CRITEO-S12] and Albertsons Media Collective, which names Commerce Yield for its onsite sponsored ads on 9 January 2024 [RTL-CRITEO-S10] and on 23 June 2026 describes sponsored products served 'Through Criteo' in its app's AI-powered conversational search as part of an 'ongoing partnership with Criteo' [RTL-ALBERTSONS-S16]; mapping that 2026 production to Commerce Yield is a | data/scores.json or data/matrix-extra.json: criteo, scale |
| C559 | criteo / governance = 2 (grade C) | Separate refute-mode cell verifier for scoring slice C. | Question 8: not a completed screen; n/e under the frozen definition and session ruling 1. | contradicted | criteo / governance -> ne | Screen incomplete (session ruling 1): the record screened the Commerce Yield help-centre titles, three article bodies, the product page and Criteo's API versioning policy, but left unread the listed governance articles 'What's New on Commerce Yield 2026', 'About the Change History Tool' and 'Why Is My Invoice Not Aligned With My Reporting?', and read no Commerce Yield terms or retailer agreement; the profile itself records 'Screen incomplete for most article bodies'. | data/scores.json or data/matrix-extra.json: criteo, governance |
| C560 | Carrot Ads / governance = 2 (grade C) | Separate refute-mode cell verifier for scoring slice C. | Question 8: bounded but not complete (terms unread); n/e under session ruling 1. | contradicted | Carrot Ads / governance -> ne | Screen incomplete (session ruling 1): the Carrot Ads API overview, ingest-orders, metrics, events and changelog pages were read (dated changelog 18 March 2024 to 10 October 2026), but no Carrot Ads terms or retailer agreement and none of the reporting-API pages the overview lists were screened. | data/scores.json or data/matrix-extra.json: Carrot Ads, governance |
| C561 | Kevel Retail Media Cloud / default_method = 2 (grade A) | Separate refute-mode cell verifier for scoring slice C. | Question 4a: an MRC under-review scope records a pre-audit status, not what Kevel's default report delivers or a design it supports; the cohort C anchor reads the platform's own documentation. Value 2/A -> 1/C. | contradicted | Kevel Retail Media Cloud / default_method -> 1 (grade C) | Anchor 1: attribution only. Kevel's own documentation names attributed outcome metrics in its standard reporting (conversions under last-touch attribution, clicks checked before views; GMV and ROAS in the Reporting API, scheduled reports and the Kevel UI) [RTL-F-S15, RTL-F-S14, RTL-F-S38] and documents no campaign counterfactual: the Retail Media Guide and attribution pages name no holdout, A/B or incrementality design [RTL-F-S16, RTL-F-S14]. MRC's services-under-review list carries 'Kevel (Pre- | data/scores.json or data/matrix-extra.json: Kevel Retail Media Cloud, default_method |
| C562 | Kevel Retail Media Cloud / attribution_disclosure = 4 (grade C) | Separate refute-mode cell verifier for scoring slice C. | Question 4b: the brief defines the sales basis as gross or net plus the treatment of returns, cancellations, discounts, taxes and shipping; 'quantity * price' gives none of the treatments, matching the standard the Criteo verifier applied. Five of eight: 4 -> 3. | contradicted | Kevel Retail Media Cloud / attribution_disclosure -> 3 (grade C) | Anchor 3: window and touch model stated, with five of eight parameters documented with values: window (1, 7, 14 or 30 days, post-click and post-view set independently per flight and ad), touch model (last touch, clicks checked before views), match basis (a consistent UserKey on decision and purchase events), halo (Same Product, Same Category and Brand, Same Brand, Same Merchant) and deduplication ('each purchase conversion is attributed to a single ad') [RTL-F-S14, RTL-F-S15]. The sales basis is | data/scores.json or data/matrix-extra.json: Kevel Retail Media Cloud, attribution_disclosure |
| C563 | Kevel Retail Media Cloud / governance = 2 (grade C) | Separate refute-mode cell verifier for scoring slice C. | Question 8: F18's statement that no changelog or release notes exist is contradicted by Kevel's own release notes, and the terms were not read: n/e. | contradicted | Kevel Retail Media Cloud / governance -> ne | Screen incomplete (session ruling 1): the record screened the documentation index and the attribution, retail media and reporting-discrepancies pages, but Kevel publishes monthly release notes (dev.kevel.com/changelog lists 'Release notes' from May 2025 to 6 March 2026) that were not screened, the GDPR and data-model pages were read only as index summaries, and the 'Legal' hub ('Kevel's agreements, policies, and notices') and the Data Processing Agreement were not read. | data/scores.json or data/matrix-extra.json: Kevel Retail Media Cloud, governance |
| C564 | Streaem / default_method = 1 (grade C) | Separate refute-mode cell verifier for scoring slice C. | Question 7: label unconfirmed; the cells cannot stand as live. | contradicted | Streaem / default_method -> ne | Not scoreable: availability unconfirmed. The product site now returns 404 (streaem.com redirects to vml.com/streaem, 'Error Not Found', 3 November 2026) and workstream U lists the row as unconfirmed; developer pages last updated 22 May 2025 show that documentation exists, not that the service is sold at the cutoff. Scored cells must be live or beta. Recorded beside the cell: post-click and post-view attribution within retailer-set windows, no control design [RTL-F-S34, RTL-F-S35]. | data/scores.json or data/matrix-extra.json: Streaem, default_method |
| C565 | Streaem / attribution_disclosure = 2 (grade C) | Separate refute-mode cell verifier for scoring slice C. | Question 7: label unconfirmed; the cells cannot stand as live. | contradicted | Streaem / attribution_disclosure -> ne | Not scoreable: availability unconfirmed. The product site now returns 404 (streaem.com redirects to vml.com/streaem, 'Error Not Found', 3 November 2026) and workstream U lists the row as unconfirmed; developer pages last updated 22 May 2025 show that documentation exists, not that the service is sold at the cutoff. Scored cells must be live or beta. Recorded beside the cell: retailer-set look-back windows with a one-day post-view window advised [RTL-F-S34, RTL-F-S35]. | data/scores.json or data/matrix-extra.json: Streaem, attribution_disclosure |
| C566 | Streaem / routes = 3 (grade C) | Separate refute-mode cell verifier for scoring slice C. | Question 7: label unconfirmed; the cells cannot stand as live. | contradicted | Streaem / routes -> ne | Not scoreable: availability unconfirmed. The product site now returns 404 (streaem.com redirects to vml.com/streaem, 'Error Not Found', 3 November 2026) and workstream U lists the row as unconfirmed; developer pages last updated 22 May 2025 show that documentation exists, not that the service is sold at the cutoff. Scored cells must be live or beta. Recorded beside the cell: a developer API for sponsored products and add-to-cart and purchase events [RTL-F-S34, RTL-F-S35]. | data/scores.json or data/matrix-extra.json: Streaem, routes |
| C567 | NIQ Activate Retail Media Intelligence / default_method = n/e | Cell verifier for scoring slice C scanned the n/e cells. | 1 (grade D, live, date 2026-11) under session ruling 8: the row's own page names the attributed outcomes its measurement delivers and no counterfactual: 'NIQ Activate closes the loop between exposure and purchase by providing AI-powered retail media measurement and attribution. This allows retail media networks and brands to clearly understand campaign performance, ROI, and contribution to sales across the retail environment.' [RTL-U-S170, re-opened 3 Nov 2026]. Borderline: ROI and contribution | structural | NIQ Activate Retail Media Intelligence / default_method: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | NIQ Activate Retail Media Intelligence, default_method |
| C568 | circana / placement_coverage = 2 (grade D) | Separate refute-mode cell verifier for scoring slice DE. | DoorDash cannot be counted: no source names the Circana service behind the DoorDash validation, the dated period belongs to the buyer-overlap analysis, and after amendment v1.1 the study could as well be NCS Sales Effect's; with Ibotta alone the cell is rung 1. | contradicted | circana / placement_coverage -> 1 (grade D) | Anchor 1 (one or two; one where 2 applies): one network has Media Lift measurement documented with the service named, the Ibotta Performance Network: Circana's Chomps case runs offers 'across the Ibotta Performance Network' 'using Circana Household Lift for independent measurement of incremental impact' (undated), and the Circana-Ibotta joint release of 30 September 2025 names 'Circana Household Lift studies' for Ibotta campaigns, rolling out in late 2025 and 'available to all brands by early 20 | data/scores.json or data/matrix-extra.json: circana, placement_coverage |
| C569 | Criteo Commerce Max / governance = 2 (grade C) | Separate refute-mode cell verifier for scoring slice DE. | The items found are real, but the screen is partial (G1's own status) and no Commerce Max terms were read; dimension 11 is 'Scored only after a completed, bounded disclosure screen of the service's documentation and terms; an incomplete screen is n/e' (ruling 3). | contradicted | Criteo Commerce Max / governance -> ne | Screen not completed: stream G1 records a partial screen of Commerce Max's help centre (categories, 2026 release notes, five guide bodies; offsite overview refused, now known to sit behind an SSO login) and no Commerce Max terms were read, so the bounded disclosure screen the dimension requires was not completed. Recorded beside the cell, not scored: dated monthly 2026 release notes, a dated reporting revision (offsite attribution and reporting moved from 9 to 72 hours in February 2026, reportin | data/scores.json or data/matrix-extra.json: Criteo Commerce Max, governance |
| C570 | Criteo Commerce Max / privacy = 1 (grade C) | Separate refute-mode cell verifier for scoring slice DE. | Commerce Max has its own privacy notice in its help centre, which the cell missed: its Privacy FAQs (re-opened here; passages verbatim) document a consent basis for the product (TCFv2; no user-level tracking for users who refuse) and name GDPR; the value is 3, not 1. | contradicted | Criteo Commerce Max / privacy -> 3 (grade C) | Anchor 3: a consent basis is documented for the product: Commerce Max's own help-centre Privacy FAQs (last published 27 March 2024, still served on 3 November 2026) say logged-in shoppers are identified by 'an agreed upon identifier, such as a hashed customer id' used 'for targeting and attribution purposes based on first-party data', that Criteo's 'preferred solution is to rely on the IAB's Transparency and Consent Framework v2.0 (TCFv2)', and that 'For users who do not consent to any form of t | data/scores.json or data/matrix-extra.json: Criteo Commerce Max, privacy |
| C571 | Intentwise Optimize / governance = 2 (grade C) | Separate refute-mode cell verifier for scoring slice DE. | G1's screen is partial (help index and seven articles; no terms), so under ruling 3 the cell is n/e; this removes Intentwise's only Standing cell. | contradicted | Intentwise Optimize / governance -> ne | Screen not completed: stream G1 records a partial screen (the help index and seven articles) and no Intentwise terms were read, so the completed, bounded disclosure screen the dimension requires is missing. Recorded beside the cell, not scored: a metric revision (ROAS redefined to 'Total Attributed Sales / Spend', undated) and disclosed Amazon-side corrections ('Negatives are corrections from Amazon's click and conversion validation process'); the index also lists change-history articles about a | data/scores.json or data/matrix-extra.json: Intentwise Optimize, governance |
| C572 | Perpetua / governance = 2 (grade C) | Separate refute-mode cell verifier for scoring slice DE. | Partial screen (home, three collections, four articles; no terms): n/e under ruling 3. | contradicted | Perpetua / governance -> ne | Screen not completed: stream G1 records a partial screen (help home, three collections, four articles) and no Perpetua terms were read, so the completed, bounded disclosure screen the dimension requires is missing. Recorded beside the cell, not scored: disclosed reporting revisions on the Walmart feed ('Conversions may be added retroactively'; a 1-2 day lag with the last two days withheld) [RTL-G1-S24, RTL-G1-S01, RTL-G1-S51; G1 table governance_screen]. | data/scores.json or data/matrix-extra.json: Perpetua, governance |
| C573 | Attain / default_method = 2 (grade D) | Separate refute-mode cell verifier for scoring slice DE. | With Sales Attribution (no counterfactual) and Incrementality (holdout, construction undisclosed) sold side by side and no default named, only the rung every offered product meets is evidenced (ruling 7); the 2 credits the optional product. | contradicted | Attain / default_method -> 1 (grade D) | Anchor 1: documented exposure-to-purchase attribution with no campaign counterfactual is the floor of what the buyer receives: the measurement page sells 'Sales Attribution' ('Did exposed consumers actually buy? Tie ad exposure to real transactions across retailers and channels') beside a separately listed 'Incrementality' product ('Measure true causal lift using advanced holdout methodologies'), and names neither as the standard design; the register's purchased service and the retail page descr | data/scores.json or data/matrix-extra.json: Attain, default_method |
| C574 | Attain / privacy = 1 (grade C) | Separate refute-mode cell verifier for scoring slice DE. | The cited policy excludes the product's data by its own scope sentence ('applies to Attain's collection and use of its Business to Business partners' personal information'); with the panel-app policies unopened the shopper-data record is silent, and a silent product record is n/e, not 1. | contradicted | Attain / privacy -> ne | Product record silent on shopper data: Attain's privacy policy (effective 16 April 2026) 'applies to Attain's collection and use of its Business to Business partners' personal information', so its CPRA, sale and 'SOC 2 Type II' statements do not cover panel members; the consumer privacy index only points to the Klover, Frisbee and Merryfield app policies on separate domains, none opened. A company policy that by its own scope excludes the data the product uses is not anchor 1's generic policy; t | data/scores.json or data/matrix-extra.json: Attain, privacy |
| C575 | Criteo Commerce Max / us_availability = n/e | Cell verifier for scoring slice DE scanned the n/e cells. | 3 (grade C, date 2026-06): Commerce Max's own 2026 release notes list a June 2026 'Digital Shelf Intelligence Report. Currently available to US & Canada users only', and Criteo's 2024 releases name Commerce Max as the self-service DSP for Albertsons and Dollar General and as the route for Belk's closed-loop reporting; US availability is documented, production use by named US advertisers in the period is not (blocks 4). The cell's 'No US statement for Commerce Max' is wrong. | structural | Criteo Commerce Max / us_availability: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Criteo Commerce Max, us_availability |
| C576 | Criteo Commerce Max / validation = n/e | Cell verifier for scoring slice DE scanned the n/e cells. | 1 (grade C, date 2026-02): accuracy claims only; the release note says the 72-hour change will 'provide a more complete and accurate view of campaign performance', the same kind of claim scored 1 for Flywheel, Quartile and Stackline. | structural | Criteo Commerce Max / validation: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | Criteo Commerce Max, validation |
| C577 | NCS Sales Effect / validation = n/e | Cell verifier for scoring slice DE scanned the n/e cells. | 1 (grade D, date 2026-11): accuracy claim only ('a precision that delivers the intelligence on how the creative, audience, frequency and media impact your campaign outcomes'), the counterpart of Media Lift's 'the most precise', which the slice scores 1. | structural | NCS Sales Effect / validation: cell added by the cell verifier where its quoted passage parses | The evidence files already held a finding that scores the cell (session ruling 6). | NCS Sales Effect, validation |
| C578 | H1a coding: "closed loop" with no metric named | The E2 verifier found uneven coding across rows (Chewy, Leap, DICK'S, Petco). | structural | A default outcome method counts as documented when the row's own documentation names what standard reporting delivers (an attributed metric or a lift design); a bare "closed loop" claim naming no metric is term only, undocumented in the primary reading and attribution in the lenient one. Both readings are reported. | Applied uniformly across rows and disclosed in the method chapter. | data/hypotheses.json H1a; fig-control-design; chapter 3 | |
| C579 | Ruling 8: rung 1 of dimension 1 | Two scoring slices coded similar wording differently (Petco vs UB Media). | structural | Dimension 1 = 1 when the row's own documentation names an attributed metric in its standard reporting and documents no counterfactual; an undesigned lift label neither raises nor blocks it; a bare "closed loop" is n/e. | Applied uniformly across rows and disclosed in the method chapter. | every dimension-1 cell; H1a | |
| C580 | Ruling 3: rows selling products at different rungs | Cell verifier for cohorts D and E (Attain). | structural | Where a row sells products, routes or formats at different rungs and names no default, only the rung every one meets is evidenced (e.g. Kroger self-serve vs managed; Walmart Store Ads). | Applied uniformly across rows and disclosed in the method chapter. | dimension 1; H1a | |
| C581 | Ruling 1: governance needs a completed screen of documentation and terms | Cell verifier for cohorts D and E. | structural | Dimension 11 is scored only after a completed, bounded screen of the service's documentation and terms (the frozen definition); a screen that did not read the terms is incomplete and the cell is n/e, including a scored 1. | Applied uniformly across rows and disclosed in the method chapter. | every dimension-11 cell | |
| C582 | Ruling 5: undated pages | Cell verifiers. | structural | A cell resting on an undated page carries the read month (2026-11) as its date. | Applied uniformly across rows and disclosed in the method chapter. | cell dates | |
| C583 | Ruling 6: cells added by cell verifiers | Cell verifiers found n/e cells the evidence already scored. | structural | A cell a verifier documents with a quoted passage from an existing finding is adopted as qualified, never verified. | Applied uniformly across rows and disclosed in the method chapter. | data/scores.json, data/matrix-extra.json | |
| C584 | Ruling 10: US availability at rung 5 | Cell verifiers B and A1 read similar filing sentences differently. | structural | Rung 5 of dimension 9 needs availability to new US buyers documented for the row itself; a filing earns grade A for that step only if it names the row's own service as offered. | Applied uniformly across rows and disclosed in the method chapter. | dimension 9 | |
| C585 | H4b coding rule | The H stream raised the boundary; the session fixed the rule on 3 November 2026 before the count. | structural | A row counts as documenting an outside identity match only when its own documentation (or its executive on record in a partner release) states that the off-site exposure-to-transaction join relies on an identity match outside its logged-in surface; network-level privacy statements and keys without a stated join do not count. Sensitivities report own pages only, current statements only and clean-room wording. | Applied uniformly across rows and disclosed in the method chapter. | H4b; fig-offsite-rails; chapter 6 | |
| C586 | H4b: Kroger offsite recoded (ruling 9) | The chapter 6 draft found the roadmap entry, which the H stream's tally had not included. | structural | Kroger Precision Marketing offsite programmatic is credited on the 84.51 roadmap entry "measure the impact of media on Meta in a secure clean room environment" (completed Q1 2025), the same kind of statement as Walmart's credited Meta clean-room sentence; the literal count moved from 4 to 5 of 13 and H4b from contradicted to unresolved. The rule did not change. | Applied uniformly across rows and disclosed in the method chapter. | H4b; fig-offsite-rails; chapter 6 | |
| C587 | Scale: residual "Other" revenue lines | The Albertsons profile coded n/e; its verifier noted the literal perimeter code. | structural | Kroger's and Albertsons' "Other" lines span non-advertising revenue, so both are coded broader parent proxies capped at 3, with n/e logged as the sensitivity; Albertsons' line is not a superset of its media revenue. | Applied uniformly across rows and disclosed in the method chapter. | dimension 8 for the Kroger and Albertsons rows | |
| C588 | Interest marks under the frozen conflict screen | Scorers set marks unevenly. | structural | Rows tied to a Samba TV listing (Walmart Sponsored Search, Walmart DSP, Walmart Offsite Display; Amazon Ads) and rows resting on Omnicom-owned sources (Flywheel, Perpetua; Instacart findings citing them) carry the interest mark; it never alters a score and the rows are excluded in the sensitivity run. | Applied uniformly across rows and disclosed in the method chapter. | data/matrix-extra.json, data/profile-overlay.json | |
| C589 | Register amendment v1.1 | The Circana profile found it separately branded and sold. | structural | NCS Sales Effect added to the register as a separate cohort E row before its scores were seen (data/register-amendments.json). | Applied uniformly across rows and disclosed in the method chapter. | data/product-register.json | |
| C590 | Consolidation record: Publicis-LiveRamp counted as open (FTC step of 19 October 2026 plotted, no close found); 8 completed, 5 open. | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-01. | Publicis Groupe's release of 2 November 2026 says it "today announced the closing of its acquisition of LiveRamp"; its release of 17 May 2026 gives the agreement (enterprise value US$2.2 billion). Catalina's Italian site, 26 February 2026: "Da oggi Catalina è entrata a far parte di Infillion". Foursquare's home page carries an undated "has joined Infillion" banner. DoorDash's 10-K records the purchase of Symbiosys, "a retail media platform company", on 28 May 2025. | contradicted | LiveRamp and Catalina moved to completed (LiveRamp: agreement 17 May, FTC step 19 October, closing 2 November 2026; Catalina: announcement 23 February, completion post 26 February 2026); Symbiosys added as a completed acquisition the consolidation search had missed; Foursquare kept open, its undated banner recorded beside it. Counts: 11 completed, 3 open, 35 dated events. | A dated statement by a party settles a completion; an undated banner does not date one. | fig-consolidation; chapters 2, 6, 10, abstract, plain edition; claims RTL-R3-F01 to F04 |
| C591 | H4b: Kroger Precision Marketing offsite programmatic credited on 84.51°'s roadmap entry ("measure the impact of media on Meta in a secure clean room environment"); Roundel credited on its executive's quote in LiveRamp's release; 5 of 13, unresolved. | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-02. | RTL-KROGER-F22 (verified as stated) names no purchase outcome; RTL-X-F08 (qualified): Roundel's executive names "first-party guest insights", the sales and ROAS wording is LiveRamp's. The credited Walmart (RTL-H-F06: closed-loop measurement of online and in-store sales), Albertsons (RTL-H-F14: exposure joined "with Albertsons transaction data") and CVS (RTL-X-F12: attributed Pinterest sales) statements name the purchase side. | contradicted | The H4b coding rule applied as written to every credited row: the purchase side of the join must be named. Kroger and Roundel lose the literal credit (kept in the looser reading). Literal 3 of 13 (23.1%): contradicted; stricter readings 1 or 2; looser 8 (unresolved); broad 3 of 22. | The rule asks for the exposure-to-transaction join; a clean room named for "impact" or "insights" does not state it. | data/hypotheses.json H4b; fig-offsite-rails; chapters 6, 13, 14, 15, abstract, plain edition; site texts |
| C592 | H1a and dimension 1: Chewy Ads, Gulp Media Network and Leap Media Group coded documented attribution under ruling 8. | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-03. | RTL-E2-F01: Chewy's pages say it "provides insights like New-to-Brand, Autoship, Lifetime Value ROAS, and incrementality" and "do not mention dashboards or reporting"; RTL-E2-F47: Gulp describes two future-tense constructs and names neither the default; RTL-E2-F52: Leap offers "Measurable Analytics, including ROAS" with no report named. | contradicted | Ruling 8's condition (a metric named in standard reporting) applied as written: the three rows coded unknown for H1a and n/e on dimension 1. H1a 36 of 37 (97.3%), network 20 of 20, lenient 41 of 42; still supported. Fig-number-kinds keeps their metric kinds as offered metrics. | An offered metric is not a documented default. | data/matrix-extra.json (three default_method cells); fig-control-design; chapters 1, 3, 12, 13, 14, 15, abstract, plain edition; site texts |
| C593 | Dimension 11: Cardlytics scored 2 for "documented reporting revisions" on data-delay statements; invoice-only dispute rules credited at Instacart and Klarna. | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-04. | RTL-E3-F14: purchases "can have up to 7-14 days delay for transaction data completeness"; performance data "updated hourly after an initial 4-day delay". Chapter 4 already read these as numbers that move without restatement. | contradicted | Ruled for every row: a schedule of when numbers settle is not a documented revision (Cardlytics governance 2 -> 1; RQ 47.99 -> 45.74, Fetch first in cohort B at 46.77, bands unchanged); an invoice dispute procedure counts as a challenge procedure, because the counts a service bills on are among its own numbers (Instacart and Klarna unchanged). | The definition covers how a service governs its own numbers, billed counts included; a latency note revises nothing. | data/scores.json cardlytics.governance; chapters 4, 12, 14 |
| C594 | The two Most documented rows rest on one counted assurance cell each; Moloco's 4 is read from the certifier's programme rules, not from the certificate. | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-05. | RTL-A-F39: the IAB Europe certificate names no standards version, metric list, deployment or geography; the programme's mandatory requirements include sale definitions, ROAS and lookback windows (RTL-A-S31, RTL-A-S30). RTL-A-F30: AAM names DoorDash's object in its own words. | supported with qualification | Basis stated beside the ranking (chapter 12) and the scope sensitivity published, each cohort recomputed: Moloco at rung 3 = 53.55, Well documented; n/e = 52.75, thin record. DoorDash at rung 3 = 51.94, Most documented; n/e = 50.85, thin record. Scores unchanged. | Each top band rests on how one cell is read; the reader gets the test. | fig-quotient-data (top_assurance_sensitivity); chapter 12, abstract, plain edition |
| C595 | H2 reported supported on the 3 November screen, with the Epsilon mapping disclosed as an inference. | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-06. | MRC's letter of 1 October 2025 names only "Epsilon"; the product identification rests on Epsilon's release of 11 May 2021 (RTL-A-F04). The decision order lets support stand only when every applicability check is resolved. | contradicted | H2 reads unresolved at step two; the supported reading, which accepts the inference, is the sensitivity. Headlines scoped to the delivery metrics of the seven mapped MRC approvals (AAM's DoorDash certification is not MRC accreditation). | An inference from an older vendor release is not a resolved applicability check. | data/hypotheses.json H2; chapters 7, 13, 14, 15, abstract, plain edition; site texts |
| C596 | Quotient deletion test held each cohort's band eligibility fixed. | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-07. | Cohort B has exactly six eligible rows; deleting the privacy cell of Chase Media Solutions, Fetch, Ibotta or Upside (each row's only Standing cell) leaves five, below the six bands need. | supported with qualification | build/quotient.mjs now flags cohort_bands_lost per deletion and counts cohort_band_losses (cohort B: 4; other cohorts: 0); each deletion case carries the row id; a quotient test covers it; chapter 12 and the instrument text report it. | The row-level band after a deletion missed the cohort gate. | build/quotient.mjs; scripts/quotient-test.mjs; data/rq-missing-evidence.json; chapter 12 |
| C597 | H6: the N = 8 sensitivity dropped R05, R07 and R11; a literal format-level shopping and US screen would drop R04, R07 and R11. Figure and plain wording said no format "has" the external functions generally available. | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-08. | RTL-K-F31 (R07: Amazon's page, shopping wording generic), RTL-K-F51 (R11: "English-speaking markets", Global data); R04's shopping use is stated for the platform (frozen population: included with that caveat). | supported with qualification | The stricter screen added as a sensitivity (N = 8): H6a supported, H6b to H6d unresolved. Wording now "explicitly described as generally available" and "documents explicit general availability". The registered H6b to H6d statements already state the rule, so they are unchanged. | The frozen population is kept; the alternative is reported. | fig-agent-surfaces; chapter 11, abstract, chapter 13, plain edition; RTL-HYP-H6a to H6d |
| C598 | Residual "Other" revenue lines: n/e sensitivity logged but not published; Cardlytics US availability 5 rested partly on vendor pages for new-buyer availability. | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-09. | Five scale cells (Kroger onsite and offsite; Albertsons onsite, offsite and in-store) rest on residual lines. Cardlytics' fiscal 2025 10-K: "Through the Cardlytics platform, our financial media network, marketers can deliver advertising content", "we are able to offer marketers the ability to optimize their marketing efforts" and "We maintain the Cardlytics platform in both the United States". | supported with qualification | Sensitivity published: with the five cells n/e and cohort A recomputed, no band moves; largest change 0.80 points (Kroger offsite 42.52 -> 43.33). Cardlytics keeps 5 (grade A) on the 10-K passages, now quoted in the cell (ruling 10 as written). | The residual-line mapping is a post-extraction judgement and is reported as such. | fig-quotient-data (residual_line_sensitivity); data/scores.json cardlytics.us_availability; chapters 12, 14 |
| C599 | Interest marks shown in the HTML matrix and figures but dropped from the Markdown matrix and data/rq.csv; the interest sensitivity and its effect on H3 not surfaced. | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-10. | The interest-excluded run moves three cohort A bands (Roundel, Orange Apron Media, Amazon display ads: Partly -> Well documented; field 38.60 -> 37.04); setting aside the H3 subset services on marked rows removes Amazon Sponsored Products, the one counterexample, and leaves two incomplete searches. | supported with qualification | The ◆ mark carried into the Markdown matrix and an interest_mark column into rq.csv; the band moves and the H3 diagnostic (unresolved without the marked rows) reported in chapters 5, 12 and 14 and RTL-HYP-H3. | A mark never alters a score; the reader needs to see what the marked rows carry. | build/instruments.mjs; build/make-csv.mjs; fig-access-ladder; chapters 5, 12 |
| C600 | Plain edition said more than the paper in places (lift "is" what the ad caused; unscoped absences; "fully launched"; no line between the rule freeze and later rulings). | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-11. | Each sentence compared with the paper's chapter text and the claim it rests on. | supported with qualification | Plain wording scoped: a sound test estimates the effect; H4b tests what services document; the market-series absence bounded to the series screened; the Amazon claim scoped to six services; measurement firms' standard tests; the lift-offer count with its one default; rules fixed before the work and dated rulings after. | The plain edition must not say more than the paper. | plain.md |
| C601 | Chapter 2 counted "three completions after 3 November"; DoorDash's 10-K recorded as not read. | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-12. | The figure has two scheduled dates and one expected window after the status date. DoorDash's fiscal 2025 10-K recognises advertising products "as part of our Marketplaces revenue" with no amount (every mention read). | supported with qualification | Chapter 2 now says 35 dated events through 3 November, then two scheduled dates and one expected window; the DoorDash scale cells and chapter 2 cite the 10-K (scale stays n/e). | The ledger had a filing-read gap. | chapter 2; data/matrix-extra.json DoorDash scale cells |
| C602 | Print edition: sparse body pages and a one-row table continuation. | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding RTL-R3-13. | The packaged PDF of 3 November 2026 was inspected page by page after the rebuild. | supported with qualification | The rebuilt PDF was checked page by page with a page-fill script (body bottom per page, and what the next page starts with). Fixed: the denominators figure (print part 1) overflowed its page by three caption lines, so its caveat now prints under part 2; chapter 3’s print figure now follows the section’s paragraphs instead of leaving a page with one paragraph; the New York privacy row was tightened so the rules table no longer carries one row onto its own page. The short pages left are chapter ends (the abstract, chapters 4 and 13, the last page) and pages before a figure or table that needs a page of its own; scripts/pdf-type.py now lists them after every build. | The PDF type gate measured type size, not page fill; it now also lists short pages to inspect. | scripts/figs/05-denominators.mjs; chapters/03-default.md; data/rules-overlay.json (New York row); scripts/pdf-type.py |
| C603 | Chapter 14 said "No outside party checked the evidence or the scores". | Final AI review of 3 November 2026 (a model from another provider, given the full package), finding Review disclosure. | The final AI review is an outside model's check, not an audit. | structural | Chapter 14, the abstract, the plain edition and the site footer now say a separate AI review spot-checked selected claims against primary sources, tested the scoring rules and inspected the print output, that every point was checked at source, and that the paper is not audited or peer reviewed. | Describe AI-assisted review honestly. | chapter 14; abstract; plain edition; paper.config.json footer |
| C604 | H2 in the assurance figure: fig-assurance-map and its data said H2 "supported" while chapter 7, the appendix and the plain edition said unresolved. | Review of the revised draft, 4 November 2026 (a model from another provider), point: H2 in the assurance figure. | scripts/figs/03-assurance-map.mjs hard-coded the reading. | contradicted | The figure now reads H2 from its own record (unresolved; supported only if the Epsilon mapping is accepted), and its title says "mapped to an eligible service"; both figure editions and data/figure-data.json regenerated. | Figures must agree with the hypothesis record. | fig-assurance-map; data/figure-data.json#assurance |
| C605 | Dates: Figure scopes and the site configuration called 20 November 2026 the evidence freeze although the evidence is dated 3 November. | Review of the revised draft, 4 November 2026 (a model from another provider), point: Dates. | paper.config.json cutoff and three figure scopes. | contradicted | Cutoff set to 3 November 2026 (publication 4 November); figure scopes say research cutoff 3 November and status rechecks on 20 November; the consolidation caveat says open statuses are rechecked on 20 November. | Keep the cutoff and the planned rechecks distinct. | paper.config.json; fig-agent-surfaces; fig-protocol-states; fig-consolidation; chapter 14 |
| C606 | H4b, Walmart row mapping: The Walmart LiveRamp sales link is documented for Social, which Walmart presents as a tactic apart from Offsite Display. | Review of the revised draft, 4 November 2026 (a model from another provider), point: H4b, Walmart row mapping. | Walmart Connect, 8 April 2026: "Closed-loop measurement via LiveRamp's data clean room is now available for Meta" under social; the frozen register defines the Offsite Display row from Walmart's offsite page, whose channels include "social media (Meta, TikTok, Pinterest)". | supported with qualification | Credit kept on the register's definition, now stated in chapter 6 and the figure; the Disney 2024 proof of concept carries it alone; without both statements the count would be 2 of 13, still contradicted. | The register row, frozen before extraction, covers Walmart's off-site social channels. | chapter 6; fig-offsite-rails notes; RTL-HYP-H4b |
| C607 | H4b, Roundel partner evidence: The review read LiveRamp's release as naming Roundel and first-party sales in one passage. | Review of the revised draft, 4 November 2026 (a model from another provider), point: H4b, Roundel partner evidence. | Re-read 4 November 2026: the first-party sales sentence speaks of retail media networks in general; Roundel is named only in its VP's quote ("first-party guest insights"). | supported with qualification | Added as a partner-release sensitivity (release read as a whole): 4 of 13, contradicted; the literal coding is unchanged. | The point holds as a sensitivity, not as a literal credit. | chapter 6; fig-offsite-rails (derived count); RTL-HYP-H4b |
| C608 | Deal dates: Catalina's 26 February date is an Italian confirmation; Foursquare is better described as "dated close unverified" than "open". | Review of the revised draft, 4 November 2026 (a model from another provider), point: Deal dates. | Catalina Italia's post is an affiliate's notice; Foursquare's banner is undated. | supported with qualification | Catalina drawn as completed by 26 February 2026 (confirmation date, not the group's legal close) and kept in the completed tally; the open lane is now "agreed or announced, no dated close", and the text says Foursquare's close may have happened but is undated. | Date what the source dates. | fig-consolidation; chapter 2; abstract; plain edition; RTL-R3-F02, F03 |
| C609 | Plain wording: The plain edition restated the H4b test as "use such a match". | Review of the revised draft, 4 November 2026 (a model from another provider), point: Plain wording. | The registered test is about what services document. | supported with qualification | Plain edition: "most services document such a match". | The plain edition must not say more than the paper. | plain.md |
The corrections ledger has 609 entries. The print edition leaves it out; it is listed in full in the web edition and in data/corrections-ledger.csv, with the original wording, the challenge, the evidence checked, the verdict, the change made, how it was implemented and how it was verified.
Data and downloads
Every data file behind the numbers, as CSV, with the print and Markdown editions.
Downloads
The data files are in the data/ folder beside this PDF, in the package and on the web page. Open the HTML edition to use these as links.
The two other editions, then the data as CSV and JSON, UTF-8. The data files are the canonical files every chart, matrix and count in the paper was built from.
- 1retail-media-2026.pdf — the print edition: A4, with page numbers and chapter bookmarks
- 2retail-media-2026.md — the reading edition in Markdown, with numbered references
- 3data/source-register.csv — every source with publication, event and access dates, type, geography, funding, access limits and reliability notes
- 4data/claim-ledger.csv — every material claim with sources, passage locator, scope, evidence class, review outcome, confidence and final wording
- 5data/capability-matrix-wide.csv — vendors × dimensions, scores only (0–5, n/e, n/a)
- 6data/capability-matrix-long.csv — one row per cell with availability, evidence grade, date, rationale and sources
- 7data/capability-dimensions.csv — the dimension definitions and anchors
- 8data/rq.csv — Receipt Quotient per vendor: group means, coverage, band, the reweightings and every dimension score
- 9data/rq-method.csv — RQ groups, weights, shrinkage, field mean, band thresholds, and the reweighting, shrinkage and missing-evidence tests
- 10data/rq-missing-evidence.csv — every scored cell turned into n/e one at a time, with the RQ before and after
- 11data/swot.csv — every SWOT line for the profiled vendors with its basis (record, vendor documentation, vendor claim, inference) and sources
- 12data/deep-profiles.csv — the deep-profile summaries: what is sold, buyers, earnings, dependencies, fit, limits
- 13data/deep-profile-records.csv — every profiled capability, performance claim, financial figure and timeline event with sources
- 14data/study-evidence.csv — studies with design, sample, sponsor and limits
- 15data/corrections-ledger.csv — every change made after verification and challenge passes, with implementation and verification
- 16data/vendor-universe.csv — the reasoned vendor sample with scope group, owner, formats, routes and status
- 17data/agentic-map.csv — AI and agentic capabilities with status labels and observability
- 18data/rules.csv — rules and governance instruments by jurisdiction
- 19data/estimates.csv — market-size estimates with inclusions, exclusions and combinability
- 20data/product-register.csv — the product register: every scored or listed product with its purchased service, operating entity, owner, cohort, inclusion reason, aliases and US availability
- 21data/hypotheses.csv — the hypotheses registered before extraction, with the rules for supported, contradicted and unresolved
- 22data/figure-data.json — the canonical dataset behind every data figure
- 23data/derived-checks.json — every derived number with its formula and inputs
- 24data/stats.json — every count the prose and figures quote about the research itself, computed from the ledgers
- 25data/scoring-rules.json — dimension definitions, anchors, cell rules and the RQ definition, fixed before scoring
Appendix
Appendix: hypotheses, universe and downloads
The registered hypotheses and how they read
The hypothesis register was frozen on 3 November 2026, after the vendor-universe sweep had built the product register and before any claim was extracted. Each entry states its population, comparator and observable outcome, with the rules for supported, contradicted and unresolved and, where needed, a decision order (chapter 14). H2, H3 and H4b test what the public record documents, so an absence in the screened documentation is their evidence; for the others an absence leaves the question unresolved. Clarifications were made after extraction began, all on 3 November 2026: the rule for a documented default under H1a, the rule for a documented identity match under H4b and the recoding of one H4b row on evidence found later. After the final AI review the same day, the H1a rule was applied as written to three rows, the H4b rule's purchase side to two credited rows, and H2's Epsilon check was read as unresolved. Each is dated in data/hypotheses.json beside the registered rule and set out in chapter 14, and none changed a threshold.
| Registered statement, in short | Reading | Basis | Chapter | |
|---|---|---|---|---|
| H1a | Among cohort A services that document a default outcome method, at least two-thirds report attributed sales with no counterfactual | Supported | 36 of 37 documented defaults (97.3%) and 20 of 20 networks; holds under the lenient reading and every sensitivity | 3 |
| H1b | Controlled lift is a conditional offer, recorded in six flags per cohort A row | No verdict (descriptive) | 22 documented offers; 21 leave assignment unknown and 1 discloses random assignment, on Instacart pages of 2022 and 2023 | 3 |
| H2 | No MRC accreditation held by an eligible cohort A, B or C service covers an outcome metric | Unresolved; supported if the Epsilon mapping is accepted | Seven eligible services hold accreditation for delivery and invalid-traffic metrics only; the one outcome-type entry, "Epsilon", names no product and is tied to a service outside the register only by inference | 7 |
| H3 | No on-site service in a pre-registered subset of major US networks documents event-level exposure data for the buyer or an independent measurer, or buyer-controlled outcome analysis | Contradicted at step 1 | Amazon Sponsored Products documents buyer-run queries in Amazon Marketing Cloud, with only aggregated outputs leaving; none of the six services documents event-level exposure data | 5 |
| H4a | In US series that split on-site, off-site and in-store spending on one definition across 2023 to 2026, the off-site share rose | Unresolved | No eligible series among the six screened | 2, 6 |
| H4b | At least two-thirds of cohort A services that document off-site outcome measurement document an identity match outside the operator's logged-in surface | Contradicted | 3 of 13 (23.1%); every stricter reading 1 or 2 of 13, the loosest 8 of 13 | 6 |
| H6a | Every explicitly generally available sponsored format on an agent surface has operator reporting | Supported | Amazon's two store prompt formats, with reporting generally available in Amazon's words | 11 |
| H6b | Third-party attribution for sponsored formats on agent surfaces is mostly none or pilot | Unresolved | None of 11 generally available; 1 stage unstated, 10 unknown | 11 |
| H6c | Causal outcome measurement for those formats is mostly none or pilot | Unresolved | None of 11 generally available; 1 pilot, 10 unknown | 11 |
| H6d | Ad-quality verification by a named third party for those formats is mostly none or pilot | Unresolved | None of 11 generally available; 1 pilot, 10 unknown | 11 |
The five coded questions carry no verdict and are answered by records and counts: Q1, what kind of number each reported outcome metric is, in chapter 3; Q2, who produces the number and on whose definitions, in chapter 4; Q3, consolidation and exits since 2023, and Q4, which series and filing lines size the market and on what definition, in chapter 2; Q5, what surveys and studies with stated methods say about retailer influence and joint business plans, in chapter 8. Two services named in the table carry the author's interest marks: Amazon Sponsored Products, because he works at Samba TV, which Amazon lists among its measurement vendors , and Instacart, which names Omnicom's Flywheel as an agency partner; his past employer, Hearts & Science, was an Omnicom agency, so that disclosure is voluntary . Chapter 14 lists every marked row, and the marks change no code or score.
The register and the universe
The product register was frozen on 3 November 2026 with 131 records and amended once the same day, before any score was seen, to add NCS Sales Effect as a cohort E row ; it now holds 132 records. Its rows are purchasable services, not networks: 107 service rows in five cohorts (49 retailer and marketplace services across 31 networks, 20 commerce media services, 20 enablement platforms, 11 cross-network optimisation platforms and 7 outcome measurement providers), 22 listed records and 3 historical ones. A network's services become separate rows only where their default method, analytical access or assurance scope differ, and evidence never transfers between rows (chapter 1). 91 rows had at least four counted cells and were admitted to the matrix; 16 were not.
The register came from one bounded sweep, workstream U, which screened every company and service named in the brief's seed list, a reviewer's leads or its own discovery for status, ownership, cohort, US availability and scale. It is a reasoned sample, not a census. The full universe table, with owner, operating entity, cohort, status and the matrix row of each scored service, is in chapter 1.
Downloads
Chapter 14 links every download. The build writes them from the same data files the text and figures are built from; the CSV files are UTF-8, comma-separated and quoted per RFC 4180.
- The editions: the print edition as an A4 PDF with page numbers and chapter bookmarks, and the reading edition in Markdown with numbered references.
- The registers: every source with its publication, event and access dates, type, geography, funding, access limits and reliability notes; every product record with its purchased service, operating entity, owner, cohort, inclusion reason, aliases and US availability; and the fifteen hypotheses and questions with their rules and dated clarifications.
- The ledgers: every claim with its sources, passage locator, scope, evidence class, review outcome, confidence and final wording, the hypothesis readings marked by hypothesis; and every entry in the corrections ledger, with the original wording or cell, what questioned it, the evidence, the verdict, the change and how it was implemented and checked.
- The matrix: scores only, one row per service; one row per cell with availability, evidence grade, date, verification status, rationale and sources; and the dimension definitions with their anchors.
- The quotient: the Receipt Quotient of each row with its group means, coverage, band, reweightings, missing-evidence fills and dimension scores; the weights, prior strength, field means and band thresholds with the reweighting, shrinkage, deletion and missing-evidence tests for each cohort; and every counted cell turned into n/e in turn, with the quotient before and after.
- The profiles: the eight deep-profile summaries, every profiled capability, performance claim, financial figure and timeline event with its sources, and every SWOT line with its basis.
- The instrument tables: the thirteen studies with design, sample, sponsor and limits; the market estimates with inclusions, exclusions and combinability; the rules by jurisdiction; the eleven agent-surface records on each coded axis with stated status, date and source; and the vendor universe.
- The data files, in JSON: the dataset behind every data figure, every derived number with its formula and inputs, every count the text quotes about the research itself, and the scoring rules fixed before scoring.
References
Numbered sources
The 910 sources cited in this paper, numbered in order of first citation. The full register of 1058 sources opened during the research, with access dates, access limits and reliability notes, is in data/source-register.csv. Click any number in the text to open the same record.
910 numbered sources
- [1]Author’s work · Retail media network measurement, compared (standards table). Evgeny Popov (No Fluff Advisory); the author of this paper. https://nofluffadvisory.com/standards/retail-media-network-measurement-comparison/ Accessed 2026-11-03.
- [2]Sponsored Display is now known as display ads (product page). Amazon Ads. https://advertising.amazon.com/solutions/products/sponsored-display Accessed 2026-11-03.
- [3]Citi Expands Customer Engagement and Commerce Media Capabilities with Addition of Kard. Citigroup Inc. (U.S. Consumer Cards business), 2026-08-13. https://www.citigroup.com/global/news/press-release/2026/citi-expands-customer-engagement-commerce-media-capabilities-kard-addition Accessed 2026-11-03.
- [4]EDGAR full-text search: "global advertising business" in Walmart Inc. filings (CIK 104169). U.S. Securities and Exchange Commission (EDGAR full-text search). https://efts.sec.gov/LATEST/search-index?q=%22global%20advertising%20business%22&ciks=0000104169 Accessed 2026-11-03. Access: partial extract.
- [5]The Kroger Co. 10-K for fiscal 2025, XBRL report R35 'ACCOUNTING POLICIES (Policies)' and R54 'ACCOUNTING POLICIES - REVENUE RECOGNITION (Details)'. The Kroger Co. (SEC EDGAR), 2026-03-31. https://www.sec.gov/Archives/edgar/data/56873/000110465926037723/R35.htm Accessed 2026-11-03. Access: partial extract.
- [6]Albertsons Companies, Inc. Form 10-K for fiscal 2025 (53 weeks ended 28 February 2026). Albertsons Companies, Inc. (SEC EDGAR, CIK 1646972), 2026-04-27 (event 2026-02-28 (period of report)). https://www.sec.gov/Archives/edgar/data/1646972/000164697226000032/aci-20260228.htm Accessed 2026-11-03.
- [7]OMNICOM MEDIA OFFICIALLY LAUNCHES HEARTS UNITED. Omnicom Media (Omnicom Media ByDesign site), 2026-08-28. https://bydesign.omnicommedia.com/?p=16241 Accessed 2026-11-03.
- [8]Citi Unveils Citi Commerce Media to Deliver More Personalized Experiences for Customers and Brands. Citigroup Inc. (U.S. Consumer Cards), 2026-09-23 (event 2026-09-23 (launch)). https://www.citigroup.com/global/news/press-release/2026/citi-commerce-media-deliver-more-personalized-customer-brand-experiences Accessed 2026-11-03.
- [9]Cardlytics, Inc. Form 10-Q for the quarter ended 30 June 2026 (cdlx-20260630.htm; accession 0001628280-26-053340). Cardlytics, Inc. (SEC EDGAR), 2026-08-05. https://www.sec.gov/Archives/edgar/data/1666071/000162828026053340/cdlx-20260630.htm Accessed 2026-11-03. Access: partial extract.
- [10]Overview of attribution (Walmart Connect Advertising Help, article 000012578). Walmart Connect (Walmart Inc.), undated on the page (describes changes effective 1 June 2026 and a deprecation on 1 September 2026, partly in future tense). https://advertisinghelp.walmart.com/s/guides?channel=Sponsored%20Products&article=000012578&language=en_US Accessed 2026-11-03.
- [11]Lowe's Launches Lowe's One Roof Media Network, A New Kind of Retail Media Sales and Marketing Service. Lowe's Companies, Inc., 2021-10-13. https://corporate.lowes.com/newsroom/press-releases/lowes-launches-lowes-one-roof-media-network-new-kind-retail-media-sales-and-marketing-service-10-13-21 Accessed 2026-11-03.
- [12]Lowe's Enhances Retail Media Network With Rebrand And Expanded Portfolio Of Omnichannel Advertising Solutions. Lowe's Companies, Inc., 2024-08-08. https://corporate.lowes.com/newsroom/press-releases/lowes-enhances-retail-media-network-rebrand-and-expanded-portfolio-omnichannel-advertising-solutions-08-08-24 Accessed 2026-11-03.
- [13]Kohl's Expands Advertising Opportunities with Kohl's Media Network. Kohl's Corporation, 2022-07-21. https://corporate.kohls.com/news/kohls-expands-advertising-opportunities-with-kohls-media-network Accessed 2026-11-03.
- [14]Albertsons Media Collective Doubles Down on Performance – Announcing Inaugural Alby Awards Honoring Top-Performing Advertising Partners (release). Albertsons Media Collective (Business Wire), 2025-12-11. https://www.albertsonsmediacollective.com/pr-details/2025/Albertsons-Media-Collective-Doubles-Down-on-Performance--Announcing-Inaugural-Alby-Awards-Honoring-Top-Performing-Advertising-Partners/default.aspx Accessed 2026-11-03.
- [15]Data use and privacy requirements (Walmart Connect Ad Policies, Advertising Help, article 000001103). Walmart Connect (Walmart Inc.), 2026-06-23 ('Last updated June 23, 2026'). https://advertisinghelp.walmart.com/s/guides?channel=Sponsored+Products&article=000001103&language=en_US Accessed 2026-11-03.
- [16]View Managed Service Campaign performance (84.51° Prism Learning Center). 84.51° LLC, 2026-01-28. https://prism.8451.com/activation/help/content/how-to/kpm-managed-service/view-performance.htm Accessed 2026-11-03.
- [17]Amazon Brand Lift: Study the impact of your advertising. Amazon Ads, undated. https://advertising.amazon.com/en-us/measurement-analytics/brand-lift Accessed 2026-11-03.
- [18]Walmart Ad Mix Modeling (Walmart Connect Advertising Help, article 000001173). Walmart Connect (Walmart Inc.), undated on the page. https://advertisinghelp.walmart.com/s/guides?channel=Sponsored+Products&article=000001173&language=en_US Accessed 2026-11-03.
- [19]Search and Compare - Terms and Conditions (Klarna partner documentation). Klarna (docs.klarna.com), 2025-07-23 (event 2025-09-01). https://docs.klarna.com/acquirer/paytrail/resources/legal-and-compliance/more-solutions-guidelines/search-and-compare-terms-and-conditions/ Accessed 2026-11-03.
- [20]Complaint for Permanent Injunction, Monetary Judgment, Civil Penalty Judgment, and Other Relief, FTC et al. v. Amazon.com, Inc., Case 2:26-cv-03097 (W.D. Wash.). Federal Trade Commission and 22 plaintiff states, 2026-08-31. https://www.ftc.gov/system/files/ftc_gov/pdf/AmazonAds-Complaint.pdf Accessed 2026-11-03.
- [21]Description of methodology for Sponsored Search (Walmart Connect Advertising Help, article 000001190). Walmart Connect (Walmart Inc.), 2026-09-09 ('Last updated September 9, 2026'). https://advertisinghelp.walmart.com/s/guides?channel=Sponsored+Products&article=000001190&language=en_US Accessed 2026-11-03. Access: partial extract.
- [22]Kroger Ad Platform Learning Center. 84.51° LLC. https://mp-help.8451.com/mp-help/content/learning-center.htm Accessed 2026-11-03. Access: partial extract.
- [23]Amazon Marketing Cloud data sources (overview). Amazon Ads (Advanced Tools Center). https://advertising.amazon.com/API/docs/en-us/guides/amazon-marketing-cloud/datasources/overview Accessed 2026-11-03.
- [24]Sponsored ads traffic (AMC data source). Amazon Ads (Advanced Tools Center). https://advertising.amazon.com/API/docs/en-us/guides/amazon-marketing-cloud/datasources/sponsored_ads_traffic Accessed 2026-11-03.
- [25]Self-Service Third Party Measurement Studies. Amazon Ads, 2026-05-15. https://advertising.amazon.com/en-us/resources/whats-new/self-service-partner-measurement-studies Accessed 2026-11-03.
- [26]The Kroger Ad Platform. Kroger Precision Marketing. https://www.krogerprecisionmarketing.com/client-resources/kroger-ad-platform/ Accessed 2026-11-03.
- [27]FAQs. Kroger Precision Marketing. https://www.krogerprecisionmarketing.com/faqs/ Accessed 2026-11-03.
- [28]Marketing Mix Modeling Data Portal (Walmart Connect Advertising Help, article 000001167). Walmart Connect (Walmart Inc.), undated on the page. https://advertisinghelp.walmart.com/s/guides?channel=Sponsored+Products&article=000001167&language=en_US Accessed 2026-11-03.
- [29]Amazon Marketing Cloud: Discover advertising insights. Amazon Ads. https://advertising.amazon.com/en-us/solutions/products/amazon-marketing-cloud Accessed 2026-11-03.
- [30]What We Offer - Roundel Media Designed by Target (solutions page). Roundel / Target Brands, Inc.. https://roundel.com/solutions/ Accessed 2026-11-03.
- [31]Measurement - Sam's Club Connect. Sam's Club / Walmart, Inc.. https://www.samsclubconnect.com/measurement Accessed 2026-11-03.
- [32]Reporting Overview (Uber Ads API, Uber Developers). Uber Technologies Inc.. https://developer.uber.com/docs/ads/reporting-metrics/overview Accessed 2026-11-03.
- [33]Offline Sales Reporting (Commerce Yield Help Center). Criteo, 2025-12-17. https://help.cyield.criteo.com/kb/guide/en/offline-sales-reporting-uTS9VpEP3H/Steps/3830483 Accessed 2026-11-03.
- [34]Privacy Policy | Perpetua. Flywheel Digital Limited, 2025-08-06. https://perpetua.io/privacy/ Accessed 2026-11-03.
- [35]Developers | Incremental (API). Incremental. https://www.incremental.com/developers Accessed 2026-11-03.
- [36]DoubleVerify Reports Third Quarter 2024 Financial Results (Form 8-K exhibit 99.1). DoubleVerify Holdings, Inc. / SEC EDGAR, 2024-11-06. https://www.sec.gov/Archives/edgar/data/1819928/000155837024014624/dv-20241106xex99d1.htm Accessed 2026-11-03.
- [37]Author’s work · The Open Web Isn't Dead. It's Uninsured.. Evgeny Popov (No Fluff Advisory); the author of this paper, 2026-08-22. https://nofluffadvisory.com/writing/the-open-web-isnt-dead-its-uninsured/ Accessed 2026-11-03.
- [38]Advertise on Gopuff — Gopuff Ads. GoBrands, Inc. d/b/a Gopuff, 2026-10-09. https://www.gopuff.com/e/ads Accessed 2026-11-03.
- [39]Staples Media Network. Staples, Inc., 2026. https://www.staples.com/sbd/cre/SMN/index.html Accessed 2026-11-03.
- [40]Grow with advertising — Wayfair supplier site. Wayfair LLC, 2026. https://sell.wayfair.com/grow-advertising Accessed 2026-11-03.
- [41]Kohl's Media Network — What we offer. Kohl's Corporation. https://advertising.kohls.com/what-we-offer.html Accessed 2026-11-03.
- [42]adretailmedia.com home page. AD Retail Media (Ahold Delhaize USA). https://www.adretailmedia.com/ Accessed 2026-11-03. Access: partial extract.
- [43]Ahold Delhaize USA Launches Edge, a New Proprietary Retail Media Ad Platform (release). Ahold Delhaize USA, 2026-01-07. https://www.adusa.com/news-releases/news-release-details/ahold-delhaize-usa-launches-edge-new-proprietary-retail-media-ad Accessed 2026-11-03.
- [44]Offsite media (solutions page). Walmart Connect / Walmart Inc.. https://www.walmartconnect.com/solutions/offsite-media Accessed 2026-11-03.
- [45]Store ads (solutions page). Walmart Connect / Walmart Inc.. https://www.walmartconnect.com/solutions/store-ads Accessed 2026-11-03.
- [46]IAB Europe In-Store Retail Media Definitions and Measurement Standards, Version 2 (Draft For Industry Review - September 2026, PDF). IAB Europe, 2026-09 (event 2026-09-17). https://iabeurope.eu/wp-content/uploads/IAB-Europe_InStore_Retail_Media_Definitions_and_Measurement_Standards_V2-PUBLIC-COMMENT-SEPT-2026.pdf Accessed 2026-11-03.
- [47]IAB Europe and IAB In-Store Retail Media Definitions and Measurement Standards (final, December 2024, PDF). IAB Europe and Interactive Advertising Bureau, 2024-12 (event 2024-12-03). https://www.iab.com/wp-content/uploads/2024/12/IAB-Europe_IAB-In-Store-Retail-Media-Standards_FINAL-DECEMBER-2024.pdf Accessed 2026-11-03.
- [48]Accredited Services - Digital (MRC accreditation list). Media Rating Council. https://mediaratingcouncil.org/accreditation/digital Accessed 2026-11-03.
- [49]Services Under Review - Digital (MRC). Media Rating Council. https://mediaratingcouncil.org/accreditation/services-under-review/digital Accessed 2026-11-03. Access: partial extract.
- [50]Retail Media Certification Programme - Certified (IAB Europe list of certified companies). IAB Europe. https://iabeurope.eu/retail-media-certification-programme/certified/ Accessed 2026-11-03.
- [51]AAM Assurance List filtered to 'Incrementality Certification' (session read). Alliance for Audited Media. https://auditedmedia.com/about/clients?industry_certifications=Incrementality+Certification Accessed 2026-11-03.
- [52]Incrementality Certification for Commerce Media (programme page). Alliance for Audited Media. https://www.auditedmedia.com/industry-certifications/retail-media Accessed 2026-11-03. Access: partial extract.
- [53]IAB Europe Advances In-Store & Travel Media Measurement with New Proposals Open for Public Comment (IAB Europe release, PressBox copy on ExchangeWire). IAB Europe via ExchangeWire PressBox, 2026-09-17. https://www.exchangewire.com/blog/2026/09/17/iab-europe-advances-in-store-travel-media-measurement-with-new-proposals-open-for-public-comment/ Accessed 2026-11-03.
- [54]With Retail Media Experiencing Meteoric Growth, ANA Survey Finds Marketers Are Optimistic But Cautious (release). Association of National Advertisers, 2024-07-23. https://ana.net/content/show/id/pr-2024-07-rmn Accessed 2026-11-03.
- [55]MRC accreditation letter: Criteo (GI031524). Media Rating Council, 2024-03-15. https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/Criteo/GI031524_Criteo%20Accreditation%20Letter.pdf Accessed 2026-11-03.
- [56]MRC accreditation letter: Walmart Connect Sponsored Search (GI092925). Media Rating Council, 2025-09-26. https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/Walmart/GI092925_Walmart%20Sponsored%20Search%20Accreditation%20Letter%5B1%5D.pdf Accessed 2026-11-03.
- [57]MRC Accreditation Status Quarterly Update, Q4 2025 (October-December 2025). Media Rating Council, 2026-01-22. https://mediaratingcouncil.org/sites/default/files/News/MRC%20Accreditation%20Status%20Quarterly%20Update%20Q4%202025.pdf Accessed 2026-11-03.
- [58]MRC Accreditation Status Quarterly Update, Q1 2026 (January-March 2026). Media Rating Council, 2026-04-17. https://mediaratingcouncil.org/sites/default/files/News/MRC%20Accreditation%20Status%20Quarterly%20Update_Q1%202026.pdf Accessed 2026-11-03.
- [59]MRC accreditation letter: Amazon DSP (GI052926). Media Rating Council, 2026-05-29. https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/Amazon/GI052926_Amazon%20DSP%20Accreditation%20Letter.pdf Accessed 2026-11-03.
- [60]MRC Accreditation Status Quarterly Update, Q2 2026 (April-June 2026). Media Rating Council, 2026-07-13. https://mediaratingcouncil.org/sites/default/files/News/MRC%20Accreditation%20Status%20Quarterly%20Update_Q2%2026%207.14.26.pdf Accessed 2026-11-03.
- [61]MRC accreditation letter: Instacart (GI060126). Media Rating Council, 2026-06-01. https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/Instacart/GI060126_Instacart%20Accreditation%20Letter.pdf Accessed 2026-11-03.
- [62]MRC accreditation letter: Amazon Sponsored Products (GI070826), with TAG Certified Against Fraud addendum. Media Rating Council, 2026-07-08. https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/Amazon/GI070826_Amazon%20Sponsored%20Products%20Accreditation%20Letter.pdf Accessed 2026-11-03.
- [63]Ghost Ads: Improving the Economics of Measuring Ad Effectiveness (working paper, NBER conference copy). Garrett A. Johnson (Simon Business School, University of Rochester), Randall A. Lewis (Netflix), Elmar I. Nubbemeyer (Google), 2016-02-18. https://conference.nber.org/conf_papers/f84173.pdf Accessed 2026-11-03.
- [64]Amazon Ads Multi-Touch Attribution (arXiv PDF v1). Lewis, Zettelmeyer, Gordon, Garib, Hermle, Perry, Romero, Schnaidt; affiliations '1Amazon 2Northwestern University 3NBER 4Work done at Amazon', 2025-08-11. https://arxiv.org/pdf/2508.08209 Accessed 2026-11-03.
- [65]Predictive Incrementality by Experimentation (PIE) for Ad Measurement (arXiv v1; abstract page lists v2 of 1 April 2026). Brett R. Gordon (Kellogg), Robert Moakler (Ads Research, Meta), Florian Zettelmeyer (Kellogg and NBER), 2023-04-13. https://arxiv.org/pdf/2304.06828v1 Accessed 2026-11-03.
- [66]Close Enough? A Large-Scale Exploration of Non-Experimental Approaches to Advertising Measurement (arXiv v2 PDF). Brett R. Gordon (Kellogg), Robert Moakler (Ads Research, Meta), Florian Zettelmeyer (Kellogg and NBER), 2022-10-04 (event 2019-11 to 2020-03 (experiments)). https://arxiv.org/pdf/2201.07055 Accessed 2026-11-03.
- [67]Close Enough? arXiv abstract page (version history). Gordon, Moakler, Zettelmeyer, 2022-10-04. https://arxiv.org/abs/2201.07055 Accessed 2026-11-03.
- [68]New tricks boosted sales for premium pet food brand (Pet Case Study). Roundel (Target), undated (campaign April-June 2021). https://roundel.com/case-studies/pets/ Accessed 2026-11-03.
- [69]LensDirect sets its sights on customer acquisition (LensDirect | Chase Media Solutions). Chase Media Solutions, 2025-07-07 (event campaign start 2024-09-15 (90 days)). https://www.chase.com/mediasolutions/resources/lensdirect Accessed 2026-11-03.
- [70]ANA: Retail media networks – a forced marriage or perfect partnership?. WARC (source: ANA), 2023-02-15. https://www.warc.com/content/paywall/article/ana-retail-media-networks--a-forced-marriage-or-perfect-partnership/en-gb/en-GB/149516 Accessed 2026-11-03. Access: partial extract.
- [71]About Attribution (Commerce Yield Help Center). Criteo, 2025-12-17. https://help.cyield.criteo.com/kb/guide/en/about-attribution-nYCDWgn3qL/Steps/3211666 Accessed 2026-11-03.
- [72]Ingest orders from a third-party system (Carrot Ads API). Instacart (Maplebear Inc.), 2026-05-14. https://docs.instacart.com/ads/api/ads/ingest_orders Accessed 2026-11-03.
- [73]Events (Carrot Ads API docs). Instacart (Maplebear Inc.), 2026-09-17. https://docs.instacart.com/ads/ads_guide/concepts/ad_events Accessed 2026-11-03.
- [74]Changelog (Carrot Ads API docs). Instacart (Maplebear Inc.), 2026-10-09. https://docs.instacart.com/ads/api/ads/changelog Accessed 2026-11-03.
- [75]Campaign Attribution Settings (Commerce Yield Help Center). Criteo, 2025-12-05 (event 2024-09-05). https://help.cyield.criteo.com/kb/guide/en/campaign-attribution-settings-eRuTQ7Ujg7/Steps/3204863 Accessed 2026-11-03.
- [76]Attribution (Koddi developer docs). Koddi Inc., 2026-01-05. https://developers.koddi.com/docs/attribution Accessed 2026-11-03.
- [77]Kevel Powers Dollar General Retail Media Network. Kevel, 2025-05-20. https://www.kevel.com/blog/kevel-dollar-general-press-release Accessed 2026-11-03.
- [78]Orange Apron Media Announces New Comprehensive Self-Service Ad Platform - Orange Access. The Home Depot, Inc. (investor relations), 2024-10-03. https://ir.homedepot.com/news-releases/2024/10-03-2024-150051692 Accessed 2026-11-03.
- [79]Moloco Commerce Media: Better discovery, better performance, and more revenue for retailers. Moloco, Inc.. https://www.moloco.com/moloco-commerce-media Accessed 2026-11-03.
- [80]Topsort: Welcome to Commerce AI. Topsort Technologies, Inc.. https://www.topsort.com/ Accessed 2026-11-03.
- [81]Circana and Ibotta Announce Strategic Partnership to Bring Sales Lift Measurement to Digital Promotions (Ibotta investor-relations copy). Circana, LLC and Ibotta, Inc. (joint release, distributed via Business Wire), 2025-09-30. https://investors.ibotta.com/news-events/press-releases/detail/127/circana-and-ibotta-announce-strategic-partnership-to-bring-sales-lift-measurement-to-digital-promotions Accessed 2026-11-03.
- [82]Sales Lift | Nielsen. The Nielsen Company (US), LLC. https://www.nielsen.com/solutions/marketing-optimization/sales-lift/ Accessed 2026-11-03.
- [83]In-Flight Measurement (Attain measurement page). Attain, undated (copyright 2026). https://www.attainoutcomes.com/measurement Accessed 2026-11-03.
- [84]In-Flight Measurement | Catalina (HUB360). Catalina. https://www.catalina.com/solutions/measurement-and-optimization/in-flight-measurement Accessed 2026-11-03.
- [85]How It Works | Incremental. Incremental. https://www.incremental.com/how-it-works Accessed 2026-11-03.
- [86]Measurement and Optimization | Catalina. Catalina. https://www.catalina.com/solutions/measurement-and-optimization Accessed 2026-11-03.
- [87]Instacart and Gopuff Announce New, First-of-Its-Kind Partnership in the U.S.. GoBrands, Inc. d/b/a Gopuff (newsroom); joint with Instacart, 2026-09-22. https://www.gopuff.com/newsroom/company-news/instacart-and-gopuff-announce-new-first-of-its-kind-partnership-in-the-u.s. Accessed 2026-11-03.
- [88]Code of Virginia § 59.1-578. Data controller responsibilities; transparency. Virginia General Assembly (Legislative Information System). https://law.lis.virginia.gov/vacode/title59.1/chapter53/section59.1-578/ Accessed 2026-11-03. Access: partial extract.
- [89]Chapter 743jj, Data Privacy (2026 Supplement to the General Statutes). Connecticut General Assembly, 2026. https://www.cga.ct.gov/2026/sup/chap_743jj.htm Accessed 2026-11-03. Access: partial extract.
- [90]Connecticut General Statutes, Chapter 743jj (Data Privacy). Connecticut General Assembly. https://www.cga.ct.gov/current/pub/chap_743jj.htm Accessed 2026-11-03. Access: partial extract.
- [91]Privacy Law FAQs for Businesses. Oregon Department of Justice, Consumer Protection. https://www.doj.state.or.us/consumer-protection/for-businesses/privacy-law-faqs-for-businesses/ Accessed 2026-11-03.
- [92]Attorney General James Demands Answers from Instacart about Algorithmic Pricing. New York State Office of the Attorney General, 2026-01-08. https://ag.ny.gov/press-release/2026/attorney-general-james-demands-answers-instacart-about-algorithmic-pricing Accessed 2026-11-03.
- [93]On Data Privacy Day, Attorney General Bonta Focuses on Surveillance Pricing Compliance. California Department of Justice, Office of the Attorney General, 2026-01-27. https://oag.ca.gov/news/press-releases/data-privacy-day-attorney-general-bonta-focuses-surveillance-pricing-compliance Accessed 2026-11-03.
- [94]Stipulated Order for Permanent Injunction, Monetary Judgment, and Other Relief, FTC v. Maplebear Inc. d/b/a Instacart, No. 4:25-cv-10783-JST (N.D. Cal.), Document 15. U.S. District Court, N.D. Cal. (Judge Jon S. Tigar); posted by the FTC, 2026-01-13. https://www.ftc.gov/system/files/ftc_gov/pdf/Instacart-StipulatedOrder_0.pdf Accessed 2026-11-03.
- [95]Instacart to Pay $60 Million in Consumer Refunds to Settle FTC Lawsuit Over Allegations it Engaged in Deceptive Tactics. Federal Trade Commission, 2025-12-18. https://www.ftc.gov/news-events/news/press-releases/2025/12/instacart-pay-60-million-consumer-refunds-settle-ftc-lawsuit-over-allegations-it-engaged-deceptive Accessed 2026-11-03.
- [96]Instacart (FTC case page). Federal Trade Commission, 2026-01-14 (event 2025-12-18). https://www.ftc.gov/legal-library/browse/cases-proceedings/instacart Accessed 2026-11-03.
- [97]Data Broker Registry — California Privacy Protection Agency (page and downloadable registration files). California Privacy Protection Agency, 2026-10-30. https://cppa.ca.gov/data_broker_registry/ Accessed 2026-11-03.
- [98]Author’s work · Measurement for Agentic Commerce. Evgeny Popov (No Fluff Advisory); the author of this paper, 2026-07-13. https://nofluffadvisory.com/writing/measurement-for-agentic-commerce/ Accessed 2026-11-03.
- [99]Product Feeds (OpenAI Ads developer documentation). OpenAI. https://developers.openai.com/ads/product-feeds Accessed 2026-11-03.
- [100]Ads in AI Mode expand to 10 new countries. Shashi Thakur, VP and GM, Google Search Ads and Ads on Google Experiences (Google), 2026-10-28 (event 2026-10-28 (expansion to ten countries; Brazil, Mexico and the UK planned for Q4 2026)). https://blog.google/products/ads-commerce/ads-in-ai-mode-global/ Accessed 2026-11-03.
- [101]UCP tag v2026-08-25 (annotated tag object) and commit f2206e68. Universal-Commerce-Protocol (GitHub API), 2026-08-25. https://github.com/Universal-Commerce-Protocol/ucp/releases/tag/v2026-08-25 Accessed 2026-11-03.
- [102]Overview of Brand Center Reporting (Ads Support, DoorDash Help Center). DoorDash, Inc.. https://help.doordash.com/en-us/advertising/article/overview-ads-manager-reporting Accessed 2026-11-03.
- [103]Retail Media Certification Programme (IAB Europe page). IAB Europe. https://iabeurope.eu/retail-media-certification-programme/ Accessed 2026-11-03.
- [104]IAB Europe Retail Media Certification Compliance Table for CMNs (including RMNs, CMNs & Ad Tech), Update May 2026, Compliance Standards v2.1 (PDF). IAB Europe, 2026-05. https://iabeurope.eu/wp-content/uploads/IAB-Europe-Retail-Media-Certification-Compliance-Table-for-Retailers-update-May-2026.pdf Accessed 2026-11-03.
- [105]FTC Launches Personalized Pricing and Compensation Inquiry. Federal Trade Commission, 2026-10-28. https://www.ftc.gov/news-events/news/press-releases/2026/10/ftc-launches-personalized-pricing-compensation-inquiry Accessed 2026-11-03.
- [106]Cardlytics, Inc. Form 10-K for the fiscal year ended 31 December 2025 (cdlx-20251231.htm; accession 0001666071-26-000010). Cardlytics, Inc., 2026. https://www.sec.gov/Archives/edgar/data/1666071/000166607126000010/cdlx-20251231.htm Accessed 2026-11-03. Access: partial extract.
- [107]Introducing New Sales Lift Metrics and A/B Testing for CPG Brands. Instacart, 2022-09-02. https://www.instacart.com/company/instacart-ads/introducing-new-sales-lift-metrics Accessed 2026-11-03.
- [108]Uncover ad value across the funnel with multi-touch attribution (beta). Amazon Ads, undated ('Announced at unBoxed 2024') (event 2024 (unBoxed 2024)). https://advertising.amazon.com/en-us/resources/whats-new/uncover-ad-value-across-funnel-with-multi-touch-attribution Accessed 2026-11-03.
- [109]Sponsored Products (product page, en-us). Amazon Ads, undated (footnote 'Amazon internal data, US, June 2026'). https://advertising.amazon.com/en-us/solutions/products/sponsored-products Accessed 2026-11-03.
- [110]Sponsored Products (product page). Amazon Ads. https://advertising.amazon.com/solutions/products/sponsored-products Accessed 2026-11-03.
- [111]Sponsored Brands (product page). Amazon Ads. https://advertising.amazon.com/solutions/products/sponsored-brands Accessed 2026-11-03.
- [112]View attribution updates for Amazon Store ads (Visualizza gli aggiornamenti sull'attribuzione per gli annunci Amazon). Amazon Ads, 2026-01-01. https://advertising.amazon.com/resources/whats-new/view-attribution-updates-for-amazon-store-ads/ Accessed 2026-11-03. Access: partial extract.
- [113]Amazon DSP is now part of Amazon Ads Agent (product page). Amazon Ads. https://advertising.amazon.com/solutions/products/amazon-dsp Accessed 2026-11-03.
- [114]Amazon Ads Agent announcement (UnBoxed 2026 news page). Amazon Ads, 2026-09-29. https://advertising.amazon.com/library/news/amazon-ads-agent Accessed 2026-11-03. Access: partial extract.
- [115]Sponsored Search ads with Walmart Connect (solutions page). Walmart Connect / Walmart Inc.. https://www.walmartconnect.com/solutions/sponsored-search Accessed 2026-11-03.
- [116]Walmart Inc. Form 8-K exhibit 99.1, Q4 and full-year fiscal 2026 earnings release (accession 0000104169-26-000032). Walmart Inc. / SEC EDGAR, 2026-02-19 (event 2026-01-31). https://www.sec.gov/Archives/edgar/data/104169/000010416926000032/earningsreleasefy26q4.htm Accessed 2026-11-03.
- [117]Omnichannel Ad Measurement & Closed-Loop Insights | Walmart Connect. Walmart Connect, Walmart Inc., undated (footnotes cite March 14, 2025 and February 2025; footer 2026). https://www.walmartconnect.com/insights/measurement Accessed 2026-11-03.
- [118]Onsite Display (solutions page). Walmart Connect / Walmart Inc.. https://www.walmartconnect.com/solutions/onsite-display Accessed 2026-11-03.
- [119]Non-endemic advertising (page). Walmart Connect / Walmart Inc.. https://www.walmartconnect.com/non-endemic-advertising Accessed 2026-11-03.
- [120]Walmart and VIZIO Scale Content to Commerce at NewFronts (release). Walmart Inc., 2026-03-23. https://corporate.walmart.com/news/2026/03/23/walmart-and-vizio-scale-content-to-commerce-at-newfronts Accessed 2026-11-03.
- [121]Walmart Sets the Vision for the Next Era of Global Commerce Media (release). Walmart Inc., 2026-06-22. https://corporate.walmart.com/news/2026/06/22/walmart-sets-the-vision-for-the-next-era-of-global-commerce-media Accessed 2026-11-03.
- [122]Instacart Ads products (grocery retail media) page. Instacart (Maplebear Inc.). https://company.instacart.com/ads/products-grocery-retail-media Accessed 2026-11-03. Access: partial extract.
- [123]Instacart Ads measurement and insights page. Instacart (Maplebear Inc.). https://company.instacart.com/ads/measurement-insights Accessed 2026-11-03. Access: partial extract.
- [124]Roundel fact sheet. Target Corporation (corporate.target.com), 2026-06-18. https://corporate.target.com/press/fact-sheet/2023/10/roundel Accessed 2026-11-03.
- [125]Measurement - Roundel Media Designed by Target. Roundel (Target Brands, Inc.), undated (footnote '2026 Target Internal Data'; copyright 2026). https://roundel.com/measurement/ Accessed 2026-11-03.
- [126]Albertsons Media Collective Introduces Onsite Incrementality Measurement to Help Advertisers Better Understand Retail Media Impact. Albertsons Companies, Inc. newsroom (Business Wire), 2026-04-16. https://www.albertsonscompanies.com/newsroom/press-releases/news-details/2026/Albertsons-Media-Collective-Introduces-Onsite-Incrementality-Measurement-to-Help-Advertisers-Better-Understand-Retail-Media-Impact/default.aspx Accessed 2026-11-03.
- [127]Albertsons Media Collective - RESULTS YOU CAN COUNT ON. (results page). Albertsons Media Collective (Albertsons Companies, Inc.), undated ('in 2025' figures). https://www.albertsonsmediacollective.com/results/ Accessed 2026-11-03.
- [128]Albertsons Media Collective Launches Incrementality Measurement to Prove True Media Impact of In-Store Campaigns and Plans Expansion. Albertsons Media Collective (Business Wire), 2026-01-06 (Business Wire id 20260106147047; the Albertsons Companies newsroom copy is dated 6 January 2026 per the brief; no date visible on this copy) (event 2026-01-06). https://www.albertsonsmediacollective.com/pr-details/2026/Albertsons-Media-Collective-Launches-Incrementality-Measurement-to-Prove-True-Media-Impact-of-In-Store-Campaigns-and-Plans-Expansion/default.aspx Accessed 2026-11-03.
- [129]DoorDash Ads Launches Sales Lift Measurement, Driving Up to 30% Incrementality Lift in Campaigns on Average for Brands. DoorDash, Inc. (DoorDash Ads), 2026-03-24 (event 2026-03-24 (launch announcement)). https://about.doordash.com/en-us/news/doordash-ads-launches-sales-lift-measurement Accessed 2026-11-03.
- [130]DoorDash Ads home page (en-us). DoorDash Ads / DoorDash, Inc.. https://advertising.doordash.com/en-us Accessed 2026-11-03.
- [131]DoorDash Earns AAM Incrementality Certification and Introduces New Ways to Measure Commerce Media. DoorDash, Inc. (DoorDash Ads), 2026-09-02 (event 2026-09-02 (announced at SHOWCASE, New York)). https://about.doordash.com/en-us/news/doordash-earns-aam-incrementality-certification Accessed 2026-11-03.
- [132]Measuring Restaurant Ads Incrementality with Ghost Ads (title tag: How DoorDash Measures Restaurant Ad Incrementality with Ghost Ads | DoorDash Ads). DoorDash Ads (no author named), 2025-10-04. https://advertising.doordash.com/en-us/resources/measuring-incrementality-with-ghost-ads-at-doordash Accessed 2026-11-03.
- [133]Get started with ads (Uber Eats merchants, US). Uber Eats / Uber Technologies Inc.. https://merchants.ubereats.com/us/en/c/get-started-with-ads/ Accessed 2026-11-03.
- [134]Uber Advertising press page. Uber Advertising / Uber Technologies Inc.. https://www.uber.com/us/en/advertising/press/ Accessed 2026-11-03. Access: partial extract.
- [135]Uber Advertising landing page (US). Uber Advertising / Uber Technologies Inc.. https://www.uber.com/us/en/advertising/ Accessed 2026-11-03.
- [136]Chewy Ads — home. Chewy, Inc. (Chewy Ads), 2026. https://discover.ads.chewy.com/ Accessed 2026-11-03.
- [137]Chewy Ads — Solutions. Chewy, Inc. (Chewy Ads), 2026. https://discover.ads.chewy.com/solutions Accessed 2026-11-03.
- [138]How we're making transparency our specialty. Best Buy Ads (Best Buy). https://www.bestbuyads.com/insights/how-we-re-making-transparency-our-specialty Accessed 2026-11-03.
- [139]Best Buy Ads — Home. Best Buy Ads, 2026. https://www.bestbuyads.com/Home Accessed 2026-11-03.
- [140]Inside The Drop: Best Buy unveils the future of commerce, media and technology. Best Buy (corporate newsroom), 2026-09-28 (event 2026-09-28 ('this year's Showcase')). https://corporate.bestbuy.com/2026/inside-the-drop-best-buy-unveils-the-future-of-commerce-media-and-technology/ Accessed 2026-11-03.
- [141]Measurement — Orange Apron Media. Orange Apron Media (The Home Depot), 2026. https://orangeapronmedia.com/solutions/measurement/ Accessed 2026-11-03.
- [142]Orange Apron Media Enhances Measurement Ecosystem to Prove Total Business Impact. Orange Apron Media (Ross Love, Director of Measurement), 2026-04-22 (date from the blog listing RTL-E2-S13; the post shows none). https://orangeapronmedia.com/oam-measurement-enhance/ Accessed 2026-11-03.
- [143]CMX Performance Dashboard. CVS Media Exchange (CVS Health). https://www.cvsmediaexchange.com/solutions/cmx-performance-dashboard Accessed 2026-11-03.
- [144]CVS Media Exchange Ads Policy. CVS Media Exchange. https://www.cvsmediaexchange.com/advertising-policy.html Accessed 2026-11-03.
- [145]In-Store Solutions | Walgreens Advertising Group. Walgreens Advertising Group (Walgreen Co.), 2024. https://www.walgreensadvertisinggroup.com/home/solution-detail.html Accessed 2026-11-03.
- [146]About — Walgreens Advertising Group. Walgreens Advertising Group (Walgreen Co.), 2024. https://www.walgreensadvertisinggroup.com/home/about.html Accessed 2026-11-03.
- [147]Offsite Solutions | Walgreens Advertising Group. Walgreens Advertising Group (Walgreen Co.), 2024. https://www.walgreensadvertisinggroup.com/home/solution-detail11.html Accessed 2026-11-03.
- [148]Macy's Media Network — News (Amazon Retail Ad Service pilot). Advertex Communications Inc. d/b/a Macy's Media Network. https://www.macysmedianetwork.com/news Accessed 2026-11-03.
- [149]Macy's Media Network – Advertise with Macy's and Bloomingdales (index). Macy's Media Network, 2022 (footer). https://www.macysmedianetwork.com/index.html Accessed 2026-11-03.
- [150]Macy's Media Network — Self-Serve Ad Terms (Terms and Conditions). Advertex Communications Inc. d/b/a Macy's Media Network. https://www.macysmedianetwork.com/termsandconditions.html Accessed 2026-11-03.
- [151]UB Media — Ulta Beauty. Ulta Beauty, Inc. (UB Media), 2026. https://www.ulta.com/company/ubmedia Accessed 2026-11-03.
- [152]UB Media — Outcomes. Ulta Beauty, Inc., 2026. https://www.ulta.com/company/ubmedia/outcomes Accessed 2026-11-03.
- [153]Costco.com Ad Specs. Costco Wholesale Corporation. https://mobilecontent.costco.com/live/resource/img/costco-ad-specs/Costco-Ad-Specs-2021.html Accessed 2026-11-03.
- [154]DG Media Network to Introduce AI-Enabled In-Store Audio Network Across Thousands of Dollar General Stores. Dollar General Corporation (investor relations), 2026-04-13. https://investor.dollargeneral.com/news-detail/dg-media-network-to-introduce-ai-enabled-in-store-audio-network-across-thousands-of-dollar-general-stores/b6876c01-c92d-49ae-807b-03a0cdc5d861 Accessed 2026-11-03.
- [155]Dollar General Introduces Evolution of Retail Media Network, DGMN. Dollar General Corporation (investor relations), 2022-06-07. https://investor.dollargeneral.com/news-detail/dollar-general-introduces-evolution-of-retail-media-network-dgmn/08de980f-5896-435e-80f6-69417f37ffe5 Accessed 2026-11-03.
- [156]Seller Announcement: Updates to eBay's Promoted Listings are coming soon. eBay Inc. (export.ebay.com) (event 2026-01-13). https://export.ebay.com/en/services-tools/advertising/seller-announcement-updates-to-ebays-promoted-listings-are-coming-soon/ Accessed 2026-11-03. Access: partial extract.
- [157]eBay Advertising attribution (Promoted Listings general, priority, Promoted Stores, Promoted Offsite). eBay Inc.. https://pages.ebay.com/sell/pl/attribution.html Accessed 2026-11-03.
- [158]How to Review the Performance of Your Etsy Ads. Etsy, Inc. (Help Center). https://help.etsy.com/hc/en-us/articles/360034223613-How-to-Review-the-Performance-of-Your-Etsy-Ads Accessed 2026-11-03.
- [159]How Etsy's Offsite Ads Work. Etsy, Inc. (Help Center). https://help.etsy.com/hc/en-us/articles/360000338367-How-Etsy-s-Offsite-Ads-Work Accessed 2026-11-03.
- [160]RedMedia Solutions | Hy-Vee. Hy-Vee, Inc. (Hy-Vee RedMedia), 2026. https://www.hredmedia.com/solutions Accessed 2026-11-03.
- [161]Hy-Vee Offers Access to its Diverse Portfolio to Launch New Kind of Retail Media Network. Hy-Vee, Inc., 2023-09-22 (event 2023-09). https://www.hy-vee.com/corporate/news-events/news-press-releases/hyvee-offers-access-to-its-diverse-portfolio-to-launch-new-kind-of-retail-media-network/ Accessed 2026-11-03.
- [162]Sponsored Listings (TravelAds and Flights Sponsored Listings) — Expedia Group Advertising. Expedia, Inc., an Expedia Group company, 2026. https://partner.expediagroup.com/en-us/solutions/advertise-with-us/sponsored-listings Accessed 2026-11-03.
- [163]Expedia Group Accelerates Vacation Rental Growth With Largest-Ever Product Launch For Vrbo And Escapia. Expedia Group, Inc. (investor relations), 2026-09-01. https://ir.expediagroup.com/news-and-events/news/news-details/2026/Expedia-Group-Accelerates-Vacation-Rental-Growth-With-Largest-Ever-Product-Launch-For-Vrbo-And-Escapia/default.aspx Accessed 2026-11-03.
- [164]GULP Media Network — home (measurement section). Gulp Media Network (7-Eleven, Inc.), 2026. https://www.gulpmedianetwork.com/ Accessed 2026-11-03.
- [165]7-Eleven, Inc. Redefines the Shopper Experience with Expansion of Gulp Radio: One of North America's Largest Commercial Radio Networks. 7-Eleven, Inc., 2024-11-12. https://corp.7-eleven.com/corp-press-releases/7-eleven-inc-redefines-the-shopper-experience-with-expansion-of-gulp-radio-one-of-north-america-s-largest-commercial-radio-networks Accessed 2026-11-03.
- [166]DICK'S Media — home. DICK'S Media (DICK'S Sporting Goods), 2025. https://www.dicksmedia.com/ Accessed 2026-11-03.
- [167]DICK'S Media — Offsite. DICK'S Media (DICK'S Sporting Goods), 2025. https://www.dicksmedia.com/offsite Accessed 2026-11-03.
- [168]DICK'S Media — Onsite. DICK'S Media (DICK'S Sporting Goods), 2025. https://www.dicksmedia.com/onsite Accessed 2026-11-03.
- [169]Petco RMN — media designed by Petco. Petco Health and Wellness Company, Inc., 2024. https://advertising.petco.com/ Accessed 2026-11-03.
- [170]About attribution for TikTok Shop Ads. TikTok for Business (ads.tiktok.com Help Center), 2025-08. https://ads.tiktok.com/help/article/about-tiktok-shop-ads-attribution?lang=en Accessed 2026-11-03. Access: partial extract.
- [171]About gross revenue for TikTok Shop Ads. TikTok for Business (ads.tiktok.com Help Center), 2025-08. https://ads.tiktok.com/help/article/about-gross-revenue-for-tiktok-shop-ads?lang=en Accessed 2026-11-03.
- [172]Leap with myPerks — Leap Media Group. Giant Eagle, Inc. (Leap Media Group), 2022. https://www.leapmediagroup.com/ Accessed 2026-11-03.
- [173]Lyft Ads | Data-driven marketing on the move. Lyft Ads. https://ads.lyft.com/ Accessed 2026-11-03.
- [174]Lyft Media expands with video, measurement & targeting. Lyft (blog), 2024-03-21. https://www.lyft.com/blog/posts/lyft-media-expands-with-video-measurement-and-targeting Accessed 2026-11-03.
- [175]Lyft Ads | Measuremet (page title as served). Lyft Ads, undated. https://ads.lyft.com/measurement Accessed 2026-11-03.
- [176]Chase Media Solutions (home). JPMorgan Chase Bank, N.A. (Chase Media Solutions), undated (copyright 2026; '2025' figures). https://www.chase.com/mediasolutions/home.html Accessed 2026-11-03.
- [177]Chase Launches Chase Media Solutions, a New Digital Media Business, Connecting 80 Million U.S. Consumers with the Brands They Love. JPMorgan Chase & Co. (media.chase.com), 2024-04-03. https://media.chase.com/news/chase-launches-chase-media-solutions Accessed 2026-11-03.
- [178]Drive incremental growth | Chase Media Solutions. JPMorgan Chase Bank, N.A., 2026. https://www.chase.com/mediasolutions/solutions Accessed 2026-11-03.
- [179]PayPal Advertising Platform | PayPal US. PayPal (PayPal Ads). https://www.paypal.com/us/advertiser Accessed 2026-11-03. Access: partial extract.
- [180]PayPal Ads Launches Transaction Graph Insights and Measurement to Improve Commerce Advertising Outcomes. PayPal (newsroom), 2026-01-06 (event 2026-01-06 (programme launch); Ulta data 'Jun-Jul 2025'). https://newsroom.paypal-corp.com/2026-01-06-PayPal-Ads-Launches-Transaction-Graph-Insights-and-Measurement-to-Improve-Commerce-Advertising-Outcomes Accessed 2026-11-03.
- [181]FAQs (Cardlytics Ads Manager). Cardlytics, Inc. (platform.cardlytics.com), 2026-10-05. https://platform.cardlytics.com/advertisers/docs/ads-manager-faqs Accessed 2026-11-03.
- [182]Cardlytics Measurement | Measurable Marketing Results. Cardlytics, Inc., 2026. https://www.cardlytics.com/marketing-solutions/campaign-measurement Accessed 2026-11-03.
- [183]Cardlytics, Inc. Form 10-K for fiscal 2025 — R5.htm CONSOLIDATED STATEMENTS OF OPERATIONS. Cardlytics, Inc. (SEC EDGAR), 2026. https://www.sec.gov/Archives/edgar/data/1666071/000166607126000010/R5.htm Accessed 2026-11-03.
- [184]Fetch FAQs. Fetch Rewards LLC, 2026. https://fetch.com/faq Accessed 2026-11-03.
- [185]KPMG Validates Fetch Methodology, Praises Platform's "Uniquely Clear View" of Campaign Impact. Fetch (Fetch Rewards LLC), business.fetch.com newsroom, 2023-06-08. https://business.fetch.com/newsroom/kpmg-validates-fetch-methodology Accessed 2026-11-03.
- [186]The Ibotta Performance Network (home). Ibotta, Inc., 2026. https://ipn.ibotta.com/ Accessed 2026-11-03.
- [187]Ibotta, Inc. Form 10-K for the fiscal year ended December 31, 2025 (ibta-20251231.htm; accession 0001628280-26-011838). Ibotta, Inc. (investors.ibotta.com copy of the EDGAR document), 2026-02-26. https://investors.ibotta.com/sec-filings/all-sec-filings/content/0001628280-26-011838/ibta-20251231.htm Accessed 2026-11-03. Access: partial extract.
- [188]Ibotta Properties. Ibotta, Inc., 2026. https://ipn.ibotta.com/properties Accessed 2026-11-03.
- [189]Marriott International Introduces MARRIOTT MEDIA, a New Media Network Shaped for the Traveler. Marriott International, Inc. (via PR Newswire), 2025-06-12. https://www.prnewswire.com/news-releases/marriott-international-introduces-marriott-media-a-new-media-network-shaped-for-the-traveler-302479904.html Accessed 2026-11-03.
- [190]Marriott Media Network (media.marriott.com home, Gatorade partner case). MARRIOTT MEDIA (Marriott International, Inc.), undated. https://media.marriott.com/ Accessed 2026-11-03. Access: partial extract.
- [191]United Launches Airline Industry's First Media Network. United Airlines (via PR Newswire), 2024-06-07. https://www.prnewswire.com/news-releases/united-launches-airline-industrys-first-media-network-302167035.html Accessed 2026-11-03.
- [192]Kinective Media by United Airlines (home page). United Airlines, Inc.. https://www.kinectivemedia.com/ Accessed 2026-11-03.
- [193]Travel Media Advertising for Brands & Ad Agencies | Expedia Group. Expedia Group Advertising (Expedia, Inc., an Expedia Group company), 2026. https://partner.expediagroup.com/en-us/industries/brands-ad-agencies Accessed 2026-11-03.
- [194]Expedia Group Advertising Introduces Travel Audience Marketing. Expedia Group, Inc. (investor relations), 2026-10-07. https://ir.expediagroup.com/news-and-events/news/news-details/2026/Expedia-Group-Advertising-Introduces--Travel-Audience-Marketing/default.aspx Accessed 2026-11-03.
- [195]Expedia Group Advertising and Redion Announce the First Travel Protection Partnership in Landmark Collaboration. Expedia Group, Inc. (investor relations), 2026-09-22. https://ir.expediagroup.com/news-and-events/news/news-details/2026/Expedia-Group-Advertising-and-Redion-Announce-the-First-Travel-Protection-Partnership-in-Landmark-Collaboration/default.aspx Accessed 2026-11-03.
- [196]Amex Offers & Ads for Merchants (American Express merchant digital media page). American Express Company. https://www.americanexpress.com/en-us/merchant/digital-media/ Accessed 2026-11-03.
- [197]Capital One Ad Solutions (capitaloneads.com), browser-pane attempt. Capital One. https://capitaloneads.com/ Accessed 2026-11-03. Access: partial extract.
- [198]Klarna expands growth tools for retailers and creators through global rollout of Creator Shops and Ads Manager. Klarna Bank AB (publ) (via PR Newswire), 2023-11-13. https://www.prnewswire.com/news-releases/klarna-expands-growth-tools-for-retailers-and-creators-through-global-rollout-of-creator-shops-and-ads-manager-301986132.html Accessed 2026-11-03.
- [199]Affirm Holdings, Inc. Form 10-K for the fiscal year ended June 30, 2026 (afrm-20260630.htm; accession 0001628280-26-059279). Affirm Holdings, Inc. (SEC EDGAR), 2026. https://www.sec.gov/Archives/edgar/data/1820953/000162828026059279/afrm-20260630.htm Accessed 2026-11-03. Access: partial extract.
- [200]Local Advertising FAQ | Zillow Group Media Manager. Zillow, Inc.. https://www.zillow.com/brands/self-service-advertising/faq/ Accessed 2026-11-03.
- [201]Acquire Customers While They Shop and Only Pay For Outcomes | Rokt Ads. Rokt, 2026. https://www.rokt.com/products/rokt-ads Accessed 2026-11-03.
- [202]Drive more gas station and c-store traffic | Upside. Upside Services, Inc, 2026. https://www.upside.com/business/fuel-and-convenience/station-owner Accessed 2026-11-03.
- [203]Commerce Yield: The retail media platform where flexibility meets control. Criteo. https://www.criteo.com/platform/commerce-yield/ Accessed 2026-11-03.
- [204]Commerce Grid: Say hello to the world's first commerce SSP. Criteo. https://www.criteo.com/platform/commerce-grid/ Accessed 2026-11-03.
- [205]CitrusAd is now part of Epsilon Retail Media (EMEA product page). Epsilon. https://www.epsilon.com/emea/products-and-services/epsilon-retail-media-citrusad Accessed 2026-11-03.
- [206]Amazon Retail Ad Service - Retail Media Advertising Solution | Amazon Ads (en-us). Amazon Ads. https://advertising.amazon.com/en-us/retail-ad-service/ Accessed 2026-11-03.
- [207]Sponsored products across retailers (beta). Amazon Ads. https://advertising.amazon.com/solutions/products/sponsored-products-across-retailers Accessed 2026-11-03.
- [208]Koddi: AI-powered commerce media built for scale. Koddi Inc.. https://koddi.com/ Accessed 2026-11-03.
- [209]Why Retail Media's Amazon Gap Comes Down to Two Fixable Problems. Sarah Mackinnon, Pentaleap, 2026-07-29. https://www.pentaleap.com/blog/why-retail-medias-amazon-gap-comes-down-to-two-fixable-problems Accessed 2026-11-03.
- [210]Orange Apron Media Announces New Comprehensive Self-Service Ad Platform – Orange Access. The Home Depot (corporate newsroom), 2024-10-03. https://corporate.homedepot.com/news/company/orange-apron-media-announces-new-comprehensive-self-service-ad-platform-orange-access Accessed 2026-11-03.
- [211]Kevel: Your audience. Your rules. Your revenue.. Kevel (Previously Adzerk). https://www.kevel.com/ Accessed 2026-11-03.
- [212]Mirakl Ads: The AI retail media solution built for retailers. Mirakl SAS. https://www.mirakl.com/products/ads/ Accessed 2026-11-03.
- [213]Carrot Ads API overview. Instacart (docs.instacart.com), 2026-09-17. https://docs.instacart.com/ads Accessed 2026-11-03.
- [214]Retail Media Intelligence (Activate). NIQ (Nielsen Consumer LLC). https://nielseniq.com/global/en/products/activate-retail-media-intelligence/ Accessed 2026-11-03.
- [215]Rippl Retail Media Network | Connecting advertisers with hard-to-reach shoppers. Rippl (operator not named on page). https://www.ripplnetwork.com/ Accessed 2026-11-03.
- [216]The Salesforce-Native Enterprise Omnichannel Retail Media Solution. ADvendio. https://www.advendio.com/omnichannel-retail-media-salesforce Accessed 2026-11-03.
- [217]Ibotta, Inc. Form 10-K for fiscal year ended 31 December 2025 (Item 1 Business). Ibotta, Inc. (SEC EDGAR), 2026-02-26. https://www.sec.gov/Archives/edgar/data/1538379/000162828026011838/ibta-20251231.htm Accessed 2026-11-03. Access: partial extract.
- [218]Microsoft looks set to shutter its retail media business. Ronan Shields, Cale Guthrie Weissman and Michael Bürgi, Digiday, 2024-10-07. https://digiday.com/media/microsoft-looks-set-to-shutter-its-retail-media-business/ Accessed 2026-11-03.
- [219]Zitcha: Just ask. (home). Zitcha. https://www.zitcha.com/ Accessed 2026-11-03.
- [220]Osmos: Retail Media Platform for Retailers and Marketplaces. Osmos, Inc.. https://www.osmos.ai/ Accessed 2026-11-03.
- [221]streaem.com -> vml.com/streaem. VML (WPP). https://www.vml.com/streaem Accessed 2026-11-03. Access: partial extract.
- [222]Getting Started with LiveRamp Clean Room for Retailers. LiveRamp, 2026-02-03. https://docs.liveramp.com/connect/en/getting-started-with-liveramp-clean-room-for-retailers.html Accessed 2026-11-03.
- [223]The Trade Desk Helps Brands Activate High-Intent Commerce and Travel Signals Throughout the Consumer Journey. The Trade Desk, Inc., 2026-06-17. https://investors.thetradedesk.com/news-and-events/news/news-details/2026/The-Trade-Desk-Helps-Brands-Activate-High-Intent-Commerce-and-Travel-Signals-Throughout-the-Consumer-Journey/default.aspx Accessed 2026-11-03.
- [224]Commerce | Demand Side Platform | Yahoo Inc.. Yahoo. https://www.yahooinc.com/yahoo-dsp/commerce Accessed 2026-11-03.
- [225]Commerce Media Suite - Accelerate with Google US. Google. https://business.google.com/us/accelerate/announcements/commerce-media-suite/ Accessed 2026-11-03.
- [226]In the News: Path to Purchase Institute Covers STRATACACHE and Albertsons Launch. STRATACACHE, 2026-01-28. https://stratacache.com/en/?p=14962 Accessed 2026-11-03.
- [227]Retail Media | Swiftly. Swiftly Systems, Inc., 2026-11-02. https://swiftly.com/products/retail-media Accessed 2026-11-03.
- [228]About Grocery TV | Leading In-Store Retail Media Platform. Grocery TV. https://grocerytv.com/about Accessed 2026-11-03.
- [229]Retail Media Solutions Overview | Skai (re-read of the page behind RTL-U-S195 for the truncated remainder). Skai (Kenshoo, Ltd.). https://skai.io/retail-media-solutions/ Accessed 2026-11-03. Access: partial extract.
- [230]Company | Skai. Skai (Kenshoo, Ltd.). https://skai.io/company/ Accessed 2026-11-03. Access: partial extract.
- [231]About Us | Pacvue. Pacvue. https://pacvue.com/about/about-us/ Accessed 2026-11-03. Access: partial extract.
- [232]Commerce Max DSP | Criteo. Criteo S.A.. https://www.criteo.com/platform/commerce-max/ Accessed 2026-11-03.
- [233]Criteo Advertising | Xnurta. Xnurta, Inc.. https://www.xnurta.com/criteo-advertising Accessed 2026-11-03.
- [234]Flywheel homepage (Commerce Cloud). Flywheel Digital (Omnicom). https://www.flywheeldigital.com/ Accessed 2026-11-03.
- [235]Omnicom Announces Closing of Acquisition of Flywheel. Omnicom Group Inc., 2024-01-02. https://investor.omc.com/news/news-details/2024/Omnicom-Announces-Closing-of-Acquisition-of-Flywheel/default.aspx Accessed 2026-11-03.
- [236]Perpetua homepage. Perpetua (Flywheel Digital Limited). https://perpetua.io/ Accessed 2026-11-03.
- [237]Quartile homepage. Quartile. https://www.quartile.com/ Accessed 2026-11-03.
- [238]Teikametrics homepage. Teikametrics, Inc.. https://www.teikametrics.com/ Accessed 2026-11-03.
- [239]Retail Media Management | CommerceIQ. CommerceIQ. https://www.commerceiq.ai/retail-media-management Accessed 2026-11-03.
- [240]Advertising (Ad Manager) | Stackline. Stackline. https://www.stackline.com/advertising Accessed 2026-11-03.
- [241]Intentwise Optimize. Intentwise, Inc.. https://www.intentwise.com/optimize Accessed 2026-11-03.
- [242]About Us | Intentwise. Intentwise, Inc.. https://www.intentwise.com/about-us Accessed 2026-11-03.
- [243]Pricing | Xnurta. Xnurta, Inc.. https://www.xnurta.com/pricing Accessed 2026-11-03.
- [244]Circana and DoorDash Partner to Validate Incremental Sales for CPG Brands. Circana LLC and DoorDash, Inc. (joint release), 2026-09-02. https://www.circana.com/post/circana-and-doordash-partner-to-validate-incremental-sales-for-cpg-brands Accessed 2026-11-03.
- [245]Gain Household-Level A/B Testing & Conduct Market/Store Tests | Circana (Media Lift) — browser-pane read relayed by the session. Circana, LLC. https://www.circana.com/solutions/media-lift Accessed 2026-11-03. Access: partial extract.
- [246]Consumer Insights for Media Companies | Numerator. Numerator, LLC and Affiliates, undated (copyright 2016-2026). https://www.numerator.com/industries/media/ Accessed 2026-11-03.
- [247]NIQ Matched Market Testing. Nielsen Consumer LLC (NIQ). https://nielseniq.com/global/en/products/niq-matched-market-testing/ Accessed 2026-11-03.
- [248]Brand and Media solutions | NIQ. Nielsen Consumer LLC (NIQ). https://nielseniq.com/global/en/solutions/brand-and-media/ Accessed 2026-11-03.
- [249]HUB360 | Catalina. Catalina. https://www.catalina.com/solutions/measurement-and-optimization/hub360 Accessed 2026-11-03.
- [250]Infillion Acquires Catalina, One of the World's Largest Sources of Deterministic Purchase Data (Catalina release). Catalina / Infillion, 2026-02-23. https://www.catalina.com/press-releases/infillion-acquires-catalina-one-of-the-worlds-largest-sources Accessed 2026-11-03.
- [251]Infillion Acquires Catalina (Infillion release). Infillion, 2026-02-23. https://infillion.com/blog/infillion-acquires-catalina-one-of-the-worlds-largest-sources-of-deterministic-purchase-data/ Accessed 2026-11-03.
- [252]Retail | Ipsos MMA. Ipsos MMA. https://mma.com/industries/retail/ Accessed 2026-11-03.
- [253]Getting Retail Media Right in MMM: New Evidence from DoorDash and Six QSR Brands (page title: Retail Media in MMM: DoorDash Ads Research | Ipsos MMA). Joe Lasala, Ipsos MMA, 2026-10-13. https://mma.com/blog/measuring-doordash-ads-marketing-mix-models/ Accessed 2026-11-03.
- [254]Nielsen Introduces Predictive Sales Lift, A New Capability To Help Advertisers and Agencies Gain Better Insights Into Media Campaign Outcomes. The Nielsen Company (US), LLC, 2026-04-27 (event 2026-04-27 (announcement); 'general availability in May' (planned)). https://www.nielsen.com/news-center/2026/nielsen-introduces-predictive-sales-lift-a-new-capability-to-help-advertisers-and-agencies-gain-better-insights-into-media-campaign-outcomes/ Accessed 2026-11-03.
- [255]LIFT ROI | Kantar. Kantar Group and Affiliates. https://www.kantar.com/campaigns/lift-roi Accessed 2026-11-03.
- [256]Retail | Attain. Attain. https://www.attainoutcomes.com/industry-retail Accessed 2026-11-03.
- [257]Our Data | Attain. Attain. https://www.attainoutcomes.com/our-data Accessed 2026-11-03.
- [258]Omnichannel Retail | Measured. Measured Inc.. https://www.measured.com/omnichannel-retail/ Accessed 2026-11-03.
- [259]Experiments | Haus. Haus Analytics, Inc.. https://www.haus.io/experiments Accessed 2026-11-03.
- [260]INCRMNTAL homepage. INCRMNTAL Ltd.. https://www.incrmntal.com/ Accessed 2026-11-03.
- [261]Master Subscription Agreement | INCRMNTAL. Incrmntal Ltd., 2022-12-01. https://www.incrmntal.com/msa Accessed 2026-11-03.
- [262]Lift Test 101 | WorkMagic Help Center. WorkMagic. https://help.workmagic.io/en/articles/11113679-lift-test-101 Accessed 2026-11-03.
- [263]Consumer Purchase Lift | Affinity Solutions. Affinity Solutions. https://www.affinity.solutions/consumer-purchase-lift/ Accessed 2026-11-03.
- [264]Affinity Solutions homepage. Affinity Solutions. https://www.affinity.solutions/ Accessed 2026-11-03.
- [265]Incremental homepage. Incremental. https://www.incremental.com/ Accessed 2026-11-03.
- [266]Retail Media Measurement & Incrementality for CPG (LiftLab blog). LiftLab Analytics, Inc., 2026-09-10. https://liftlab.com/blog/retail-media-measurement-cpg/ Accessed 2026-11-03.
- [267]ncsolutions.com (redirect). NCSolutions / Circana. https://www.ncsolutions.com/ Accessed 2026-11-03. Access: partial extract.
- [268]Circana to Acquire NCSolutions and Nielsen's Marketing Mix Modeling Business. Nielsen (The Nielsen Company (US), LLC site), 2024-08-26. https://www.nielsen.com/news-center/2024/circana-to-acquire-ncsolutions-and-nielsens-marketing-mix-modeling-business/ Accessed 2026-11-03.
- [269]Amazon.com, Inc. Form 10-K for fiscal 2025, XBRL R88 'Segment Information - Disaggregation of Revenue (Details)' (accession 0001018724-26-000004, filed 6 February 2026). Amazon.com, Inc. / SEC EDGAR, 2026-02-06 (event 2025-12-31). https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/R88.htm Accessed 2026-11-03.
- [270]Maplebear Inc. Form 10-K for fiscal 2025, XBRL R47 'Revenue - Schedule of Reconciliation of Revenue from Segments to Consolidated (Details)' (accession 0001579091-26-000018, filed 26 February 2026). Maplebear Inc. / SEC EDGAR, 2026-02-26 (event 2025-12-31). https://www.sec.gov/Archives/edgar/data/1579091/000157909126000018/R47.htm Accessed 2026-11-03.
- [271]Criteo S.A. FY2025 10-K, XBRL R60: Segment information - Schedule of Segment Reporting Information (Details). Criteo S.A. (SEC EDGAR), 2026-02-26. https://www.sec.gov/Archives/edgar/data/1576427/000157642726000014/R60.htm Accessed 2026-11-03.
- [272]Cardlytics, Inc. FY2025 10-K, XBRL R77: SEGMENTS - Revenue by Segment (Details). Cardlytics, Inc. (SEC EDGAR), 2026-03-04. https://www.sec.gov/Archives/edgar/data/1666071/000166607126000010/R77.htm Accessed 2026-11-03.
- [273]Weekly Bankruptcy Alert June 1, 2026 (For the Week Ending May 31, 2026). Pierce Atwood LLP (Bankruptcy & Creditors' Rights), via The National Law Review, 2026-06-01 (event 2026-05-29). https://natlawreview.com/article/weekly-bankruptcy-alert-june-1-2026-week-ending-may-31-2026 Accessed 2026-11-03.
- [274]shopkick.com (home). Shopkick, Inc. (domain). https://www.shopkick.com/ Accessed 2026-11-03. Access: partial extract.
- [275]About Us (timeline and login menu). Neptune Retail Solutions. https://neptuneretailsolutions.com/about-us/ Accessed 2026-11-03.
- [276]Numerator | Kantar H1 2026 interim results (PDF). Kantar Global Holdings S.a r.l., 2026-08-20 (event 2026-01-01). https://www.kantar.com/-/media/E7E16CF3DDB84DAA990EAFC4B582A26E.ashx Accessed 2026-11-03.
- [277]How to take the guesswork out of ad measurement. Kroger Precision Marketing, 2024-07-19. https://www.krogerprecisionmarketing.com/knowledge-hub/blog/how-to-take-the-guesswork-out-of-ad-measurement/ Accessed 2026-11-03.
- [278]New Incremental Sales Measurement Enhances KPM Programmatic Campaigns. Kroger Precision Marketing, 2025-01-14. https://www.krogerprecisionmarketing.com/knowledge-hub/blog/new-incremental-sales-measurement-enhances-kpm-programmatic-campaigns/ Accessed 2026-11-03.
- [279]Programmatic Advertising. Kroger Precision Marketing. https://www.krogerprecisionmarketing.com/our-solutions/retail-media/programmatic Accessed 2026-11-03.
- [280]New Ways to Activate and Extend Reach in Offsite Media. Kroger Precision Marketing, 2026-08-11. https://www.krogerprecisionmarketing.com/knowledge-hub/blog/new-ways-to-activate-and-extend-reach-in-offsite-media/ Accessed 2026-11-03.
- [281]Informazioni su Pacvue (About Pacvue, Italian locale). Pacvue. https://pacvue.com/it/about/about-us/ Accessed 2026-11-03.
- [282]IAB/PwC Internet Advertising Revenue Report: Full-year 2025 results (April 2026), commerce media pages. IAB and PwC, 2026-04. https://iabhongkong.com/sites/default/files/2026-04/IAB_PwC_Internet_Ad_Revenue_Report_Full_Year_2025_April_2026.pdf Accessed 2026-11-03.
- [283]IAB Internet Advertising Revenue Report: Full-year 2024 results (PDF, April 2025). PwC for IAB, 2025-04 (event 2024-12-31). https://www.iab.com/wp-content/uploads/2025/04/IAB_PwC-Internet-Ad-Revenue-Report-Full-Year-2024.pdf Accessed 2026-11-03.
- [284]Citi's financial media network brings cardholder data to marketers. EMARKETER (Katherine Smith), 2026-09-23. https://www.emarketer.com/content/citi-s-financial-media-network-brings-cardholder-data-marketers Accessed 2026-11-03.
- [285]US Retail Media Ad Spending Forecast (page). EMARKETER. https://www.emarketer.com/content/us-retail-media-ad-spending-forecast Accessed 2026-11-03. Access: partial extract.
- [286]Amazon gives advertisers access to ChatGPT ads with OpenAI partnership. EMARKETER (Rachel Wolff and Marisa Jones), 2026-09-10. https://www.emarketer.com/content/amazon-gives-advertisers-access-chatgpt-ads-with-openai-partnership Accessed 2026-11-03.
- [287]This Year Next Year (hub page). WPP Media. https://www.wppmedia.com/this-year-next-year Accessed 2026-11-03. Access: partial extract.
- [288]This Year Next Year: Mid-Year 2026. WPP Media. https://www.wppmedia.com/thought-leadership/this-year-next-year/mid-year-2026 Accessed 2026-11-03. Access: partial extract.
- [289]Forecasts (page). MAGNA. https://magnaglobal.com/forecasts/ Accessed 2026-11-03. Access: partial extract.
- [290]dentsu Global Ad Spend Forecasts | Mid Year 2026 (our-latest-thinking page). dentsu. https://www.dentsu.com/our-latest-thinking/dentsu-ad-spend-forecast-mid-year-2026 Accessed 2026-11-03. Access: partial extract.
- [291]dentsu Ad Spend Forecasts - Mid Year 2026 (report page). dentsu. https://www.dentsu.com/hu/en/reports/dentsu_ad_spend_forecasts___mid_year_2026_aa1cb9b Accessed 2026-11-03. Access: partial extract.
- [292]Expedia Group, Inc. Form 10-K for fiscal 2025 — R92.htm Segment Information - Schedule of Revenue by Services and Geographic Area (Details). Expedia Group, Inc. (SEC EDGAR), 2026-02-13. https://www.sec.gov/Archives/edgar/data/1324424/000132442426000008/R92.htm Accessed 2026-11-03.
- [293]Walmart Reports Fourth Quarter Results (fiscal 2024 Q4 earnings release, 8-K exhibit 99.1). Walmart Inc. (SEC EDGAR), 2024-02-20. https://www.sec.gov/Archives/edgar/data/104169/000010416924000019/earningsreleasefy24q4.htm Accessed 2026-11-03.
- [294]Walmart reports fourth quarter results (fiscal 2025 Q4 earnings release, 8-K exhibit 99.1). Walmart Inc. (SEC EDGAR), 2025-02-20. https://www.sec.gov/Archives/edgar/data/104169/000010416925000010/earningsreleasefy25q4.htm Accessed 2026-11-03.
- [295]Maplebear Inc. Form 10-K for fiscal 2025, XBRL R12 'Significant Accounting Policies' (note text; accession 0001579091-26-000018, filed 26 February 2026). Maplebear Inc. / SEC EDGAR, 2026-02-26 (event 2025-12-31). https://www.sec.gov/Archives/edgar/data/1579091/000157909126000018/R12.htm Accessed 2026-11-03.
- [296]Maplebear Inc. Form 10-K for fiscal 2025, XBRL R13 'Revenue' (note; accession 0001579091-26-000018). Maplebear Inc. / SEC EDGAR, 2026-02-26 (event 2025-12-31). https://www.sec.gov/Archives/edgar/data/1579091/000157909126000018/R13.htm Accessed 2026-11-03.
- [297]Instacart Announces Third Quarter 2026 Financial Results (Form 8-K exhibit 99.1, accession 0001193125-26-437014). Maplebear Inc. d/b/a Instacart / SEC EDGAR, 2026-10-29 (event 2026-09-30). https://www.sec.gov/Archives/edgar/data/1579091/000119312526437014/d24470dex991.htm Accessed 2026-11-03.
- [298]Criteo S.A. FY2025 10-K, XBRL R110: Breakdown of Revenue and Noncurrent Assets - Schedule of Revenue by Product and Services (Details). Criteo S.A. (SEC EDGAR), 2026-02-26. https://www.sec.gov/Archives/edgar/data/1576427/000157642726000014/R110.htm Accessed 2026-11-03.
- [299]CRITEO REPORTS SECOND QUARTER 2026 RESULTS (Form 8-K exhibit 99.1). Criteo S.A. (SEC EDGAR), 2026-08-05. https://www.sec.gov/Archives/edgar/data/1576427/000162828026052913/exhibit991-8xkq22026.htm Accessed 2026-11-03.
- [300]Cardlytics, Inc. FY2025 10-K, XBRL R79: SEGMENTS - Geographical Information (Details). Cardlytics, Inc. (SEC EDGAR), 2026-03-04. https://www.sec.gov/Archives/edgar/data/1666071/000166607126000010/R79.htm Accessed 2026-11-03.
- [301]Cardlytics, Inc. Form 8-K, Item 2.01 Completion of Acquisition or Disposition of Assets (Bridg Sale). Cardlytics, Inc. (SEC EDGAR), 2026-03-24. https://www.sec.gov/Archives/edgar/data/1666071/000162828026020811/cdlx-20260324.htm Accessed 2026-11-03.
- [302]Amazon.com, Inc. Form 10-K for fiscal 2025, XBRL R19 'Segment Information' (note text; accession 0001018724-26-000004, filed 6 February 2026). Amazon.com, Inc. / SEC EDGAR, 2026-02-06 (event 2025-12-31). https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/R19.htm Accessed 2026-11-03.
- [303]Amazon.com, Inc. Form 10-Q for the quarter ended 30 September 2026, XBRL R54 'Segment Information - Disaggregation of Revenue (Details)' (accession 0001018724-26-000040, filed 30 October 2026). Amazon.com, Inc. / SEC EDGAR, 2026-10-30 (event 2026-09-30). https://www.sec.gov/Archives/edgar/data/1018724/000101872426000040/R54.htm Accessed 2026-11-03.
- [304]eBay Inc. Reports Fourth Quarter and Full Year 2025 Results (Form 8-K exhibit 99.1, accession 0001065088-26-000017). eBay Inc. / SEC EDGAR, 2026-02-18 (event 2025-12-31). https://www.sec.gov/Archives/edgar/data/1065088/000106508826000017/exhibit991erebayq42025.htm Accessed 2026-11-03.
- [305]The Kroger Co. Form 10-K for the fiscal year ended January 31, 2026. The Kroger Co. (SEC EDGAR), 2026-03-31 (event fiscal year ended 2026-01-31). https://www.sec.gov/Archives/edgar/data/56873/000110465926037723/kr-20260131x10k.htm Accessed 2026-11-03. Access: partial extract.
- [306]DoorDash, Inc. shareholder letter, fourth quarter 2025 (Form 8-K exhibit 99.2, accession 0001792789-26-000012). DoorDash, Inc. / SEC EDGAR, 2026-02-18 (event 2025-12-31). https://www.sec.gov/Archives/edgar/data/1792789/000179278926000012/proddashq4ex992-shareholde.htm Accessed 2026-11-03.
- [307]DoorDash Releases Fourth Quarter and Full Year 2025 Financial Results (Form 8-K exhibit 99.1). DoorDash, Inc. / SEC EDGAR, 2026-02-18. https://www.sec.gov/Archives/edgar/data/1792789/000179278926000012/prodq4dashex991-pressrelea.htm Accessed 2026-11-03.
- [308]EDGAR full-text searches: "advertising" and "advertisers" in DoorDash filings of 10-25 February 2026; "advertising" in Uber filings of February 2026. SEC EDGAR (efts). https://efts.sec.gov/LATEST/search-index?q=%22advertising%22&dateRange=custom&startdt=2026-02-10&enddt=2026-02-25&ciks=0001792789 Accessed 2026-11-03. Access: partial extract.
- [309]Neptune Retail Solutions Completes Acquisition Of Quotient Technology Inc.. Neptune Retail Solutions (PR Newswire), 2023-09-05. https://www.prnewswire.com/news-releases/neptune-retail-solutions-completes-acquisition-of-quotient-technology-inc-301917787.html Accessed 2026-11-03.
- [310]Vizio Holding Corp. Form 8-K (Item 1.01 Agreement and Plan of Merger with Walmart Inc.). Vizio Holding Corp. (SEC EDGAR), 2024-02-20 (event 2024-02-19). https://www.sec.gov/Archives/edgar/data/1835591/000119312524040047/d794432d8k.htm Accessed 2026-11-03.
- [311]Walmart Completes Acquisition of VIZIO. Walmart Inc., 2024-12-03. https://corporate.walmart.com/news/2024/12/03/walmart-completes-acquisition-of-vizio Accessed 2026-11-03.
- [312]Walmart Inc. fiscal 2025 Form 10-K (main document). Walmart Inc. (SEC EDGAR), 2025-03-14 (event 2024-12). https://www.sec.gov/Archives/edgar/data/104169/000010416925000021/wmt-20250131.htm Accessed 2026-11-03. Access: partial extract.
- [313]Circana Officially Acquires NCSolutions | Circana. Circana, LLC, 2025-06-02. https://www.circana.com/post/circana-completes-acquisition-of-ncsolutions Accessed 2026-11-03. Access: partial extract.
- [314]Circana Completes Acquisition of NCSolutions (GlobeNewswire syndication). Circana via GlobeNewswire (Mammoth Times reprint), 2025-06-02. https://business.mammothtimes.com/mammothtimes/article/gnwcq-2025-6-2-circana-completes-acquisition-of-ncsolutions Accessed 2026-11-03. Access: partial extract.
- [315]Circana Completes Acquisition of Nielsen’s Marketing Mix Modeling Business (circana.com post). Circana, LLC, 2025-08-21. https://www.circana.com/post/circana-completes-acquisition-of-nielsen-s-marketing-mix-modeling-business Accessed 2026-11-03.
- [316]DoorDash, Inc. Annual Report on Form 10-K for the fiscal year ended December 31, 2025. DoorDash, Inc. (SEC filing), 2026-02-18. https://www.sec.gov/Archives/edgar/data/1792789/000179278926000013/dash-20251231.htm Accessed 2026-11-03.
- [317]Infillion blog (listing page). Infillion. https://infillion.com/blog Accessed 2026-11-03.
- [318]Acquisizione di Catalina da parte di Infillion. Catalina (Italian site, catalinamarketing.it), 2026-02-26. https://catalinamarketing.it/news/acquisizione-di-catalina-da-parte-di-infillion Accessed 2026-11-03.
- [319]Publicis to acquire LiveRamp to accelerate data co-creation for smarter agents. Publicis Groupe, 2026-05-17. https://www.publicisgroupe.com/en/news/press-releases/publicis-to-acquire-liveramp-to-accelerate-data-co-creation-for-smarter-agents Accessed 2026-11-03.
- [320]FTC Order Preserves Competition in Digital Advertising Ecosystem as Part of Publicis's Acquisition of LiveRamp. Federal Trade Commission, 2026-10-19. https://www.ftc.gov/news-events/news/press-releases/2026/10/ftc-order-preserves-competition-digital-advertising-ecosystem-part-publiciss-acquisition-liveramp Accessed 2026-11-03.
- [321]LiveRamp Expands Meta Insights Available to Retail Media Networks. LiveRamp Holdings, Inc. (Business Wire release on LiveRamp's newsroom), 2025-10-23. https://liveramp.com/news/liveramp-expands-meta-insights-available-to-retail-media-networks Accessed 2026-11-03.
- [322]Publicis Groupe Completes LiveRamp Acquisition, Launches the Industry’s First Plug-And-Play AI Growth Engine that Clients Fully Own and Control. Publicis Groupe, 2026-11-02. https://www.publicisgroupe.com/en/news/press-releases/publicis-groupe-completes-liveramp-acquisition-launches-the-industry-s-first-plug-and-play-ai-growth-engine-that-clients-fully-own-and-control Accessed 2026-11-03.
- [323]NIQ Global Intelligence plc Form 10-Q for the quarter ended 30 June 2025 (Note 1, 'Initial Public Offering'). NIQ Global Intelligence plc / SEC EDGAR, 2025-08-15. https://www.sec.gov/Archives/edgar/data/2054696/000205469625000012/niq-20250630.htm Accessed 2026-11-03.
- [324]NIQ Announces Pricing of Initial Public Offering. NIQ (NielsenIQ), 2025-07-22. https://nielseniq.com/global/en/news-center/2025/niq-announces-pricing-of-initial-public-offering/ Accessed 2026-11-03.
- [325]NIQ Global Intelligence plc Form 8-K, date of report 24 July 2025 (accession 0001193125-25-164427). NIQ Global Intelligence plc / SEC EDGAR, 2025-07-24. https://www.sec.gov/Archives/edgar/data/2054696/000119312525164427/d54238d8k.htm Accessed 2026-11-03.
- [326]EDGAR full-text search: "NIQ Global Intelligence", form 8-K, 20 July to 10 August 2025. SEC EDGAR (efts). https://efts.sec.gov/LATEST/search-index?q=%22NIQ%20Global%20Intelligence%22&forms=8-K&dateRange=custom&startdt=2025-07-20&enddt=2025-08-10 Accessed 2026-11-03.
- [327]Form 8-K, Walgreens Boots Alliance, Inc. (completion of merger). Walgreens Boots Alliance, Inc. (SEC EDGAR), 2025-08-28. https://www.sec.gov/Archives/edgar/data/1618921/000119312525190603/d87240d8k.htm Accessed 2026-11-03.
- [328]Cardlytics Completes Sale of Bridg Assets to PAR Technology. Cardlytics, Inc., 2026-03-24. https://ir.cardlytics.com/news/news-details/2026/Cardlytics-Completes-Sale-of-Bridg-Assets-to-PAR-Technology/default.aspx Accessed 2026-11-03.
- [329]PAR Technology to Acquire Bridg, Bringing Loyalty and Non-Loyalty Data Together for Smarter Retail and Restaurant Activation (8-K exhibit). PAR Technology Corporation, 2026-01-26. https://www.sec.gov/Archives/edgar/data/708821/000070882126000007/parxbridgrelease12626.htm Accessed 2026-11-03.
- [330]Infillion to Acquire Foursquare, Closing the Loop Between Advertising and Real-World Results. Infillion, 2026-09-18. https://infillion.com/blog/infillion-to-acquire-foursquare-closing-the-loop-between-advertising-and-real-world-results/ Accessed 2026-11-03.
- [331]Foursquare home page (banner on Infillion). Foursquare. https://foursquare.com/ Accessed 2026-11-03.
- [332]DoorDash, Inc. Form 8-K, Item 1.01 Agreement and Plan of Merger with Maplebear Inc. (accession 0001140361-26-041111). DoorDash, Inc. / SEC EDGAR, 2026-10-29 (event 2026-10-28). https://www.sec.gov/Archives/edgar/data/1792789/000114036126041111/ef20084547_8k.htm Accessed 2026-11-03.
- [333]DoorDash to Acquire Instacart to Help Local Stores Win (Maplebear Form 8-K exhibit 99.1, accession 0001193125-26-437075). DoorDash, Inc. and Maplebear Inc. (joint release, filed by Maplebear), 2026-10-29 (event 2026-10-28). https://www.sec.gov/Archives/edgar/data/1579091/000119312526437075/d171568dex991.htm Accessed 2026-11-03.
- [334]The Kroger Co. Form 8-K (Item 1.02 Termination of a Material Definitive Agreement). The Kroger Co. (SEC EDGAR), 2024-12-11. https://www.sec.gov/Archives/edgar/data/56873/000110465924127669/tm2427516d10_8k.htm Accessed 2026-11-03.
- [335]Albertsons Companies, Inc. Form 8-K (Item 1.02 Termination of a Material Definitive Agreement). Albertsons Companies, Inc. (SEC EDGAR), 2024-12-11 (event 2024-12-10). https://www.sec.gov/Archives/edgar/data/1646972/000164697224000231/aci-20241210.htm Accessed 2026-11-03.
- [336]Kroger Reiterates Its Commitment to Lower Prices and Initiates New $7.5B Share Buyback Program (8-K exhibit 99.1). The Kroger Co. (SEC EDGAR), 2024-12-11. https://www.sec.gov/Archives/edgar/data/56873/000110465924127669/tm2427516d10_ex99-1.htm Accessed 2026-11-03.
- [337]IRI and NPD Rebrand as Circana (circana.com post). Circana. https://www.circana.com/post/iri-and-npd-rebrand-as-circana-the-leading-advisor-on-the-complexity-of-consumer-behavior Accessed 2026-11-03. Access: partial extract.
- [338]IRI and NPD Rebrand as Circana (circana.com post). Circana (event 2023-03-07). https://www.circana.com/es-es/post/iri-and-npd-rebrand-as-circana-the-leading-advisor-on-the-complexity-of-consumer-behavior Accessed 2026-11-03. Access: partial extract.
- [339]IRI and NPD Rebrand as Circana, the Leading Advisor on the Complexity of Consumer Behavior (Hellman & Friedman news post). Hellman & Friedman LLC, 2023-03-07. https://hf.com/?p=10309 Accessed 2026-11-03.
- [340]Pacvue Unveils the Industry's First Commerce Acceleration Platform (Italian locale). Pacvue, 2023-03-27. https://pacvue.com/it/blog/pacvue-unveils-the-industrys-first-commerce-acceleration-platform/ Accessed 2026-11-03.
- [341]Amazon Ads introduces Full-Funnel Campaigns, DVA+, and enhanced chat experience in AI-powered ad platform, Amazon Ads Agent. Amazon Ads, 2026-09-29. https://advertising.amazon.com/en-us/library/news/amazon-ads-agent Accessed 2026-11-03.
- [342]In-House Retail Media: Control Data with Kevel (PromoteIQ shutdown post). Kevel Team. https://kevel.com/blog/promoteiq-shutdown Accessed 2026-11-03.
- [343]promoteiq.com. PromoteIQ / Microsoft. https://www.promoteiq.com/ Accessed 2026-11-03. Access: partial extract.
- [344]Maplebear Inc. Form 8-K, Item 2.05 Costs Associated with Exit or Disposal Activities (accession 0001579091-26-000050). Maplebear Inc. / SEC EDGAR, 2026-10-26 (event 2026-10-25). https://www.sec.gov/Archives/edgar/data/1579091/000157909126000050/cart-20261025.htm Accessed 2026-11-03.
- [345]Message from the Founder: Launching Pentaleap. Andreas Reiffen, Pentaleap, 2023-09-03. https://www.pentaleap.com/blog/message-from-the-founder Accessed 2026-11-03.
- [346]Criteo Completes Redomiciliation from France to Luxembourg; Board Approves Next Step Toward U.S. Domicile (release PDF). Criteo S.A., 2026-07-29. https://www.criteo.com/wp-content/uploads/2026/07/Criteo-Completes-Redomiciliation-from-France-to-Luxembourg_FINAL.pdf Accessed 2026-11-03.
- [347]Criteo S.A. Form 8-K dated 5 August 2026 (Item 1.01 Entry into a Material Definitive Agreement; Item 8.01). Criteo S.A. (SEC EDGAR), 2026-08-05. https://www.sec.gov/Archives/edgar/data/1576427/000162828026053414/crto-20260805.htm Accessed 2026-11-03.
- [348]Criteo S.A. Form 8-K dated 28 October 2026 (Item 5.02). Criteo S.A. (SEC EDGAR), 2026-10-29 (event 2026-10-28). https://www.sec.gov/Archives/edgar/data/1576427/000157642726000101/crto-20261028.htm Accessed 2026-11-03.
- [349]EDGAR filing list: Maplebear Inc. 10-Q filings. SEC EDGAR. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=1579091&type=10-Q&dateb=&owner=include&count=5 Accessed 2026-11-03.
- [350]Reporting API v1 overview: 'What has changed in the new Reporting API' (Amazon Ads Advanced Tools Center). Amazon Ads, undated (footer '© 2023'). https://advertising.amazon.com/API/docs/en-us/guides/reporting/ads-v1/overview Accessed 2026-11-03.
- [351]Purchases metric family (Reporting API v1 metrics library: Amazon retail conversions). Amazon Ads, undated. https://advertising.amazon.com/API/docs/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/purchases Accessed 2026-11-03. Access: partial extract.
- [352]Metrics: all metrics available for the version 3 reporting API (Amazon Ads Advanced Tools Center, Reporting, Version 3, Columns). Amazon Ads, undated (footer '© 2023 Amazon.com, Inc. or its affiliates.'). https://advertising.amazon.com/API/docs/en-us/guides/reporting/v3/columns Accessed 2026-11-03.
- [353]View Attribution Updates for Amazon Store ads. Amazon Ads, 2026-01-01. https://advertising.amazon.com/en-us/resources/whats-new/view-attribution-updates-for-amazon-store-ads/ Accessed 2026-11-03.
- [354]Sponsored Products prompts and Sponsored Brands prompts. Amazon Ads, 2026-03-10. https://advertising.amazon.com/en-us/resources/whats-new/unboxed-2025-sponsored-products-and-sponsored-brands-prompts Accessed 2026-11-03.
- [355]Approved third-party providers. Amazon Ads, 2026-06-01. https://advertising.amazon.com/en-us/resources/ad-policy/approved-3p-ad-servers Accessed 2026-11-03.
- [356]Parameters Generated by Snapshot Reports v2 (Sponsored Search APIs, Walmart Connect Developer Portal). Walmart Connect Developer Portal (Advertising Partners - Search), 2026-07-16 (updatedAt 2026-07-16T02:23:14Z). https://developer.walmart.com/advertising-partners-search/docs/parameters-generated-by-snapshot-reports-v2-updated Accessed 2026-11-03.
- [357]Advertiser reports (Walmart Connect Advertising Help, article 000008390). Walmart Connect (Walmart Inc.), undated on the page. https://advertisinghelp.walmart.com/s/guides?channel=Sponsored%20Products&article=000008390&language=en_US Accessed 2026-11-03.
- [358]Glossary (Walmart Connect Advertising Help, Sponsored Products, article 000008385). Walmart Connect (Walmart Inc.), undated on the page. https://advertisinghelp.walmart.com/s/guides?channel=Sponsored%20Products&article=000008385&language=en_US Accessed 2026-11-03.
- [359]How we're bringing incremental sales measurement to Sponsored Search. Jeff Clark, VP of Product & Commercialization, Walmart Connect, 2025-04-28 (event 2025-04-28 (announcement; broader rollout to managed advertisers planned 'this summer')). https://www.walmartconnect.com/resources/articles/2025/sponsored-search-incremental-sales-measurement Accessed 2026-11-03.
- [360]Definition of Various Parameters Generated Across the Snapshot Reports (Display APIs, Walmart Connect Developer Portal). Walmart Connect Developer Portal (Advertising Partners - Display), 2026-10-28 (updatedAt 2026-10-28T12:48:30Z). https://developer.walmart.com/advertising-partners/docs/definition-of-various-parameters-generated-across-the-snapshot-reports Accessed 2026-11-03.
- [361]June 2026 Release Notes (Display APIs changelog, Walmart Connect Developer Portal). Walmart Connect Developer Portal (Advertising Partners - Display), 2026-06 (event 2026-06-01 (identity and attribution upgrades)). https://developer.walmart.com/advertising-partners/changelog/june-2026-release-notes Accessed 2026-11-03.
- [362]Integrated measurement solutions, tailored to your goals (insights page). Walmart Connect (Walmart Inc.). https://www.walmartconnect.com/insights/integrated-measurement-solutions Accessed 2026-11-03.
- [363]Walmart Connect Launches Its New Demand-Side Platform, Walmart DSP, to Expand Its Off-Site Media Offerings at Scale (post by Rich Lehrfeld). Walmart Inc. (Rich Lehrfeld, SVP and GM, Walmart Connect), 2021-08-25 (event 2021-10 (planned availability 'by the end of October')). https://corporate.walmart.com/news/2021/08/25/walmart-connect-launches-its-new-demand-side-platform-walmart-dsp-to-expand-its-off-site-media-offerings-at-scale Accessed 2026-11-03.
- [364]Expanding retail-powered social media. Walmart Connect / Walmart Inc., 2026-04-08. https://www.walmartconnect.com/resources/articles/2026/expanding-retail-powered-social-media Accessed 2026-11-03.
- [365]Sam's Club Connect (home page). Sam's Club / Walmart, Inc.. https://www.samsclubconnect.com/ Accessed 2026-11-03.
- [366]Measuring sponsored product performance (Instacart Ads Manager docs). Maplebear Inc. d/b/a Instacart (docs.instacart.com), 2026-09-17. https://docs.instacart.com/ads_manager/sponsored_products/measuring_sponsored_product_performance Accessed 2026-11-03.
- [367]Linear attribution reporting (Instacart Ads Manager). Maplebear Inc. d/b/a Instacart, 2026-08-12 (event 2022-08-01). https://docs.instacart.com/ads_manager/data_and_analytics/linear_attribution_reporting Accessed 2026-11-03.
- [368]Introducing New Sales Lift Metrics and A/B Testing for CPG Brands (Instacart Ads). Maplebear Inc. d/b/a Instacart, 2022-09-02 (event 2022-09). https://web.archive.org/web/20251212091556/https://www.instacart.com/company/instacart-ads/introducing-new-sales-lift-metrics Accessed 2026-11-03.
- [369]Reckitt Achieves 16.5% Incremental Sales Lift from Instacart Ads. Maplebear Inc. d/b/a Instacart, 2023-05-05 (event 2022 Q3-Q4 (test)). https://web.archive.org/web/20260306110304/https://www.instacart.com/company/instacart-ads/reckitt-achieves-16-5-incremental-sales-lift-from-instacart-ads Accessed 2026-11-03.
- [370]How Instacart Measures the True Value of Advertising: The Methodology of Ad Incrementality. Jason Kim, Instacart (tech-at-instacart, Medium), 2023-06-30. https://tech.instacart.com/how-instacart-measures-the-true-value-of-advertising-the-methodology-of-ad-incrementality-aac6c58e627c Accessed 2026-11-03.
- [371]Instacart Ads measurement and insights page (Internet Archive capture of 28 September 2026, raw HTML). Maplebear Inc. d/b/a Instacart. https://web.archive.org/web/20260928080759/https://company.instacart.com/ads/measurement-insights Accessed 2026-11-03. Access: partial extract.
- [372]Measuring display ad performance (Instacart Ads Manager docs). Maplebear Inc. d/b/a Instacart (docs.instacart.com), 2026-08-12. https://docs.instacart.com/ads_manager/display_ads/measuring_display_ad_performance Accessed 2026-11-03.
- [373]Report Center (84.51° Prism Learning Center). 84.51° LLC, 2026-08-17. https://prism.8451.com/activation/help/content/about/report-center.htm Accessed 2026-11-03. Access: partial extract.
- [374]Glossary (84.51° Prism Learning Center). 84.51° LLC, 2026-05-22. https://prism.8451.com/activation/help/content/resources-faqs/glossary.htm Accessed 2026-11-03. Access: partial extract.
- [375]Create a Managed Service Campaign (84.51° Prism Learning Center). 84.51° LLC, 2026-04-16. https://prism.8451.com/activation/help/content/how-to/kpm-managed-service/create-campaign.htm Accessed 2026-11-03. Access: partial extract.
- [376]Performance Metrics Are a Governance Issue. Retail Media Must Elevate How It Defines Impact (blog post by Liz Roche, Vice President of Media & Measurement). Albertsons Media Collective, 2026-03-23. https://www.albertsonsmediacollective.com/news/blog/blog-details/2026/Performance-Metrics-Are-a-Governance-Issue--Retail-Media-Must-Elevate-How-It-Defines-Impact/default.aspx Accessed 2026-11-03.
- [377]Sponsored Search Stays Top of Mind (Bob's Red Mill case study). Albertsons Media Collective (event campaign August to December 2023). https://www.albertsonsmediacollective.com/results/sponsored-search-stays-top-of-mind/ Accessed 2026-11-03.
- [378]Albertsons Media Collective Launches Collective TV - Powered by First-Party Data, Closed-Loop Measurement, Shoppable Ads and Premium Inventory (release). Albertsons Media Collective, 2024-06-13. https://www.albertsonsmediacollective.com/pr-details/2024/Albertsons-Media-Collective-Launches-Collective-TV---Powered-by-First-Party-Data-Closed-Loop-Measurement-Shoppable-Ads-and-Premium-Inventory/default.aspx Accessed 2026-11-03.
- [379]Digital Out-of-Home (offerings page). Albertsons Media Collective (Albertsons Companies, Inc.). https://www.albertsonsmediacollective.com/offerings/digital-out-of-home/default.aspx Accessed 2026-11-03.
- [380]Albertsons Media Collective - Offsite Display & Collective TV. Albertsons Media Collective (Albertsons Companies, Inc.). https://www.albertsonsmediacollective.com/offerings/offsite-display-and-collective-tv/default.aspx Accessed 2026-11-03.
- [381]In-Store (offerings page). Albertsons Media Collective (Albertsons Companies, Inc.). https://www.albertsonsmediacollective.com/offerings/in-store/default.aspx Accessed 2026-11-03.
- [382]Understanding Key Metrics (Ads Support, DoorDash Help Center). DoorDash, Inc.. https://help.doordash.com/en-us/advertising/article/understanding-key-metrics Accessed 2026-11-03.
- [383]How to Create and Manage Sponsored Listings (Merchant Support, DoorDash Help Center, en-US). DoorDash, Inc.. https://help.doordash.com/merchants/s/article/Getting-started-with-Self-Serve-Sponsored-Listing?language=en_US Accessed 2026-11-03.
- [384]Metrics Glossary (Uber Ads API, Uber Developers). Uber Technologies Inc.. https://developer.uber.com/docs/ads/reporting-metrics/metrics-glossary Accessed 2026-11-03.
- [385]Uber Advertising Launches Cross-Channel Incrementality Measurement, Powered by Affinity Solutions, to Prove Sponsored Listings Drive Brick-and-Mortar Sales. Uber Advertising (Uber Technologies, Inc.), 2026-10-05. https://www.uber.com/us/en/blog/sponsored-listings-cross-channel-incrementality/ Accessed 2026-11-03.
- [386]The 5 biggest announcements from the 2026 Best Buy Ads Showcase, The Drop.. Best Buy Ads, 2026 (undated on the page; the corporate post RTL-U-S048 is dated 2026-09-28). https://www.bestbuyads.com/insights/the_5_biggest_announcements_from_the_drop Accessed 2026-11-03.
- [387]How Lowe's is building and measuring a full-funnel media network. Marketing Dive (Chris Kelly, Senior Reporter; Informa TechTarget), 2026-10-13. https://marketingdive.com/news/how-lowes-is-building-and-measuring-a-full-funnel-media-network/832747 Accessed 2026-11-03.
- [388]Walgreens Advertising Group — home. Walgreens Advertising Group (Walgreen Co.), 2024 (footer). https://www.walgreensadvertisinggroup.com/ Accessed 2026-11-03.
- [389]Wayback Machine capture of Kohl's corporate release 'Kohl's Expands Advertising Opportunities with Kohl's Media Network'. Internet Archive. https://web.archive.org/web/2026/https://corporate.kohls.com/news/kohls-expands-advertising-opportunities-with-kohls-media-network Accessed 2026-11-03. Access: partial extract.
- [390]On-site Retail Media Solutions | DGMN. DG Media Network, LLC, 2026. https://www.dgmedianetwork.com/on-site-solutions Accessed 2026-11-03.
- [391]Off-site Retail Media Solutions | DGMN. DG Media Network, LLC, 2026. https://www.dgmedianetwork.com/off-site-solutions Accessed 2026-11-03.
- [392]DG Media Network Offers New Ad Solutions for Meta Platforms. DG Media Network (Dollar General Corporation), 2023-03-27. https://www.dgmedianetwork.com/news/dg-media-network-offers-new-ad-solutions-for-meta-platforms Accessed 2026-11-03.
- [393]Seller Announcement: Updates to eBay's Promoted Listings are coming soon (community announcement). eBay Inc. (seller_news_team, eBay Staff), 2025 (shown only as 'Posted 1 year ago') (event 2026-01-13 (effective)). https://community.ebay.com/forum/announcements-57928/topic/seller-announcement-updates-to-ebays-promoted-listings-are-coming-soon-8506/ Accessed 2026-11-03.
- [394]eBay advertising (Promoted Listings hub, export site). eBay Inc.. https://export.ebay.com/en/marketing/promoted-listings-hub/ Accessed 2026-11-03.
- [395]How to Set Up and Manage an Etsy Ads Campaign. Etsy, Inc.. https://help.etsy.com/hc/articles/360033701174 Accessed 2026-11-03.
- [396]How to Use Etsy Stats for Your Shop. Etsy, Inc.. https://help.etsy.com/hc/articles/115015774268 Accessed 2026-11-03.
- [397]Hy-Vee RedMedia — home. Hy-Vee RedMedia (Hy-Vee, Inc.), 2026. https://www.hredmedia.com/ Accessed 2026-11-03.
- [398]How TravelAds features help drive online bookings. Expedia Group Advertising, 2024-04-25. https://partner.expediagroup.com/en-us/resources/blog/how-travelads-features-help-drive-online-bookings Accessed 2026-11-03.
- [399]The Guide to TravelAds Sponsored Listings with Expedia Group Media Solutions (PDF). Expedia Group Media Solutions, 2021-03 (PDF creationDate 2021-03-03). https://go2.advertising.expedia.com/rs/185-EIA-216/images/TravelAds-Best-Practices_GUIDE-Small.pdf Accessed 2026-11-03.
- [400]Top of search, top of mind: A guide to Vrbo Sponsored Listings. Vrbo (Expedia Group), 2026-08-06. https://partner.expediagroup.com/en-us/resources/blog/top-of-search-top-of-mind-guide-to-vrbo-sponsored-listings Accessed 2026-11-03.
- [401]Leap Media Group — Solutions. Leap Media Group (Giant Eagle, Inc.), 2022 (copyright). https://www.leapmediagroup.com/solutions Accessed 2026-11-03.
- [402]Understand the value of new-to-brand shoppers beyond immediate sales (long-term sales launch announcement). Amazon Ads, 2024-10-15. https://advertising.amazon.com/en-us/resources/whats-new/long-term-sales/ Accessed 2026-11-03.
- [403]What a 95% drop in Search investment revealed about incremental growth (article). Walmart Connect; analysis by QBR Media, 2026-10-30 (event 2026-08 (one reduced-spend week; analysis window August 1-31, 2026)). https://www.walmartconnect.com/resources/articles/2026/retail-media-incrementality-search-advertising Accessed 2026-11-03.
- [404]Albertsons Media Collective Pioneers Bleeding Edge Incremental Multi-Touch Attribution (release). Albertsons Media Collective (Albertsons Companies, Inc.) via Business Wire, 2026. https://www.albertsonsmediacollective.com/pr-details/2026/Albertsons-Media-Collective-Pioneers-Bleeding-Edge-Incremental-Multi-Touch-Attribution/default.aspx Accessed 2026-11-03.
- [405]Albertsons Media Collective Launches In-Store Digital Display Network (release). Albertsons Media Collective, 2025-06-16 (event 2025 summer (pilot start stated as future)). https://www.albertsonsmediacollective.com/pr-details/2025/Albertsons-Media-Collective-Launches-In-Store-Digital-Display-Network/default.aspx Accessed 2026-11-03.
- [406]View Direct Connect Campaign performance (84.51° Prism Learning Center). 84.51° LLC, 2026-05-08. https://prism.8451.com/activation/help/content/how-to/direct-connect/view-performance.htm Accessed 2026-11-03. Access: partial extract.
- [407]Amazon Ads API reporting v3: Sponsored Products report types (developer documentation). Amazon Ads. https://advertising.amazon.com/API/docs/en-us/guides/reporting/v3/report-types/sponsored-products Accessed 2026-11-03. Access: partial extract.
- [408]Partner Types and API Access (Sponsored Search APIs, Walmart Connect Developer Portal). Walmart Connect Developer Portal (Advertising Partners - Search), 2026-04-12 (updatedAt). https://developer.walmart.com/advertising-partners-search/docs/partner-types-and-api-access Accessed 2026-11-03.
- [409]Reporting Only Access (Sponsored Search APIs, Walmart Connect Developer Portal; slug 'overview'). Walmart Connect Developer Portal (Advertising Partners - Search), 2026-04-12 (updatedAt). https://developer.walmart.com/advertising-partners-search/docs/overview Accessed 2026-11-03.
- [410]Instacart Ads API (landing: overview, API terms and conditions, required minimum functionality, versioning). Maplebear Inc. d/b/a Instacart, 2022-10-04. https://api.ads.instacart.com/docs/landing Accessed 2026-11-03.
- [411]Advertising on Instacart (Ads Manager documentation index). Maplebear Inc. d/b/a Instacart, 2026-10-29. https://docs.instacart.com/ads_manager Accessed 2026-11-03.
- [412]Instacart Data Hub. Maplebear Inc. d/b/a Instacart, 2026-07-20. https://docs.instacart.com/data_hub Accessed 2026-11-03.
- [413]Configure and view clean rooms (Instacart Data Hub). Maplebear Inc. d/b/a Instacart, 2026-05-14. https://docs.instacart.com/data_hub/cleanroom Accessed 2026-11-03.
- [414]Instacart Debuts Data Hub, a Clean Room Offering for Enhanced Media Performance (syndicated copy). Instacart (Maplebear Inc.), via Perishable News, 2026-01-08. https://perishablenews.com/?p=423316 Accessed 2026-11-03.
- [415]84.51 API Catalog (sign-in page). 84.51° LLC. https://developer.8451.com/ Accessed 2026-11-03. Access: partial extract.
- [416]84.51° Collaborative Cloud (Kroger Precision Marketing solutions page). Kroger Precision Marketing / 84.51° LLC. https://www.krogerprecisionmarketing.com/our-solutions/insights/8451-collaborative-cloud Accessed 2026-11-03.
- [417]Albertsons Media Collective - Retail Media Advertising Solutions (home page). Albertsons Media Collective (Albertsons Companies, Inc.). https://www.albertsonsmediacollective.com/ Accessed 2026-11-03.
- [418]Albertsons Media Collective Enables Advertisers to Measure Performance with Greater Ease and Frictionless Access Via New API Offering and TransUnion Partnership (release). Albertsons Media Collective (Business Wire), 2025-01-07. https://www.albertsonsmediacollective.com/pr-details/2025/Albertsons-Media-Collective-Enables-Advertisers-to-Measure-Performance-with-Greater-Ease-and-Frictionless-Access-Via-New-API-Offering-and-TransUnion-Partnership-2/default.aspx Accessed 2026-11-03.
- [419]Standard Terms and Conditions (Standard Terms and Conditions for Internet Advertising for Media Buys One Year or Less, Version 3.0, as adopted). Albertsons Media Collective. https://www.albertsonsmediacollective.com/standard-terms-and-conditions/default.aspx Accessed 2026-11-03.
- [420]Walmart Connect's Sponsored Search Earns MRC Accreditation (post). Walmart Connect / Walmart Inc., 2025-10-16 (event 2025-09). https://www.walmartconnect.com/walmart-connects-sponsored-search-earns-mrc-accreditation Accessed 2026-11-03.
- [421]Walmart Connect help centre (Advertiser Community). Walmart Connect. https://advertisinghelp.walmart.com/ Accessed 2026-11-03. Access: partial extract.
- [422]Walmart Connect home page. Walmart Connect (Walmart Inc.). https://www.walmartconnect.com/ Accessed 2026-11-03.
- [423]Instacart Partners with Third-Party Verification Leaders, DoubleVerify (DV) and Integral Ad Science (IAS). Maplebear Inc. d/b/a Instacart, 2023-10 (event 2023-10-20). https://company.instacart.com/instacart-ads/instacart-partners-with-third-party-verification-leaders Accessed 2026-11-03. Access: partial extract.
- [424]84.51° Introduces the Collaborative Cloud. Jennifer Strailey, Supermarket News, 2021-11-19 (event 2021-11). https://supermarketnews.com/grocery-technology/84-51-introduces-the-collaborative-cloud Accessed 2026-11-03.
- [425]Albertsons Media Collective - Premium Onsite. Albertsons Media Collective (Albertsons Companies, Inc.). https://www.albertsonsmediacollective.com/offerings/on-site/default.aspx Accessed 2026-11-03.
- [426]Albertsons Media Collective - How we Work together Says it all. (difference page). Albertsons Media Collective (Albertsons Companies, Inc.). https://www.albertsonsmediacollective.com/difference/default.aspx Accessed 2026-11-03.
- [427]Amazon Advertising Agreement (formerly the Sponsored Ads Agreement). Amazon (the applicable 'Amazon Contracting Party'), 2026-07-23 ('Posted: July 23, 2026'). https://advertising.amazon.com/terms Accessed 2026-11-03.
- [428]Amazon Ads reporting overview. Amazon Ads, undated (footer '© 2023'); content refers to Amazon Marketing Stream Version 2 and the reporting API v1, so it post-dates their release. https://advertising.amazon.com/API/docs/en-us/guides/reporting/overview Accessed 2026-11-03.
- [429]Data Freshness in RMP | Commerce Max Help Center (Understanding Data Freshness). Criteo S.A.. https://help.retailmedia.criteo.com/kb/guide/en/understanding-data-freshness-JN4YIAR6by/Steps/1033092 Accessed 2026-11-03.
- [430]About Sales Channels | Commerce Max Help Center. Criteo S.A.. https://help.retailmedia.criteo.com/kb/guide/en/about-sales-channels-yyYvZ0VwLe/Steps/1853109 Accessed 2026-11-03.
- [431]Demand-Side Metrics | Commerce Max Help Center. Criteo S.A.. https://help.retailmedia.criteo.com/kb/guide/en/demand-side-metrics-WUR7bClZmA/Steps/1033012 Accessed 2026-11-03.
- [432]What's New on Commerce Max 2026 (Commerce Max Help Center guide). Criteo S.A.. https://help.retailmedia.criteo.com/kb/guide/en/whats-new-on-commerce-max-2026-xJBAchLLZZ/Steps/5684477 Accessed 2026-11-03.
- [433]Disney Advertising and Walmart Connect to Bring Closed-Loop Attribution To Streaming Advertisers (release). Disney Advertising (The Walt Disney Company), 2024-05-08 (event 2024-05-08 (announcement); 'This proof of concept launches in Q2' (2024)). https://press.disneyadvertising.com/disney-advertising-and-walmart-connect-to-bring-closed-loop-attribution-to-streaming-advertisers Accessed 2026-11-03.
- [434]Roadmap (84.51° Prism Learning Center). 84.51° LLC, 2026-09-09. https://prism.8451.com/activation/help/content/prism-roadmap.htm Accessed 2026-11-03. Access: partial extract.
- [435]Programmatic by Roundel (what-we-offer page). Roundel (Target Brands, Inc.). https://roundel.com/what-we-offer/programmatic/ Accessed 2026-11-03.
- [436]CTV by Roundel (what-we-offer page). Roundel (Target Brands, Inc.). https://roundel.com/what-we-offer/ctv/ Accessed 2026-11-03.
- [437]CMX Social solutions (Attributed sales on Pinterest). CVS Media Exchange (CVS Health). https://www.cvsmediaexchange.com/solutions/social Accessed 2026-11-03.
- [438]CVS Media Exchange (CMX) partners with Pinterest. Mass Market Retailers (reprint of a CVS Media Exchange announcement, dateline Woonsocket, R.I.), 2024-02-29. https://www.massmarketretailers.com/cvs-media-exchange-cmx-partners-with-pinterest/ Accessed 2026-11-03.
- [439]Advertisers Gain Real-Time Customizable Audiences and Always-On Measurement with Instacart and The Trade Desk (PR Newswire release, reprinted). Instacart (Maplebear Inc.) via PR Newswire; reprint host stocktitan.net, 2025-06-10. https://www.stocktitan.net/news/CART/advertisers-gain-real-time-customizable-audiences-and-always-on-0t4ry9nvbd4e.html Accessed 2026-11-03.
- [440]Advertisers Gain Real-Time Customizable Audiences and Always-On Measurement with Instacart and The Trade Desk (issuer release, Internet Archive capture). Instacart (Maplebear Inc.), 2025-06-10. https://web.archive.org/web/20260712150159id_/https://company.instacart.com/pressreleases/customizable-audiences-with-instacart-and-the-trade-desk Accessed 2026-11-03.
- [441]Instacart Debuts Data Hub, a Clean Room Offering for Enhanced Media Performance. Instacart (PR Newswire release, StockTitan copy), 2026-01-06. https://www.stocktitan.net/news/CART/instacart-debuts-data-hub-a-clean-room-offering-for-enhanced-media-ixo3uqg1ehgp.html Accessed 2026-11-03.
- [442]Instacart Ads off-platform partnerships. Instacart (Maplebear Inc.). https://web.archive.org/web/20260301022943/https://www.instacart.com/company/ads/off-platform Accessed 2026-11-03.
- [443]Instacart Announces First End-to-End Retail Media Solution on TikTok. Instacart (Maplebear Inc.; PR Newswire release, StockTitan copy), 2025-10-07. https://www.stocktitan.net/news/CART/instacart-announces-first-end-to-end-retail-media-solution-on-tik-40ktxsb2ce2x.html Accessed 2026-11-03.
- [444]Advertisers Gain Real-Time Customizable Audiences and Always-On Measurement with Instacart and The Trade Desk (release). Instacart (Maplebear Inc.), 2025-06-10. https://company.instacart.com/pressreleases/customizable-audiences-with-instacart-and-the-trade-desk Accessed 2026-11-03. Access: partial extract.
- [445]Expanding where and how advertisers can use Walmart audiences and measurement (article by Khurrum Malik). Walmart Connect (Khurrum Malik, VP of Business & Product Marketing), 2026-10-06. https://www.walmartconnect.com/resources/articles/2026/walmart-audiences-measurement-across-media Accessed 2026-11-03.
- [446]DoorDash Advances Transparency in Incrementality Measurement with AAM Certification (AAM blog post). Alliance for Audited Media, 2026-09-10. https://blog.auditedmedia.com/newsviews/doordash-advances-transparency-in-incrementality-measurement-with-aam-certification Accessed 2026-11-03.
- [447]AAM Assurance Brief: Increasing Confidence in Incrementality Measurement. Alliance for Audited Media, 2026-10-14. https://blog.auditedmedia.com/newsviews/assurance-brief-incrementality Accessed 2026-11-03.
- [448]How we measure clicks and impressions (Instacart Ads Manager docs). Maplebear Inc. d/b/a Instacart (docs.instacart.com), 2026-08-12. https://docs.instacart.com/ads_manager/data_and_analytics/how_we_measure_clicks_and_impressions Accessed 2026-11-03.
- [449]Kroger Precision Marketing home page. Kroger Precision Marketing (84.51°, The Kroger Co.). https://www.krogerprecisionmarketing.com/ Accessed 2026-11-03.
- [450]IAB Europe Retail Media Certification certificate: Moloco (PDF). IAB Europe, 2026-10-07. https://iabeurope.eu/wp-content/uploads/IAB-A4-Retail-Media-Certificate-MOLOCO.pdf Accessed 2026-11-03.
- [451]MRC accreditation letter: Epsilon (GI100125). Media Rating Council, 2025-10-01. https://mediaratingcouncil.org/sites/default/files/Accreditation%20Letters/Epsilon/GI100125_Epsilon%20Accreditation%20Letter.pdf Accessed 2026-11-03.
- [452]Epsilon PeopleCloud Digital Media Solutions Receives Accreditation from the Media Rating Council for Correlated Outcomes (Epsilon pressroom). Epsilon, 2021-05-11. https://epsilon.com/us/about-us/pressroom/epsilon-peoplecloud-digital-media-solutions-receives-accreditation-from-the-media-rating-council-for-correlated-outcomes Accessed 2026-11-03.
- [453]Albertsons Media Collective Brings Sponsored Product Discovery to AI-Powered Conversational Search (release). Albertsons Media Collective, 2026-06-23. https://www.albertsonsmediacollective.com/pr-details/2026/Albertsons-Media-Collective-Brings-Sponsored-Product-Discovery-to-AI-Powered-Conversational-Search/default.aspx Accessed 2026-11-03.
- [454]AAM Assurance List with the filter 'Retail Media Certification' requested by URL. Alliance for Audited Media. https://auditedmedia.com/about/clients?industry_certifications=Retail+Media+Certification Accessed 2026-11-03.
- [455]Marketers Spending Significantly More on Retail Media Networks Despite Underlying Concerns: ANA Report (release). Association of National Advertisers, 2023-01-30. https://www.ana.net/content/show/id/77513 Accessed 2026-11-03.
- [456]IAB/MRC Retail Media Measurement Guidelines - Draft for Public Comment (September 2023). Interactive Advertising Bureau and Media Rating Council, 2023-09 (event 2023-09-13). https://www.iab.com/wp-content/uploads/2023/09/IAB-MRC-Retail-Media-Measurement-Guidelines_For-Public-Comment-1.pdf Accessed 2026-11-03.
- [457]IAB/MRC Retail Media Measurement Guidelines (final, January 2024). Interactive Advertising Bureau and Media Rating Council, 2024-01. https://www.iab.com/wp-content/uploads/2024/01/IAB_Retail_Media_Measurement_Guidelines_January2024.pdf Accessed 2026-11-03.
- [458]IAB Europe's Retail Media Measurement Standards for Europe (knowledge-hub page, V1). IAB Europe, 2024-04-25. https://iabeurope.eu/knowledge_hub/iab-europes-retail-media-measurement-standards-for-europe/ Accessed 2026-11-03.
- [459]Final IAB/MRC Retail Media Measurement Guidelines Available for Download (IAB page). Interactive Advertising Bureau, 2024-02-28. https://www.iab.com/guidelines/final-iab-mrc-retail-media-measurement-guidelines-available-for-download/ Accessed 2026-11-03.
- [460]Retail Media Advanced Measurement and Data Collaboration (IAB page). Interactive Advertising Bureau, 2024-10-31. https://www.iab.com/guidelines/retail-media-advanced-measurement-and-data-collaboration/ Accessed 2026-11-03. Access: partial extract.
- [461]In-Store Retail Media: Definitions and Measurement Standards (IAB page). Interactive Advertising Bureau, 2024-12-03. https://www.iab.com/guidelines/in-store-retail-media/ Accessed 2026-11-03. Access: partial extract.
- [462]Retail Media Internal Management Guidance (ANA research report page). Association of National Advertisers, 2025-09-26. https://www.ana.net/miccontent/show/id/rr-2025-09-retail-media-internal-management Accessed 2026-11-03. Access: abstract only.
- [463]Commerce (including Retail) Media Measurement Standards V2 - Open for Public Comment. IAB Europe, 2025-10-09. https://iabeurope.eu/commerce-including-retail-media-measurement-standards-v2-open-for-public-comment/ Accessed 2026-11-03.
- [464]4 Quick Steps to Becoming Certified in Retail Media Measurement (AAM blog post). Alliance for Audited Media, 2025-10-29. https://blog.auditedmedia.com/newsviews/4-quick-steps-to-retail-media-measurement-certification Accessed 2026-11-03.
- [465]Guidelines for Incremental Measurement in Commerce Media (IAB page). Interactive Advertising Bureau, 2025-11-03. https://www.iab.com/guidelines/guidelines-for-incremental-measurement-in-commerce-media/ Accessed 2026-11-03.
- [466]A Viable Framework for Maturing In-Store Media Measurement (IAB page). Interactive Advertising Bureau, 2025-12-09. https://www.iab.com/guidelines/framework-for-maturing-in-store-media-measurement/ Accessed 2026-11-03. Access: partial extract.
- [467]IAB Europe Releases Updated Commerce (Incl. Retail) Media Standards & Flexi Ad Sizes Guidelines for Retail Media Networks (release). IAB Europe, 2026-01-22. https://iabeurope.eu/iab-europe-releases-updated-commerce-incl-retail-media-standards-flexi-ad-sizes-guidelines/ Accessed 2026-11-03.
- [468]IAB Europe Commerce (Incl. Retail) Measurement Standards V2.1 (PDF, cover May 2026; body 'V2', 'updated as of January 2026'). IAB Europe, 2026-01 (event 2026-01-22). https://iabeurope.eu/wp-content/uploads/IAB-Europes-Commerce-Media-Measurement-Standards-V2.1-May-2026.pdf Accessed 2026-11-03.
- [469]IAB Europe's Commerce (Incl. Retail) Media Measurement Standards V2.1 (knowledge-hub page). IAB Europe, 2026-01-22. https://iabeurope.eu/knowledge_hub/iab-europes-commerce-incl-retail-media-measurement-standards-v2/ Accessed 2026-11-03.
- [470]Alliance for Audited Media Launches Incrementality Certification to Support Transparency and Growth in Commerce Media (release). Alliance for Audited Media, 2026-06-09. https://auditedmedia.com/about/press/aam-launches-incrementality-certification Accessed 2026-11-03.
- [471]Retail Media Measurement Standardization (ANA research report page). Association of National Advertisers, 2026-08-18. https://www.ana.net/miccontent/show/id/rr-2026-08-retail-media-measurement Accessed 2026-11-03. Access: abstract only.
- [472]IAB Europe Travel Media Measurement Standards: An Addendum to the IAB Europe Commerce (incl. Retail) Media Measurement Standards V2 (First Draft For Industry Review - September 2026, PDF). IAB Europe, 2026-09 (event 2026-09-17). https://iabeurope.eu/wp-content/uploads/IAB_Europe_Travel_Media_Standards-PUBLIC-COMMENT-SEP-2026.docx.pdf Accessed 2026-11-03.
- [473]MRC Accreditation Status Quarterly Update, Q3 2026 (July-September 2026). Media Rating Council, 2026-10-16. https://mediaratingcouncil.org/sites/default/files/News/Accreditation-Matters/MRC_Accreditation_Status_Quarterly_Update_Q3_2026.pdf Accessed 2026-11-03.
- [474]Retail Media iROAS Demystified: Understanding Your Brand's Incremental ROAS (full report). Ovative Group and Albertsons Media Collective, 'In partnership with professors from Northwestern University Kellogg School of Management'; authors Sophie Armor, Eric T. Anderson, Kate Bante, Levi Dantzinger, Brett R. Gordon, Derek Nelson, 2026-03. https://ovative.com/wp-content/uploads/2026/03/Retail-Media-iROAS-Demystified.pdf Accessed 2026-11-03.
- [475]Ghost Ads: Improving the Economics of Measuring Online Ad Effectiveness (bibliographic record only). Garrett A. Johnson, Randall A. Lewis, Elmar I. Nubbemeyer, 2017. https://api.semanticscholar.org/graph/v1/paper/DOI:10.1509/jmr.15.0297?fields=title,abstract,authors,year,venue,journal,externalIds,openAccessPdf Accessed 2026-11-03. Access: partial extract.
- [476]A Comparison of Approaches to Advertising Measurement: Evidence from Big Field Experiments at Facebook (working-paper PDF, Kellogg). Brett R. Gordon (Kellogg), Florian Zettelmeyer (Kellogg and NBER), Neha Bhargava (Facebook), Dan Chapsky (Facebook), 2018-04-12 (event 2015 (experiments January to September 2015)). https://www.kellogg.northwestern.edu/faculty/gordon_b/files/fb_comparison.pdf Accessed 2026-11-03.
- [477]Consumer Heterogeneity and Paid Search Effectiveness: A Large Scale Field Experiment (NBER Working Paper 20171, page and PDF). Tom Blake (eBay Research Labs), Chris Nosko (Chicago Booth; eBay Research Labs), Steven Tadelis (Berkeley Haas and NBER), 2014-05 (event 2012 (brand test March 2012; non-brand geo test 60 days, April to July)). https://www.nber.org/papers/w20171 Accessed 2026-11-03.
- [478]The Unfavorable Economics of Measuring the Returns to Advertising (QJE article page). Randall A. Lewis, Justin M. Rao, 2015-07-06. https://academic.oup.com/qje/article/130/4/1941/1916003 Accessed 2026-11-03. Access: partial extract.
- [479]THE UNFAVORABLE ECONOMICS OF MEASURING THE RETURNS TO ADVERTISING (journal-index record with full abstract). Randall A. Lewis and Justin M. Rao (record lists 'Alphabet Inc.; Google Incorporated'), 2015. https://academicnewsletter.sufe.edu.cn/info/345393 Accessed 2026-11-03. Access: abstract only.
- [480]The Unfavorable Economics of Measuring the Returns to Advertising (typeset QJE article, third-party mirror). Randall A. Lewis and Justin M. Rao, 2015-07-06. https://www.gwern.net/doc/economics/advertising/2015-lewis.pdf Accessed 2026-11-03.
- [481]Online ads and offline sales: measuring the effect of retail advertising via a controlled experiment on Yahoo! (QME article page). Randall A. Lewis and David H. Reiley (Google, Inc. at publication; 'This work was completed while both authors were employees at Yahoo! Research.'), 2014-05-28. https://link.springer.com/article/10.1007/s11129-014-9146-6 Accessed 2026-11-03.
- [482]Are consumers averse to sponsored messages? The role of search advertising in information discovery (IDEAS/RePEc record). Navdeep S. Sahni and Charles Zhang (Stanford GSB), 2024-03. https://ideas.repec.org/a/kap/qmktec/v22y2024i1d10.1007_s11129-023-09270-z.html Accessed 2026-11-03. Access: abstract only.
- [483]Not-so-Classical Measurement Errors: A Validation Study of Homescan (NBER Working Paper 14436). Liran Einav, Ephraim Leibtag, Aviv Nevo, 2008-10. https://www.nber.org/papers/w14436 Accessed 2026-11-03. Access: abstract only.
- [484]The Impact of Retail Media on Online Marketplaces: Insights from a Field Experiment (journal index record with abstract). Vibhanshu Abhishek (UC Irvine), Kinshuk Jerath (Columbia), Siddhartha Sharma (Indiana University Kelley), per the record's affiliations, 2025. https://academicnewsletter.sufe.edu.cn/info/406211 Accessed 2026-11-03. Access: abstract only.
- [485]California Consumer Privacy Act (Civil Code 1798.100 et seq.), effective January 1, 2026. California Privacy Protection Agency (statute compilation), 2026-01-01. https://cppa.ca.gov/pdf/20260101_ccpa_statute.pdf Accessed 2026-11-03.
- [486]California Consumer Privacy Act Regulations (Title 11, Division 6, Chapter 1), effective 1/1/2026. California Privacy Protection Agency, 2026-01-01. https://cppa.ca.gov/regulations/pdf/ccpa_statute_eff_20260101.pdf Accessed 2026-11-03.
- [487]Chapter 19.373 RCW — Washington My Health My Data Act (full chapter text). Washington State Legislature (app.leg.wa.gov), 2023 (event 2024-03-31). https://app.leg.wa.gov/RCW/default.aspx?cite=19.373&full=true Accessed 2026-11-03. Access: partial extract.
- [488]Colorado Privacy Act (CPA) — Colorado Attorney General resource page. Colorado Attorney General (coag.gov), 2026 (event 2023-07-01). https://coag.gov/resources/colorado-privacy-act/ Accessed 2026-11-03.
- [489]H.B. No. 4 (88R), enrolled — Texas Data Privacy and Security Act (Business & Commerce Code chapter 541). Texas Legislature Online (capitol.texas.gov), 2023 (event 2024-07-01). https://capitol.texas.gov/tlodocs/88R/billtext/html/HB00004F.htm Accessed 2026-11-03.
- [490]Data Brokers — Texas Secretary of State. Texas Secretary of State, 2026. https://www.sos.texas.gov/statdoc/data-brokers.shtml Accessed 2026-11-03.
- [491]Minnesota Session Laws 2024, Chapter 121 (H.F. 4757), Article 5 — Minnesota Consumer Data Privacy Act. Office of the Revisor of Statutes, Minnesota Legislature, 2024 (event 2025-07-31). https://www.revisor.mn.gov/laws/2024/0/Session+Law/Chapter/121/ Accessed 2026-11-03.
- [492]Minn. Stat. § 325M.16 (Responsibilities of controllers) and § 325M.17 (Requirements for small businesses). Office of the Revisor of Statutes, Minnesota, 2025. https://www.revisor.mn.gov/statutes/cite/325M.16 Accessed 2026-11-03.
- [493]Chapter 455 (Senate Bill 541), 2024 Laws of Maryland — Maryland Online Data Privacy Act of 2024. Maryland General Assembly (mgaleg.maryland.gov), 2024-05-09 (event 2025-10-01). https://mgaleg.maryland.gov/2024RS/Chapters_noln/CH_455_sb0541e.pdf Accessed 2026-11-03.
- [494]45 CFR 164.508 — Uses and disclosures for which an authorization is required (HIPAA Privacy Rule), 2024 annual edition. U.S. Government Publishing Office (govinfo.gov), Department of Health and Human Services, 2024-10-01. https://www.govinfo.gov/content/pkg/CFR-2024-title45-vol2/xml/CFR-2024-title45-vol2-sec164-508.xml Accessed 2026-11-03.
- [495]FTC Enforcement Action to Bar GoodRx from Sharing Consumers' Sensitive Health Info for Advertising. Federal Trade Commission, 2023-02-01. https://www.ftc.gov/news-events/news/press-releases/2023/02/ftc-enforcement-action-bar-goodrx-sharing-consumers-sensitive-health-info-advertising Accessed 2026-11-03.
- [496]FTC Surveillance Pricing Study Indicates Wide Range of Personal Data Used to Set Individualized Consumer Prices. Federal Trade Commission, 2025-01-17. https://www.ftc.gov/news-events/news/press-releases/2025/01/ftc-surveillance-pricing-study-indicates-wide-range-personal-data-used-set-individualized-consumer Accessed 2026-11-03.
- [497]Order to File A Special Report - 6(b) Surveillance Pricing Intermediaries (template). Federal Trade Commission (Chair Khan; Commissioners Slaughter, Bedoya, Holyoak, Ferguson), 2024-07-23. https://www.ftc.gov/system/files/ftc_gov/pdf/sp6b_order_surv_pricing.pdf Accessed 2026-11-03.
- [498]Federal Trade Commission's Proposed Enforcement Policy Statement Regarding Personalized Pricing. Federal Trade Commission, 2026-08-19. https://www.ftc.gov/system/files/ftc_gov/pdf/p034101-ftc-enforcement-policy-statement-re-personalized-pricing-proposed-for-public-comment.pdf Accessed 2026-11-03.
- [499]Agentic shopping is here. What does that mean for advertisers?. Amazon Ads, 2026-06-11. https://advertising.amazon.com/en-us/library/news/agentic-shopping-advertising Accessed 2026-11-03.
- [500]Our approach to advertising and expanding access to ChatGPT. OpenAI (Fidji Simo), 2026-01-16. https://openai.com/index/our-approach-to-advertising-and-expanding-access/ Accessed 2026-11-03.
- [501]Testing ads in ChatGPT. OpenAI, 2026-02-09. https://openai.com/index/testing-ads-in-chatgpt/ Accessed 2026-11-03.
- [502]New ways to buy ChatGPT ads. OpenAI, 2026-05-05. https://openai.com/index/new-ways-to-buy-chatgpt-ads/ Accessed 2026-11-03.
- [503]A milestone in expanding access to AI. OpenAI, 2026-08-31. https://openai.com/index/expanding-access-to-ai-with-chatgpt-ads/ Accessed 2026-11-03.
- [504]ChatGPT Ads expands to 26 new countries as over 100,000 advertisers have used the platform. OpenAI, 2026-10-21. https://openai.com/index/chatgpt-ads-expands-26-new-countries/ Accessed 2026-11-03.
- [505]Building advertising for the way people use AI. OpenAI, 2026-10-05 (event 2026-10-05 (announcement; visual-format test planned 'later this month')). https://openai.com/index/new-chatgpt-ads-format-and-measurement/ Accessed 2026-11-03.
- [506]Reimagining advertising with AI. OpenAI, 2026-09-16. https://openai.com/index/reimagining-advertising-with-ai/ Accessed 2026-11-03.
- [507]Amazon Ads announces new integration to support advertising in ChatGPT. Amazon Ads, 2026-09-10. https://advertising.amazon.com/en-us/library/news/amazon-ads-chat-gpt-advertising-integration Accessed 2026-11-03.
- [508]About ads in AI-powered experiences in Google Search. Google Ads Help. https://support.google.com/google-ads/answer/18826097?hl=en Accessed 2026-11-03. Access: partial extract.
- [509]Ads in AI Mode expand to new markets.. Google (Shashi Thakur, VP and GM, Google Search Ads and Ads on Google Experiences), 2026-10-28. https://blog.google/intl/en-africa/products/ads-in-ai-mode-expand-to-new-markets/ Accessed 2026-11-03.
- [510]New tech and tools for retailers to succeed in an agentic shopping era. Google (Vidhya Srinivasan), 2026-01-11. https://blog.google/products/ads-commerce/agentic-commerce-ai-tools-protocol-retailers-platforms/ Accessed 2026-11-03.
- [511]Kroger Precision Marketing Launches Shopping Assistant PLAs (Product Listing Ads). Kroger Precision Marketing, 2026-08-03. https://www.krogerprecisionmarketing.com/knowledge-hub/blog/kroger-precision-marketing-launches-shopping-assistant-plas-product-listing-ads/ Accessed 2026-11-03.
- [512]Win across all three eras of the web. Microsoft Advertising (Tim Frank, CVP Microsoft AI Monetization), 2026-04-21. https://about.ads.microsoft.com/en/blog/post/april-2026/win-across-all-three-eras-of-the-web Accessed 2026-11-03.
- [513]73% higher CTRs: Why advertisers need to pay attention to conversational AI. Microsoft Advertising (Kevin Klein and Paul Longo), 2025-08-06. https://about.ads.microsoft.com/en/blog/post/august-2025/73-higher-ctrs-why-advertisers-need-to-pay-attention-to-conversational-ai Accessed 2026-11-03.
- [514]Conversations that Convert: Copilot Checkout and Brand Agents. Microsoft Advertising (Nayna Sheth and Ravi Yada), 2026-01-08. https://about.ads.microsoft.com/en/blog/post/january-2026/conversations-that-convert-copilot-checkout-and-brand-agents Accessed 2026-11-03.
- [515]agentic-commerce-protocol repository README (main). Agentic Commerce Protocol (OpenAI and Stripe), 2026-07-18. https://github.com/agentic-commerce-protocol/agentic-commerce-protocol Accessed 2026-11-03.
- [516]ACP changelog, Version 2026-04-17. Agentic Commerce Protocol (OpenAI and Stripe), 2026-04-17. https://github.com/agentic-commerce-protocol/agentic-commerce-protocol/blob/main/changelog/2026-04-17.md Accessed 2026-11-03.
- [517]Products (File Upload) product feed specification. OpenAI. https://developers.openai.com/commerce/specs/file-upload/products.md Accessed 2026-11-03.
- [518]Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol. OpenAI, 2025-09-29. https://openai.com/index/buy-it-in-chatgpt/ Accessed 2026-11-03.
- [519]ACP releases page. Agentic Commerce Protocol (OpenAI and Stripe). https://github.com/agentic-commerce-protocol/agentic-commerce-protocol/releases Accessed 2026-11-03.
- [520]Powering Product Discovery in ChatGPT. OpenAI, 2026-03-24. https://openai.com/index/powering-product-discovery-in-chatgpt/ Accessed 2026-11-03.
- [521]Instacart and OpenAI partner on AI shopping experiences. OpenAI, 2025-12-08. https://openai.com/index/instacart-partnership/ Accessed 2026-11-03.
- [522]Universal Commerce Protocol releases. Universal-Commerce-Protocol (GitHub organisation), 2026-08-25. https://github.com/Universal-Commerce-Protocol/ucp/releases Accessed 2026-11-03.
- [523]Universal Commerce Protocol repository README. Universal-Commerce-Protocol (GitHub organisation). https://github.com/Universal-Commerce-Protocol/ucp Accessed 2026-11-03.
- [524]How we're helping retailers thrive with new Universal Commerce Protocol features and AI tools on Google. Google (The Keyword), 2026-05-20. https://blog.google/products-and-platforms/products/shopping/shopping-updates-google-marketing-live/ Accessed 2026-11-03.
- [525]AP2 releases. google-agentic-commerce (GitHub organisation), 2026-04-28. https://github.com/google-agentic-commerce/AP2/releases Accessed 2026-11-03.
- [526]AP2 repository README. google-agentic-commerce (GitHub organisation). https://github.com/google-agentic-commerce/AP2 Accessed 2026-11-03.
- [527]AdCP releases. AgenticAdvertising.org / adcontextprotocol, 2026-10-31. https://github.com/adcontextprotocol/adcp/releases Accessed 2026-11-03.
- [528]AdCP documentation introduction. AgenticAdvertising.org / adcontextprotocol. https://docs.adcontextprotocol.org/docs/intro Accessed 2026-11-03.
- [529]AdCP docs: Commerce Media (3.2.8). AgenticAdvertising.org / adcontextprotocol. https://docs.adcontextprotocol.org/dist/docs/3.2.8/media-buy/commerce-media.md Accessed 2026-11-03.
- [530]AdCP docs: Media Channel Taxonomy (3.2.8). AgenticAdvertising.org / adcontextprotocol. https://docs.adcontextprotocol.org/dist/docs/3.2.8/reference/media-channel-taxonomy.md Accessed 2026-11-03.
- [531]AdCP compliance specialism: sales-retail-media (index.yaml, main). AgenticAdvertising.org / adcontextprotocol. https://github.com/adcontextprotocol/adcp/blob/main/static/compliance/source/specialisms/sales-retail-media/index.yaml Accessed 2026-11-03.
- [532]AdCP specialism enum (specialism.json, main). AgenticAdvertising.org / adcontextprotocol. https://github.com/adcontextprotocol/adcp/blob/main/static/schemas/source/enums/specialism.json Accessed 2026-11-03.
- [533]AdCP sponsored-placement-type enum (main). AgenticAdvertising.org / adcontextprotocol. https://github.com/adcontextprotocol/adcp/blob/main/static/schemas/source/enums/sponsored-placement-type.json Accessed 2026-11-03.
- [534]AdCP canonical format: sponsored_placement.json (main). AgenticAdvertising.org / adcontextprotocol. https://github.com/adcontextprotocol/adcp/blob/main/static/schemas/source/formats/canonical/sponsored_placement.json Accessed 2026-11-03.
- [535]AdCP docs: Sponsored placement adapter contracts (main). AgenticAdvertising.org / adcontextprotocol. https://github.com/adcontextprotocol/adcp/blob/main/docs/creative/sponsored-placement-adapter-contracts.mdx Accessed 2026-11-03.
- [536]AdCP repository files at release tag v3.2.8 (sales-retail-media index.yaml, specialism.json, sponsored-placement-type.json, canonical sponsored_placement.json, sponsored-placement-adapter-contracts.mdx). AgenticAdvertising.org / adcontextprotocol (GitHub API), 2026-10-28. https://github.com/adcontextprotocol/adcp/tree/v3.2.8 Accessed 2026-11-03.
- [537]IAB Tech Lab Introduces AAMP 3.0 to Standardize the RFP-to-Buy Process for Agentic Advertising. IAB Tech Lab, 2026-09-22. https://iabtechlab.com/press-releases/iab-tech-lab-introduces-aamp-3-0-with-openproposal/ Accessed 2026-11-03.
- [538]IABTechLab/OpenProposal repository (README and release v3.0-draft-1). IAB Tech Lab, 2026-09-24. https://github.com/IABTechLab/OpenProposal Accessed 2026-11-03.
- [539]GitHub code search for 'retail' and 'commerce' in IAB Tech Lab AAMP repositories. GitHub code search (via gh api). https://github.com/IABTechLab Accessed 2026-11-03. Access: partial extract.
- [540]IAB Tech Lab Releases AAMP 2.3, Bringing Enterprise-Grade Infrastructure and Privacy Diligence to Agentic Advertising. IAB Tech Lab, 2026-07-30. https://iabtechlab.com/press-releases/iab-tech-lab-releases-aamp-2-3-bringing-enterprise-grade-infrastructure-and-privacy-diligence-to-agentic-advertising/ Accessed 2026-11-03.
- [541]AAMP: Agentic Advertising Management Protocols (standards page). IAB Tech Lab, 2026-10-27. https://iabtechlab.com/standards/aamp-agentic-advertising-management-protocols/ Accessed 2026-11-03.
- [542]Expanding access to our measurement solutions (article). Walmart Connect (Walmart Inc.), 2025-04-04. https://www.walmartconnect.com/resources/articles/2025/partner-network-expanding-measurement-solutions Accessed 2026-11-03.
- [543]Instacart Ads Attribution Window (Perpetua help article). Perpetua (author David Scalise), 2022-07-22. https://help.perpetua.io/en/articles/4480611-instacart-ads-attribution-window Accessed 2026-11-03.
- [544]Why Retail Media iROAS Might Be Misleading You (landing page for the report). Ovative Group. https://ovative.com/impact/expert-insights/retail-media-iroas-demystified/ Accessed 2026-11-03.
- [545]Transforming Retail Media Through Standardization: Albertsons Media Collective Introduces Best Practices Framework (white paper). Albertsons Media Collective, 2023-06. https://s29.q4cdn.com/239956855/files/doc_downloads/2023/FINAL_TheCollective_Whitepaper.pdf Accessed 2026-11-03.
- [546]Case Study: Converting Reach into Revenue with In-Store Activation (Mondelēz, Sargento Cheese Bakes). Albertsons Media Collective, 2026-01-06. https://www.albertsonsmediacollective.com/news/blog/blog-details/2026/Case-Study-Converting-Reach-into-Revenue-with-In-Store-Activation-/default.aspx Accessed 2026-11-03.
- [547]Big Screens, Measurable Impact (Primo Water digital out-of-home case study). Albertsons Media Collective, 2025-09-24 (event campaign April to May 2025). https://www.albertsonsmediacollective.com/news/blog/blog-details/2025/-Big-Screens-Measurable-Impact/default.aspx Accessed 2026-11-03.
- [548]Albertsons Media Collective - Privacy Policy. Albertsons Media Collective (Albertsons Companies, Inc.), 2026-01-12. https://www.albertsonsmediacollective.com/privacy-policy/default.aspx?ItemId=ed6e59f7-c220-448e-97ad-0b6325b3e1c7 Accessed 2026-11-03.
- [549]Sponsored ads reporting FAQ (version 3). Amazon Ads, undated (footer '© 2023'); mentions a 30 March 2023 deprecation in the future tense. https://advertising.amazon.com/API/docs/en-us/guides/reporting/v3/faq Accessed 2026-11-03.
- [550]IAB/MRC Retail Media Measurement Guidelines (January 2024), served by MRC. IAB Retail Media Measurement Working Group with the Media Rating Council, 2024-01. https://mediaratingcouncil.org/sites/default/files/Standards/IAB_Retail_Media_Measurement_Guidelines_January2024.pdf Accessed 2026-11-03.
- [551]FTC, States Sue Amazon Over Secret Ad Surcharge Scheme (press release). Federal Trade Commission, 2026-08-31. https://www.ftc.gov/news-events/news/press-releases/2026/08/ftc-states-sue-amazon-over-secret-ad-surcharge-scheme Accessed 2026-11-03.
- [552]FTC Sues Amazon for Illegally Maintaining Monopoly Power (press release). Federal Trade Commission, 2023-09-26. https://www.ftc.gov/news-events/news/press-releases/2023/09/ftc-sues-amazon-illegally-maintaining-monopoly-power Accessed 2026-11-03.
- [553]Amazon.com, Inc. (Amazon eCommerce) case page, FTC matter nos. 1910129 / 1910130. Federal Trade Commission, 2024-10-31. https://www.ftc.gov/legal-library/browse/cases-proceedings/1910129-1910130-amazoncom-inc-amazon-ecommerce Accessed 2026-11-03.
- [554]Frequently Asked Questions (cardlytics.com). Cardlytics, Inc., 2026 (footer copyright; no page date). https://cardlytics.com/faqs Accessed 2026-11-03.
- [555]Glossary (Cardlytics Ads Manager User Guide). Cardlytics, Inc., 2026-02-19. https://platform.cardlytics.com/advertisers/docs/ads-manager-glossary Accessed 2026-11-03.
- [556]How to Access (Cardlytics Ads Manager User Guide). Cardlytics, Inc., 2026-09-29. https://platform.cardlytics.com/advertisers/docs/ads-manager-how-to-access Accessed 2026-11-03.
- [557]Overview (Cardlytics Advertiser API Integration Guide). Cardlytics, Inc., 2026-02-13. https://platform.cardlytics.com/advertisers/docs/api-overview Accessed 2026-11-03.
- [558]Master Advertiser Terms & Conditions (CPS). Cardlytics, Inc., 2026 (footer copyright; no version or date). https://www.cardlytics.com/advtscscps Accessed 2026-11-03.
- [559]Cardlytics, Inc. Form 8-K filed 3 June 2026 (Items 3.01 and 3.03: Nasdaq notice; reverse stock split). Cardlytics, Inc. (SEC EDGAR), 2026-06-03. https://www.sec.gov/Archives/edgar/data/1666071/000162828026040351/cdlx-20260603.htm Accessed 2026-11-03.
- [560]Cardlytics, Inc. Form 8-K, date of report 4 September 2026 (filed 11 September 2026; Item 1.01, settlement with Amit Jain). Cardlytics, Inc. (SEC EDGAR), 2026-09-11 (event 2026-09-04). https://www.sec.gov/Archives/edgar/data/1666071/000162828026061459/cdlx-20260904.htm Accessed 2026-11-03.
- [561]Circana meta-study 2026 (Ibotta resource hub). Ibotta, Inc. (reporting a Circana study), 2026-06-19. https://ipn.ibotta.com/resource-hub/circana-meta-study-2026 Accessed 2026-11-03.
- [562]Understanding IRI Household-Based and Store-Based Scanner Data (Technical Bulletin 1942). Muth, Sweitzer, Brown, Capogrossi, Karns, Levin, Okrent, Siegel and Zhen (RTI International and USDA Economic Research Service), 2016-04. https://ers.usda.gov/sites/default/files/_laserfiche/publications/47633/57105_tb-1942.pdf?v=11117 Accessed 2026-11-03.
- [563]Circana Expands Largest U.S. Receipt Panel to 200,000 Static Panelists (circana.com post). Circana, LLC, 2025-02-06. https://www.circana.com/post/circana-expands-largest-u-s-receipt-panel-to-200-000-static-panelists Accessed 2026-11-03.
- [564]NCS Measurement (NCSolutions portfolio page on circana.com: 'NCS Sales Effect'). Circana, LLC, undated (footer '© 2026 Circana' in site template). https://www.circana.com/solutions/ncs-measurement Accessed 2026-11-03.
- [565]The Circana Group Global Privacy Policy and Notice. Circana, LLC (Delaware) and its subsidiaries and affiliates, 2025-10-23. https://www.circana.com/the-circana-group-global-privacy-policy-and-notice Accessed 2026-11-03.
- [566]How Identity Signals Support Attribution (Commerce Yield Help Center). Criteo, 2026-08-11. https://help.cyield.criteo.com/kb/guide/en/how-identity-signals-support-attribution-pgC8GqHo1X/Steps/5877097 Accessed 2026-11-03.
- [567]Criteo S.A. Annual Report on Form 10-K for the fiscal year ended 31 December 2025 (main document crto-20251231.htm). Criteo S.A. (SEC EDGAR), 2026-02-26. https://www.sec.gov/Archives/edgar/data/1576427/000157642726000014/crto-20251231.htm Accessed 2026-11-03.
- [568]CRITEO REPORTS STRONG THIRD QUARTER 2025 RESULTS. Criteo S.A., 2025-10-29. https://criteo.investorroom.com/2025-10-29-CRITEO-REPORTS-STRONG-THIRD-QUARTER-2025-RESULTS Accessed 2026-11-03.
- [569]Commerce Yield Help Center (home page and article index). Criteo. https://help.cyield.criteo.com/kb/en/ Accessed 2026-11-03. Access: partial extract.
- [570]Criteo Methodologies for Ad Measurement (Commerce Yield Help Center). Criteo, 2026-08-11 (event 2026-07-13). https://help.cyield.criteo.com/kb/guide/en/criteo-methodologies-for-ad-measurement-T7vVSGy9kc/Steps/5789733 Accessed 2026-11-03.
- [571]Understanding incrementality: The key to measuring a campaign’s true impact (Criteo blog). Criteo (Carmen Chan, per the author box), 2024-09-18. https://www.criteo.com/blog/understanding-incrementality/ Accessed 2026-11-03.
- [572]Smarter ways to prove ad impact: Criteo's flexible approach to incrementality (blog). Criteo S.A. (author Elizabeth Kim), 2025-07-22. https://www.criteo.com/blog/criteos-flexible-approach-to-incrementality/ Accessed 2026-11-03.
- [573]Criteo Collaborates with Microsoft Advertising to Drive Retail Media Growth. Criteo S.A., 2024-07-11. https://criteo.investorroom.com/2024-07-11-Criteo-Collaborates-with-Microsoft-Advertising-to-Drive-Retail-Media-Growth Accessed 2026-11-03.
- [574]Reshaping the retail media ecosystem with scale, simplicity, and innovation. Lynne Kjolso, Microsoft Advertising, 2024-07-11. https://about.ads.microsoft.com/en/blog/post/july-2024/reshaping-the-retail-media-ecosystem-with-scale-simplicity-and-innovation Accessed 2026-11-03.
- [575]Change to impression reporting for sponsored products (Instacart Ads Manager docs). Maplebear Inc. d/b/a Instacart (docs.instacart.com), 2026-08-12 (event 2022-01-18 (change); reporting affected from 2022-01-01). https://docs.instacart.com/ads_manager/sponsored_products/change_to_impression_reporting_for_sponsored_products Accessed 2026-11-03.
- [576]New Circana Case Studies Underscore Instacart's Incremental Sales. Maplebear Inc. d/b/a Instacart, 2024-05-02. https://web.archive.org/web/20251224075703/https://www.instacart.com/company/updates/new-circana-case-studies-underscore-instacarts-incremental-sales Accessed 2026-11-03.
- [577]Instacart Receives Expanded MRC Accreditation for Carrot Ads, Bringing Verified Ad Metrics to More Than 240 Ecommerce Sites (PR Newswire release, reprinted). Maplebear Inc. d/b/a Instacart (via stocktitan.net reprint), 2025-11-06. https://www.stocktitan.net/news/CART/instacart-receives-expanded-mrc-accreditation-for-carrot-ads-ih55iehgac69.html Accessed 2026-11-03.
- [578]Associated Food Stores Expands Instacart E-Commerce and Retail Media Solutions to Owned Banners and Member Retailers (PR Newswire release, reprinted). Maplebear Inc. d/b/a Instacart and Associated Food Stores (via stocktitan.net reprint), 2025-11-11. https://www.stocktitan.net/news/CART/associated-food-stores-expands-instacart-e-commerce-and-retail-media-c1xd5x2vodde.html Accessed 2026-11-03.
- [579]Carrot Ads: retail media, built to scale. Maplebear Inc. d/b/a Instacart (Carrot Ads). https://company.instacart.com/enterprise-platform/retail-media/carrot-ads Accessed 2026-11-03. Access: partial extract.
- [580]Kroger Precision Marketing (KPM) Managed Service FAQs (84.51° Prism Learning Center). 84.51° LLC, 2026-10-20. https://prism.8451.com/activation/help/content/resources-faqs/kpm-faqs.htm Accessed 2026-11-03. Access: partial extract.
- [581]What's new in 84.51° Prism — Release Notes. 84.51° LLC, 2026 (event entries January to September 2026). https://prism.8451.com/activation/help/content/release-notes/whats-new.htm Accessed 2026-11-03. Access: partial extract.
- [582]Kroger Outlines Plan to Build America's Favorite Grocer and Introduces New Long-Term Financial Targets (8-K exhibit 99.1). The Kroger Co. (SEC EDGAR), 2026-10-20 (event 2026-10-20 (Investor Update, New York)). https://www.sec.gov/Archives/edgar/data/56873/000110465926117103/tm2627753d1_ex99-1.htm Accessed 2026-11-03. Access: partial extract.
- [583]Sponsored Products: Cost-Per-Click Search Ads (solutions page). Walmart Connect (Walmart Inc.), undated (footer '© 2026 Walmart Inc.'). https://www.walmartconnect.com/solutions/sponsored-products Accessed 2026-11-03.
- [584]Reporting FAQs (Sponsored Search APIs, Walmart Connect Developer Portal). Walmart Connect Developer Portal (Advertising Partners - Search), 2026-07-10 (updatedAt). https://developer.walmart.com/advertising-partners-search/docs/reporting-faqs Accessed 2026-11-03.
- [585]Walmart Customer Privacy Notice (Online and In-Store). Walmart Inc., 2026-08-20. https://corporate.walmart.com/privacy-security/walmart-privacy-notice Accessed 2026-11-03.
- [586]Privacy Policy - Walmart Connect. Walmart Connect (Walmart Inc.), 2026-03-17. https://www.walmartconnect.com/privacy-policy Accessed 2026-11-03.
- [587]Product Release Notes (Sponsored Search APIs changelog, Walmart Connect Developer Portal). Walmart Connect Developer Portal (Advertising Partners - Search), undated page; entries April 2025 to September 2026 (event 2025-04 to 2026-09). https://developer.walmart.com/advertising-partners-search/changelog/product-release-notes Accessed 2026-11-03.
- [588]Evolving Walmart Connect's identity graph to power stronger signals (article). Walmart Connect (Walmart Inc.), 2026-04-03. https://www.walmartconnect.com/resources/articles/2026/evolving-walmart-connects-identity-graph-to-power-stronger-signals Accessed 2026-11-03.
- [589]Order to File a Special Report (shell), FTC Matter No. P246202. Federal Trade Commission (Chairman Andrew N. Ferguson; Commissioner Mark R. Meador), 2026-10-28. https://www.ftc.gov/system/files/ftc_gov/pdf/PP6b-Special-Order-Shell.pdf Accessed 2026-11-03.
- [590]Albertsons Media Collective Debuts Preliminary Standardization Framework to Improve Transparency in Retail Media and Remove Points of Friction for Advertisers (release). Albertsons Media Collective, 2023-06-22. https://www.albertsonsmediacollective.com/pr-details/2023/Albertsons-Media-Collective-Debuts-Preliminary-Standardization-Framework-to-Improve-Transparency-in-Retail-Media-and-Remove-Points-of-Friction-for-Advertisers/default.aspx Accessed 2026-11-03.
- [591]Retail Media iROAS Demystified: Executive Summary. Ovative Group and Albertsons Media Collective, 2026-03. https://ovative.com/wp-content/uploads/2026/03/Retail-Media-iROAS-Demystified-Executive-Summary.pdf Accessed 2026-11-03.
- [592]Albertsons Companies Unveils New Retail Media Network (release). Albertsons Companies, Inc. (Business Wire 20211111006049), 2021-11-11 (event 2021-11-11 (announcement); brand campaigns to begin 2022-02-27). https://www.albertsonsmediacollective.com/pr-details/2021/Albertsons-Companies-Unveils-New-Retail-Media-Network/default.aspx Accessed 2026-11-03.
- [593]Albertsons Companies, Inc. Form 10-Q for the quarter ended 12 September 2026. Albertsons Companies, Inc. (SEC EDGAR), 2026-10-20 (event 2026-09-12 (period of report)). https://www.sec.gov/Archives/edgar/data/1646972/000164697226000065/aci-20260912.htm Accessed 2026-11-03.
- [594]Albertsons Companies, Inc. Reports Fourth Quarter and Full Year Results (Form 8-K exhibit 99.1). Albertsons Companies, Inc. (SEC EDGAR), 2026-04-14. https://www.sec.gov/Archives/edgar/data/1646972/000164697226000028/aciq425earningsrelease.htm Accessed 2026-11-03.
- [595]Albertsons Companies, Inc. Reports Second Quarter Fiscal 2026 Results (Form 8-K exhibit 99.1). Albertsons Companies, Inc. (SEC EDGAR), 2026-10-13. https://www.sec.gov/Archives/edgar/data/1646972/000164697226000059/aciq226earningsrelease.htm Accessed 2026-11-03.
- [596]Amazon Ads Multi-Touch Attribution (arXiv abstract page). Randall Lewis, Florian Zettelmeyer, Brett R. Gordon, Cristobal Garib, Johannes Hermle, Mike Perry, Henrique Romero, German Schnaidt, 2025-08-11. https://arxiv.org/abs/2508.08209 Accessed 2026-11-03.
- [597]Hourly Metrics: Amazon Marketing Stream and Budget Consumption (Intentwise help article). Intentwise, Inc.. https://help.intentwise.com/amazon-marketing-stream-explained Accessed 2026-11-03.
- [598]Amazon.com Privacy Notice. Amazon.com, Inc. and affiliates, 2026-06-30. https://www.amazon.com/gp/help/customer/display.html?nodeId=GX7NJQ4ZB8MHFRNJ Accessed 2026-11-03.
- [599]Interest-Based Ads — Amazon Customer Service. Amazon.com, Inc., 2026. https://www.amazon.com/gp/help/customer/display.html?nodeId=GLVB9XDF9M8MU7UZ Accessed 2026-11-03.
- [600]Consumer Health Data Privacy Disclosure (Amazon). Amazon.com, Inc. and affiliates, 2026-06-30. https://www.amazon.com/gp/help/customer/display.html?nodeId=TnACMrGVghHocjL8KB Accessed 2026-11-03.
- [601]Additional State-Specific Privacy Disclosures (Amazon). Amazon.com, Inc. and affiliates, 2026-06-30. https://www.amazon.com/gp/help/customer/display.html?nodeId=GC5HB5DVMU5Y8CJ2 Accessed 2026-11-03.
- [602]Performance Reporting, REST API (Cardlytics Advertiser API Integration Guide). Cardlytics, Inc., 2026-02-13. https://platform.cardlytics.com/advertisers/docs/api-performance-reporting Accessed 2026-11-03.
- [603]Campaign Performance (Cardlytics Ads Manager User Guide). Cardlytics, Inc., 2026-08-13. https://platform.cardlytics.com/advertisers/docs/ads-manager-campaign-performance Accessed 2026-11-03.
- [604]Reporting Hub (Cardlytics Ads Manager User Guide). Cardlytics, Inc., 2026-10-05. https://platform.cardlytics.com/advertisers/docs/ads-manager-reporting-hub Accessed 2026-11-03.
- [605]Redemption Feed (Cardlytics Advertiser API Integration Guide). Cardlytics, Inc., 2026-02-13. https://platform.cardlytics.com/advertisers/docs/api-redemption-feed Accessed 2026-11-03.
- [606]Dunkin': Driving consistent incremental returns (Cardlytics case study). Cardlytics, Inc.. https://www.cardlytics.com/case-studies/dunkin-driving-consistent-incremental-returns Accessed 2026-11-03. Access: partial extract.
- [607]How We Power Merchant Offers (Cardlytics publisher integration documentation). Cardlytics, Inc., 2025-08-28. https://platform.cardlytics.com/publisher-integrations/docs/learn-more-about-cardlytics-rewards-platform Accessed 2026-11-03.
- [608]Get Started (Cardlytics Ads Manager User Guide). Cardlytics, Inc., 2026-01-22. https://platform.cardlytics.com/advertisers/docs/ads-manager-user-guide Accessed 2026-11-03.
- [609]Cardlytics Privacy Policy. Cardlytics, Inc., 2025-12-19. https://www.cardlytics.com/privacy-policy Accessed 2026-11-03.
- [610]Privacy Notice for UK Bank Customers. Cardlytics UK Ltd., 2024-03-13. https://cardlytics.com/privacy-notice-bank-customers Accessed 2026-11-03.
- [611]4.5x category benchmark sales lift: Validating incremental growth with Circana’s Household Lift™ (case study). Circana, LLC, undated. https://www.circana.com/case-studies/4-5x-category-benchmark-sales-lift Accessed 2026-11-03.
- [612]Circana Market Lift - Enabling brands to test marketing hypothesis before committing dollars (case study, ZA locale). Circana, LLC, 2024-07-17. https://www.circana.com/za/case-studies/circana-market-lift-enabling-brands-to-test-marketing-hypothesis-before-committing-dollars Accessed 2026-11-03.
- [613]Unlock Growth with Advanced Media Measurement (Mars case study, ZA locale). Circana, LLC, 2024-10-25. https://www.circana.com/za/case-studies/unlock-growth-with-advanced-media-measurement Accessed 2026-11-03.
- [614]Liquid Mix: Self-Serve AI Marketing Mix Modeling | Circana. Circana, LLC, undated. https://www.circana.com/solutions/liquid-mix Accessed 2026-11-03.
- [615]Consumer Panel Behavior Data, Insights, & Analysis Solutions | Circana (Complete Consumer). Circana, LLC, undated. https://www.circana.com/solutions/complete-consumer Accessed 2026-11-03.
- [616]Data-Driven Audience Targeting Solutions for Retailers & Brands | Circana (Complete Audiences). Circana, LLC, undated. https://www.circana.com/solutions/complete-audiences Accessed 2026-11-03.
- [617]Danone x LiveLift case study (Ibotta resource hub). Ibotta, Inc., 2026-09-21. https://ipn.ibotta.com/resource-hub/danone-livelift-case-study Accessed 2026-11-03.
- [618]New Circana Case Studies Underscore Instacart's Incremental Sales. Maplebear Inc. dba Instacart, unknown (not read; a search snippet cites a February 2024 Circana analysis). https://company.instacart.com/updates/new-circana-case-studies-underscore-instacarts-incremental-sales Accessed 2026-11-03. Access: partial extract.
- [619]New Circana studies show Instacart display ads drive in-store sales. Maplebear Inc. dba Instacart, unknown (search snippet: August 2026). https://company.instacart.com/instacart-ads/new-circana-studies-show-instacart-display-ads-drive-in-store-sales Accessed 2026-11-03. Access: partial extract.
- [620]Understand Your Target Audiences with Retail Media | Circana (Retail Media solution page). Circana, LLC, undated. https://www.circana.com/solutions/retail-media Accessed 2026-11-03.
- [621]Optimize Marketing Efforts & Drive ROI with Marketing Mix Modeling | Circana (Marketing Mix solution page). Circana, LLC, undated. https://www.circana.com/solutions/marketing-mix Accessed 2026-11-03.
- [622]Demand Side Analytics (DSP) (Criteo Retail Media API docs). Criteo. https://developers.criteo.com/retail-media/docs/analytics Accessed 2026-11-03.
- [623]Supply Side Analytics (SSP) (Criteo Retail Media API docs). Criteo. https://developers.criteo.com/retail-media/docs/supply-side-reporting-ssp Accessed 2026-11-03.
- [624]Criteo Launches Commerce Max DSP into General Availability and Announces Next-Gen Retailer Monetization Solution Suite. Criteo S.A., 2023-09-12. https://criteo.investorroom.com/2023-09-12-Criteo-Launches-Commerce-Max-DSP-into-General-Availability-and-Announces-Next-Gen-Retailer-Monetization-Solution-Suite Accessed 2026-11-03.
- [625]Criteo investor relations: news releases list (100 most recent, 14 March 2023 to 29 October 2026). Criteo S.A., 2026-10-29. https://criteo.investorroom.com/releases?l=100&o=0 Accessed 2026-11-03. Access: partial extract.
- [626]Get Started (Criteo Retail Media API docs). Criteo. https://developers.criteo.com/retail-media/docs Accessed 2026-11-03.
- [627]API Versioning Policy. Criteo. https://developers.criteo.com/criteo-apis/docs/versioning-policy.md Accessed 2026-11-03.
- [628]Albertsons Media Collective Selects Criteo to Power its Retail Media Ecosystem. Criteo S.A., 2024-01-09. https://criteo.investorroom.com/2024-01-09-Albertsons-Media-Collective-Selects-Criteo-to-Power-its-Retail-Media-Ecosystem Accessed 2026-11-03.
- [629]Dollar General Selects Criteo to Enhance its Retail Media Offering. Criteo S.A., 2024-05-31. https://criteo.investorroom.com/2024-05-31-Dollar-General-Selects-Criteo-to-Enhance-its-Retail-Media-Offering Accessed 2026-11-03.
- [630]Belk Teams Up with Criteo to Connect Brands with Retailer's Loyal Customer Base. Criteo S.A., 2024-10-08. https://criteo.investorroom.com/2024-10-08-Belk-Teams-Up-with-Criteo-to-Connect-Brands-with-Retailers-Loyal-Customer-Base Accessed 2026-11-03.
- [631]CRITEO REPORTS STRONG SECOND QUARTER 2025 RESULTS. Criteo S.A., 2025-07-30. https://criteo.investorroom.com/2025-07-30-CRITEO-REPORTS-STRONG-SECOND-QUARTER-2025-RESULTS Accessed 2026-11-03.
- [632]Criteo: Markets supported (Intentwise help article). Intentwise, Inc.. https://help.intentwise.com/list-of-markets-and-retailers-supported-on-criteo Accessed 2026-11-03.
- [633]Criteo Secures Its First MRC Accreditation for Retail Media Measurement. Criteo, 2024-03-28. https://www.criteo.com/news/press-releases/2024/03/criteo-secures-its-first-mrc-accreditation-for-retail-media-measurement/ Accessed 2026-11-03.
- [634]Privacy FAQs | Commerce Max Help Center. Criteo, 2024-03-27. https://help.retailmedia.criteo.com/kb/guide/en/privacy-faqs-gcztz7NE5f/Steps/1535799 Accessed 2026-11-03.
- [635]Criteo Privacy Policy. Criteo (Criteo Corp. for the US site; data protection office in Paris), 2026-07-17. https://www.criteo.com/privacy/ Accessed 2026-11-03.
- [636]Criteo Joins OpenAI Advertising Pilot in ChatGPT (press release PDF). Criteo S.A., 2026-03-02. https://www.criteo.com/wp-content/uploads/2026/02/Criteo-Joins-OpenAI-Advertising-Pilot-in-ChatGPT.pdf Accessed 2026-11-03.
- [637]Data and analytics (Instacart Ads Manager). Maplebear Inc. d/b/a Instacart, 2026-08-12. https://docs.instacart.com/ads_manager/data_and_analytics/overview Accessed 2026-11-03.
- [638]Instacart ads terms and conditions (Instacart Ads Manager docs). Maplebear Inc. d/b/a Instacart (docs.instacart.com), 2026-05-14 (event 2026-05-06 (terms 'UPDATED MAY 6, 2026')). https://docs.instacart.com/ads_manager/instacart_ads_terms_and_conditions Accessed 2026-11-03.
- [639]DoubleVerify Unveils New Partnership To Authenticate Instacart Ads. DoubleVerify, 2023-10-12. https://doubleverify.com/newsroom/doubleverify-unveils-new-partnership-to-authenticate-instacart-ads/ Accessed 2026-11-03.
- [640]Maplebear Inc. Form 10-K for fiscal 2025, main document cart-20251231.htm (accession 0001579091-26-000018). Maplebear Inc. / SEC EDGAR, 2026-02-26 (event 2025-12-31). https://www.sec.gov/Archives/edgar/data/1579091/000157909126000018/cart-20251231.htm Accessed 2026-11-03. Access: partial extract.
- [641]How to Connect Your Instacart Advertising Account to Perpetua (Perpetua help article). Perpetua (author Priyanka Ramjutun), 2024-05-01. https://help.perpetua.io/en/articles/4490259-how-to-connect-your-instacart-advertising-account-to-perpetua Accessed 2026-11-03.
- [642]Instacart Ads: Get Started on Intentwise (Intentwise help article). Intentwise, Inc.. https://help.intentwise.com/instacart-ads-explained Accessed 2026-11-03.
- [643]Instacart Advertising: Unified Reporting & Insights with Skai. Skai (Kenshoo, Ltd.). https://skai.io/capabilities/instacart/ Accessed 2026-11-03.
- [644]Ads Manager changelog (Instacart docs). Maplebear Inc. d/b/a Instacart (docs.instacart.com), 2026-10-29. https://docs.instacart.com/ads_manager/changelog Accessed 2026-11-03.
- [645]Instacart Privacy Policy. Maplebear Inc. d/b/a Instacart and Maplebear Canada Inc., 2026-01-15. https://www.instacart.com/privacy Accessed 2026-11-03.
- [646]84.51° and Vestcom Announce Expanded Collaboration Enabling CPGs to Extend Brand Equity Media at the Shelf Edge and Attribute In-Store Sales. Vestcom (release on vestcom.com, with 84.51° boilerplate), 2022-09-16. https://vestcom.com/resources/press-releases/84-51-and-vestcom-announce-expanded-collaboration-enabling-cpgs-to-extend-brand-equity-media-at-the-shelf-edge-and-attribute-in-store-sales/ Accessed 2026-11-03. Access: partial extract.
- [647]Onsite Search & Display. Kroger Precision Marketing. https://www.krogerprecisionmarketing.com/our-solutions/retail-media/search-display Accessed 2026-11-03.
- [648]Retail Media Solutions. Kroger Precision Marketing. https://www.krogerprecisionmarketing.com/our-solutions/retail-media/ Accessed 2026-11-03.
- [649]Kroger Reports Second Quarter 2026 Results and Updates Guidance for 2026 (8-K exhibit 99.1). The Kroger Co. (SEC EDGAR), 2026-09-11 (event quarter ended 2026-08-15). https://www.sec.gov/Archives/edgar/data/0000056873/000110465926106890/tm2625060d1_ex99-1.htm Accessed 2026-11-03. Access: partial extract.
- [650]EDGAR full-text search: "alternative profit" in The Kroger Co. Form 10-K filings. US Securities and Exchange Commission (EDGAR full-text search) (event 2026-11-03 (query)). https://efts.sec.gov/LATEST/search-index?q=%22alternative%20profit%22&ciks=0000056873&forms=10-K Accessed 2026-11-03.
- [651]EDGAR full-text search: "retail media" in The Kroger Co. Form 8-K filings, 1 January 2025 to 3 November 2026. US Securities and Exchange Commission (EDGAR full-text search) (event 2026-11-03 (query)). https://efts.sec.gov/LATEST/search-index?q=%22retail%20media%22&ciks=0000056873&forms=8-K&dateRange=custom&startdt=2025-01-01&enddt=2026-11-03 Accessed 2026-11-03.
- [652]KPM Direct Connect reporting (ad policy, 84.51° Prism Learning Center). 84.51° LLC, 2026-05-08. https://prism.8451.com/activation/help/content/resources-faqs/ad-policies/dc-reporting.htm Accessed 2026-11-03. Access: partial extract.
- [653]84.51° Prism Learning Center (home). 84.51° LLC. https://prism.8451.com/activation/help/content/learning-center.htm Accessed 2026-11-03. Access: partial extract.
- [654]Walmart Connect's Sponsored Search Earns MRC Accreditation (article, resources path). Walmart Connect (Khurrum Malik), 2025-10-16 (event 2025-09 (accreditation letter 26 September 2025)). https://www.walmartconnect.com/resources/articles/2025/walmart-connects-sponsored-search-earns-mrc-accreditation Accessed 2026-11-03.
- [655]Helping advertisers make quicker, data-driven decisions for Sponsored Search campaigns (article). Walmart Connect (Walmart Inc.), 2024-07-31. https://www.walmartconnect.com/resources/articles/2024/sponsored-search-powered-by-data Accessed 2026-11-03.
- [656]Acquire New Customers: Retail Media Strategies with Walmart Connect (goals page). Walmart Connect (Walmart Inc.), undated (footer 2026). https://www.walmartconnect.com/goals/acquire-new-customers Accessed 2026-11-03.
- [657]Walmart Connect Partner Network: Agencies & Tech Solutions (partners page). Walmart Connect (Walmart Inc.), undated (footer 2026; footnote period calendar 2025). https://www.walmartconnect.com/partners Accessed 2026-11-03.
- [658]Walmart: Setup Walmart Connect Ads API Access (Intentwise help article). Intentwise, Inc.. https://help.intentwise.com/process-to-setup-walmart-connect-ads-api-access Accessed 2026-11-03.
- [659]Walmart Inc. Form 8-K exhibit 99.1, second-quarter fiscal 2027 earnings release (earningsreleasefy27q2.htm; accession 0000104169-26-000145). Walmart Inc., 2026-08-20 (event quarter ended 31 July 2026). https://www.sec.gov/Archives/edgar/data/104169/000010416926000145/earningsreleasefy27q2.htm Accessed 2026-11-03.
- [660]MRC Accreditation Status Quarterly Update, Q3 2025 (July-September 2025). Media Rating Council, 2025-10-16. https://www.mediaratingcouncil.org/sites/default/files/News/MRC%20Accreditation%20Status%20Quarterly%20Update_Q3%2025_10.15.25%20.pdf Accessed 2026-11-03.
- [661]Advertising Partners (Search): Walmart Connect Developer Portal, project home and navigation. Walmart Connect Developer Portal, undated (pages carry updatedAt stamps to 2026-10-21). https://developer.walmart.com/advertising-partners-search Accessed 2026-11-03. Access: partial extract.
- [662]Advertising Partners (Display): Walmart Connect Developer Portal, project home and navigation. Walmart Connect Developer Portal, undated (pages carry updatedAt stamps to 2026-10-29). https://developer.walmart.com/advertising-partners Accessed 2026-11-03. Access: partial extract.
- [663]Why do Attributed Sales Differ Between Perpetua's Dashboard vs. Walmart's Ad Console? (Perpetua help article). Perpetua (author Claire Lee), 2026-05-13. https://help.perpetua.io/en/articles/5581185-why-do-attributed-sales-differ-between-perpetua-s-dashboard-vs-walmart-s-ad-console Accessed 2026-11-03.
- [664]PayPal Ads Brings Purchase Attribution to Connected TV Through PayPal Curated Ads. PayPal Holdings, Inc. (newsroom), 2026-04-22. https://newsroom.paypal-corp.com/2026-04-22-PayPal-Ads-Brings-Purchase-Attribution-to-Connected-TV-Through-PayPal-Curated-Ads Accessed 2026-11-03.
- [665]Fetch capabilities (business.fetch.com/capabilities). Fetch Rewards LLC, 2026. https://business.fetch.com/capabilities Accessed 2026-11-03.
- [666]Conversion tracking (Rokt Ads integration guides). Rokt, 2026-10-02. https://docs.rokt.com/integration-guides/ads/conversion-tracking/ Accessed 2026-11-03.
- [667]Upside Privacy Policy. Upside Services, Inc., 2026-09-02. https://www.upside.com/data-and-security/privacy-policy Accessed 2026-11-03. Access: partial extract.
- [668]Digital Out-Of-Home (DOOH) Measurement Guide (PDF). IAB, 2025-07 (event 2025-07-08 (publication per the location paper's LOC-A-F23)). https://assets.contentstack.io/v3/assets/bltbeaed4aed52c223a/blt494e044b82a9dcb0/68cb0a53da7537f88855f47c/iab-dooh-measurement-guide.pdf Accessed 2026-11-03.
- [669]Roundel home page. Roundel (Target Brands, Inc.). https://roundel.com/ Accessed 2026-11-03.
- [670]Grocery TV Media Kit 2026 - Website (PDF). Grocery TV, 2026. https://4675579.fs1.hubspotusercontent-na1.net/hubfs/4675579/Grocery%20TV%20Media%20Kit%202026%20-%20Website.pdf Accessed 2026-11-03.
- [671]Stingray Advertising Marks 100th Matched Panel Test, Showcasing an Average 20.3% Sales Lift with In-Store Audio Ads (release). Stingray Group Inc. (Stingray Advertising), 2024-06-20. https://corporate.stingray.com/?p=10475 Accessed 2026-11-03.
- [672]IAB Europe Commerce Media Measurement Standards V2 (body: 'updated as of January 2026'; file created 19 May 2026). IAB Europe, 2026-01. https://iabeurope.eu/wp-content/uploads/IAB-Europes-Commerce-Media-Measurement-Standards-V2-Jan-2026.pdf Accessed 2026-11-03.
- [673]Guidelines for Incremental Measurement in Commerce Media (PDF). IAB and IAB Europe, 2025-11 (event 2025-11-03). https://www.iab.com/wp-content/uploads/2026/01/IAB_and_IAB_Europe_Guidelines_Incremental_Measurement_Commerce_Media_November_2025.pdf Accessed 2026-11-03.
- [674]EDGAR filing list: Amazon.com, Inc. 10-Q filings. SEC EDGAR. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=1018724&type=10-Q&dateb=&owner=include&count=5 Accessed 2026-11-03.
- [675]Uber Announces Results for Fourth Quarter and Full Year 2025 (Form 8-K exhibit 99.1, accession 0001543151-26-000011). Uber Technologies, Inc. / SEC EDGAR, 2026-02-04 (event 2025-12-31). https://www.sec.gov/Archives/edgar/data/1543151/000154315126000011/uberq425earningspressrelea.htm Accessed 2026-11-03.
- [676]NIQ News Center (listing). NIQ (Nielsen Consumer LLC / NIQ Global Intelligence plc). https://nielseniq.com/global/en/news-center/ Accessed 2026-11-03. Access: partial extract.
- [677]Macy's partners with Amazon for retail ads ahead of holiday season. Retail Dive (Aaron Baar), 2025-08-27 (event 2025-08). https://www.retaildive.com/news/macys-partners-amazon-retail-ads-holidays/758719/ Accessed 2026-11-03.
- [678]Who Owns Numerator. Numerator, LLC. https://www.numerator.com/who-owns-numerator/ Accessed 2026-11-03.
- [679]Investor Relations (The Kantar Group). Kantar. https://www.kantar.com/Campaigns/Investor-Relations Accessed 2026-11-03.
- [680]Announcement of the Numerator | Kantar H1 2026 results. Kantar Global Holdings S.a r.l., 2026-08-20. https://www.kantar.com/investor-relations/financial-announcement/announcement-of-the-numerator-kantar-h1-2026-results Accessed 2026-11-03.
- [681]Press release list (Q4 feed GetPressReleaseList) served on albertsonsmediacollective.com. Albertsons Companies, Inc. / Albertsons Media Collective (Q4 Inc. platform). https://www.albertsonsmediacollective.com/feed/PressRelease.svc/GetPressReleaseList?LanguageId=1&bodyType=0&pressReleaseDateFilter=3&categoryId=1cb807d2-208f-4bc3-9133-6a9ad45ac3b0&pageSize=-1&pageNumber=0&tagList=&includeTags=true&year=-1&excludeSelection=1 Accessed 2026-11-03.
- [682]EDGAR filing list, Cardlytics, Inc., form 10-Q. SEC EDGAR. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=1666071&type=10-Q&dateb=&owner=include&count=40 Accessed 2026-11-03.
- [683]EDGAR filing list, Cardlytics, Inc., form 8-K. SEC EDGAR. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=1666071&type=8-K&dateb=&owner=include&count=40 Accessed 2026-11-03.
- [684]IRI and NPD Complete Merger, Creating a Leading Global Technology, Analytics and Data Provider (circana.com post). Circana (then IRI and The NPD Group), 2022-08-01. https://www.circana.com/post/news-press-releases-iri-and-npd-complete-merger Accessed 2026-11-03.
- [685]Circana (Hellman & Friedman portfolio page). Hellman & Friedman LLC, 2021-12-15 (date shown on the page). https://hf.com/portfolio/circana/ Accessed 2026-11-03.
- [686]EDGAR company filing lists for Criteo S.A. (CIK 1576427): all filings (40 most recent) and 10-Q filings. SEC EDGAR. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=1576427&type=&dateb=&owner=include&count=40 Accessed 2026-11-03.
- [687]EDGAR company filing list, Maplebear Inc. (CIK 1579091), all forms, 20 most recent. SEC EDGAR. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=1579091&type=&dateb=&owner=include&count=20 Accessed 2026-11-03.
- [688]Walmart Inc. Annual Report to Shareholders, fiscal 2026 (Form ARS; wmtfy26annualreport-final.pdf). Walmart Inc., 2026-04-23. https://www.sec.gov/Archives/edgar/data/104169/000010416926000091/wmtfy26annualreport-final.pdf Accessed 2026-11-03. Access: partial extract.
- [689]Walmart Inc. definitive proxy statement 2026 (DEF 14A; wmt-20260423.htm). Walmart Inc., 2026-04-23. https://www.sec.gov/Archives/edgar/data/104169/000119312526173673/wmt-20260423.htm Accessed 2026-11-03. Access: partial extract.
- [690]Ibotta Performance Network: proof page. Ibotta, Inc., 2026. https://ipn.ibotta.com/proof Accessed 2026-11-03.
- [691]Welcome to Skai (blog post by Margo Kahnrose). Skai, 2021-06-08. https://skai.io/blog/welcome-to-skai/ Accessed 2026-11-03.
- [692]CRITEO REPORTS FOURTH QUARTER 2025 RESULTS. Criteo S.A., 2026-02-11. https://criteo.investorroom.com/2026-02-11-CRITEO-REPORTS-FOURTH-QUARTER-2025-RESULTS Accessed 2026-11-03.
- [693]Uber Eats ads lifted in-store sales 5% among exposed eaters, Uber says. PPC Land (Luis Rijo), 2026-10-05. https://ppc.land/uber-eats-ads-lifted-in-store-sales-5-among-exposed-eaters-uber-says/ Accessed 2026-11-03.
- [694]Results | Walgreens Advertising Group. Walgreens Advertising Group, 2024 (footer). https://www.walgreensadvertisinggroup.com/home/results.html Accessed 2026-11-03.
- [695]Solutions | UB Media | Ulta Beauty. Ulta Beauty, Inc., 2026 (footer). https://www.ulta.com/company/ubmedia/solutions Accessed 2026-11-03.
- [696]Self Service Retail Media | DGMN. DG Media Network, LLC, 2026. https://www.dgmedianetwork.com/self-service Accessed 2026-11-03.
- [697]Reckitt achieves 16.5% incremental sales lift from Instacart Ads. Maplebear Inc. dba Instacart, unknown (not read). https://www.instacart.com/company/instacart-ads/reckitt-achieves-16-5-incremental-sales-lift-from-instacart-ads Accessed 2026-11-03. Access: partial extract.
- [698]Marketing API Overview (eBay Developers Program). eBay Inc., undated (copyright 1999-2026). https://developer.ebay.com/api-docs/sell/marketing/static/overview.html Accessed 2026-11-03.
- [699]Promoted Listings help (ebay.com/help id=4658). eBay Inc.. https://www.ebay.com/help/selling/listings/promoted-listings?id=4658 Accessed 2026-11-03. Access: partial extract.
- [700]Orange Apron Media — In the News (blog listing, page 1). Orange Apron Media, 2026. https://orangeapronmedia.com/blog/ Accessed 2026-11-03. Access: partial extract.
- [701]Chewy Ads FAQ (guessed path). Chewy, Inc.. https://discover.ads.chewy.com/faq Accessed 2026-11-03. Access: partial extract.
- [702]Instacart Docs (index). Maplebear Inc. d/b/a Instacart. https://docs.instacart.com/ Accessed 2026-11-03.
- [703]CVS Media Exchange — home. CVS Media Exchange, 2026. https://www.cvsmediaexchange.com/ Accessed 2026-11-03.
- [704]About | Chase Media Solutions. JPMorgan Chase Bank, N.A. (Chase Media Solutions), 2026. https://www.chase.com/mediasolutions/about Accessed 2026-11-03.
- [705]KPMG and Fetch: key takeaways for brand partners (business.fetch.com trends-and-insights post). Fetch Rewards LLC, 2023-06-08. https://business.fetch.com/trends-and-insights/kpmg-fetch-3-key-takeaways-brand-partners Accessed 2026-11-03.
- [706]Measurement methodology (Upside for business). Upside Services, Inc., 2026. https://www.upside.com/business/how-it-works/measurement-methodology Accessed 2026-11-03.
- [707]Grocery methodology overview for contracts (Upside). Upside Services, Inc., 2023-07. https://upside.com/data-and-security/grocery-contract-methodology Accessed 2026-11-03.
- [708]Reporting overview (Rokt Ads platform guides). Rokt, 2026-08-04. https://docs.rokt.com/platform-guides/rokt-ads/reporting/overview/ Accessed 2026-11-03.
- [709]Conversion reporting (Rokt Ads platform guides). Rokt. https://docs.rokt.com/platform-guides/rokt-ads/reporting/conversion-reporting/ Accessed 2026-11-03.
- [710]Amazon Ads Advanced Tools Center: AMC overview (developer documentation). Amazon Ads. https://advertising.amazon.com/API/docs/en-us/amc/overview Accessed 2026-11-03. Access: partial extract.
- [711]Overview of DoorDash Ads (Ads Support, DoorDash Help Center). DoorDash, Inc.. https://help.doordash.com/en-us/advertising/article/doordash-ads-overview Accessed 2026-11-03.
- [712]2025 innovations that are launching an even more impactful 2026.. Best Buy Ads, undated (footer ©2026). https://www.bestbuyads.com/insights/2025-innovations-that-are-launching-an-even-more-impactful-2026 Accessed 2026-11-03.
- [713]SecureMeasure — client success story (PDF). Numerator, 2023-06-30 (PDF creation date; uploaded 2024-05 path). https://www.numerator.com/wp-content/uploads/2024/05/SecureMeasure-Non-Branded-Case-Study.pdf Accessed 2026-11-03.
- [714]Measurement Panel | Numerator (Omnipanels). Numerator, LLC. https://www.numerator.com/omnipanels/measurement-panel/ Accessed 2026-11-03.
- [715]Amazon Ads launches Amazon Retail Ad Service. Amazon Ads, 2025-01-09. https://advertising.amazon.com/en-us/library/news/retail-ad-service-announcement Accessed 2026-11-03.
- [716]Walmart Announces Expanded Vision and New Name for Its Media Business (release). Walmart Inc., 2021-01-28. https://corporate.walmart.com/news/2021/01/28/walmart-announces-expanded-vision-and-new-name-for-its-media-business Accessed 2026-11-03.
- [717]Orange Apron Media Unveils New Partnerships, Onsite Offerings, and Certification Program at Third Annual InFronts (release). The Home Depot (Orange Apron Media), 2026-04-22. https://corporate.homedepot.com/news/company/orange-apron-media-unveils-new-partnerships-infronts-2026 Accessed 2026-11-03.
- [718]STRATACACHE press-release URL for the Albertsons in-store digital display network (served content: 'In the News: Path to Purchase Institute Covers STRATACACHE and Albertsons Launch', 28 January 2026). STRATACACHE, 2026-01-28 (event 2026-01 (network 'officially debuted')). https://stratacache.com/en/news/press-releases/stratacache-and-albertsons-media-collective-team-up-to-launch-new-in-store-digital-display-network/ Accessed 2026-11-03. Access: partial extract.
- [719]In-Store Retail Media Networks (STRATACACHE solutions page). STRATACACHE. https://stratacache.com/en/solutions/in-store-retail-media-networks/ Accessed 2026-11-03.
- [720]Grocery TV for brands (page). Grocery TV. https://grocerytv.com/brands Accessed 2026-11-03.
- [721]Grocery TV home page. Grocery TV. https://grocerytv.com/ Accessed 2026-11-03.
- [722]KPM quick reference guide (84.51° Prism Learning Center). 84.51° LLC, 2026-05-08. https://prism.8451.com/activation/help/content/resources-faqs/kpm-guide.htm Accessed 2026-11-03. Access: partial extract.
- [723]Walmart and Roku Debut First-of-its-Kind Partnership to Bring Commerce to TV Ads (joint Business Wire release, reprinted). Walmart Inc. and Roku, Inc. (Business Wire); reprint host 01net.it, 2022-06-16. https://www.01net.it/walmart-and-roku-debut-first-of-its-kind-partnership-to-bring-commerce-to-tv-ads/ Accessed 2026-11-03.
- [724]How the Walmart Connect Partner Lab Is Building Omnichannel Connections (post by Sylvia Yam). Walmart Inc. (Sylvia Yam, Senior Director of Strategy and Business Development, Walmart Connect), 2023-06-15 (event 2022-11-22 to 2023-02-10 (Roku pilot per footnotes 2 and 3); 2023-06-15 (expansion announced)). https://corporate.walmart.com/news/2023/06/15/how-the-walmart-connect-partner-lab-is-building-omnichannel-connections Accessed 2026-11-03.
- [725]Expanding access to Walmart first-party data across leading platforms (article by Ryan Mayward). Walmart Connect (Ryan Mayward, GM & SVP), 2026-05-28. https://www.walmartconnect.com/resources/articles/2026/expanding-access-to-walmart-first-party-data-across-leading-platforms Accessed 2026-11-03.
- [726]Samba TV newsroom (listing). Samba TV. https://samba.com/newsroom Accessed 2026-11-03. Access: partial extract.
- [727]Catalina and Samba TV to Provide Unprecedented TV and Digital Measurement Services to Georgia-Pacific (release). Catalina and Samba TV, 2020-08-12. https://samba.com/newsroom/catalina-and-samba-tv-to-provide-unprecedented-tv-and-digital-measurement-services-to-georgia-pacific Accessed 2026-11-03.
- [728]Samba TV Announces Exclusive Integration of Affinity Solutions' Deterministic Purchase Media Metrics to Usher in the Next Generation of Advertising Measurement and Activation (release). Samba TV, 2023-10-16. https://samba.com/newsroom/samba-tv-announces-exclusive-integration-of-affinity-solutions-deterministic-purchase-media-metrics-to-usher-in-the-next-generation-of-advertising-measurement-and-activation Accessed 2026-11-03.
- [729]More than Half of Retail Media Networks Advertisers Are Reallocating Budgets To RMNs, According to New IAB Study (release announcing the draft guidelines). Interactive Advertising Bureau, 2023-09-13. https://www.iab.com/news/iab-and-mrc-releases-retail-media-measurement-guidelines/ Accessed 2026-11-03.
- [730]Measurement That the Market Can Trust: Reflections from the IAB Europe Retailer Leaders Council. IAB Europe, 2026-03-04 (event 2026-03-03). https://iabeurope.eu/?p=62311 Accessed 2026-11-03. Access: partial extract.
- [731]AAM Incrementality Certification Compliance Guide. Alliance for Audited Media. https://support.auditedmedia.com/aam-incrementality-certification-compliance-guide Accessed 2026-11-03.
- [732]New Retail Media Measurement Report Highlights Need for Industry Standardization and Assurance (AAM blog post on the ANA report). Alliance for Audited Media, 2026-09-02. https://blog.auditedmedia.com/newsviews/new-retail-media-measurement-report Accessed 2026-11-03.
- [733]Retail Media Networks: A Comprehensive Research Report (report page). Association of National Advertisers, 2023-01-30. https://www.ana.net/miccontent/show/id/rr-2023-01-ana-retail-media-networks-partnership Accessed 2026-11-03. Access: partial extract.
- [734]Retail Media Networks: Optimism Tempered with Caution (report page). Association of National Advertisers, 2024-07-24. https://www.ana.net/miccontent/show/id/rr-2024-07-retail-media-networks-optimism-caution Accessed 2026-11-03. Access: partial extract.
- [735]How to navigate Retail Media success with your Joint Business Plan (blog post). Nate Pinkston, head of growth, Microsoft Retail Media (Microsoft Advertising), 2024-08-08. https://about.ads.microsoft.com/en/blog/post/august-2024/how-to-navigate-retail-media-success-with-your-joint-business-plan Accessed 2026-11-03.
- [736]Joint Business Planning (JBP) Between Retailers and Brands (DSI & Microsoft Advertising report page). Digital Shelf Institute with Microsoft Advertising. https://www.digitalshelfinstitute.org/report-dsi-microsoft-advertising-joint-business-planning Accessed 2026-11-03. Access: partial extract.
- [737]"We've got to call their bluff": Amidst retail media spend scrutiny, advertisers pull out of negotiations. Kimeko McCoy, Digiday (republished on Modern Retail), 2025-03-25. https://www.modernretail.co/marketing/weve-got-to-call-their-bluff-amidst-retail-media-spend-scrutiny-advertisers-pull-out-of-negotiations/ Accessed 2026-11-03.
- [738]Marketing Briefing: Marketers push for more flexible retail media deals amidst economic uncertainty. Kimeko McCoy, Digiday, 2025-04-01. https://digiday.com/marketing/marketing-briefing-marketers-push-for-more-flexible-retail-media-deals-amidst-economic-uncertainty/ Accessed 2026-11-03.
- [739]About Attribution (Commerce Max Help Center). Criteo. https://help.retailmedia.criteo.com/kb/guide/en/about-attribution-sZH1iCyq5L/Steps/1034909 Accessed 2026-11-03.
- [740]Retail Marketing Platform: Media Platform for Advertisers (Epsilon Retail Media, US 'For advertisers' page). Epsilon (Epsilon Data Management, LLC). https://www.epsilon.com/us/products-and-services/retail-media-network/advertisers Accessed 2026-11-03.
- [741]Retail media (Epsilon Retail Media, US product page). Epsilon Data Management, LLC. https://www.epsilon.com/us/products-and-services/retail-media-network Accessed 2026-11-03.
- [742]Gopuff Goes Off-Site; Becomes First Retail Media Network to Tap into the Full Capabilities of CitrusAd, powered by Epsilon. Epsilon Marketing, 2022-08-18. https://www.epsilon.com/us/about-us/pressroom/gopuff-goes-off-site-becomes-first-retail-media-network-to-tap-into-the-full-capabilities-of-citrusad-powered-by-epsilon Accessed 2026-11-03.
- [743]Kevel Attribution FAQ. Kevel, 2026-08-18. https://dev.kevel.com/docs/attribution-faq.md Accessed 2026-11-03.
- [744]Kevel Attribution. Kevel, 2026-10-02. https://dev.kevel.com/docs/attribution-overview.md Accessed 2026-11-03.
- [745]Retail Media Guide (Kevel docs). Kevel, 2026-10-02. https://dev.kevel.com/docs/retail-media.md Accessed 2026-11-03.
- [746]Retail - Koddi | Ad Technology for Driving Revenue and Growth. Koddi Inc.. https://koddi.com/retail/ Accessed 2026-11-03.
- [747]Attribution: Overview (Topsort knowledge base). Topsort Technologies, Inc., 2025-11-18. https://docs.topsort.com/en/knowledge-base/ad-server/attribution.md Accessed 2026-11-03.
- [748]Report events (Topsort API reference). Topsort Technologies, Inc.. https://docs.topsort.com/en/api-reference/events/ Accessed 2026-11-03.
- [749]Direct Attribution (Topsort knowledge base). Topsort, 2025-11-18. https://docs.topsort.com/en/knowledge-base/ad-server/attribution/direct-attribution Accessed 2026-11-03.
- [750]Topsort Developer Documentation (landing). Topsort Technologies, Inc.. https://docs.topsort.com/ Accessed 2026-11-03.
- [751]How RMNs Use Pentaleap. Pentaleap GmbH. https://pentaleap.com/how-rmns-use-pentaleap Accessed 2026-11-03.
- [752]Unify Organic and Sponsored Ranking, Connect to Demand (Pentaleap home). Pentaleap GmbH. https://www.pentaleap.com/ Accessed 2026-11-03.
- [753]Pentaleap is Selected by Macy's, Inc. to Power Its Next-Generation Retail Media Network. Pentaleap GmbH, 2026-01-22. https://pentaleap.com/press-release/pentaleap-is-selected-by-macys-inc-to-power-its-next-generation-retail-media-network Accessed 2026-11-03.
- [754]Attribution Methodology (Moloco Commerce Media docs). Moloco, Inc., 2026-04-17. https://mcm-docs.moloco.com/docs/attribution-methodology Accessed 2026-11-03.
- [755]Solution components (Moloco Commerce Media docs). Moloco, 2025-02-07. https://mcm-docs.moloco.com/v1.0/docs/solution-components Accessed 2026-11-03.
- [756]Retail Media Platform for Enterprise Retailers (Vantage platform page). Vantage. https://www.gotvantage.com/platform Accessed 2026-11-03.
- [757]Incrementality Measurement for Retail Media | Osmos. Osmos, Inc.. https://www.osmos.ai/platform/measurement/incrementality Accessed 2026-11-03.
- [758]Regional Retail Media | Grow Your Business with Rippl Network (for retailers). Rippl (contact [email protected]; no legal entity on page). https://www.ripplnetwork.com/for-regional-retailers Accessed 2026-11-03.
- [759]Rippl Network | Retail Media for Advertisers & Agencies. Rippl (footer email [email protected]; no legal entity). https://www.ripplnetwork.com/for-advertisers-and-agencies Accessed 2026-11-03.
- [760]STRATACACHE home and news listing. STRATACACHE. https://stratacache.com/en/ Accessed 2026-11-03.
- [761]Zitcha | Retail Media Management & Margin Optimization Platform (platform page). Zitcha. https://www.zitcha.com/retail-media-activation-platform Accessed 2026-11-03.
- [762]Transform Your Retail Media Operations with a Comprehensive OMS Solution. ADvendio (by Alex), 2023-06-07. https://www.advendio.com/transform-retail-media-operations-comprehensive-oms-solution Accessed 2026-11-03.
- [763]Post Click Attribution (Streaem developer docs). Streaem (developers.streaem.com; VML not named on the page), 2025-05-22. https://developers.streaem.com/docs/post-click-attribution Accessed 2026-11-03.
- [764]Post View Attribution (Streaem developer docs). Streaem (developers.streaem.com), 2025-05-22. https://developers.streaem.com/docs/post-view-attribution Accessed 2026-11-03.
- [765]GMV and ROAS (Kevel docs). Kevel, 2026-08-01. https://dev.kevel.com/docs/gmv-and-roas.md Accessed 2026-11-03.
- [766]Data & Analytics - Stratacache. STRATACACHE. https://stratacache.com/en/solutions/data-and-analytics/ Accessed 2026-11-03.
- [767]How top brands cracked the code on retail media incrementality this Prime Day (blog). CommerceIQ (author Daniel Wong, Senior Marketing Manager), 2025-08-15. https://www.commerceiq.ai/blog/how-top-brands-cracked-the-code-on-retail-media-incrementality-this-prime-day Accessed 2026-11-03.
- [768]What's New? (Commerce Max Help Center category). Criteo S.A.. https://help.retailmedia.criteo.com/kb/en/whats-new-419962 Accessed 2026-11-03.
- [769]Commerce Growth Help Center (help.criteo.com home). Criteo S.A.. https://help.criteo.com/ Accessed 2026-11-03.
- [770]Retail Media Software | Flywheel. Flywheel Digital (Omnicom). https://www.flywheeldigital.com/retail-media Accessed 2026-11-03.
- [771]Walmart: Understanding In-Store Attribution (Intentwise Knowledge Base > Intentwise Ad Optimizer > Walmart). Intentwise, Inc.. https://help.intentwise.com/understanding-walmarts-in-store-attribution Accessed 2026-11-03.
- [772]Criteo Ads: How to Get Started on Intentwise (Intentwise help article). Intentwise, Inc.. https://help.intentwise.com/criteo-ads-intentwise Accessed 2026-11-03.
- [773]Intentwise Ad Optimizer (help category: sections and article titles). Intentwise, Inc.. https://help.intentwise.com/intentwise-ad-optimizer Accessed 2026-11-03.
- [774]Intentwise Help Center (home: categories). Intentwise, Inc.. https://help.intentwise.com/ Accessed 2026-11-03.
- [775]Pacvue iROAS page (Internet Archive capture attempted). Pacvue. https://web.archive.org/web/2026/https://pacvue.com/iroas/ Accessed 2026-11-03. Access: partial extract.
- [776]Pacvue Debuts New Incrementality Console ... (newsroom; Internet Archive capture attempted). Pacvue. https://web.archive.org/web/2026/https://pacvue.com/newsroom/pacvue-debuts-new-incrementality-console-to-help-advertisers-grow-incremental-revenue-with-accurate-and-actionable-insights/ Accessed 2026-11-03. Access: partial extract.
- [777]Pacvue help centre (help.pacvue.com). Pacvue. https://help.pacvue.com/ Accessed 2026-11-03. Access: partial extract.
- [778]Amazon Advertising | Xnurta. Xnurta, Inc.. https://www.xnurta.com/amazon-advertising Accessed 2026-11-03. Access: partial extract.
- [779]Xnurta Help Center (help.xnurta.com). Xnurta, Inc.. https://help.xnurta.com/ Accessed 2026-11-03. Access: partial extract.
- [780]Attribution | Stackline. Stackline. https://www.stackline.com/attribution Accessed 2026-11-03.
- [781]About | Attain. Attain. https://www.attainoutcomes.com/about Accessed 2026-11-03.
- [782]Privacy Policy | Attain. Attain, 2026-04-16. https://www.attainoutcomes.com/privacy-policy Accessed 2026-11-03.
- [783]Consumer Privacy Policy (index) | Attain. Attain. https://www.attainoutcomes.com/consumer-privacy-policy Accessed 2026-11-03.
- [784]About | Incremental. Incremental. https://www.incremental.com/about Accessed 2026-11-03.
- [785]GeoLift | Haus. Haus Analytics, Inc.. https://www.haus.io/geolift Accessed 2026-11-03.
- [786]Incrementality Testing Engine | LiftLab. LiftLab Analytics, Inc.. https://liftlab.com/platform/incrementality-testing-engine/ Accessed 2026-11-03.
- [787]LiftLab home page. LiftLab Analytics, Inc.. https://liftlab.com/ Accessed 2026-11-03.
- [788]Metrics (Carrot Ads API docs). Instacart (Maplebear Inc.), 2026-09-17. https://docs.instacart.com/ads/ads_guide/concepts/ad_metrics Accessed 2026-11-03.
- [789]Privacy Notice - Legal Policies (Amazon Ads; country index). Amazon Ads. https://advertising.amazon.com/en-us/legal/privacy-notice Accessed 2026-11-03. Access: partial extract.
- [790]Interest-Based Ads (Amazon.com help notice). Amazon.com, Inc.. https://www.amazon.com/gp/help/customer/display.html?nodeId=202075050 Accessed 2026-11-03. Access: partial extract.
- [791]Amazon.com Privacy Notice (US). Amazon.com, Inc.. https://www.amazon.com/gp/help/customer/display.html?nodeId=468496 Accessed 2026-11-03. Access: partial extract.
- [792]Privacy Notice - Sam's Member Access Platform. Walmart Inc. (corporate.walmart.com), 2026-02-23. https://corporate.walmart.com/about/samsclub/privacy-notice-sams-member-access-platform Accessed 2026-11-03.
- [793]Target Privacy Policy. Target Corporation, 2026-08-15. https://www.target.com/c/target-privacy-policy/-/N-4sr7p Accessed 2026-11-03.
- [794]FTC Issues Orders to Eight Companies Seeking Information on Surveillance Pricing. Federal Trade Commission, 2024-07-23. https://www.ftc.gov/news-events/news/press-releases/2024/07/ftc-issues-orders-eight-companies-seeking-information-surveillance-pricing Accessed 2026-11-03.
- [795]FTC Surveillance Pricing 6(b) Study: Research Summaries — A Staff Perspective. Federal Trade Commission staff (Office of Technology), 2025-01 (event 2025-01-17). https://www.ftc.gov/system/files/ftc_gov/pdf/p246202_surveillancepricing6bstudy_researchsummaries_redacted.pdf Accessed 2026-11-03.
- [796]Target Health Privacy Policy (consumer health data). Target Corporation, 2025-06-30. https://www.target.com/c/health-policy/-/N-4gc0k Accessed 2026-11-03.
- [797]FTC Seeks Comment on Enforcement Policy Statement Regarding Personalized Pricing. Federal Trade Commission, 2026-08-19. https://www.ftc.gov/news-events/news/press-releases/2026/08/ftc-seeks-comment-enforcement-policy-statement-regarding-personalized-pricing Accessed 2026-11-03.
- [798]FTC Extends Public Comment on Proposed Policy Statement Regarding Personalized Pricing. Federal Trade Commission, 2026-09-03. https://www.ftc.gov/news-events/news/press-releases/2026/09/ftc-extends-public-comment-proposed-policy-statement-regarding-personalized-pricing Accessed 2026-11-03.
- [799]New ways for marketers to reach customers with AI Overviews and Lens. Google (Shashi Thakur), 2024-10-03. https://blog.google/products/ads-commerce/google-lens-ai-overviews-ads-marketers/ Accessed 2026-11-03.
- [800]More opportunities for your business on Google Search. Google (The Keyword, Google Ads), 2025-05-21. https://blog.google/products/ads-commerce/google-search-ai-brand-discovery/ Accessed 2026-11-03.
- [801]Why Perplexity Has Stopped Accepting New Advertisers. TipRanks (Vince Condarcuri), 2025-10-10 (event 2025-10). https://www.tipranks.com/news/why-perplexity-has-stopped-accepting-new-advertisers Accessed 2026-11-03.
- [802]Introducing Perplexity Shopping (hub post). Perplexity, 2024-11-18. https://www.perplexity.ai/hub/blog/introducing-perplexity-shopping Accessed 2026-11-03. Access: partial extract.
- [803]Agentic Commerce (developer documentation hub). OpenAI. https://developers.openai.com/commerce Accessed 2026-11-03.
- [804]Fetch Privacy Policy. Fetch Rewards LLC, 2026-10-13 (event effective 2026-01-22). https://fetch.com/privacy-policy Accessed 2026-11-03.
- [805]Return on Ad Spend reporting (Rokt Ads platform guides). Rokt. https://docs.rokt.com/platform-guides/rokt-ads/reporting/roas/ Accessed 2026-11-03.
- [806]Measuring return on investment: ROAS and iROAS (business.fetch.com trends-and-insights post). Fetch Rewards LLC, 2023-02-03. https://business.fetch.com/trends-and-insights/measuring-return-on-investment-roas-iroas Accessed 2026-11-03.
- [807]DG Media Network | National Scale. Unique Reach. (home). DG Media Network, LLC (Dollar General), 2026. https://www.dgmedianetwork.com/ Accessed 2026-11-03.
- [808]Programmatic advertising (Epsilon 'Digital media' US product page). Epsilon Data Management, LLC. https://www.epsilon.com/us/products-and-services/epsilon-digital Accessed 2026-11-03.
- [809]FAQs | Incremental. Incremental. https://www.incremental.com/faqs Accessed 2026-11-03.
- [810]Beyond Last-Click: How We Prove Incremental ROAS in Retail Media. Kevin Wu, Moloco. https://www.moloco.com/blog/proving-incremental-roas-retail-media-advertising Accessed 2026-11-03.
- [811]Regulations — California Privacy Protection Agency. California Privacy Protection Agency, 2026 (event 2026-01-01). https://cppa.ca.gov/regulations/ Accessed 2026-11-03.
- [812]Solo Stove Fires Up Sales With Chase Customers. Chase Media Solutions, undated (copyright 2026) (event campaign in Q4 2023). https://www.chase.com/mediasolutions/resources/solostove Accessed 2026-11-03.
- [813]Numerator Launches 200,000 U.S. Household Static Panel, Setting New Standard for Data Quality (press release). Numerator, LLC, 2025-04-10. https://www.numerator.com/press/numerator-launches-200000-u-s-household-static-panel-setting-new-standard-for-data-quality/ Accessed 2026-11-03.
- [814]Total Commerce Panel WHITEPAPER: Statics, Projection & Calibration (PDF). Numerator, 2023. https://www.numerator.com/wp-content/uploads/2023/11/TOTAL-COMMERCE.pdf Accessed 2026-11-03.
- [815]Albertsons Companies, Inc. Reports First Quarter Fiscal 2026 Results (Form 8-K exhibit 99.1). Albertsons Companies, Inc. (SEC EDGAR, accession 0001646972-26-000041), 2026-07-23. https://www.sec.gov/Archives/edgar/data/1646972/000164697226000041/aciq126earningsrelease.htm Accessed 2026-11-03.
- [816]DG Media Network Expands Commerce Media with Google's Commerce Media Suite and YouTube. DG Media Network (Dollar General), undated (footer © 2026). https://www.dgmedianetwork.com/news/dgmn-expands-commerce-media-with-google-youtube Accessed 2026-11-03.
- [817]In-Store Advertising - CVS Media Exchange. CVS Media Exchange. https://www.cvsmediaexchange.com/solutions/instore Accessed 2026-11-03.
- [818]Expanding access to display advertising for the Walmart Connect Partner Network & the advertisers they support (article). Walmart Connect (Walmart Inc.), 2025-02-26. https://www.walmartconnect.com/resources/articles/2025/partner-network-display-advertising-api Accessed 2026-11-03.
- [819]Brand Center Sales Dashboard (Ads Support, DoorDash Help Center). DoorDash, Inc.. https://help.doordash.com/en-us/advertising/article/ads-manager-dashboard Accessed 2026-11-03.
- [820]Explore DoorDash Ad Products. DoorDash, Inc.. https://advertising.doordash.com/en-us/products Accessed 2026-11-03.
- [821]Marketing 101: Grow Your Business with DoorDash Marketing (Merchant Learning Center). DoorDash, Inc., 2026-10-26. https://merchants.doordash.com/en-us/learning-center/how-to-get-new-customers Accessed 2026-11-03.
- [822]GULP Media Network — Non-endemic advertising. Gulp Media Network, 2026. https://www.gulpmedianetwork.com/non-endemic-advertising Accessed 2026-11-03.
- [823]Orange Apron Media — home. Orange Apron Media, 2026. https://orangeapronmedia.com/ Accessed 2026-11-03.
- [824]Giant Eagle Selects Grocery TV To Expand Retail Media Network. The Shelby Report, 2025-02-06. https://theshelbyreport.com/2025/02/06/giant-eagle-selects-grocery-tv-to-expand-retail-media-network/ Accessed 2026-11-03.
- [825]Uber Advertising: Measurement. Uber Technologies Inc. (Uber Advertising), 2026. https://www.uber.com/us/en/advertising/measurement/ Accessed 2026-11-03.
- [826]Uber Advertising custom AU (Uber blog, mx/en locale). Uber Technologies Inc. (Uber Advertising), 2025-11-04 (event press dateline 2025-10-31 (San Francisco, CA)). https://www.uber.com/mx/en/blog/uber-advertising-custom-au Accessed 2026-11-03.
- [827]Uber Advertising: Journey Ads page. Uber Technologies Inc.. https://www.uber.com/us/en/advertising/journey-ads/ Accessed 2026-11-03. Access: partial extract.
- [828]Resources | Chase Media Solutions (case-study and event listing). JPMorgan Chase Bank, N.A. (Chase Media Solutions), 2026. https://www.chase.com/mediasolutions/resources Accessed 2026-11-03.
- [829]Verified Incremental from Fetch: A Superior Measurement Methodology. Fetch Rewards LLC, unknown. https://business.fetch.com/trends-and-insights/verified-incremental-from-fetch-a-superior-measurement-methodology Accessed 2026-11-03. Access: partial extract.
- [830]Fetch for business (home page). Fetch Rewards LLC, 2026. https://business.fetch.com/ Accessed 2026-11-03. Access: partial extract.
- [831]Commissioned Nielsen study (MMM): Fetch iROAS outperforms traditional channels (business.fetch.com post). Fetch Rewards LLC, 2025-01-06. https://business.fetch.com/trends-and-insights/commissioned-nielsen-study-mmm-fetch-iroas-outperforms-traditional-channels Accessed 2026-11-03.
- [832]Expedia Group: offsite advertising. Expedia, Inc., an Expedia Group company, 2026. https://partner.expediagroup.com/en-us/solutions/advertise-with-us/offsite-advertising Accessed 2026-11-03.
- [833]Ad Policies for Klarna Advertisers (docs.klarna.com). Klarna. https://docs.klarna.com/resources/legal-and-compliance/more-solutions-guidelines/ad-policies-for-klarna-advertisers Accessed 2026-11-03. Access: partial extract.
- [834]Rokt Ads platform guides (index). Rokt, 2026-08-04. https://docs.rokt.com/platform-guides/rokt-ads/ Accessed 2026-11-03. Access: partial extract.
- [835]Rokt Ads FAQ. Rokt, 2026-08-04. https://docs.rokt.com/platform-guides/rokt-ads/faq/ Accessed 2026-11-03.
- [836]Capital One Ad Solutions home page (Internet Archive capture, 12 May 2026). Capital One. https://web.archive.org/web/20260512034918/https://capitaloneads.com/ Accessed 2026-11-03. Access: partial extract.
- [837]Demand-Side Metrics (Commerce Yield Help Center). Criteo, 2026-08-13. https://help.cyield.criteo.com/kb/guide/en/demand-side-metrics-MmKu6As0MP/Steps/3211686 Accessed 2026-11-03.
- [838]Kevel Documentation (index). Kevel. https://dev.kevel.com/docs Accessed 2026-11-03.
- [839]Reporting Discrepancies (Kevel docs). Kevel, 2026-09-29. https://dev.kevel.com/docs/reporting-discrepancies.md Accessed 2026-11-03.
- [840]Commerce Max Help Center (home: categories). Criteo S.A.. https://help.retailmedia.criteo.com/ Accessed 2026-11-03.
- [841]Reporting & Insights (Commerce Max Help Center category: article titles). Criteo S.A.. https://help.retailmedia.criteo.com/kb/en/reporting-insights-316769 Accessed 2026-11-03.
- [842]Understand Sales Attribution | Commerce Max Help Center (category: sub-article titles). Criteo S.A.. https://help.retailmedia.criteo.com/kb/en/understand-sales-attribution-127275 Accessed 2026-11-03.
- [843]Understand Your Data (Commerce Max Help Center category: article titles). Criteo S.A.. https://help.retailmedia.criteo.com/kb/en/understand-your-data-127274 Accessed 2026-11-03.
- [844]Perpetua Help Center (home: collections). Perpetua (Flywheel Digital Limited). https://help.perpetua.io/ Accessed 2026-11-03.
- [845]Amazon Advertising - Reporting & Analytics Tools (Perpetua help collection: article titles). Perpetua (Flywheel Digital Limited). https://help.perpetua.io/en/collections/2968136-amazon-advertising-reporting-analytics-tools Accessed 2026-11-03.
- [846]Sponsored Display new-to-brand metrics (what's new; en-us path). Amazon Ads. https://advertising.amazon.com/en-us/resources/whats-new/sponsored-display-new-to-brand-metrics Accessed 2026-11-03. Access: partial extract.
- [847]Introducing CorIQ: Built on ExtraCare, engineered for results.. CVS Media Exchange. https://www.cvsmediaexchange.com/insights/introducing-CorIQ-from-CMX Accessed 2026-11-03.
- [848]Site Search Advertising - CVS Media Exchange (sponsored product search). CVS Media Exchange. https://www.cvsmediaexchange.com/solutions/sponsored-product-search Accessed 2026-11-03.
- [849]CMX Programmatic Self-Service. CVS Media Exchange. https://www.cvsmediaexchange.com/solutions/cmx-programmatic-self-service Accessed 2026-11-03.
- [850]Kevel, Dollar General, and The Trade Desk Introduce First-of-its-Kind Retail Media Solution Connecting Onsite and Offsite Activation. Kevel (release hosted by DG Media Network), 2026-05-04 (event 2026-05-04 (announcement)). https://www.dgmedianetwork.com/news/kevel-dollar-general-and-the-trade-desk-introduce-first-of-its-kind-retail-media-solution-connecting-onsite-and-offsite-activation Accessed 2026-11-03.
- [851]DGMN news listing. DG Media Network, LLC. https://www.dgmedianetwork.com/news Accessed 2026-11-03. Access: partial extract.
- [852]TravelAds Sponsored Listings (advertising.expedia.com). Expedia Group Advertising. https://advertising.expedia.com/solutions/sponsored-listings Accessed 2026-11-03. Access: partial extract.
- [853]TravelAds Tips and Tricks to Help Boost Hotel Bookings. Haley White, Director of Business Development, Expedia Group Media Solutions (on Hospitality Net), 2022-06-17. https://www.hospitalitynet.org/opinion/4111085.html Accessed 2026-11-03.
- [854]Hy-Vee RedMedia Partners with Instacart to Further Retail Media Capabilities. Maplebear Inc. d/b/a Instacart and Hy-Vee (PR Newswire), 2025-03-25. https://www.prnewswire.com/news-releases/hy-vee-redmedia-partners-with-instacart-to-further-retail-media-capabilities-302409812.html Accessed 2026-11-03.
- [855]New Studies Show Instacart Ads Influencing In-Store Purchases. Maplebear Inc. d/b/a Instacart, 2024-12-17. https://web.archive.org/web/20260711165924id_/https://company.instacart.com/instacart-ads/new-studies-show-instacart-ads-influencing-in-store-purchases Accessed 2026-11-03.
- [856]Instacart Ads products page (Internet Archive capture, 2026, raw HTML). Maplebear Inc. d/b/a Instacart. https://web.archive.org/web/2026/https://company.instacart.com/ads/products-grocery-retail-media Accessed 2026-11-03. Access: partial extract.
- [857]Shoppable display ads on Caper Carts (Instacart Ads Manager docs). Maplebear Inc. d/b/a Instacart (docs.instacart.com), 2026-08-12. https://docs.instacart.com/ads_manager/display_ads/shoppable_display_ads_on_caper_carts Accessed 2026-11-03.
- [858]Create a Premium Placement Ad (84.51° Prism Learning Center). 84.51° LLC, 2026-01-28. https://prism.8451.com/activation/help/content/how-to/kpm-managed-service/channels/premium-placement.htm Accessed 2026-11-03.
- [859]Kroger Precision Marketing (KPM) Managed Service (84.51° Prism Learning Center, concept page). 84.51° LLC, 2026-05-08. https://prism.8451.com/activation/help/content/about/kpm-managed-service.htm Accessed 2026-11-03.
- [860]Direct Connect (84.51° Prism Learning Center, concept page; deprecated topic). 84.51° LLC, 2026-05-08. https://prism.8451.com/activation/help/content/about/direct-connect.htm Accessed 2026-11-03.
- [861]Lowe's Media Network site. Lowe's Companies, Inc.. https://lowesmedianetwork.com/ Accessed 2026-11-03. Access: partial extract.
- [862]Wayback Machine capture of lowesmedianetwork.com. Internet Archive. https://web.archive.org/web/2026/https://lowesmedianetwork.com/ Accessed 2026-11-03. Access: partial extract.
- [863]MMN Privacy Notice. Macy's Media Network (Advertex Communications Inc.), 2023-02-17. https://www.macysmedianetwork.com/privacy.html Accessed 2026-11-03.
- [864]Macy's Media Network Legal Notice (page headed 'MMN Privacy Notice'). Advertex Communications Inc. d/b/a Macy's Media Network, 2023-06-12. https://www.macysmedianetwork.com/legal.html Accessed 2026-11-03.
- [865]Orange Apron Media Expands Pinterest Partnership to Non-Endemic Advertisers. Orange Apron Media (Stephenie Cattonar, Senior Director, Strategy, Product & Analytics), 2026-04-22 (date from the blog listing RTL-E2-S13; the post shows none) (event 2026-04-22 (announcement, per listing)). https://orangeapronmedia.com/oam-expand-pinterest-partnership/ Accessed 2026-11-03.
- [866]Solutions (Sam's Club Connect). Sam's Club / Walmart, Inc.. https://www.samsclubconnect.com/solutions Accessed 2026-11-03.
- [867]O & O Solutions | Walgreens Advertising Group. Walgreens Advertising Group, 2024 (footer). https://www.walgreensadvertisinggroup.com/home/solution-detail1.html Accessed 2026-11-03.
- [868]Approach | Walgreens Advertising Group. Walgreens Advertising Group, 2024 (footer). https://www.walgreensadvertisinggroup.com/home/approach.html Accessed 2026-11-03.
- [869]Walgreens Advertising Group terms of use. Walgreens Advertising Group. https://www.walgreensadvertisinggroup.com/terms-of-use.html Accessed 2026-11-03. Access: partial extract.
- [870]Walmart Advertising | Xnurta. Xnurta, Inc.. https://www.xnurta.com/walmart-advertising Accessed 2026-11-03.
- [871]How VIZIO helped serve extra shots of new product awareness (Cafe Bustelo case study). Walmart Connect (Walmart Inc.), 2026 (event 2025-07-25 to 2025-08-22 (campaign)). https://www.walmartconnect.com/resources/case-studies/2026/cafe-bustelo-vizio Accessed 2026-11-03.
- [872]How we're helping brands create more omni customer connections with one of Walmart's greatest strengths: Our stores (article). Walmart Connect (Whitney Cooper), 2023-05-05 (event 2023-04 (demo pilot start)). https://www.walmartconnect.com/resources/articles/2023/taking-omnichannel-capabilities-even-further-with-our-stores Accessed 2026-11-03.
- [873]walmartconnect.com sitemap. Walmart Connect. https://www.walmartconnect.com/sitemap.xml Accessed 2026-11-03.
- [874]Advertiser API Terminology. Cardlytics, Inc., 2026-02-25. https://platform.cardlytics.com/advertisers/docs/api-terminology Accessed 2026-11-03.
- [875]Master Advertiser Terms & Conditions (CPT). Cardlytics, Inc.. https://www.cardlytics.com/advtscscpt Accessed 2026-11-03.
- [876]Advertiser API FAQs. Cardlytics, Inc., 2026-02-13. https://platform.cardlytics.com/advertisers/docs/api-faqs Accessed 2026-11-03.
- [877]Master Advertiser Terms & Conditions (CPR). Cardlytics, Inc.. https://www.cardlytics.com/advtscscpr Accessed 2026-11-03.
- [878]Master Advertiser Terms (CPS, UK). Cardlytics UK. https://www.cardlytics.com/uk/advtscscps Accessed 2026-11-03.
- [879]Chase Online Privacy Policy. JPMorgan Chase Bank, N.A., 2023-09. https://www.chase.com/digital/resources/privacy-security/privacy/online-privacy-policy Accessed 2026-11-03.
- [880]U.S. Consumer Privacy Notice (Chase). JPMorgan Chase Bank, N.A., 2025-12. https://www.chase.com/digital/resources/privacy-security/privacy/consumer-privacy-notice Accessed 2026-11-03.
- [881]Citi Commerce Media | Advertise with Citi (commercemedia.citi.com). Citigroup Inc.. https://commercemedia.citi.com/ Accessed 2026-11-03. Access: partial extract.
- [882]Klarna Group plc Form 20-F for fiscal 2025 — R76.htm Operating segments - Disclosure of disaggregation of revenue (Details). Klarna Group plc (SEC EDGAR), 2026-02-26. https://www.sec.gov/Archives/edgar/data/2003292/000200329226000007/R76.htm Accessed 2026-11-03.
- [883]Klarna Search price list (US). Klarna, 2025-03-26. https://docs.klarna.com/marketing/solutions/klarna-search/price-list/ Accessed 2026-11-03.
- [884]PayPal Advertising solutions (PayPal US). PayPal (PayPal Ads). https://www.paypal.com/us/advertiser/solutions Accessed 2026-11-03. Access: partial extract.
- [885]PayPal Privacy Statement (US legal hub). PayPal. https://www.paypal.com/us/legalhub/paypal/privacy-full Accessed 2026-11-03. Access: partial extract.
- [886]Experiments: getting started (Rokt Ads platform guides). Rokt, 2026-08-04. https://docs.rokt.com/platform-guides/rokt-ads/experiments/getting-started/ Accessed 2026-11-03.
- [887]Giant Eagle validation (Upside retailer blog). Upside Services, Inc., 2025-11-04. https://www.upside.com/business/retailer-blog/giant-eagle-validation Accessed 2026-11-03.
- [888]External audit of Upside's methodology (Upside knowledge center). Upside Services, Inc.. https://knowledgecenter.upside.com/hc/en-us/articles/10715366947867-External-audit-of-Upside-s-methodology Accessed 2026-11-03. Access: partial extract.
- [889]Schnucks Increases Retail Media Revenue by 7x with Instacart Carrot Ads Technology. Maplebear Inc. d/b/a Instacart (enterprise blog), 2025-08-19. https://web.archive.org/web/20260713151850id_/https://company.instacart.com/enterprise-blog/schnucks-increases-retail-media-revenue-by-7x-with-instacart-carrot-ads-technology Accessed 2026-11-03.
- [890]Osmos Developer Hub. Osmos, Inc.. https://developers.osmos.ai/ Accessed 2026-11-03. Access: partial extract.
- [891]Zitcha: About us. Zitcha. https://www.zitcha.com/about-us Accessed 2026-11-03.
- [892]Walmart: New-to-Brand (NTB) Metrics (Intentwise help article). Intentwise, Inc.. https://help.intentwise.com/understanding-walmarts-new-to-brand-ntb-metrics Accessed 2026-11-03.
- [893]Walmart Advertising with Perpetua (help collection: article titles). Perpetua (Flywheel Digital Limited). https://help.perpetua.io/en/collections/3003429-walmart-advertising-with-perpetua Accessed 2026-11-03.
- [894]Instacart Advertising with Perpetua (help collections: article titles). Perpetua (Flywheel Digital Limited). https://help.perpetua.io/en/collections/2565187-instacart-advertising-with-perpetua Accessed 2026-11-03.
- [895]How to Connect Your Walmart Account to Perpetua (Perpetua help article). Perpetua (author Andrea Mamone), 2026-03-27. https://help.perpetua.io/en/articles/5589444-how-to-connect-your-walmart-account-to-perpetua Accessed 2026-11-03.
- [896]Stackline and Amazon Launch Groundbreaking Multi-Retailer Attribution Solution (Business Wire release, syndicated copy). Stackline via Business Wire; copy hosted by Silicon UK, 2024-06-17. https://www.silicon.co.uk/press-release/stackline-and-amazon-launch-groundbreaking-multi-retailer-attribution-solution-allowing-brands-to-see-the-sales-impact-of-their-retail-media-investment-across-all-retailers-and-channels Accessed 2026-11-03.
- [897]Legal (cookie policy, terms of service, privacy notice) | Catalina. Catalina. https://www.catalina.com/legal Accessed 2026-11-03. Access: partial extract.
- [898]Incremental API Docs. Incremental. https://docs.incremental.com/ Accessed 2026-11-03. Access: partial extract.
- [899]Media Solutions - Purchase Data to Optimize Marketing Dollars | Numerator. Numerator, LLC. https://www.numerator.com/solutions/media-solutions/ Accessed 2026-11-03.
- [900]Consumer Data Privacy | Numerator. Numerator, LLC. https://www.numerator.com/consumer-data-privacy/ Accessed 2026-11-03.
- [901]Privacy Policy | Numerator. Numerator, LLC, 2026-03-09. https://www.numerator.com/privacy-policy/ Accessed 2026-11-03.
- [902]Receipt Hog Privacy Policy. Numerator, LLC (operator of the Receipt Hog website and app), 2026-06-11. https://receipthog.com/privacy/ Accessed 2026-11-03. Access: partial extract.
- [903]EDGAR submissions record, Albertsons Companies, Inc. (CIK 0001646972). U.S. Securities and Exchange Commission. https://data.sec.gov/submissions/CIK0001646972.json Accessed 2026-11-03.
- [904]EDGAR filing list: Amazon.com, Inc. 10-K filings. SEC EDGAR, 2026-02-06. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=1018724&type=10-K&dateb=&owner=include&count=5 Accessed 2026-11-03.
- [905]Cardlytics, Inc. Form 10-K filing index (accession 0001666071-26-000010). Cardlytics, Inc. (SEC EDGAR), 2026-03-04. https://www.sec.gov/Archives/edgar/data/1666071/000166607126000010/0001666071-26-000010-index.htm Accessed 2026-11-03.
- [906]CRITEO REPORTS STRONG SECOND QUARTER 2022 RESULTS. Criteo S.A., 2022-08-03 (event 2022-08-01). https://criteo.investorroom.com/2022-08-03-CRITEO-REPORTS-STRONG-SECOND-QUARTER-2022-RESULTS Accessed 2026-11-03.
- [907]Criteo and DoorDash Sign Multi-Year Partnership to Expand Retail Media Access. Criteo S.A., 2025-10-06. https://criteo.investorroom.com/2025-10-06-Criteo-and-DoorDash-Sign-Multi-Year-Partnership-to-Expand-Retail-Media-Access Accessed 2026-11-03.
- [908]EDGAR filing list: Criteo S.A. 10-K filings. SEC EDGAR, 2026-02-26. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=1576427&type=10-K&dateb=&owner=include&count=5 Accessed 2026-11-03.
- [909]Maplebear Inc. Form 10-K filing index (accession 0001579091-26-000018). Maplebear Inc. (SEC EDGAR), 2026-02-26. https://www.sec.gov/Archives/edgar/data/1579091/000157909126000018/0001579091-26-000018-index.htm Accessed 2026-11-03.
- [910]Kroger Precision Marketing Shares In-Store Advertising Solution. The Shelby Report (staff), 2025-06-18. https://theshelbyreport.com/2025/06/18/kroger-precision-marketing-shares-in-store-advertising-solution Accessed 2026-11-03.