{ "note": "Blind second scoring pass, 30 September 2026 (internal review, carried over from the series method). A separate AI scorer, given the published scoring rules and each row's cited sources and verified claims but not the first scores, grades or rationales, re-scored every applicable cell of the rows whose band sits within 2 points of a band line or moves under a K or reweighting variant. The second pass is reported beside the first and replaces no score. Material disagreements (n/e against a score, or two or more points apart) were adjudicated by re-reading the deciding sources: two cells, both upheld for the second pass (AJ-01, AJ-02). Agreement statistics compare the second pass with the first pass as scored before adjudication.", "rows": [ "adelaide", "tvision", "eye-square-incontext", "lumen-spotlight", "neurons-ai", "tobii-sticky" ], "cells": [ { "id": "adelaide", "vendor": "Adelaide", "product": "Adelaide AU", "cohort": "media", "dimension": "method_validity", "first": 2, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-ADELAIDE-S06", "ATT-ADELAIDE-S10", "ATT-ADELAIDE-S11", "ATT-E-S34", "ATT-C-S10" ], "second_reason": "Anchor 3: AU uses eye-tracking reference data (fully opted-in panels licensed from Amplified, Lumen, Tobii, TVision and Viomba plus Adelaide's own research, TVision panel weights for CTV), but no reference sample size or recruitment and no held-out agreement statistic are published, and AU is a model estimate that blends attention with outcome prediction." }, { "id": "adelaide", "vendor": "Adelaide", "product": "Adelaide AU", "cohort": "media", "dimension": "outcome_validation", "first": 2, "first_grade": "D", "second": 2, "second_grade": "D", "second_sources": [ "ATT-C-S10", "ATT-ADELAIDE-S08", "ATT-C-S06", "ATT-C-S07", "ATT-ADELAIDE-S33", "ATT-C-S12" ], "second_reason": "Anchor 2: vendor and commissioned case work states samples and outcomes (Google-commissioned 395 US campaigns with 5.6x conversion rate for high- vs low-AU OLV; Outcomes Guides of 52 and 60 case studies with brand and lower-funnel lifts; iFIT Upwave lifts) without a published method or defined comparator, and the Adelaide-commissioned MediaSense rebuild of AU-to-brand-KPI curves (Media Leader page would not open without sign-in, so relied on the claims) states no sample, used Adelaide-supplied data and reports no holdout, so anchor 3 is not met." }, { "id": "adelaide", "vendor": "Adelaide", "product": "Adelaide AU", "cohort": "media", "dimension": "granularity", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-ADELAIDE-S06", "ATT-V-ADELAIDE-S03", "ATT-E-S15", "ATT-ADELAIDE-S41" ], "second_reason": "Anchor 3: vendor technical documentation shows placement- and campaign-level AU reporting, placement-level pre-bid segments and an impression-level AU signal in DV360 custom bidding, all labelled as model estimates; impression-level output at a second stage (post-campaign in Amazon DSP) rests only on a vendor release, so anchor 4 would be capped at 3 by grade D." }, { "id": "adelaide", "vendor": "Adelaide", "product": "Adelaide AU", "cohort": "media", "dimension": "activation", "first": 3, "first_grade": "D", "second": 4, "second_grade": "C", "second_sources": [ "ATT-V-ADELAIDE-S03", "ATT-E-S18", "ATT-E-S15", "ATT-V-ADELAIDE-S05", "ATT-E-S14", "ATT-ADELAIDE-S16" ], "second_reason": "Anchor 5 is met on substance (live in three or more named buying platforms: DV360 'Maximize Adelaide AU' custom bidding per Google's help centre, The Trade Desk self-serve tiers since October 2024, Amazon DSP tiers and Quality Floor since June 2026, plus Yahoo DSP since 2023), but the strongest source is a platform help centre (C) and the trade reports relay vendor statements with funding unknown, so the C cap limits the score to 4." }, { "id": "adelaide", "vendor": "Adelaide", "product": "Adelaide AU", "cohort": "media", "dimension": "channel_coverage", "first": 3, "first_grade": "D", "second": 3, "second_grade": "C", "second_sources": [ "ATT-ADELAIDE-S06", "ATT-E-S34", "ATT-G-S01", "ATT-ADELAIDE-S35", "ATT-ADELAIDE-S12", "ATT-ADELAIDE-S41" ], "second_reason": "Anchor 3: vendor technical documentation shows live measurement methods for display, online video, CTV and podcast audio (four channels); social rests only on vendor marketing (2020 release, report uploads), so a fifth channel would be capped at 3 by grade D, and DOOH has no documented method." }, { "id": "adelaide", "vendor": "Adelaide", "product": "Adelaide AU", "cohort": "media", "dimension": "data_access", "first": 3, "first_grade": "D", "second": 3, "second_grade": "D", "second_sources": [ "ATT-U-S34", "ATT-V-ADELAIDE-S11", "ATT-E-S15", "ATT-ADELAIDE-S41" ], "second_reason": "Anchor 3: one or two named live routes beyond Adelaide's dashboard are documented (AU in Nielsen ONE's Outcomes Marketplace since October 2025; post-campaign impression-level AU visibility in Amazon DSP), both in vendor or partner releases, and no API reference, log-level export, ad-server or clean-room route is documented." }, { "id": "adelaide", "vendor": "Adelaide", "product": "Adelaide AU", "cohort": "media", "dimension": "production_scale", "first": 3, "first_grade": "D", "second": 3, "second_grade": "D", "second_sources": [ "ATT-ADELAIDE-S33" ], "second_reason": "The Google-commissioned case study reports 32B impressions across 395 US campaigns in 2022-2024, a derived rate of about 10.7B impressions a year (anchor 4, not dated before 2024 since the period runs into 2024), capped at 3 because the only source is a vendor case study (grade D)." }, { "id": "adelaide", "vendor": "Adelaide", "product": "Adelaide AU", "cohort": "media", "dimension": "us_availability", "first": 3, "first_grade": "D", "second": 4, "second_grade": "C", "second_sources": [ "ATT-E-S34", "ATT-V-ADELAIDE-S03", "ATT-E-S18", "ATT-E-S14", "ATT-ADELAIDE-S33", "ATT-C-S12" ], "second_reason": "Anchor 5 is met on substance (sold through US buying platforms; CTV and linear calibration on TVision US panel data and US tune-in attribution; US production in 395 US campaigns, iFIT and Verizon), but US production evidence is vendor marketing and the calibration and availability evidence is vendor or platform documentation, so the C cap limits the score to 4." }, { "id": "adelaide", "vendor": "Adelaide", "product": "Adelaide AU", "cohort": "media", "dimension": "assurance", "first": 1, "first_grade": "A", "second": 1, "second_grade": "A", "second_sources": [ "ATT-A-S10", "ATT-A-S09", "ATT-C-S10" ], "second_reason": "MRC's own records show AU as 'Pre-audit reviewed 5/2026, Audit not begun' and no MRC accreditation of AU or its viewability or IVT inputs, so anchors 3 and 2 are not met; the remaining assurance is Adelaide's assertion that AU 'passed' a review it commissioned from MediaSense, which is not accreditation (anchor 1)." }, { "id": "adelaide", "vendor": "Adelaide", "product": "Adelaide AU", "cohort": "media", "dimension": "method_disclosure", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-ADELAIDE-S06", "ATT-E-S34", "ATT-A-S19", "ATT-ADELAIDE-S10", "ATT-ADELAIDE-S11", "ATT-C-S10" ], "second_reason": "Anchor 3: four of seven elements are published (construct, signals, reference labels from named gaze and outcome sources, and a summary model and four-step scoring procedure), while panel size and recruitment, validation statistics and recalibration cadence are not, and algorithm weights are withheld." }, { "id": "adelaide", "vendor": "Adelaide", "product": "Adelaide AU", "cohort": "media", "dimension": "privacy", "first": 4, "first_grade": "C", "second": 4, "second_grade": "C", "second_sources": [ "ATT-H-S24", "ATT-ADELAIDE-S06" ], "second_reason": "Anchor 4: the product-specific AU Ratings privacy policy documents data minimisation (device type, IP, location to city/DMA/zip, inferences, no sensitive data, no cookies or device storage access) and addresses CCPA, GDPR, UK GDPR and ePrivacy, and Adelaide states its eye-tracking data come from fully opted-in panels; no independent compliance record exists for 5." }, { "id": "tvision", "vendor": "TVision", "product": "TVision", "cohort": "media", "dimension": "method_validity", "first": 3, "first_grade": "C", "second": 3, "second_grade": "B", "second_sources": [ "ATT-A-S23", "ATT-B-S06", "ATT-V-TVISION-S01", "ATT-H-S26", "ATT-TVISION-S24" ], "second_reason": "Anchor 3: attention is observed directly by camera gaze inference (eyes-open plus eyes-on-screen) in a stated opt-in US panel of about 5,000 homes recruited from top DMAs and weighted to Census/ACS, but no agreement or error statistic for the gaze classifier on held-out reference data is published by TVision or in the independent paper." }, { "id": "tvision", "vendor": "TVision", "product": "TVision", "cohort": "media", "dimension": "outcome_validation", "first": 3, "first_grade": "B", "second": 4, "second_grade": "B", "second_sources": [ "ATT-B-S06", "ATT-V-TVISION-S10", "ATT-C-S25", "ATT-TVISION-S48" ], "second_reason": "Brand: the independent McGranaghan-Liaukonyte-Wilbur study (no funding; TVision supplied data) finds TVision attention explains the most variance in brand-search lift across 180 national ads, but in-sample, without holdout, on 2016-17 data and not replicated, so it sits above commissioned-no-holdout (3) but short of 5; memory: TVision/Upwave regression on 40+ US campaigns with aided awareness, no sample or holdout (anchor 2); sales: Viant's 42 pilot campaigns with 1.4x conversion lift, no design or control (anchor 1-2)." }, { "id": "tvision", "vendor": "TVision", "product": "TVision", "cohort": "media", "dimension": "granularity", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-TVISION-S16", "ATT-TVISION-S15", "ATT-TVISION-S14", "ATT-TVISION-S12", "ATT-TVISION-S18" ], "second_reason": "Anchor 3: TVision delivers placement-level planning benchmarks (network, daypart, program), mid-campaign and post-campaign campaign-level reports from panel observation, and DSP scores only at daypart and channel level from historic data; impression-level output exists only as panel data feeds, at one stage, and stages are not labelled by measurement basis." }, { "id": "tvision", "vendor": "TVision", "product": "TVision", "cohort": "media", "dimension": "activation", "first": 3, "first_grade": "D", "second": 3, "second_grade": "A", "second_sources": [ "ATT-E-S25", "ATT-E-S22", "ATT-V-TVISION-S09", "ATT-TVISION-S26", "ATT-TVISION-S14" ], "second_reason": "Anchor 3: OpenX launched TVision-powered pre-bid CTV attention targeting in OpenXSelect (one named platform, March 2026, post-acquisition status undocumented) and Viant's 10 August 2026 filing says TVision pre-bid attention is still in testing in Viant's stack (beta; the Havas Attention Marketplace claim is vendor marketing outranked by the filing); TVision's own DSP scoring page names no DSP, so two live platforms are not documented." }, { "id": "tvision", "vendor": "TVision", "product": "TVision", "cohort": "media", "dimension": "channel_coverage", "first": 1, "first_grade": "C", "second": 1, "second_grade": "C", "second_sources": [ "ATT-TVISION-S12", "ATT-TVISION-S16", "ATT-TVISION-S15", "ATT-TVISION-S48" ], "second_reason": "Anchor 1: measurement is live only on the TV screen; of the seven channels only CTV counts (including YouTube and other walled-garden apps viewed on TV), while linear TV is not one of the seven and earns no points." }, { "id": "tvision", "vendor": "TVision", "product": "TVision", "cohort": "media", "dimension": "data_access", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-TVISION-S12", "ATT-TVISION-S13", "ATT-TVISION-S15", "ATT-G-S20" ], "second_reason": "Anchor 3: beyond dashboards, TVision documents one live named route, licensed person-level data feeds (sessionized raw data, calculated probabilities, behavioural projections; licensee VideoAmp named), while the DoubleVerify integration was only a 2023 closed-beta plan with no documented GA, and no API reference, ad-server or clean-room route is documented." }, { "id": "tvision", "vendor": "TVision", "product": "TVision", "cohort": "media", "dimension": "production_scale", "first": 1, "first_grade": "D", "second": 1, "second_grade": "C", "second_sources": [ "ATT-TVISION-S12", "ATT-U-S24", "ATT-TVISION-S54" ], "second_reason": "Anchor 1: the only volume figures are 'millions of program and ad impressions' and a cumulative counter of about 25 billion measured seconds, both with no period or start date (and the latter in seconds, not impressions); no impressions- or campaigns-per-year figure is stated." }, { "id": "tvision", "vendor": "TVision", "product": "TVision", "cohort": "media", "dimension": "us_availability", "first": 4, "first_grade": "C", "second": 4, "second_grade": "C", "second_sources": [ "ATT-A-S23", "ATT-V-TVISION-S01", "ATT-H-S26", "ATT-TVISION-S16", "ATT-C-S25", "ATT-B-S06" ], "second_reason": "Anchor 5 is met on vendor documentation (sold in the US, a 5,000-home US panel weighted to Census/ACS, US production with named US clients such as Anheuser-Busch and Vevo and 40+ US campaigns in the Upwave study), but current availability and use rest on vendor sources (the independent US evidence is the 2016-17 panel extract), so the grade C cap holds it at 4." }, { "id": "tvision", "vendor": "TVision", "product": "TVision", "cohort": "media", "dimension": "assurance", "first": 1, "first_grade": "A", "second": 1, "second_grade": "A", "second_sources": [ "ATT-A-S09", "ATT-TVISION-S44", "ATT-TVISION-S47", "ATT-A-S10", "ATT-A-S11", "ATT-V-TVISION-S04", "ATT-V-TVISION-S03" ], "second_reason": "Anchor 1: MRC's accredited and under-review lists (digital, television, multimedia) carry no TVision or Viant attention entry and no accredited input metric, and TVision only self-attests, saying it supports the IAB/MRC attention guidelines, finds they validate its approach, and holds ISO 27001 (an information-security standard, not attention)." }, { "id": "tvision", "vendor": "TVision", "product": "TVision", "cohort": "media", "dimension": "method_disclosure", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-A-S23", "ATT-TVISION-S54", "ATT-V-TVISION-S01", "ATT-B-S06" ], "second_reason": "Anchor 3: five of seven elements are public (construct: time looking at the screen; signals: optical sensors and audio fingerprints; panel size and recruitment; scoring procedure: eyes-open plus eyes-on-screen co-occurrence with daily weighting projection; cadence: daily in-tab and weighting), but attention reference labels and validation statistics are not published." }, { "id": "tvision", "vendor": "TVision", "product": "TVision", "cohort": "media", "dimension": "privacy", "first": 4, "first_grade": "C", "second": 4, "second_grade": "C", "second_sources": [ "ATT-TVISION-S22", "ATT-A-S23", "ATT-H-S26", "ATT-V-TVISION-S03" ], "second_reason": "Anchor 4: the panel-specific privacy policy documents the consent model (opt-in, compensated, withdrawable consent to sharing) and the signals collected (including face geometry and periodic snapshot images, with a three-year biometric destruction rule), the method and panel pages document minimisation (no video or images leave the home once in-tab), and COPPA, CCPA and VCDPA are addressed, though the policy also allows sharing hashed, re-identifiable personal data, and the ISO 27001 certificate is reported only in a vendor release, not an independent privacy record, so 5 is not reached." }, { "id": "eye-square-incontext", "vendor": "eye square", "product": "eye square InContext", "cohort": "creative", "dimension": "method_validity", "first": 3, "first_grade": "D", "second": 3, "second_grade": "D", "second_sources": [ "ATT-EYESQUARE-S05", "ATT-EYESQUARE-S23", "ATT-EYESQUARE-S19", "ATT-EYESQUARE-S25" ], "second_reason": "Anchor 3: InContext Eye Tracking measures gaze directly with the webcam-based SEAL algorithm after on-screen calibration, but the only validation is a bare '40-60px accuracy' with no reference tracker, reference sample or held-out statistic (the patent describes a calibration method without results, and the base platform without eye tracking is on-screen signals only)." }, { "id": "eye-square-incontext", "vendor": "eye square", "product": "eye square InContext", "cohort": "creative", "dimension": "outcome_validation", "first": 2, "first_grade": "D", "second": 3, "second_grade": "D", "second_sources": [ "ATT-EYESQUARE-S19", "ATT-EYESQUARE-S26" ], "second_reason": "Anchor 3: sales, the vendor-run ATLAS study (May 2024, SEAL) states its design (12 test cells plus control), sample (N=1,203, Dynata, 18-49, smartphone) and outcome (simulated Amazon purchase) but shows the attention-purchase link only as three platform-level points with no holdout or statistic; memory, the vendor's 2020 meta-analysis reports dwell time explaining 10.4% of recall and 52.3% of recognition variance across 2003-2019 in-context studies that predate SEAL; brand, no independent or attention-linked evidence; grade D caps at 3." }, { "id": "eye-square-incontext", "vendor": "eye square", "product": "eye square InContext", "cohort": "creative", "dimension": "granularity", "first": 3, "first_grade": "D", "second": 3, "second_grade": "D", "second_sources": [ "ATT-EYESQUARE-S02", "ATT-EYESQUARE-S04", "ATT-EYESQUARE-S05", "ATT-EYESQUARE-S19", "ATT-EYESQUARE-S06" ], "second_reason": "Anchor 4 facts (viewer-retention curves over video time, gaze heatmaps, AOI and branded-moment attention per asset, with benchmarks only at ad level, not at that resolution) are capped at 3 by the grade-D vendor marketing sources." }, { "id": "eye-square-incontext", "vendor": "eye square", "product": "eye square InContext", "cohort": "creative", "dimension": "creative_diagnostics", "first": 3, "first_grade": "D", "second": 3, "second_grade": "D", "second_sources": [ "ATT-EYESQUARE-S02", "ATT-EYESQUARE-S03", "ATT-EYESQUARE-S19", "ATT-EYESQUARE-S04" ], "second_reason": "Anchor 4 facts (a described pre-launch design: replicated feed, one ad version or control, monadic with about 100 per stimulus, post-exposure survey; norms of 'over 1000 previous tests in 11 sectors' with no per-sector breakdown) are capped at 3 by grade D." }, { "id": "eye-square-incontext", "vendor": "eye square", "product": "eye square InContext", "cohort": "creative", "dimension": "channel_coverage", "first": 3, "first_grade": "D", "second": 3, "second_grade": "D", "second_sources": [ "ATT-EYESQUARE-S07", "ATT-EYESQUARE-S06", "ATT-EYESQUARE-S05", "ATT-EYESQUARE-S19", "ATT-EYESQUARE-S12" ], "second_reason": "Anchor 3: social, online video, TV/CTV spots and website display are live in InContext, while audio rests only on a demo link and a Spotify logo, and grade D would cap the cell at 3 even if audio counted." }, { "id": "eye-square-incontext", "vendor": "eye square", "product": "eye square InContext", "cohort": "creative", "dimension": "data_access", "first": 1, "first_grade": "D", "second": 1, "second_grade": "D", "second_sources": [ "ATT-EYESQUARE-S05", "ATT-EYESQUARE-S23", "ATT-EYESQUARE-S08", "ATT-EYESQUARE-S11", "ATT-EYESQUARE-S03" ], "second_reason": "Anchor 1: results come as scorecards, standardised reports or dashboards; the claimed 'API access' has no published or cited developer documentation and no creative tool, asset-management system or survey platform is named as an integration." }, { "id": "eye-square-incontext", "vendor": "eye square", "product": "eye square InContext", "cohort": "creative", "dimension": "production_scale", "first": 1, "first_grade": "D", "second": 2, "second_grade": "D", "second_sources": [ "ATT-EYESQUARE-S26", "ATT-EYESQUARE-S03", "ATT-EYESQUARE-S21" ], "second_reason": "Anchor 2 (derived): the 2020 benchmark counts 3,283 stimuli from in-context studies 2003-2019, a mean of about 200 assets a year, and a pre-2024 figure caps at 2; InContext Express's 'over 1000 previous tests' has no period and the January 2026 'nearly 13,000 analyzed advertisements' is not tied to InContext by method." }, { "id": "eye-square-incontext", "vendor": "eye square", "product": "eye square InContext", "cohort": "creative", "dimension": "us_availability", "first": 3, "first_grade": "D", "second": 3, "second_grade": "D", "second_sources": [ "ATT-EYESQUARE-S11", "ATT-EYESQUARE-S04", "ATT-EYESQUARE-S17", "ATT-V-EYESQUARE-S01", "ATT-V-EYESQUARE-S03" ], "second_reason": "Anchor 4 facts (sold in the US through eye square INC with US in-context consultants, and US production use in Snap-commissioned InContext studies that sampled US users, but no US norms or calibration data) are capped at 3 by the grade-D vendor sources." }, { "id": "eye-square-incontext", "vendor": "eye square", "product": "eye square InContext", "cohort": "creative", "dimension": "assurance", "first": 4, "first_grade": "A", "second": 4, "second_grade": "A", "second_sources": [ "ATT-EYESQUARE-S10", "ATT-EYESQUARE-S09" ], "second_reason": "Anchor 4: Austrian Standards' database lists eye square GmbH's ISO 20252:2019 certificate S 001201 (full-service market research, valid to 13 November 2026), which covers research-process quality but not the InContext attention metric or SEAL, and no audit of the method itself was found." }, { "id": "eye-square-incontext", "vendor": "eye square", "product": "eye square InContext", "cohort": "creative", "dimension": "method_disclosure", "first": 3, "first_grade": "D", "second": 3, "second_grade": "D", "second_sources": [ "ATT-EYESQUARE-S04", "ATT-EYESQUARE-S05", "ATT-EYESQUARE-S19", "ATT-EYESQUARE-S23" ], "second_reason": "Anchor 3: four of seven elements are published (construct, observed signals, calibration-target reference labels, and sample size with named panel recruiters), while model type is only partial ('thousands of clickworkers', no model named), validation is a bare '40-60px' figure and no recalibration cadence is given." }, { "id": "eye-square-incontext", "vendor": "eye square", "product": "eye square InContext", "cohort": "creative", "dimension": "privacy", "first": 3, "first_grade": "C", "second": 4, "second_grade": "C", "second_sources": [ "ATT-EYESQUARE-S17", "ATT-EYESQUARE-S18", "ATT-EYESQUARE-S05", "ATT-EYESQUARE-S16" ], "second_reason": "Anchor 4: the InContext platform participant policy (v1.0, March 2026) documents consent screen, pseudonymous identifiers, signals not collected, retention, GDPR and CCPA, and the eye-tracking policy documents Art. 6(1)(a) consent and the face-video data collected, though it conflicts with the marketing claim that only X/Y gaze coordinates leave the device; no independent product-scoped record." }, { "id": "lumen-spotlight", "vendor": "Lumen Research", "product": "Lumen SPOTLIGHT", "cohort": "creative", "dimension": "method_validity", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-H-S27", "ATT-LUMEN-S19", "ATT-LUMEN-S51" ], "second_reason": "Anchor 3: SPOTLIGHT measures attention directly with webcam eye tracking calibrated to each participant's eyes (stated accuracy about a 2.5cm circle), but no held-out agreement or error statistic against a reference is published, participant recruitment is described only as opt-in (150 per market in the YOC study), and calibration is not documented by environment." }, { "id": "lumen-spotlight", "vendor": "Lumen Research", "product": "Lumen SPOTLIGHT", "cohort": "creative", "dimension": "outcome_validation", "first": 2, "first_grade": "D", "second": 4, "second_grade": "B", "second_sources": [ "ATT-B-S09", "ATT-LUMEN-S51" ], "second_reason": "Anchor 4: an independent academic experiment (Simonov et al., funded by Imperial and the NET Institute, Lumen webcam eye tracking supplied in kind, 1,013 US/UK participants) found eye-tracked dwell associated with aided recall (memory) and a purchase proxy, but in-sample, not replicated and with no increment test over visibility; the YOC SPOTLIGHT study reports only recall against norms from an interested party.", "adjudication": { "verdict": "second pass upheld", "error_type": "anchor application", "published": 4, "reason": "The first pass excluded Simonov et al. by analogy with Lumen’s media model; SPOTLIGHT sells the eye-tracking measurement the study tested, so independent validation against memory (anchor 4) is met." } }, { "id": "lumen-spotlight", "vendor": "Lumen Research", "product": "Lumen SPOTLIGHT", "cohort": "creative", "dimension": "granularity", "first": 4, "first_grade": "C", "second": 4, "second_grade": "C", "second_sources": [ "ATT-LUMEN-S19", "ATT-G-S15", "ATT-LUMEN-S51" ], "second_reason": "Anchor 4: outputs include element-level diagnostics per asset (heat maps quantifying attention to ad elements, view order, feature analysis, gaze linked to on-screen HTML elements), but norms are documented only for the asset-level attention funnel, not at element resolution." }, { "id": "lumen-spotlight", "vendor": "Lumen Research", "product": "Lumen SPOTLIGHT", "cohort": "creative", "dimension": "creative_diagnostics", "first": 3, "first_grade": "D", "second": 4, "second_grade": "C", "second_sources": [ "ATT-LUMEN-S19", "ATT-G-S15", "ATT-LUMEN-S51" ], "second_reason": "Anchor 4: pre-launch testing with a published participant method (survey-embedded webcam eye tracking, stimulus in context then isolation, recall/perception/purchase-intent questions, attention funnel metrics, 150 respondents per market in the YOC study) benchmarked against Lumen norms whose base (number of assets or tests) and category or market breakdown are not stated." }, { "id": "lumen-spotlight", "vendor": "Lumen Research", "product": "Lumen SPOTLIGHT", "cohort": "creative", "dimension": "channel_coverage", "first": 3, "first_grade": "D", "second": 4, "second_grade": "C", "second_sources": [ "ATT-G-S15", "ATT-LUMEN-S19" ], "second_reason": "Anchor 5 met on the record (display, online video, TV/VOD, social, in-game and OOH testing offered, five or more of the seven channels), capped at 4 because the evidence is vendor documentation only." }, { "id": "lumen-spotlight", "vendor": "Lumen Research", "product": "Lumen SPOTLIGHT", "cohort": "creative", "dimension": "data_access", "first": 1, "first_grade": "D", "second": 1, "second_grade": "C", "second_sources": [ "ATT-G-S15", "ATT-LUMEN-S19" ], "second_reason": "Anchor 1: results are delivered as reports (attention funnel, heat maps, gaze plots, view order, questionnaire data); no named creative-tool or asset-management integration or documented API is cited." }, { "id": "lumen-spotlight", "vendor": "Lumen Research", "product": "Lumen SPOTLIGHT", "cohort": "creative", "dimension": "production_scale", "first": "n/e", "first_grade": "", "second": "n/e", "second_grade": "D", "second_sources": [ "ATT-LUMEN-S51" ], "second_reason": "No count of SPOTLIGHT assets or tests in any period or with a start date is documented; the YOC study is a single study (n=300) and norms are described only as drawn from various studies in the last two years, and LAMP session totals are not inherited." }, { "id": "lumen-spotlight", "vendor": "Lumen Research", "product": "Lumen SPOTLIGHT", "cohort": "creative", "dimension": "us_availability", "first": 1, "first_grade": "D", "second": 1, "second_grade": "B", "second_sources": [ "ATT-B-S09" ], "second_reason": "Anchor 1: the only US fact is a single academic study with US participants using Lumen webcam eye tracking; SPOTLIGHT pages do not document US availability, US norms or US clients, and the US panel and TVision CTV arrangements belong to Lumen's media products and are not inherited." }, { "id": "lumen-spotlight", "vendor": "Lumen Research", "product": "Lumen SPOTLIGHT", "cohort": "creative", "dimension": "assurance", "first": "n/e", "first_grade": "", "second": 1, "second_grade": "D", "second_sources": [ "ATT-H-S27", "ATT-C-S14" ], "second_reason": "Anchor 1: the commissioned PwC review covers Lumen's media attention models and campaign analysis, not the SPOTLIGHT testing method, so what remains is vendor self-attestation (eye tracking 'approved for use by academics', 'validated and audited' marketing) with no audit record scoped to SPOTLIGHT.", "adjudication": { "verdict": "second pass upheld", "error_type": "anchor application", "published": 1, "reason": "Lumen’s company-level ‘validated and audited’ statement covers the technology SPOTLIGHT uses but names an audit outside its scope: self-attestation (anchor 1), as Tobii’s academic-validation statement is scored." } }, { "id": "lumen-spotlight", "vendor": "Lumen Research", "product": "Lumen SPOTLIGHT", "cohort": "creative", "dimension": "method_disclosure", "first": 3, "first_grade": "D", "second": 3, "second_grade": "C", "second_sources": [ "ATT-H-S27", "ATT-LUMEN-S19", "ATT-LUMEN-S51" ], "second_reason": "Anchor 3: four of seven elements are published (construct, webcam gaze signals, sample size with opt-in recruitment, and the attention-funnel scoring formula), while reference labels, validation statistics and a recalibration cadence are not, and no data are released." }, { "id": "lumen-spotlight", "vendor": "Lumen Research", "product": "Lumen SPOTLIGHT", "cohort": "creative", "dimension": "privacy", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-H-S27", "ATT-LUMEN-S19", "ATT-LUMEN-S16" ], "second_reason": "Anchor 3: data minimisation is documented for the eye tracking (on-device processing, no video or PII uploaded, software deleted from respondents' devices after upload), but the participant consent model is described only as opt-in, no named legal framework is addressed for the product, and the privacy policy covers the website." }, { "id": "neurons-ai", "vendor": "Neurons", "product": "Neurons AI", "cohort": "creative", "dimension": "method_validity", "first": 3, "first_grade": "C", "second": 4, "second_grade": "C", "second_sources": [ "ATT-NEURONS-S02", "ATT-NEURONS-S03", "ATT-V-NEURONS-S01" ], "second_reason": "Anchor 4: vendor documentation names Tobii lab eye-tracking as the reference with sample and recruitment stated (30-40 per test, 8,000+ ads, 20,000+ participants; lab, online panels, third-party recruiters) and reports agreement statistics (CC/SIM/KLD) on a random train/validation split, but calibration is reported by modality (image, video, beta mixed media), not per channel, so (c) is unmet and grade C caps at 4 anyway." }, { "id": "neurons-ai", "vendor": "Neurons", "product": "Neurons AI", "cohort": "creative", "dimension": "outcome_validation", "first": 1, "first_grade": "D", "second": 1, "second_grade": "D", "second_sources": [ "ATT-NEURONS-S23", "ATT-NEURONS-S16", "ATT-NEURONS-S22" ], "second_reason": "Anchor 1: brand evidence is a vendor case (Teads brand lift, +64% awareness, +29% recall, three cells with an unoptimised comparator) with no sample, method or date, so anchor 2's sample-and-outcome test is not met; memory evidence validates the separate Memory model against Neurons' own recall test (100+ videos, 1,000+ participants, '90%' accuracy), not the attention measure; sales evidence is case-title claims only; nothing independent." }, { "id": "neurons-ai", "vendor": "Neurons", "product": "Neurons AI", "cohort": "creative", "dimension": "granularity", "first": 4, "first_grade": "C", "second": 4, "second_grade": "C", "second_sources": [ "ATT-NEURONS-S12", "ATT-NEURONS-S16", "ATT-NEURONS-S08", "ATT-V-NEURONS-S06" ], "second_reason": "Anchor 5 is met on vendor documentation (element-level AOI diagnostics per asset with element-level norms for images, and for video since the 13 May 2026 launch), but grade C caps the cell at 4; per-second video norms are not documented." }, { "id": "neurons-ai", "vendor": "Neurons", "product": "Neurons AI", "cohort": "creative", "dimension": "creative_diagnostics", "first": 4, "first_grade": "C", "second": 4, "second_grade": "C", "second_sources": [ "ATT-NEURONS-S12", "ATT-NEURONS-S13", "ATT-NEURONS-S14", "ATT-NEURONS-S16", "ATT-NEURONS-S02" ], "second_reason": "Anchor 5 is met on vendor documentation (published model design, sample and metrics; norms with a stated base of 10,000+ images and 7,000+ videos broken down by industry and use case, though the norms are model-predicted scores and the paper's own figures differ), capped at 4 by grade C." }, { "id": "neurons-ai", "vendor": "Neurons", "product": "Neurons AI", "cohort": "creative", "dimension": "channel_coverage", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-NEURONS-S13", "ATT-NEURONS-S16", "ATT-NEURONS-S06" ], "second_reason": "Anchor 3: four channels are live through benchmark use cases (display, online video, TV/CTV spots, social with YouTube counted as social per the vendor's labelling); OOH is static images only and DOOH, audio and in-game are not documented." }, { "id": "neurons-ai", "vendor": "Neurons", "product": "Neurons AI", "cohort": "creative", "dimension": "data_access", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-NEURONS-S06", "ATT-NEURONS-S07", "ATT-NEURONS-S11", "ATT-NEURONS-S29", "ATT-NEURONS-S28" ], "second_reason": "Anchor 3: a documented public REST API (Customers API v2 with OpenAPI reference) and a live Figma plugin are two qualifying routes; the third, a Chrome extension (v1.0.3, last updated 2022), is a browser tool rather than a creative tool or asset-management system, so anchor 5 (capped 4) is not reached." }, { "id": "neurons-ai", "vendor": "Neurons", "product": "Neurons AI", "cohort": "creative", "dimension": "production_scale", "first": 1, "first_grade": "C", "second": 1, "second_grade": "C", "second_sources": [ "ATT-NEURONS-S12", "ATT-NEURONS-S24", "ATT-NEURONS-S22" ], "second_reason": "Anchor 1: the only volume figures are a norm-base total (10,000+ images, 7,000+ videos) with no period and one undated client engagement (Mindshare Lithuania, 1,800+ ads), both totals with no period; no product-wide assets-per-year figure is given." }, { "id": "neurons-ai", "vendor": "Neurons", "product": "Neurons AI", "cohort": "creative", "dimension": "us_availability", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-V-NEURONS-S02", "ATT-NEURONS-S31" ], "second_reason": "Anchor 3: a US version of the 2026 Neurons terms (Texas law, Travis County venue, US Central Time CSM support) and an Austin office reported in April 2025 document US availability, but neither US calibration data nor named US clients or campaigns are documented; the contracting entity is the Danish Neurons Inc ApS." }, { "id": "neurons-ai", "vendor": "Neurons", "product": "Neurons AI", "cohort": "creative", "dimension": "assurance", "first": 1, "first_grade": "D", "second": 1, "second_grade": "D", "second_sources": [ "ATT-NEURONS-S17", "ATT-NEURONS-S18", "ATT-NEURONS-S20" ], "second_reason": "Anchor 1: the vendor asserts a science-validated method and SOC 2 Type II certification (report on request, auditor and period not public), but SOC 2 covers security controls, not the testing method, and no named independent audit of the method exists." }, { "id": "neurons-ai", "vendor": "Neurons", "product": "Neurons AI", "cohort": "creative", "dimension": "method_disclosure", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-NEURONS-S02", "ATT-NEURONS-S03", "ATT-V-NEURONS-S01" ], "second_reason": "Anchor 3: six of seven elements are published in vendor documentation (construct, eye-tracking signals, reference heatmap labels, sample and recruitment, named model architecture, CC/SIM/KLD validation statistics), but no update or recalibration cadence is given and no data or code is released." }, { "id": "neurons-ai", "vendor": "Neurons", "product": "Neurons AI", "cohort": "creative", "dimension": "privacy", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-NEURONS-S19", "ATT-NEURONS-S18", "ATT-NEURONS-S20" ], "second_reason": "Anchor 3: product-specific data handling is documented (a GDPR DPA for the Neurons Platform and API with data categories and sub-processors, Neurons AI terms, models built on anonymised data), but the panel/participant consent model is documented only in the 2019 company-wide privacy policy, and client uploads are used for ML and benchmarking by default with opt-out, so anchor 4's product-specific consent model is not met." }, { "id": "tobii-sticky", "vendor": "Tobii", "product": "Tobii Sticky", "cohort": "creative", "dimension": "method_validity", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-TOBII-S04", "ATT-TOBII-S05", "ATT-TOBII-S03", "ATT-TOBII-S13" ], "second_reason": "Anchor 3: webcam gaze is calibrated and checked against known on-screen validation points with a published mean-validation-error grading, but no held-out agreement or error statistic for the current algorithm against a lab or infrared reference tracker, with reference sample and recruitment stated, is documented (the 2014 Burton/Albert/Flynn paper page returned 403 and pre-dates the current algorithm, so it is not relied on for more)." }, { "id": "tobii-sticky", "vendor": "Tobii", "product": "Tobii Sticky", "cohort": "creative", "dimension": "outcome_validation", "first": 1, "first_grade": "C", "second": 1, "second_grade": "C", "second_sources": [ "ATT-TOBII-S02", "ATT-TOBII-S08", "ATT-TOBII-S16" ], "second_reason": "Anchor 1: the only outcome statements for Sticky's own attention measures are help-centre vendor claims (longer viewing 'often' correlating with likeability and recall); memory: vendor claim only; brand: none; sales: none; the emotion white paper's figures are third-party (Unruly, Affectiva) and the 2021 P&G/Dentsu case study reports no outcome data." }, { "id": "tobii-sticky", "vendor": "Tobii", "product": "Tobii Sticky", "cohort": "creative", "dimension": "granularity", "first": 4, "first_grade": "C", "second": 4, "second_grade": "C", "second_sources": [ "ATT-TOBII-S02", "ATT-TOBII-S20", "ATT-D-S38" ], "second_reason": "Anchor 4: element-level AOI diagnostics per asset (Seen by %, Time until noticed, Time viewed, Seen first, Revisited, time-interval filtering, moving-AOI tracking, second-based emotion curves), compared only against fixed rule-of-thumb ranges with no stated base rather than norms at that resolution; grade C also caps at 4." }, { "id": "tobii-sticky", "vendor": "Tobii", "product": "Tobii Sticky", "cohort": "creative", "dimension": "creative_diagnostics", "first": 3, "first_grade": "C", "second": 4, "second_grade": "C", "second_sources": [ "ATT-D-S38", "ATT-TOBII-S12", "ATT-TOBII-S05", "ATT-TOBII-S02" ], "second_reason": "Anchor 4: pre-launch testing of ads, video, websites and packaging with a published method (dot calibration, session quality thresholds, about 30 usable sessions, formula-level metrics), with comparison ranges published but no norm database with a stated base of assets or tests or a category or market breakdown." }, { "id": "tobii-sticky", "vendor": "Tobii", "product": "Tobii Sticky", "cohort": "creative", "dimension": "channel_coverage", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-TOBII-S12", "ATT-TOBII-S03", "ATT-D-S38", "ATT-TOBII-S15", "ATT-TOBII-S16" ], "second_reason": "Anchor 3: three channels live in help documentation (display, online video and simulated social placements); no TV/CTV-spot, in-game or DOOH testing documented, and supported stimuli include no audio-only format; packaging, websites and shelf are attributes only." }, { "id": "tobii-sticky", "vendor": "Tobii", "product": "Tobii Sticky", "cohort": "creative", "dimension": "data_access", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-TOBII-S10", "ATT-TOBII-S11", "ATT-D-S38" ], "second_reason": "Anchor 3: a published REST API documentation is the one live output route; the six survey-platform snippets (including Alchemer and Qualtrics) return only completion status and participant ID, and no creative-tool or asset-management integration is documented." }, { "id": "tobii-sticky", "vendor": "Tobii", "product": "Tobii Sticky", "cohort": "creative", "dimension": "production_scale", "first": 1, "first_grade": "D", "second": 1, "second_grade": "D", "second_sources": [ "ATT-D-S38" ], "second_reason": "Anchor 1: the only volume figure is the product page's undated '10,000+ studies run globally', a total with no period or start date that counts studies, not assets." }, { "id": "tobii-sticky", "vendor": "Tobii", "product": "Tobii Sticky", "cohort": "creative", "dimension": "us_availability", "first": 3, "first_grade": "C", "second": 4, "second_grade": "C", "second_sources": [ "ATT-V-TOBII-S02", "ATT-V-TOBII-S01", "ATT-D-S38", "ATT-TOBII-S09" ], "second_reason": "Anchor 4: generally available in the US (US and Canadian clients contract with Tobii Technology Inc, Reston, under the 20 March 2026 terms) with US participants offered from the panel list (USA among 35 countries), but no named US study, campaign or client is documented as US production use." }, { "id": "tobii-sticky", "vendor": "Tobii", "product": "Tobii Sticky", "cohort": "creative", "dimension": "assurance", "first": 1, "first_grade": "D", "second": 1, "second_grade": "D", "second_sources": [ "ATT-D-S38", "ATT-V-TOBII-S04" ], "second_reason": "Anchor 1: the product page asserts 'academic validation' of accuracy without naming a study or auditor, and the ISO/IEC 27001:2022 certificate held by Tobii AB covers company information security, not Sticky's testing method, panel or fieldwork, so it earns no method-assurance credit." }, { "id": "tobii-sticky", "vendor": "Tobii", "product": "Tobii Sticky", "cohort": "creative", "dimension": "method_disclosure", "first": 3, "first_grade": "C", "second": 3, "second_grade": "C", "second_sources": [ "ATT-TOBII-S04", "ATT-TOBII-S03", "ATT-TOBII-S12", "ATT-TOBII-S10" ], "second_reason": "Anchor 3: five of seven elements are published (construct, signals, reference labels as known calibration/validation points, sample guidance and recruitment, formula-level scoring procedure), while validation statistics for the current algorithm and an update or recalibration cadence are not." }, { "id": "tobii-sticky", "vendor": "Tobii", "product": "Tobii Sticky", "cohort": "creative", "dimension": "privacy", "first": 4, "first_grade": "C", "second": 4, "second_grade": "C", "second_sources": [ "ATT-TOBII-S09", "ATT-V-TOBII-S03", "ATT-TOBII-S02", "ATT-V-TOBII-S04" ], "second_reason": "Anchor 4: the Sticky data-protection white paper documents a product-specific consent model and data minimisation (face/head framing, no direct identifiers, scrubbed reports, recordings deleted within 90 days, opt-in emotion analysis) and names GDPR, but the only certification (Tobii AB ISO/IEC 27001) does not name Sticky, so no product-scoped independent record supports 5." } ], "bands": [ { "id": "adelaide", "vendor": "Adelaide", "product": "Adelaide AU", "cohort": "media", "first_q": 53, "first_band": "Strong", "second_q": 57, "second_band": "Strong" }, { "id": "tvision", "vendor": "TVision", "product": "TVision", "cohort": "media", "first_q": 53, "first_band": "Strong", "second_q": 55, "second_band": "Strong" }, { "id": "eye-square-incontext", "vendor": "eye square", "product": "eye square InContext", "cohort": "creative", "first_q": 54, "first_band": "Strong", "second_q": 58, "second_band": "Strong" }, { "id": "lumen-spotlight", "vendor": "Lumen Research", "product": "Lumen SPOTLIGHT", "cohort": "creative", "first_q": 53, "first_band": "Strong", "second_q": 56, "second_band": "Strong" }, { "id": "neurons-ai", "vendor": "Neurons", "product": "Neurons AI", "cohort": "creative", "first_q": 53, "first_band": "Strong", "second_q": 55, "second_band": "Strong" }, { "id": "tobii-sticky", "vendor": "Tobii", "product": "Tobii Sticky", "cohort": "creative", "first_q": 53, "first_band": "Strong", "second_q": 55, "second_band": "Strong" } ], "both_scored": 64 }