Méthodologie de l'Attention
Principes méthodologiques
 Updated July 2026 | Adelaide |  |  |  | xpln.ai |
|---|---|---|---|---|---|
| Does your Attention solution rely on a Data Signals approach (which uses servers/SDKs for Time-in-View, Scroll Depth, clicks, publisher metadata, audibility, etc.)? (Yes/No) | Yes | Yes | Yes | Yes | Yes |
| Does your Attention solution utilize a Visual Tracking/Biometric approach (Eye Tracking, cameras)? (Yes/No) | Yes | No | For the open internet (websites and standard applications), the measurement is based on direct measured signals. However, for certain specific environments where tags cannot be deployed (walled gardens, connected TV), DoubleVerify supplements these methodologies with partners that use on partners that use panel-based methodologies and visual tracking technologies | Yes | Yes |
| Does your solution incorporate Panel and Survey-Based Methods? (Yes/No) | Yes | No | For the open internet (websites and standard applications), the measurement is based on direct measured signals. It relies on the DoubleVerify tag, which collects real-time signals (touch, scroll, viewability, etc.) without using panels or surveys to generate the attention score. However, for certain specific environments where tags cannot be deployed (walled gardens, connected TV), DoubleVerify supplements these methodologies with partners that use on partners that use panel-based methodologies and visual tracking technologies. | Yes | Yes |
| Does your Attention solution rely on Predictive Modeling (Machine learning, feature engineering / AI)? (Yes/No) | Yes | No | DV Attention performs direct measurement combined with predictive modeling to increase coverage only in environments where direct signals are not available. | Yes | Yes |
| Does your Attention solution cover both web and in-app or web only? (Both, web only)? | Both | Both | Both | Both | Both |
| What devices do you measure (e.g. mobile, desktop, smartTV, CTV/OTT, gaming consoles…) List devices | 1. Mobile 2. Desktop 3. CTV/OTT 4. Television 5. Radio 6. DOOH/OOH 7. Cinema | - Desktop - Mobile (Mobile web and in-app environments) - CTV | Mobile (web and in-app) Desktop CTV/OTT | Desktop Mobile and Tablet (web and In app) CTV/ OTT | All devices |
| How do you ensure that your Attention measurement solution is GDPR compliant? | Adelaide designs and operates its platform in alignment with the principles of the General Data Protection Regulation (GDPR), including data minimization, purpose limitation, security, and privacy by design. Adelaide does not collect, process, or store personal data or personally identifiable information (PII) as defined under GDPR. All data processed is anonymized and aggregated prior to use and cannot be used to identify any individual. As a result, Adelaide’s services generally operate outside the scope of GDPR’s personal data processing requirements. | Yes, AdLoox whole verification solution is GDPR compliant. | Attention measurement does not require or use any PII data. It is a impression-based measurement that doesn't involve tracking users. Attention measures exposure and engagement of the creative actions once loaded. DoubleVerify has a dedicated privacy team that is responsible for maintaining privacy controls that are required as part of local regulations and contractual obligations. DV holds an industry-leading set of certifications, including: ISO/IEC 27001 (Security Management) ISO/IEC 27701 (Privacy Extension) SOC 2 Type II APEC CBPR & APEC PRP Privacy Frameworks TRUSTe International Privacy Verification TrustArc Responsible AI Governance Certification — DV was the first organisation in the industry to achieve this Compliance with GDPR, CCPA/CPRA, LGPD, PIPEDA, VCDPA, and EU-US/UK/Swiss Data Privacy Frameworks For more information please visit trust.doubleverify.com | Solution is fully GDPR compliant https://integralads.com/ias-legal-portal/ | xpln.ai is privacy-by-design: no PII, no cookies, no device IDs, no persistent identifiers collected at scale. Compliant with GDPR, UK GDPR, and CCPA/CPRA. TCF v2 member (Vendor ID 1142). Dedicated DPO. Panel respondents are opted-in and explicitly consented. Infrastructure hosted on Google Cloud Platform (ISO 27001, SOC 2 Type II); AES-256 at rest, TLS 1.2+ in transit; RBAC, MFA, annual penetration tests. Privacy policy: xpln.ai/privacy. |
Signaux de données
Updated July 2026 | Adelaide | | | | xpln.ai |
|---|---|---|---|---|---|
| If you use data signals in your Attention measurement, please specify which ones? | Viewability Share of screen Clutter (number of ads on screen) Sound on/off Ad placement/position Ad format (video/static) Completion (for video) User interaction Context quality Content type (news, sports…) Day part | Viewability Share of screen Sound on/off Ad placement/position Ad format (video/static) Completion (for video) User interaction Context quality | Viewability Share of screen Clutter (number of ads on screen) Sound on/off Ad placement/position Ad format (video/static) Completion (for video) User interaction Context quality Content type (news, sports…) | Viewability Share of screen Clutter (number of ads on screen) Sound on/off Ad placement/position Ad format (video/static) Completion (for video) User interaction Context quality Content type (news, sports…) Day part |
Tracking visuel
Updated July 2026 | Adelaide | | | | xpln.ai |
|---|---|---|---|---|---|
| If you use Visual Tracking / Biometric Tracking in your Attention measurement, please specify the technology you use? | Infrared glasses Infrared bar on computer (partner data) Webcam | N/A | Infrared glasses (for OOH and POS inventory data collection) Infrared bar on computer (in the past, only for desktop advertising) Webcam | Infrared glasses Infrared bar on computer Webcam (Proprietary webcam eyetracking technology) |
|
| What is the technical threshold you use to start counting attention? | Adelaide partners may use a minimum saccades threshold approach | N/A | DV is compliant with the standard introduced in the IAB and MRC Attention Guidelines, which requires that some portion of the ad be visible for any non-zero time (and non-mute audible for audio). In addition to this, DV also requires that the ad be fraud free, brand suitable, and served in the intended geo area. | 100ms (i.e. 1/10th of a second) | Technical threshold used is 0.2 seconds |
| Is the duration of attention calculated by summing discontinuous periods? (Yes/No) | Yes (partner data) | N/A | DV Attention Measurement uses a combination of both consecutive and cumulative (discontinuous) signals. | Yes, but only on the same ad in the same viewable session. If the ad is on screen continuously, but the gaze is split across 2 discontinuous sessions, then the aggregate of the 2 sessions is reported; if the ad is viewable twice (e.g. people scroll down past the ad and then scroll up, revealing it for a second time) then the gazefor the two sessions is not summed | Yes |
| Do you report results from visual tracking or is visual tracking used to develop your model or as one of the inputs of your model? (Yes/No) | Yes visual tracking is used to develop our model but we do not report visual tracking results | No | No for the core solution on the Open Web, and Yes for Social and CTV solutions via partners. | Yes | Yes |
| If you use a Predictive Modeling in your Attention measurement, what input data is used to develop and train your predictive model? If yes, what is the size of the panels you used to develop the model (number of panelists)? | User Interaction and Behavior Signals (scroll, clicks…) Exposure/Viewability Measures Metadata describing the ad Device Panel / Survey Each of our 19 channel models is built from a unique combination of data ingredients. Outcomes data provider panels vary in size. TVision panel is 5000 HHs in US/approx 15k respondents + 500 HHs/approx 750 respondents in UK (gaze only). Viomba panel is approx 1000 panelists, Total number of panelists from all studies used is approx 20k individuals in total. | User Interaction and Behavior Signals (scroll, clicks…) Exposure/Viewability Measures Device | The models rely on data from DV direct measured signals, expanding the same quality of insights as that gained from the deterministic Attention measurement. User Interaction and Behavior Signals (scroll, clicks…) Exposure/Viewability Measures Metadata describing the ad Device Panel / Survey 5000 households for CTV, 750,000+ eye-tracking sessions for social | User Interaction and Behavior Signals (scroll, clicks…) Exposure/Viewability Measures Metadata describing the ad Device Panel / Survey Eye Tracking Real conversations Context quality Dataset composition: - 750,000+ eye-tracking sessions across individuals globally - Thousands of inlab studies across Europe, Middle- East, APAC and US, LATAM, in english and local languages, across social, openweb, CTV, OOH, DOOH, Cinema, Retail Media - Thousands of creative tests through our eye-tracking proprietary technology | User Interaction and Behavior Signals (scroll, clicks…) Exposure/Viewability Measures Metadata describing the ad Device Panel / Survey +100,000 cumulated panellists across 14 markets (may 2026) |
Modélisation et validation
Updated July 2026 | Adelaide | | | | xpln.ai |
|---|---|---|---|---|---|
| If you use a Predictive Modeling in your Attention measurement, what input data is used to develop and train your predictive model? If yes, what is the size of the panels you used to develop the model (number of panelists)? | User Interaction and Behavior Signals (scroll, clicks…) Exposure/Viewability Measures Metadata describing the ad Device Panel / Survey Each of our 19 channel models is built from a unique combination of data ingredients. Outcomes data provider panels vary in size. TVision panel is 5000 HHs in US/approx 15k respondents + 500 HHs/approx 750 respondents in UK (gaze only). Viomba panel is approx 1000 panelists, Total number of panelists from all studies used is approx 20k individuals in total. | User Interaction and Behavior Signals (scroll, clicks…) Exposure/Viewability Measures Device | The models rely on data from DV direct measured signals, expanding the same quality of insights as that gained from the deterministic Attention measurement. User Interaction and Behavior Signals (scroll, clicks…) Exposure/Viewability Measures Metadata describing the ad Device Panel / Survey 5000 households for CTV, 750,000+ eye-tracking sessions for social | User Interaction and Behavior Signals (scroll, clicks…) Exposure/Viewability Measures Metadata describing the ad Device Panel / Survey Eye Tracking Real conversations Context quality Dataset composition: - 750,000+ eye-tracking sessions across individuals globally - Thousands of inlab studies across Europe, Middle- East, APAC and US, LATAM, in english and local languages, across social, openweb, CTV, OOH, DOOH, Cinema, Retail Media - Thousands of creative tests through our eye-tracking proprietary technology | User Interaction and Behavior Signals (scroll, clicks…) Exposure/Viewability Measures Metadata describing the ad Device Panel / Survey +100,000 cumulated panellists across 14 markets (may 2026) |
| How many observations is your training data set based on? | Adelaide uses approx 1m survey-based data points and over 100m mid- and lower-funnel data points to establish our training set. | Each impression is considered in AdLoox validation model | DV measures billions (>65bn) of impressions every month for attention measurement. | The training data set for the IAS Quality Attention predictive model is based on a massive foundation comprising billions of impressions and millions of conversions. Additionally, the research and models are refined using millions of advertising data signals. | 7M to 8M observations |
| Which country.ies do these observations come from? | Channel models include gaze data from TVision (US and UK) and Viomba (US, Finland, Germany, Denmark). Outcomes data are generated in the countries where our measured campaigns run. (Our conversion and survey training is more broad) | All countries covered by AdLoox JS tracking tech | DV Attention measurement works globally, across virtually every country. | All markets | USA, UK, FR, IT, ES, DE, BE, NL, CH, AT, BR, JP, SE, CA |
| Have you performed a cross-validation of your predictive model to prove its reliability? (Yes/No) | Yes | N/A | Yes. DV regularly validates their models, and validate reliability of attention measurement with individual clients | Yes: validation exercises have been performed by PwC and Imperial College, University of London | Yes |
Médias couverts
Updated July 2026 | Adelaide | | | | xpln.ai |
|---|---|---|---|---|---|
| Is your solution designed and equipped to measure advertising attention on Linear TV: If yes, specify the different methodologies employed for measurement: | Yes Data Signals Visual Tracking Panel and Survey-Based Methods Predictive Models | No | No | No | No |
| Is your solution designed and equipped to measure advertising attention on Connected TV (replay of linear TV and SVOD): If yes, specify the different methodologies employed for measurement: | Yes Data Signals Visual Tracking Panel and Survey-Based Methods Predictive Models | Yes Data Signals | Yes Data Signals Visual Tracking Panel and Survey-Based Methods Declined to comment for Predictive Models | Yes Data Signals Visual Tracking Predictive Models | Yes Data Signals Visual Tracking Predictive Models |
| Is your solution designed and equipped to measure advertising attention on Facebook: 1 - If so, do you have a contractual partnership with Facebook that notably gives you access to granular data for the calculation of metrics? 2 - If so, what type of platform reporting access do you use to calculate attention metrics? If yes, specify the different methodologies employed for measurement: | Yes 1 - No (not today, but we expect to soon) | Yes 1 - No Data Signals Predictive Models (through Ad Exposure Measurement Studies) | No | Yes 1 - Yes 2- Platform reporting API Data Signals Visual Tracking Predictive Models | Yes 1 - No 2 - Platform reporting API and Platform reporting interfaces or scheduled data exports Data Signals Visual Tracking Predictive Models |
| Is your solution designed and equipped to measure advertising attention on Instagram: 1 - If so, do you have a contractual partnership with Instagram that notably gives you access to granular data for the calculation of metrics? 2 - If so, what type of platform reporting access do you use to calculate attention metrics? If yes, specify the different methodologies employed for measurement: | Yes 1 - No (not today, but we expect to soon) Data Signals Visual Tracking Panel and Survey-Based Methods Predictive Models | Yes 1 - No Data Signals Predictive Models (through Ad Exposure Measurement Studies) | Declined to comment | Yes 1 - Yes 2- Platform reporting API Data Signals Visual Tracking Predictive Models | Yes 1 - No 2 - Platform reporting API and Platform reporting interfaces or scheduled data exports Data Signals Visual Tracking Predictive Models |
| Is your solution designed and equipped to measure advertising attention on Snapchat: 1 - If so, do you have a contractual partnership with Snapchat that notably gives you access to granular data for the calculation of metrics? 2 - If so, what type of platform reporting access do you use to calculate attention metrics? If yes, specify the different methodologies employed for measurement: | Yes 1 - No Data Signals Visual Tracking Panel and Survey-Based Methods Predictive Models | Yes Predictive Models (through Ad Exposure Measurement Studies) | Yes 1 - https://doubleverify.com/company/newsroom/doubleverify-debuts-first-of-its-kind-attention-measurement-for-social-launching-with-snap Data Signals Visual Tracking Panel and Survey-Based Methods Predictive Models | Yes 1 - Yes 2- Platform reporting API Data Signals Visual Tracking Predictive Models | Yes 1 - No 2 - Platform reporting API and Platform reporting interfaces or scheduled data exports Data Signals Visual Tracking Predictive Models |
| Is your solution designed and equipped to measure advertising attention on TikTok: 1 - If so, do you have a contractual partnership with TikTok thatnotably gives you access to granular data for the calculation of metrics? 2 - If so, what type of platform reporting access do you use to calculate attention metrics? If yes, specify the different methodologies employed for measurement: | Yes 1 - No Data Signals Visual Tracking Panel and Survey-Based Methods Predictive Models | Yes Predictive Models (through Ad Exposure Measurement Studies) | Yes 1 - https://doubleverify.com/company/newsroom/doubleverify-expands-dv-authentic-attention-to-tiktok-as-the-first-badged-tiktok-marketing-partner-for-impression-level-attention-insights Data Signals | Yes 1 - Yes 2- Platform reporting API Data Signals Visual Tracking Predictive Models | Yes 1 - No 2 - Platform reporting API and Platform reporting interfaces or scheduled data exports Data Signals Visual Tracking Predictive Models |
| Is your solution designed and equipped to measure advertising attention on X: 1 - If so, do you have a contractual partnership with X that notably gives you access to granular data for the calculation of metrics? 2 - If so, what type of platform reporting access do you use to calculate attention metrics? If yes, specify the different methodologies employed for measurement: | Yes 1 - No Data Signals Visual Tracking Panel and Survey-Based Methods Predictive Models | No | No | Yes 1 - Yes 2- Platform reporting API Data Signals Visual Tracking Predictive Models | Yes 1 - No 2 - Platform reporting API and Platform reporting interfaces or scheduled data exports Data Signals Visual Tracking Predictive Models |
| Is your solution designed and equipped to measure advertising attention on YouTube: 1 - If so, do you have a contractual partnership with YouTube that notably gives you access to granular data for the calculation of metrics? 2 - If so, what type of platform reporting access do you use to calculate attention metrics? If yes, specify the different methodologies employed for measurement: | Yes 1 - Yes 2 - Platform reporting interfaces or scheduled data exports Data Signals Visual Tracking Panel and Survey-Based Methods Predictive Models | Yes Data Signals Predictive Models (through Ad Exposure Measurement Studies) | Declined to comment | Yes 2- Platform reporting API Data Signals Visual Tracking Predictive Models | Yes 1 - No 2 - Platform reporting API and Platform reporting interfaces or scheduled data exports Data Signals Visual Tracking Predictive Models |
| Is your solution designed and equipped to measure advertising attention on Open Web: If yes, specify the different methodologies employed for measurement: | Yes Data Signals Visual Tracking Panel and Survey-Based Methods Predictive Models | Yes Data Signals | Yes Data Signals Predictive Models | Yes Data Signals Visual Tracking Predictive Models | Yes Data Signals Visual Tracking Predictive Models |
| Is your solution designed and equipped to measure advertising attention on Terrestrial Radio: If yes, specify the different methodologies employed for measurement: | Yes Data Signals Panel and Survey-Based Methods Predictive Models | No | No | No | No |
| Is your solution designed and equipped to measure advertising attention on Digital Audio: If yes, specify the different methodologies employed for measurement: | Yes Data Signals Panel and Survey-Based Methods Predictive Models | No | No | No | No |
| Is your solution designed and equipped to measure advertising attention on Print: If yes, specify the different methodologies employed for measurement: | No | No | No | No | No |
| Is your solution designed and equipped to measure advertising attention on OOH: If yes, specify the different methodologies employed for measurement: | No | No | No | No | No |
| Is your solution designed and equipped to measure advertising attention on DOOH: If yes, specify the different methodologies employed for measurement: | Yes Data Signals Visual Tracking Panel and Survey-Based Methods Predictive Models | No | No | No | No |
| Is your solution designed and equipped to measure advertising attention on Cinema: If yes, specify the different methodologies employed for measurement: | Yes Data Signals Visual Tracking Panel and Survey-Based Methods Predictive Models | No | No | No | No |
Research partners
Updated July 2026 | Adelaide | | | | xpln.ai |
|---|---|---|---|---|---|
| Do you have research partners to measure attention? (Yes/No) If yes, which ones? | Yes Viomba, Tobii, Tvision, Vertical Impressions, Amazon Ads, Attain, Kantar, Microsoft Advertising, Nielsen, Quantcast, Cint Group, DISQO, Dynata, Foursquare, Ipsos, Lucid, OnDevice, Quantcast, Upwave, Attain, IRI, Data Plus Math (LiveRamp) | No | Yes - Lumen - Tvision - Kantar - Upwave - Eye Square - Attain | Yes - Lumen Research - Snap Inc - YouGov | Yes |