Métriques et reporting
Métriques
 Updated July 2026 | Adelaide |  |  |  | xpln.ai |
|---|---|---|---|---|---|
| Which key attention metrics do you report on? | AU (Attention Unit) | In View Stride Score Engagement Score Viewport Size Score Attention Score Attention Score Measured Attention Rate Maximum Possible (Attention) Score | DV Attention Index Exposure Index Engagement Index | Quality Attention Score Visibility Metrics Situation Interaction Attention metrics % viewed, viewing time, Attentive seconds and Attentive seconds per 1000 impressions (APM) Buying metrics aCPM, rCPM | Attention Time Attention Rate (n)s Attention Time on Attentive Impressions Cost per Attentive Second (CPaS) aCPM: Attentive CPM |
| When applicable, do you report metrics - With MRC threshold (you discount impressions generating attention which are viewed less than one continuous second for display and two for video) (Yes/No) - Without MRC threshold (Yes/No) | N/A | - Yes - Yes | DV scores both IAB viewable and non-IAB viewable impressions. Users can choose to analyze the data holistically or view each category separately. | Without MRC thresholds: the attention calculations start at 100ms | - Yes - Yes |
| Are the results based on actual observations of the campaign or on a predictive score calculated by a model? (Yes/No) | On a predictive score calculated by a model which includes actual campaign observations | Yes, on actual observations but not on predictive score (for open web), using a predictive score and modelling for walled gardens. | It depends on the measured environment. Open Web (Desktop, Mobile, In-App): Results are based on real observations (deterministic), leveraging analysis of more than 50 data points collected in real time by the DV tag directly on the user’s device, combined with predictive modeling to improve accuracy where direct signals may not be available. Social and CTV (Modeled): On closed platforms and Connected TV, results are model-based. DV combines data measured on the device, feeds it into trusted third-party partner managed eye-tracking panels to model and estimate attention across these environments. Activation and Optimization: Separately from measurement, DV offers power pre-bid optimization using Attention data to optimize media buying toward higher-attention inventory. This is an activation capability — it does not influence how attention is measured or scored. | Yes | On a predictive score calculated by a model which includes actual campaign observations |
| Do you provide attention Benchmarks (average) (by media / format /country) allowing users to evaluate their campaign performance? (Yes/No) | Yes | Yes | Yes | Yes | Yes |
| If yes, which data are they based on? (size of database, countries covered and recency) | Currently by media channel and placement. At this time we do not report country-level benchmarks. | - Size of database: The benchmark is calculated using the compiled and weighted average of attention scores from all campaigns across Adloox's client base (more than 300 brands and a large volume of daily impressions). - Countries covered: The benchmark covers more than 200 countries worldwide. - Recency: The benchmark is continuously updated and calculated as a rolling average, available over 3-month windows based on a 13-month data retention framework. Within this window, end users can break down the data by day, week, or any custom period of their choosing. | DV measures billions (>65bn) of impressions every month to maintain our global attention benchmarks. DV provides Industry Benchmarks where clients can compare performance against 11 different industry verticals with filtering by buying channel, region, market, device delivery, media type, ad size, and ad duration for apples-to-apples comparison. DV Attention dashboards are updated hourly, and the Industry Benchmarks dashboard is updated monthly. | Total dataset: 1m+ individual user sessions, data on over 5m ads, 6bn gaze points; data collected from 37 countries; entire data database used to create models, with dataset updated periodically as additional data collected | Over 150BN impressions, across 50+ markets, rolling average of last 12 months, updated twice a year. |
| Do you predict business outcomes (brand lift, conversions…) based on attention measurement? (Yes/No) | Yes | No | DV does not predict business outcomes; however, Authentic Attention has been correlated to real business outcomes with our advertiser clients. | Yes | Yes |
Score
If you provide an Attention Score
Updated July 2026 | Adelaide | | | |
|---|---|---|---|---|
| Is the score expressed on a universal scale of 0 to 100 or as an index (Base 100)? (0-100 / Base 100 Index / Other) | 0-100 | 0-100 | Base 100 Index | A unified predictive score ranging from 1 to 100, calculated using an advanced machine learning model that weights deterministic media quality signals (Visibility, Context, and Interaction) alongside probabilistic biometric eye-tracking data from Lumen Research. |
| Do you report a composite Attention score? (Yes/No) | No | Yes | Yes | Yes |
| Does the distribution of weighting among the different score components vary according to the format, device, or campaign objective? (Yes/No) | Yes | No | N/A | Yes |
| What is the average percentage weighting attributed to Viewability/Exposure in the calculation of your score? (Provide a number in %) | Adelaide does not make its score component weights available outside product audit contexts. | Not willing to share this information. | N/A | IAS does not wish to share this information |
| Do user interactions (scroll, cursor, etc.) represent more than 25% of the total score weighting? (Yes/No) | Adelaide does not make its score component weights available outside product audit contexts. | Not willing to share this information. | DV indices are not weighted. User interactions are reported under the Engagement index, which analyzes key user-initiated events that occur while the ad creative is displayed, including user touches, screen orientation, video playback, and audio control interactions. | IAS does not wish to share this information. These interactions are represented in the % of pixels viewable + viewable time factors |