For years, viewability was the baseline: a display ad just had to have half its pixels on screen for one continuous second. Attention metrics change the question from “could this be seen?” to “did a person actually look?”
That shift is no longer a measurement side conversation. In 2026, attention scores are moving into the bidstream, with Amazon DSP and Google’s DV360 now accepting Adelaide’s Attention Unit (AU) as a pre-bid signal. If you run programmatic, this changes how you should think about inventory quality.
The weakness viewability leaves open
Viewability records a condition of the ad. Attention tries to measure a behaviour of the audience. The Media Rating Council and IAB made that distinction explicit in guidelines finalised on November 12, 2025. Under the draft released in May 2025, measurement splits into four methods, plus predictive models.
- Data signals: time in view, scroll depth, share of screen, ad density, audibility and interactions.
- Visual tracking: eye tracking, gaze tracking and facial coding among consenting panellists.
- Physiological and neurological observation: heart rate variability, skin conductance and EEG.
- Panel and survey methods: asking people what they remember.
- Predictive models: trained on deterministic data to score every impression without a live panel.
Predictive models are the ones that reached bidding systems first. A panel can observe a sample, but only a model can attach a score to every bid request. Lumen Research and Adelaide typify that approach: Lumen trains on eye-tracking data, while Adelaide’s AU predicts attention likelihood from signals like ad size, time in view, clutter and position.
Attention is moving into the transaction
Attention touches a campaign before, during and after the buy. Planners use placement benchmarks to shortlist inventory. Bidders use scores as filters or Custom Bidding signals. Verification tags report delivered attention afterwards.
Recent integrations show the momentum:
- Adelaide’s Amazon DSP segments launched on June 10, 2026, tiering inventory into high, average and low attention bands and removing the bottom 10% of AU-scored placements.
- In DV360, AU became a Custom Bidding signal in mid-July 2026, so an advertiser can value each impression against its attention score.
- InMobi launched FRAME Attention on July 24, 2026, with an 8,000 APM floor across in-app video and CTV.
That APM floor is useful shorthand: 8,000 APM equals eight attentive seconds per impression on average. InMobi’s Attention Per Mille multiplies share of impressions viewed by average view time and normalises to 1,000 impressions. It turns a cheap banner and an expensive CTV spot into a shared “attentive seconds” unit.
What to do before you lean in
The guidelines standardise vocabulary and disclosure, not methodology. A score of 60 in one vendor’s system is not a 60 in another. DoubleVerify reports an index, Adelaide reports a placement score, and InMobi reports volume in seconds. That means you cannot port benchmarks blindly across vendors.
The link to sales is also contested. A Kantar analysis of 873 campaigns and $3.2 billion in media spend found overall channel attention levels did not directly correlate with brand contribution or cost-effectiveness. Kroger Precision Marketing said in October 2024 it was “not even seeing a sales correlation to high attention metrics.” The IAB/MRC draft itself says attention should not be used as a measure of campaign outcomes.
So the practical play is to treat attention as an exposure-quality signal, not a proxy for revenue. Test it as a bid modifier. Use one vendor’s score consistently. Ask whether the metric is MRC-accredited and how often the predictive model is revalidated. And watch for a hidden risk: optimising for attention can reward large, intrusive formats if you are not careful.
The opportunity is real. EssenceMediacom France reported attention-based custom bidding in DV360 during July 2024 delivered CPMs 7.3% more cost-effective than standard automated bidding. Publicis Next says attention has become “easier than ever to test.” That is the right framing: test, compare within one measurement system, and keep sales outcomes as your true north.
Source: PPC Land
