According to eMarketer, 90% of digital display media will be purchased using programmatic pipes in 2022. As attention metrics mature from research environments to broader applications, developing strategies to apply them within programmatic tools is critical.

As attention metrics mature from research environments to broader applications, developing strategies to apply them within programmatic tools is critical.

This article will explore the metrics advertisers can use to measure attention and the programmatic applications they can support.

Which Attention Data?

Deciding which attention data to use is just as important as how it’s used. There are three main types of attention data that can be deployed programmatically.

Verification Metrics: Metrics such as viewability and video completion rate are sometimes combined with mouseovers, audibility, or other engagement metrics to approximate attention. These metrics lack many nuances that eye-tracking and outcomes training bring to measurement.

The benefit of working with verification metrics is mostly tied to ease of execution as new tags and reporting probably won’t be required.

Eye-tracking Measurement and Prediction: Literal eye-tracking data measures the length of time people look at ads. This data is typically gathered in eye-tracking labs or via a panel using front-facing cameras on devices. In some cases, measures are expressed as observed averages on a particular channel or placement; in others, predictions are made based on commonalities with observed placements.

Eye-tracking data is extremely precise but often hamstrung by noise from creative and audience.

Synthetic (or ML) Attention Metrics: Machine learning (ML) has been used to create synthetic media quality metrics that are less susceptible to noise from audience and creative. These metrics combine eye-tracking with device signals to attempt to measure the probability of attention to ads in a placement or the total amount of attention paid to an impression. In some cases, outcome-based training of the ML maximizes the predictability of outcomes.

Some buyers are uncomfortable with ML’s inherent opacity, leading to synthetic metrics being labeled as “black box.”

Programmatic Applications of Attention Data

Pre-Bid Segments: Attention prebid segments are formed from placements grouped into sets of varying quality. Leveraging these segments, advertisers can focus spend on higher-quality inventory, ensuring their programmatic media investment drives efficient attention and outcomes.

Prebid segments are the easiest way to incorporate attention metrics into programmatic media buying but have more of a blunt force effect on pricing than custom algorithms. Each segment represents a group of media placements within a particular quality range, enabling advertisers to simply choose the levels of quality they’d like to prioritize within their budget. Of course, a range has both a high and a low end, so prebid segments cannot adjust bids with the level of granularity that custom algorithms offer. 

Curated Inventory via SSPs: SSPs can curate supply based on attention data and expose different tranches of quality as separate deal IDs. Advertisers can also choose to combine attention data with other targeting tactics. 

Curated inventory via an SSP delivers similar functionality as prebid segments without any integration required from the DSP.  The downsides of deal ID-based attention strategies are added complexity to configuring the segments and friction from siloed data across SSPs.

Custom Algorithms: Advertisers can use attention data to inform custom bidding algorithms. This approach maximizes media value dynamically to drive efficient attention and outcomes.

Custom algorithms should use attention data as an input to influence bidding instead of as the main optimization KPI. This is because systems should still optimize to business outcomes or reach and frequency, and impression-based attention maximization can lead to suboptimal outcomes, as detailed in The Attentive Audience Paradox.

Deploying custom algorithms is more involved than prebid filtering, but they offer a far more precise use of attention data.

Programmatic Attention Metrics Already on the Rise

At Adelaide, we’ve partnered with a premier professional sports league to build custom algorithms across display, OLV, and CTV for programmatic activation of our attention-based metric AU. For each channel, our AU-based algorithm enabled the brand to secure higher-quality media with a greater likelihood of attention at a lower cost than its standard bidding counterpart.

Attention-based media optimization provides a layer of quality control currently lacking in programmatic given the limitations of binary metrics.

Across programmatic display, custom bidding line items saw a 10% increase in high-AU formats, while standard bidding line items tracked increased delivery against lower-AU formats.

On OLV, we found a 35% increase in average AU compared to standard bidding tactics and a 17% more efficient cost per AU (CPAU) (see chart).

Finally, by favoring bids and delivery across more attentive dayparts and higher-quality apps, our custom algorithm achieved nearly a 5% higher average AU across CTV placements than the standard algorithm that did not use attention metrics (see chart).

Attention-based media optimization provides a layer of quality control currently lacking in programmatic given the limitations of binary metrics. Whether advertisers activate via pre-bid segments, curated inventory, or custom algorithms, programmatic attention metrics break down deployment barriers and offer a more efficient way to secure high-quality media.