How We Use Data at Adelaide

‍

Eye-Tracking

When developing an AU model for a channel, we begin by understanding how people consume ads on that medium through eye-tracking data. In Adelaide’s five-year history, we’ve become one of the most prolific users of eye-tracking data, utilizing data from nearly every leading provider, including Lumen, Amplified Intelligence, Tobii, TVision, Viomba, and even our own eye-tracking tools.
‍

_{The Tech Behind AU: Eye-Tracking Partners}‍

‍We use eye-tracking data for AU differently than attention vendors who attempt to predict the duration of attention to an impression. While our models consume and learn from gaze duration data, they prioritize identifying the likelihood of any attention to an impression and examining its context within the surrounding media environment to evaluate the quality of that placement.

As part of this methodology, we incorporate multiple sources of eye-tracking data into our models, including TVision's panel data for CTV and linear, as well as Viomba's extensive dataset spanning web and social platforms.

Exposure Data

Adelaide uses exposure data to build models and generate AU ratings for campaigns. Exposure data describes the placements from which impressions came. It can be as detailed as telemetry data generated by a JavaScript tag, as simple as a URL and ad format, or as broad as the genre of a podcast. Once built, AU models can process any exposure data describing a placement to generate as precise a rating as possible based on the granularity of the data supplied.

Sources of exposure data are diverse, including log files, Adelaide’s tag and pixels, offline scraping, ad server data, Automatic Content Recognition (ACR) data, verification data, and any other source of information that helps describe the ads people are exposed to.

Outcome Data

To ensure AU is predictive of outcomes through the funnel, we train our models with outcome data—from upper-funnel awareness to purchase data.

Outcome data includes:

AU ratings themselves are outcome-agnostic*, meaning they adhere to the principle that placements serve as conduits for creative and, for the most part, don't inherently favor certain advertising objectives. The exception is mediums that allow for specific types of engagement more suitable to particular outcomes. 

Our focus on outcomes is predicated on the belief that attention is only an effective measure so long as it drives positive impact for the advertiser—and not merely attention for attention’s sake.

In August 2024, we bolstered this capability by acquiring Rita Data. With Rita's privacy-first, cross-channel outcome data—ranging from searches to purchases and in-app actions—we've strengthened our models' predictive power even further.

Our Stance on Privacy

A lot of sensitive data is used to generate AU, so it's important to be clear that we use fully opted-in eye tracking and outcome data and don't capture PII when measuring campaigns.

We believe better media quality measurement will create alternatives to some of the less privacy-friendly practices used in the industry today, allowing marketers to opt out of the identity and tracking cat-and-mouse game with regulators and consumers.

^‍

^{*We offer several tools to help advertisers identify and target the optimal AU required to achieve an outcome.} ‍