“Data-Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.
Today’s column is written by Seraj Bharwani, Chief Strategy Officer at AcuityAds
The behavior-tracking infrastructure that fueled the ad tech industry over the past two decades is undergoing a major overhaul. And the remodeled tracking pathways must be in place before the consumer traffic goes “dark” in the next 12 months. Google’s recent decree that email-based ID solutions aren’t viable over the long-term threatens to cut off some of these potential pathways.
While imperfect, traditional third-party cookies have been a scalable, efficient, and proven infrastructure to track and develop behavior-based audience models, advertise across channels, and measure effectiveness of campaigns on the Open Web. In their absence, online publishers, advertisers, and the ad tech ecosystem at large needs to find alternative ways to reach consumers and measure effectiveness of media investments with potentially less precise and, thus far, unproven data infrastructures.
The leading solutions, which are at various stages of development, fall into three major classifications that I refer to as the 3Cs for cookie replacement: Context, Cohorts, and Coop. None is a perfect replacement for all use cases supported by third-party cookies—yet., But collectively they offer the promise of a more sustainable solution for both publishers and advertisers.
While the Coop solution has gained substantial momentum in recent months, Google’s latest announcement poses challenges with respect to the use of deterministic data for behavior tracking.
- Context. Advertising in context is back in vogue. The idea of delivering ads based on what a person might be watching, listening to or reading isn’t new. Aligning advertising with the editorial content makes ads more relevant even if one knows nothing else about the user (which preserves privacy). By that logic, auto advertising would coexist with auto content, athletic gear ads with sports content, apparel ads with fashion, and so on. The advent of machine learning and AI has brought far greater precision by matching creative ad formats to the keywords, syntax, and semantics of the content.
While being relevant is clearly important for ads to draw viewer attention, knowing if the ads are being exposed to users who are in-market to buy, would be even better, which is something algorithms cannot determine from context alone. The other major issue with contextual ads is measurement. Last-touch attribution is usually the only option, which is quite limiting for a vast majority of the advertisers who need accountability for conversion over longer, multi-touch attribution windows. In order for contextual advertising to be a viable alternative to third-party cookies alone, initiatives to enhance contextual placements with multiparty, authenticated, data overlays will be imperative.
- Cohorts. Developing look-alike audience segments with machine learning applied to internet browsing habits at a cohort level is another way to preserve user privacy. The idea has its roots in differential privacy whereby random noise is introduced as part of aggregating user interests at the cohort level. Federated Learning of Cohorts (FLoC) is a major cohort advertising initiative being socialized by the team at Google Privacy Sandbox. Access to FLoC data will be available through multiple APIs that can support use cases for both targeting and measurement at the cohort level.
Initial reaction to FLoC has been mixed even though Google claims that early tests have delivered results (conversions per dollar) comparable to that with cookies. Since FLoC data and APIs are not publicly available for testing, the assessments, to date, are based mainly on various whitepapers and subsequent proposals for enhancements from publishers and ad tech platforms engaged with the Privacy Sandbox working group.
For FLoC cohorts to be broadly accepted by the industry, an independent validation by trusted, third parties will be critical to proving effectiveness and incremental outcomes. The FLoC APIs will need to support view-through attribution critical to multi-touch, journey activations and assure unduplicated reach and frequency. The segments are currently limited to click-through attribution which will not be effective for a variety of creative formats (videos, banners, audio ads). A vast majority of the conversions require accounting for media allocation over a multi-week lookback window.
- Coop. Publishers on the Open Web (outside the walled gardens) have begun to develop their own, proprietary, first-party subscriber data platforms as a privacy protected replacement for third-party cookies that have, so far, been integral to monetizing media inventory. Major publishers like The New York Times, The Wall Street Journal, and Forbes have established a clear value exchange with readers behind a registration wall and are using their authenticated data to build direct relationships with the advertisers. While advertisers love the privacy protected access to premium inventory, this approach does not scale beyond a few select publishers.
Advertisers need privacy protected access to a broader, publisher marketplace that includes the ComScore 100 and a full spectrum of mid-tier and niche publishers in a coop that collectively offers a scalable pool of live, authenticated inventory that can be traded programmatically.
With the emergence of ATS/IDL from LiveRamp, UID2 from Prebid, and other comparable solutions, there exists a clear opportunity for publishers to coalesce around a common set of trusted, identity solutions that are interoperable.
These solutions can help protect consumer privacy, enable rich identity graphs, support multichannel measurement and attribution, and enhance pricing power of the publisher inventory. Individual publishers are empowered to determine their own method of obtaining explicit consent from viewers on the use of a unique identifier that assures anonymity of the user.
The prevailing ID solutions rely on identifiers such as email address, IP address, timestamp and others. If Google Chrome pushes back on the use of email-based identifiers, the industry will default to probabilistic methods of resolving identity.
While there are several key requirements for a solution to be considered viable, the most important of those are scalability and measurability. The race to build, test, and learn the most effective and ‘future-proof’ strategy for the ad-supported Open Web is on.