Interviews, insight & analysis on digital media & marketing

The early signs are positive for Seller-Defined Audiences 

By Kacper Polewiak, Project Specialist at RTB House

In February 2022, IAB Tech Lab published its specifications for a cookieless approach to addressability, known as Seller Defined Audiences (SDA). The SDA specification is intended to provide publishers, DMPs and data providers with a way to scale first-party data responsibly without data leakage or additional IDs or black box technologies.

It provides a framework around which the concept of contextual and audience ‘cohorts’ can be utilised on the open web, i.e. beyond the walled gardens of platforms such as Google, Amazon or Meta. As such, the independent publishing community has shown considerable interest in exploring the use of SDA over the past year.

What are SDAs?

Essentially, SDAs permit the sell-side to provide three pieces of meaningful information about their inventory to buyers: the SDA provider, the content topic or user’s interest and the leveraged taxonomy. SDAs do this by enhancing the OpenRTB architecture with additional elements, which can be used through a Prebid adapter or SSPs.

Firstly, publishers or their data partners analyse their content and first-party data to later determine relevant contextual elements and user interaction-based  audience attributes on their properties. Next, they label their content with adequate labels and map similar groups of users to broad cohorts, using  a standardised Content and Audience Taxonomies, and document data characteristics/metadata via the Data Transparency Standard (DTS). Finally, they relay taxonomy IDs within OpenRTB to inform downstream signalling by buyers. 

There are two main types of SDA signals, each conveying a different type of data:

  • Contextual SDA signals – placed in site, content or data of the bid request, they assign topics that are supposed to reflect the website’s content.
  • User SDA signals – placed in user data of the bid request, they assign interests to users according to their historical behaviour on the publisher’s domains.

A single bid request can include multiple SDA signals that are different and/or the same type. 

A definition of success?

While SDAs are increasingly being spoken about across the industry, there has been some uncertainty around their quality on the buy side. However, early evidence suggests that advertisers and demand side platforms (DSPs) should be excited about what the solution can deliver, without being too worried about its reliability.

To prove this, RTB House ran a pair of tests to establish the quality of contextual SDA signals. These tests used two solutions to check Seller-Defined Audiences against – audience-based deals set up by SSPs (deal IDs) and its proprietary ContextAI engine – to establish the level of quality displayed by the signals.

RTB House worked with a few different SDA providers and SSPs, receiving over 670 million bid requests with simultaneously requested deal IDs and compliant contextual SDA signals. All of the investigated cases presented a match rate of 83% or higher and in some cases contextual SDAs proved to be more precise than deal IDs. 

For the second test, RTB House relied on mapping SDA segments into ContextAI labels, and then identifying match rates between the ten most common categories of contextual SDA signals and ContextAI labels. In this instance, only two of the two SDA categories presented lower-than-expected match rates, with the weighted average match rate across all 10 categories compiled coming out over 85%. 

It is worth noting that these tests only focused on compliant contextual SDA signals, which make up around 67% of the signals RTB House receives. A compliant SDA signal, as per IAB Tech Lab’s specification, must contain the name of the business that produced the signal, the topic/segment that the content or user’s interest fits into, and highlight the taxonomy of that topic/segment.

RTB House also conducted two further tests to check that the theoretical mechanisms of assigning users to interest groups held true in practice, and if the captured audience with the selected Seller-Defined Audiences segment is more relevant than the one without. The outcome of both user SDA tests was positive, confirming that the vast potential of Seller-Defined Audiences as an efficient, scalable and cookieless method of targeted advertising is not jeopardised by lack of reliability.

The future of SDAs

It’s essential to keep in mind that leveraging user SDAs will only be privacy-preserving if they’re not passed along with other user-specific signals, such as external cross-site IDs, 1st-party IDs, 3rd-party cookies, and others. This currently doesn’t hold true for the vast majority of user SDAs. 

In addition, the quality of user SDA signals should not be taken for granted as many websites still lack any SDAs whatsoever and, where they don’t, their quality may vary depending on the publisher and the mechanism used to generate them. On top of that, a large portion of Seller-Defined Audiences signals is incomplete, which makes them useless for buyers. The Data Transparency Standard mentioned above should assist in achieving a larger level of buyers’ certainty around data quality, but to achieve that, certification checks should be performed more frequently than on an annual basis.

Overall, though, DSPs should be encouraged by RTB House’s research results, and begin investing resources into testing SDAs themselves. This is vital for the solution to really take off because, should they not begin to see some demand, publishers will stop investing time in SDAs – despite the influence it provides them over their inventory and audience labelling.


More posts from ->