Tech Lab Releases “Seller Defined Audiences”, Aims to Democratize Cohort Development for the Open Ecosystem

I’m excited to announce that Tech Lab has finalized its first addressability specification incubated within Project Rearc: Seller Defined Audiences (SDA). SDA was first proposed to the industry in March 2021 as part of a portfolio of addressability and accountability solutions. It’s been a labor of love for a number of forward-looking technologists who – with an eye on browser/OS maneuvering and shifting privacy and data security expectations – are attempting to make a dent in some big legacy issues within our industry. 

Here are the details …

What is Seller Defined Audiences?

The SDA specification allows publishers, DMPs and data providers to scale first-party data responsibly and reliably without data leakage or reliance on deprecated IDs and/or new, untested technologies. It aims to 1) democratize the concept of audience cohorts for the open ecosystem, 2) provide a conceptual foundation and technical scaffolding to do so, and 3) support ongoing innovation, growth, and open-market competition on top of a shared, flexible framework. SDA leverages Prebid infrastructure – the open source header bidding solution – which provides out-of-the-box scale and speed to market across Prebid’s install base.  

The basic concept is straightforward. Publishers or their data partners determine audience attributes based on user interactions on their properties, map similar groups of users to broad, standardized taxonomy attribute descriptions (Audience Taxonomy), document audience characteristics/metadata via a standardized transparency schema (the Data Transparency Standard aka DTS), then relay taxonomy IDs within OpenRTB to inform downstream signaling by buyers. 

Key Benefits 

SDA is the only addressability system design that focuses exclusively on empowering individual publishers to develop and scale anonymized first party data sets – across browser, app, and OTT/CTV environments – instead of relying on external systems that aggregate and normalize audience data points across publisher domains. Over time, publishers can compete with each other based on the underlying quality of their audiences, as defined by over 20 disclosures within the Data Transparency Standard available to buyers via Tech Lab’s centralized repository called “Transparency Center.” And they can do so with a high level of control over data leakage (which historically has commoditized premium audiences), while minimizing opaque consumer data collection. 

Buyers also benefit from SDA in a number of ways. Notably, SDA streamlines PMP-like implementations by removing the need for manual deal-ID creation. It can also more efficiently train DSP machine learning systems – via the rich set of DTS metadata available via API – to learn over time which cohorts generate the best marketing outcomes and optimize and reward publishers accordingly. 

The Whole is Greater Than The Sum of Its Parts

SDA relies on a number of existing, widely adopted specifications that have been updated to work together in new ways. Notably, a lot of work was done on Prebid integrations, OpenRTB 2.6 community extensions, Transparency Center and Datalabel.org backend mergers, API development, and Data Transparency Standard schema updates from 1.0 to 1.1 to accommodate signals that are important to make SDA valuable. While that might sound like tinkering around the edges, the working groups involved had to develop a number of creative solutions to some hairy problems.

It’s Not Just About Audiences

While much of the attention of the spec is focused on privacy-centric audience signaling, SDA also supports contextual signaling (by simply conveying Content taxonomy IDs instead of Audience Taxonomy IDs), and/or content signaling (should publishers work with a proprietary content labeling service vs a standardized taxonomy). The key difference is that context and content labels have yet to be standardized for the industry – as we have with audience metadata via the Data Transparency Standard – but that’s something Tech Lab is exploring.

Built for Flexibility + Extensibility (BYO Taxonomy)

Adopters don’t need to use standardized taxonomies to benefit from SDA. They can also choose to use internal, proprietary taxonomy ranges by simply registering their taxonomy within a new Tech Lab maintained enumerated list within OpenRTB. This allows for industry extensibility and proprietary applications of SDA while still maintaining the efficiencies provided by a standardized signaling framework. This also means that SDA has the flexibility to account for other widely-discussed, taxonomy-based approaches to cohort signaling, but that would require a bit more collaboration with the browser API teams. 

Moving forward, there’s no doubt that advertisers and publishers will need to test and learn from a portfolio of solutions to addressability. SDA makes no claims of being a silver-bullet for the industry, but should be a valuable tool in our toolbox. And given that this is Version 1 of SDA, much work will continue as adoption scales and we learn how the industry is using it in the wild. It’s heartening to already see tools and features built on top of SDA within our testing phase, and you’ll undoubtedly see more announcements in the coming months. It’s an exciting time to be on the front lines as this work evolves, and we hope to see you there.


About the Author

Ben Dick
Senior Director of Product – Consumer Privacy, Identity and Data
IAB Tech Lab