Seller-Defined Audiences (SDA) let publishers, DMPs and data providers describe audience groups with standardized taxonomy IDs and relay those IDs inside OpenRTB. IAB Tech Lab says SDA was first proposed in March 2021 and was finalized as its first addressability specification from Project Rearc. A small OpenRTB extension called segtax tells the buyer which taxonomy each ID belongs to.

How does a publisher build a seller-defined audience?

The IAB Tech Lab announcement, published in 2022, describes the basic concept in four steps:

  1. The publisher or its data partner decides audience attributes from what users do on its own properties.
  2. It maps similar groups of users to broad, standardized descriptions in the Audience Taxonomy.
  3. It documents the audience’s characteristics in the Data Transparency Standard, a schema that Tech Lab says includes over 20 disclosures.
  4. It relays the taxonomy IDs in OpenRTB, so the buyer sees the cohort in the bid request.

The announcement adds that SDA builds on Prebid infrastructure and can carry contextual signals too, by sending Content Taxonomy IDs instead of Audience Taxonomy IDs. A publisher who prefers its own labels can register a proprietary taxonomy in a list that Tech Lab maintains.

Buyers can reach the Data Transparency Standard disclosures through Tech Lab’s Transparency Center and an API, according to the same post. Tech Lab suggests a DSP could use that metadata to learn which cohorts work best over time. The post also calls this Version 1 of SDA and expects the work to continue as adoption grows.

How does segtax tell a buyer which taxonomy applies?

The segtax extension aims to support publisher, proprietary and standardized segments with a simple structure that avoids repeating text. It adds an integer field, ext.segtax, to the OpenRTB Data object. Without it, segment IDs should be treated as fully proprietary, so the same number from two providers can mean two different things. With it, the buyer reads the ID against a named taxonomy.

The extension works wherever Data objects appear, including user.data, site.content.data and app.content.data. A minimal example follows. The provider name and segment IDs are placeholders, not real entries.

{
  "user": {
    "data": [{ "name": "provider.example", "ext": { "segtax": 4 }, "segment": [{ "id": "1001" }, { "id": "1002" }] }]
  }
}

The registry in the same file assigns these values:

  • Value 4: IAB Tech Lab Audience Taxonomy 1.1.
  • Values 5, 6 and 7: IAB Tech Lab Content Taxonomy 2.1, 2.2 and 3.0, per the same file.
  • Values 500 and up: vendor-specific taxonomies, added to the file through pull requests.
  • Values 1 and 2: older Content Category Taxonomies that the file marks as deprecated.

The registry is a Markdown file in the OpenRTB repository. Anyone can submit a pull request to add a vendor taxonomy, and the approvers check for ID conflicts without a working group review. As of October 2, 2026, the segtax file still lists two Chromium Topics taxonomies, values 600 and 601, even though Chrome’s feature status page lists Topics under Deprecate and remove.

Where does SDA fit next to identity-based targeting?

An SDA cohort describes a group the publisher defined. An identity token points at an identity. They answer different questions, and a campaign can use one, both or neither.

The announcement is candid about limits. It says SDA makes no claim of being a silver bullet and that advertisers and publishers will need to test a portfolio of addressability solutions. It also credits SDA with streamlining PMP-like setups by removing manual deal ID creation. That is Tech Lab’s claim, and a buyer should confirm it publisher by publisher.

What should a buyer ask a publisher?

  • Which taxonomy and version does each segment ID use, and is segtax sent with it?
  • What do the Data Transparency Standard disclosures say about how the audience was built?
  • Is the audience drawn from the publisher’s own first-party data?
  • Does the same segment ID mean the same thing on every placement?

How does Vectravia handle this?

Audience Lab reads seller-defined audiences from the OpenRTB data object using segtax values. A segment that arrives with no segtax value is, under the extension, fully proprietary, so it needs its provider’s documentation before anyone relies on it. This article makes no claim about how many publishers send these fields.