An ad network aggregates hundreds or thousands of sites — most of them long-tail properties with no first-party data, no research budget and no audience story of their own. Sold as undifferentiated reach, that inventory earns reach prices. Cookieless Audience changes what a network can say about itself: pre-computed audience attributes for 102 million domains — demographics, 285 sub-interests, 283 purchase-intent segments, personas, all aligned with IAB Audience Taxonomy 1.1 — plus a real-time API for page-level segmentation. Classify every member site, package the network into audience products, and give the sales team evidence instead of adjectives.
Networks know their sites by vertical — “lifestyle”, “home”, “gaming” — because vertical is what you can see from the outside. But buyers budget against audiences, not verticals, and on the 40%+ of traffic where Safari, Firefox and iOS block third-party cookies (cookies remain on Chrome), no identifier is going to describe those readers for you.
Joining the member roster against the domain database replaces the vertical label with a full audience profile per site: 8 age brackets, 5-point gender skew, 6 income bands, 14 life stages, household composition, 285 sub-interests, 283 purchase-intent segments, B2B firmographics and one of 1,667 personas — every attribute a coded value with banded confidence. Suddenly the network is not “900 lifestyle sites” but a measured distribution of audiences you can package, price and declare.
For member sites that span topics, the real-time API classifies individual URLs with the same vocabularies, so a large member’s parenting section can sit in a different package than its recipes section.
// Roster classification, one member site { "domain": "example-parenting-blog.com", "life_stage": "new_parent", "household_composition": "family_with_children", "gender_skew": "female_lean", "interests": [ "INT.family_relationships.parenting_babies_toddlers" ], "purchase_intent": [ "PI.family_parenting.childcare", "PI.cpg.non_edible" ], "confidence": "high" } // Repeat for every site on the roster — // it's a join, not a research project.
Audience classification of the roster feeds three commercial motions — packaging, pricing and pitching.
Cluster the roster by shared attributes into named products — “new & expecting parents”, “fitness enthusiasts 25–44”, “B2B software evaluators” — each with an explicit rule set and confidence gate. Crosswalk the attributes to Audience Taxonomy 1.1 nodes and the packages can carry seller-defined audience declarations.
Reprice inventory by the audience it carries rather than its vertical label. Sites indexing on high-value intent segments — finance, autos, B2B — move out of run-of-network floors; genuinely undifferentiated reach stays priced as reach. Audience-based domain valuation covers the method.
Give sellers per-package one-sheets: the rule logic, the member domains, the confidence bands, the taxonomy crosswalk. An RFP response that cites coded evidence reads differently from one that says “highly engaged moms”. See publisher pitch decks for the deck workflow.
One join, then queries. The roster changes weekly; the process doesn’t.
Match member domains against the licensed file. A top-1M or vertical slice usually covers a network roster — see pricing.
Profile what the network actually holds: which personas, interests and intent segments concentrate where, at what confidence.
Define each product as attribute rules with a confidence gate; materialize member lists per package.
Set floors per package by intent value, and emit taxonomy-aligned SDA declarations on package inventory through your ad server or SSP partners.
New member site? Look it up — in the file, or via the per-URL API for section-level detail — and slot it into packages the day it joins.
A network of roughly 900 lifestyle sites wants a parenting product for CPG and childcare advertisers. Classification of the roster surfaces the cluster; the rules below define the package.
| Package rule (coded values) | Reads as | Role in the package |
|---|---|---|
life_stage ∈ {expecting_parent, new_parent, family_young_children} | Expecting through early-family readerships | Core theme — the axis the package is sold on |
INT.family_relationships.parenting_babies_toddlers | Parenting-babies interest | Confirms the editorial-audience match |
PI.family_parenting.childcare or PI.cpg.non_edible | In-market for childcare and family CPG | The intent layer CPG buyers pay for |
household_composition = family_with_children | Family households | Reinforcing demographic evidence |
confidence = high on life stage | Strongly supported classification | Quality gate — keeps marginal sites in run-of-network instead |
segtax: 4 through its SSP partners, and hands sales an evidence-backed product for CPG RFPs. Sites that miss the confidence gate stay in run-of-network — which keeps the package’s claim honest and its price defensible. Next quarter’s refresh re-scores the roster automatically.| Run-of-network reach | Vertical packaging | Audience-attribute packaging (this dataset) | |
|---|---|---|---|
| What the buyer buys | Impressions, undifferentiated | Topical adjacency (“home & garden sites”) | A described audience: life stage, income, interests, intent |
| Evidence behind the claim | None needed — and none offered | Site categorization | Coded attributes with confidence bands, reproducible per domain |
| Pricing power | Floor of the market | Modest premium on endemic verticals | Premium tracks intent value; floors set per package |
| Cookieless dependence | None, but earns least | None | None — domain-based, works identically on Safari, Firefox and iOS |
| Sales conversation | Price and volume | “We have sites about X” | “We have measured audiences of X — here is the rule set” |
That is the point of a 102M-domain corpus: it reaches far beyond the head of the market where audience data usually stops. Typical network rosters — niche blogs, hobbyist communities, regional properties — resolve against the file, each with the same coded attribute schema as a major publisher. Coverage is exactly what you can verify before licensing: look your members up in the audience demo.
No. Classification is domain-level and pre-computed — no tags on member pages, no data-sharing agreements, no PII at any step. Members benefit from the packaging without any integration work, which matters when your roster is hundreds of independently owned sites.
Use both layers. The domain record gives the property-wide profile; the real-time API classifies representative URLs per section with the same v1.0 vocabularies, so one member can contribute different sections to different packages. Large multi-topic members are usually where the API earns its keep.
Yes — that is the natural endpoint. Package rules are already expressed in vocabularies aligned with IAB Audience Taxonomy 1.1, so each retained attribute crosswalks to a taxonomy node, and your ad server or SSP partners emit the node IDs under ext.segtax: 4 on package inventory. The declaration mechanics are covered on the seller-defined audiences page; the network’s job is the evidence, which is what the dataset supplies.
Declaring network packages under segtax 4.
Turning classifications into sales materials.
Pricing inventory by the audience it carries.
The marketplace-scale version of the same motion.
Look up a handful of member sites in the audience demo, then license the file and classify the whole roster in an afternoon.
Open the audience demo See database pricing