Offsite is where retail media grows next — and where its data advantage traditionally stops at the property line. Cookieless Audience extends shopper logic to the open web: 102 million domains, each pre-scored against 283 purchase-intent segments (PI.* codes) plus full demographics and interests, all from fixed v1.0 vocabularies aligned with IAB Audience Taxonomy 1.1. The purchase-intent codes are the join key: map a retailer category to its PI segments, and the database returns every domain whose audience is in-market for exactly what the sponsoring brand sells — with no cookies, no identity graph, and no PII.
A retail media network’s pitch is closed-loop shopper data — and onsite, that pitch is unbeatable. Offsite, the picture changes. Extending shopper segments to the open web through identifiers means matching against third-party cookies and ID graphs: coverage collapses on Safari, Firefox and iOS, where cookies are blocked by default — roughly 40%+ of traffic. Cookies remain on Chrome, but a curation strategy that only works on half the web is not a strategy, and privacy regulation keeps raising the cost of the identifier path.
Domain-level audience match is the complementary path: instead of following the shopper off the property, characterise the destinations. Every domain in this database carries coded purchase-intent segments — is this audience in-market for home improvement, pet supplies, baby furniture, cruises? — alongside demographics, interests and a resolved persona. Selecting offsite inventory becomes a query: which domains’ audiences are in-market for the sponsoring brand’s category?
Because both sides of that query speak in fixed codes, the whole offsite curation layer is reproducible, auditable and explainable to the brand paying for the campaign.
PI.hardware_supplies.*
PI.pet_services.pet_stores
PI.furniture.baby_and_toddler_furniture
One coded vocabulary on both sides of the match — no identity join, no modelled lookalikes.
The dataset slots in wherever the network decides which open-web inventory deserves the brand’s budget.
For each sponsored category, build the allow-list of domains whose audiences carry matching PI.* intent — then hand it to your DSP or curation partner as the campaign’s inventory universe. The full workflow: retail media offsite planning.
Package category-matched domain sets as named PMP / Deal-ID products — “Home Improvement Intenders, open web” — that brands buy off the rate card. Construction details in inventory curation.
Every domain on the plan carries checkable codes and confidence bands from the published taxonomy — so the answer to a brand’s “why is my budget on this site?” is a row of data, not a modelling story.
The unit of work is a sponsored category, not a cookie pool. Per-URL granularity for large multi-topic domains is available through the real-time API.
Translate the retailer category (and the sponsoring brand’s range) into PI.* segments from the 34-group intent vocabulary.
Select domains whose audiences carry those intent codes at medium+ confidence.
Refine with demographics that mirror the retailer’s shopper base — life stage, income band, urbanicity, home ownership.
Rank by intent concentration, apply brand-suitability checks, and cut the list into campaign allow-lists or Deal-ID packages.
Attach the coded rationale to campaign reporting; re-run the identical query at each quarterly data refresh.
A home-improvement RMN is selling an offsite package to a power-tool brand. The target: homeowners actively renovating. Here is the offsite selection query — codes on the left, what each filter means on the right.
| Field | Filter value (code) | Reads as | Confidence floor |
|---|---|---|---|
| purchase intent | PI.home_garden_services.home_improvement_and_repair | In-market: home improvement & repair | high |
| purchase intent | PI.home_garden_services.remodeling_and_construction | In-market: remodeling & construction | medium |
| interest | INT.home_garden.home_improvement | Home improvement content | high |
| interest | INT.home_garden.remodeling_and_construction | Remodeling & construction content | medium |
| home_ownership | owner | Homeowner-skewed audiences | medium |
| life_stage | family_young_children, established_professional | Prime renovation life stages | medium |
| urbanicity | suburban, small_town | Where the stores are | medium |
| ID-matched shopper segments | Contextual keyword targeting | Domain-level audience match (this dataset) | |
|---|---|---|---|
| Cookieless coverage | Collapses where cookies are blocked (Safari, Firefox, iOS) | Full | Full — profile is a property of the domain |
| Selection unit | Matched user IDs | Page keywords at bid time | Domains, chosen before the campaign |
| Category precision | High where matched, biased sample | Keyword ≈ topic, not audience | 283 coded intent segments per audience |
| Explainability to brands | Match-rate caveats, modelled extensions | Keyword lists | Fixed v1.0 codes + confidence bands per domain |
| Privacy surface | Identity joins, consent-dependent | None | None — no PII, nothing joins to shopper identity |
| Works pre-bid per impression | Yes, where IDs match | Yes | No — this is the planning/curation layer, by design |
No, deliberately. There is no PII anywhere in the pipeline and nothing to key against a customer record. The join happens at the category level: your taxonomy maps to PI.* intent codes, and those codes select domains. That keeps the offsite curation layer entirely outside your shopper-data governance perimeter — no consent implications, no data-sharing agreements with a third party about your customers.
The 34 intent groups and 283 segments are published in full on the taxonomy page, and the group structure — CPG, consumer electronics, furniture, hardware, apparel, pet services, travel and so on — deliberately mirrors how retail categories are organised, so most mappings are one-to-few and a category manager can review them in an afternoon. For large catalogues, custom enrichment including a category-mapping layer is available under a custom licence.
Yes — the output of a selection query is a plain domain list with attached codes, which loads into any DSP as an allow-list or into a curation platform as the basis of a Deal-ID package. If you operate offsite through an agency or a curation partner, you can share campaign-specific lists; redistributing the dataset itself or embedding it in a commercial product requires an OEM/custom licence (from $15,000/year).
Offsite curation benefits from tail coverage, so the common starting point is the top-1M file at $1,990 one-time ($590/quarter refresh, instant card checkout). Networks concentrated in one market often add country or vertical slices at $190–$490. Larger cuts — 5M domains up to the full 102M corpus, recurring feeds, custom enrichment — are quoted individually. Per-URL analysis via the API starts at $99/month.
The full offsite workflow, step by step.
Turn matched domains into named deal packages.
Reach the shopper traffic ID matches never see.
How buying teams plan against the same coded segments.
Map one sponsored category to its PI codes in the demo dashboard and see the offsite domain list assemble — then license the tier that fits your network.
Open the audience demo See database pricing