Offsite retail media needs a way to identify which open-web domains actually reach your shopper segments. Cookieless Audience provides a 102M-domain database where every domain carries purchase-intent codes, interests, demographics and personas aligned with IAB Audience Taxonomy 1.1. The purchase-intent branch — 283 PI.* segments in 34 groups — is the natural join key between retailer categories and web audiences.
Onsite, the network knows what shoppers search, browse and buy. Offsite display, native and video bought across the open web has none of that signal — buyers fall back on broad extensions they cannot inspect.
Safari, Firefox and iOS already block third-party cookies, leaving roughly 40%+ of open-web traffic cookieless. An offsite plan that only works in one browser is not a plan.
Instead of asking “which users can we follow?”, domain-level audience data asks “which domains does our shopper segment read?” That works on every browser because it needs no identifier.
Every domain in the database carries purchase-intent segments from a fixed vocabulary of 283 codes in 34 groups — the complete IAB Audience purchase-intent branch, expressed as stable PI.* identifiers.
A grocery or drug retailer's food, beverage and household aisles map onto the CPG and food-service branches of the intent vocabulary.
PI.cpg.edible Edible CPG
PI.food_beverage.food_delivery_services Food Delivery Services
PI.cpg.non_edible Non-edible CPG
Department-store and marketplace categories join to the clothing, electronics and home-goods branches with segment-level precision.
PI.clothing_accessories.footwear Footwear
PI.consumer_electronics.televisions Televisions
PI.furniture.outdoor_furniture Outdoor Furniture
Specialty verticals get their own groups, so a pet or sporting-goods retailer isn't forced into a generic “shopping” bucket.
PI.pet_services.pet_stores Pet Stores
PI.sporting_goods.exercise_and_fitness_equipment Exercise & Fitness Equipment
PI.beauty_services.spas Spas
Because the codes are fixed and versioned, the mapping from your category tree to PI.* segments is a one-time exercise that survives every data refresh. The full vocabulary is published on the audience segmentation taxonomy page.
The same workflow serves an in-house retail media team, the agency planning against it, or a curation desk packaging the result as a deal.
Translate the campaign's shopper segment into vocabulary terms: purchase-intent codes plus supporting demographics such as life stage or income band.
Query the 102M-domain database for domains carrying those PI.* segments, restricted to your geo or vertical slice.
Rank candidates by attribute fit and banded confidence (low / medium / high), and drop domains whose demographics contradict the segment.
Export the shortlist as an inclusion list, PMP / Deal-ID package or curated marketplace seat — with the audience evidence attached for the brand.
The domain database is built for planning, curation and enrichment. For page-level granularity — a specific article or section URL — the real-time audience segmentation API classifies individual URLs on demand. Neither claims impression-level pre-bid classification in the bidstream.
The shopper segment: households that cook, skew to young families, mid-to-upper income. Here is how three candidate domain profiles from the database compare — codes shown with their human-readable labels.
| Attribute | Recipes & cooking publisher | Parenting & family magazine | General tech-news site |
|---|---|---|---|
| Purchase intent | PI.cpg.edible Edible CPGPI.food_beverage.food_delivery_services Food Delivery Services |
PI.family_parenting Family & ParentingPI.cpg.edible Edible CPG |
PI.consumer_electronics.computers Computers |
| Interests | INT.food_drink.cooking CookingINT.food_drink.healthy_cooking_and_eating Healthy Cooking & Eating |
INT.family_relationships.parenting_babies_toddlers Parenting Babies & Toddlers |
INT.tech_computing.computing Computing |
| Age bracket | 25_34 / 35_44 |
25_34 / 35_44 |
25_34 |
| Life stage | family_young_children Family, young children |
new_parent New parent |
young_professional Young professional |
| Income level | upper_middle Upper middle |
middle Middle |
upper_middle Upper middle |
| Confidence | high |
high |
medium |
| Verdict for this plan | Core inclusion | Include — secondary | Exclude for this segment |
The tech-news site is perfectly good inventory — for a different campaign. That is the point of joining on intent codes rather than reach: the plan inherits the segment definition, and every inclusion can be explained to the brand line by line.
Each method has a legitimate role; this table shows what each one gives you when the brief is “reach our shopper segment offsite, on all browsers, with evidence”.
| Criterion | Identity-based extension | Keyword contextual | Black-box shopper segments | Domain-level audience data |
|---|---|---|---|---|
| Works on Safari / Firefox / iOS | Partially — depends on match rates | Yes | Varies by vendor method | Yes — no identifiers used |
| Maps to retailer categories | Indirectly, via seed audiences | Loosely, via keyword lists | Claimed, rarely inspectable | Directly, via 283 PI.* codes |
| Inspectable by the brand | No | Yes, but shallow | No | Yes — fixed vocabularies, per-attribute confidence |
| Granularity | User level | Page level | Segment level | Domain level; page level via the real-time API |
| Best used for | Retargeting known buyers | Moment-level adjacency | Quick scale | Planning, curation and Deal-ID packaging |
The general form of this workflow: build curated domain packages for any audience, not just shopper segments.
Extend the same segment definitions to digital-out-of-home content feeds and omnichannel plans.
Profile a retail prospect's audience landscape before you pitch the offsite budget.
Plan against 1,667 deterministic personas instead of raw attribute filters.
The publisher-side mirror image: package your own inventory with taxonomy-coded audiences.
The industry view: how RMNs use the database and API across the whole offsite stack.
The purchase-intent vocabulary is the complete IAB Audience purchase-intent branch: 283 segments in 34 groups, each with a stable PI.* code and label. Most retailer trees map with a simple lookup table — grocery aisles to PI.cpg.edible and the food-service group, electronics departments to the PI.consumer_electronics group, and so on. Because the vocabulary is fixed and versioned (v1.0), the mapping is done once and survives refreshes. The full list is browsable on the taxonomy page.
No, and we don't claim it. The database is a planning, curation and enrichment instrument: you use it to decide which domains belong in an offsite plan, inclusion list or Deal-ID package before the campaign runs. For per-URL granularity — classifying a specific article or section page — the real-time API returns audience attributes for individual URLs, again for planning and analysis rather than in-bidstream decisioning.
The top 100k domains with full audience attributes are $490 one-time ($190/quarter refresh); the top 1M is $1,990 one-time ($590/quarter) with instant card checkout and immediate download. Vertical or country slices run $190–$490. Larger cuts — 5M up to the full 102M corpus, custom enrichment or feeds — are quoted individually. See pricing for details.
No. There is no PII anywhere in the pipeline. Attributes describe the audience a domain's content attracts — expressed as coded values from fixed vocabularies with banded confidence (low / medium / high) — not individual users. That is why the data behaves identically on Safari, Firefox, iOS and Chrome: it never depends on an identifier in the first place.
Open the demo and filter real domains by purchase-intent codes, or buy the Top 1M database with instant download and join it to your own category tree this week.