Every publisher domain, one row: who sells it, how much of it, how clean it is, and who reads it

The problem this solves

Programmatic buying runs on lists. A curator packages a few hundred sites into a deal. An agency maintains inclusion lists for every client. A supply-path team decides which exchanges to keep for which publishers. A brand safety team maintains exclusion lists that grow but never shrink. Every one of those lists is a set of publisher domains, and every domain on them was once qualified by a person: someone opened the site, judged it, guessed at its scale, asked the publisher or an SSP contact which pipe to buy it through, and wrote the answer into a spreadsheet.

That process breaks in three places. It does not scale, so teams review the same few thousand well-known publishers while the long tail goes unexamined. It goes stale, because publishers change their monetisation partners, add and remove ad positions and get acquired, and nobody re-reviews a site that was approved two years ago. And it is blind to the sites nobody has heard of, which is exactly where made-for-advertising inventory lives and where an audience that a brief actually wants may also live.

Publisher supply intelligence replaces the manual review with a row of data that exists for any domain on demand. The row answers the six questions a buyer asks before a site goes into a package, an inclusion list or a supply path, and it answers them the same way for a national newspaper and for a site launched last month.

What is in a row

Active versus authorised

Publishers permit many more sellers than they use. A regional news site in the sample lists 113 authorised sellers and 989 reseller relationships, yet only 24 exchanges actually carried its ad requests on the day it was profiled. If a curator builds a deal in one of the other 89, the deal never fills, and the usual explanation is "the publisher must have turned it off". The row separates the two: active SSPs, with the seat IDs that belong to them, and the dormant remainder as a count. That single distinction removes the most common cause of curated deals that never deliver.

Reading the impression range

The impression figure is deliberately a range, not a point estimate. Its job is to sort sites into the right order of magnitude before a conversation, not to replace the delivery report an SSP produces after a deal is live. In practice the decisions a buyer makes at this stage are coarse: is this a site that can carry three million impressions a month for a regional campaign, or one that will exhaust itself in a week. The range answers that, and it comes with the number of ad positions per page so the reasoning is visible.

What the quality score means

The score runs from 0 to 100 and higher is better. Editorial publishers in the sample sit between 79 and 93. The three made-for-advertising sites score 28, and they look ordinary from the outside: they have owners, dozens of active exchanges, and audiences the profile can describe. What they do not have is a reason to exist other than to carry ads, and the score captures the signals of that: page structure, ad density, content origin and the way traffic is acquired. A threshold of 50 separates the two groups cleanly across the sample, and a buyer can set a stricter threshold for a premium package.

Audience without cookies

The audience profile is inferred from what the site publishes, not from who visited it. That has two consequences that matter for curation. First, it exists for every domain, including sites that have never been in a data partner's panel. Second, it survives the end of third-party cookies, because it never depended on them. Each profile carries likely age groups, gender skew, income level, education, life stage, household type, interest segments, in-market signals and named personas, all coded to the IAB Audience Taxonomy so segments are filterable and joinable across the whole database. The beauty site in the sample profiles as female-leaning, 25 to 44, with beauty products in market and makeup artist and beauty blogger personas; the football site as male-leaning with sports merchandise in market; the parenting site as expectant mothers and new parents with baby products in market. Those are the labels a curated package is sold under.

Four workflows, end to end

1. Qualifying a media owner for a curated deal

A client brief names a region and an audience. The curator pulls the candidate publishers for that region from the database, filtered by category and language, and sorts them by the impression range. The quality score removes the made-for-advertising sites at once. For each remaining site, the active SSP list shows where to build the deal and the seat ID to build it against, and the audience profile confirms the site attracts the audience the brief describes. What used to be a week of emails to publishers becomes an hour with a table.

2. Refreshing an inclusion list every week

An agency keeps an inclusion list of 4,000 domains per market. Once a week the whole list is re-run. Sites whose quality score dropped below the threshold are flagged for review. Sites that changed their active SSPs are flagged so the buying path is updated. New candidates that meet the scale and quality thresholds in the target categories are appended. The list stays current without anyone re-reviewing sites by hand, and every change carries the numbers that caused it.

3. Cleaning supply before it enters a marketplace

An exchange runs a curated marketplace and wants a guarantee that no made-for-advertising inventory reaches it. Every publisher that applies is profiled. The score gates admission, the impression range sets the expected volume so under-delivery is caught early, and the audience profile decides which packages the publisher belongs in. The same profile is re-run monthly so a publisher that changes character after admission is caught.

4. Building an audience-first package without cookies

A retail brand wants to reach expectant parents in the United Kingdom without relying on identifiers. The profiles are filtered for the expectant mother and new parent personas with baby products in market, restricted to UK sites above a scale threshold with a quality score above 70. The result is a package of domains selected by the audience their content attracts, each with the exchange and seat ID to buy it through. The package is defined by the content, so it does not decay as cookies disappear.

Coverage and the long tail

The row is available for any domain in the underlying database of more than 100 million domains, which covers the sites behind 99.99 percent of active web usage. That is the point of the product: the well-known publishers are already known to every buyer; the value is in answering for the sites nobody on the team has reviewed, which is where both the risk and the unused opportunity sit. Long-tail sites are profiled the same way as the largest, with the same fields, so a list can grow into the tail without growing the review team.

Publishers that carry no advertising at all are still in the database with their category, language, country, rank and audience profile; they simply have no active seller list and no impression estimate, which is itself a useful answer when a client asks about a site that turns out not to sell ads programmatically.

How it is delivered

Three delivery forms cover the three ways teams work. The lookup API returns a row as JSON for one domain per request and suits tools and dashboards that qualify sites as people work. The list job takes a file of domains and returns the rows as CSV or JSON, with an optional weekly refresh that re-runs the same list and reports what changed. The database licence delivers the full set for on-premises use, joinable with whatever the team already holds, with the deep layer for the largest sites refreshed monthly. Every delivery uses the same field names, so a team can start with the API and move to a licence without changing anything downstream.

Why one row instead of five vendors

Each dimension in the row can be bought separately today from a different specialist: supply-chain data from one, traffic estimates from another, quality lists from a third, contextual categories from a fourth, audience data from a fifth. Buyers who assemble those feeds spend most of their effort on joining them: different domain normalisations, different refresh cycles, different definitions of the same thing, and a licensing conversation for each. The decision a buyer makes needs all of them at once, for the same domain, on the same day. That is what a single row provides, and it is why the audience profile is part of it rather than a separate product: a curated package is sold as an audience, and the audience and the supply must be described together.

What it is not

It is not a delivery report. Impressions are an estimated range from the outside, and once a deal is live the SSP's own numbers are the record. It is not a bid stream or a real-time signal; profiles are refreshed weekly or monthly, which matches how quickly publishers actually change. And it is not a replacement for a verification tag on the impression; it is the layer before the impression, deciding which sites are worth bidding on in the first place. Used that way, it removes most of the sites a verification vendor would later flag, which is the cheaper place to remove them.