One row per domain that tells you who reads the site, what they care about and what they are about to buy: demographics, interests, purchase intent, business audiences and personas, each from a fixed vocabulary with its own code. No cookies, no device identifiers and no personal data.
Every value the file can contain is listed with its code: the demographic and business fields, all 29 interest groups with their 288 sub-interests, all 34 purchase-intent groups with their 288 segments, and the complete list of 1,667 personas. Check it before you buy, and keep it open while you build.
Most audience data is tied to people: cookies, device identifiers and logged-in users. That kind of data is harder to collect every year, harder to share with partners and harder to defend in a privacy review. The audience database takes a different route. It describes the audience of a website as a whole, so the unit of data is a domain, not a person.
For each of one million popular domains you get one row. The row tells you which age brackets the site appeals to, whether its readers lean male or female, their typical income and education, their life stage, the interests they share and the products and services they are in the market for. Where a site serves professionals, the row adds business attributes such as company size, seniority and job function. Every attribute group carries its own confidence band, so you always know how strong the signal behind a value is.
Because every value comes from a closed, versioned vocabulary, the file behaves like reference data rather than like a text dump. You can filter on it, group by it and join it to your own tables without cleaning free text first. The codes stay stable between releases: a new release can add values, but it does not rename the ones your queries and models already use.
The file is delivered as CSV, JSON or Parquet, in the same 34 columns as the free sample, so it loads straight into a warehouse, a notebook, a search index or a lookup service. It sits naturally next to the categorization data you may already use from us, because both are keyed by the same registered domain.
The 34 columns fall into six groups. Fields that can hold several values separate them with a vertical bar, codes and labels come in matching pairs, and an empty field always means that the site gave no reliable signal for it.
Every value of every field is listed in the audience taxonomy.
Eight rows from the free 200-domain sample, which is drawn at random from the one-million-domain file. The table shows eight of the 34 columns; the download has all of them.
| Domain | Popularity band | Audience | Age | Gender | Interests | Purchase intent | Personas |
|---|---|---|---|---|---|---|---|
| 1ak-group.com | 1,000,001-5,000,000 | mixed | 25-34, 35-44, 45-54 | balanced | Automotive, Technology & Computing | Automotive Parts and Accessories, Power and Electrical Supplies | Auto Industry Professional, Auto Mechanic |
| 60fps.design | 1,000,001-5,000,000 | mixed | 18-24, 25-34, 35-44 | balanced | Technology & Computing, Careers | Online Education, Computer Software | Graphic Designer, Technology Enthusiast |
| advancedclinicals.com | 1,000,001-5,000,000 | b2c | 25-34, 35-44, 45-54 | female lean | Healthy Living, Shopping | Skin Care Treatments, Gift Baskets | Skin Care Enthusiast, Sensitive Skin Individual |
| alanrevista.org | 1,000,001-5,000,000 | mixed | 25-34, 35-44, 45-54 | balanced | Medical Health, Academic Interests | Publishing and Self-Publishing Services | Dietician, Nutrition Enthusiast |
| alesis.com | 500,001-1,000,000 | b2c | 18-24, 25-34, 35-44 | balanced | Music & Audio, Hobbies & Interests | Musical Instruments and Accessories, Audio | Music Producer, Music Enthusiast |
| amherstma.gov | 1,000,001-5,000,000 | b2c | 25-34, 35-44, 45-54 | balanced | News & Politics, Events & Attractions | Water Services, Gyms and Health Clubs | Community Activist, Event Enthusiast |
| andhravilas.net | 1,000,001-5,000,000 | b2c | 18-24, 25-34, 35-44 | balanced | Movies, Television | Ticket Services, Music and Video Streaming Services | Movie Buff, Foreign Film Enthusiast |
| annovera.com | 1,000,001-5,000,000 | b2c | 18-24, 25-34, 35-44 | female strong | Medical Health, Family & Relationships | Drugstores and Pharmacies, Health Services | Birth Control Explorer, Contraception Researcher |
Content categories tell you what a site is about. Audience data tells you who reads it. Both are useful on their own, and together they answer questions neither can answer alone. A cooking site and a restaurant review site share a category, but one may skew toward parents planning family meals and the other toward young professionals who eat out. A software review site and a gaming site may both be technology, yet their readers are in the market for very different products.
The audience file already carries the IAB content category of every domain and its web filtering category, so you can segment by topic and audience in one pass. If you also license our URL categorization database or use the categorization API, the registered domain links the two, and you can extend audience attributes to any domain you categorize.
Topic plus intent. Start from a content category, then keep only domains whose audience shows purchase intent in your product area. A travel category combined with air travel and hotel intent gives a much tighter list than the category alone.
Audience plus safety. Use the web filtering category to exclude sensitive content, then select on demographics and interests. You get audience reach without the brand-safety surprises.
Business audiences across topics. Professionals read far more than trade press. Filter on business seniority and job function across all categories to find the general-interest sites your target accounts actually visit.
The file is built to be easy on pipelines. There is exactly one row per registered domain, so the domain column is a primary key. Join it to your own data on the registered domain of each referrer, placement, bid request or click; if your logs hold full hostnames, reduce them to the registered domain first so that news.example.com and www.example.com both match example.com.
Fields with several values, such as age brackets, interests or personas, separate the values with a vertical bar. Most warehouses can split them into arrays on load, which makes filtering and counting simple. Codes and labels come in pairs: query on the codes, because they never change, and show the labels to people.
For example, a query that finds sites whose audience is interested in travel and in the market for flights, with at least medium confidence, can look like this:
Empty values are deliberate. When a site does not give a reliable signal for a field, the field is left empty rather than filled with a guess, so a count of non-empty values is an honest measure of coverage. Confidence bands let you choose your own balance: include low-confidence values when you want reach, keep only medium and high when you want precision.
The vocab_version column tells you which version of the vocabulary each row uses. Within a major version the vocabulary only grows, so a refresh never breaks the codes you already depend on.
Most teams start with the free sample. Load the 200 rows into the table structure you plan to use for the full file, write the joins and filters you need, and check the results against a handful of domains you know well. When the order is placed, the full file arrives in exactly the same columns, so the same code runs unchanged on one million rows.
For recurring use, the quarterly refresh replaces the file with a new release. Because the codes are stable, a refresh is a straightforward table swap: load the new file next to the old one, compare the counts, and switch over. Domains that have expired or dropped out of the popularity ranking are gone, and newly popular domains arrive with complete profiles.
If you only need part of the file, for example one vertical or one country, a slice costs less than the full tier and comes in the same columns, so it fits the same pipeline.
The database describes audiences at rest, which suits planning, packaging and enrichment work.
The tier holds one million domain profiles, prioritised by popularity. It starts with the most visited domains and continues down the popularity ranking until one million profiles are reached, and the popularity_rank_band column shows where each row comes from.
Every domain is a main, registered domain. Adult and gambling sites are not included, and neither are parked, for-sale, inactive or redirecting domains, or service domains without content of their own. Each order gets a fresh build, so domains that have expired since the last release are removed before delivery.
Every attribute uses a closed vocabulary aligned with the IAB Audience Taxonomy 1.1, so values are consistent across the whole file and across releases. Free text never appears in coded fields.
Each attribute group carries a confidence band, an empty field always means no reliable signal, and every row records the vocabulary version it was written with. The same inputs always produce the same personas, so segments built today can be rebuilt tomorrow.
A one-time license with an optional quarterly refresh. We send a checkout link by email; card and PayPal are accepted.
Write to [email protected] with the size or slice you need, and we send a checkout link or a quote within one business day.
No. Each profile describes the audience of a website as a whole. No cookies, device identifiers or personal data are used or delivered, so the file can be stored and shared like any other reference data.
Yes. The sample is 200 rows drawn at random from the one-million-domain file, with all 34 columns in the same order and with the same codes. Whatever you build on the sample works on the full file.
That the site gave no reliable signal for it. We leave the field empty instead of guessing, so you can count coverage honestly and decide yourself how to treat missing values.
Every coded field comes with a matching label field in the file, and the full list of codes and labels is on the audience taxonomy page.
Each order gets a fresh build. The download link follows within 2 business days of payment.
Yes. The optional quarterly refresh ($590 per quarter) delivers each new release in the same columns, with expired domains removed and new popular domains added.
The one-time license covers internal use. If you want to embed the data in a product you sell, or need more than one million domains, write to us for a quote.
CSV, JSON and Parquet. All three hold the same rows and fields. Parquet is the most compact and loads fastest into a warehouse; CSV is the easiest to open anywhere; JSON suits services that read the file directly.
Yes. Vertical and country slices cost $190 to $490, depending on their size, and come in the same 34 columns as the full file. Tell us the interest groups, categories or countries you need and we send a checkout link.
Yes. The file already includes the content category and web filtering category of each domain, and the registered domain links it to our URL categorization database and categorization API.