Datasets
An audience is built against a dataset - the underlying data domain you are targeting. Most datasets are a kind of real-world behavior the Large Behavioral Model learns from, such as the places devices visit or the apps they use; others add demographics, home locations, or your own cohorts. Each dataset accepts its own set of filters, and each filter is populated from reference data so you send the exact ids and values the builder expects.
The datasets
The Intuizi API supports the datasets below. More datasets will follow.
| Dataset | What it targets | Typical filters |
|---|---|---|
Web (WebDomain) | Web-domain audiences - people by the domains they visit, plus device and location. | Countries, states, cities, IAB categories and subcategories, web domains, referring domains, device types, makes, OSes, browsers, languages. |
CTV (CTV) | Connected-TV audiences - people by the streaming content and devices they use. | CTV vendors, content genres, channel names. |
Cohorts (Cohorts) | Your own saved cohorts, reused as audience building blocks. | Your company’s completed cohorts. |
Apps (Apps) | Mobile-app audiences - people by the apps they have and the categories those apps belong to. | App categories, tags, OS, bundle ids, taxonomies. |
POI (POI) | Points-of-interest audiences - people by the places they physically visited. | POI segments, categories, brands, locations, or POI attributes and their values; countries (required), states, cities, DMAs, ZIP codes. |
Transactions (AffinityTransactions) | Purchase audiences - people by what they actually bought (United States only). | Transactions purchase categories, sub-categories, brands; income, age, gender, ethnicity; spend and transaction bands; purchase channel. |
Demographics (Demographics) | Household-demographic audiences, from the latest monthly vintage of the demographics data. | Gender, age range, marital status, income range; countries. |
Deidentified (Deidentified) | Delivers the deidentified signals themselves for a country and date window, rather than a targeted device list. | Signal field groups (geoLocation, advertising, userDetails, ipDetails, privacy, general). |
Profile Attributes (ProfileAttributes) | Attribute audiences - people by the profile attributes they carry. | Attribute categories, keys and values, combined as All or Any of groups. |
Origin (Origin) | Home-location audiences - people by where their device originates, resolved weekly. | Countries (required), states, cities, DMAs, ZIP codes. |
Connectivity (Connectivity) | Network audiences - devices by the home internet network or mobile carrier they used on most days of a week, resolved weekly (United States, available on request). | Network kind (required), networks, technologies, minimum days per week. Countries (required), states, cities, DMAs, ZIP codes. |
SKUs (SKUs) | Online-purchase audiences - the devices seen on the home networks of the purchases of a product brand, or from a merchant, from a panel of affiliate-tracked orders that is strongest at large United States retailers (available on request). | Merchants, product brands (at least one of the two), Match Strictness. Countries and states. |
How a dataset relates to reference data
Each dataset’s filters are not free text - they expect specific ids or values
drawn from a catalog. You fetch those catalogs from the
/api/v2/analyses/reference/* reads before you build. See
Working with Reference Data for the full
flow and the per-dataset list of which reference reads each one needs.
Building an audience on a dataset
Once you have the ids, you create an audience against the dataset with
POST /api/v2/analyses/audiences/create. The
Create an Audience guide walks through the request
body and documents the fields each dataset type accepts.