Traditionally, audiences have been built using data from various sources (e.g. mobile apps, IP address, web browsing) and stored in cloud databases. A unique identifier is then used to combine these different data points to determine which audiences a user might belong to, such as: EA7583CD-A667-48BC-B806-42ECB2B48606 → Auto Intender, Car Enthusiast.

With on-device audiences, there is no centralized data repository because audiences are generated and stored within the app itself. There is no need for a uniquely identifiable identifier to determine a user’s audience memberships, since all behavioural data already resides within the app, such as: Music App: Joggers, Walkers, Early Risers.

These on-device audiences can be activated by advertisers to reach relevant audiences without requiring any app-level identifiers for targeting, thereby preserving the user’s identity.