Glossary
Lookalike audience
Published · Last updated
Next scheduled review:
A lookalike audience is a group of people chosen because they resemble a "seed" list of known customers or converters, found by a model that compares their traits and behaviour with the seed.
The IAB Tech Lab's PAIR specification calls the idea audience augmentation (IAB Tech Lab, PAIR specification, glossary). It means expanding an advertiser's audience by finding people who share traits with those on a known seed list. Its clean room guidance lists lookalike modelling, based on conversion and attribution analysis, among the uses of a clean room (IAB Tech Lab, Data Clean Rooms Guidance).
The process has four steps. The advertiser uploads a seed, usually hashed emails or IDs. The platform matches the seed to its own users and learns what sets them apart. It scores everyone else for similarity. The buyer then picks a size: a small lookalike is close to the seed, a large one adds reach and loses precision. Models can copy bias in the seed, and for credit, housing and job ads that can raise legal questions, covered in fair lending advertising rules.
Example: A meal-kit company seeds a model with 20,000 subscribers who stayed longer than six months. The platform returns a 1% lookalike of 2 million people and a 5% lookalike of 10 million. In a split test, the 1% group converts at twice the rate, and the 5% group brings more total sign-ups at a higher cost per sign-up.
Related terms: audience extension, second-party data, clean room, PAIR, third-party data, hashed email.