Lookalike Audience

A lookalike audience is a modeled audience built by finding people who share behavioral or demographic signals with a seed group. The seed is typically a CRM list (existing customers), a pixel-fired conversion cohort (recent purchasers), or a high-value retention segment. The model expands the seed by 5x-100x using whatever signals the platform has, first-party engagement, third-party data overlays, contextual signals, identity graphs.

Lookalikes are the workhorse of upper-funnel acquisition. They're more efficient than broad demographic targeting because they're conditioned on a behavior the advertiser actually cares about (the seed conversion event) rather than a proxy. They're also more efficient than retargeting because they extend reach beyond the addressable retargeting pool.

In the cookie-deprecated post-2024 web environment, lookalikes are increasingly built from first-party data rather than third-party graphs. The seed-to-lookalike model runs inside a clean room (Google ADH, AWS Clean Rooms, Habu) where the seed list never leaves the advertiser's environment.

On DOOH and CTV, lookalikes work differently because the impression is to a shared screen, not an individual. The equivalent is "venue lookalike", finding screens whose ambient audience matches the seed cohort's demographics. Trillboards' cohort composition signal makes this directly addressable: a buyer can target screens whose recent audience matches their seed's group composition and intent stage.