Unity & In-App UA
Unity Vector
Unity's machine-learning campaign model, which allocates spend and matches players to games from behavioural signals instead of manually defined interests or lookalike audiences.
Unity Vector is the machine-learning model behind Unity Ads campaigns. Instead of describing an audience with interests, lookalikes or manual segments, you give it an optimization event, a target cost and a creative pack, and it decides who sees the ad from behavioural signals across the Unity network. For a buyer arriving from Meta this is the main mental shift: there is no audience to build, and most of the levers you are used to pulling simply do not exist.
What you do control is narrow, which is exactly why it is worth doing well: the optimization event (Install, Registration D0/D1/D3, Payer), the target cost against that event, the creatives, and the placement lists. Every new optimization event restarts learning for roughly two weeks, and the campaign should not be touched during it. The most common way buyers break Vector is treating it like Facebook, with daily restructuring, budget jerks and a fresh campaign on every dip, none of which the model rewards.
In buyer speech
“Stop rebuilding it every morning, Vector is still learning on the Payer event, give it the full two weeks.”