How lookalikes are built
A lookalike audience is a targeting option on Meta and similar ad platforms. You upload or select a seed audience, such as customers from the last 180 days, and the platform finds other users whose behaviour and profile resemble that seed. On Meta you choose a size as a percentage of a country's users: a one percent lookalike is the closest match, a ten percent lookalike is broader and less similar. The quality of the seed decides the quality of the result.
Seed with people you actually want more of. A list of repeat buyers or high order values usually produces better lookalikes than a list of everyone who ever visited the site.
Lookalikes in the era of broad targeting
Ad platforms increasingly recommend broad targeting and let the algorithm find buyers from conversion data, so lookalikes matter less than they did a few years ago. Many accounts now test a lookalike against broad and keep whichever wins; neither is guaranteed.
When targeting is broad, the creative does the targeting. An ad whose first frame shows a specific person and a specific problem tells the algorithm who responds, which is one reason brands buy UGC ads from creators who resemble the customer they want.
Matching creators to the audience
If the seed is mothers of toddlers in the US, the creator on screen should plausibly be one. Use your customer data to brief the creator search: age range, country, life stage and the problem they bought for. Then let creative testing confirm which faces and angles convert for that audience.
Refresh seed lists regularly. A lookalike built from last year's buyers keeps finding last year's customer.
How it's used
The one percent lookalike of repeat buyers beat broad on cost per purchase, but only with the nurse creator's video.
Our seed list is mostly discount hunters, so the lookalike keeps finding more of them.