Lookalike Audience Build
Lookalikes are only as good as the seed you feed them. Here's how we choose seeds, size percentages, and layer targeting for scalable prospecting.
A lookalike audience is a model Meta builds from a seed list, finding new people who resemble them. The mistake most brands make is seeding from too broad a group — all customers — which produces a fuzzy, average audience that scales but barely converts.
We seed from quality, not quantity. The strongest seeds are your top 10% of customers by LTV, or anyone who's made a repeat purchase, or a high-value action like a completed quiz. We build the audience at 1% first for precision, then expand to 3% and 5% once the 1% has proven spend. We avoid stacking narrow interest targeting on top, which starves the algorithm of signal — let the lookalike do the work.
Refresh seeds quarterly, because customer quality drifts as you scale. And always benchmark a lookalike against a broad Advantage+ audience in the same account; sometimes the algorithmic broad audience outperforms a hand-built lookalike, and the only way to know is to test both with equal budget.