The lookalike audience mistake that shrinks your reach over time

A lookalike audience is only as good as the seed list it's built from, and most accounts set one up once, early in an account's life, and never revisit it. The seed list stops reflecting the business's actual current customer base within months, and the lookalike quietly drifts further from who's actually converting today.

The compounding problem: as a lookalike audience ages without a refresh, the platform's modeling is working from an increasingly outdated signal, and the audience's real-world reach and relevance shrink even though the audience size setting in the ad account hasn't changed at all. The account looks the same on paper while performing against a stale target underneath.

The fix is a scheduled refresh, not a one-time setup: rebuilding the seed list from the last 90 to 180 days of actual purchasers on a recurring cadence, ideally automated, so the lookalike is always modeled against who's buying now rather than who was buying whenever the audience happened to get created.

This is a five-minute maintenance task that most accounts never schedule, and it's a common, quiet explanation for a lookalike audience that performed well at launch and has been slowly declining ever since, with no obvious single cause anyone can point to in the interface.