Most creative testing on Meta fails because the test is set up to be inconclusive. Too many variables, not enough budget per ad, and results read at day three instead of day ten.
Our default structure is 3-2-1: three angles, two hooks per angle, one format. That's six ads in a single campaign with budget concentrated enough to actually learn something.
We give each test two full weeks unless an ad is clearly dead by day four. The signal we watch first is hook rate (3-second video views over impressions), then thumb-stop CPMs, then CVR downstream.