Meta's ranking and delivery systems have changed faster than most testing playbooks. Stacked ad sets, granular interest splits, and look-alike towers no longer behave the way they did before broad targeting and Advantage+ became the default. The result is that creative is now the largest controllable lever in a Meta account.
Use the A/B Test tool, not duplicated campaigns
Meta's A/B Test feature splits audiences so each variant receives a comparable share of users, then reports a confidence-based winner. Duplicating campaigns instead introduces auction overlap, where your ad sets bid against each other and inflate CPMs — Meta's own documentation warns against this.
Structure tests around one variable
- Hook variant: the first 3 seconds of a video or the headline of a static.
- Offer variant: same creative format, different offer or guarantee.
- Format variant: static vs short-form video vs carousel for the same message.
- Audience variant: broad vs interest-stacked, only when creative tests have stabilised.
Advantage+ changes the testing surface
Advantage+ shopping and lead campaigns combine prospecting and retargeting in a single optimisation, and Meta recommends giving the system at least the 50-conversion learning window per ad set before judging results. Testing too early — or pausing winners before learning exits — resets the model and burns budget.
Ship 3-5 new creative concepts per week, keep tests running until statistical significance or the learning phase exits, and only then make a call.
Where this is drawn from.
This article references the following publicly available documentation. Always verify against the original source — platform features change frequently.
- [1] Meta Business Help CentreAbout A/B Testinghttps://www.facebook.com/business/help/1738164643098669
- [2] Meta Business Help CentreAbout Advantage+ shopping campaignshttps://www.facebook.com/business/help/1170590960567013
- [3] Meta Business Help CentreAbout the learning phasehttps://www.facebook.com/business/help/112167992830700
