How do I test two ads properly?
Use the A/B test tool in Ads Manager, change only one thing between the two versions, and let it run at least 7 days. Meta splits your audience into separate groups so the two versions never compete, then names the winner by cost per result.
ShoaibUpdated 3 min read
Why not just run two ads and watch?
Because two ads in the same ad set share one audience and one budget, so Meta pushes spend to whichever starts faster, which isn't always the better ad. Meta's A/B test tool instead shows each version to random, non-overlapping groups.
How do I run one?
- In Ads Manager, select the campaign or ad set and choose A/B test.
- Pick one variable: creative, audience, or placement.
- Choose the result to judge by, for example cost per purchase or cost per chat.
- Set the length: at least 7 days, up to 30.
- Launch, then don't edit anything while it runs.
What should I test first, in order?
| Priority | What to test | Why |
|---|---|---|
| 1 | The hook (first 3 seconds, or the headline) | Biggest swing in results |
| 2 | The offer | Changes who replies |
| 3 | The format (video vs image) | Cheap to try |
| 4 | The audience | Meta often finds buyers by itself |
| 5 | Placements | Smallest effect |
How do I read the result?
Meta shows cost per result for each version plus a confidence level, which tells you how likely you'd see the same outcome if you ran it again. Low confidence means "not proven", so keep the current winner and test something bigger next.
One rule to remember
One change per test. Two changes at once mean you'll never know which one worked.
New ads to test each week are the real bottleneck: our Creative Production service keeps the queue full.