Short answer: Meta’s A/B test feature, found in Experiments in Ads Manager, splits your audience into groups that do not overlap, shows each group a different version, and tells you which one won and how confident Meta is in the result. Test one variable at a time, such as creative, audience or optimisation event, give the test enough budget and at least a week, and choose the winner on cost per result, not on clicks.
Why use the A/B test tool
If you simply run two ad sets side by side, the same people can see both, Meta can push budget to one early, and the comparison is not fair. The A/B test tool avoids this by:
- Splitting the audience randomly so groups do not overlap
- Giving each version its own budget
- Reporting a winner with a confidence level
What you can test
| Variable | Example question |
|---|---|
| Creative | Does a founder video beat a product photo? |
| Audience | Does broad targeting beat a lookalike? |
| Placement | Do Advantage+ placements beat Reels only? |
| Optimisation event | Does optimising for purchases beat add to cart? |
| Offer or landing page | Does free shipping beat 10% off? |
How to set up a test
- Go to Experiments in Ads Manager, or use the A/B test option when duplicating a campaign or ad set.
- Choose the campaigns or ad sets to compare, or create a copy and change one variable.
- Pick the key metric, usually cost per purchase or cost per lead.
- Set the schedule. Run for at least 7 days so weekdays and weekends are included.
- Make sure each version has enough budget to get a meaningful number of results.
How to read the results
- Look at the winner and the confidence percentage Meta shows. A low confidence means the difference may be chance.
- Check the actual numbers, not just the label. A small difference may not be worth changing your whole account.
- Confirm the result in your store or CRM where possible.
Testing tips
- Test big ideas first. A new angle or offer usually matters more than a button colour.
- Keep a log of every test and result.
- Do not edit tests while they run.
- Retest winners later, because audiences and seasons change.
Other experiment types
Experiments also includes lift studies, which measure how many extra sales or conversions your ads caused compared with a group that did not see them. Eligibility depends on your spend and account, and they are useful when you want to know the true impact of Meta Ads.
Frequently asked questions
How do I run an A/B test in Meta Ads?
How long should a Facebook A/B test run?
What should I test first?
Why not just run two ad sets?
What is a lift study?
Who can plan Meta Ads tests for my business?
Want a testing plan?
I can plan and run tests that answer real business questions. See my Meta Ads management service or request a free ads audit. Also read Facebook ad creative testing and Facebook lookalike audiences.