About A/B testing in Meta Ads Manager, explained
Meta’s help article “About A/B testing” only loads with JavaScript. Here it is in plain words, with the setup steps and Meta’s best practices, from Meta’s help center as read on October 9, 2026.
On this page · 8 sections
A/B testing in Meta Ads Manager compares two versions of an ad strategy that differ by one variable, such as the image, the text, the audience or the placement. Meta shows each version to a separate part of your audience, makes sure nobody sees both, and then tells you which one performed best, on a cost per result or a cost per conversion lift basis. You start it from the A/B Test button in the Ads Manager toolbar, either by copying an existing ad and changing one variable, or by picking two or more existing ad sets or campaigns. Meta advises against testing informally by switching ad sets on and off, because audiences can overlap and results become unreliable. Its best practices: test one variable, write a measurable hypothesis, use an audience no other campaign is using, run for at least 7 days and at most 30, and give the test enough budget to find a winner.
TL;DR
- Meta’s A/B test splits your audience so nobody sees both versions, then names the version that performed best.
- Start it from A/B Test in the Ads Manager toolbar: copy an ad, or select two existing ads.
- Test one variable, with a measurable hypothesis and the same budget on each version.
- Meta recommends at least 7 days. A test runs for 30 days at most.
- Adwize reads your results each day and explains them in plain words, with changes waiting for your approval.
What is an A/B test in Meta Ads Manager?
It is Meta’s built-in way to compare two versions of an ad strategy. You change one variable, such as the ad image, the ad text, the audience or the placement. Meta shows each version to its own segment of your audience, makes sure nobody sees both, and then determines which version performs best.
Meta measures the result on a cost per result basis or a cost per conversion lift basis. So the winner is not the ad people liked most, but the one that brought your chosen result for less.
This article covers Meta’s own tool. For the wider method of testing creatives on a small budget, without the formal tool, read our guide to testing creatives without wasting budget.
When should you use Meta’s A/B test?
Meta recommends it when you want to measure a change to your advertising or quickly compare two strategies. Typical questions: does this new audience beat my current one, does a video beat a photo, does one performance goal give a lower cost per result than another.
Meta explicitly advises against testing informally, for example by turning ad sets or campaigns on and off by hand. That can lead to inefficient delivery and unreliable results, because the two audiences may overlap. The A/B tool splits the audience evenly so the two groups are statistically comparable.
Example. A skincare brand runs a new ad set for a week, pauses it, then runs the old one for a week. The second week had a holiday. The comparison says more about the calendar than about the ads. An A/B test runs both at the same time on separate groups.
Why start with a hypothesis?
Because a test only helps if you know what decision it will drive. Meta asks you to choose a hypothesis before you select a variable. Its own example: a custom audience strategy will outperform an interest-based audience strategy for your business. That sentence tells you what to build and what to read.
Meta’s best practices page shows how to sharpen a vague question. Start with something general, like whether changing the performance goal gives better results. Then narrow it to a measurable statement, such as: my cost per result will be lower when I optimize for landing page views than for link clicks.
A precise hypothesis makes the result easy to act on. If it holds, you switch the goal on future campaigns. If it does not, you keep what you had and test something else.
- General question
- = does the performance goal matter?
- Narrower
- = link clicks or landing page views?
- Hypothesis
- = landing page views give a lower cost per result
How do you create an A/B test in Ads Manager?
Go to Ads Manager and click A/B Test in the toolbar. Choose Make a copy of this ad or Select two existing ads, pick the ad from the dropdown and click Next. The rest of the steps depend on that choice. You can also build a test in Meta’s Experiments tool instead of Ads Manager.
With Make a copy of this ad, you select the variable to change, choose the ad set to copy, name the test, choose how the winner will be determined, set the test dates, click Duplicate Ad Set, name the new ad set, then select Publish.
With Select two existing ads, you choose whether to compare campaigns or ad sets, pick them, click Add Ad Set if you want more than two, then name the test, choose the winning metric, set the dates and select Publish Test.
- Click A/B Test in the Ads Manager toolbar
- Choose Make a copy of this ad or Select two existing ads
- Pick the variable or the ad sets
- Name the test, choose the winning metric
- Set the test dates, then publish
Which option should you choose: copy or compare?
Make a copy of this ad is the clean option when you are creating something new. It uses an existing campaign, ad set or ad as a template, so everything stays identical except the variable you change. That is exactly what Meta’s first best practice asks for: one variable, everything else equal.
Select two existing ads suits strategies you already run, such as two campaigns built differently. It is quicker, but the two may differ in more than one way, which makes the result harder to explain.
Budget follows the choice. If you duplicate a campaign, Meta duplicates its budget too. If you compare existing ones, Meta uses their current budgets, which you can change at the ad set level. Meta recommends the same budget on both versions for a fair comparison.
How long and how big should the test be?
Meta recommends a minimum of 7 days for the most reliable results. A test can run for 30 days at most, and in Ads Manager the schedule must be between 1 and 30 days. Tests shorter than 7 days may end without a clear answer.
Adjust for how your customers buy. If they usually take more than 7 days to convert after seeing an ad, Meta suggests a longer test, such as 10 days, so those conversions have time to happen. The test dates are separate from the campaign dates but should cover them.
Size matters too. The audience should be large enough for the test and not used by any other campaign running at the same time on Facebook, Instagram or other Meta technologies. And the budget should produce enough results to name a winner with confidence.
| Setting | What Meta recommends |
|---|---|
| Variables | One at a time, everything else identical |
| Hypothesis | Specific and measurable, written before the test |
| Audience | Large enough, not used by another campaign |
| Budget | Same on both versions, enough to find a winner |
| Length | At least 7 days, 30 days at most |
| Method | The A/B tool, not switching ad sets on and off |
How do you read the winner?
When you set up the test, you choose how the winner will be determined from the options on screen, and you can add more metrics to follow. Pick the metric that matches your hypothesis, usually cost per result, before the test starts, so you are not tempted to change the yardstick afterward.
Then read the result against the sentence you wrote. If your hypothesis was that landing page views give a lower cost per result, the answer is in that column, not in clicks or likes.
Example. A test shows the copied ad set with a new audience reaching a lower cost per purchase, while the original got more link clicks. The hypothesis was about purchases, so the new audience wins, and the next campaigns use it.
How does Adwize help with testing?
Adwize prepares variants of your ads in your brand’s colors, so you have real versions to compare, and shows every campaign in plain words before anything goes live. You approve, edit or refuse. Nothing launches, changes or spends without your yes. See the Create page.
Once ads run, Adwize reads the results each day and tells you in one sentence which version costs less per result, then proposes a change for your approval. Its reports put the numbers in plain words. See the Report page.
Adwize works with Meta ads only, inside your own ad account. When a decision deserves a formal test with a split audience, Meta’s A/B Test tool in Ads Manager remains the way to run it.
Frequently asked questions
How long should a Meta A/B test run?
Meta recommends at least 7 days and allows 30 days at most. In Ads Manager the schedule must be between 1 and 30 days. If your customers take longer than 7 days to convert, run it longer.
Can people see both versions of my A/B test?
No. Meta shows each version to a separate segment of your audience and makes sure nobody sees both. That is why it advises against testing by turning ad sets on and off yourself.
Does Meta duplicate my budget in an A/B test?
If you duplicate an existing campaign, yes, the budget is duplicated too. Otherwise Meta uses your current budgets, which you can change at the ad set level. Meta recommends the same budget on both versions.
Can I test more than two versions?
Yes. With Select two existing ads, you can click Add Ad Set to add another ad set to the comparison. Meta’s best practice still asks that the versions differ by one variable only.
Does Adwize run Meta A/B tests for me?
Adwize prepares ad variants and reads your results daily, explaining which version costs less in one sentence. Any change waits for your approval. For a formal split test, use Meta’s A/B Test tool.
Sources
- Meta Business Help Center, About A/B testingAccessed 10/09/2026
- Meta Business Help Center, Create an A/B test in Ads ManagerAccessed 10/09/2026
- Meta Business Help Center, Best practices for A/B testingAccessed 10/09/2026
