How to Run an A/B Test in Meta Ads Manager (2026)
How do I run an A/B test in Meta Ads Manager?
Click A/B Test in the Ads Manager toolbar, copy an ad or choose two existing ad sets, change one variable, pick the metric that decides the winner, and schedule the test. Meta recommends at least 7 days, allows up to 30, and suggests at least 80% estimated power. Nobody sees both versions.
Meta’s built-in A/B test does one thing a manual test cannot: it splits your audience so each person sees only one version, then names the version with the lower cost per result. You set it up from the Ads Manager toolbar or the Experiments tool. What decides whether the answer is worth anything is the setup around it: one variable, equal budgets, an audience no other campaign uses, and at least 7 days.
What is a Meta A/B test, and why not just compare two ad sets?
An A/B test compares two or more versions of an ad strategy that differ in one variable, such as the image, the text, the audience or the placement. Meta shows each version to a separate segment of your audience, makes sure nobody sees both, and then determines which performed best (About A/B testing).
Meta advises against informal testing, such as turning ad sets on and off by hand, because it can cause inefficient delivery and unreliable results; two ad sets running side by side can also reach overlapping audiences. Its A/B test keeps the audiences evenly split and statistically comparable. If you duplicate a campaign for the test, Meta duplicates the budget too; if you compare existing campaigns, it uses their current budgets, and it recommends the same budget for both versions.
Which ways can I create an A/B test?
Meta describes four main routes, plus test prompts it sometimes suggests in Experiments; all of them use the same underlying technology and report results in Experiments (Ways to create an A/B test).
| Route | Where you start | Best when | Limits Meta states |
|---|---|---|---|
| Ads Manager toolbar | A/B Test button above the table | Testing one variable from an existing ad set or campaign | Meta’s general recommendation |
| Duplicate an ad set or ad | Duplicate > New split test campaign | A quick change to a live ad | Rolling out gradually; 1 to 30 days |
| Experiments tool | Experiments > Create test > A/B Testing | Comparing campaigns you set up in full first | Up to 5 campaigns; none already in a test |
| New campaign | The campaign creation flow | Testing while you build a campaign | May not be available to you |
How do I run an A/B test from the Ads Manager toolbar?
These steps follow Meta’s Create an A/B test in Ads Manager for the “Make a copy of this ad” route:
- Open Ads Manager and click A/B Test in the toolbar above the campaigns table.
- Select Make a copy of this ad, choose the ad from the dropdown, and click Next.
- Select the one variable to change in the new version, and choose the ad set to copy.
- Name the test, then choose the metric that decides the winner, such as cost per purchase; add any extra metrics.
- Schedule the test’s start and end dates; Meta says they should cover your campaigns’ run, or enough time for reliable results.
- Click Duplicate Ad Set, name the new ad set, and make the change your variable calls for.
- Review every setting in both versions, then click Publish to start the test.
To compare ad sets you already have, choose Select two existing ads in step 2, pick campaigns or ad sets, click Add Ad Set for each extra one, and finish with Publish Test.
In the Experiments tool the flow is similar: go to Experiments, select Create test, find A/B Testing and click Get started, then choose Test another campaign or Duplicate an existing campaign (Create an A/B test in Experiments). You can compare up to 5 existing campaigns, but not campaigns already in another test or using the reservation buying type.
Which variable should I test?
Meta’s variables are creative, audience, placement and custom (Choosing a variable). Creative variables are set at the ad level, audience variables at the ad set level, and custom lets you change several things at once. Meta recommends a single variable if you are new to testing, because it gives the clearest read.
Make the versions different enough to matter. Meta’s own warning example is two audiences of past website shoppers aged 18 to 20 against 20 to 22: too similar to produce a conclusive result. A better test is past shoppers against an interest-based audience (Tips for improving A/B tests). For how to build the cells themselves, see how do I structure a clean creative test on Meta? For an audience test specifically, should I use broad or narrow targeting on Meta? walks through one.
How long and how big should the test be?
Meta’s numbers (A/B testing best practices):
- Length. Run at least 7 days; tests shorter than that may be inconclusive. Ads Manager allows a schedule of 1 to 30 days. If your customers usually take more than 7 days to buy after seeing an ad, run longer, for example 10 days.
- Power. When you duplicate an ad set for a test, Ads Manager shows estimated power, the likelihood of a statistically significant result for your schedule and budget. Meta recommends at least 80% (Create an A/B test by duplicating).
- Audience. Use an audience large enough to split, and don’t use it for any other campaign running at the same time, because overlap can contaminate the result.
- Budget. Give the test enough budget to produce results for each version. If the test under-delivers, Meta suggests broadening the audience or raising the budget.
How do I read the result?
When the test ends, Meta emails you a summary and a link to the full results in Experiments. You can also go to Experiments > View results, or find the campaign in Ads Manager, where a beaker icon marks campaigns in a test, and select View charts (Understand your A/B test results).
The result names the version with the lowest cost per result for your chosen key metric and marks it with a green trophy. If there is no clear winner, Meta shows the top-performing versions on other metrics. If it found neither, you get a recap and, where possible, a next step, such as extending a test that did not collect enough data. Age and gender charts break down the result as well.
Then act on it. Meta’s own example: if product images beat lifestyle photos, turn off the lifestyle ads, move their budget to the winner, and test the next question, such as the same creative with a different age range. If you use Advantage+ campaigns, where the platform reallocates budget during a test, read how do I test creative inside Advantage+ campaigns? before trusting the read. More guides sit in the Facebook ad creative testing hub.
FAQ
How long should a Meta A/B test run?
At least 7 days, by Meta’s guidance, and no more than 30, which is the limit Ads Manager allows. Tests shorter than 7 days may be inconclusive. If your customers usually take longer than a week to buy after seeing an ad, extend the test, for example to 10 days, so their purchases land inside it.
Can people see both versions of my A/B test?
No. Meta says it shows each version to a separate segment of your audience and makes sure nobody sees both. That is the main difference from running two ad sets side by side, where the same people can be reached by both and the comparison stops being clean.
Why did my A/B test not find a winner?
Usually not enough results, or versions too alike. Meta’s suggestions are to run longer, raise the budget, broaden audiences to avoid under-delivery, and make sure the versions differ enough. Aim for at least 80% estimated power before you start, and change only one variable so a difference is readable.
What budget does an A/B test use?
The ad budgets of the versions in it. If you duplicate a campaign for the test, Meta duplicates its budget as well, so check total spend before publishing. If you compare existing campaigns, they keep their current budgets, and Meta recommends setting them equal.
Sources
Sources: Meta Business Help Center, About A/B testing, Create an A/B test in Ads Manager, Ways to create an A/B test, Create an A/B test by duplicating an ad set or ad, Create an A/B test in Experiments, Choosing a variable, A/B testing best practices, Tips for improving A/B tests and Understand your A/B test results (checked 2 Oct 2026).