Meta Pixel, Conversions API and attribution

How do I enable automatic A/B testing in Facebook Ads Manager?

Ads Manager has no single A/B switch. Use Creative testing in the ad editor or the A/B test in Experiments, and set the budget and the decision rules first.

Published

There is no single switch called automatic A/B testing in Ads Manager. Meta gives you two tools instead: Creative testing, which sits in the ad editor and splits spend between new ads for a week on its own, and the A/B test, which runs from the Ads Manager toolbar or the Experiments tool and compares two versions on a separate audience. The tool decides how the test runs; the budget and the decision rules decide whether the result means anything.

Which tool fits the test

Creative testingA/B test
Where it startsinside the ad editorthe Ads Manager toolbar or Experiments
What it comparestwo to five new ads, up to 10 in many accounts, created as duplicates of the ad you are editingversions of a campaign, ad set or ad
Audienceone audience, each person sees one versionseparate segments of the audience, so nobody sees both
Spendroughly even during the test, from a test budgetthe same budget recommended for each version
Length7 days by defaultat least 7 days, 30 at most
Resultsin Experimentscost per result, side by side
Best whenthe budget is too small to feed a separate testing ad settwo concepts or two strategies go head to head

The ad you are editing is not part of a Creative test. The tool duplicates it so the original keeps delivering, which is the property that makes the setup safe on an account that is already spending.

How to start a Creative test

  1. Open the ad you want to build from, in the ad editor.
  2. Choose two to five new ads, created as duplicates of that ad.
  3. Set the test budget. Meta recommends no more than 20% of the campaign or ad set budget.
  4. Leave the length at the 7 days it suggests, unless you have a reason to go longer.
  5. Pick the comparison metric, usually cost per result.
  6. Launch and leave it alone. Each person sees one version, and the results land in Experiments.

Meta’s own guidance puts the ceiling at 20% of the campaign budget for a reason: the test should not starve the ads that already pay for themselves. At launch the tool required a daily budget and the Highest volume bid strategy. Since then the limits have moved, including how many ads an account can test, so check what your own editor offers rather than a screenshot from last year.

How to start an A/B test

  1. Create it from the Ads Manager toolbar or the Experiments tool.
  2. Pick the level: campaign, ad set or ad.
  3. Change one variable between the two versions.
  4. Give each version the same budget, which is what Meta recommends.
  5. Run it for at least 7 days and no more than 30.
  6. Read the result on cost per result.

Meta advises against testing informally by switching ad sets or campaigns on and off. Switching on and off is not a test, it is two campaigns with a story attached.

Both tools answer a narrow question well. Neither decides anything for you: they report cost per result on Meta’s count, and the winner needs to be confirmed on the leads you actually received.

The budget that makes a test readable

Leads are counts, and small counts are noisy. Budget each creative in multiples of your target CPL, because the multiple says how many leads you would expect if the creative were exactly on target.

StageSpend per creativeLeads at target CPLWhat you can decide
Screen3× target CPL3zero leads means cut, which wrongly kills an on-target creative 5% of the time
Stricter screen5× target CPL5zero or one lead means cut
Rank10× target CPL10move budget towards the leaders, crown nobody
Call a winner25 to 30× target CPL, finalists only25 to 30separate creatives whose CPL differs by about 2×

The weekly budget then follows from Meta’s learning guidance. Five creatives at 10× target CPL is 50× target CPL a week, and that same volume lets a testing ad set reach the 50 results Meta mentions.

The noise is worse than most accounts assume. Ten leads at a €30 CPL are consistent with a true CPL anywhere from about €16 to €63. Two creatives at 12 and 8 leads give the leader about an 81% probability of being genuinely cheaper, which means crowning it would be wrong about one time in five. At 20 leads against 10 the probability is about 96.5%.

Write the rules before launch

  1. Fix the target CPL, the screen threshold, the rank threshold and the winner rule.
  2. Cut at 3× target CPL with zero leads.
  3. Take no verdict under about 10 leads per creative.
  4. Call a winner at 20 to 30 leads, with roughly two leads for every one of the runner-up, or a win probability of 95% or more.
  5. Judge on first-party leads, then on quality.
  6. Do not extend a test because you dislike the result, or stop it early because you like it.

Changing the rules after seeing the data is how noise becomes strategy. If an AI agent reads the results for you, give it the same rules first.

The last rule is the one worth pinning above the desk. Both tools will produce a ranking of five ads after a week, and a ranking is not a verdict: in a simulation of five identical creatives with 10× target CPL each, the median run had the best one 29% below target and the worst 67% above. In 61% of runs the worst looked twice as expensive as the best. Nothing separated them except luck.

After the test

Winners graduate to the scaling ad set, and there they need to arrive in batches. Adding an ad to an ad set is a significant edit that sends it back into learning, so five winners added on five days mean five resets. Meta’s learning guidance puts exit at about 50 results in the week after the last significant edit.

A test also says nothing about whether the winner is still good next month. The signals to watch afterwards, and the ones that look like fatigue but are not, are in creative fatigue in Meta ads. The full framework, from what to test first to how to join leads on the ad ID, is in the Meta ads creative testing framework, and CBO vs ABO explains which budget type leaves a test with the fixed spend it needs.

One more thing the tools will not tell you: which landing page the winner should point to. A creative that wins on one landing page can lose on another, so the pair is the unit worth judging. DropCPL keeps the ad and the landing page on every lead and runs the landing page side as its own test, so the creative test and the landing page test answer two different questions instead of one blurred one.

Questions on this topic

All questions on this topic →

FAQ

Does Creative testing change my scaling ad set?+

It adds the new ads to the ad set you are editing, so the ad set does receive them. Spend stays roughly even during the test and each person sees one version, then the ads deliver normally. Adding ads counts as a significant edit, which is why batches are better than one ad at a time.

How long should a Facebook A/B test run?+

Meta's current best practices recommend at least 7 days, with 30 as the maximum. Creative testing defaults to 7 days. Anything shorter usually ends before the numbers separate, and stopping early because you like the result inflates false winners.

Can I run a test with a lifetime budget?+

At launch, Creative testing required a daily budget and the Highest volume bid strategy, so check what your ad editor offers before you plan around it. The limits have moved since, including how many ads an account can test.

What is the minimum budget for a readable test?+

Ten leads per creative is the floor for moving budget, and 20 to 30 for calling a winner. Since leads are counts, budget in multiples of your target CPL: 3x to screen, 10x to rank, 25 to 30x for finalists.

Ready to find out what
converts, for real?

Free to start.

Start free →