Short answer: use two methods, for two jobs. For the weekly flow of new ads, put several new ads into a testing ad set, let Meta allocate spend between them, and judge on cost per conversion. For a big question, such as whether a new positioning beats the old one, run a true equal-split test with Meta’s A/B test tool. The first is an operational screen, not a randomized experiment. Reserve controlled tests for questions worth funding carefully.
Page Contents
Why not give every new ad an equal budget?
Because funding many variants simultaneously can become expensive. At an illustrative $20 a day per ad, 6 ads cost $120 a day; 60 ads cost $1,200 a day; 600 ads cost $12,000 a day. Those are illustrative budgets, not required minimums. You can instead test fewer variants at a time or reserve a controlled comparison for the strongest candidates.
Compare variants concurrently when possible: results from different weeks also reflect changes in the auction and audience.
How does Meta decide which ad gets the spend?
Meta’s auction explanation describes bids, estimated action rates and ad quality as factors. Its breakdown explanation says delivery considers predicted future opportunities and pacing, so the ad with the lowest historical cost does not necessarily get the next impression. Neither source says every new ad gets equal initial exposure or that an unfunded ad has been proven inferior.
So the spend split inside an ad set is a signal, not a verdict. Read it together with cost per conversion over enough days to settle.
How do you run a Meta-allocated creative test?
- Keep testing separate from scaling. A separate testing budget gives new ads funding outside your scaling budget, though it does not guarantee each new ad receives meaningful delivery.
- Fund it in conversions, not dollars. Ideally a campaign gets about 30 conversions a day, and 10 is good. A small campaign can run on 1 or 2 a day. Separately, for each creative test aim for 10 a day; 2 or 3 is the bare minimum only when cost per conversion is very high. These are our planning guides, not guarantees of statistical certainty.
- Test directions, with several ads each. Within each creative direction, test several ads across angles, hooks and formats. One ad per idea tells you about that execution, not the idea. There is no fixed rule for ad set size, but typically give Meta at least about 8 ads to choose between.
- Leave proven winners out of the testing ad set. This is our operational preference because an established ad can take spend that might otherwise expose new ideas. It is not a platform requirement or a substitute for a randomized control.
- Optimize for the conversion your economics run on, not clicks or video views. The system finds more of whatever you ask it for.
- Review weekly, allowing results to mature. Move promising ads into scaling campaigns when their economics justify it. An ad with little exposure is inconclusive: fund a deliberate retest if the hypothesis matters, or replace it for operational reasons without labeling it a proven loser.
How to structure creative tests for Meta ads and how to test ad creatives on Meta cover the setup in more detail.
When is an equal-split A/B test the right call?
When you need a clean answer to one big question: a new positioning against the old one, a new format family, a different landing experience. Meta’s A/B test splits the audience so the versions do not overlap. Its best practices recommend one variable and at least 7 days. The tool allows 1 to 30 days; 7 is a recommendation, not the shortest selectable duration. Allow longer when your conversion cycle requires it.
A controlled test is useful at any account size when the decision warrants the budget and sufficient sample. It can be inconclusive; equal budgets alone do not guarantee a meaningful answer. Meta’s A/B testing tool for app campaigns covers how to read its results.
What should you measure?
- Cost per conversion, on the event your business runs on. Click-through rate is a weak guide: an ad can earn clicks from people who never pay.
- For subscription apps, read trials by concept. Cost per conversion at the trial stage is useful, but compare trial-to-paid conversion between concepts too. A cheaper trial does not necessarily produce cheaper paying customers.
- Distinguish iOS reporting methods. In SKAN 4, conversion windows cover days 0 to 2, 3 to 7 and 8 to 35 after first launch. After each window closes or locks early, random delays are 24 to 48 hours for the first postback and 24 to 144 hours for later ones. Eligible Meta AEM reporting can be near real-time; it does not make delayed SKAN cohorts complete. Our iOS attribution guide covers the distinction.
What should you test next?
Both new concepts and variants of what is winning. A variant (a new hook, new on-screen text, a different visual treatment) gets more out of a winner. A new concept (a different claim, pain or reason to install) finds the next one. Count them separately so a week of variants doesn’t pass for a week of new ideas: see creative concepts versus variants and, for scale, how to scale your creative strategy to thousands of ads.
Frequently asked questions
Should I include a control ad in a creative test? We usually keep established ads out of a new-creative screening ad set so they do not dominate its budget. If you need a controlled comparison with your current best, include it as the control in a formal A/B test.
How much budget does a creative test need? Our guide is ideally 10 conversions a day per creative test, with 2 or 3 the bare minimum only at very high cost per conversion. Multiply the intended daily volume by your expected cost per conversion to estimate daily funding. Plan the total duration and sample separately.
Is automatic delivery an A/B test? No. Unequal delivery within an ad set helps the platform pursue its objective, but does not isolate a creative’s causal effect. A formal experiment separates the audience groups and compares a planned variable; it can still finish without a clear winner.
What tools do you need for asset-level creative testing? Ad-level reporting in each platform plus your mobile measurement partner covers most of it; a creative analytics tool helps once you run hundreds of ads. How to choose a creative analytics tool for mobile ads covers what to look for.
