Formulas used by this calculator
- Monthly testing budget = monthly ad budget × testing share
- Spend per ad before a verdict = target CPA × judging multiple
- New ads you can judge per month = testing budget ÷ spend per ad, rounded down
- Days to a verdict for a round = ads per round × spend per ad ÷ (testing budget ÷ 30)
- ROAS at target CPA = value per conversion ÷ target CPA
Worked example
Illustrative inputs: $6,000 monthly budget, 20% for testing, $45 target CPA, $90 average order value, each ad judged after 2× target CPA, three new ads per round.
The testing budget is $1,200. Each ad needs $90 before a verdict, so the month can judge 13 new ads (1,200 ÷ 90 = 13.3, rounded down).
Three ads share $40 a day, so a round needs 270 ÷ 40 = 6.75 days. An ad that hits the $45 target would show only two conversions by then, which is enough to cut clear losers but not to prove a winner.
Why the spend per ad is tied to your CPA
An ad that has spent less than one target CPA has not had a fair chance to produce even one conversion at the cost you can afford. Judging it earlier mostly measures luck. Giving each ad two or three target CPAs lets a working ad show a few conversions and lets a clearly failing ad show that it is failing.
The trade-off is speed. A higher multiple gives steadier verdicts but fewer ads per month and longer rounds. A multiple of 1 is useful for cutting obvious losers fast; it is not evidence that the survivor is better than your current ad.
Meta's Marketing API documentation for split tests notes that larger reach, longer schedules or higher budgets tend to give more statistically significant results, and that a test should change one variable at a time. Meta for Developers: split testing.
Small budgets should test fewer ads
If the testing budget funds three judged ads a month, launching ten spreads the money so thin that none reaches a verdict. Launch fewer ads at once and make each one a genuinely different idea (a new angle, offer or format), not ten colour variations of one ad.
The calculator warns when one round of your chosen size cannot be funded in a month. Lower the round size, lower the judging multiple, or raise the testing share, and see which trade-off you can live with.
Screening is not proof
This calculator plans a screening budget: enough spend to keep or cut ads sensibly. To claim that one ad truly beats another, use a controlled experiment with a sample size planned in advance. The A/B test sample-size calculator in the free tools library plans that for a two-version test.
Platforms rarely split spend evenly across ads in the same ad set, so some ads will reach their spend threshold much later than others. Treat the days figure as the fastest possible verdict, not a schedule the platform will follow.
Common questions
What judging multiple should I use?
Two times target CPA is a common middle ground. Use 1 when you need to cut clear losers quickly and accept noisy calls. Use 3 or more when ads are expensive to make or a wrong call is costly. Write the rule down before launch so you do not move it after seeing results.
Should I use expected CPA or target CPA?
Use the CPA you can afford if the account has little history. If the account has a steady CPA from mature data, that number tells you what a typical ad costs per conversion, which is a fair bar for new ads.
Is 20% the right testing share?
It is an example, not a rule. Accounts with no proven ads may spend most of the budget on testing; accounts with steady winners often spend less. Pick a share you can sustain and review it monthly.
