Plug in reach and conversions for each variant. The calculator compares them to your control and tells you if the gap is real or still noise.
This compares conversion rates between your control and each variant with a two-proportion z-test. You get a confidence score for whether a challenger actually beats the control, or just got lucky on a thin sample.
It also nags you about the stuff that breaks Meta tests in practice: cutting a test short, calling winners while the ad set is still in learning, scaling before you have enough conversions, or comparing variants that did not get a fair split of reach.
Reach for this when:
For purchase campaigns, do not make the call on CTR or CPC alone. Judge on conversions (purchases, leads, or whatever you optimize for) once you have enough volume behind the number.
The full workflow (what to test, budget split, kill rules) lives in our Meta ads testing framework guide.
There is no one number that fits every account. For most ecommerce Meta tests, these floors are a sane starting point:
The calculator turns those into warnings on the results panel so you do not scale on a coin flip.
Most operators want at least 50 conversions per variant on purchase campaigns before they crown a winner. This tool uses 50 as a warning threshold and 95% confidence before it calls a result significant. Below that, a lift can look real and still flip if you wait another week.
It means the gap between variants is probably not random noise. This calculator uses 95% confidence: hit that bar, clear the guardrail flags, and you can treat the leading variant as a real win.
When the challenger beats the control, you have roughly 50+ conversions per variant, the test has run about a week, reach was split evenly, and the ad set is out of learning. A hot CTR day is not enough.
Plan for at least seven days and enough optimization events to exit learning. Meta often wants on the order of 50 weekly optimization events per ad set. CTR in the first 48 hours is a hint; conversions are the verdict.
Better hypotheses beat better spreadsheets. The sample reports show how we break down a brand's live Meta ads, so you can see the format before you order one on your own competitor and sketch your next variant from real market data.