How to use it
Enter total visitors and conversions for each variant. The tool returns both conversion rates, the difference, and a significance value indicating how likely that difference is to be chance.
← Tools center
A/B Test Significance Calculator
Enter the relevant inputs, run the tool safely, and review a measurable result immediately.
Local utility
This tool does not call an external service. Sensitive inputs are not persisted.
What this tool checks
The difference between the two rates is evaluated with a two-proportion z-test. The calculation uses only the four numbers you enter.
How to read the result
Significance does not prove the difference is real; it states how unlikely the observed gap is under chance alone. A 95% confidence threshold is common convention, not a mathematical requirement. A test that does not reach significance usually means "not enough data yet" rather than "no difference".
Examples
Compare two landing page variants, two headlines, or two form lengths.
Use cases
Useful for campaign teams, landing page owners and conversion optimisation work.
Common mistakes
The most damaging mistake is stopping the moment significance appears, which sharply increases false positives. The second is ending mid-week: weekday and weekend behaviour differ. The third is changing more than one variable and attributing the difference to a single cause.
What this tool cannot see
The tool knows nothing beyond the numbers you type. It cannot see whether traffic was actually randomised, whether the same user was counted twice, seasonality, or a campaign running at the same time. Without those, the result is arithmetically correct but practically misleading.
Privacy and data handling
The numbers you enter are not stored; only an anonymised run record is kept.
Frequently asked questions
When should I stop the test?
Decide the duration and sample size before starting and hold to it. Stopping as soon as significance appears inflates the false positive rate.
Does 95% significance prove the difference is real?
No. It says the observed difference is unlikely under chance alone; it is not a claim of certainty.
Can I use it with low traffic?
You can, but reaching significance is hard at low volumes and any result you do get is fragile.