Skip to content
TextArray1.41.0
100% local

A/B test significance calculator

Calculate conversion rates, lift and statistical significance across A/B test variants.

Input
Output
Related tools
Sum and number statisticsNumbers
Percentage calculatorNumbers
Percentage changeNumbers
Birth number validatorNumbers
Inline math evaluatorNumbers
Percent calculatorNumbers
A/B test significance calculator
Paste your test data as one line per variant — a name, the number of visitors and the number of conversions — and this tool runs a two-proportion z-test against the first line, which it treats as the control. It reports each variant's conversion rate, the absolute and relative lift over the control, a confidence interval and whether the difference is statistically significant.
The confidence level sets how strict the test is: 90%, 95% or 99%. A two-tailed test asks whether a variant differs from the control in either direction; a one-tailed test asks only whether it is better, which reaches significance sooner for the same data. Turning on the confidence interval adds a range for every variant's true conversion rate, not just the control. Choose a table for a compact side-by-side comparison, or a sentence for a plain-language readout ending in a stated winner.
Real traffic is messy, so the tool also checks sample size: when the expected number of conversions or non-conversions for any variant falls below five, it adds a note that the normal approximation behind the test may be unreliable, rather than reporting a confident-sounding number that isn't. You can compare more than two variants at once — every line after the control is tested independently against it, and the winner is the highest-converting variant that beats the control by a statistically significant margin.
Everything runs locally in your browser. Traffic and conversion numbers can be commercially sensitive, and none of it is ever uploaded — the calculation happens on your device and nowhere else. Copy the result, download it as a .txt file, or send it back into the input to adjust a number and see the significance update.
FAQ
How is significance calculated?
The tool runs a two-proportion z-test between the control (the first line) and each other variant, using a pooled standard error, and compares the resulting p-value against your chosen confidence level.
What's the difference between one-tailed and two-tailed?
A two-tailed test checks for any difference from the control, better or worse. A one-tailed test only checks whether the variant is better, so it needs less extreme data to call a result significant.
Why does it warn about sample size?
The z-test assumes a roughly normal distribution, which needs at least about five expected conversions and five expected non-conversions per variant. Below that, the tool flags the result as potentially unreliable instead of hiding the caveat.
Can I test more than two variants?
Yes. Add as many lines as you like after the control — each one is compared against the control independently, and the winner is the best variant that clears your significance threshold.
Is my traffic and conversion data uploaded anywhere?
No. The calculation runs entirely in your browser, so your visitor counts and conversion numbers never leave your device.