A/B Test Calculator

Enter your control and variant counts. The calculator runs a two-proportion z-test with pooled standard error — the same test Optimizely, VWO, and most growth teams report as their default p-value. Significance is declared at p < 0.05.

Control (A)

Total visitors (or sessions) shown the control.

How many of those visitors converted.

Variant (B)

Total visitors shown the variant.

How many of those visitors converted.

Don't peek. Decide the sample size before you launch, then read the result once. Repeatedly checking p-values mid-test inflates your false-positive rate — a fancy way of saying you'll declare wins that aren't real.

The A/B Test Calculator runs the same statistical test that Optimizely, VWO, and most growth teams report on their dashboards: a two-proportion z-test with pooled standard error. Paste your control visitors + conversions and your variant visitors + conversions, and the calculator returns each arm's conversion rate, the absolute and relative lift, the z statistic, the two-tailed p-value, 95% confidence intervals for each rate, and a clear verdict at α = 0.05. Flip to the sample-size planner to find out how many visitors per arm you need before launching — given your baseline rate, the smallest lift you want to detect, your significance threshold, and your statistical power. No spreadsheet, no R session, no waiting for the data team.

Built by Bob QA by Ben Shipped

How to use

  1. 1

    In Significance test mode, enter total visitors and conversions for both the control (A) and variant (B). Visitors must be a positive whole number; conversions can be zero or higher, but never more than visitors.

  2. 2

    Read the verdict card: green when p < 0.05 (variant wins or control wins, depending on direction), navy when the data is consistent with random noise. The card also surfaces the exact p-value and a plain-English interpretation.

  3. 3

    Below the verdict, you get each arm's rate with a 95% confidence interval, the relative and absolute lift, the z-score, and the p-value. The confidence intervals are Wald-style; significance is judged at α = 0.05.

  4. 4

    Switch to Sample-size planner mode to size your next test. Enter the expected baseline conversion rate, the minimum lift you'd want to detect (as a relative percent — 10 means '10% relative lift over baseline'), and optionally adjust α (default 0.05) and power (default 0.80).

  5. 5

    Don't peek mid-test. Choose your sample size up front, run to that size, then read the result once. Repeated mid-test checks inflate your false-positive rate — that's why this calculator is two separate panels, not a real-time monitoring tool.

Frequently asked questions

Ratings & Reviews

Rate this tool

Sign in to leave a written review.
Loading reviews…