Skip to content

A/B test sample size calculator.

Work out how many visitors each version needs before you launch, and how long the test has to run at your traffic. Sizing a test first is what stops it ending without an answer.

Confidence level
Hypothesis
Percent, so 3 means 3%
Relative to your current rate, so 10 means 3% to 3.3%
2 to 8, split evenly
Visitors who actually enter the test

Traffic needed at other uplifts

Expected uplift Per variant Total visitors Days

Already ran the test? Check whether the result holds

How these numbers are calculated
Sample size

The same equation solved for visitors per variant, given a baseline rate, the uplift you want to detect, your confidence level, and statistical power fixed at 80%. Power is the chance of spotting a real effect of that size, and 80% is the planning convention, so it is set rather than exposed as an option.

n = ( z_crit + z_power )^2 x ( p_a(1-p_a) + p_b(1-p_b) ) / ( p_b - p_a )^2
Multiple comparisons

Sidak correction across k comparisons against the control, holding the test-wide false positive rate at the level selected.

alpha_adj = 1 - (1 - alpha)^(1 / k)
Limits
  • Fixed-horizon test. Use this figure as the sample size, run to it, and evaluate once on the significance calculator. Stopping at the first significant reading inflates false positives beyond the stated level.
  • Binary outcomes only. Revenue per visitor, average order value and items per order are skewed by large orders and need different methods.
  • The normal approximation needs roughly 5 conversions and 5 non-conversions per variant as a floor.
  • Significance is not effect size. A significant result can still be too small to be worth shipping.

Get started
with Wisepops

Book a demo to get a personalized audit of your traffic.

Help