A/B Test Sample Size Calculator

How many users per variant do you actually need? Two-sided test at 95% confidence.

Baseline conversion (%)

Min detectable effect — rel. lift (%)

Statistical power

Per variant (frequentist)—

Total (both variants)—

Per variant (Bayesian)—

Detecting lift to—

Frequentist uses the two-proportion z-test (alpha=0.05, two-sided). Relative MDE: 20% on a 3% baseline targets 3.6%. Don't peek before hitting n.