Free calculator

Sample size calculator for A/B tests

Estimate the visitors per variant you need to catch a minimum detectable effect at a given confidence and power. Two-proportion formula, industry-standard defaults.

Your current control rate. 5 = 5%.

Relative lift you want to catch. 10 = detect ≥ +10% over baseline.

95 is standard. Higher = harder to reject the null.

80 is standard. Chance of detecting a real effect.

Visitors per variant

31,235

Total across both

62,470

Reach this per variant before you trust the result. Halt the test earlier and you risk a false positive.

Rules of thumb

  • Sample size scales inverse-quadratically with MDE. Halving your MDE target quadruples the sample size you need.
  • Low baselines are expensive. A 1% baseline needs ~25× the traffic of a 20% baseline for the same relative MDE.
  • 95% confidence + 80% power are the defaults everyone uses. Higher = tighter but longer.
  • Round up. Never launch a test with fewer visitors than the calculator says.

Related