A/B test sample size calculator

The calculator uses the standard formula for comparing two proportions with a two-sided test:

n = (z₁₋α/₂ · √(2·p₁(1−p₁)) + z₁₋β · √(p₁(1−p₁) + p₂(1−p₂)))² ÷ (p₂ − p₁)²

• p₁ is the baseline rate and p₂ the rate you want to be able to detect (baseline + effect).
• z₁₋α/₂ comes from the significance level: 1.645 for 90%, 1.960 for 95%, 2.576 for 99%.
• z₁₋β comes from the power: 0.842 for 80%, 1.282 for 90%.

It is the same formula used by Evan Miller’s widely cited calculator; we round up to the next whole visitor. Example: a 20% baseline and a 5-point absolute effect at 95% significance and 80% power gives about 1,030.2, so 1,031 visitors per variant.

• Minimum detectable effect: the smallest lift that would change your decision. Halving it roughly quadruples the sample.
• Significance (95% is standard): how sure you want to be that a “winner” isn’t a fluke.
• Power (80% is standard): the chance of detecting the effect if it is really there.

Decide the sample size up front and don’t stop the test the moment it looks significant — “peeking” inflates false positives dramatically.

If the duration comes out in months, test bolder changes (a new offer, a shorter signup) instead of button colours, or test higher in the funnel where there is more traffic. Analisa’s funnels show where most visitors drop off, which is usually the best place to test. Get your baseline with the conversion rate calculator.