How it works
The method
It uses the usual two-proportion z-test. You give the baseline conversion rate, the smallest lift worth detecting (in percentage points or as a percent of the baseline), the significance level and the power. The tool returns the visitors needed per variant.
The total is that number times the variants, and if you enter daily visitors it also gives the days the test must run.
n = (zα·√(2p̄(1−p̄)) + zβ·√(p₁(1−p₁)+p₂(1−p₂)))² ÷ (p₂−p₁)²
Choosing the inputs
Significance 0.05 and power 0.8 are common defaults. A one-tailed test needs fewer visitors but only looks for a change in one direction. A smaller lift or a lower baseline rate needs many more visitors.
