Marketing A/B Test Sample Size Estimator

1. Enter the baseline rate
Use a recent conversion rate from traffic similar to the planned experiment.

2. Choose the minimum lift
Enter the smallest relative improvement worth detecting.

3. Set confidence and power
Higher values reduce false positives or missed effects but require more observations.

4. Read the per-variant requirement
Allocate at least the displayed sample to both control and test variants.

5. Allow for data loss
Increase the target if exclusions, tracking loss, or uneven allocation are expected.

p₂ = p₁ × (1 + Relative lift)
n ≈ [zα/2 × √(2p̄(1−p̄)) + zβ × √(p₁(1−p₁)+p₂(1−p₂))]² ÷ (p₂−p₁)²

p₁ is the baseline conversion rate, p₂ is the target test rate, p̄ is their average, zα/2 corresponds to the two-sided confidence level, and zβ corresponds to power.

What the result means

The output is the approximate sample required in each equally sized variant.

This normal approximation can be less reliable for very rare events, small samples, repeated peeking, clustered users, or multiple comparisons.

Given: 5% baseline conversion, 15% relative lift, 95% confidence, and 80% power.

Calculation: Target rate = 5% × 1.15 = 5.75%, an absolute difference of 0.75 percentage points. Applying the two-sided normal approximation gives about 14,758 observations per variant.

Result: Plan for roughly 29,516 total observations across two equal variants.