A/B Test Sample Size

Compute the required sample per variant for a two-proportion test. Choose relative or absolute lift, significance, and power. Runs fully in your browser.

Inputs

Used to estimate test duration with 50/50 split.
We’ll flag if estimated duration exceeds this.

Results

Per variant
—
Equal allocation
Total sample
—
A + B
Absolute Δ
—
percentage points
Zα / Zβ
—
critical values
Est. duration
—
with 50/50 split

Summary

—

Notes

Formula

Two-proportion z-test with pooled variance: n per variant = 2·(Zα√(2 p̄q̄) + Zβ√(p₁q₁ + p₂q₂))² / Δ².

Relative vs absolute

Relative MDE (%) applies a % lift to the baseline; absolute MDE is the change in percentage points.

Power & α

Higher power or lower α requires larger samples. One-sided tests need fewer samples than two-sided.

Duration

Duration assumes 50/50 split and constant traffic; real-world guardrails (seasonality, novelty) still apply.