Check if an A/B test is significant

The job

Variant A converted 4.2%, variant B 4.6% — better, or luck? A two-proportion z-test compares the rates against the noise their sample sizes imply, and returns a z-score and p-value.

Run the test

Enter visitors and conversions for each variant. The tool reports each rate, the absolute lift, relative lift, the pooled z-score, and the two-sided p-value, then calls significance at α = 0.05.

• P < 0.05 reads as "the difference is unlikely under no real effect" — not "B is better by X".
• Not significant is not proof of no effect; it can also mean the sample is too small.

Worked example

A: 10,000 visitors, 420 conversions (4.2%). B: 10,000, 460 (4.6%). z ≈ 1.38, p ≈ 0.17 — not significant; the gap is consistent with noise.

Where it stops

Two variants only, two-sided, no continuity correction. Peeking daily and stopping when p dips under 0.05 inflates false positives — decide the sample first (use the sample-size tool), then measure once.