← Back to Tools

A/B Test Duration Calculator

Running tests too short? Making decisions on noise? Calculate exactly how long your A/B test needs to run for reliable results.

%

Your current trial start rate, paywall conversion, or whatever metric you're testing.

A 10% relative lift on a 5% baseline means detecting 5% → 5.5%.

Daily users who see your paywall, onboarding screen, or test variant.

95% is industry standard. Higher = longer test but fewer false positives.

How This Calculator Works

The Math

This calculator uses the standard sample size formula for two-proportion tests with:

  • Two-tailed test (detecting improvements OR declines)
  • 80% statistical power (industry standard)
  • Equal traffic split between control and variant

Why Sample Size Matters

Smaller effects need larger samples to detect reliably. A 5% → 5.5% lift is real, but you need thousands of users to distinguish it from random noise. A 5% → 6.5% lift is easier to spot with fewer users.

When to Adjust

  • Low traffic? Accept a larger MDE (15-20%) or run longer
  • High stakes? Use 99% confidence to reduce false positives
  • Exploratory? 90% confidence is fine for initial tests