Minimum detectable effect

The minimum detectable effect is the smallest change in conversion rate an A/B test is designed to detect. The smaller it is, the more visitors the test needs.

Why it matters

It's the input that drives sample size most. Halving the effect you want to detect roughly quadruples the visitors you need, which is why low-traffic pages can only reliably test big changes.

How to get it right

• Set it to the smallest lift that would change your decision, not the lift you hope for.
• Express it as a relative change: +20% takes a 3% rate to 3.6%.
• If the required sample is out of reach, test a bigger change instead.

Common mistakes

• Choosing a tiny effect and then stopping the test early because it takes too long.
• Mixing up relative and absolute changes.