Minimum Detectable Effect (MDE) defines the smallest change in your can reliably identify. Below this threshold, improvements become statistically indistinguishable from random variation in the data.
What is MDE?
MDE represents the minimum lift you can detect given your sample size and statistical settings. A 10% MDE means your test can detect a 10% relative improvement (e.g., from 2.0% to 2.2% conversion rate) but not smaller changes.
The relationship is inverse-square: halving your MDE roughly quadruples the required sample size. This is why choosing the right MDE matters for test planning.
How MDE Affects Sample Size
- 5% MDE requires around 50,000+ visitors per variant
- 10% MDE requires around 12,000+ visitors per variant
- 20% MDE requires around 3,000+ visitors per variant
Lower MDE values detect smaller improvements but need more traffic and longer test duration. To calculate exact sample size for your MDE and conversion rate, use an A/B test sample size calculator.
Choosing the Right MDE
Consider your traffic volume first. Limited traffic requires higher MDE (15-20%). Attempting to detect small effects with low traffic leads to tests that run for months without reaching .
Then consider whether the MDE represents meaningful business impact. If baseline conversion is 2% and MDE is 20%, you're testing for a lift to 2.4%. Calculate whether that difference justifies the test.
Common Mistakes
- Setting MDE too low results in tests that never complete
- Ignoring MDE means running tests without understanding what you can actually detect
- Using same MDE for all tests ignores that different changes have different expected effect sizes
