The minimum detectable effect (MDE) is the smallest lift a test can reliably identify. Before running an experiment, you set an MDE to calculate how much traffic and time you need. If the true effect of your change is smaller than your MDE, the test will likely miss it.
Why MDE Matters
Setting the right MDE forces a critical question: how big does the effect need to be to matter?
A landing page with a 2% baseline conversion rate might not care about a 0.1% absolute improvement — that's noise in your data and negligible revenue impact. But a 0.3 percentage point lift (15% relative improvement) could be meaningful. Setting MDE = 15% relative lift tells your sample size calculator what it needs.
Choosing too small an MDE means you need enormous sample sizes and long runtimes. Choosing too large an MDE means you'll miss real but modest improvements.
MDE and Sample Size
MDE, sample size, confidence level, and statistical power are all linked. Holding power and confidence constant:
| MDE (relative lift) | Approximate traffic multiplier |
|---|---|
| 5% | ~16× baseline sample |
| 10% | ~4× baseline sample |
| 20% | ~1× baseline sample |
| 30% | ~0.5× baseline sample |
Halving the MDE roughly quadruples the required sample size.
How to Choose an MDE
- Revenue threshold — What lift is worth shipping? Calculate the revenue impact of a 5%, 10%, and 20% improvement on current traffic and decide the minimum business-meaningful threshold.
- Traffic reality — Run a sample size calculator with your baseline conversion rate and traffic volume. If detecting a 5% lift requires 6 months of data, you may need to raise the MDE or limit the test to high-traffic pages.
- Historical test data — Look at past winning tests. If your typical winning lift is 15–25%, calibrate your MDE to that range.
MDE vs. Observed Lift
MDE is a planning parameter set before the test. The observed lift is what you measure after. If your observed lift is smaller than your pre-set MDE, you likely ran an underpowered test — the result may be unreliable even if it reaches significance.
Always specify your MDE in the test brief before launching.