Loading...
Loading...
Estimate the minimum detectable effect for your A/B test. Know your experiment's sensitivity before you launch.
The MDE Calculator estimates the smallest effect size your A/B test can reliably detect given your current traffic and statistical parameters. Understanding your minimum detectable effect before running a test helps you set realistic expectations. If the MDE is larger than the effect you expect from your change, consider increasing traffic, reducing variants, or accepting lower confidence.
The Minimum Detectable Effect is the bridge between statistical theory and business decisions. Before launching any A/B test, you should know what your experiment is capable of detecting. If your MDE is 10% but you expect a 2% lift, you're running a test that will almost certainly fail to reach significance, even if the treatment genuinely works.
MDE depends on three factors: sample size, statistical power, and confidence level. For a fixed significance threshold (typically 95%) and power (typically 80%), MDE is entirely determined by your available traffic and the natural variability of your metric. Understanding this relationship helps you make informed decisions about experiment duration and which metrics to test.
The MDE vs Sample Size chart shows how sensitivity improves with more data. The curve follows an inverse square root relationship. The first few thousand users provide the biggest improvement, while further gains require exponentially more traffic. The dashed line marks your current sample size.
For conversion rate metrics, MDE is expressed as a relative percentage lift from the baseline. A 5% baseline with a 10% relative MDE means you can detect a shift from 5.0% to 5.5%. For continuous metrics like revenue or session duration, MDE is in absolute units. Cohen's d provides a standardized comparison across different metric types.
Explore more A/B testing and statistics tools
Calculate the minimum sample size for your A/B test.
Calculate standard deviation, variance, and coefficient of variation.
Analyze A/B test results with frequentist Z-tests and T-tests.