Minimum Detectable Effect Calculator

The Minimum Detectable Effect is the smallest true difference in conversion rates that your A/B test has sufficient statistical power to reliably detect. If the real effect is smaller than the MDE, your test is likely to miss it (a false negative). Setting a smaller MDE requires a larger sample size.

Relative MDE expresses the detectable change as a percentage of the baseline — for example, a 10% relative MDE on a 10% baseline means detecting a change to 11%. Absolute MDE is the raw percentage-point change, so a 1 pp absolute MDE on the same baseline also means detecting a shift to 11%. Both convey the same threshold; the framing differs.

Start by asking what change would be meaningful or actionable for your business. If a 5% relative improvement in conversions would justify rolling out a change, set your MDE to 5%. Smaller MDEs demand more traffic and longer test durations, so balance statistical rigor with practical constraints.

95% (α = 0.05) is the widely accepted industry standard and means you accept a 5% chance of a false positive. Higher significance (99%) reduces false positives but requires larger samples. Lower levels (90%) are used when speed matters more and the cost of a false positive is low.

Statistical power (1 − β) is the probability that your test will correctly detect a true effect when one exists. A power of 80% means there is a 20% chance of missing a real difference (Type II error). Higher power is safer but requires more visitors. 80% is the conventional minimum for A/B tests.

The required sample size depends on your baseline conversion rate, the MDE you want to detect, your significance level, and your desired power. Generally, lower baseline rates or smaller MDEs require more visitors. Use this calculator in reverse — set a target MDE based on your available traffic and see what effect you can realistically detect.

Run your test until each variation reaches the required number of visitors per variation. Stopping early — even if results look significant — inflates false positive rates. As a rule of thumb, run for at least one full business cycle (typically 1–2 weeks) to account for day-of-week effects.

A high MDE with a large sample often results from a very low baseline conversion rate. When conversions are rare, variance is high and you need more data to distinguish signal from noise. Try increasing traffic to each variation or, where possible, choose a higher-converting funnel step as your primary metric.