Experiment planning tool

How much evidence do you need?

Set a baseline conversion rate and the smallest relative lift worth detecting. This calculator estimates the users needed in each variant before you start a fixed-horizon binary A/B test.

Four assumptions

Plan the test

50 / 50

Use a binary conversion metric and unique users as the sample unit. Allocation is fixed at 50 / 50 to keep the estimate easy to audit.

The current Control conversion rate.

10 means detecting a 10% relative improvement.

Two-sided threshold for a false positive.

Chance of detecting the chosen MDE if real.

This is a normal-approximation estimate for a two-sided, fixed-horizon, two-proportion test with equal allocation. It is not a substitute for a power analysis designed for revenue, ratios, retention, sequential testing, or clustered users.

Plan before the first user Ready to plan

Your test needs a meaningful target.

Update the assumptions to estimate the required sample size.

Treatment target
—
Absolute MDE
—
Users / variant
—
Total users
—
What this number means

The estimate is the minimum rounded-up sample for each equally sized variant. Give the test its planned duration before interpreting the result.

Calculated locally in your browser. No planning inputs are sent anywhere.

Term guide

Plain English
MDE
Minimum detectable effect: the smallest change this plan is designed to reliably distinguish from no change.
Relative lift
The change divided by Control. A move from 10% to 11% is a 10% relative lift.
Absolute MDE
The same target expressed in percentage points. A move from 10% to 11% is +1 percentage point.
Significance level
The false-positive threshold for the two-sided test. A 5% level corresponds to a 95% confidence interval.
Power
The chance of detecting the chosen MDE if that change is real under the planning assumptions.
Equal allocation
Half of the planned users go to Control and half to Treatment. This calculator keeps that split fixed.