Sample Size Calculator
Plan a two-group experiment around the smallest effect worth detecting — for conversion rates or numeric metrics.
Why use this calculator
Set the test size before the result can influence the decision
A sample size calculator turns a business-relevant minimum detectable effect into a concrete experiment plan. It helps avoid tests that finish with too little evidence to detect a real win, while keeping the required traffic proportional to the decision you need to make.
How to use this estimate
For conversion metrics, enter the rate among all assigned users — including zeroes for people who do not convert. For numeric metrics, use a standard deviation estimated at the same unit of analysis as the final test.
The smaller allocation is not a reason to wait for a balanced sample. The estimate above calculates the observations required in each arm at the allocation you selected.
How to calculate sample size for an A/B test
An A/B test sample size calculation answers a practical planning question: how many assigned users are needed to reliably detect the smallest effect that would matter? The answer depends on the metric, its baseline variability, the minimum detectable effect (MDE), your significance level, and the power you want.
Start with the decision, not the traffic
For a conversion rate, choose a baseline and a relative lift worth acting on. For example, a 5% conversion rate and a 10% relative MDE means planning to detect a change from 5.00% to 5.50%. For a numeric metric, use the baseline mean and standard deviation from comparable historical data, then enter the smallest absolute difference that matters.
Power and alpha set the evidence threshold
Alpha controls the planned false-positive rate; 5% is a common two-sided default. Statistical power is the chance of detecting the chosen effect if it is real; 80% is a common starting point. A smaller MDE, lower alpha, or higher power requires more observations.
Unequal traffic changes each arm’s requirement
A 50/50 split is efficient, but experiments sometimes allocate less traffic to a risky treatment. In that case, the calculator increases the larger arm so the smaller arm still has enough observations. If you enter daily eligible traffic, the result also estimates how long the planned allocation will take to fill.