Detecting a smaller change requires sharply more observations.
A/B test sample size and duration calculator
Calculate required sample per variant and estimated duration from baseline conversion, MDE, significance, power, and daily traffic.
No CSVNo sign-upRuns in your browser
YOUR ASSUMPTIONSChange a number. The answer updates immediately.
Sample per variant31,235
Estimated duration63 days
Fix sample and duration before reading results to reduce early-stopping errors.
FORMULA
Continue to experiment design →Normal approximation for two proportions · two-sided testHOW TO READ IT
Use the answer as a decision boundary
Reducing missed effects requires more observations.
Questions before using this number
The minimum change worth detecting, fixed before the test. Changing it after seeing results breaks the design.
Repeated peeking inflates false positives. Fill the planned sample unless you use a sequential design.