A/B Test Sample Size Calculator
Work out how many visitors each variant needs before you start an A/B test, based on your baseline conversion rate and the smallest uplift you care about.
Enter your baseline conversion rate and the smallest uplift worth detecting. Assumes 95% confidence and 80% power.
Why sample size comes first
Deciding the sample size before the test starts is what keeps the result honest. If you run until the numbers look good, you will mostly catch noise. This calculator uses the standard two-proportion formula at 95% confidence and 80% power, the defaults most testing tools use.
Choosing the minimum detectable effect
The minimum detectable effect (MDE) is the smallest relative uplift you would act on. It drives the sample size more than anything else: detecting a 5% uplift needs roughly four times the traffic of detecting a 10% uplift, because required sample size scales with the inverse square of the effect.
Be realistic about it. Most site changes move conversion by single-digit percentages, but if your traffic only supports detecting a 30% uplift in a reasonable time, test bolder changes rather than running an underpowered test for months.
Working out test duration
Divide the total required sample (both variants combined) by your daily visitors to the page under test. If the answer is more than four to six weeks, the test will drag through seasonal noise and team patience; raise the MDE, test a higher-traffic page, or pick a metric earlier in the funnel where conversions are more frequent.
Once you know the number, run the test to completion and check the result once with the significance calculator.
Questions fréquentes
What is statistical power?
Power is the probability that your test detects a real effect of the size you specified. At the standard 80% power, a true uplift at your MDE will be detected four times out of five. Higher power needs more traffic.
What baseline conversion rate should I enter?
The current conversion rate of the page or step you are testing, measured over a recent representative period. If your checkout converts at 3.2% over the last month, enter 3.2.
Is the sample size per variant or total?
The result shown is per variant. A two-variant test needs twice that number in total visitors, split evenly by random assignment.
What if I cannot reach the required sample size?
Test a bigger change (raise the MDE), test on a higher-traffic page, or measure a more frequent event earlier in the journey. An underpowered test rarely produces a trustworthy answer, however long it runs.
Why 95% confidence and 80% power?
They are the established defaults that balance false positives against traffic cost. Nothing stops you using stricter values for high-stakes changes; the required sample grows accordingly.
La boîte à outils complète, dans un seul onglet
Analyse combine analytique sans cookies, tunnels de conversion, rétention et un moteur SEO qui rédige et publie des articles documentés. Essayez toutes les fonctionnalités gratuitement pendant 14 jours.
Démarrer l'essai gratuit