A/B Testing Calculator
Calculate the sample size and estimated duration for a statistically significant A/B test. Two-proportion formula with 80% statistical power.
Test parameters
95% is the benchmark for most CRO tests.
The percentage of users who complete the desired action in the control variant.
The minimum relative variation you want to detect. For example: 20% means detecting an increase from 5% to 6.00%.
The total of visitors distributed between variant A and variant B (50/50).
Results
ValidHow to Interpret Results
- Campione per variante: quanti utenti devono vedere ciascuna variante (A e B) prima di poter trarre conclusioni.
- Minimum Detectable Effect (MDE): minimum detectable relative variation. A 20% MDE on a 5% rate means detecting a change up to 6%.
- Giorni stimati: calcolato dividendo il campione totale per i visitatori giornalieri inseriti.
- Statistical power 80%: the test has an 80% probability of detecting a real effect when it exists.
Come utilizzare Calcolatore A/B Test
Set the current exchange rate
Enter your control variant's conversion rate, for example, the percentage of visitors who complete a purchase or registration today.
Define Minimum Detectable Effect (MDE)
The MDE is the minimum relative variation you want to be able to detect compared to baseline. Smaller values require larger samples.
Choose confidence level
Select 90%, 95% or 99% statistical confidence. 95% is the most commonly used in conversion rate optimization (CRO) tests.
Enter daily visitors and read result
With total daily traffic, the calculator estimates how many days are needed to reach both variants' required sample size.
Suggerimenti
- Don't interrupt A/B tests before reaching required samples even if partial results seem already favorable: risk "peeking" and wrong conclusions.
- If traffic's low, consider raising MDE (detect only larger variations) instead of stretching the test too long.
- Run tests for complete weeks (multiples of 7 days) to neutralize behavior differences between weekdays and weekends.
Domande frequenti
What's a sample size (sample size) in an A/B test?
It's the minimum number of users each variant must receive before the test results are statistically reliable. Too small a sample exposes you to the risk of concluding that one variant is better when observed differences are just random.
What does Minimum Detectable Effect (MDE) mean?
It's the smallest relative conversion rate variation that the test can detect with statistical certainty. A 20% MDE on a base rate of 5% means the test is designed to detect an increase up to about 6%. Smaller MDEs require much larger samples.
Why does confidence level change sample size?
Higher confidence levels (e.g., 99% instead of 95%) reduce false positives but require more data for the same reliability, thus increasing required sample and test duration.
What happens if a test lasts over 30-90 days?
Tests that last too long risk being influenced by external factors (seasonality, marketing campaigns, holidays) affecting results. If estimated duration exceeds 30-90 days, consider boosting traffic to the page, raising MDE or reducing confidence level.
What does an 80% statistical power indicate?
Statistical power indicates the probability that a test will detect real effects when they actually exist. 80% is industry standard: it means if variant B really outperforms, this test has an 80% chance of showing so with calculated sample size.