Calculate the required duration of your next A/B test
A/B Test Duration Calculator
Find out exactly how many days your A/B test needs to run to get reliable results.
%
%
%
%
Test duration
—
Sample size per variation
—
Total visitors needed
—
Conversions needed
—
Estimated end date:
FAQ
Frequently asked questions
How long should an A/B test run?
There is no single answer that fits every test. The right duration depends on your daily traffic, baseline conversion rate, and the size of the change you want to detect. Use the calculator above to get a number specific to your test. As a general rule, most website tests should run between 2 and 6 weeks.
What happens if I stop my test early?
Stopping early is one of the most common A/B testing mistakes. If you check results daily and stop the moment a variation looks like a winner, you are almost guaranteed to see false positives. The effect you see early in a test is often noise, not signal. This is called the "peeking problem." Let the test run to its planned duration before drawing conclusions.
What is the minimum detectable effect (MDE)?
The MDE is the smallest improvement in conversion rate you want your test to be able to detect. If you set an MDE of 10%, the test is designed to find lifts of 10% or larger. Smaller MDEs require longer tests because small differences are harder to distinguish from random variation. Set your MDE based on what lift would actually make a business difference, not the smallest possible number.
Does running more variations make a test longer?
Yes. Each variation needs its own sample of visitors. A test with 3 variations (A/B/C) needs 50% more total visitors than a simple A/B test. A test with 4 variations needs twice as many. Focus on your strongest hypotheses first and test fewer variations at once to get results faster.
Why should I run my test for full business cycles?
A business cycle is typically one week: Monday through Sunday. User behavior shifts significantly between weekdays and weekends. A test that only runs Monday to Thursday will be biased toward weekday users. Running for at least two full weeks ensures your results reflect a representative mix of all user types, not just a snapshot of one behavioral pattern.
What is the minimum duration for an A/B test?
You should run an A/B test for at least 14 days, even if you reach statistical significance sooner. Two full weeks captures a complete weekday and weekend cycle, which matters because user behavior on Mondays looks very different from Saturdays. Stopping earlier increases the risk of false positives.
What if my test is taking too long?
If the calculator shows your test needs more than 8 weeks, you have a few options. First, check whether you can increase traffic to the tested page. Second, consider whether you really need to detect a 5% lift or whether a 15% lift would also be meaningful. Raising your MDE significantly shortens the required duration. Third, check if a lower confidence level (for example, 90% instead of 95%) is acceptable for this decision.
How does traffic volume affect A/B test duration?
Directly and proportionally. If your calculator says a test needs 28,000 total visitors and you get 500 per day, the test will take 56 days. If you drive 1,000 visitors per day, it takes 28 days. Increasing traffic is the most practical lever for shortening test duration when you cannot change the MDE or confidence requirements.
What confidence level should I use for an A/B test?
95% is the standard for most web experiments. It means there is a 5% chance the result you see happened by chance. For tests where the cost of a wrong decision is high (major redesigns, pricing changes), use 99%. For low-stakes tests where speed matters more, 90% is acceptable. Lowering the confidence level shortens your required test duration but increases your false positive rate.
Can I run multiple A/B tests at the same time?
Yes, if they test different pages or elements that do not interact with each other. Running two tests on the same page at the same time can cause interaction effects where the result of one test influences the other. If you must test the same page simultaneously, use a multivariate test or run the tests on separate, isolated traffic segments.
