Free tool
A/B test sample size calculator
Find out how many visitors each version of your landing page needs before a test can tell them apart, and how many weeks that takes at your traffic.
The share of visitors who convert on the page today.
Relative change. 20% takes a 3% rate to 3.6%.
Optional. Used to estimate how long the test runs.
Confidence
Power
Visitors needed per variant
13,914
- Total for A and B
- 27,828
- Test length
- 6 weeks
To tell 3% from 3.6% with 95% confidence and 80% power.
How to read the result
The number is how many visitors each version needs. A standard A/B test splits traffic evenly, so the test needs twice that in total. If your page gets fewer visitors than that in a few weeks, the test is unlikely to give a reliable answer.
The input that matters most is the smallest lift you want to detect. Small changes, like a button colour, tend to have small effects, and small effects need very large samples. Bigger changes to the offer, the headline or the page structure are more likely to move the conversion rate enough to measure.
Before you start a test
- Fix the sample size first, and don't stop the test early because one version looks ahead.
- Run whole weeks, so weekday and weekend visitors are both included.
- Change one idea per test, so you know what caused the result.
- Pick the test worth running. Each test takes weeks, so spend them on the ideas most likely to matter.
Decide what to test faster
Most landing pages can run only a few tests a year. Klar's Copy Lab puts up to 5 versions of your headline, subline and CTA in front of a panel of simulated buyer personas in minutes, so you can take the strongest to a real test. The panel is simulated, so its picks are a way to choose what to test, not a replacement for the test. For more, read how to get honest feedback on your landing page.
Questions, answered.
- What is a minimum detectable effect?
- The smallest change in conversion rate you want the test to be able to spot. A smaller effect needs many more visitors: halving it roughly quadruples the sample size.
- What do confidence and power mean?
- Confidence (95% is standard) limits how often you'd call a winner when there's no real difference. Power (80% is standard) is how often the test finds a real difference of the size you set.
- Can I stop the test early if one variant is winning?
- Not safely with this kind of test. Checking results repeatedly and stopping at the first good-looking moment inflates false positives. Decide the sample size up front and wait for it.
- What if I don't have enough traffic?
- Test bigger changes, which are easier to detect, test on a page with more traffic, or use faster methods to decide what's worth testing at all.
- How is this calculated?
- With the standard formula for a two-sided test of two proportions, using the normal approximation. It's the same method most sample size calculators use.
Find what's worth testing on your page.
Paste your URL. Five buyer personas read it and tell you what to fix first. Free while in beta.