Conversion rates with confidence intervals
Breakdown by variant
| Variant | Visitors | Conversions | Rate | Confidence interval |
|---|
Copy the verdict and figures to paste into a report or ticket.
What other lift sizes would cost
| Relative lift | Target rate | Per variant | All variants | Days |
|---|
Copy the plan to share with your team before the test starts.
What This Calculator Tests
An A/B test shows two versions of something - a landing page, an email subject line, a checkout button - to randomly split groups of visitors, then compares how many people in each group completed the goal, such as signing up or buying. The raw gap between the two conversion rates is only part of the story, because a small sample can produce a large gap purely by chance.
This tool runs a two-proportion z-test. It pools both groups to estimate the shared rate you would expect if the versions were identical, measures how far the observed difference sits from that expectation, and converts the distance into a p-value. A result counts as significant when the p-value is at or below 1 minus your chosen confidence level - so at 95% confidence, a p-value of 0.05 or less.
How to Use It
Practical Use Cases
Limits and Notes
| Situation | What to keep in mind |
|---|---|
| Very few conversions | When a variant has fewer than about 5 conversions or non-conversions, the normal approximation is unreliable. The calculator warns you in that case. |
| Stopping early | Checking the result repeatedly and stopping the moment it turns significant inflates false positives. Decide the sample size in advance with the planner. |
| Three variants | Each challenger is tested against the control separately. Testing several comparisons raises the chance that one looks significant by luck. |
| Unit of counting | Enter visitors and conversions on the same basis, such as unique users or sessions. Mixing the two gives meaningless rates. |
A significant result tells you a difference is unlikely to be pure chance under this model. It does not tell you how big the effect will be over the long run, or whether it is worth the cost of the change.
Frequently Asked Questions
No. The visitor and conversion counts you enter are processed by JavaScript in your own browser and nothing is uploaded. The figures are saved in this browser's local storage only, so they are still there when you come back to the page.
A higher conversion rate is not enough on its own. The test asks whether a gap of this size would be unlikely if both versions truly performed the same. With small samples, ordinary random variation can produce a large gap, so the p-value stays above the threshold until enough visitors have been counted.
A two-tailed test checks whether the variant differs in either direction, which is the safer default. A one-tailed test only checks for a change in one direction and reaches the threshold with less data, but it cannot tell you that the variant is worse. Choose the direction before you look at the data.
Use the sample size planner. Enter your current conversion rate, the smallest relative lift you care about, the significance level and the power. The planner returns the number of visitors each variant needs, and converts that into days if you enter your daily traffic.