Confidence and p-value
The confidence shown is 1 − p-value. The result is significant when it exceeds your chosen level: the difference is then unlikely to be due to chance.
RESULTS CALCULATOR
Is your A/B test really a winner?
Enter the figures for each variant. You’ll know whether the difference is significant, how much traffic is missing, and what the variant would earn once rolled out.
How to read these results
The confidence interval
It gives the plausible range of the true uplift. If it includes 0%, the observed difference could still be explained by chance: keep the test running.
Traffic split
If A and B don’t receive their planned share of traffic (Sample Ratio Mismatch), the test is probably biased. Check your setup before drawing conclusions.
Method: two-proportion Z-test (pooled variance). Expected split: 50 / 50. Impact is estimated on your current traffic, excluding seasonality.
Want an expert eye on your tests?
We’ll frame your hypotheses, test duration and how to read the results, in 30 minutes.
