Statistics Didn’t Have to Be This Hard · Chapters 13 and 14

A/B Test

Did version B really beat version A? Enter visitors and conversions for each and see the size of the difference, how uncertain it is, and how surprising it would be by luck alone.

Home › Calculators › A/B Test

Planning: how big a test do you need?
Enter the visitors and conversions, then press Calculate.
From the book, Chapters 13 and 14. The book’s checkout test: 6.0% against 7.8% with 2,000 visitors each gives z = 2.25 and p = 0.025, a rise of 1.8 points (95% interval about 0.2 to 3.4). A p-value is the chance of a gap this big if the versions were equally good; it is not the chance that B is better. Decide the test length in advance and look once: checking every day inflates false alarms. A test of 2,000 per version has only a 33% chance of detecting a real rise from 6% to 7.2%.

Illustrative arithmetic only. Everything stays in your browser: no sign-up, nothing stored or sent.