A/B tests

1. A/B tests

   Test homepage banners against each other, or show a new one to a growing share of visitors. Each visitor keeps the same version. No personal data is used, and search engines see the first version. This tour walks through the page.

2. New test

   Name the test, choose its type and the share of visitors, pick the banners, and create it. A new test starts as a draft until you start it.

3. Type

   An A/B test shows banners A or B to each visitor. A gradual rollout shows the new banners to the share of visitors you choose, and you can raise that share later.

4. A test

   Each test with its type and status: draft, running or ended. Start a draft to run it. While it runs, end it by choosing the version to keep; the banners of the other version are switched off.

5. Results

   Visits, paid orders and conversion for each version. The line below says whether the difference is unlikely to be chance or there is not enough evidence yet.