A technique that involves trying several alternative designs for a feature and seeing which works best. It is often used for designing user interfaces for web applications and is a great way to settle disputes about what design is best. For example, if you want customers to click on a button and are not sure whether the button should have a picture of a bobblehead or an actual baseball player on it, you can try both. Set it up so that some users get the bobblehead and some the baseball player, and measure which gets more clicks. In the digital world where small changes can be made quickly and cheaply, and where A/B testing can easily be automated, this has become a very effective way to make design decisions.
— War and Peace and IT: Glossary, Mark Schwartz