Knowledge base · Data & measurement
A/B test
In an A/B test two variants run at the same time for comparable groups, to measure which performs better.
The idea is simple: change one thing, run it long enough and see whether the difference beats chance. In practice it goes wrong at that last point, because most tests are stopped too early.
Test things that matter. A different headline, a different opening frame or a different offer yields more than a button two shades off.
Keep what you learn as well. A collection of outcomes across a year is worth more than any single test.
How we do it
We test with a fixed duration and record every outcome, including the tests that changed nothing.
Related service
Data & Visibility
View the service →Curious what this means for your brand?
We look at your numbers with you and show where the gains are, in plain language.