A/B testing

How a test works

A test sends part of your traffic to your original page and part to a variation. Each visitor sees one version, and the test counts who clicked through on the page's call to action. When enough visitors have been through, the statistics tell you which version performed better.

Availability depends on your plan. The app shows your test usage, and your plan sets how many tests can run at once. Starter accounts can view tests but not create them.

Create a test

After the setup, design the variation in the page editor. The goal of every test is click-through on the page's call to action.

Run, pause, stop

Start the test when the variation is ready. A test needs at least 2 active variants and a published page before it can start.

While it runs you stay in control. Pause a test to hold both versions, resume it when you are ready, or stop it when you have your answer.

Every test card shows participants, days running, significance, and confidence, so you can check progress at a glance. The dashboard's Active A/B tests widget links straight into running tests.

Read the results

Results come as two views of the same data. The Summary view is a Bayesian analysis, and the Z-test view is a frequentist analysis. Both show confidence intervals, the p-value, the improvement, and a plain-language recommendation: collect more visitors, run longer, or call it.

Each variant card shows its traffic share, visitors, conversions, rate, and improvement. Badges mark the control and, once the numbers support it, the winner.

Declare and apply the winner

When you have an answer, stop the test and declare the winner in one step. You can also pick the winner yourself from any variant.

On a finished test, Apply Winner to Page puts the winning content on your live page and archives the test. Tests you no longer need can be deleted.