Skip to main content
An A/B test splits traffic between two variants of a version so you can find what performs best, then rolls out the winner. This is the short version; for the detail, start with how A/B tests work.
A/B testing applies to brand referral campaigns. A product referral campaign uses a holdout instead, which answers a different question.
1

Start the Test

To test against your live version, click Set up A/B test on it (or the recommendation in the insights panel), name the new version and click Create test. To test two fresh ideas on a draft, click + A/B test on the node you want to test.
2

Choose How It Runs

In the wizard, pick the test metric (conversion rate or share rate) and the test speed, then set what happens when it concludes. See set up an A/B test.
3

Differentiate the Variants

Edit Variant B so it differs from the control in one way. See manage an A/B test.
4

Launch and Monitor

Launch the version like any other, then follow the result on monitor an A/B test.
For ideas on what to test, see A/B testing best practices.
Last modified on September 25, 2026