A/B test runs several versions of one bundle at the same time and splits shoppers across them, so real sales rather than guesswork tell you which offer performs best. Test a different discount, a different product mix, a different layout, or different wording.

It suits a bundle that already gets steady traffic — a variant with almost no views will not produce a result you can act on.

Open the bundle in the editor and click Add variant. You can create up to four variants of one bundle, labelled A to D. Each has its own configuration, its own preview, and its own tab in the editor.

Adding a variant does not start a test. Nothing is split until you run one.

Run A/B test splits shoppers across whichever variants are active. From then on the bundle list marks the bundle with an A/B test badge, and you can expand or collapse its variants there.

End A/B test reports revenue and purchases for each variant. This is where you read the result — the per-variant figures live on the test itself, not on the Analytics dashboard.

Choose the winner and all traffic moves to that variant; the remaining variants are archived.

While a test is running, some fields are shared across every variant and can only be edited on the primary variant, A. On the other variants they are read-only:

If you need to change one of these on a bundle that is being tested, end the test or choose a winner first.

Change one thing per variant. If a variant differs in both its discount and its products, a win tells you nothing about which change caused it.

Let it run. Give each variant a meaningful number of views before you end the test.

Decide your measure up front. End A/B test gives you revenue and purchases per variant, so settle beforehand on which of the two decides the winner.