A/B testing
Run up to three versions of a builder against each other, and let visitor and revenue numbers pick a winner.
A/B testing lives inside the editor, above the Builder Settings / Products / Design tabs — it isn't a separate tool, it's a second thing every builder can do.
Start a test
Open a builder. Above the three tabs is the variant bar, starting with a single A variant.
Click Add variant to create a B variant (and, if you want a third, a C variant — three is the maximum).
The new variant starts as a copy of the one you branched from. Switch to it and change whatever you're testing — a different Sidebar template, box sizes, product card, wording, anything on any tab.
Save and publish. Traffic splits evenly across every variant from that point on.
Each variant keeps its own Builder Settings, Products, and Design — they're independent builders under one roof, not one builder with a few swapped fields.
Editing a live test
You can keep editing any variant while a test is running. Changes take effect the next time that variant serves a visitor.
Where you see it running
The Bundle Builders list badges a builder running a test as A/B test (N variants), with the date it started. Analytics for that builder — both on the list and on the dedicated Analytics page — break out per variant while the test runs.
End the test
Click End A/B test on the builder's row in the Bundle Builders list.
The modal shows a table with one row per variant: Visitors, Extra Rev., Total rev., Orders, AOV, CR, and Rev. per visitor.
Check the Statistical Significance line above the table.
Click Choose on the variant you want to keep.
Click End test.
Ending the test promotes your chosen variant to Variant A and archives the others — their settings aren't lost, but they stop serving traffic and the builder goes back to being a single, non-testing builder.
Reading significance
| What you see | What it means |
|---|---|
| Need more data | A winner needs every variant to have reached a minimum visitor count and the test to have a minimum number of total orders. Keep the test running. |
| N% confident X variant wins | Enough data to call it. The table also marks the leader with a Winner badge and shows each other variant's difference "vs A". |
| Not significant | Enough data by volume, but the difference between variants isn't reliable enough to act on. |
Picking a non-winning variant
If you choose a variant other than the one with the highest revenue per visitor, a warning asks you to confirm — naming the leader and what you're choosing instead. You can still go ahead; it's a check against an accidental click, not a restriction.
Next
Analytics — the dashboard and per-builder figures a test's numbers come from.