A/B testing a cart

An A/B test splits your visitors between two and five versions of the same cart, so you can compare which look and content convert better. Set one up from the A/B test card in the editor's left column.

Variants differ in appearance and content only. The cart itself — its structure, its behaviour and its activation — is shared by every variant, so a test never changes who sees the drawer or how it opens.

Setting up a test

1. Open the cart and find the A/B test card.
2. Click Set up A/B test.
3. Turn on Test different looks. Two variants, A and B, are created for you.
4. Add up to three more with Variant if you want them (A through E).
5. Choose a Traffic split: Even split, or Custom with your own percentages. Custom weights must total 100 — the modal tells you when you are short or over.
6. Click Save.

The card header then reads A/B test · On and shows a pill for each variant.

Editing a variant

Variant content is not edited in the modal. Instead:

1. Click a variant's pill in the A/B test card header. A is the original.
2. The editor's panels switch to that variant. A banner confirms it: "Editing variant B. Design, text, drawer layout and modules apply to this variant only. The cart name, structure, behavior and activation are shared; switch back to A to change them."
3. Make your changes, then click Save test.

Variant changes save with Save test, not the ordinary save bar. That is a deliberate separation: the base cart and the variants are saved by different buttons so one cannot quietly overwrite the other.

What a variant can and cannot change

Can change, per variant:

• The design — template, colors, fonts, radius, spacing.
• Drawer width, panel style and launcher settings.
• Every piece of shopper-facing text, in every language.
• Which modules are switched on, and their order and settings.

Cannot change — these are shared and stay editable only on variant A:

• The cart's name.
• Which modules exist at all (adding and removing).
• The Behavior and Advanced panels.
• Activation — markets and schedule.

While a variant pill is selected, those sections are locked and the + and delete controls disappear from the module list.

Reading results

Once orders start converting, the Experiment results table appears in the same card, with a row per variant showing Orders, Revenue and AOV. Before that it says "No attributed orders yet. Results appear as variants convert."

One number to read carefully: the note under the table says "Attributed line revenue per variant, not full order totals." Compare variants against each other; do not compare these figures against your Shopify order totals.

The home page's per-cart columns cover the cart as a whole, not per variant. See Reading the numbers.

Running a test well

• Change one thing at a time. A variant that changes the template, the copy and the module order tells you that something worked, not what.
• Let it run. Cart drawer conversion differences are small; a day of traffic on a small store is noise.
• Do not restart the split mid-test. Changing the traffic weights part-way makes the comparison unreliable.
• Finish by making the winner the base. Switch to variant A, apply the winning changes there, then turn Test different looks off.

Related

• The editor
• Activating and scheduling a cart
• Reading the numbers