Pages & Funnels

A/B Testing for Store Pages

Updated on 11/3/2026

Compare two versions of a page, split the traffic and promote the one that converts better.

What it's for

Horsee A/B testing shows the original version of a page to some visitors and a variant to the rest, then compares conversions and revenue for both. The feature is in beta and is not open to every store yet: if the Create A/B test option doesn't appear in your page menu, it isn't available to you yet.

Where to find it

The list of tests lives under Marketing → A/B Testing. The item only shows up in the sidebar after you finish your store's first steps. Without the feature enabled, the screen says "No A/B tests yet" and the page menu has no option to create a test.

Step by step (stores with the feature enabled)

  1. Publish the page you want to test. A draft page can't be the original in a test.
  2. In Sell → Pages, open the page's three-dot menu and choose Create A/B test, or use Test a variation inside the editor.
  3. Set the name, the duration and the traffic split (for example, 50%/50%). Horsee creates the Variant, a hidden copy of the Original.
  4. Edit the variant and click Start test. The store is republished when needed, and the split starts once publishing finishes (1 to 3 minutes).
  5. Follow users, conversions, revenue, conversion rate and lift (% gain) for each version on the test dashboard. Use Pause test if you need to stop it.
  6. When you have a result, use Promote B to make the variant the official page, or Discard test to keep the original.

Checkout A/B testing

Also in beta, a checkout test changes only the shell of the raw checkout (header, copy, badges, colors); the payment form always stays the same. There, conversion means purchases ÷ visitors who reached the checkout in each version, not every store visitor. Starting or resuming the test republishes the store automatically.

Common mistakes

  • The original isn't published: publish the page before creating the test.
  • A test already exists for that page: each page can have one open test at a time. Finish or discard the current one.
  • Calling it too early: a verdict needs at least 7 days and 30 conversions in each version. Promoting before that asks for a second confirmation, because the result isn't conclusive yet.
  • Editing the variant while the test runs: use Update variant; the store is republished and the count starts over.
  • Forgetting that promotion is final: the winner becomes the official page.

Until the feature reaches your store

You can compare two pages by hand: publish both versions at different addresses, send each campaign to one of them with its own UTM parameters, and compare sales in UTM analytics. See the Help Center for the UTM analytics guide.

Was this article helpful?

Your feedback helps us improve our guides