An A/B test splits real storefront visitors between two to four versions of a live page and counts which one performs best. The control is your current page. Each other version is a copy you edit.
Before You Start
- Experiments need the full analytics tier, which is on the Core, Unlimited and Agency plans. On Free, the screen says Split tests need the full analytics tier.
- The page must be published. Otherwise you see Publish this page before running a test.
- Only one test can run on a page at a time.
Create a Test
- Go to Pages, open More actions on the page's row and click Experiments.
- Under New test, enter a Test name.
- Pick the metric to compare: Orders, Add to carts or Checkouts started.
- Choose how many versions: 2 versions, 3 versions or 4 versions.
- Give each version a label and a Share of visitors (%), for example 50 and 50.
- Click Create test.
Visitors are split between versions by cookie, in the shares you set.
Edit the Versions
Click Edit beside a version in the test's table. It opens that version on the canvas, where you edit it like any page. The control stays your current page.
Start and Stop
- Click Start to begin splitting traffic.
- Watch the results table: Version, Weight, Views, Visitors and Result.
- Click Stop to end the test without choosing a winner.
Declare a Winner
Click Declare winner on the best version. If the difference is not yet statistically significant, the app asks This result is not significant yet. Declare the winner anyway? The winning version gets the Winner badge.
A test measures orders and checkouts credited to the page, using the same 30-minute rule as page analytics.