Analytics and testing
How to create your first A/B test
Publish the page, open A/B tests, and create a named test for that page. Edit Variant B, review it against Original A, then choose Go Live and complete the safety check. Eligible visitors receive a stable approximately 50/50 assignment between the two versions.
Choose one question to test
Start with a specific hypothesis and a measurable action. For example: “A clearer explanation above the quote form will encourage more people to submit it.” Keep prices and other important behavior consistent so the comparison is understandable. Publish the selected page and verify its live version before creating the test. A/B tests operate on unlisted shared pages; they do not send traffic to a private draft.
Create a paused variant
- Open your project’s A/B tests tool.
- Choose Create A/B Test.
- Enter a descriptive Test name and select Page to test.
- Choose Create test, then Edit variant.
The new test starts paused. Creating it does not immediately change what visitors see. If the tool reports that tests are unavailable for this project, resolve that availability message before continuing.
Edit Variant B
Select Variant B and describe the change in the variant prompt bar, then choose Update variant B. The prompt applies to that test’s variant, not the main project. Wait for the saved result and test its behavior. Saved variant edits pause the test until you choose Go Live again. You can leave during a build and return to check its saved state instead of repeatedly resubmitting.
Review both versions and go live
Switch between Original A and Variant B. Test the primary action, navigation, and a mobile layout in each. Choose Go Live from the test’s status menu and wait for its saved-page and image safety check. Confirm the status becomes Live. Only one test can be live for the same page; activating another test pauses the earlier one for that page.
Share the normal link and observe
Share the page’s ordinary live link. Assignment stays consistent for a browser while its assignment cookie remains available, so refreshing is not a way to alternate versions. Small samples need not be evenly divided. Open Analysis to inspect recorded activity and Pause to return visitors to the original page. Higher observed activity alone does not establish a statistically reliable winner; review the metric definitions before deciding what to keep.