Testing experiences

Context

You want to maximize conversions but do not know which version of a page or component performs best. A/B/n testing lets decisions rest on visitor behavior rather than internal speculation. It is particularly useful after launch, during optimization sprints, and ahead of rollouts across many sites or geographies.

Execution

Testing or personalization

The two are confused often enough to separate clearly. Personalization delivers different content to specific user groups based on factors such as location, device, or how they arrived. A/B/n testing compares versions to see which performs best, without needing to identify the visitor. Use personalization when you know who you are speaking to. Use testing when you want to compare ideas across all visitors.

A page runs a test or a personalization, not both at once. You cannot create an A/B/n test on a page that has personalization configured. If you are unsure which experience will perform better, start with a test, then turn the winning version into a personalized variant.

Plan the test

Two stages shape a testing strategy.

Business objectives alignment. Select the components and content to test, write hypotheses, and agree success metrics in advance. Prioritize by potential impact so the high-value work comes first. This keeps the team from spending time on experiments that cannot inform a decision.

The test-and-learn cycle. Run variations and measure their effect against performance goals. Over time you build a picture of what resonates, and each result rests on observed behavior rather than guesswork.

When you brainstorm tests:

• Identify what you want to achieve: higher conversion, lower bounce, or better engagement.
• Focus on the areas with the most influence on that goal, such as key landing pages, checkout steps, or calls to action.
• Write a clear, testable hypothesis based on your goals and audience insight. For example, "Changing the color of the call to action button will increase click-through rate."

Steps to run a test

1. Define the business goal and pick the page or component with the most influence on it.
2. Write the hypothesis.
3. Create the test from the component in Page builder. A test supports up to six variants, and a page can run up to six tests at once, with extra tests queued.
4. Set the goal. Its metrics are used to calculate the number of visits needed to declare a winner or call the test inconclusive.
5. Assign traffic to variants, set automated actions for the outcome if you want them, and edit advanced options such as the percentage of visitor traffic in the test.
6. Preview and start the test, then publish the page.
7. Read the result against the original hypothesis. Apply the winner, set up personalization if the learning supports it, and plan the next test.

Keep a record of each test with its hypothesis, variants, result, and what you learned. A shared log or board is enough. It lets you refine the testing strategy over time.

Review Get started with A/B/n testing for the in-product steps, and Best practices for A/B/n testing for further guidelines.

Insights

When a test declares a winner, move the winning version into the component or personalization rules. Leaving a finished test running at 100/0 traffic is a workaround nobody governs.

Related recipes

Related documentation