TypeUI logoTypeUI

A/B Testing

Compare two or more versions of the same page. For example, find out whether more visitors click your pricing button when it appears above or below the feature list.

Create your test

  1. Open A/B Testing in your dashboard and choose Create test.
  2. Select a website, enter the page path, and name your versions. Version A is your baseline.
  3. Choose a goal, such as a checkout-button click. Reuse an existing custom event or create a dedicated A/B goal.
  4. Copy SETUP.md into your AI coding tool or give it to your developer.
  5. Publish the tracking changes, start the test, and check that each version is receiving exposures.

Your website decides which version each visitor sees. TypeUI measures the results; it does not change your page or split your traffic. All versions can use the same URL.

Read your results

An exposed session saw a version of your page. A converted session completed the selected goal afterward, in the same session and within 24 hours. Each session converts only once per goal; Goal actions also includes repeated actions.

Dates use UTC and the session's first exposure. A goal completed after midnight still belongs to that exposure date. Recent totals may change as delayed events arrive. Sessions that saw multiple versions are shown separately and excluded from conversion rates.

Use the datepicker and comparison range to explore different periods. You can change the baseline, filter by device, or choose separate date windows for versions tested at different times. When comparing different weeks, remember that traffic sources, campaigns, and seasonality can also affect the result. TypeUI does not automatically declare a winner or claim statistical significance.

Pause or finish

Pause stops new exposures from entering the test. Goals from already exposed sessions can still arrive within the measurement window. Resume continues the same test. Complete or archive a test to free its page for another test. Create a new test when changing versions or measurement rules.

Paid workspaces can save up to 100 tests per website and 500 per workspace, with up to 5 active or paused tests per website and 50 per workspace. A test has 2–5 versions and up to 3 goals. Trials allow one test with two versions and one goal.

Tracking follows your existing Insights consent settings. Do not include personal information in test names, version names, or goals. Exposures do not add pageviews. Dedicated goal actions share your custom-event allowance; existing custom events are not charged again just because a test uses them.

To ask your AI about a test, enable A/B Testing in the website's MCP settings. MCP returns aggregate results only and uses your normal MCP request allowance. A/B results are not included in public shared dashboards.

HeatmapsCodex