Test two versions (A/B)

An A/B test shares a single link between two live prototypes. Each tester sees version A or version B, drawn at random, and always gets the same one if they come back. You then compare the two versions on equivalent audiences.

Prepare a test well

• Publish both versions from Figma, as two prototypes: each keeps its own results.
• Change only one thing at a time between A and B, to know what makes the difference.
• Give equivalent screens the same name and set the same goal screen: Compare matches screens by name and compares completion rates.
• Let the test run until the “Small sample” label disappears from both sides.

Create a test

Compare tab, A/B tests sub-tab, Test two versions button (also in the Segments tab). The open prototype is version A: choose version B and the test’s name. Both prototypes must be live. A comparison named after the test is also saved in Comparisons, to find its differences in one click.

The test adds to each of the two prototypes an audience named after it, tagged A/B testing, which receives its sessions. Its link is grayed out: always share the test’s link, never a version’s. These audiences open and close with the test, always together, so that neither version gets more exposure than the other.

Follow and read the results

Each test shows the sessions received by A and by B. See the differences opens Compare on the test’s sessions only. DASHUND shows the differences for each indicator, without naming a winner: it’s up to you to judge based on what you’re testing.

Suspend or close

• Suspend: the test link shows a waiting page, then Resume restarts it. Pausing A or B, or archiving it, also suspends the test.
• Close the test (press and hold): its link and its two audiences are turned off for good. Results remain available.