You can randomly assign users to a test variant to evaluate the effectiveness of different treatments while on a journey.
- You cannot add an A/B test after an End or Jump event.
- If a parent segment has not run and a journey with an A/B test has launched, the number of profiles within the segment is unavailable. Rerun the parent segment to calculate the number of profiles in the segment.
- If stale profiles are not removed from an A/B variant, rerun the parent segment and journey workflow to refresh the profile count.
- Jumped profiles are not removed from the A/B test variant profiles.
- You can add a maximum of four layers of A/B testing (add an A/B test to an A/B test).
- You can add unlimited A/B tests within the journey.
- When a parent segment updates, the system-generated ID for all profiles is not persistent. Therefore, you might use the optional unique ID field in your A/B tests to ensure a persistent ID is retained in all variants. You use a unique ID field to specify a stable key (attribute) to be used as a seed for random assignment.
You can add an A/B test to any decision point, and the number of A/B tests within a journey is unlimited.
- Open TD Console
- Open Audience Studio.
- Open an existing journey or create a new one.
- Select and drag the A/B Test event into your journey stage.
- Select the A/B icon to create your A/B test.
| Field | Action | Additional Information |
|---|---|---|
| Name | Create a unique name for your A/B test | A/B Test names cannot be the same within a journey stage; however, you can duplicate A/B Test names if the test is not within the same journey stage. |
| Number of variants | Identify how many individual tests you want to run. | The default is 2, but you can choose a single variant with a control group. You can add eight test groups (plus one control group). By default, Treasure Data names A/B test variants A, B, C, D, E, F, G, and H. However, you can overwrite the default test names. |
| Customized Split | Move the toggle to Yes if you want a customized split. | By default, Treasure Data splits the percentages equally among variants; however, if you move the toggle to Yes, you can specify specific percentages for each test variant. You might see a slight difference between the percentage of the splits you set and the actual split displayed in the A/B test. This difference can occur when a variant receives less than 600 profiles. |
| Total % to test | Use the arrows to increase or decrease the percentage or enter a number. | The percentage you enter is split evenly between the number of variants. In the screen capture, Variant A is 37.5 %, and Variant B is 37.5%, which equals the 75% entered for the Total % to test. 25% is added to the control, making the count 100%. |
| (Optional) | ||
| Unique ID Field | Specify a stable key (attribute) to be used as a seed for random assignment. | The unique ID fields are the attributes associated with the journey. These unique ID fields are the same as the attributes you use to create stage rules. Select an attribute representing a unique, persistent ID, such as a customer id or email. |
- (Optional) You can add additional events to this stage. You cannot add an A/B test after an End or Jump event.
- Select Save A/B Test.
After running an A/B test, you might want to get a final list of customer IDs for each variant. You can add the drag-on rule of A/B Test Variants - Journey to get a list of profiles and then export the content for further analysis.
See also Journey FAQs.