This page explains the mechanics behind how users are distributed to test groups in A/B testing.
Note: Click on the image for a larger view.
The traffic distribution into different A/B test group is as follows:
1. Check if the A/B test exists.
- If the A/B test does not exist, the default group, Test Group A, is determined.
2. Make sure the user is manually assigned.
- A/B test setup - Refers to the user whose user identifier is registered in the test device registration menu.
- In the case of a manually assigned user, the test group is determined by the set test group.
3. Check if the A/B test status is in draft mode or paused.
- If the A/B test status is draft mode or paused, the default group, test group A, is determined.
4. Ensure that A/B testing is complete.
- If the A/B test is completed, the winning test group selected at the end of the test will be determined for all oncoming traffic.
5. Make sure it is allocated to traffic via bucketing.
- For unassigned users, the default group, Test Group A, is determined.
- For the meaning of bucketing, please refer to the Principles behind Traffic Distribution.
6. Make sure the assigned test group is not dropped from the experiment.
- If the user is assigned to an excluded or dropped test group, the user will be determined as the default group, test group A.
- If not excluded, the user will be allocated to the corresponding test group.
Updated about 1 year ago