Editing and Canceling MVT Tests

Overview 

This page describes what editing is allowed once a test has been made active as well as canceling an active test.

Editing Active Tests

Editing a test once it is created is not supported. Changes to the meaning of the treatments, or the percentages of traffic, can interfere with a meaningful test with results that can be interpreted. If you should make a mistake in setting up a test with a future start date, cancel the incorrect test, and start again. If you should need to change a test that is already running, cancel the current test and create a new one; consult with the Analytics team to see if they can assist with a manual analysis to merge the results of the two tests.

The only change that can be made to an active test if you are using legacy, cookie-based treatment assignment (which is not recommended) is the traffic percentage assigned to a treatment. Changing the treatments once a test has been activated can cause customer behavior to change, which will produce inconclusive test results. This is why it is important to define the test clearly before activating it for your customers. Modifying the traffic percentage will not change the customers who have already been assigned to a treatment, even when they come back on future days. It will only impact the assignment of treatments to new customers that arrive after the change is made.

Canceling an MVT

If the test has proven your hypothesis, and you have decided that a particular treatment (or the control) produces the best results, you do not need to continue to run the MVT. You can cancel and make the appropriate changes to your site, so that all customer traffic sees the winning behavior.

Note: Once a test has been canceled, it cannot be restarted. You can, however, create a new test with the same treatment definitions.