Split-testing functionality is available for instant and scheduled emails in the “Message Editor” step.
You can add up to 5 variants of emails, for each of which you can individually configure the subject, sender email, sender name, email content (layout, UTM tags and URL parameters). Reply tags and email address are common for all versions.
When adding a new version, the content of the previous version is copied for convenience, but you can edit it if necessary.
After all version templates have been configured, you need to set the split testing settings.
Selecting a winner
Three options are available to determine the winner:
When selecting the clicks or opens options, you should set the duration of testing in hours (minimum value - 1 hour, maximum - 1000 hours). During this time, the mailing will be analysed, and at the end of this period, based on the statistics of the selected indicator, the sending of the winning mailing will start.
The testing period is not configurable for manual determination.
Sampling for testing
After setting up the conversion rate you want to test against, you need to configure the size of the test and control groups. You can assign a test group size from 5% to 100% of the total number of recipients. The test group will be equally divided by the number of versions of the test.
The remaining subscribers who are not in the test group will automatically be sent the winning version after the test is complete. If you have chosen to determine the winner manually, you will need to analyse the statistics by version in the message details (after you have finished sending to the test group and changed the message status to “Select a winner”), select the desired test version yourself and send it to the remaining recipients.
If no targeted actions were performed in the test emails (no emails were opened/clicked) during the automatic winner determination - the winning version will be determined randomly.
In the email details you can use the split test version toggle to see all templates and content of the messages participating in the test.
All split tests are automatically tagged with an A/B tag to make them easier to find in the list of messages.