Einstein Content Testing

Content Testing consumes Einstein Content Selection Impressions, which is a type of super message. Images are selected in real time, so super messages are consumed only when an email open event is received. For more information, see Multipliers for Marketing Cloud Engagement and Marketing Cloud Engagement Data Usage Types.

With automated testing, Einstein runs a multi-armed bandit test that balances optimization of winning content with open-time experimentation for the duration of the test. The winning content can be dynamic, and is defined as content with the highest click-to-open rate at the time an open event is received. Winning content is selected more often than the remaining experimental content. This experiment type is intended for continual self-regulation, although you can opt to select a winner so that all subsequent opens receive the same content. The automated test is best for longer-duration sends, such as a long-running journey or a send that’s managed by Einstein Send Time Optimization, or when you don’t want to self-manage the experiment. See the model card for more details.

The manual approach equally distributes content until a winner is determined. You can either identify a winner manually after reviewing the results or have the winner determined after a certain number of clicks is reached. If you want the winner to be determined, the content with the highest click-to-open rate is selected when the click threshold is reached. Manual tests work well when you’re experimenting with a test audience and the winner determines which content to include in subsequent large batch sends. Manual testing is also useful when experimenting for a long-running journey and if you want more control when a winner is determined.

• Create a Content Test
  Use Einstein Content Testing to compare up to 32 images to determine what works best. You can create automated tests or manual A/B/n tests. You can also access previous tests on the Content Test page. Draft, completed, and archived tests aren’t available.
• Delete a Content Test Draft
  When a test draft is no longer needed, you can delete it.
• Stop a Content Test
  You can stop a manual or automated content test at any time. For example, you might want to stop a manual test after you’ve received engagement data on a few thousand opens. Generally, you don’t need to stop an automated test unless you want to override the continual testing portion and ensure that all images are optimized to a single asset.
• Archive a Content Test
  Clean up your list of content tests by archiving tests that are no longer relevant.
• View Content Test Performance
  To increase your subscriber interactions, use the performance results data to review the click rate and open rate and guide your messaging.