Overview
The Audience Recording feature allows for recording the exact state of a user's recipients when a Campaign is sent, which helps to make customer data more reliable and actionable. When using Audience Recording with Phrasee there are a few different ways to run the feature when sending to non-targeted segment groups. Before starting, it is important to note that Audience Recording functionality works by performing a random split when running a Job. If you are using Multivariate Testing, this will function in a different way, by launching multiple individual jobs. As a result, any Audience Recording that is run on a Multivariate Test will produce results that are not consistent across each iteration.
Table of Contents
Option 1: Use a Random Split to create a Holdout Group
Option 2: Create a Random/Pseudo-Generated Solution
Note: Using Phrasee React (Option 3) is the easiest of these three options to set up, however, it limits your ability to view the different variations of the MessageGears platform, as everything will appear as a single job. Instead, you would need to view the test results in Phrasee's platform which would enable you to use AR as expected. |
Option 1: Use a Random Split to create a Holdout Group
This two-step process works by recording the output of a Blueprint's randomization for the Holdouts Group. The process then takes that output group and randomly splits it against the variants groups and remaining groups.
-
Execute the Blueprint you provided as an External Campaign using Audience Recording, saving which users were participating in the campaign and which were in the Holdout group.
-
Create a second Blueprint that queries the results of the Audience Recording output in order to target just the recipients who should participate in the actual campaign.
Option 2: Create a Random/Pseudo-Generated Solution
This method involves re-executing the split for each existing variant in a way that is consistent across each execution. Using this option might allow you to skip the need for the External Campaign/AR, as the Holdout Group could be predictable within the Reporting / Analytics. This reduces the need for a two-step process.
-
Create a Test Group field in the SQL output, for example RecipientId % 10, which is random/pseudo-random, but reproducible upon each execution.
-
Execute the Blueprint as an External Campaign using Audience Recording to record the full state of the users.
-
In your Blueprint, rather than using our Split functionality, create a Target Group like Test Group > 1 to generate a similar percentage-based split, that also happens to be reproducible.
Option 3: Use Phrasee React
If you only need to view the test results in Phrasee and do not require additional information, you can do this by utilizing Phrases React.
-
Using Phrasee's React product, the entire Blueprint can be launched as a traditional Marketing Campaign (Not using our MVT approach), but set the subject line as a Phrasee placeholder to be filled in for each user.
Comments
0 comments
Please sign in to leave a comment.