The first part of setting up a Phrasee's Acoustic integration is to define 11 database segments.
Ten segments are for the split test, nine of which will receive a Phrasee variant and one which will receive your human control variant. The eleventh and final query is to select the remaining bulk of the members of the database who will receive the winning variant with the best performing language from the test.
You need to provide Phrasee with the IDs of these queries prior to your first send so they can be recorded in your profile settings. You only need to do this once, as long as you do not delete or recreate the queries for any reason.
The following instructions detail how to create the queries and update your database using the built in Acoustic segmentation tool and field type to populate the queries.
You will need to run the Acoustic segmentation tool prior to the deploying a Phrasee experiment to ensure the audiences selected for the queries are newly randomized for each send. We suggest scheduling this soon after the conclusion of your previous experiment to provide enough time for Acoustic to complete the operation before your next experiment.
Setting up the database segmentation
Log in to Acoustic and open your main database.
Select your Master Database.
Click Add Field.
Give the field a valid meaningful name and select the Segmenting Field Type.
Run the segmentation tool to randomly select your split test audiences.
Complete Step 1 of the Segment contact source by filling in the fields as indicated:
Set the Base Segmenting Value to 1
Select the field your created above for the segmentation field
Give the data job a meaningful name
Complete Step 2 by:
Selecting conduct sampling by selecting the following option and providing values
Selecting Number of segments and entering 10 (or the number of splits advised by your Customer Success Manager)
Selecting Percent of contacts per segment and entering the number confirmed by your Phrasee Customer Success Manager (usually in the neighborhood of 3)
Unchecking Automatically generate each segment as a
Clicking Next to move to Step 3
Click Submit to run the data segmentation job.
Acoustic should then confirm the job has been submitted.
Creating queries
To use the segmentation you created in your Phrasee experiments, you now need to create queries to select audience members based on your segmentation field.
Go to Queries in Acoustic.
Select Create Query.
Complete the Save Query As form by:
Selecting the Master Database where you created the split number field
Giving the Query a meaningful name and including the appropriate split number at the end to help you keep track
Selecting the location as Shared and the type as Query
In the query creation form select Profile > Data field under Add Criteria.
In the Data Field selection, find the Split Number field you created earlier.
Fill in the Operator Field as is equal to and then enter the value that corresponds to the Query being created (e.g. 1, 2, 3, 4...10). Then select click Done.
You should have the following setting for the query which you can then Save & Close.
Complete the query creation for all of your remaining splits by repeating the steps above. When you're finished, you should have a query for each split number one through 10 (or for whichever number of splits your Phrasee Customer Success Manager suggested).
Finally, create the final send query.
For this final query, give it a different name like Phrasee_Final and use the Operator is blank.
The final send query settings should look like the below before clicking Save & Close.
Getting the Query IDs
Once all the queries have been created, you need to record the query IDs to send to Phrasee to include in your integration setup. You can do this by hovering over the query name and recording the query ID that pops up in the tooltip. If you click on the query, you can also find the ID in the address bar of your browser.
Record the query ID for each split test (e.g. Phrasee_Split_1 = 555621) and be sure to include the final send query, too.
Before each send, you will need to follow the procedure for Segmenting your database up until the data job has been submitted to properly randomize your audiences.