Customize the Factors for the Churn Model

If necessary, you can extract and use factors from your data model by using the current Churn Prediction implementation.

Here's an example of customizing factors for the churn model:

To generate the new training and scoring datasets, update the existing set of recipes. For example, to remove Tenure and add Customer Satisfaction as a factor, replace Tenure with Customer Satisfaction by supplying the new field in your data model that stores the Customer Satisfaction value.
Update the preconfigured Story template.
Use Customer Satisfaction instead of Tenure as a factor for the machine learning model.

Update Out-of-the-Box Recipes

After an update, rerun the recipes. Ensure that the recipes are run in the order listed in this table.

Recipe Name	Description	Output
Dates and CLV Recipe	Processes the daily usage data to generate a snapshot date and the activity start date. Then, the system calculates the customer lifetime value from inception to the snapshot date for training and scoring windows by aggregating the top-up amount.	Scoring activity dates Customer Lifetime Value
Training Label Recipe	Processes the snapshot date and subscription end date to create churned labels.	Churn field names
Scoring Trend Recipe	Generates a scoring trend based on the subscription data, cases, and daily usage activity.	Scoring dataset for churn trends
Training Trend Recipe	Generates a training trend based on the subscription data, cases, and daily usage activity.	Training dataset for churn trends

Example: Scoring Trend Recipe

To get insights and predictions based on your custom features, edit and update the Minimize Churned 1 model settings.

Edit and update the Minimize Churned 1 model settings.
For more information on editing and updating model settings, see Edit Model Settings.
Redeploy the model after updating the required factors.