Predictive Churn

With Predictive Churn, you can define what churn means for your business and identify the users you want to retain. When you create a prediction, Braze trains a machine learning model using gradient boosted decision trees to recognize at-risk users by analyzing patterns from past behavior—both from users who churned and those who didn’t.

About Predictive Churn

After the prediction model is built, users in the prediction audience will be assigned a churn risk score between 0 and 100 denoting how likely they are to churn according to your definition. The higher the score, the more likely a user is to churn.

Updating the risk scores of the prediction audience can be done at a frequency you choose. This way, you can reach out to users who are at risk of churning before they actually do and prevent it from happening in the first place. Using up to three active predictions, you can leverage Predictive Churn to tailor individual models to help prevent churn within specific segments of your users that you deem to be the most valuable.

Accessing Predictive Churn

The Predictions page is located in the Analytics section. For full access, contact your account manager.

Prior to purchasing this feature, it is available in preview mode. This will allow you to see a demo churn prediction with synthetic data and create one churn prediction model based on your user data at a time. This preview will not allow you to target users for messaging according to churn risk and will not regularly update after creation.

With the preview, you can also edit and rebuild your one prediction or archive it and create others to test the expected prediction quality of different definitions.