Churn Predictions for Communications with Einstein Discovery

Use real-time predictions to provide intelligent visual insights that your agents can use to minimize subscriber churn. You can calculate predicted churn scores by using only the customer subscription data or by using Feedback Management and Sentiment Insights data.

Your agents can use the predicted churn scores to identify subscribers who are likely to churn in the next three months and understand the factors that impact churn, such as changes in the frequency of customer payments or data usage trends.

Insights are influenced by the factors recommended by the TM Forum, the Feedback Management factors, and the Sentiment Insights factors.

For example, if a customer reports a lost SIM card, their score in the Likelihood to Churn card becomes higher. Your agent can get a recommendation to ship a new SIM card as the next best action to retain the customer. Your agents can also use the churn score for upselling. If a customer has a low churn score and the data usage is tending up, the agent sees a recommendation to offer the customer a product that best meets the usage history and optimizes the spending.

Recommendations appear in the Next Best Action section on the agent console.

A business analyst can get valuable insights into the factors that influence churn and how they correlate in the Minimize Churned 1 or the Subscriber Churn Indicators model.

If only the customer subscription data is used, the analyst can refer to the Minimize Churned 1 model. If Feedback Management and Sentiment Insights data is used with the customer subscription data, the analyst can refer to the Minimize Subscriber Churn model.

The analyst can use different thresholds for each factor and evaluate how the factors contribute to the churn score. Analysts can also customize the predictive model by adding new factors or removing existing factors via Discovery Model Manager. By using the model metrics, analysts can easily assess the accuracy of the model and understand the related quality measures that they can work with to improve churn prediction.

To set up real-time churn predictions, your Salesforce administrator can use the out-of-the-box (OOTB) templates:

• Subscriber Churn Predictions with Customer Subscription Data
  Predict your subscribers’ churn with customer subscription data. You can consume the external subscription data to generate training or scoring datasets and then use these datasets to predict subscriber churn.
• Subscriber Churn Predictions with Customer Subscription, Feedback Management, and Sentiment Insights Data
  Predict your subscribers’ churn using customer subscription data, Feedback Management data, and Sentiment Insights data. The Feedback Management and Sentiment Insights data is based on contacts associated with your subscribers’ accounts.
• Churn Score After Customer Interaction Ends
  Instead of getting the customer churn score updated based on the frequency defined by your Salesforce admin, get the latest churn score after each customer interaction by using AI Accelerator.