Churn prediction / risk alerts

Churn prediction watches signals such as falling logins, fewer features used, unanswered emails or unpaid invoices, and raises an alert when a customer looks at risk of leaving. For example, a software company might get a warning that a client has not opened the product for three weeks, so an account manager can call before the renewal date. ChurnZero, Custify, Gainsight, Planhat, Totango and Vitally are tools in the library that list this feature.

Keeping an existing customer usually costs less than finding a new one, yet cancellations are often noticed only when the customer has already decided. Early warnings give you time to fix a problem, offer help or adjust the plan. You need this when you have a recurring revenue model, enough customers that no one can watch each account by hand, and data on how customers use the product. You can skip it with a handful of large accounts, because the account manager already knows who is unhappy.

• Which signals feed the prediction, and whether you can choose or weight them.
• Whether the alerts explain why a customer is flagged, or only show a score.
• How the tool performs on your own past cancellations when you test it on historical data.
• Who receives an alert, and whether it creates a task in the tools your team already uses.