Modeling Client Churn for Small Business-to-Business Firms

Abstract

With the widespread adoption of customer relationship management (CRM) systems such as Salesforce, HubSpot and Oracle, businesses are becoming increasingly aware of their customer churn rates. Churn rates describe how many customers stop using a product or service within a certain time period and provide a sense of the businesses’ long-term viability. Business-to-Business (B2B) firms place high value on the ability to predict individual customer churn, as it presents an opportunity to retain key clients in an inherently limited customer portfolio. These predictions must be both actionable and timely if a manager hopes to retain their client, since a client’s churn decision occurs months before the observed churn event. This study explores the HubSpot data of a B2B organization. The objective is to determine the client characteristics that predict sustained product usage and to analyze the indicators of potential churn. Our approach was to model the predictive features of client churn, which would allow managers to directly map churn probability to business strategies. Our final models flagged a handful of management-adjustable features that were significant for predicting customer churn and survival times.