Customer Churn Analysis for a Software-as-a-service Company

SaaS companies generate revenues by charging recurring subscription fees for using their software services. The fast growth of SaaS companies is usually accompanied with huge upfront costs in marketing expenses targeted at their potential customers. Customer retention is a critical issue for SaaS companies because it takes twelve months on average to break-even with the expenses for a single customer. This study describes a methodology for helping SaaS companies manage their customer relationships. We investigated the time-dependent software feature usage data, for example, login numbers and comment numbers, to predict whether a customer would churn within the next three months. Our study compared model performance across four classification algorithms. The XGBoost model yielded the best results for identifying the most important software usage features and for classifying customers as either churn type or non-risky type. Our model achieved a 10-fold cross-validated mean AUC score of 0.7941. Companies can choose to move along the ROC curve to accommodate to their marketing capability. The feature importance output from the XGBoost model can facilitate SaaS companies in identifying the most significant software features to launch more effective marketing campaigns when facing prospective customers.