Objective and Scope:

The client, an e-commerce retailer, sought to develop a predictive analytics framework to identify customers at high risk of churn. The engagement aimed to support strategic decision-making through:

• Identify at-risk customer segments using machine learning models to predict churn likelihood across customer cohorts, with focus on behavioral patterns and engagement metrics
• Quantify churn drivers and patterns across customer acquisition channels, subscription plans, feature adoption, and support interactions to enable targeted intervention strategies

Approach:

Benori combined automated data extraction from CRM systems and transaction logs with advanced machine learning to identify and rank churn predictors. Custom Python scripts were developed to aggregate customer behavioral signals (usage frequency, support interactions, payment delays) and engineer predictive features. Leveraging gradient boosting and ensemble modeling techniques, Benori trained multiple models to classify churn probability, cross-validated results with business rules, and created a structured dataset ready for real-time churn scoring and dashboard deployment.

Impact:

The study enabled the client to:

• Achieve early churn detection by assigning propensity scores to every active customer, enabling the commercial team to prioritize at-risk accounts for targeted retention campaigns
• Enable data-driven retention strategies through identification of key behavioral signals and root cause analysis, allowing to proactively address pain points