The company has a large customer base, however, customers leaving them was a major concern. They wanted to prevent churn by identifying reasons that stopped them from making subsequent purchases and predict customers who are at high risk of churning to take preventive actions and improve customer lifetime value. The client also wanted to improve their sales via e-commerce platforms, retail stores, and drive subscriptions while improving the demand for their products.
We integrated 15 different data sources consisting of customer information to create a strong audience profiling and segment customers based on subscription plans at a granular level. A prediction model for each of these segments was created based on the historical analysis of the churned customers. This gave the client a detailed insight into the subscription trends of their customers. We further did feature engineering using XGBoost.
The next step was to train and test each of these models separately and then combine the predictions to identify the customers with the highest probability of churning. The machine learning model training and prediction were done on Amazon Sagemaker. Sigmoid also created a dashboard on Tableau as a part of historical data analysis on the month-on-month retention rate, churn reasons, revenue generated, and more.
The ML model created by Sigmoid increased the accuracy of predicting customers likely to churn by 2.5x times. We could also ensure 70% improvement in customer retention while enabling the client to design marketing campaigns to retain customers and increase the customer lifetime value.