Customer retention and addressing customer churn through Predictive Analytics in telecom industry

Abstract

Churn has become a major issue to almost all the telecom companies. Predicting churn draws considerably a higher importance which would help retain the existing customers of the telecom organization. Cost of acquiring a new customer is always higher than retaining an already existing customer who is about to leave the company. In order to predict the potential customers who would churn, past data must be analyzed to build the relationships between the derived variables. This is quite a challenging task as this entire exercise is based on the production dataset provided by the telecom organization. It should contain some knowledge within the multi-dimensional set of data, and this will be possible only if a proper exploratory data analysis is done. The purpose of this research is to identify the potential churners in the current customer base who would leave the company in the time to come. That entire knowledge is hidden in the production dataset and by using machine learning models, we will identify the set of customers who has a higher potential to leave the company. Out of the many machine learning models in existence, Regression Logistic model, Decision Tree model and Multi-Layer Perceptron model will be used in this study and based on the descriptive metrics of evaluation such as accuracy, recall, precision and F1 score, the best model will be identified. Once the model is identified, it will be able to intake any production dataset arranged as per the specification and to predict the potential churners in a targeted proactive manner who would leave the company. Once identified, the telecom organization will have the liberty to retain those potential churners by offering various types of offers, discounts and benefits only to those targeted customers. By making this prediction as accurate as possible, it will not only retain the existing customers, but also it will save a lot of money from untargeted and mass advertising on offers and other service-related discounts.