Customer churn prediction for telecommunication industry: A Malaysian Case Study

Area of Science:

• Data Science
• Machine Learning
• Telecommunications Industry Analysis

Background:

• Customer churn, the rate at which customers leave a business, impacts revenue and requires proactive retention strategies.
• High customer satisfaction correlates with increased sales, emphasizing the importance of analyzing customer behavior.
• Understanding churn drivers is crucial for telecommunications companies to maintain long-term customer relationships.

Purpose of the Study:

• To identify key variables influencing customer churn within Malaysia's telecommunications sector.
• To develop and propose a predictive churn framework for the telecommunications industry.
• To analyze the Net Promoter Score (NPS) dataset for churn prediction insights.

Main Methods:

• Applied data mining techniques to a Malaysian telecommunications NPS dataset (2019-2020) with 7776 records.
• Utilized machine learning algorithms including Logistic Regression, Linear Discriminant Analysis, K-Nearest Neighbours, CART, Gaussian Naïve Bayes, and Support Vector Machine.
• Developed a customer churn propensity model using 33 variables to predict churn behavior.

Main Results:

• Customer churn rates are significantly higher among customers with low Net Promoter Scores (NPS).
• Immediate helpdesk intervention can mitigate churn by addressing customer needs and ensuring satisfaction.
• Classification and Regression Trees (CART) demonstrated the highest accuracy in churn prediction at 98%.

Conclusions:

• Classification and Regression Trees (CART) offer a highly accurate method for predicting customer churn.
• Malaysia's data protection policy restricts access to personal customer information, posing a limitation.
• NPS scores can be leveraged by businesses to gauge potential customer churn and gather valuable feedback.