Customer Retention Analysis for a Telecommunication Company

Kaplan-Meier Model Implementation for Customer Churn Prediction

Project Overview

This project applies Kaplan-Meier survival analysis to estimate customer retention rates and predict churn for a telecommunication company. By analyzing historical data, we aim to quantify retention trends and calculate Customer Lifetime Value (CLV) using three different modeling approaches:

1. Simple Retention Model – A basic estimation of retention rates and churn times.
2. Generalized Retention Model – Advanced survival analysis using the lifelines library.
3. Generalized Retention Model with Grouping – Segmentation by Internet Service type to compare customer loyalty across groups.

Key Assumptions

• Payments are made in arrears.
• The average net income per customer is $25 per period.
• A 2% discount rate is applied for CLV calculations.

Methodology

1. Simple Retention Model

• Retention Rate: Calculated as the ratio of active customers over time.
• Time to Churn: Average and median time before customer departure.
• Expected CLV: Derived using retention rate, net income, and discount rate.

2. Generalized Retention Model

• Utilizes the Kaplan-Meier survival function to estimate churn probability.
• Computes average and median churn times using survival analysis.
• Predicts CLV dynamically by accounting for a variable retention rate.

3. Generalized Retention Model with Grouping

• Applies the Kaplan-Meier model to different customer groups (e.g., Fiber Optic vs. DSL users).
• Identifies differences in customer loyalty based on service type.

Data Preparation

The dataset (telco_customers.csv) includes essential features for retention analysis:

• tenure: Customer’s service duration (in months).
• cancel: Churn indicator (1 = churned, 0 = active).

Data is sorted by tenure to align with the sequential nature of survival analysis.

Key Findings

• High Retention Rate (0.977): Most customers stay with the company.
• Median Time to Churn (44 months): Customers typically leave after nearly 4 years.
• Service-Based Loyalty: Fiber Optic users exhibit higher retention rates.

Conclusion

The telecommunication company demonstrates strong customer retention, particularly among fiber optic users. Insights from this analysis enable strategic marketing decisions and better budgeting for customer acquisition.

How to Run the Project

Prerequisites

• Python 3.6+
• Required libraries:

  pip install pandas lifelines matplotlib

• Place telco_customers.csv in the project directory.

Execution

Run the Jupyter Notebook or Python script to generate retention insights and visualize Kaplan-Meier survival curves.