📉 Customer Churn Prediction & Retention Strategy

This project presents a full end-to-end data science solution to a real-world churn problem in the telecommunications industry. It includes business-driven data exploration, churn segmentation, model development, and actionable recommendations — packaged in a professional executive report and supported by machine learning deployment.

---

🗂️ Table of Contents

• Problem Statement
• Project Overview
• EDA Analysis & Key Insights
• Modeling Approach
• Business Impact
• Results Summary
• Tech Stack
• Files & Structure
• Author
• Status

---

🧠 Problem Statement

Customer churn directly impacts recurring revenue and profitability. The objective of this project is to predict which customers are likely to churn and to provide data-driven strategies for reducing churn and preserving lifetime value (LTV).

---

📂 Project Overview

Phase	Description
Business Understanding	Defined churn problem and goals aligned with stakeholder needs
Exploratory Data Analysis	Uncovered churn patterns by contract type, payment method, age group, etc.
Feature Engineering	Created behavioral and financial indicators (e.g., Engagement Score)
Model Building	Trained Logistic Regression, Random Forest, and XGBoost on 3 dataset versions
Evaluation	Used precision, recall, and F1-score to focus on churn prediction quality
Business Impact	Estimated model-driven churn reduction of 7.34% and retention of $317K+ in LTV

---

🔍 EDA Analysis & Key Insights

• High churn rates were observed among:
  • Month-to-month contract users
  • Solo customers not enrolled in group plans
  • Customers paying via Direct Debit or Paper Check
• Higher monthly charges and frequent customer service calls were strong predictors of churn.
• Younger customers (<30) and seniors showed slightly elevated churn risks.
• Unlimited data plan users also churned at a higher rate than expected, indicating potential dissatisfaction.
• Business Segments Identified for targeted retention campaigns based on churn drivers and LTV risk.

---

🤖 Modeling Approach

Three models were tested across multiple dataset versions (full data, feature-selected, and feature-engineered):

• Logistic Regression
• Random Forest
• XGBoost (final selected model)

Focus was placed on maximizing recall and F1-score for churners to prioritize early intervention and revenue preservation.

---

📈 Business Impact

• Estimated Churn Reduction: 7.34%
• Estimated Revenue Retention: ~$317,000
• Model Recall: 81% (catching majority of true churners)
• Precision: 85% (avoiding unnecessary interventions)

The model enables efficient retention targeting and protects customer lifetime value through early churn risk detection.

---

📈 Results Summary

• Best Model: XGBoost
• Accuracy: 91.13%
• Precision (Churn): 85%
• Recall (Churn): 81%
• F1 Score (Churn): 83%

---

🛠️ Tech Stack

• Python, Pandas, Scikit-learn, XGBoost
• Matplotlib, Seaborn, Plotly
• Streamlit (upcoming frontend deployment)
• FastAPI (upcoming API deployment)

---

---

👤 Author

Vijay Kumar
📫 LinkedIn

---

✅ Status

✅ Report complete
✅ Model trained & evaluated
🔜 API + Streamlit deployment