📊 Customer Lifetime Value Prediction & Churn Analysis

ML-Powered Customer Segmentation and Retention Strategy

End-to-end machine learning pipeline for predicting customer churn, estimating lifetime value, and generating actionable retention strategies using three classification models with SHAP-based interpretability.

---

🎯 Project Overview

This project builds a complete Customer Lifetime Value (CLV) prediction system that enables businesses to:

• Identify at-risk customers before they churn
• Segment customers by predicted lifetime value (Low → Premium tiers)
• Quantify churn drivers using SHAP feature importance
• Guide retention spend by targeting high-CLV, high-churn-risk customers

The analysis processes customer behavioural data including purchase history, engagement metrics, satisfaction scores, and support interactions to build predictive models that directly inform marketing and retention strategy.

---

📋 Methodology

Feature Engineering

Feature Category	Variables	Purpose
Purchase Behaviour	total_purchases, avg_purchase_value, estimated_clv	Revenue contribution profile
Engagement Metrics	website_visits, avg_time_per_visit, engagement_score	Digital interaction intensity
Recency Signals	days_since_last_purchase	Churn proximity indicator
Service Interactions	support_tickets_last_6_months, support_burden	Friction measurement
Satisfaction	satisfaction_score	Direct sentiment gauge
Membership	membership_status (Bronze → Platinum)	Loyalty tier encoding
Referral Activity	referred_friends	Advocacy measurement

Engineered Features

# Core derived metrics
estimated_clv = total_purchases × avg_purchase_value
revenue_per_visit = estimated_clv / website_visits
engagement_score = f(visits, time_per_visit, referred_friends)
support_burden = support_tickets / total_purchases

---

🤖 Models

Three classification models trained with stratified cross-validation and class-weight balancing:

Model	Configuration	Strengths
Logistic Regression	max_iter=1000, class_weight=balanced	Interpretable coefficients, baseline
Random Forest	n_estimators=300, max_depth=10	Robust to outliers, feature importance
Gradient Boosting	n_estimators=300, max_depth=4, lr=0.05	Best predictive performance

Evaluation Metrics

• AUC-ROC — Primary metric for ranking ability
• Cross-Validation AUC — Generalisation assessment (5-fold stratified)
• Classification Report — Precision, recall, F1 per class
• Confusion Matrix — Error pattern analysis

---

📈 Key Outputs

1. Churn Distribution Analysis

Comprehensive analysis of active vs. churned customer populations, examining demographic and behavioural differences across segments.

2. CLV Segmentation

Customers segmented into four CLV tiers using quartile-based bucketing:

Tier	Description	Strategic Priority
Premium	Top 25% by CLV	Retention-critical: loyalty programmes
High	50th–75th percentile	Growth opportunity: upsell candidates
Medium	25th–50th percentile	Engagement focus: activation campaigns
Low	Bottom 25%	Cost management: targeted win-backs

3. Feature Importance (SHAP)

SHAP-based feature attribution revealing the top 20 predictors of churn, ranked by absolute contribution to model output. Critical for translating ML output into actionable business decisions.

4. Customer Risk Map

Interactive scatter visualisation plotting Estimated CLV vs. Churn Probability, colour-coded by segment and membership tier — the strategic decision matrix for retention spend allocation.

---

📁 Repository Structure

clv-analysis/
├── README.md                       # This file
├── LICENSE                         # MIT License
├── requirements.txt                # Python dependencies
├── .gitignore                      # Git ignore rules
│
├── notebooks/
│   └── clv_analysis.ipynb          # Complete analysis notebook
│
└── docs/
    └── methodology.md              # Detailed methodology notes

---

🛠️ Tech Stack

Category	Tools
Language	Python 3.12
Data Processing	Pandas, NumPy
Machine Learning	Scikit-learn (LogReg, RF, GBM)
Interpretability	SHAP
Visualisation	Plotly, Seaborn, Matplotlib
Environment	Jupyter Notebook

---

🚀 Getting Started

# Clone the repository
git clone https://github.com/svanscreates/clv-analysis.git
cd clv-analysis

# Install dependencies
pip install -r requirements.txt

# Launch the analysis
jupyter notebook notebooks/clv_analysis.ipynb

---

💡 Business Applications

This framework is directly applicable to:

• Subscription businesses — Predict churn timing, optimise retention offers
• E-commerce platforms — Segment customers for targeted marketing campaigns
• Financial services — Identify high-value client attrition risk
• Real estate — Client relationship management and lead scoring
• SaaS — Revenue forecasting and expansion revenue targeting

---

📝 Citation

@misc{rastogi2025clv,
  title={Customer Lifetime Value Prediction and Churn Analysis: 
         An ML-Powered Segmentation Framework},
  author={Rastogi, Svanik},
  year={2025},
  howpublished={GitHub repository},
  url={https://github.com/svanscreates/clv-analysis}
}

---

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

---

Svanik Rastogi · Christ University, Bangalore · BBA Strategy & Business Analytics