📊 Customer Churn Prediction | Product Data Science Case Study

Product Data Science | Retention Analytics | Machine Learning | Experimentation Strategy | Revenue Optimization

---

🚀 Executive Summary

Customer churn directly impacts revenue, growth, and long-term product sustainability. This project approaches churn not just as a classification problem — but as a product retention and revenue optimization challenge.

---

The goal is to:

• Identify high-risk churn segments
• Understand behavioral and contractual drivers
• Quantify business impact
• Recommend product-level interventions
• Enable experimentation-driven retention strategies

This mirrors how churn is handled in SaaS, FinTech, HealthTech, and Telecom product organizations.

---

🎯 Product Problem Statement

Subscription-based products experience revenue leakage when customers cancel early.

• Key Product Questions:
• Which customers are most likely to churn?
• What signals predict churn behavior?
• When is the highest-risk churn window?
• What retention experiment should be prioritized?

---

📈 North Star Metric (NSM)

- Retention Rate / Active Subscription Rate

Supporting Metrics:

• Monthly Recurring Revenue (MRR)
• Customer Churn Rate
• Customer Lifetime Value (CLV)
• Average Revenue Per User (ARPU)
• Tenure Distribution
• Contract Conversion Rate

---

📊 Business Context Simulation

Assume:

• 10,000 active customers
• 26% churn rate
• $70 average monthly revenue
• A 5% reduction in churn leads to significant annual revenue preservation.

---

This project provides predictive modeling to enable proactive intervention before churn occurs.

🔍 Exploratory Product Insights

Key findings from EDA:

• Month-to-month contracts have the highest churn probability
• Short-tenure customers churn within the early lifecycle stage
• Higher monthly charges correlate with increased churn
• Electronic check payment method shows elevated churn behavior

These insights inform targeted retention strategies.

---

🤖 Machine Learning Approach

Models Evaluated:

• Logistic Regression
• Random Forest
• Decision Tree
• Gradient Boosting (if applicable)

Evaluation Metrics:

• Accuracy
• Precision
• Recall
• F1 Score
• ROC-AUC

The selected model achieved strong ROC-AUC performance and improved churn risk identification.

---

🧠 Feature Importance & Driver Analysis

Top predictors of churn:

• Contract Type
• Tenure
• Monthly Charges
• Internet Service
• Payment Method

Interpretation:

Churn behavior is influenced primarily by lifecycle stage and contract structure rather than demographics — indicating product and pricing optimization opportunities.

---

🧪 Experimentation Strategy (Product Lens)

• Instead of stopping at prediction, this project proposes actionable experiments:

1️⃣ Contract Conversion Incentive

• Offer discounted annual plans to month-to-month customers.

Hypothesis:

• Customers transitioning to long-term contracts will reduce churn probability.

2️⃣ Early Lifecycle Engagement Experiment

• Target customers within the first 90 days with onboarding nudges.

Hypothesis:

• Improved onboarding engagement increases long-term retention.

3️⃣ High-Value Customer Retention Offer

• Provide personalized loyalty offers to high ARPU customers flagged as high churn risk.

Hypothesis:

• Targeted retention incentives preserve revenue efficiently.

All experiments can be validated using A/B testing frameworks.

🏗️ End-to-End Workflow

• Data Cleaning & Preprocessing
• Exploratory Data Analysis (EDA) with business interpretation
• Feature Engineering (contract type, tenure segmentation, billing patterns)
• Encoding & Scaling
• Train-Test Split
• Model Training (Logistic Regression, Random Forest, etc.)
• Model Evaluation (ROC-AUC, Precision, Recall, F1)
• Feature Importance Analysis
• Product-Level Interpretation & Strategy Recommendation

---

🛠️ Tech Stack

• Python
• Pandas
• NumPy
• Scikit-learn
• Matplotlib
• Seaborn
• Jupyter Notebook
• Streamlit (optional deployment)

---

📂 Project Structure

Customer-Churn/
│
├── data/
├── notebooks/
├── models/
├── app.py
├── requirements.txt
└── README.md

📊 Business Impact Summary

• If implemented in production, this system enables:
• Proactive churn detection
• Targeted retention campaigns
• Revenue preservation
• Experiment-driven product decisions
• Lifecycle-based segmentation

Even a 3–5% reduction in churn materially increases annual recurring revenue and customer lifetime value.

---

🔮 Future Enhancements

• SHAP-based model explainability
• Survival analysis for churn timing
• Uplift modeling for retention targeting
• Real-time churn scoring API
• Automated A/B testing simulator integration
• Deployment with FastAPI

---

🎯 Skills Demonstrated

• Product Data Science
• Customer Retention Analytics
• Churn Modeling
• Predictive Modeling
• Feature Engineering
• Machine Learning
• Experiment Design
• Revenue Optimization
• Business Analytics
• Python

---

How to Run this

Clone the repository:

• git clone https://github.com/Denis0242/Customer-Churn.git
• cd Customer-Churn

Install dependencies:

• pip install -r requirements.txt

Run the notebook:

• jupyter notebook

If using Streamlit (optional):

• streamlit run app.py

---

👤 Author

Denis Agyapong

Product Data Scientist | Data Analyst

Oakland, CA