Telco Customer Churn Analysis (Week 1)

This project presents an end-to-end customer churn analysis using a real-world telecom dataset.
It focuses on business understanding, clean data preparation, baseline modeling, and actionable insights.

---

📌 Business Problem

Customer churn occurs when customers discontinue a service.
For telecom companies, churn directly impacts revenue, growth, and customer lifetime value.

The goal of this project is to:

• Understand churn drivers
• Predict customers likely to churn
• Provide recommendations to reduce churn proactively

---

📊 Dataset

• Source: Telco Customer Churn Dataset
• Size: 7,043 customers, 21 raw features
• Target Variable: Churn (Yes / No)

The dataset includes:

• Customer demographics
• Subscription details
• Service usage
• Billing and tenure information

---

🛠️ Project Workflow

1. Problem framing and business understanding
2. Data cleaning and preprocessing
3. Feature engineering
4. Exploratory analysis
5. Baseline churn modeling (Logistic Regression)
6. Model evaluation
7. Business recommendations

---

🔧 Key Data Preparation Steps

• Converted TotalCharges from string to numeric to avoid incorrect encoding
• Handled missing values using median imputation
• Engineered a new feature: num_services
• One-hot encoded categorical variables
• Ensured proper target variable mapping
• Applied stratified train-test split

---

🤖 Model Used

Logistic Regression (Baseline Model)

Chosen for:

• Interpretability
• Strong baseline performance
• Industry relevance for churn analysis

---

📈 Model Performance (Test Set)

• Accuracy: ~80%
• Recall (Churn): ~52%
• Precision (Churn): ~65%
• ROC-AUC: 0.84

Recall was prioritized because missing a churn-risk customer directly results in revenue loss.

---

💡 Key Insights

• Customers with shorter tenure and fewer services are more likely to churn
• The model shows strong ability to distinguish churners from non-churners
• Logistic Regression provides a solid baseline for further improvements

---

📌 Business Recommendations

1. Prioritize customers predicted to churn for retention campaigns
2. Offer personalized incentives to low-tenure customers
3. Encourage bundling of services to increase engagement
4. Monitor churn predictions regularly and retrain models as data evolves

---

🚀 Next Steps

• Train advanced models (Random Forest, Gradient Boosting)
• Perform feature importance analysis
• Build churn dashboards (Power BI / Tableau)
• Improve recall through threshold tuning and class weighting