Customer churn is a critical issue for telecom companies, directly impacting revenue and profitability. Without clear insights into why customers leave and which current customers are at risk, companies struggle to implement effective retention strategies.
The goal of this project was to:
- Analyze historical churn patterns
- Identify key drivers of churn
- Build a predictive model to score current customers by churn risk
- Provide business-friendly dashboards to support proactive retention efforts
- Power BI (interactive dashboards, Q&A)
- Python (data preprocessing, machine learning)
- scikit-learn (ML modeling)
- Pandas (data wrangling)
- Jupyter Notebook (analysis and modeling)
- Kaggle Telecom Customer Churn Dataset
The project was completed in four stages:
Segmented churned customers and analyzed their behavior using Power BI. Key variables included contract type, payment method, tenure, and monthly charges. Insights were visualized to help the business understand churn patterns.
Analyzed the full customer base to identify broader trends. Created segmentation views and demographic overlays in Power BI to compare churned vs. non-churned customers.
Developed a classification model (Logistic Regression) to predict churn risk for current customers:
- Engineered relevant features (tenure, contract type, service usage, etc.)
- Trained and validated the model
- Scored current customers and categorized them by low, medium, or high risk
- Integrated risk scores into Power BI dashboards
Enabled Power BI’s Q&A feature to allow ad-hoc analysis by business users using natural language queries.
- Data quality: The dataset required significant preprocessing (missing values, categorical encoding, outlier handling).
- Feature selection: Identifying the most predictive features for the ML model required multiple iterations and tuning.
- Model interpretability: Translating model results into business-friendly insights was key to adoption.
- Visualization: Designing dashboards that are both insightful and actionable for business teams required thoughtful iteration.
-
Identified key churn drivers, including contract type, short tenure, and automatic payment methods.
-
Built a churn risk prediction model with strong classification performance.
-
Scored current customer base for future churn risk.
-
Delivered an interactive Power BI report with:
- Churn trend visualizations
- Key driver analysis
- Current customer churn risk scores
- Natural language Q&A interface
-
Created a foundation for targeted retention campaigns and data-driven decision-making.