Customer Churn Analysis & Prediction Overview
This project explores customer churn behaviour using a real-world telecom dataset. The goal was to identify key factors influencing churn and build a machine learning model to predict customers at risk of leaving.
Objectives Clean and preprocess raw customer data Perform exploratory data analysis (EDA) Identify key drivers of churn Build a predictive model using Logistic Regression Generate churn probabilities for each customer Visualise insights using Power BI
Tools & Technologies Python (Google Colab) Pandas & NumPy Scikit-learn Power BI
Dataset Telco Customer Churn Dataset Includes customer demographics, contract details, billing information, and churn status
Data Cleaning & Preparation Removed irrelevant columns (e.g. location data, IDs) Converted Total Charges to numeric format Handled missing values Encoded categorical variables using one-hot encoding
Model Development Model used: Logistic Regression Train/test split: 80/20 Features included: Tenure Months Monthly Charges Total Charges Contract Type Payment Method Paperless Billing
Results Model Accuracy: ~79% Strong performance in predicting non-churn customers Moderate performance in identifying churn customers
Key Insights Customers on month-to-month contracts have higher churn risk Certain payment methods are associated with higher churn Higher monthly charges can correlate with increased churn probability Churn risk can be segmented into High / Medium / Low categories
Power BI Dashboard An interactive dashboard was created to:
Visualise churn probability by contract type Identify high-risk customer segments Explore trends across payment methods and tenure Highlight customers with high churn probability
Future Improvements Apply advanced models (Random Forest, XGBoost) Improve recall for churn prediction Perform feature importance analysis Deploy as a simple web app or dashboard
Project Structure customer_churn_analysis.ipynb → Data cleaning, modelling, and analysis churn_with_probabilities.csv → Final dataset with predictions Power BI dashboard → Visual insights
Conclusion This project demonstrates an end-to-end data analysis workflow, from raw data to actionable business insights, combining Python and Power BI to support data-driven decision making.
Business Recommendations Based on the analysis and model predictions, the following actions could help reduce customer churn:
1. Target High-Risk Customers Customers with a churn probability above 70% should be prioritised for retention strategies such as: Personalised offers or discounts Proactive customer support outreach Contract upgrade incentives
*2. Encourage Long-Term Contracts Customers on month-to-month contracts showed higher churn risk. Offer incentives to move customers onto 1-year or 2-year plans Provide pricing benefits for longer commitments
3. Optimise Payment Methods Certain payment methods (e.g. electronic check) are associated with higher churn. Encourage customers to switch to automatic payments Offer small discounts for auto-pay adoption
4. Monitor High Monthly Charges Customers with higher monthly charges may be more likely to churn. Introduce loyalty discounts for long-term customers Offer bundled services to increase perceived value
5. Implement a Churn Monitoring System Use the model to regularly score customers Create a live dashboard (Power BI) for tracking churn risk Enable customer service teams to act on real-time insights
Business Impact Implementing these strategies could: Reduce customer churn rate Increase customer lifetime value (CLTV) Improve customer satisfaction and retention Support more data-driven decision making