A Business Analyst & APM portfolio project analyzing customer churn patterns in retail banking — combining data analysis, ML-based risk prediction, and actionable business recommendations.
Live App: bank-churn-dashboard.streamlit.app
A retail bank is experiencing a 20.4% customer churn rate, resulting in significant revenue loss through lost balances and reduced lifetime value. This project identifies who is churning, why they are churning, and what the business should do about it.
- Quantify churn rate and revenue impact across customer segments
- Identify high-risk cohorts by geography, age, product usage, and engagement
- Build a real-time churn risk calculator to flag individual at-risk customers
- Translate findings into actionable business recommendations
High-level KPIs for leadership and stakeholders:
| Metric | Description |
|---|---|
| Total Customers | Count of customers in selected segment |
| Churn Rate (%) | % of customers who have left the bank |
| Revenue at Risk ($) | Total balance lost to churned customers |
| Avg Credit Score | Financial health indicator of the segment |
Deep-dive cohort analysis across demographics:
- Active vs. Inactive member churn comparison
- Segment breakdown by Geography × Gender
- Avg Balance, Avg Credit Score, and Churn Rate per cohort
On-the-fly ML prediction using a Random Forest model:
- Input any customer profile and get an instant churn probability
- Risk categorization: Low (<30%) | Medium (30–60%) | High (>60%)
- Germany has the highest churn rate (~32%) vs France and Spain (~16%)
- Age group 40–60 churns significantly more than younger customers
- Inactive members are 2× more likely to churn than active ones
- Customers with 3–4 products show unexpectedly high churn — suggesting product-fit issues
- High-balance customers represent disproportionate revenue at risk
- Targeted retention campaigns for the 40–60 age group with personalized offers
- Re-engagement program for inactive members — triggered email/SMS within 30 days of inactivity
- Germany market review — investigate local competitive landscape and pricing
- Product bundling audit — review why multi-product customers churn more; assess onboarding quality
| Tool | Purpose |
|---|---|
| Python | Data analysis and ML modeling |
| Pandas | Data manipulation |
| Plotly | Interactive visualizations |
| Scikit-learn | Random Forest churn prediction model |
| Streamlit | Dashboard deployment |
bank-churn-dashboard/
│
├── data/
│ └── bank_churn.csv # Source dataset (Maven Analytics)
├── app.py # Main Streamlit dashboard
├── requirements.txt # Python dependencies
└── README.md
This project is supported by full business documentation:
- BRD (Business Requirements Document) — stakeholder requirements, AS-IS/TO-BE process, functional requirements
- PRD (Product Requirements Document) — feature specs, user stories, success metrics (Available on request)
- Source: Maven Analytics — Bank Customer Churn
- Size: 10,000 customers
- Features: Credit Score, Geography, Age, Tenure, Balance, NumOfProducts, IsActiveMember, EstimatedSalary, Exited
Aditya — 3rd Year CSE Student | Aspiring APM & Business Analyst