Bank Customer Churn Analysis
End-to-end Python analysis identifying key factors driving customer churn — from data cleaning through EDA to actionable business recommendations.
Overview
End-to-end Python analysis identifying the key drivers of customer churn for a bank with 10,000 customers across France, Germany, and Spain. Covers the full data science workflow — raw data ingestion, cleaning, EDA, and actionable business recommendations.
Dataset: 10,002 rows × 14 columns → cleaned to 10,000 rows × 11 features
Target variable: Exited (1 = churned, 0 = retained)
Objective
Identify the key factors driving customer churn and deliver data-backed recommendations to reduce attrition and improve retention strategy.
Phase 1 — Data Understanding & Preparation
- Initial Exploration — loaded dataset, checked dimensions, reviewed data types and column structure
- Quality Assessment — found 2 duplicate rows, 4 columns with 1 missing value each, Age stored as float
- Cleaning — removed duplicates and 3 irrelevant columns (RowNumber, CustomerId, Surname); filled missing values with mode/median; corrected data types
- Output — clean dataset: 10,000 rows × 11 columns, zero missing values, CreditScore range 350–850 valid ✅
Phase 2 — Exploratory Data Analysis (EDA)
Univariate Analysis
- France: 50.14% of customers · Germany: 25.09% · Spain: 24.77%
- Age distribution: 31–40 is largest group (44.5%); 51–60 is smallest active segment (8.0%)
- 36.16% of customers have zero balance
- Overall churn rate: 20.37% (2,037 of 10,000 customers)
Bivariate Analysis — Churn by Geography
| Country | Customers | Churn Rate |
|---|---|---|
| France | 5,014 | 16.15% |
| Germany | 2,509 | 32.44% |
| Spain | 2,477 | 16.67% |
Germany churns at double the rate of France and Spain despite fewer customers.
Bivariate Analysis — Churn by Age Group
| Age Group | Churn Rate |
|---|---|
| 18–30 | 7.52% |
| 31–40 | 12.08% |
| 41–50 | 33.97% |
| 51–60 | 56.21% |
| 60+ | 24.78% |
Bivariate Analysis — Other Features
- Female churn: 25.1% vs male: 16.5% — 8.6pp gap
- Inactive members churn at 26.9% vs active members — engagement directly reduces attrition
- 1-product customers: 50.84% of base, highest churn; 2-product: lowest churn; 3–4 products: disproportionately high churn despite only 3.26% of base
- Churned customers average age 44.84 vs 37.41 retained — age is the strongest predictor
- CreditScore difference: only 6.5 points between churned and retained — weak predictor
Correlation Analysis
- Age: strongest positive correlation with churn
- IsActiveMember: negative correlation — inactive = higher churn
- CreditScore & EstimatedSalary: minimal correlation with churn
Key Findings
- Germany is a critical risk market — 32.44% churn vs ~16% in France and Spain
- Middle-to-older customers (41–60) are the highest risk — 51–60 age band peaks at 56.21%
- Female customers churn at 25.1% vs 16.5% for males — 8.6 percentage point gap
- Inactive members churn at 26.9% — engagement directly reduces attrition
- Customers with 3–4 products have extremely high churn despite being a small segment (3.26%)
- CreditScore is a weak predictor — only 6.5 point difference between churned and retained
Business Recommendations
- Targeted retention campaigns in Germany with localised, personalised offers
- Age-segmented loyalty programmes for customers aged 41–60
- Re-engagement campaigns for inactive members (26.9% churn risk)
- Gender-specific engagement strategies for female customers
- Cross-sell products to single-product customers to increase stickiness
- Review product bundling strategy — 3–4 product customers show disproportionate churn