Bank Customers Churn Analysis
- Michael Olaniyi Jeremiah
- Aug 14
- 5 min read
Updated: Aug 15
By Michael Olaniyi Jeremiah
Project Background
Customer churn is one of the most critical challenges in the banking sector. Losing customers not only affects revenue but also increases acquisition costs to replace them.
This project was initiated to understand the patterns, behaviors, and factors that contribute to customer churn, enabling proactive strategies to retain valuable customers.
Objective
To perform a comprehensive analysis of bank customer data, identify churn trends, and present actionable insights through an interactive dashboard.
Data Collection & Overview
The dataset consisted of 10,000 bank customer records with attributes such as demographics, account information, customer activity, and satisfaction scores.
Key original columns included:
CustomerId, Surname, Geography, Gender, Age, Credit Score, Tenure, Balance, NumOfProducts, HasCrCard, IsActiveMember, EstimatedSalary, Satisfaction Score, Card Type, Point Earned, and Exit status (Exited).
Data Cleaning
The raw dataset contained inconsistencies and redundancy. The cleaning process included:
Removing duplicate entries based on CustomerId.
Standardizing text fields (e.g., ensuring consistent spelling in “Geography” and “Gender”).
Validating numerical ranges (e.g., credit score between 300–900, satisfaction scores between 1–5).
Checking for missing values and applying appropriate treatments (imputation or removal).
Data Enrichment
To improve analysis and storytelling, three new key columns were engineered:
Age Group: Customers were categorized into groups like Young Adult, Middle Age, Senior, etc., to spot churn trends by life stage. The Excel function used was Switch:
=SWITCH(TRUE,[@Age]>=70,"Very Old",[@Age]>=50,"Old",[@Age]>=35,"Middle Age","Young") Churned: A human-readable label based on the Exited flag, allowing clearer communication in reports. The Excel function used was IF
=IF([@Exited]=1,"Yes","No") Complained: A derived field capturing whether a customer had lodged a complaint, enabling correlation analysis between complaints and churn likelihood. The Excel formular used was
=IF([@Complain]=1,"Yes","No") These enrichments transformed the dataset into a more decision-friendly format.
Key Performance Indicators (KPIs)
To provide a high-level view of customer dynamics, KPIs were calculated as follows:
KPI | Value |
Total Customers | 10,000 |
Retained Customers | 7,962 |
Retention Rate | 79.62% |
Churn Customers | 2,038 |
Churn Rate | 20.38% |
Average Satisfaction Score | 3.01 ★★★☆☆ |
Average Tenure (Years) | 5.01 |
Average Age (Years) | 38.92 |
These KPIs revealed that roughly 1 in 5 customers had churned, and the average satisfaction score was slightly above the midpoint, signaling room for improvement.
Data Analysis
A combination of descriptive and diagnostic analysis was conducted:
Churn Rate Calculation: Determined overall churn percentage.
Demographic Trends: Analyzed churn patterns by gender, geography, and age group.
Behavioral Insights: Studied the impact of credit score, tenure, balance, and product usage on churn.
Complaint Correlation: Identified strong churn associations with customers who had lodged complaints.
Customer Value: Explored the link between churn and high-value customers based on balance and points earned.
Correlation Analysis (What most drives churn?)
I quantified linear relationships using Pearson correlation between Churn (0/1) and each variable.
Gender Churn: +0.50 (moderate positive)
Interpretation: churn differs meaningfully by gender; the segment encoded as “Female” showed higher churn (25.07%) than “Male” (16.47%).
Takeaway: gender-targeted retention plays (messages, offers, service touchpoints) are justified.
Credit Score Churn: –0.44 (moderate negative)
Interpretation: higher credit score → lower churn.
Takeaway: customers with lower credit scores are a higher-risk segment; prioritize financial coaching, fee waivers, or product-fit reviews.
Tenure Churn: –0.42 (moderate negative)
Interpretation: the longer a customer has banked with us, the less likely they are to churn.
Takeaway: invest in early-life onboarding and first-year loyalty nudges to push new customers past the high-risk window.
Note: Positive means churn rises as the variable increases (or for the encoded group); negative means churn falls as the variable increases.
Data Visualization
KPI cards (churn, retention, averages), trend and segment charts, and slicers for Age Group, Geography, Churned, and Complained.
A “Drivers” view highlights the Gender (+0.50), Credit Score (–0.44), and Tenure (–0.42) relationships alongside segment distributions.
Using Excel’s Pivot Tables, Charts, and Slicers, the analysis was brought to life:
Churn by Age Group & Gender
Churn by Geography
Churn vs Complaint History
Customer Tenure Distribution
Product Usage Patterns
Satisfaction Score vs Churn
Demographics Dashboard:
Others Dashboard (This includes Key findings and recommendation)
Bank Customers Churn Key Insights & Recommendations
1. Churn by Age
Insight: Higher coefficient (0.50) suggests that age is positively correlated with churn rate.
Recommendation: Develop tailored retention campaigns for older customers, such as loyalty rewards, personalized financial advice, and exclusive offers.
2. Churn by Geography
Insight: Germany shows a slightly higher churn rate compared to Spain and France, indicating a potential regional issue.
Recommendation: Conduct a localized survey in Germany to identify customer dissatisfaction factors. Introduce region-specific retention programs and targeted marketing campaigns.
3. Churn by Average Balance
Insight: Customers with higher average balances are more likely to churn.
Recommendation: Provide premium customers with enhanced benefits such as priority service, relationship managers, and exclusive investment opportunities.
4. Churn by Number of Products
Insight: Customers with 3–4 products have significantly higher churn than those with 1–2 products.
Recommendation: Review cross-selling strategies to ensure added products provide value and do not overwhelm customers. Offer personalized product bundles based on actual customer needs.
5. Churn by Tenure
Insight: Churn occurs across all tenure levels, with a slight spike for customers with 2–3 years of tenure.
Recommendation: Implement mid-tenure engagement programs, such as annual account reviews or exclusive benefits after 2 years.
6. Tenure & Multiple Factors
Insight: Early tenure churn remains notable despite satisfaction increasing over time.
Recommendation: Strengthen onboarding programs to build trust early. Provide early-tenure incentives and regular follow-ups in the first 6 months.
7. Churn by Active Membership
Insight: Active members have a higher churn rate (26.07%) than inactive members (16.47%).
Recommendation: Investigate service gaps for active users and address them with better engagement support and feedback loops.
8. Churn by Credit Card Type
Insight: Gold cardholders have the lowest churn rate; highest churn rate is 31.19% among other customers.
Recommendation: Promote upgrades to Gold membership with targeted offers and incentives.
9. Churn by Credit Score
Insight: Customers with lower credit scores are more likely to churn.
Recommendation: Offer financial coaching, credit-building products, and better repayment options to at-risk customers.
10. Churn by Complaint Status
Insight: Customers with complaints show a sharp rise in churn, complaint handling is the leading churn indicator.
Recommendation: Strengthen complaint resolution processes and introduce real-time issue tracking. Conduct root-cause analysis of frequent complaints to prevent recurrence.
-----------------------------------------------------------------------------------------------------------------------
Thank you for taking the time to review this project. I welcome your comments, suggestions, and feedback.
For any project discussions or job opportunities, please feel free to contact me at:
Email Address: michaeljeremiah124@gmail.com
Phone: 234706664402
GitHub: https://github.com/mikeolani
Comments