Analyzing Bank Customer Churn: A triple approach to predictive and analyzing customer churn

Churn. The dreaded "C" word for any business. Losing customers is costly, and understanding why they leave is crucial. As an analyst, I'm constantly seeking ways to predict and mitigate churn, and recently, I dove deep into this challenge using a powerful combination of SQL, predictive analytics, and Power BI. Let me walk you through my process.

The Data Foundation: SQL for Efficiency

First things first: the data. For this analysis, I focused on three key factors that often influence churn: demographics, credit score, financial history and bank product usage. Accessing and preparing this data efficiently is paramount. That's where SQL shines. Instead of wrestling with clunky spreadsheets, I leveraged the power of SQL to extract and transform the relevant data directly , think of it as precisely carving out the information I needed, nothing more, nothing less.

SQL

WITH Income_Group AS (
    SELECT  
        CASE 
            WHEN Balance < 50000 THEN 'Low_Income'
            WHEN Balance < 100000 THEN 'Mid_Income'
            WHEN Balance < 150000 THEN 'High_Income'
            ELSE 'Ultra_Income'
        END AS Income_Category,
        IsActiveMember,
        Exited,
        Geography,
        Gender
    FROM Bank_Cust
)
SELECT Income_Category, Geography, Gender, COUNT(*) AS Group_Count
FROM Income_Group
WHERE IsActiveMember = 1 AND Exited = 1
GROUP BY Income_Category, Geography, Gender
ORDER BY Group_Count DESC;

This snippet, for example, demonstrates how I could pull specific customer attributes and their churn status. SQL's ability to filter and aggregate data is a game-changer for speed and efficiency. It allows me to perform complex data manipulation tasks without exporting massive datasets, saving significant time and resources.

Visualizing the Story: Power BI for Insight

Once the data was prepped, I turned to Power BI to bring it to life. Power BI's interactive dashboards allowed me to visualize the relationships between my chosen factors and churn. I created charts and graphs that revealed trends and patterns that would have been difficult to spot in raw data. For instance, I could easily see how churn rates varied across different age groups, countries, gender or credit score ranges.

Power BI’s strength lies in its ability to translate complex data into easily digestible visual stories. This is crucial for communicating findings to stakeholders who may not be data experts. A well-crafted dashboard can quickly highlight key insights and drive data-informed decision-making.

Predicting the Future: Predictive Modeling with Random Forest and Decision Trees

Now, for the exciting part: predicting churn. While SQL and Power BI helped me understand what was happening, predictive modeling allowed me to anticipate what might happen. I employed two powerful algorithms: Random Forest and Decision Tree.

• Decision Tree:  This algorithm creates a tree-like structure to classify customers based on their attributes. It's relatively easy to interpret, providing clear rules for why a customer might be at risk of churning.

  

  Decision Tree Classification Report

• Random Forest:  This algorithm takes the Decision Tree concept a step further by building multiple trees and combining their predictions. This approach often leads to higher accuracy and is less prone to overfitting.

  

  Random Forest Classification Report

Both algorithms offered valuable insights, with Random Forest generally providing more accurate predictions in this specific use case. However, the interpretability of the Decision Tree was beneficial for understanding the key drivers of churn.

Key Takeaways and Recommendations

This analysis reinforced the power of combining different analytical tools. SQL provided the foundation for efficient data preparation, Power BI enabled clear visualization and communication, and predictive modeling allowed me to anticipate churn.

For complex predictive models, tools like Random Forest and Decision Trees are essential. However, for initial exploration and simpler analyses, SQL and Power BI offer a cost-effective and time-efficient solution.

As an analyst, these tools are invaluable for gaining granular insights and making data-driven recommendations. In the case of churn, this analysis empowers us to proactively target at-risk customers with retention strategies, ultimately reducing customer loss and boosting the bottom line. Moving forward, I'll continue to refine these models and explore additional factors that might influence churn, ensuring we stay ahead of the curve.