Objective

To help a leading U.S. retail bank increase cross-sales and reduce marketing expenses by creating a machine-learning-based model to identify deposit holders/customers most likely to accept a personal loan.

CRISIL's Solution

• Implemented various data mining techniques to identify customers who most likely to accept a personal loan
• Developed Exploratory Data Analysis (EDA) to find the outliers, trends and patterns in customer data
• Developed Tree-Based classification models (classification and regression trees, unbiased recursive partitioning: a conditional Inference and random forests) to analyze the relationship between the target variable and covariates
• Developed a support vector machine (SVM) algorithm to classify customers 
• Developed a k-means clustering approach to group customers to study their grouping properties
• Employed open-source software R was for this data-intensive implementation

Client Impact

• Increased revenues by enhancing cross-selling success rates. New model improved immediate response rate by 21 percent and the long-term response rate by 34 percent
• Allowed bank to reduce marketing expenses through effective sales targeting
• Model integrated into the client’s systems and is now a key driver of sales & marketing strategies
  

Questions

Looking for high-end research and risk services? Reach out to us at: