Direct Marketing Model for Student Loan and Credit Card
The Challenge
- A large financial institution wanted to optimize their direct to consumer marketing campaign for credit cards and student loans.
- They needed a machine learning model that could predict the responses of the potential customers, maximize the response rate, and optimize the offer window.
- They also wanted to maximize the total profit of the lifetime value of the customers, not just the accuracy of the selection.
The Solution
- We developed a boosted tree based classification model that used recency/frequency analysis to select the top potential responders and the best timing of the offer.
- We also customized the objective function of the model to maximize the overall lifetime profit of the institution, taking into account the cost of the marketing campaign, the default rate and amount, and the customer acquisition value.
- We used third party information at zip code level to enrich the data and improve the model performance.
The Results
- Our machine learning model increased the response rate for student loans by 23% and credit cards by 20%, compared to the institution’s previous model.
- More importantly, our model maximized the quality of the targeted customers, resulting in:
- Higher profit, as we increased the customer acquisition value by 15% and reduced the marketing cost by 10%.
- Lower default risk, as we decreased the default rate by 12% and reduced the default amount by 8%.
- Better customer retention, as we increased the customer loyalty by 18% and reduced the churn rate by 10%.