RPA and Smart Automation Case Studies

Predictive Modeling: A Marketing Solution

Every month the client considers a large group of people who have mortgages to potentially send out offers about refinancing to. Previously, hand-written rules were manually created by subject matter experts to reduce this large pool of potential recipients to a smaller number of borrowers to send offers to. Of these thousands of offers, which they send monthly, typically less than 10% of the recipients subsequently initiate a loan application to refinance with the client. In this case, Machine Learning can more intelligently identify which borrowers are most likely to refinance therefore allowing the client to specifically target those most likely to respond. This helps the clients increase the number of borrowers that refinance with them and helps them to spend their marketing budget more effectively.

  • Our Data Scientists worked with subject matter experts to understand the business model, mathematically define the problem performance metrics, and weigh the costs and benefits at play for the given problem.
  • Exploratory Data Analysis (EDA) was conducted, which looked at what data was available in order to understand its volume, as well as what patterns and relationships existed among the variables being considered.
  • The data selected during the EDA was processed into a format that the machine learning algorithms used to perform the predictions.
  • Basic algorithms are trained, and rudimentary experiments were performed, to identify the best subset of data and which models would be used in the final application.
  • The identified best models were optimized to maximize the predictive power of the model.
  • The model was packaged into an application which was then delivered to and deployed for the client.

The overall solution to the problem was a machine learning application which the client can use each month to identify the borrowers most likely to refinance. This enables them to reduce the number of offers they send to people that are unlikely to refinance as well as improves their ability to follow up with those most likely to refinance.

Reduced the number of offers sent by 20-25% each month
Increased number of responses by 10% each month