4most have a long-standing relationship working with a leading automaker that provides vehicles worldwide and offers financing to its customers.
Our client approached us to help them enhance their credit application underwriting process. Their primary goal was to automate the application process and subsequently improve customer experience, reduce costs and obtain more consistent decisions in line with regulations and compliance.
The project was delivered in phases; their current decision strategy was optimised through identification of consistent behaviour while enabling the creation of a comprehensive and transparent database, that could then feed the data science exercise. This alone resulted in a 30% reduction of underwriting effort. This task identified a population that became the target of the data science exercise.
A horse race between several algorithms was performed to establish the model structure that best described the relationships in the data to predict the target (Accept/Reject). A key advantage of using machine learning techniques over more traditional methods was the ability to handle large volumes of data effectively and therefore, outperforming the existing policy rules in the identification of desired customers.
For this, we used our specialised tool, Columbus. Columbus is a flexible and customisable R/Python based tool, which allows the user to explore a model and identify relationships from the data. Regardless of model complexity, the tool enables users to understand the model output in many forms such as a decision, a ranking or a probability.
Columbus focuses on variable interpretation, understanding how certain profiles influence the modelling target, making it a comprehensive model monitoring solution. Its capability to explore individual cases and identify the reasons underlying model decisions and recommendations is a distinguishing feature. This benefit not only enhances transparency and understanding but also supports compliance in heavily regulated industries - the techniques used are unique since they are based on bespoke metrics that not only work but are easy to explain.
We reduced our client’s overall underwriting effort by 80%. We were aware of the challenges involved in fully integrating ML techniques in an underwriting process, but our expertise and persistence resulted in impressive results.
In addition to cost efficiency, we delivered a streamlined and efficient underwriting process, which subsequently allowed:
Faster application processing for better customer experience
More extensive data available to drive more informed decisions
Consistent decisions - resilience to biases and oversights
Accept rates potentially lower than current level to maintain bad rate
Simplified decline process - reducing underwriter effort in processing recommended decline
Flexible approach - can be quickly and easily adapted to changing scenarios.
Underwriting is a costly part of any business and can be a regular source of complaints. Being able to automate it, was a huge advantage to the company and its customers alike.
If you would like further information, please contact our Head of Data Science:
Fabrizio Russo, email@example.com