Use Case: Predicting loan defaults

cancel
Showing results for 
Search instead for 
Did you mean: 

Use Case: Predicting loan defaults

Risk assessment tools for small and medium enterprise (SME) fintech lenders


What’s the problem?

Digital lending is a significant part of the fintech industry. In 2015, digital lending originations in the United States were $25 billion. By 2020, originations are expected to grow to $90 billion. As digital lending becomes a more prominent part of the fintech ecosystem, the need to appropriately price financial products for risk of default becomes increasingly important. While risk-based pricing is a common underwriting operation of large financial institutions, many small and medium enterprise (SME) fintech lenders are still developing the tools needed to appropriately price their risks. These smaller lenders often lack the resources to develop state-of-the-art underwriting platforms. As a result, default risk for SME lenders continues to be a paramount concern.

The challenge and solution

SME lenders face two challenges when building underwriting platforms to address default risk. The first is to acquire data on an applicant that is sufficient for developing a business profile of that applicant. This challenge is more easily addressed by the multitude of third-party resources available for purchase. The more pressing challenge is a method for using this data to make accurate default risk predictions in real-time. An underwriting platform can address this challenge by incorporating automated machine learning algorithms that take advantage of applicant data, building an appropriate risk-based pricing scheme.

For example…

Sally is an underwriter for an SME lender specializing in business loans. She receives hundreds of applications per day. Her job is to determine if an application will be approved. Sally gathers several variables about each applicant and starts to use common business rules to exclude applications not meeting basic criteria for funding. However, Sally is then left with a multitude of applications which appear to be eligible for approval but still need to be appropriately priced for their default risk. This is where Sally turns to DataRobot to score these applications, in near real-time. She can then provide credit decisions to each applicant in an effort to win their business, before they shop for a loan from another lender.


Training Datasets: https://s3.amazonaws.com/datarobot-use-case-datasets/Lending+Club+Dataset+Train.csv
Prediction Datasets: https://s3.amazonaws.com/datarobot-use-case-datasets/Lending+Club+Dataset+Pred.csv
Data Dictionary: https://s3.amazonaws.com/datarobot-use-case-datasets/Predicting+Loan+Defaults+-+Data+Dictionary.pdf 

(Originally posted August 2017)

Version history
Revision #:
8 of 8
Last update:
‎01-24-2020 11:36 AM
Updated by: