New Data Sources' Contribution to Auto Lending

AI and ML models enable lenders to derive reliable credit risk insights based on encounters between heaps of new data variables previously inaccessible to conventional lending models, such as application data or CRM data already accessible to the lender

FREMONT, CA: Machine learning is in the process of changing the auto lending market, mainly as other conventional credit scores have done before. Auto lenders who are capable of implementing artificial intelligence could attract more consumers and take less risk at the same time.

Below is how auto lenders can use emerging data sources to accept more applicants and take less risk.

New Data Sources for Auto Lending Decisions

Auto lenders are accustomed to employing logistic regression models that weigh in one to two dozen variables when deciding whether or not to grant a loan to the borrower. In general, the applicant's credit

 background and present use, wages, workplace, down payment, and other financial and credit factors are considered when determining whether or not to accept an application.

However, Machine Learning (ML) algorithms can process hundreds or thousands of variables and look at an almost infinite number of relationships between these variables to determine creditors' risk. Several data points might inform an MI lending model, many of which are ignored by most conventional models.

AI and ML models enable lenders to derive reliable credit risk insights based on encounters between heaps of new data variables previously inaccessible to conventional lending models, such as application data or CRM data already accessible to the lender. For example, widening the data collection helps the lender factor in if the borrower was bankrupt and, in the case of bankruptcy, what kind of bankruptcy it was, and how bad it was.

An ML model can also affect the applicant's customer contact center background, the history of residency, or whether the candidate has an open court proceeding that may impact the applicant's financial status based on how they are settled. Auto companies with comprehensive digitized historical transaction data could feed the data into an algorithm to gain insight into who their individual dealerships can accept loans for specific vehicles. This instance has been achievable for previous statistical approaches but is brought to a higher degree of granularity with new data sources and ML.

For example, it could be that, on average, around the U.S., people who buy one company's vehicle are less able to repay their loans than people who buy a different firm's automobile. However, a particular dealership can use this data and algorithm to determine their customer base is above average for loan repayment.

As a consequence of the fine-tuning that ML provides, dealerships may accept more loans than they would generally use national data from dealerships and lenders worldwide. In comparison, they will take on less risky candidates who would have been accepted using conventional credit scores.