Harnessing the Reins of Powerful AI/ML Technology

Irresistible website designs and kickass marketing programs that deliver results

Learn More
hero-jobbies-7

Like a high-performance vehicle, automated scoring solutions are incredibly powerful engines that take you where you tell it to take you. Also like a high-performance vehicle, if you're steering full-speed into a ditch, the machine will not know that and you will be virtually guaranteed to crash. In the lending industry, the end goal is almost always to maximize profits, but in order to do so, lenders need to strategically select targets along the way, with many complex and nuanced considerations. 

Machine Learning in Credit Scoring: A Simplified Briefer

Without getting too technical, machine learning (ML) is a subset of AI that improves and optimizes a software's outputs (generally predictions) without explicit (human) programming. In a broad sense, this is done through the use of historical data with known outcomes as input and identifying patterns consistent with certain outputs. 

Credit scoring, which is essentially predictive analytics in the form of probability of default scores, is extremely well suited to the application of machine learning.  By analyzing historical loan data where the borrower is known to be either a "good" or "bad" borrower, ML is able to develop a model that can predict the probability that applicant will turn out to be either a "good" or "bad" borrower from loan applicant data. This gives us a credit score which represents that probability in an understandable and operationalized manner. Seems simple, right?

The "Good", The "Bad", and The "Ugly"

The reality is that "good" and "bad" borrowers are only objectively defined at the extreme ends of super-prime borrowers that are virtually guaranteed to pay back loans on time and the correctly-scored ultra-deep subprime borrowers that are virtually guaranteed to default on the first payment. The business of subprime lending and sifting out Invisible Primes, however, is much more like the "ugly" when it comes to the big picture of identifying a profitable origination opportunity. 

What Makes "Good" and "Bad" So Difficult to Define?

There are a number of nuanced factors which make defining borrower categories difficult. Being mindful of these factors, and partnering with a knowledgeable decision management system provider, can help ensure that automating your scoring is done for the right targets to provide you with the most meaningful scores for your purposes.

  1. Internal Differences 

    Different people within your organization are highly likely to have different ideas as to what an ideal borrower/outcome will be, from your executives, to your employees. Oftentimes, this stems from qualitative understandings of "good" and "bad" borrowers as it relates to your overall business strategy. Good decision management system providers will work with you to better understand your strategic needs and develop data-driven understandings to help align your team.

  2. Dynamic of Time

    A fundamental piece of debt repayment is that payments are due at certain points. But which is preferable in this example: a borrower that consistently missed payments but eventually cures their payments, or a borrower that is highly consistent with on-time payments, until halfway through the repayment period when they default entirely? The answer to this question is ultimately dependent on your business structure and strategy. A good decision management system provider will make sure that the dynamic of time and its potential impacts are well understood in the context of your business, and will develop recommendations for alignment with their system.

  3. Point of Profitability

    Despite common belief, a loan does not need to be paid back in full in order to be profitable for your business. Depending on interest rates, payment amounts, and overall loan structure, a loan can be profitable past a certain point in the repayment period even if the borrower defaults.  With this in mind, sometimes purely profit-driven criteria are not desirable due to other business factors that are important to overall success. Good decision management system providers, however, will work with you to identify these points and optimize their systems with the best outcomes for your business in mind. 

  4. Asset Valuation and Depreciation

    Further on the idea of profitability, for asset-backed loans, such as auto loans, there is an additional complexity to projecting the value of the asset (a vehicle in auto lending). Accurately understanding the value of an asset, in the event that it needs to be repossessed, is incredibly important to determining a risk-mitigated and profitable loan structure. If you are an auto lender or another asset-backed lender, a good decision management system provider will understand the unique market you operate in and the nuances surrounding valuation, and pair that with optimized ML modelling outcomes to provide you with the best insights available even in volatile market conditions. For this subsector of lending, especially look for providers with subsector experience and knowledge. 

Contact us today to learn more about how Trust Science can help your business pick the right targets to capitalize on the ML-powered models offered in Credit Bureau +. 

Contact Us