Demystifying Non-Prime Auto Loans: How to Predict Performance with Ridiculous Accuracy

Demystifying Non-Prime Auto Loans: How to Predict Performance with Ridiculous Accuracy

When you think of non-prime auto loans, what are the first words that come to mind? Here are a few of my answers: Predictable. Reliable. Low risk. That’s right. Your eyes aren’t playing tricks on you. The traits you would normally associate with prime paper are now a perfect fit for the non-prime world.

Listen … I understand full well that the non-prime market segment has traditionally been seen as high risk. In fact, the unpredictability of non-prime loan performance has plagued the auto industry for decades. But that was the past.

Today, artificial intelligence and machine learning can now consistently predict the performance of non-prime auto loans with 98.5% accuracy. That margin for error is because our models have performed even better than expected. This timeframe of predictability includes a global pandemic, business lockdowns, sky-high inflation, and multiple interest rate hikes. These loans have seen a lot of curveballs – and yet they still perform as predicted.

Applying a fresh perspective and an open mind, non-prime auto loans can represent an enormous opportunity for the auto industry. This is truly an untapped market with big potential, made possible by data science and predictive analytics.

It’s expensive to be a non-prime buyer

While prime customers enjoy low interest rates as a reflection of their minimal risk, non-prime buyers find themselves on a hamster wheel of high interest rates and fees or even completely non-bankable.

Don’t get me wrong, there is a reason why non-prime loans have had such a questionable reputation. They’re more likely to default, and lenders must charge higher interest to account for the increase in risk.

That difference can equate to a monthly car payment that is several hundred dollars higher than their prime peers. Non-prime buyers may also need to make larger down payments or pay more fees, which can push paycheck-to-paycheck budgets to the limit. Throw in a medical emergency, job loss, or record inflation, and you’ve got a recipe for missed payments and the potential for repossession.

Ridiculously accurate performance is now mandatory

Thanks to inflation, everyone is paying significantly more for housing, groceries and gas than a couple of years ago. This reality hits harder as 60 percent of the United States population is living paycheck-to-paycheck, resulting in less disposable income in households. This pinch and the reality of economic struggles applies across the board, including lenders who must anticipate and adjust for these constantly changing factors.

Not being comfortable with the risk in non-prime, many finance companies have decided to decrease or even eliminate their non-prime business lines of business. Unfortunately, they will miss out on a very profitable market segment.

Want to know the prerequisite to successfully optimize the non-prime auto market? Predict loan performance with ridiculous accuracy. When you can do that, it doesn’t matter if inflation comes in hot or the hospital bills are piling up. Your portfolio forecast can account for these occurrences.

With a high degree of accuracy, you can be proactive instead of reactive to your business needs. Plan your cash flow instead of scrambling every 30 days, or worse, wonder if you can even make payroll. Get forecasts you can use to help plan your business instead of pie-in-the-sky projections you can’t trust.

Most of all, dependable predictions allow you to distinguish between the “non-prime customers who pay” vs. “non-prime customers who default.” Before this type of sophisticated modeling and segmenting was available, non-prime was just one big bucket and you took the good with a lot of bad. The “everyone is in the same bucket” traditional approach takes a big ax to segmenting and overlooks the scalpel-like nuances that can identify buyers that have a higher likelihood of consistently paying their loans.

I think we can all agree that the non-prime sector is a mixed bag. And as a result, finance companies miss out on a viable buyer who got into some surprise medical debt, has a stable job and home life, and will make every payment on time.

Just imagine, how would your business grow if you could consistently identify the ideal non-prime customer and decline the rest? It’s now possible, thanks to predictive analytics.

Deciphering what happened vs. what will happen

Conventional methods of predicting loan performance (i.e., credit scores) are based on what occurred in the past. This strategy served its purpose when it was the main evaluation tool available. But times have changed. You should care more about what will happen tomorrow than what already transpired in a customer’s life.

That’s why advanced data science blends the most helpful traditional data points with alternative data proven to predict non-prime auto loan performance. The result? You can precisely determine which customers will pay – and which ones you should confidently decline.

Predictive models that deliver 98.5 percent accuracy aren’t built out of thin air. At Agora Data, for example, we’ve back-tested more than $300 billion of non-prime paper to develop our predictive models, and we regularly adjust the inputs based on out of the ordinary incidents. This strategy is a new approach. So if you currently base your risk assumptions primarily off of credit scores, this adjustment will feel like moving from a typewriter to a computer. It’s a black-and-white methodology changed to full technicolor.

Here are just a few pieces of qualitative data proven to predict a customer’s future payments:

  • What do they do with their money on payday – cash the check or deposit in a bank account? Maintaining a bank account is a sign of financial stability.
  • Do they stay current on necessities like utility bills? This provides insight into consistency and reliability.
  • How often do they change their phone number or email address? Getting a new one every six months may be a clue that the customer is avoiding creditors.

Even something as minor as the battery life on their phone screen can also help predict ability to pay.  Lack of a consistent routine and where they charge their phone daily can be in interesting indicator of loan performance if their battery life dips quite low.

Alternative data and the right mix of behavioral data combined with conventional data points like credit score, amount financed, and vehicle mileage can paint a revealing picture of a buyer’s potential loan performance. You can start to see with clarity which non-prime buyers you should do business with. You can also better pinpoint the risk each buyer brings to your business and structure their loan appropriately.

Double your potential customer base

Financial institutions that ignore the non-prime segment for fear of how the loans will perform are passing over nearly half the country’s car buyers. It makes sense to avoid the non-prime sector if that paper was actually unreliable. But once you know how to finance non-prime loans with prime predictability, the uncertainty is gone.

I can’t emphasize enough what a paradigm shift this approach is having on auto finance. Just mentioning the word “non-prime” automatically change the tone in a room. Today, I live in a world where non-prime is simply one of several loan tiers. Non-prime auto loans are now capable of the same accurate forecasting you can achieve higher up the credit spectrum.

This potential doesn’t just apply to auto finance. Predictive analytics can accomplish the same for home mortgages, insurance, and more. As a professional data scientist, it is thrilling to watch how machine learning and artificial intelligence are completely revolutionizing multiple industries. It’s also refreshing to see the non-prime customer segment destigmatized since so many of those loans perform well. And now we can demystify non-prime and identify those that deserve credit for their auto purchases.

A new era for used car dealers

With this knowledge, I expect to see more used car dealers develop long-term, mutually beneficial relationships with their customers. What was once a repossession merry-go-round has become a recurring revenue machine producing consistent monthly payments, repeat buyers, and never-ending referrals.

Auto finance innovation is ushering in a new era for the used car industry. The most profitable dealers will have the most successful customers. Repos will be rare. Dealers will originate for long-term performance, and predictability will be the name of the game. No doubt, data science will become further integrated into our business models, making the potential for growth greater than ever.

Read the original article published by using this link.

UCD Magazine is published by NIADA.

Start Making More With AgoraCapital

Get access to unlimited, low-cost capital to use any way you want. See if AgoraCapital would benefit your business.

Leading-edge analytics gives you leading-edge power

Sign up for a free account and gain access to our state-of-the-art auto loan portfolio analytics tools.