.png)
The US mortgage industry has just been through the best 18 months in terms of volumes and profitability in a decade. As we head into 2022, the old issues of managing expenses and scalability come back into focus. Many firms are already selling servicing assets to raise cash, a sure sign that profit margins are getting tight. Other firms are buying servicing at record prices to perhaps capture refinance opportunities. Successful firms must find a way to achieve cost savings in heretofore sacred parts of the mortgage factory such as loan underwriting, which is the most costly and also important part of the lending process. But every loan is different. A typical loan has literally tens of thousands of variables that an underwriter must navigate on the pathway to approval. Most firms have error rates for loan underwriting in low single digits. To be successful in 2022 and beyond, mortgage lenders must move that error number a couple digits to the right of the decimal point. In this issue of The Institutional Risk Analyst, we speak with a revolutionary in the world of lending, Tom Showalter, Founder and CEO of Candor Technology.
The IRA: Tom, you started Candor back in 2017 to solve several problems with the loan manufacturing process. You identified rising costs and shrinking margins, a high error rate in underwriting with the potential for bias, and as a result, a poor experience for the consumer. Talk about how you came to this vision.
Tom Showalter
Showalter: Prior to joining the mortgage industry, I worked for NASA in the aerospace industry. The mortgage industry has huge amounts of complexity and spends aggressively on technology, but has yet to see that technology spend returned in the form of increased productivity and profits. I began to think about how we can transform the process of manufacturing a loan into something closer to the zero-error tolerance world of aircraft and space travel.
The IRA: That sounds like a heavy lift for an industry that has spent a decade retooling after the 2008 financial crisis. Despite the overlay of robotics and automation, lending remains a very labor-intensive manufacturing model more akin to Henry Ford in 1910 than Boeing Company (BA) today.
Showalter: At the heart of the problem lies something very fundamental. The integrity of the mortgage asset, the loan, has been suspect since the 2008 meltdown and is still suspect today. We will have 100% due diligence on private label mortgage deals. The only ways of increasing the loan integrity are manual, which means slow and expensive. In our analysis, it was the lack of integrity in the mortgage loan that we sought to correct. We thought, if we could find a way to use a machine to dramatically upgrade the integrity of the mortgage loan, we can achieve gains in speed, productivity and profit levels, but only if the loan produced by our methods met the highest standards of integrity. We viewed this problem at two levels. One was to improve the integrity of the data used to manufacture a loan. Secondly, using that high integrity data, we would ensure that the loan met all guidelines, ensuring a high level of decision integrity.
The IRA: You’ve stated that until Candor, technology solutions in the mortgage industry did not have a meaningful impact on cycle time or net profit per loan. The aerospace industry has made improved asset efficiency and the avoidance of error the rule going back to WWII. How do we make similar progress in mortgage lending?
Showalter: What we’ve learned is that once we taught a machine how to ensure the integrity of the data and the decision, the mortgage origination process became incredibly fast, shrinking loan origination cycles. Lee Smith from Flagstar Bancorp (FBC) told you back in August that “I don’t think we’ve seen anybody really come to the fore in terms of growing market share based upon technology.” That is another way of saying that mortgage firms basically have no upside in terms of operating leverage. The industry has spent billions on robotic systems that automate what is still a largely manual process. Loan underwriting is too complex, too variable for such an approach. That’s why we decided to apply the basic tools of behavioral science to underwriting mortgage loans.
IRA: Talk about the specific tools employed in the Candor process to get the task to completion.
Showalter: At the heart of the aerospace industry approach is its reliance on engineering and mathematics to craft the technology necessary to advance its fortunes. Just as you wrote in your “Ford Men” book, Ford Motor Co (F) embraced assembly lines to meet demand for cars. Aircraft manufacturers had to meet demand for civilian and military aircraft, but with zero tolerance for error. Two forms of technology leveraged constantly by the aerospace industry (and not leveraged by the mortgage industry) are “expert systems” and “system safety engineering.”
The IRA: Never heard those terms used in the context of residential mortgage lending. Talk about what this means. What is the aerospace industry doing that the mortgage industry is not?
Showalter: “Expert systems” is a technology where you teach a machine how to think like an expert. Expert systems are used by the intelligence community to replicate the thought process of the 95th percentile intelligence analyst. They are used by the medical industry to replicate the thought life of a diagnostician and by aircraft manufacturers to replicate the diagnostics needed to repair a jet engine. “Systems safety engineering” is a design methodology supported by a special form of mathematics that enables an aerospace engineer to design a very complex system, one with over a million interacting components and ensure that it operates flawlessly for hours, days, weeks on end. Used in this context, “expert systems” means the art of teaching a computer how to think like a particular kind of expert, such as a loan underwriter. With an expert system approach, for example, a less seasoned team can produce a fully underwritten file without an underwriter, then present the loan file for approval with a complete history of all of the data and decisions in that process.
The IRA: Traditional systems and controls in the world of banking and consumer finance are focused on managing decisions by people. How is your approach different?
Showalter: Using technology to define and limit the lending process results in greater efficiency, with more loans, fewer errors and, importantly, less opportunities for lender bias. We not only make the underwriting process more disciplined, but we are able to improve initial to final approval, pull through and deliver a significant increase in asset turns via a decrease in time to funding. In people-based systems, machines do a portion of the data aggregation and analysis, but people do the final analysis and make the decisions. This also creates a huge bottleneck. There are too few people with high levels of final analysis and decision-making skills.
The IRA: Better asset turns are always good. We wrote favorably about Western Alliance Bancorp (WAL) earlier this year when they acquired AmeriHome from Apollo (APO). These folks understand asset turns. WAL was the best performing large bank so far stock in 2021. Mortgage lenders would certainly welcome ways to make the underwriting process more efficient. Is it really possible for Candor to improve operating leverage in areas such as underwriting?
Showalter: Our expert system simulates the decision-making ability and high-level cognitive tasks of human experts – but without human bias. Humans approach tasks by using a narrative, a model of past loans, that is used to assess a new borrower. And again, this is not robotics or “AI” but rather a way of engineering a very narrow and specific decision process.
The IRA: That sounds impressive. How does this work in practice as part of a loan origination system (LOS)?
Showalter: Think about how we underwrite a loan today. There is tremendous complexity in the process. Every loan has a unique combo of 1003 loan application information + supporting data. As a result, each loan begins at a different point of departure. The underwriter can’t follow the same process to solution. When new data & information refute previous conclusions, underwriters must pivot and begin the underwriting process all over again in real time. Candor automates ~ 90,000 real time micro decisions on every loan. By using the machine to perform these tasks, we speed the underwriting process and leverage your most valuable people, namely the experienced underwriters, in reviewing and providing a formal approval.
The IRA: How does this application of decision technology address the issue of loan quality? At present we are in a period where credit risk appears to have no cost, as in 2005, but we all know that after a period of high lending volumes credit problems appear. The GSEs and FHA start making loan repurchase demands due to defects. How does you address this issue of loan quality?
Showalter: In 2019 and 2020, when volumes were at the highest levels in history, technology could not handle the load. The technology of the day could not provide the critical thinking necessary to accomplish the underwriting challenges of record volumes. Technology deficiencies forced lenders into adding droves of critical thinking talent (e.g. underwriters, loan processors). Lenders saw a large portion of the profits go to paying exorbitant salaries as the industry drained the labor pool of underwriting talent. Loan manufacturing slowed considerably, costs grew and treatment of borrowers become less consistent.
The IRA: In other words, we increased the potential for error as volumes grew? Sounds like the early 2000s.
Showalter: By embracing decision technology, we enable lenders to perform a more complex transaction, faster, less expensively and with much greater consistency. We believe this new approach will produce loans that the secondary market would be happy to buy and, indeed, will prefer over time. Just as investors shy away from MBS with high default or prepayment rates, we believe that investors and guarantors like the GSEs or FHA will gradually come to favor loans created with technology like Candor.
The IRA: That is a pretty bold statement. Why are you so confident that applying the technology used in the world of aerospace can truly change the mortgage industry?
Showalter: In the past, one industry, the aerospace industry has been asked to do something of immense complexity, to do it safely and to accomplish things previously thought impossible like putting a man on the moon, providing extremely safe commercial jet travel, and supersonic flight and even space slight. The aerospace industry had to invent the technology necessary to manufacture and fly the planes and space craft necessary to accomplish these most difficult objectives and do it safely. The aerospace industry successfully developed the technology that would transform it from one of the riskiest on the planet to one of the most reliable. Aerospace technology can produce a plane that contains over 1 million interacting components, all of which operate flawlessly. To accomplish this miracle, aerospace technology has not only made aircraft and spacecraft extremely reliable, it made these machines smarter. We have to make the mortgage lending process smarter and less biased by human limitations just as the aerospace industry did with aircraft and space travel.
The IRA: Enabling lenders to leverage parts of the process map such as underwriting would bring some truly revolutionary changes to the industry. Is this revolution really possible in the world of residential lending?
Showalter: I believe so. Decision technology can be applied to any type of credit creation. It is possible to port aerospace technology into the mortgage space and do so without requiring the industry to completely rethink and reinvent itself. What the mortgage industry needs is a technology that will embed the critical thinking skills necessary to make and underwrite a mortgage with the same degree of perfection that we see in the manufacture of complex machines such as aircraft and space vehicles. We can do this.
The IRA: Thank you for your time, Tom.
