“Does it make sense?”
That’s the mantra and ultimate question Candor Technology Inc. seeks to answer with their highly advanced automated underwriting technology. Former NASA Rocket Scientist and current CEO of Candor Technology, Tom Showalter, applied his skills and knowledge to solve long-standing problems in the mortgage industry, such as capacity issues and lack of ample talent in the underwriting sector.
The pandemic stirred up a housing frenzy, with interest rates dipping to historic lows and demand skyrocketing to new heights, which caused pipelines to bulge and underwriters to work at full speed to push loans through. Typically, mortgage companies will hire hundreds of new underwriters when the housing market heats up — many of whom are unseasoned and eventually get canned when the market cools back down. Such is the way for many sectors of the mortgage industry, which is why many companies look to automate wherever they can.
“We need thousands of new underwriters now and it takes five years to just make one,” Showalter said. “Underwriting is naturally prone to human error, but it happens more frequently among inexperienced underwriters. If an error is found, the servicers need to conduct a quality control re-underwrite, which costs hundreds of dollars and prolongs the closing time. There will always be a need for confidence in the way data is verified.”
Mistakes or errors in underwriting can also lead to loans getting wrongfully denied or wrongfully accepted, which stirs up major problems for investors in the secondary market. Nearly all loans (97%) are sold as mortgage-backed securities in the secondary market, where they get packaged and sold to international investors. These international investors are promised a specified interest rate and return of principal after a certain amount of time. They’re not concerned with the particulars of U.S. mortgage politics, such as moratoriums, forbearances, or the pandemic’s impact on the domestic economy. Someone needs to fund these shortfalls if these loans default, and this burden typically falls on the shoulders of mortgage servicers. For this reason, it is important to be confident that these loans are eligible for investors to purchase.
“Investors need to be confident in the bonds they’re purchasing and our technology has the ability to vet these loans more thoroughly than any human underwriter,” Showalter said. “We ensure cost reduction, speed, and integrity with all loans verified through us.”
The first product Candor launched in September in 2020 was the Loan Engineering System — a loan scoring machine that thoroughly performs the duty of an underwriter by conducting cross-checks, pin-pointing defects and mitigating them. So far, the company has been successful in onboarding over 50 clients within their first year of launching.
Specifically, the loan scoring engine operates by running the data through a 45,733 pivot point architecture to assess investor guideline eligibility, then through a corroboration engine that conducts over 1,100 cross-checks to identify and mitigate defects. By the time the process in complete, only one underwriter needs to review the loan for conditions before it's sent out to close.
In comparison to the manual underwriting process, underwriters complete about 2 to 3 loans per day, conducting 32 cross-checks for each loan. However, the initial tests of the Loan Engineering System were able to conduct 10 to 15 loans per day and reduce the cost to originate by $1,600.
The machine also has a newly integrated Metascore system that will score the confidence level for every loan that’s been manufactured by the Loan Engineering System. A Metascore of 85 or higher trigger eligibility for investors. Moreover, all the metadata used during the autonomous process is saved to a Blockchain-type database providing unprecedented transparency for investors.
“Our first client came to us with a fair amount of tech skepticism,” Showalter said. “He essentially wanted to know if the Loan Engineering System worked as promised, so he challenged us by giving us 38,000 loans to underwrite overnight… and we got it done.”
Prior to the founding of Candor, Showalter and his current chief marketing officer, Brandie Young, worked together at Digital Risk to conduct forensic underwriting and understand why most loans failed during the crash. “Most of the time it was a failure in the manufacturing process and the fact we’re reliant on underwriters,” Young said. “An inconsistent manufacturing process can be remedied with AI, but that term is also very broad. It’s essential to choose the right kind of technology for the process you’re looking to emulate or improve.”
“AI is a buzz-term that’s been floating around the media for some time,” Young continued. “For example, machine-learning technology is another kind of AI technology and a buzz-term people have been using for some time. But, that isn’t a viable solution when it comes to underwriting loans. Every loan that comes in is different and takes its own unique path to becoming verified, but machine-learning technology is conducted through pattern recognition. Only so much can be repeatable because loans vary so much, so that wouldn’t be helpful with pin-pointing unique defects.”
Model expert thinking is the right technological solution for underwriting loans, according to Showalter, who has been studying and developing this kind of technology since the 1980’s when he was building complex space programs for NASA. “We wanted to emulate the skills of a seasoned underwriter, or in other words, think like a seasoned underwriter,” he said.
Over the past few years, AI technology has been introduced and touted by a number of mortgage software companies in the industry, but not at the level of complexity as Candor’s Loan Engineering System. “You need to know aerospace principles and incorporate model expert thinking which is extremely complex and very costly,” Showalter said. “In terms of competition, no one else in the industry has what we have.”
Showalter estimates it took $25 million merely to get the Loan Engineering System to function as promised. “We were angel and mainly self-funded until we got a series A for $12.5 million. We launched with a million lines of code that are still being added to with our monthly sprints, so the technology is constantly developing. We’re able to make it affordable, though, because it runs at a zero variable cost,” Showalter said.
As of now, Candor can conduct 110,000 intelligent acts — including edits, updates, additions, calculations, verifications, and validations — within 7 seconds. Candor is the only solution able to autonomously conduct nuanced cross-checks across a vast and complex set of dynamic data. These painstaking corroborations exceed guideline requirements, setting a new standard for underwriting in the mortgage industry.
Asked if these model expert thinking solutions could assist the appraisal industry in the midst of modernizing, Showalter said, “Yes, I believe our kind of technology could potentially help appraisers conduct and analyze home inspections, especially for appraisal underwriters.”
Win silently also seems to be the motto for this company as they gather heavy hitter clients and potentially become the next ‘unicorn’ (billion-dollar startup) of the mortgage industry. However, CEO and founder Showalter humbly stated, “We don’t subscribe to those terms, necessarily. I will say we are unique to the industry, bringing extremely valuable products, and right now we have a significant valuation.”
"There will always be a place for seasoned underwriters," Young added. "Their roles will change as technology becomes more integrated in what they do, but this will allow them to ask more critical questions about the loan rather than do the basic cross-checks. It will become a better, higher-value job for those already in the industry."