Transforming Mortgage Banking

When Rajesh Bhat and his wife tried to buy a house a few years ago, and then later refinance it, it was a stressful and confusing series of processes, including an unexplained $2,500 check in the mail after closing and a surprise mortgage broker’s fee. Bhat decided to do more than just survive that “sufficiently traumatic experience” of buying a home. Following along the lines of any good Silicon Valley entrepreneur, he decided to create a startup that would transform the experience.

An early version of Bhat’s startup, Roostify, focused on the real estate transaction-but that effort “flopped,” says the CEO and co-founder of the company. San Francisco-based Roostify later evolved to provide an online mortgage application for lenders instead. The company sells the platform to about 50 lenders, including JPMorgan Chase & Co. and Guild Mortgage, and it facilitated more than $10 billion worth of mortgages last year. Roostify has reduced the time it takes to fill out an application from about two hours to 20 minutes. Meanwhile, from its headquarters in Detroit, Quicken Loans offers an entirely online application with pre-approval in as little as eight minutes and remote closing in some states via video conference.

Digital technology is significantly changing the mortgage industry, making the process of obtaining a mortgage quicker and easier. Financial technology startups and longtime lenders alike are making huge improvements to the online and mobile application process. But when it comes to simplifying the underwriting, valuation and closing process while reducing costs for the lender, the home mortgage-one of the most complicated transactions in modern finance-has proven difficult to transform. “When you think about the total power digital could do, it’s probably not going to move at the pace some of us would like,” says John Newlin, a managing director at the consulting firm Accenture.

The mortgage industry could use some help. The average production cost of a home loan rose to $8,082 in 2017, up 35 percent from four years earlier, according to the Mortgage Bankers Association (MBA). That has been driven largely by increased regulations and falling refinancing volumes as interest rates rise. As volume falls but production costs remain high, that makes the per-loan cost higher. Mortgage lenders’ profits are getting squeezed. “They are really struggling right now,” says Marina Walsh, vice president of industry analysis for the MBA.

Banks, beleaguered by regulation from the Dodd-Frank Act and mortgage-related lawsuits and fines that followed the financial crisis, have been losing market share to nonbanks such as Quicken Loans and San Diego-based Guild Mortgage. In 2016, for the first time, nonbanks originated more mortgages by loan value than banks and made up 12 of the top 20 mortgage originators, according to S&P Global Market Intelligence. Quicken Loans became the largest mortgage originator in the country in the fourth quarter of 2017 and the first quarter of 2018, surpassing the previous top originator, Wells Fargo & Co. Nonbanks have grown so much, they handled about 75 percent of Federal Housing Administration (FHA) loans in 2015, according to a March 2017 study by researchers at Columbia University, the University of Chicago and Stanford University. The researchers found that the growth in nonbank lenders was largely caused by regulation and secondly, by a technological advantage that increased convenience for consumers.

Part of that convenience is the online mortgage application. J.D. Power found that 47 percent of nonbank borrowers in 2017 said they completed a detailed application online, versus 37 percent getting a mortgage from a regional or big bank. “What does Guild do? What does Quicken do? It’s mortgages,” says Craig Martin, J.D. Power’s senior director of wealth and lending. “That’s it.” Although getting a mortgage quickly is far from the only thing borrowers want, it’s one area in which fintech lenders excel. In a separate study published in February 2018 by researchers at the Federal Reserve, fintech companies, which don’t have a branch network or deposits, are shaving the time it takes to get a mortgage down by about 20 percent, cutting the process by about 9.2 days for a purchase loan and 14.6 days for a refinance, all without offering lower interest rates or experiencing higher default than traditional lenders.

Some of the nonbanks are financial technology companies that lend directly to consumers, such as SoFi, LendingHome, Clara and Lenda. LendingHome, for example, originated more than $2 billion in loans in 2017 for real estate investors and homeowners, building its system from the ground up with the latest technology. There was no need to tap into legacy loan origination platforms that are decades old, says LendingHome Executive Vice President Josh Stech.

Other nonbanks lenders that are speeding up the digital mortgage are some that have been around for decades, such as Quicken Loans. The company was keen to transform the mortgage application at a time when other lenders were simply struggling to comply with new mortgage regulations stemming from the Consumer Financial Protection Bureau. About six years ago, company executives sat in a room pinned with paper on all four walls and painstakingly wrote down a step-by-step process of what it’s like, from a client’s perspective, to get a mortgage, says Regis Hadiaris, a senior executive at Quicken Loans. It didn’t look pretty. After years of work, the company launched the Rocket Mortgage product in a media blitz and Super Bowl ad in 2016. Now, almost the entire mortgage process can be completed online or on a phone.

Rocket Mortgage customers can upload documents online, or they’re prompted to save time and temporarily allow access to bank and investment accounts to pre-fill their application with income, asset and employment history through third-party data providers. The application can even pull a customer’s salary and bonus. Quicken says it can connect with 98 percent of the financial institutions in the country this way, and it has access to the employment status of about half of the customers who apply.

Applicants are prompted with a series of automated questions to navigate a “choose your own adventure” style mortgage, with recommendations for particular products such as an FHA or a VA loan, Hadiaris says. The customer can test out different options and rates by clicking through them. You can learn, for example, what happens to your payment if you finance your closing fees or switch to a 15-year fixed rate mortgage, questions normally answered by a loan officer. Each screen lists definitions of each product and term. If the customer isn’t ready to make a decision, they can save their information and come back later to lock in a rate online. Customers who want help can click a button that leads to a phone call or an online chat. In some states, consumers meet settlement agents in a coffee shop or at home. In a handful of other states, they can close remotely via video.

Quicken is privately owned and doesn’t disclose its advertising budget, but spends heavily. Unlike a bank, which might simply promote its rates, Quicken has Avengers characters blowing up stuff in the background of its advertisements and comedian Keegan-Michael Key explaining mortgages. The culture inside the organization also is key to its success. Quicken has a 136-page training guide filled with aphorisms such as “every client, every time, no exceptions, no excuses,” and practical examples of what those aphorisms look like in real life. Employees commit to answering every phone call and email on the same day, for example. The walls inside Quicken’s Detroit offices are brightly colored and the executives have the enthusiasm to match. When Terry Wakefield, a consultant, visited the lender a few years ago, he wondered if he had walked into a mortgage company “or a high school pep rally.”

So while it’s clear that Quicken has become the No. 1 mortgage lender for more than just its technology, banks feel compelled to keep up with the whiz bang of nonbanks like Quicken and others. Several fintech companies, including Roostify, CloudVirga and Blend, don’t compete with banks but sell their application platforms to banks instead, which speed up the process of getting a mortgage in a similar way to Quicken Loans, says Celent Senior Analyst Craig Focardi.

The platforms use automation and application programming interfaces to tap into data providers such as Yodlee, Plaid and FormFree to verify income, assets and employment and then send the application information directly into the bank’s loan origination system. Buyers of those loans, including government-sponsored agencies, are growing more comfortable with the accuracy and reliability of those verification systems, which cut down on the work of making phone calls and checking income and employment. For example, as part of its Day 1 Certainty program, Fannie Mae certifies a list of vendors that offer automated verifications, including Blend, providing representation and warranty relief to lenders who use them. Basically, lenders that use those vendors with Fannie Mae’s underwriting technology and guidelines don’t have to worry about having to buy back loans because of inaccurate verification.

San Francisco-based Blend’s platform is gaining credibility with banks. It handled about $60 billion worth of applications last year, and estimates that it provides a platform for about 25 percent of the mortgage market, including large lenders such as Wells Fargo and U.S. Bancorp. Over time and with a huge data set, the system gets better at offering products that more precisely fit the customer’s needs, a process called machine learning, says Brian Kneafsy, Blend’s head of client operations.

But not all banks are keen on using third-party platforms to serve their customers. David Doyle, product executive for Bank of America Corp.’s new digital mortgage that was rolled out this year, says executives didn’t want to be dependent on a third party or have the bank’s loan application become a commodity. By building its own platform, the bank was able to integrate its digital mortgage with other product offerings, such as rewarding mortgage customers who also are Merrill Lynch customers or part of the bank’s Preferred Rewards program, where they get perks in exchange for average combined balances of $20,000 or more. But Doyle admitted that creating the new digital loan platform was only possible because the bank upgraded its loan origination system from the 1980s in 2016. “[The previous loan origination system] couldn’t come close to supporting a digital mortgage,” he says.

Still, the industry hasn’t adopted a fully digital mortgage. The online application is only one part of a complicated transaction involving multiple parties and reports, including title insurance, flood reports, property appraisals and closing agents, which is why it still takes on average 40 to 55 days for a lender to close on a purchase loan, according to the MBA. The transformation of loan closing to a remote, fully digital process is hampered by the customs of settlement agents and state-by-state laws regarding notary practices. In New York, for example, a lawyer must be present at closing, says Focardi.

“The front-end solutions help streamline the application process, but that’s only the first 20 minutes in a 45-day process,” says Joseph Tyrrell, executive vice president of corporate strategy at Ellie Mae, which sells software and loan origination service platforms to the mortgage industry. The opportunity to create the most efficiency is really in the automation of back-end loan processing, but that’s easier said than done. Tyrrell estimates that automation could eliminate $2,600 from the average $8,000 that it costs to originate a loan. In automation, the only time a person should have to look at a loan is when there is an exception to a rule, he says. But lenders and their technology providers have been sidetracked by increasing regulation and haven’t devoted a lot of resources to automation during the last few years. “It’s really all about automation and leveraging data to a much greater extent than the industry has,” he says.

Focardi also identifies automation as one of the most significant opportunities for the industry right now, using technology such as optical character recognition to read data off documents and eliminate data entry errors or the necessity of someone having to input information. “Only a few companies are doing it end-to-end from loan application to post closing,” he says. Guild Mortgage, for example, is partially using optical character recognition to read documents coming back from settlement attorneys and classifying them automatically. If someone changed the document, the system will recognize and flag that. No one has to re-read 96 documents before closing, in other words.

Stech at LendingHome estimates that the process of automation could cut the cost of a mortgage in half and shave three to four weeks off the process. LendingHome, which was founded in 2013, is working to build its own automated “decisioning” models and says loan buyers and government agencies are beginning to come around to the idea that loans that meet specified risk profiles, such as low loan-to-value ratios, could be approved by a computer.

Wakefield, who has built digital mortgage processes for lenders, says 70 to 80 percent of loans could be handled by a computer underwriter. The typical mortgage requires about 350 to 400 human tasks, but could be reduced to about 120, he adds. But Wakefield has met resistance to his recommendations that the entire process must be revamped from beginning to end. “Mortgage loan originators don’t accept the fact that a human doesn’t need to look at this document,” he says. He has since given up on consulting for the digital loan industry, tired of the feeling that he was in the desert preaching “that the prophet is coming.”

Persuading an industry that experienced the financial crisis 10 years ago that a computer can handle underwriting is no small task. “[Underwriters’]sense of extreme accountability makes it so they want to check the checker,” says Gabe Minton, executive vice president and chief information officer for Guild Mortgage. “That’s not bad.”

The value of the data and the parameters going into the underwriting decision will influence the quality of that loan. “If the source data is wrong, that increases the risk,” says Focardi. It’s important also for banks to use fraud detection verification systems. But digitization of data offers clear improvements from a risk perspective. “Most lenders are thinking of [automation] as something that will increase data quality and reduce risk,” says Newlin. With so many parties involved in a mortgage purchase transaction, there is always the threat that documents at closing don’t match what was agreed upon. Digitization helps solve that problem. “You can’t create a document that’s out of sync with the data,” says Rick Hill, vice president of industry technology for the MBA. “That gives a level of comfort to everyone downstream that they can trust what’s there. We’ve had checkers checking checkers. Those inefficiencies will start to go away.”

The technology raises difficult questions about what to do with the checkers who lose their jobs, and other employees replaced by digitalization. Some are optimistic that the industry will retrain employees for other jobs. Whether or not that happens, competition may provide answers of its own for companies that don’t adapt. “The transformation that’s underway is phenomenal,” says Hill.