What’s the key to effectively automating the mortgage process, from application through post-closing?
According to David Sohm, chief operating officer for mortgage document and data firm Capsilon, the process has to start with accurate and verified data.
Lenders that have deployed Web-based portals through which borrowers can upload applications and critical loan documents have taken the first big step toward collecting the necessary data to automate. Once the documents are uploaded, document recognition software, such as Capsilon’s DocVelocity product, can be used to identify the documents and properly “name” or index them for the lender. From there, a data extraction program can be used to extract just the data that is needed from the forms in order to complete the rest of the process. That data can then be verified using either a manual process or automated verification software.
As Sohm tells MortgageOrb, the most advanced data extraction systems today scan only the pertinent fields in the forms. The software, he says, uses optical character recognition (OCR) to “read” the data – but more importantly, it “knows” where to locate the pertinent data on the form through a process known as document mapping. Extracting just the needed data, he says, is much more efficient than running OCR on the entire document, extracting all of the data and then tossing what’s not needed.
Once a lender has extracted the data and it has been verified and compared against the original forms, the original forms can then be securely archived and, in theory, should basically no longer be needed. Sohm, however, points out that because most lenders today have not gone fully digital with their mortgage processes, the original documents are usually reviewed and compared during the post-closing audit process.
Still, lenders that are using a mostly digital mortgage process today have a distinct advantage over lenders that are using traditional, paper-based processes in that they can leverage data for “straight-through processing” of mortgage loans.
At this point, Sohm says, there’s no question that technology and automation are the key to reducing pre- and post-close audit times and smoothing out the entire mortgage process.
“As you might guess, we see automation as a path to becoming more efficient and cost-effective,” Sohm says. “We see automation as the key to solving many problems, to reducing errors. It’s not that people don’t want to do a good job – but things do sometimes get overlooked or done incorrectly when you have manual processes.”
“For example, wholesale lenders have a unique problem in that they are getting loan files from a lot of different companies,” Sohm continues. “So, they can’t always control the differences between all the documents that they’re getting or what people are submitting. We see the ability to gather those documents, name them, extract the data and compared it, and then flip that back to the broker with a list of what’s missing. To do that quickly is a real advantage in the marketplace.
“The automation of that process speeds up the process of getting loan files to the wholesale lenders,” he adds. “[Looking at today’s manual processes], it’s not hard to see how [they can be automated]. Lenders have a checklist that’s on their desk – but they have to get around to looking at the loan file, checking to see if the documents are named the same thing or whether they got the right document but the lender didn’t provide the correct full name. Being able to sort those things out of the way quickly is a real advantage.”
As Sohm points out, many lenders today “say they are fully electronic, but they’re really dealing in scanned documents – they’re sending the docs in PDF form and they’re saying, ‘We’re all electronic.’” Although these lenders still need to make the final leap to become all digital, that doesn’t mean they cannot extract the needed data from the scans they are receiving and create a demarcation point in the process where it becomes entirely data-driven.
“You don’t want to work with the documents throughout the entire process,” Sohm says. “You want to get data from the documents that has full integrity and which can be put through a data-driven process.”
Where the use of data from the start of the process really pays off, Sohm says, is in the time saved during post-closing audits.
“We’ve been talking about the gathering of data so that the processes can be automated for file intake or submission of loan packages – but as we move into the post-close audit process, that’s where the data really becomes important,” he says. “Again, the documents aren’t so important in the auditing of the loan. What’s important is the data that the underwriter used – and knowing where it came from.
“One of the problems that often happens is that [a lender] will get multiple applications, and it’ll [accidentally] pull a piece of data about income from one application and a collateral address from another,” he says. “That’s a really bad thing that could pop up at a post-close audit. By getting the data verified up front, and through automation, these types of errors can be practically eliminated.”
Not only does having the data help speed the process of auditing a loan package, but it also enables a lender to audit more of its portfolio for compliance and quality control.
“One of the problems in the post-close audit today is that it is very labor-intensive – and what that leads to is sampling,” Sohm says. “Only 10 percent or 20 percent of the loans are audited, post-close. Well, we don’t think that’s good enough. Lenders should be able to do 100 percent audit throughout the process. And the way to do that is to present the people who make the decisions with only the data that they need to make a decision on. For us, it’s figuring out what are the automation steps that will allow lenders to reduce labor – which, by the way, doesn’t mean eliminating people – and gaining new efficiencies.”
Sohm says not only is it possible to audit 100% of the loans in a package, but it’s also possible to do it using a government-sponsored enterprise checklist.
“Again, the foundation of being able to do this type of work is data integrity, knowing where the data came from and knowing that it is internally consistent within the loan package, and the rest is just details,” he says.
Interestingly, just as much as data presents new opportunities to automate the mortgage process and gain new efficiency, it also present lenders with a bit of a quandary: What do they do with the data after it has been extracted? He points out that even once most lenders know what they want to do with the data, “the hard part comes in figuring out how to do that.”
“If a lender can extract the data and validate it, then it knows that the data it has with a document is correct,” he says. “But the problem is the consuming of that data. We’re finding that lenders have different IT skills, and many of them have a difficult time – not so much knowing what to do with the data, but how to do it.”
“That’s the biggest problem: ‘Now that I have the data extracted from the document, how do I use it?’ In general, it resides in the LOS – but it takes a lot of thought to figure out the best way to put it to work.”