Mortgages Are Data-rich but Insight-poor

This is the final post in a five-part series explaining that, at Blend, we work on mortgages because they’re HARD: Human, Astronomically Large, Regulated and Data-rich (Insight-poor). Check out our first, second, third and fourth posts.

One of the most pressing problems that we are tackling at Blend centers around the sheer volume of data that goes into the creation of a mortgage application. The average mortgage application today consists of over 400 pages of raw documentation, covering personal details, asset information, sources of income, employment history, and much, much more. This information usually takes at least a week of painstaking effort by the borrower to locate, retrieve, and consolidate; and then transfer to the lender via phone, email, snail mail, fax, and sometimes even in-person when the other methods fail. And the worst part is that much of this information does not even apply to the average person seeking a mortgage! This process has been so historically painful that it is commonly referred to as a “financial colonoscopy.”

Making the process even more difficult is the fact that there are thousands of data points that are trapped on a sheet of paper or in a PDF; data that can hardly be leveraged effectively. Currently, the process of originating a single mortgage involves days, weeks, and months of human labor required to sort through individual elements of data. Lenders are looking at asset statements to see if there are any transactions that require explanation, tax returns to determine your ability to repay your obligations, and credit reports to understand your ability to manage debt, just to name a few. This process takes dozens of humans, hundreds of hours, and thousands of dollars; all to originate a single loan. And as with any human-centric process involving reams of documents and data?—?there will also be errors and delays due to staffing levels, limited working hours, sick and vacation days.

At Blend, we aim to change this. We can dramatically improve the ‘data-rich’ element of the mortgage application process while also driving key workflow improvements and tangible insights for both the borrower and the lender. These improvements will help borrowers spend less time collecting documents and data and more time focusing on the important stuff: getting their perfect home and a loan that makes sense for them, while using technology that empowers their lenders to originate better, more efficient mortgages.

1-fxh12d5mopg9p3jgyqojcw

Tap, Connect, Done

At Blend, we have an internal philosophy about letting borrowers use their data to get a better mortgage. It’s called Tap, Connect, Done. The principle here is that the majority of data fields on a loan application can be filled automatically using borrower-provisioned data. By employing this principle, the time it takes a borrower to complete an application and provide required documentation is significantly reduced. By importing financial statements directly using online banking credentials, tax statements using turbotax, HR Block, etc., and pay stubs using ADP or Gusto, the borrower is able to collect almost all of their application documentation through Blend in a matter of minutes, and the time it takes to complete an application goes from hours that might span days or weeks to under 20 minutes that can be completed in a single sitting. Movement Mortgage, a large lender that is using Blend to power its originations operation, recently saw a 76-year-old borrower finish his entire application in Blend in just over an hour?—?a process that would have previously taken well over week.

By gathering data directly from source institutions, we create “living data” by extracting data from documents, and allowing automated rules to be run against this data. Currently, lenders use a complex set of rules based on income history, asset and net worth information, credit reports, and a host of other factors to identify a borrower’s creditworthiness. And despite all of this data, the decision to approve a borrower’s loan is still ultimately made by a human being. By automatically extracting the data from the PDFs that it was previously trapped in, storing it in machine-readable format, and applying rules to this data, we are able to automate a great number of tasks that are currently being done manually, and supercharge the humans that are currently performing many of these low-leverage tasks. For example, we are able to automatically and programmatically identify which loan products a borrower might qualify for, dynamically identify and generate requirements that a borrower needs to satisfy to get approved, ascertain borrower creditworthiness, and much more. By using this “living data”, we have just turned the mortgage application process into one that is proactive and streamlined rather than reactive and manual. In a world where it takes 50 days to actually obtain a mortgage, we believe we can decrease this number to under 2 weeks by simply leveraging the data that we are collecting, and the technology at our disposal.

1-77aww2hxt2qh9uzeyz-hnw

From Data To Insights

Another area that we are investing in at Blend is the application of anonymized, aggregate-level data to understand trends and patterns across borrowers and lenders. By securely and automatically collecting data directly from authoritative sources, and applying a set of flexible business rules to this data, we believe that we can not only improve the efficiency of originating a single mortgage, but also vastly improve the functioning of the home lending ecosystem as a whole.

By collecting data across millions of borrowers, and clustering these borrowers based on shared characteristics, we can apply these aggregate learnings to the individual, resulting in a dramatically faster and more personalized application experience for each singular borrower. Imagine a scenario where we are looking at millions of borrowers who have specific types of income, certain types of transactions within their financial accounts, and specific events in their credit history that require follow-up documentation from the borrower. For this cohort of similarly-situated borrowers, we are able to understand the additional requirements that a borrower is required to provide, and can apply these learnings to a new borrower that satisfies these criteria, resulting in a dramatically-streamlined application experience. Using this aggregated, anonymized data, instead of spending days to weeks going back and forth with their lender, the borrower could theoretically complete their entire mortgage application, including all documentation, in a single day.

In addition to providing benefit to borrowers, Blend can help lenders create immense value in the efficiency of application evaluation, while also driving insights for their business. Blend can help lenders understand patterns of borrower behavior, the relative popularity of different loan product offerings, and the efficiency of their originations process. In addition to helping lenders improve the borrower experience, Blend can also help them make decisions faster and with greater confidence, all while helping borrowers obtain a mortgage with greater success, better terms, and that is tailored to their needs.

Policy Implications

The work we are doing at Blend also has the potential to help support better national housing policy and improve the fairness and functionality of the home lending ecosystem as a whole. By collecting and analyzing demographic data, Blend can help policy makers level the lending playing field so every person, without regard to demographic profile, is extended the best possible loan that s/he is qualified for. After all, two individuals with similar credit profiles and transaction parameters should receive the same pricing and approval rates, irrespective of demographic data such as sex, marital status, race and ethnicity. In addition, mortgage loan data can help lenders objectively monitor, assess, and revise their policies, procedures, and practices so they can track their own efforts to achieve fair and responsible lending. By helping improve the process of collecting and analyzing borrower data, technology can assist policy makers and lenders in effecting better decision-making, oversight, and outcomes.

Mortgage lending has and continues to be plagued by issues of a previous century. We firmly believe that by efficiently collecting data, and effectively applying the learnings of this data to the mortgage process, we can take mortgages into the 21st century. The current state of lending, one that requires over 400+ pages of documentation, costs lenders $7,000, requires 25+ human beings, and takes 50 days to complete, all to originate a single loan, is one that we at Blend are working hard to make a distant memory.

David St. Geme is a Product Manager at Blend. In his role, he defines and builds out the borrower experience for the product.