Unlocking the Loan File: The Untapped Intelligence Inside (Sponsored by Paradatec)
Lee Gillispie, Managing Principal at Paradatec
Mortgage originators and servicers have invested heavily in building increasingly sophisticated systems of record for data.
For originators, the LOS knows the finalized results of income, asset, credit, and property calculations required to support a loan approval decision.
For servicers, the servicing system knows the current balance, payment history, interest rate, escrow position, delinquency status, and hundreds of other attributes associated with a loan.
For investors determining the right price for a pool or portfolio, data files are critical for their modeling systems, to identify value and possible risk.
But behind that structured data sits something potentially even more valuable.
The loan file documents = the source of truth
The loan application, income and asset documentation, appraisal and other property valuation and assessment documents, title work, property, and mortgage insurance policies, closing documents and other records contain the source information from which the summary loan decisions originate.
That includes the loan approval to the borrower, the loan purchase approval by a correspondent lender or secondary market investor, and the loan file onboarding into the servicing system.
Historically, besides using these documents to validate that the loan is a compliant and enforceable lien asset, they have primarily been treated as records to retain and retrieve when necessary. But they are much more than that. The documents are the original source of truth about the loan.
What if we started treating them as portfolio intelligence?
Start With Data Integrity
The first opportunity is determining how much confidence a lender, servicer or investor should have in the data it already has.
A servicing system might identify a property as owner occupied. But what do the underlying documents say?
The loan application may contain an occupancy designation. So may the appraisal. Insurance documentation may provide another indication.
If those sources agree with one another and with the servicing system, confidence increases.
If they don’t, that doesn’t automatically mean something is wrong. It means there is an exception worth understanding.
Apply that concept programmatically across thousands—or millions—of loans and the original loan file becomes another layer of portfolio surveillance.
The same principle can be applied to property characteristics, insurance information, loan terms, borrower information, liens, taxes, HOA obligations and potentially thousands of other data elements.
Instead of asking “What does our system-of-record say?” a loan purchaser, investor, or servicer can begin asking:
“What do the source documents say, and does our current data agree where applicable?”
From Data Validation to Risk Identification
That creates a natural second use case: identifying risk.
The same data element often appears on multiple documents within a mortgage file. Historically, that redundancy has been a processing challenge.
It can also be an advantage.
Technology can identify a data element wherever it appears, compare the values across documents and then compare those findings with the servicing system.
The objective isn’t necessarily to have technology decide whether a loan presents a compliance or credit problem.
It’s to surface inconsistencies for the people qualified to make that determination, with the applicable data so they can make that decision quickly and confidently.
That changes the exercise from searching loan files for potential problems to managing identified exceptions.
For large investment and servicing portfolios, that distinction can be significant; especially when identifying issues can be performed in real time.
Creating Portfolio Intelligence
There is an even broader opportunity.
Some of the information contained in the original loan documents may never have made it into the systems-of-record at all. That means it’s not available in any data file outputs from those systems, either.
An appraisal contains considerably more than a property value. It may contain detailed information about condition, quality, improvements, effective age, and property characteristics.
A loan application can contain information about the borrower’s home life including dependent needs, employment history, asset mixes, and other real estate owned.
Insurance, title, and closing documents contain still other information about the borrower, collateral, and transaction.
Not every historical data point remains relevant. And historical information should never automatically be assumed to describe the borrower or property today.
But making that information structured, searchable and comparable creates two advantages most organizations have never really had:
- the ability to interrogate the collective intelligence contained in the original loan files across an entire portfolio
- the ability to create agentic and other workflow automation based on data modeling
And Then There Is the Customer
Once that foundation of clean labeled data exists, including data provenance and context, another possibility emerges.
The same intelligence may help identify relevant customer opportunities.
Consider the original appraisal.
Rather than simply knowing that a property was valued at $350,000 ten years ago, the underlying document may tell us that the kitchen and bathrooms had not been updated, the property’s effective age, its condition, and other characteristics.
Combine that historical information with today’s loan balance, a current AVM and other appropriately obtained current data, and a servicer may be able to identify homeowners for whom a home-improvement HELOC could be particularly relevant.
Or consider other real estate owned.
The original loan file may establish that a borrower owned several investment properties at origination. Combined with current information, that historical insight might help reveal a customer with broader real-estate financing needs.
Similarly, comparing the borrower’s historical debt profile with appropriately obtained current credit information could potentially identify situations where a consolidation product deserves consideration.
The distinction is important.
The objective isn’t to turn old loan documents into a sales-lead database.
It is to create better portfolio intelligence—and then determine where that intelligence can responsibly improve both risk management and the customer relationship.
Unlocking What We Already Own
One of our industry colleagues, Mirza Hodzic of Black Wolf Advisory, made an important observation as I was discussing this with him: the opportunity should begin with data integrity, risk identification, and portfolio intelligence, with customer opportunities viewed as an additional potential use case.
I think he’s exactly right.
That framing points toward a larger evolution in closed-loan data management.
The industry has spent decades digitizing loan documents and building systems capable of storing enormous amounts of loan-related data.
The next question is:
What information exists inside the source documents that mortgage finance systems-of-record don’t know—and what could we learn if they did?
The original loan file of documents can help:
- An investor collect and triangulate more data when confirming the loan’s delivery price, find risk, and assess potential future value for a borrower
- A servicer validate the accuracy of its current servicing data, identify inconsistencies, and better understand risk across the portfolio.
And once that capability exists, the same intelligence may reveal opportunities to create additional value for the customer and the institution.
The mortgage industry has already digitized the loan file.
Now it’s time to unlock it.
(Sponsored content includes material submitted independently of the Mortgage Bankers Association and MBA NewsLink and does not connote an MBA endorsement of a specific company, product or service. For more information about sponsored content opportunities, contact Bill Farmakis at bill@jlfarmakis.com or 203/834-8832.)
