IDP is the fastest path to no-touch mortgage

Amit Jnagal
Founder & CEO
October 28, 2021

When we started our IDP business, our assumption was that cost savings would be the main driver for purchasing decisions - in other words, how we can help our customers save 30-50 percent on data processing costs. But the last six years have taught us those cost savings are the number two priority for most companies looking for IDP solutions.

Any guesses at number 1?

Our customers have taught us that the number one thing is being able to effectively scale their operations. Today, data processing is heavily dependent on manual labor. We’ve worked with customers who were worried about their business picking up more volume because they just did not have the capacity to handle that increase. There are two aspects to this challenge.

Data Processing Department

Most companies that handle complex documents need to train their new employees from six to nine months before they are no longer required to review their data extraction work. It takes this much time for them to understand and see all document variations and to become productive and accurate. Unfortunately, data processing is not a dream job for a lot of people, resulting in a good amount of churn in this job segment. Worse yet, when a data processing person leaves the job, all the knowledge and the training leaves with them.

Data Decision Department

In addition to data processing employees, scaling presents a challenge when it comes to knowledge workers, the people who need to make decisions based on the extracted data. Your underwriters, claims adjusters, actuaries, etc. Ask anyone who manages a team of underwriters and they will tell you how difficult it is to recruit and scale that team. Mortgage companies need to be extremely judicious of how they use their time.

The Case for No and Low Touch Mortgage

As your mortgage business scales, the first departments that struggle with additional pressure are underwriting and data processing. Being able to manage efficient mortgage processing is a huge competitive advantage. Your SLAs can almost single-handedly dictate how much business you will pick up. Mortgage companies need to invest in technology that can help them scale their operations without having to scale headcount in these departments. The leadership teams of most of the mortgage companies that Infrrd works with are committed to minimizing the number of documents touched by their data processing and decisions teams.

Path To Reducing Touches For Mortgage

A no-touch mortgage is a state that all mortgage companies should aspire to attain but it is a journey with a lot of milestones. You cannot switch from 100% touch to no-touch operations with the flip of a switch. Here are the broad milestones of making this pilgrimage:

  • Data Processing - Low Touch
  • Data Decisions - Low Bounce
  • Data Processing - No Touch
  • Data Decisions - Low To No Touch

Goal: Data Processing - Low Touch

The first step in reducing touches is to automate the related data processing work. This is where our IDP solution helps a great deal by automatically handling incoming documents. The data processing team only looks at documents that our AI algorithms cannot handle with high enough confidence. These are given to the data processing team for a second look. Every time they make a correction, our algorithms learn and errors are reduced. . The feedback-based learning loop puts you on the track to no-touch processing over time.

The Rub

The challenge of low-touch processing is the quality of documents. Mortgage and mortgage insurance companies often do not deal with the borrowers directly. They get documents from other originators who have scanned them in low resolution and sent them for processing. Machine learning-based pre-processing, training, and image touch up can improve processing for low-resolution documents and decrease the low touch.

Goal 2: Data Decisions - Low Bounce

Before you get to low touch on the data decision side, you should plan to reduce the bounce between the decision teams and the data source. This bounce is the back and forth that you go through between the underwriters and the borrower because the data in the documents does not add up. For example, a borrower mentioned that he has a gross pay of $80,000 but the payslips only add up to $65,000. By the time this data goes to the underwriters and they take a look at it and send it back, a lot of valuable time has been wasted.

The Rub

IDP solutions that are simply focused on extracting data from complex documents do not have the capability to correlate this data. But this functionality is so tightly integrated with data extraction that it is counterproductive to use two different products. The amount of touch time that you will save in the overall process will be offset by the time you spend in integrating these systems and keeping them current. Our Intelligent Workflow and business rules system lets you configure these correlation rules to decrease the touchpoints in data processing.

Goal 3: Data Processing - No Touch

After you roll out a machine learning-based IDP system, it immediately gives you a reduction in touch point with the data processing team. With time, as it keeps relearning from corrections made by your data processing team, the touchpoints become less frequent and you are well on your way to a touchless data processing system. The key here is to make sure you choose an IDP platform with machine learning built into its core and c send the correction feedback to this system for relearning.

The Rub

The biggest complication of no-touch processing is the confidence score of the IDP platform. You should be able to rely on this confidence score. It is extremely important that when the IDP system tells you that it is confident of its processing results, you should be able to skip the manual data processing step completely. Our IDP system uses a combination of three algorithms to figure out when it has extracted the correct values. Though it is unable to extract 100% of the data all the time, it is almost 100% accurate when it tells you that you do not need to review this data.

Here is actual data from one of our implementations:

As you can see, the reliability of the confidence score is fundamental to reducing the touch in the system.

Goal 4: Data Decisions - Low Touch to No Touch

In order to get to low-touch decisions on the decisions side, you need to invest in a machine learning-based prediction system that can present a few decision options based on what it learned from past data. Like the IDP system, it should learn continuously from the decision corrections to reduce the work of the underwriters.

The Rub

If you need to get to a no-touch decision system, then you need to make sure you have more than one decision system or algorithm. No-touch decisions should only be taken when 3 or more independent algorithms come up with the same decision. If one of the algorithms disagrees, then it needs to be presented to the underwriter for review.

Where there is a will…

We are very excited about the goals that the mortgage industry has set for itself in terms of automation leading to low touch and no-touch processing. We are also honored to partner with some of the most innovative mortgage companies that are pushing the envelope of how fast this new technology can become a reality. This year has already seen some very encouraging developments in mortgage processing, and the next one looks even more promising.

Here’s to a lower-touch mortgage in 2022, which will benefit both mortgage companies and their customers.

Frequently asked questions

What does your pricing model look like?

We price based on the annual volume of pages and complexity of document type.  We can get you preliminary pricing once we outlined a solution.  Let's do this.

To know more, book a 15-min session with an IDP expert

How can I try Infrrd before I commit to a full deployment?

Sure.  The first step is to schedule a guided demo where you get to jump into the thick of it.  After you explore our solution you can try a proof of concept. When you're ready, you can deploy the system to one use case.  Then more use cases.  Then across your enterprise.

To know more, book a 15-min session with an IDP expert

How does your system integrate with others in my enterprise?

We play nice.  Our solutions are API-based.  Your documents are feed into the solution using APIs. And extracted data is sent out through APIs.  We use REST APIs.

To know more, book a 15-min session with an IDP expert

Does your solution run in the cloud or on premise?

Our solution is cloud-native but is also design for premise deployments.  Your choice on how you want to deploy it.

To know more, book a 15-min session with an IDP expert

Does Infrrd run on mobile or desktop device?

Glad you asked.  Our data extraction process runs on servers.  We have found performance and accuracy decline when running on a desktop or mobile device. (Remember Infrrd is running a powerful AI stack).

To know more, book a 15-min session with an IDP expert

Does your system work out of the box or does it require training?

Common documents and use cases work out of the box.  The cool thing is your solution will improve as the system learns from your documents upfront and over time.

To know more, book a 15-min session with an IDP expert

How does your solution handle corrections?

Did you know no system is 100% accurate all the time?  When extraction errors occur you want to correct them.  We provide a simple UI that your business analyst will use to make corrections.

To know more, book a 15-min session with an IDP expert

Does your solution work with handwriting?

Our solution excels at data extraction from handwriting.  We've got proprietary methods and techniques that do the trick.  It's pretty cool.  See for yourself.

To know more, book a 15-min session with an IDP expert