top of page

Why Lenders Need Custom Data Models

It’s a fact. A mortgage loan demands more data than almost every other consumer financial transaction. And the data required to get a mortgage loan ready for closing generates still more throughout the origination cycle. Lenders are awash in data. So much so, in fact, that it is difficult to make sense of it, let alone put it to practical use.

The opportunity is immense. Properly leveraged mortgage data provides insights that have the power to improve borrower satisfaction and increase profitability, plus a hundred other useful tips that help everyone responsible for getting the borrower and their loan to the closing table to do so faster and with lower friction.

Properly leveraging mortgage data is the key to success. Yet there is no easy, straight line from data to information to action, and it is easy for that line to become blurry. Becoming overwhelmed by the volume of data, what to do with it and how to use it is entirely normal. It is also far too easy to be misled by points in the data, often down paths that obfuscate rather than elucidate. The prime example of this is the tendency of lenders to concentrate on metrics that are focused on discreet parts of the mortgage process. While use of data in this way can help improve those processes it often creates problems and bottlenecks in other parts of the origination cycle. The far too often result is poor productivity, unsatisfactory customer experiences and low profitability.

This is one of the key reasons we believe mortgage data can and should be put to better use. Doing so is daunting, time consuming and difficult, none of which is made better by off-the-shelf, out-of-the-box data models. These solutions, ubiquitous in the marketplace, promise often promise rapid implementation but they just as often fail to address the fact that, while one mortgage loan is very much like another, no two lenders are alike.

Another way to think about this is your operation is not a loan, it is a business. Most off-the-shelf products look exclusively at well-defined loan attributes, which is helpful though woefully myopic. Other data points like staffing fluctuations, ebbs and flows in productivity, cashflows, market fluctuations, partner performance and other shifts in your business must be included in your data model in order for it to deliver ongoing, real-time value. Custom data models do exactly this, and more.

Custom data models don’t take any longer to implement and the cost of ownership is far less than off-the-shelf solutions. These models are tailored to individual lender business models and practices, markets, borrowers, products, business mix (think QM v. non-QM, as an example) and operations in order to thrive in today’s competitive and contracting market. Implementing a custom model addresses three key elements that out-of-the-box solutions overlook, or skim over:

  1. The Right Data and Metrics: Quality over Quantity. Deciding which data is important and useful is the first step in constructing a custom data model. Quantity is not the problem, nor is it necessarily any data scientist or analyst’s friend. Quality is what matters; these decisions need to be made first and thoughtfully in order for the model to deliver lasting value. Lenders should begin the Quality over Quantity Exercise by focusing on a few, key, macro metrics. Productivity, cost-to-close, pull-through, closing velocity and market share are good starting points. These five metrics highlight that profit is so much more than loan pricing. In fact, loan pricing is dependent on nailing these five metrics. The next step, once the macro metrics are defined, are the micro metrics that focus on discreet steps in the mortgage origination cycle. The list here could be endless; this is where many lenders fail. Micro metrics should be limited in number just like their macro counterparts. And even though they are necessarily focused on discreet processes, they must be defined starting when the borrower first makes an inquiry all the way through closing, post-closing and delivery. This prevents one process from dominating and degrading others. It will take some time to choose the right data and the correct metrics. When done correctly the custom data model becomes immediately useful and valuable.

  1. Data Sources. The data required to manufacture a mortgage comes from many sources other than the borrower, including credit reporting agencies, the GSEs / investors, flood zone providers, appraisers, title companies and the IRS to name many of them. All of this, in turn, is further augmented through the processes of fraud detection, regulatory compliance checks and other verifications. Profitable mortgage lending does not turn solely on production. With that in mind, there are other sources of data lenders must consider while building a custom data model. So much of mortgage, housing, interest rate, GSE/Investor and competitor financial information is publicly available for ready download. Considering how these sources can be integrated with the sources mentioned above add dimensionality to data models, which, in turn, both increases and improves operational and competitive insight. These sources, those used to ‘create’ a mortgage plus those that are publicly available, must be thoughtfully linked if they are to be properly mined.

  2. Data Repositories. The debate about ‘system of record / single source of truth’ versus best of breed origination technologies is about as tired and worn out as a subject can be. And it appears settled. Mortgage data in the 2020s resides in multiple systems given all the component technologies available to lenders today. Each of these component technologies must be included when designing and building custom data models. As with the sources of data, the locations of data must also be thoughtfully linked to bring your entire operation into the picture.

At Grind Analytics we know that a custom data model puts your data in your hands in the unique ways that align with your business model, customers, products, pricing and markets. We make that easy. We are data and IT experts who work with you to design and build custom data models in no time flat. Our data models automate report production, make creating new reports easy with just the click of a mouse, and eliminate SQL queries and the need for advanced Excel skills. We get data right, we get data from everywhere and we make data meaningful and useable by your entire team.

Want to learn more? Contact Brian Benson, COO, at


bottom of page