The Data Deluge: A Cautionary Tale for the Future of Credit Underwriting

Chris Joles is senior vice president of enterprise and credit risk at Planet.

Credit officers across the mortgage industry can feel the ground shifting beneath their feet. Not because of inflation, home prices, or geopolitics, but because of something more subtle yet just as disruptive: the sudden arrival of too much data. Utility histories, cash‑flow feeds, BNPL obligations, trending‑behavior scores, and alternative credit signals are pouring into the credit score system faster than ever before.

For decades, underwriting was a discipline of documentation. Today, it is becoming a discipline of interpretation. And in that transition lies the risk that the industry, in its pursuit of inclusion, may unintentionally build a credit ecosystem that is harder to navigate, harder to price, and harder to defend.

The goal of new scoring models like FICO 10T and VantageScore was to expand access to homeownership by capturing a fuller picture of consumer behavior. But with that expansion comes complexity and risk. These models are too young for us to understand how they behave across unique economic cycles. Investors typically need five to fifteen years of performance data before they trust a model’s predictive power. With 10T and Vantage, that seasoning simply doesn’t exist.

This lack of history creates a paradox. Classic FICO hasn’t been updated in decades, but lenders and investors know how it behaves. The new models are more sophisticated but also more opaque. They incorporate trending data, utility histories, and rent signals that often appear only when they’re late, because that’s when they get reported. The result is a credit file that looks richer but may actually be more uneven.

Consider the borrower who pays off credit cards with a personal loan right before applying for a mortgage. Classic FICO sees a drop in revolving utilization and rewards the behavior with a score that jumps to 700. FICO 10T, however, sees the trend line—the run‑up in balances, the consolidation pattern—and lands the borrower closer to 660 or 680. After COVID, this pattern became common. When disposable income tightened again, Classic scores fell back to earth. 10T had already priced that risk in.

This divergence could affect how investors price mortgage‑backed securities. When they don’t understand how a score behaves, they price conservatively. And conservative pricing means higher costs for borrowers and tighter credit for everyone.

Utility and rent data were supposed to help thin‑file borrowers, but they’ve introduced their own complications. As an industry, we haven’t agreed on how to classify these obligations. Are they tradelines? Are they installment loans? Should they be included in DTI? If they reflect payment ability, why wouldn’t they be? And if day care (sometime equal to a 5-year monthly mortgage payment) is excluded, why would utilities be included?

The danger isn’t the data itself. It’s the absence of a shared framework for how to use it. Without that clarity, lenders risk misclassifying liabilities, LOS systems risk pulling the wrong data into the wrong buckets, and investors risk losing confidence in the consistency of the credit box. When that happens, spreads widen. Pricing worsens. And the very borrowers the industry is trying to help end up paying the price.

But the most immediate pressure point isn’t investor behavior. It’s the operational burden placed on credit officers. The shift from document‑based underwriting to data‑driven underwriting has created a new reality: once a data point is visible, we’re going to need it validated.

This is where the cautionary tale becomes most acute.

Imagine an underwriter reviewing a file enriched with 24 months of cash‑flow data. They see a large deposit from 14 months ago. It’s outside the required window. But it’s there. And once it’s there, is it a question? A condition? Multiply that by every NSF, every irregular transfer, every unexplained payment, and the underwriting process slows to a crawl.

Even if the GSEs say, “You don’t need to validate everything,” the psychological and legal pressure will push underwriters to do exactly that. Repurchase anxiety doesn’t disappear just because the rulebook says it should. Data is like rabbits to the underwriting greyhound. The more we see, the more we must chase. And the more we chase, the more the credit box tightens because of operational drag.

BNPL adds another layer of uncertainty. It’s the loudest topic in credit circles and the least understood. There is no centralized, reliable reporting. Borrowers can have dozens of BNPL obligations that never appear in a bureau file. If an underwriter suspects BNPL exists, what is their burden? Do they ask? Do they verify? Do they include it in DTI? What happens when the data is incomplete?

BNPL is the perfect example of the slippery slope: if the industry begins including everything it can see, where does it stop?

The path forward requires discipline. Not just in how data is used, but in how the industry defines the credit box itself. Lenders need explicit guidance on what counts and what doesn’t. LOS systems need standardized liability guardrails. And investors need transparency into how new models and new data sources are being interpreted.

The industry is standing at a crossroads. The push for financial inclusion is real and necessary. But inclusion without clear intent, industrywide agreement, and standardized infrastructure can lead to chaos because more data is only an asset if it can be interpreted correctly, validated efficiently, priced confidently, and used equitably. Otherwise, the data deluge becomes a flood that washes away the very borrowers the industry is trying to help.

The future of credit underwriting should be shaped not by how much data we can collect, but by how wisely we use the data. The guardrails we build before the new ecosystem hardens around us could determine whether we expand opportunity or unintentionally narrow it.

(Views expressed in this article do not necessarily reflect policies of the Mortgage Bankers Association, nor do they connote an MBA endorsement of a specific company, product or service. MBA NewsLink welcomes submissions from member firms. Inquiries can be sent to Editor Michael Tucker or Editorial Manager Anneliese Mahoney.)