The Donut Hole in Lending Technology: The Credit Risk Analytics Gap

Lending has evolved rapidly over the past decade, but not evenly.

Most innovation has concentrated at the edges of the lending workflow. On one side, data access has improved dramatically. Open banking, aggregation, and enrichment tools make it easier than ever to ingest transaction-level data. On the other, workflow and decisioning platforms have matured, offering configurable rules, orchestration, and automation at scale. What’s often missing is credit risk analytics: the layer that turns raw data into decision-ready risk and capacity insight.

Between those two layers sits the hardest problem in lending, and the one that most stacks still struggle to solve – turning raw data into a rigorous, defensible view of credit risk, capacity, and value remains a gap. 

This is the donut hole in lending technology: the credit risk analytics gap between data access and decision execution.

Why credit risk analytics in the middle of the stack matters

Access to data doesn’t answer underwriting questions on its own. 

A decision engine, however flexible, doesn’t determine what the right decision should be on its own either.

The critical question lenders face is unchanged:

Should we extend credit to this customer, for this amount, on these terms, given our risk appetite and economics?

Answering that requires more than categorization, rules, or static scores. It requires credit risk analytics that can interpret behavior, assess capacity, and quantify risk in ways that reflect how borrowers actually manage money.

When this middle layer is weak or missing, lenders tend to fall back on blunt tools, bureau scores designed for a different era, rigid policy bands, or over-simplified ratios. The stack may look modern, but decision quality often lags behind.

How to tell if your stack has a donut hole

Many lenders already sense this gap, even if they don’t label it as such. These are usually signs the credit risk analytics layer is thin, generic, or missing.:

• You ingest transaction data, but primarily use it for categorization or simple affordability ratios.
• Your decision engine executes policy well, but relies on static or generic scores underneath.
• You approve or decline confidently, but struggle to size limits or terms precisely.
• You optimize models for AUC or Gini, but struggle to connect improvements to margin or loss outcomes.
• You detect borrower stress only after missed payments or delinquency signals appear.
• Manual overrides are common, compensating for weak separation in edge cases.

If several of these feel familiar, the issue is not data access or workflow; it may be the analytics in between. 

The real cost of the gap

The donut hole is not a theoretical problem. It shows up directly in portfolio outcomes:

• False declines, where resilient borrowers are turned away because capacity isn’t visible
• Under-lending, where strong customers receive conservative limits and generate less value than they could
• Over-extension, where borrowers with acceptable bureau profiles take on more credit than their cash flow can sustain
• Delayed stress detection, increasing roll rates and servicing costs
• Operational friction, as teams rely on manual reviews and overrides to “fix” model blind spots.

Individually, these might look like edge cases. In aggregate, they could shape growth, loss, and margin for your business. 

What a strong credit risk analytics layer looks like

Filling the donut hole is not only about adding more rules or more features. It’s about strengthening credit risk analytics so decisions are informed by behavior.

A robust middle layer typically has a few defining characteristics:

• Interprets transaction data behaviorally, not just by category.
• Focuses on stability, volatility, timing, buffers, and interactions, not totals alone.
• Produces calibrated risk or capacity estimates tied to real outcomes.
• Connects outputs directly to decisions around approval, limits, pricing, and tenure.
• Is explainable enough to support governance, adverse action, and internal oversight.
• Integrates alongside existing systems rather than replacing them.

This is where underwriting moves from “Can we automate this?” to “Are we making the right decision?”

Why a strong analytics layer matters now

As open banking levels access to data, advantage no longer comes from who can get information, but from who can interpret it best. At the same time, margin pressure, rising delinquencies, and greater regulatory scrutiny are pushing lenders toward precision rather than blunt conservatism.

In this environment, tightening policy alone suppresses growth. Better analytics create optionality.

How Carrington Labs fills the donut hole

Carrington Labs is built to strengthen the missing middle and provide the credit risk analytics between data access and decisioning and monitoring.

Our focus is on cash flow underwriting and credit risk analytics that sit between data access and decision execution, in addition to post-origination monitoring. We build models that translate transaction-level behavior into decision-ready insights, capturing stability, volatility, liquidity management, and capacity in ways that static scores and category summaries cannot.

Just as importantly, our approach is designed for real lending environments. Models are explainable, configurable, and aligned to lender economics. They can be run in shadow mode, deployed as challenger signals, or gradually weighted into live decisions, allowing teams to prove impact safely and incrementally.

Rather than forcing binary approve-or-decline outcomes, this approach enables more precise questions:

What is the right offer? 

What is the right exposure? 

Where does this customer create value within our risk appetite?

The takeaway

As data access becomes commoditized, credit risk analytics that inform decision quality and monitoring becomes the differentiator.

The lenders that perform best will not be those with the most data or the most rules, but those that can interpret behavior accurately and connect it to decisions they can defend, commercially, operationally, and regulatorily.

That is the donut hole in lending technology.

And it is the gap Carrington Labs is designed to fill.