If your institution is experiencing elevated losses alongside declining efficiency, one of three feedback loop gaps is almost certainly the root cause: your data infrastructure prevents a shared view of credit performance across origination, servicing, and portfolio monitoring; your decisioning framework lacks real-time signal processing to catch emerging risk before it becomes a charge-off; or your governance execution produces dashboards that inform but do not trigger action.
The financial industry has a measurement confidence problem. Bank Director’s 2025 Technology Survey shows that roughly half of financial institution executives believe their organization uses analytics effectively for risk management. Yet only 39% use a data lake or warehouse, and 41% still run lending analytics through spreadsheets. You cannot have both realities at once. Cornerstone Advisors’ 2025 Data EQ study puts the average institution at 241 out of 500 on data maturity, with financial institutions specifically averaging just 211. As Ron Shevlin of Cornerstone noted, “Without a strong data strategy, even the best AI tools will underperform or backfire.”
The cost of not measuring shows up across the portfolio simultaneously: slower approvals, wider loss variance, higher origination costs, and declining borrower satisfaction. These are not separate problems. They are the same problem expressed through different symptoms.
Most financial institutions track what is easy to count: loan volume, delinquency rates, and charge-offs. High performers track what is hard to count but just as important for growth: where decisions degrade in the credit process, where borrower friction drives abandonment, where policy exceptions cluster, and whether portfolio outcomes match original underwriting assumptions.
Borrower friction measurement for example is one area that is underdeveloped. The Fed’s 2024 Small Business Credit Survey found that 22% of large-institution applicants reported a difficult application process, worse than smaller institution applicants. JD Power’s 2025 mortgage study found that satisfaction is 32 points higher when lender engagement starts early and 64 points lower when first engagement begins at application. Those gaps represent retention and referral losses that never appear in a charge-off report.
A closed feedback loop in lending requires four structural elements: integrated data across origination, credit, servicing, and portfolio monitoring; stage-level metrics for process quality and borrower experience; formal review cadences with defined escalation thresholds; and a mechanism pushing findings back into policy, pricing, workflow, and model adjustments. Most institutions have pieces of this. Almost none have all four operating together continuously.
The regulatory expectation for closed-loop measurement is not optional. SR 11-7 requires ongoing monitoring and outcomes analysis comparing model outputs with actual outcomes, along with model recalibration when results fall outside acceptable thresholds. FDIC interagency guidance calls for “independent, ongoing credit risk review” with communication to management and the board. The closed-loop structure is embedded in regulatory expectation, whether institutions recognize it as such or not.
The gap is not that financial institutions measure nothing. RMA’s 2024 model risk management survey found that more than 90% of respondents formally review highest-risk models annually. The real problem is that measurement is disconnected, slow, or not translated into operational decisions quickly enough to prevent loss.
McKinsey documented a middle-market financial institution that rebuilt its credit underwriting process with end-to-end time, cost, and quality measurement embedded in portfolio tracking. The result: an expected $15 million loss reduction in year one and a 3% improvement in return on capital over three years. The improvement came not from tightening credit standards, but from giving decision-makers visibility to act on portfolio signals before they became losses. Freddie Mac’s research reinforces the point: after strengthening lender feedback and QC review processes, Non-Acceptable Quality rates fell roughly 30% from peak and repurchase requests declined approximately 60%.
The efficiency and quality gains from feedback-driven lending compound from the same system improvements.
McKinsey’s digitized credit-risk case provides the clearest operational before-and-after in public research: time to a credit decision fell from 24 to 48 hours down to four minutes for fully automated cases; cost per origination dropped 30% to 40%; and the application experience shrank from 50 screens to five. Freddie Mac’s 2024 Cost to Originate study shows the average retail lender was losing approximately $600 per loan, with origination costs rising 35% over the prior three years. Operating at the bottom quartile of performance is no longer a sustainable position.
Technology alone does not close the loop. McKinsey explicitly warns that vendor platforms fail when institutions do not redesign end-to-end processes and governance around them. Bank Director confirms it: of the institutions that set clear technology objectives, 41% still reported that an initiative fell short in the prior 18 months. Buying software without building measurement architecture is the most expensive way to avoid the problem. What high performers do differently is build feedback into the platform design itself, so that every decisioning step generates data, every data point reaches the right owner, and every owner has a defined threshold for action.
Days 1 to 30: Data and Decisioning Audit
Assess your current position on the measurement maturity ladder: ad hoc and manual (spreadsheets, after-the-fact problem solving); periodic reporting (monthly dashboards, limited cross-functional action); managed cross-functional (common data model, standardized exception taxonomies); or closed-loop (near-real-time monitoring, stage-level alerts, continuous policy and model recalibration).
Map your data infrastructure to identify how many source systems require manual reconciliation at handoffs and whether your warehouse or data lake spans origination, servicing, and portfolio monitoring. Identify the metrics you currently cannot report without a manual extraction. Those gaps are your highest-priority feedback loop investments. Key benchmarks to measure against: straight-through processing rate, cost per loan by product type, exception rate by loan officer, time from application to first decision, and portfolio performance variance against original underwriting assumptions by vintage.
Days 31 to 60: Governance Structure and Escalation Design
Define the accountability structure that makes measurement actionable. Named owners, defined thresholds, and explicit decision rights over policy, pricing, and model changes are the mechanism that turns measurement into correction. Build escalation thresholds that answer three specific questions: what exception volume level triggers a policy review, what early-warning signal triggers a portfolio manager call, and what model drift level triggers an out-of-cycle validation event.
Days 61 to 90: Platform Integration and Execution Roadmap
Prioritize three integration investments: a unified data model connecting origination, servicing, and portfolio performance; workflow tooling that captures stage-level process metrics automatically; and decision engine connectivity that routes model outputs, exceptions, and alerts to the appropriate owner in real time.
Set 12-month performance targets against peer benchmarks. For cost per loan, target movement toward the top-quartile benchmark of approximately $6,900. For straight-through processing, set a minimum target of 50% for straightforward credit decisions with a 70% longer-term goal. Track learning rate, not just outcomes. McKinsey estimates simplification at scale can generate lasting productivity gains of up to 15% in two years and increase ROE by 1.0 to 1.5 percentage points. That is the financial return of a system that learns continuously.
The OCC warned in fall 2025 that lack of investment in new technologies, products, and services may threaten long-term institutional viability. Fintech application share among small-business borrowers is rising. Margin pressure is compressing. Your institution cannot improve credit quality, reduce cost per loan, and retain borrowers simultaneously by doing more of what produced the current performance gaps. Those outcomes require measurement infrastructure, governance frameworks that convert signals into action, and execution platforms that close the loop continuously.
You cannot improve what you do not measure. The financial institutions compounding advantage right now are not the ones with the tightest credit policies. They are the ones with the most precise feedback loops. The diagnostic is clear. The performance gap is measurable. The question is whether your institution builds feedback loops into lending now, or continues managing outcomes it cannot fully explain.
Baker Hill is a leading provider of lending solutions for financial institutions across the United States, helping institutions lend better, lend faster, and lend more through integrated loan origination, risk management, and portfolio analytics platforms.