The “High-Performance Lending Report” reveals significant gaps between the lending priorities of community-based financial institutions and their current operational capabilities. The following sections extract and categorize the key data points and statistics found in the sources.
Top Executive Priorities (2021–2024)
Financial institutions have shifted their focus toward efficiency and cost management. The percentage of executives citing these as a top concern nearly doubled from 2023 to 2024.
| Concern | 2021 | 2022 | 2023 | 2024 |
|---|---|---|---|---|
| Cost of funds | 8% | 9% | 45% | 70% |
| Efficiency, noninterest expenses, costs | 31% | 36% | 29% | 53% |
| Interest rate environment | 55% | 46% | 56% | 49% |
| Deposit gathering | — | — | — | 48% |
| New customer/member growth | 33% | 34% | 30% | 48% |
Credit Modeling and Data Landscape
While almost 90% of institutions have at least one credit model, only 10% consider their approach “industry-leading”.
• Model Sources:
→ 63% use syndicated scores (e.g., FICO or VantageScore).
→ 24% use custom in-house models.
→ 21% use models from their primary Loan Origination System (LOS) provider.
• Use of Nontraditional Data: Two-thirds of institutions use at least one nontraditional source.
→ 41% use employment history.
→ 35% use deposit account transaction patterns.
→ 23% use FCRA-compliant alternative data or bill payment history.
• Satisfaction and Preparedness:
→ Only 13% are “very satisfied” with the availability of credit data.
→ Only 13% feel “very prepared” to adapt models to changing market conditions.
→ 66% admit to having difficulty updating their credit models.
Operational Speed and Auto-Decisioning
There is a stark contrast in efficiency between institutions that automate and those that do not.
• Efficiency Gain: Institutions using automated decisioning review 3.5 times more loan applications per underwriting full-time equivalent (FTE) employee per month compared to those that do not.
• Model Cycle Times:
→ 35% can build and validate a new model in under a month, but over 50% take six weeks or longer.
→ Deployment times vary: 39% can deploy a model in 1–2 weeks, but 13% take 6–8 weeks.
• Adoption of Auto-Decisioning:
→ 80% of institutions generate automated underwriting decisions.
→ However, 40% of institutions auto-decision only 20% or fewer of their total loans.
→ Staff Confidence: Only 25% of origination staff are “very confident” in automated outputs.
Artificial Intelligence (AI) and Machine Learning (ML)
Despite high belief in AI’s potential, actual deployment remains low.
• Current Adoption: Only 13% have already deployed AI in credit and lending; 29% do not have it on their radar at all.
• Top Concerns Regarding AI:
→ 70% cite regulatory scrutiny as a major concern.
→ 45% worry about the cost to acquire.
→ 42% are concerned about transparency, bias, or lack of knowledge.
• Future Investment (Next 3 Years):
→ 12% plan to invest more than $500,000 in AI-driven models.
→ 25% plan to invest less than $50,000.
Performance Impacts of AI/ML Implementation
The sources highlight specific, measurable benefits for institutions that adopt machine learning.
• Approval Rates: ML models using alternative data approve 23% to 29% more applicants compared to traditional models.
• Interest Rates: These models can lower the average annual percentage rate (APR) by 15% to 17% for approved loans.
• Collections: Using ML to optimize contact strategies for delinquent accounts can reduce net charge-off losses by 4% to 5%.
• Implementation Speed: High-risk account identification models can be implemented in 12–16 weeks and offer a 30%–40% improvement in accuracy over existing models.
4 mins
Talk to Our
Consultants
Chat with
Our Experts
Write us
an Email
