The lending industry has seen drastic changes in the past 15 years. Lenders are turning to advanced analytics to compete and keep up with economic changes, new regulatory requirements, and new competition from more nimble digital operators.
With the growth of analytics demand and data science research across all industries, AI model solutions are being offered as the ultimate product to lenders. US lending industry regulators have published concerns on model explainability in a recent fair lending report.
At the same time, all lenders need a lending decision strategy. Lending strategies may drive the eligibility criteria for a loan, pricing, loan amount assignment, collection calls, and other key lending processes. Recent progress in AI/ML for lending strategies with particular attention to regulatory compliance is getting traction.
Below are 5 reasons why leading Financial Institutions are focusing on lending strategies to drive business resiliency, inclusivity, and growth.
1. Strategy first, models second
A model by itself doesn’t add any value. A model needs a strategy to understand how to use it. For example: do we want to approve all customers above a 600 score derived from the model?
On the flip side, a strategy can be defined and may perform well without a model.
In addition, a robust model development methodology requires a large enough sample size and target values, for example, underperforming loans. This is not always available, especially when exploring a new lending segment or launching a new fintech business.
Initially, innovative lenders usually start deploying a pilot strategy on the target segment to capture enough data to eventually build a custom model.
After building experience and further development, lenders may add custom models to their strategy. However, models developed on historical data or simulations have training “blind spots” and require additional strategy input to refine them.
Strategies are even more critical in lending as uncontrolled “blind spots” or what is called “tail risk” in credit may have disproportionate consequences on revenue and drive business net losses.
2. Fast review, backtest, and adjustments
A lending strategy is usually defined by a discrete criteria or a set of rules. These rules have the main purpose of managing lending risk and reward: for underwriting, the goal is to optimize revenue within the business credit risk appetite.
The number of rules in a strategy is usually kept within the low double-digit to increase statistical significance, reduce overfitting, be human-explainable, and easy to understand.
These characteristics help lending executives and analysts get a fast understanding of strategy impacts, validate the strategy performance, and propose enhancements.
On the other hand, model review, backtesting, and adjustments after implementation may take several weeks or months based on the model complexity.
3. Strategies + AI + compliance = competitive advantage
Lending regulators have recently shared concerns about the use of AI/ML to develop models, especially for credit underwriting. Model decisions are usually based on complex interrelationships that are difficult to translate into easy-to-understand customer decline reasons.
By contrast, strategies high explainability promotes them as the ideal candidate to drive the next lending evolution.
With the recent progress in artificial intelligence and automated machine learning, advanced techniques may be used to develop lending strategies and replace the historically manual process.
An AI-driven strategy builder with automated compliance testing provides a clear competitive advantage while maintaining fair lending standards.
4. Highly adaptable
Leading lenders are prioritizing strategies because of their adaptability.
While models often need rigorous and lengthy governance standards, a strategy can be adapted with relatively little impact on compliance.
This helps lenders better adapt to market volatility or new portfolio management goals.
The strategies ease of use also helps lenders drive inclusivity: get to market faster with testing and validation of new alternative data that can be leveraged to offer credit to underserved segments.
Meanwhile, models strongly rely on historical data. Models add value to strategies when the economy and the targeted customer base have the same characteristics as its development data.
When a change happens, lenders need to budget and plan for expert resources and long development/validation time cycles to adapt.
5. Simple to implement
The nature of lending strategies is characterized by a small number of rules. Their high explainability makes them easier to review by lending executives, fair lending compliance teams, and implementation teams.
With the support of a digital-ready Loan Origination System, strategies can be reviewed and implemented within days.
In contrast, models may take months of development, testing, and implementation based on the complexity of the algorithms.
Now you know 5 reasons why lending success comes with strategy. To read more on strategies and AI, I’d recommend reading our 10 Insider Tips For Lending Success Using AI.
This article will help you learn more about what AI for lending strategies is, what it can accomplish, and the steps for reaching your own lending success.
Ready to try the first Lending Strategy System? Request a demo of Prometics Lending NavigatorTM and subscribe below to get the latest industry insights, products updates, community events, and other news.