Credit Modelling 101: Credit Where Credit’s Due

Abstract: The 2024 Canadian Lenders Summit panel, “Credit Modelling 101: Credit Where Credit’s Due” discusses the rapid evolution of credit scoring, new tools, techniques, and data sources reshaping how lenders assess risk and opportunity. Panelists discuss the changes to credit scoring, how to build credit models, and opportunities that can be find in underserved customer bases.   👉 Check out the full video here.   👀  

Credit scoring has evolved beyond traditional FICO models. How do you view this evolution?

Paul: Credit bureau data is still king. The best predictor of future credit behavior is past credit behavior. However, orthogonal data points—like consistent contributions to an RRSP—can provide additional insights. The challenge lies in correlating these data points to individual customers.

Samantha: Data, tools, and techniques are now more democratized. However, alternative data, such as social media mining, is often overhyped and rarely replaces traditional financial data.

Tom: While data, techniques, and technology have always been the pillars of credit modeling, the pace of change has accelerated. Open banking is particularly exciting, offering new ways to leverage traditional data sources.

Matt: At Trust Science, we focus on combining alternative data with traditional methods to identify subprime consumers who perform like prime borrowers. The key is using data in novel ways to enhance resolution and decision-making.

For lenders launching new credit products, the lack of initial performance data presents a unique challenge. How can they build effective credit models from scratch?

Samantha: Start simple. Use segmentation or application data to categorize customers into risk tiers. As data accumulates, evolve to proprietary risk models and advanced techniques. For example, a first-generation model might use linear regression, while a third-generation model might leverage gradient boosting.

Paul: At Wealthsimple, buying initial credit bureau data helped us define our target population and shape our product. While correlating internal data takes time, the effort pays off in improved decision-making.

Subprime markets are often underserved due to limited credit data. What new approaches are bridging this gap?

Matt: In Canada, privacy laws restrict alternative data compared to the U.S. However, open banking and consumer-permissioned data can unlock valuable insights, especially for subprime consumers.

Testing is essential for optimizing credit strategies and understanding the difference between causation and correlation. How can organizations build a culture of testing?

Tom: At TD, we’ve established a ‘Test and Learn’ center of excellence to standardize frameworks and foster a culture of experimentation. Incremental testing allows us to identify opportunities and mitigate risks.

Matt: Small-scale tests and retro studies are critical. They enable lenders to evaluate models and strategies without major disruptions.

Building a sophisticated model is crucial, but how do you ensure it’s effectively applied?

Samantha: Risk models drive segmentation, but valuation methodologies help monetize those segments. Monitoring and adjusting policies are equally important for maintaining effectiveness.

Tom: Understanding the full cash flow and performance of loans is vital. At TD, our ‘Evaluation Center of Excellence’ complements testing efforts to ensure data-driven decision-making.

What actionable advice do you have for lenders new to credit modeling?

Samantha: Launching with short-term or secured products provides faster data collection and learning cycles.

Tom: Consider external factors like inflation and COVID-19 when analyzing historical data. Use sensitivity analyses to account for biases.

Matt: Incremental testing and continuous monitoring are essential for refining models and strategies. Additionally, open banking and consumer-permissioned data offer opportunities to enhance decision-making, especially for underserved markets.

Brent: Thank you to our panelists, Paul, Samantha, Tom, and Matt, for sharing their expertise. And thanks to the audience for your insightful questions. As the credit landscape continues to evolve, embracing innovation while staying grounded in proven practices will be key to success. Thank you to the CLA for the wonderful event! Check out the full video here.   👀  

Sign up for our 2025 Summit Series