Subramanian Narayanaswamy, Financial Services Risk Management Executive
Credit is the bloodline of our economy. Yet, for 22 million Americans in 2022 per TransUnion, that lifeblood flows sluggishly – if at all. These thin- or no-file consumers are often invisible to traditional credit scoring models despite potentially being reliable borrowers. But a decisive shift is underway, fueled by alternative data, and it’s forcing lenders, from nimble fintechs to established credit unions and banks, to rethink how they assess risk.
The Credit Invisible: Thin-File and No-File Consumers
So, what exactly is alternative data? In essence, it’s any consumer information valuable for risk decisions that fall outside the purview of the traditional Big Three credit bureaus. Think beyond just loan and credit card payment history and consider patterns in rent and utility payments, bank account cash flows and even data points used primarily for fraud prevention like device fingerprints, geolocation and the age and consistency of email addresses and phone numbers.
For years, fintechs have leveraged this broader data spectrum as their bread and butter, successfully challenging traditional institutions by reaching customers the old models overlooked. Now, the pressure is on for banks and credit unions. Why? Because failing to underwrite these segments, particularly younger generations like Gen Z and millennials (who often have thin files), means potentially losing them forever to competitors who see their potential first.
The Benefit for Lenders and Applicants
The benefits are compelling. For lenders, especially when evaluating borderline cases or applicants below traditional FICO cutoffs, alternative data can tip the scale, revealing insights that justify approval. This isn’t recklessly lowering standards; it’s about gaining a more complete picture. One key insight: lenders can often approve more applicants in near-prime or thin-file segments using alternative data without taking on significantly higher default risk compared to traditional scoring methods for those same segments. It helps uncover the ‘hidden’ primes or individuals who are responsible but lack conventional credit history.
For borrowers, the impact is profound. It offers a pathway to mainstream financial services, a potential stepping stone to financial stability. Access to a small-dollar loan or a secured credit card, underwritten using a more holistic view of their financial habits, can be the first rung on the credit ladder, preventing a slide toward costly payday lenders.
AI + ML: The Power Behind the Data
Driving the effective use of this data deluge is the power couple of modern finance: Artificial Intelligence (AI) and Machine Learning (ML). These technologies excel at sifting through hundreds of data points, both traditional and alternative, to build highly predictive risk models. The results can be dramatic. In some cases, ML models achieve four times the predictive accuracy of traditional scores for credit risk.
In the realm of fraud detection, AI/ML is even more crucial. Fraudsters are dynamic, constantly devising new schemes. AI models can be retrained frequently (even weekly) to spot emerging patterns humans might miss. This involves triangulation and cross-referencing various data points (phone, email, device, location) to answer the fundamental question: Are you who you say you are? The power here is significant, with examples of ML models identifying fraud segments with 15 times greater precision than standard methods.
Compliance: Regulation, Fairness and Explainability
However, navigating this new landscape requires careful footing, particularly around regulations. The Fair Credit Reporting Act (FCRA) and Equal Credit Opportunity Act (ECOA) are paramount in the US. Any data used for lending decisions must be proven non-discriminatory, and crucially, the reasoning behind a declined loan (adverse action) must be clearly explainable to the consumer. This allows for disputes and gives individuals actionable feedback. It’s why inherently opaque or potentially biased data points, like social media activity or mobile phone usage patterns, are generally off-limits for US credit decisions. Explainability remains a hurdle, especially for more complex AI models, but advancements in interpretable AI, for which humans can understand its reasoning, are paving the way for broader adoption.
The market itself reflects this shift. Major bureaus like Experian and TransUnion have acquired leading alternative data providers (Clarity, FactorTrust, DataX) and integrated these insights into their core offerings. This makes it easier for lenders to access a blended view of consumer risk.
From Credit Access to Portfolio Profitability
Challenges remain, of course. Making small-dollar loans profitable is notoriously difficult, often relying on fostering long-term relationships (Lifetime Value – LTV) and encouraging repeat business from customers who prove their creditworthiness (and frequently exhibit 50% lower risk on subsequent loans). Stress-testing these portfolios, often through A/B testing different loan structures during good times, is vital. And macroeconomic factors remain critically important, perhaps even more so for community institutions with geographically concentrated portfolios, influencing everything from underwriting adjustments to CECL (Current Expected Credit Losses) provisions.
Ultimately, alternative data is rapidly becoming just data. It’s an essential toolkit for any lender looking to grow responsibly, promote financial inclusion, manage risk effectively, and compete in the modern financial ecosystem. The era of relying solely on the traditional credit score is fading; a more nuanced, data-rich approach to understanding consumer risk is not just the future – it’s rapidly becoming the present.
You can reach Subbu at snsubbu@gmail.com