The Economics of Credit Scoring Algorithms in Fintech

How fintech lenders use alternative data to assess credit risk, and what that means for people traditional scoring overlooks.

Traditional credit scores, covered in this curriculum’s credit-debt module, rely almost entirely on a person’s history with loans and credit cards - a system that works reasonably well for people who already have that history, and works poorly for people who don’t. A wave of fintech lenders has built alternative approaches to fill exactly that gap.

The limits of traditional credit scoring

Traditional credit scoring models are built almost entirely from data about past borrowing: credit card payment history, loan repayment records, and total amounts owed. This approach is genuinely predictive for people with an established credit history, but it structurally excludes anyone who hasn’t yet built one - immigrants new to a country’s financial system, young adults just starting out, and people who’ve simply avoided using credit, whether by choice or circumstance.

Thin-file borrowers and the gap they create

Financially responsible, but invisible to a traditional score

Consider someone who has consistently paid rent and utility bills on time for years, holds a stable job, and has simply never taken out a loan or opened a credit card. Under a traditional scoring model, this person is a **thin-file borrower** - someone with too little credit history to generate a reliable score at all - and may be rejected for a loan not because of any actual financial risk, but simply because the traditional system has no data to evaluate them on. This person may, in reality, be a genuinely low-risk borrower who traditional scoring simply can't see.

Alternative data: filling in the picture

Many fintech lenders now incorporate alternative data - information beyond traditional credit history, such as rent payment records, utility bill payment history, cash flow patterns visible in a linked bank account, or even employment stability - to build a more complete risk assessment for borrowers a traditional score can’t evaluate well. This approach can meaningfully expand access to credit for creditworthy people the traditional system overlooks, addressing part of the credit-invisible problem discussed in this curriculum’s credit-debt module.

Algorithmic underwriting and its own risks

The broader shift toward using this expanded data through algorithmic underwriting - automated systems evaluating loan applications using far more data inputs and more complex models than a traditional credit score - brings real benefits in speed and access, but also raises its own version of the algorithmic bias concerns discussed in this curriculum’s ethics module: a model trained on alternative data can still encode unfair patterns if that data itself reflects existing social or economic inequality, even while genuinely expanding access for many thin-file borrowers overall.

Weighing the tradeoff

Alternative data-driven lending represents a genuine improvement in access for many people traditional scoring structurally excludes, but it also requires real scrutiny to ensure the new data sources and models used are actually more accurate and fair, not simply different, and that the expanded access doesn’t come bundled with predatory pricing or terms specifically targeting borrowers who have fewer traditional alternatives to compare against.

Key takeaways

Traditional credit scoring relies on past borrowing history, structurally excluding people who haven't built that history.
Thin-file borrowers can be creditworthy but invisible to traditional scoring simply due to lack of loan or credit card data.
Alternative data like rent, utility payments, and cash flow patterns helps assess risk for borrowers traditional scores miss.
Algorithmic underwriting expands access but carries its own risk of encoding bias present in the alternative data used.
Expanded credit access through alternative data needs scrutiny to ensure fairness and avoid predatory terms.