← All sectors / The AI transformation

140 · Consumer credit & underwriting technology

Lending to people the bureau cannot see

Curve position

Launch pad

Binding constraint

Regulatory scrutiny of alternative data in credit decisions.

Lending to people the bureau cannot see

Traditional credit scoring works well for people with long credit histories and poorly for everyone else. Cash flow data from bank accounts describes ability to repay far more directly than a score built on borrowing history, and open banking made that data accessible.

Historically thin file borrowers were either rejected or priced at rates that guaranteed poor outcomes. That was a data problem dressed up as a risk problem.

The structural driver is access to transaction data plus competitive pressure. A lender that can price a thin file borrower accurately wins volume that incumbents decline reflexively.

The technology layer spans bank account data aggregation, cash flow analysis and income verification, alternative scoring models, fraud and synthetic identity detection at application, and the adverse action explanations regulation requires.

Adoption economics work through approval rate at constant loss rate. Approving more borrowers without raising losses is directly measurable and immediately valuable.

The beneficiaries include underwriting technology vendors, data aggregation providers, identity verification firms, and the lenders willing to underwrite differently.

The value chain runs from data aggregation through modelling to lending decision and servicing. Aggregation is concentrated among a few providers, which is a structural chokepoint.

The overlooked layer includes income and employment verification specialists, debt collection technology, loan servicing platforms, and the compliance vendors handling adverse action requirements.

Competitive dynamics favour lenders and vendors with proven performance through a full credit cycle, which many newer models have not yet experienced.

Risks: credit losses rise sharply in downturns and untested models fail first, regulators scrutinise alternative data for disparate impact, data aggregation access can be restricted, and funding costs move against lenders quickly.

What to watch: approval rates at constant loss, model performance through a credit cycle, regulatory guidance on alternative data, and open banking rule implementation.