Fintech Takes Facing Credit: Credit Underwriting as a System
Mar 25, 2026
Michelle Young, Plaid credit product lead who builds cash-flow and permissioned data tools. Kiran Aware, LendingClub credit chief who builds decisioning systems. They discuss treating underwriting as a full system, why cash-flow and ongoing signals matter, how lending platforms run dozens of models and experiments, and practical AI impacts on verification, fraud, and future signal generation.
AI Snips
Chapters
Transcript
Episode notes
Motivated Middle Borrower Profile
- LendingClub targets the "motivated middle": digitally savvy, higher-income borrowers with ~720 FICO who prioritize managing credit and paying down debt.
- During inflation LendingClub spotted stress via leading indicators like falling cash balances and pulled back to recalibrate credit strategies.
Trajectory Beats Snapshot In Risk
- LendingClub evaluates borrower trajectories, not snapshots; two borrowers with identical FICO can have opposite risk if one is deleveraging and the other is accelerating debt.
- They embed trajectory-based attributes and custom signals into models to distinguish behavior over time.
Credit As An Operating System
- Credit is an operating system, not a single model: data, multiple models, pricing, verification, fraud, servicing and real-time feedback loops must integrate across the lifecycle.
- LendingClub runs ~60 production models, 240 decision strategies and ~200 concurrent experiments to continuously optimize.