Risk-Based Pricing Engine
Project Information
- Category: Algorithmic Lending
- Client: Personal Loan Fintech
- Stack: PD Model, Dynamic Rate Engine, Loan Book Training
- Delivery: Integrated Credit + Pricing API
From flat-band pricing to granular PD-based rates—in the same sub-2-second decision window.
A personal loan fintech was pricing all borrowers within a credit tier at the same rate. Best-in-tier applicants were being underpriced; riskiest were underpriced relative to their actual PD. We built a risk-based pricing engine trained on their own loan book that prices each borrower on their individual predicted default probability.
The Challenge
The fintech assigned every approved borrower within a credit tier a single flat interest rate. A borrower at the top of the band and one at the bottom received identical pricing—meaning the best borrowers were being offered rates that were uncompetitive, and the riskiest were being approved at rates that didn't adequately price their default probability. The result was margin compression at the top end, higher-than-necessary expected credit loss at the lower end, and no mechanism to use pricing as a competitive lever to improve conversion among low-risk borrowers.
What We Built
We built a risk-based pricing engine that maps each borrower's predicted probability of default to a dynamic interest rate band. The PD model is trained on the client's own historical loan book and repayment data. The pricing engine is integrated directly with the underwriting engine so that pricing runs in the same API call as the credit decision—returning both the credit outcome and the borrower-specific rate in under 2 seconds. Pricing band definitions and rate floor/ceiling constraints are editable by the product team without engineering involvement.
Key Capabilities Delivered
- PD-to-rate mapping with configurable risk band definitions and rate constraints
- Integrated with underwriting engine: credit decision and pricing returned in a single API call
- PD model trained on client's own loan book and repayment behaviour
- Pricing rules editable by product team without engineering changes
- Sub-2-second combined credit + pricing decision under production load
- Full audit trail on every pricing decision with PD score and applied rate band
The Outcome
The fintech moved from flat-band pricing to granular PD-based rates across their entire approved borrower population. Margin capture improved on lower-risk approvals as competitive rates reduced rate-shopping drop-off. Higher-risk approvals are now priced more accurately relative to their predicted default probability. The pricing engine operates within the same sub-2-second window as the credit decision, so applicants receive both their approval and their personalised rate instantly.