ClearRate: Building a Personal Lines Rating Engine
Insurance pricing requires a balance of transparency and mathematical rigor. I developed ClearRate to demonstrate how a modern, multiplicative rating engine can be built using a pure Python stack for instant premium generation.
The Objective
The goal was to create a functional underwriting tool that:
- Calculates: Computes premiums based on driver, vehicle, and territory factors.
- Adjusts: Integrates a GLM-based credibility model for rate refinement.
- Analyzes: Provides a suite to test how factor changes impact the final quote.
The Tech Stack
| Component | Tool |
|---|---|
| Language | Python (Pandas, NumPy) |
| Modeling | Poisson Log-Link GLM |
| Interface | Streamlit |
| Logic | Multiplicative Rating Model |
Challenges Faced
- Data Normalization: Managing high-dimensional rating tables (Age x Territory x Vehicle Value) using CSV backends required strict schema validation to prevent calculation errors.
- Performance Optimization: Ensuring sub-millisecond premium calculation as users toggled dashboard inputs necessitated highly vectorized Pandas operations.
What I Learned
- Actuarial Modeling in Production: I gained experience translating theoretical GLM credibility models into a functional Python library.
- User-Centric Design: Building the Streamlit interface taught me how to present complex actuarial factor breakdowns in a way that is intuitive for a non-technical underwriter.