Beyond Wildfire: How Jessica Leong and Octagram Analytics Are Rethinking Fire Risk in Insurance
Beyond Wildfire: How Jessica Leong and Octagram Analytics Are Rethinking Fire Risk in Insurance  
Podcast: Full Episodes | Insurtech Leadership Podcast
Published On: Sat Apr 11 2026
Description: Introduction In this episode of the Insurtech Leadership Podcast, host Josh Hollander welcomes back Jessica Leong, co-founder of Octagram Analytics, to discuss FireRQ — a non-catastrophe fire risk model delivering actionable risk scores for any U.S. address. While the industry fixates on headline-grabbing catastrophes, Jessica and her team are tackling the everyday fire risk that quietly drives loss ratios, underwriting decisions, and portfolio performance. Guest Bio Jessica Leong is co-founder of Octagram Analytics, an actuarial analytics firm. Before founding Octagram, she served as Head of Data & Analytics at Zurich North America, where she led the team that built all predictive models for pricing and claims. She is also a former President of the Casualty Actuarial Society. Jessica brings over a decade of experience in insurance predictive analytics to the problem of non-catastrophe fire risk. Key Topics Non-cat fire risk: the overlooked loss driver — Fire (excluding wildfire) accounts for 15–30 points of property loss ratio in homeowners and commercial lines, yet most carriers treat it as a solved problem. Jessica explains why it isn't. The dataset advantage: 1.7 million fires — Octagram built FireRQ on the National Fire Incident Reporting System (NFIRS), a publicly available dataset of fires reported by U.S. fire departments. Even Fortune 500 carriers only see ~1% of this data in their own books. Repeat fires and fire clusters — The data reveals that buildings with prior fires are significantly more likely to burn again, and that fires cluster by geography and occupancy type. The Myrtle Beach hotel cluster (10–15 hotel fires per year in a single zip code) is a striking example. Machine learning for fire prediction — FireRQ uses a gradient boosting machine (GBM) that starts with building-level history, then branches outward to area and occupancy-level fire experience. The model captures 80% of fires in the worst 20% of buildings. How underwriters use FireRQ — Carriers apply the score for pricing adjustments, risk selection (declining high-score accounts), and early warning. Octagram offers a free proof of concept using an older model version so clients can validate on their own loss data. Model transparency and explainability — As larger accounts adopt FireRQ, demand for "why" behind scores is growing. Octagram is adding context layers: prior fires at the location, area-level fire frequency, occupancy benchmarks. What's next for Octagram — LiabilityRQ and CrashRQ are in development, extending the same data-driven approach to liability and auto crash risk. Quotes "We can look at 100% of the data where you're staring at 1% of the data." "If we tell you these buildings are the worst 20% buildings in the U.S., we do see they have 80% of the fires." "No one talks about [non-cat fire] anymore, but it's still a very, very real risk." Resources Octagram Analytics website: octogramanalytics.com The Little Book of Fires: Free resource available on the Octagram Analytics website National Fire Incident Reporting System (NFIRS): Publicly available U.S. fire data Subscribe & Review If you enjoyed this episode, subscribe to the Insurtech Leadership Podcast on YouTube, Apple Podcasts, Spotify, or wherever you listen. Leave a review — it helps other insurance and technology professionals find the show.