How the book behaves, and how we keep it diversified

Two working models behind Raincheck's underwriting. Both are built from the live FloodNet sensor record, priced at today's rate, and refreshed as the data changes.

448live sensors
6,714NYC block groups
19.4%price, % of limit / yr
43%median-year loss ratio
9 scenarios

Loss scenario tree

Storm season (low / medium / heavy) × loss severity (low / median / high) → loss ratio at today's price, with the simulated histogram and the storm days behind every leaf. Includes 2024, 2025 and 2026 year-to-date replays.

 3 rule sets

Capacity & waitlist model

How a sale is decided block group by block group: a daily cap, a 5% single-block-group share rule, and a 1-km radius rule that slows a hot corner's neighbours before they hit their own limits. Day-by-day launch replay on a map of all 6,714 block groups, with a 24-hour waitlist.
What these pages are for. They show how we think about loss and concentration before a single policy is sold. The scenario tree uses real sensor history; the capacity model uses real geography and hazard with simulated demand. Both will be re-run on real sales once the book exists.