Every policy is tied to one of New York City's 6,714 census block groups. The book stays diversified by three rules that decide, block group by block group and day by day, whether a new buyer gets a policy now or a place on a 24-hour waitlist. This page replays a four-month launch under each rule set on simulated demand.
Think of it as a turnstile at the block-group level. A buyer arrives, the rules for that block group are checked against the book as it stands that morning, and the buyer either walks through or joins the line for that block group.
Share rules are measured against the larger of the live book and a -policy seed book, so the first few hundred sales are not blocked by their own smallness. All thresholds are parameters, not conclusions.
Same demand in all three scenarios: a slow citywide ramp plus a storm on day that floods the hottest corner and multiplies sign-ups there. Scrub the day and switch scenarios to see which block groups sell, slow or pause, and where the waitlist forms.
Lines are the three scenarios over the 120-day launch. The radius rule is the one that changes the shape of the book: it caps the hottest 1-km cluster near the line and converts the overflow into a waitlist that clears within weeks.
Why all three end in the same place: demand is identical and the waitlist clears once the book is big enough, so the rules change the path, not the destination. What they protect is the first six weeks, when a flooded corner could otherwise become half the book. Loss volatility is the book's loss standard deviation relative to a book spread evenly over every block group, using a km correlation length between block groups.
Ten largest block groups at day 120 in the selected scenario. Click a block group on the map for the same detail anywhere in the city.
The network rate the quote API prices on, by monthly as-of date. Each block group's rate is the network rate times a local multiplier (its sensors' payout rate relative to the network, clipped to 0.5–2.0×). New sensor information moves the price once a month, never mid-policy.
Room per block group is recomputed daily from the book as it stands. A block group's availability goes down when its share or its 1-km cluster share rises, and up as the rest of the city buys. Price and availability are separate levers: price pays for hazard, availability protects diversification.
This is a mechanism demonstration on synthetic demand. The block-group geography, households, sensor hazard and price history are real; every sale, waitlist entry and offer is simulated. Numbers will change as soon as real sales replace the demand model.