The model
The Day Zero Index is a modeled likelihood, not an observed probability: the chance of a 72+ hour disruption to water, power, or payments in a county within the next 90 days. Every input, weight, and change is published here. Read it before you trust it.
Status, 17 Sep 2026. The public panel runs on sample figures for 8 states. No data feed is connected and no forecast has been scored yet. The sections below describe the method the live model will follow, and what exists today.
How the index is built
Each county gets one figure between 0 and 100 percent for the coming 90 days. It is built from five drivers, each scored 0 to 5 and weighted, then calibrated against the past 40 years of events that actually cut households off. The five drivers, in fixed order:
- Financial and payment systems
- Water and drought
- Grid load and outages
- Conflict and sanctions
- Freight and fuel
The figure is banded the same way everywhere it appears: 35 percent and above is Elevated, 20 to 34 is Moderate, below 20 is Low. Elevated is the only band drawn in red, because red is reserved for conditions that affect a real household.
The index says how likely you are to need your supplies. Days of supply says how long they would last. The two are kept separate on purpose: a household target of 14 days is a preparedness goal, not a point on a risk scale.
Track record
A forecast only earns trust by being scored against what happened. This section will hold, for every 90-day window the live model has issued, the counties where a Day Zero event occurred, the likelihood the model gave them beforehand, and the calibration across all counties.
No forecast has been issued or scored yet. The first entry appears here when the live model ships.
Data sources and weights
The model reads public data only. Each driver is fed by a source that any reader can check, and each carries a published weight that sums to one across the five. Today none of the five feeds is connected, so no weights are set.
| Driver | Source | Weight | Status |
|---|---|---|---|
| Financial and payment systems | To be published | — | Not connected |
| Water and drought | To be published | — | Not connected |
| Grid load and outages | To be published | — | Not connected |
| Conflict and sanctions | To be published | — | Not connected |
| Freight and fuel | To be published | — | Not connected |
Method changelog
- 17 Sep 2026 — Public Index panel published on the landing page with state and county drill-down for 8 sample states. Figures are illustrative. Bands fixed at 35 / 20. Wording changed from “chance” to “modeled likelihood” throughout.
- 17 Sep 2026 — This method page published with the intended five-driver structure, an empty track record, and no connected feeds.