Where a first site could work: an early screen from public data
Every US county ranked against the others on how often and how long the power goes out, on public EV charging, and on who would need a place to charge. It is a screen for choosing where to look closer, not a market forecast.
How to read this: each score is a percentile of US counties from 0 to 100. Utilities report reliability for their whole area, not by county, so every county figure is an estimate; the hover gives each county's confidence.
The county map loads when you reach it.
Top 25 counties
How the score works, sources and limits
Outage score = the mean of two percentiles among US counties: SAIFI (outages a year per customer) and SAIDI (minutes without power a year per customer), both with major events, each the median of 2021–2025. Outage score v2 adds how often outages run long: the mean of the same two percentiles and the percentile of outages lasting 8 hours or more a year in the county, 2020–2024, from EAGLE-I's 15-minute county outage record (an outage here is a stretch with at least 0.5% of the county's customers out, and at least 50). EV score v1 = 100 minus the percentile of public charging ports per 100,000 residents, so fewer ports ranks higher: it measures room for a new site among existing chargers, not demand. About 860 counties have no public ports, so they tie at the top of v1. EV score v2 leans on demand: the mean of three parts, the v1 score, homes (the mean of the percentiles of renter households and of homes in buildings of five or more units, people less likely to charge at home), and whether an EV-designated Alternative Fuel Corridor crosses the county. Combined = the mean of outage score v2 and EV v2; the hover and the extract also give it with outage v1, to compare. Ties share a percentile.
Shown, not in the score: EVs per 1,000 residents and ports per EV (open county counts exist only for Washington, whole fleet, and Connecticut, EVs newly registered since 2021, so they would rank a few counties on a different basis), DC fast share, planned ports, long commutes, January mean temperature (1991–2020, derived from NOAA's county series, not an official normal), and grocery stores and gas stations (Census County Business Patterns 2023; not available for Connecticut's old counties). New housing permits are not used: the licence is unclear.
How a county gets its figures. Each utility's customers in a state are spread over the counties its EIA-861 service territory lists there, in proportion to each county's customers as EAGLE-I models them for 2022, or to its homes (Census 2020 housing units) where a county has no modelled count. A county's figure is the mean of its utilities' figures, weighted by those estimated customers. Confidence is the share of the county's estimated customers whose utility files to the IEEE 1366 standard: high at 80% or more, medium at 50%, else low. EAGLE-I coverage is flagged low where EAGLE-I's own state coverage rate or data quality index for the state is low, or the county has no modelled customer count; there, few outages on record may mean few reports, not few outages. Counties where no utility files reliability (Puerto Rico, the territories and some small co-op and city utilities) have no outage score.
Context, not in the score: days each county was under each kind of NWS warning, 2014–2025, in the hover and the extract. FEMA disaster declarations are not loaded yet; they will be context too.
Sources. Reliability, service territories and customers: U.S. Energy Information Administration Form EIA-861, via Catalyst Cooperative's PUDL (CC BY 4.0), read Nov 15 2026. NWS warnings: Iowa Environmental Mesonet, Iowa State University. Homes: U.S. Census Bureau. County shapes: U.S. Census Bureau 2017 boundaries via us-atlas (ISC). Charging stations: U.S. Department of Energy, Alternative Fuels Data Center, via the National Laboratory of the Rockies (formerly NREL), read Nov 16 2026. People, renters, apartments, commutes: U.S. Census Bureau, American Community Survey 5-year 2024. Businesses: U.S. Census Bureau, County Business Patterns 2023. This product uses the Census Bureau Data API but is not endorsed or certified by the Census Bureau. Corridors: FHWA Alternative Fuel Corridors via BTS NTAD. Temperature: NOAA NCEI nClimDiv. EV registrations: Washington State Department of Licensing (ODbL 1.0) and Connecticut DMV (public domain). Outage history: EAGLE-I, Oak Ridge National Laboratory for the US Department of Energy (CC BY 4.0), figshare release version 3 (2014–2024), years 2020–2024, county summaries only.
Download the data
- swap-counties.csv: one row per county, every input and score (county summaries only).
- swap-sources.json: every source, its owner, licence and credit.
- eagle-coverage-by-state.json: EAGLE-I coverage by state and the low-coverage rule.
- us-counties-10m.json and its licence: county shapes (us-atlas 3.0.1, ISC).
The Washington EV counts come from the Washington State Department of Licensing under the Open Database License (ODbL 1.0); the columns built from them are shared under the same licence. EAGLE-I figures: credit EAGLE-I, Oak Ridge National Laboratory for the US Department of Energy (CC BY 4.0).