Methodology

The data

Every figure comes from Form EIA-861, the annual filing US electric utilities make to the Energy Information Administration. We use the 2025 reliability file, published 2026-10-07. It is US government work and in the public domain.

The three measures

• SAIDI — total minutes the average customer went without power over the year. This is what we rank on.
• SAIFI — how many separate interruptions the average customer had.
• CAIDI — average length of one interruption (SAIDI divided by SAIFI).

We include storms

Utilities report each measure twice: with and without “major event days” — the hurricanes, ice storms and wildfires that cause the longest outages. Every ranking here uses the figure including them. A Florida ranking that excluded hurricanes would describe a year no Florida customer actually experienced.

States rank by median, not average

A single rural cooperative can post an extraordinary year — one Tennessee utility reported over 12,000 minutes — and an average would let it define its entire state. Michigan’s median utility is at 592 minutes against a 742-minute average for exactly this reason. We rank states on the median so the number describes a typical customer.

What we leave out

752 utilities filed reliability data. Those that did not are absent from this site entirely — never shown as zero, never estimated from neighbours. We also do not publish a ranking page for a state with fewer than 5 reporting utilities, because a list of three is not a ranking. That excludes 10 states and territories, including West Virginia, which has the worst average outage time in the country across its four reporting utilities.

What these numbers cannot tell you

Utilities self-report, and definitions of a sustained interruption vary somewhat between them. Rural systems maintain far more line per customer than urban ones, so a high SAIDI often reflects geography rather than management. The national median is 3 hours; treat that as the comparison line, not a pass mark.