How it works
Every status on this site comes from the company's own public status page. We check 1,441 of them every 5 minutes and keep every incident they publish.
What we collect
- Current status — the overall indicator each status page publishes (operational, minor, major, critical, maintenance) and the state of each component.
- Incidents — name, impact, start and resolution time, affected components and every update the company posted. Status pages show only recent history; we keep all of it. Our archive reaches back to August 26, 2013.
- Scheduled maintenance — announced planned work.
How we calculate
- Uptime
- 1 minus the time with a major or critical incident open, over the period. Overlapping incidents count once. Minor incidents and maintenance do not reduce uptime but are listed. At most the first 24 hours of any one incident count: week-long notices are almost always advisories or problems in one region, not the whole service being down.
- Forgotten incidents
- Some incidents stay open on a status page long after anyone updates them. An open incident with no update for three days counts as ended one day after its last update, and is shown as "no update since".
- Complete history
- Status pages list their latest 50 incidents. For services that report very often, that list reaches back only weeks; until our own archive covers a full period, we do not show or rank uptime for it.
- Time to fix
- The median time from an incident's start to its resolution.
- Big outages
- We measure how many services usually report a new incident in an hour. An hour in which far more distinct services than that report incidents (the usual rate plus four standard deviations, and at least four services) is listed as a big outage.
Limits
Companies write their own status pages. Some report every slowdown, others only large outages, and some report late. Our figures describe what each company published, not independent measurements.