How IncidentHub Counts GitHub Outages
The rules behind the figures on the GitHub reliability page.
How We Count
Outages are included in the count when their start time falls in the window, in UTC. Maintenance events are excluded.
Major means GitHub labeled the outage critical or major. Minor means GitHub labeled the outage as minor. Any other label, including none, is unranked and stays in the total.
Median length uses resolved outages only.
Source
The major/minor labels are from GitHub's own public status page data, recorded as the highest severity GitHub published for the outage. IncidentHub does not assign a separate grade. Severity can change over an outage's lifecycle. Peak severity is the highest impact GitHub had published at the times the data was collected.
Critical and major are counted together as major. Minor is recorded as minor. Any impact marked as none, a missing label, and any other values are unranked.
Windows
The 90-day figures are the 90 days ending at the as-of time.
The 12-month figures, including the chart and the component counts, are the last 12 calendar months in UTC, including the current month. The current month contains only the days up to the as-of time.
The as-of time is when these figures were computed (the results might be cached for up to an hour on the GitHub Reliability page). It is the date to cite if you use these figures in your work.
Time
The page also shows how many major outages are still open, if any.
Median length uses resolved outages of every severity in the window. Open outages are left out of the median.
Components
A component's count is the number of outages in the window that list it. An outage that lists two components counts once in this table for each (but is recorded as one outage overall).
The most affected component is the one with the most outages. Note that an outage can affect multiple components.
A component name links to its page on incidenthub.cloud.