A service runs normally: users send requests, an app answers, a database keeps the data. An incident is any unplanned event that harms that service for its users: errors, slowness, lost data, an outage. A planned maintenance window is not an incident, and a failing disk nobody feels is not one yet.
This lesson follows one made-up incident from the first error to the written review. The numbers (times, error rates) are illustrative. The point is the shape of a good response: stop the harm quickly, then learn from it.
💡Good to know
Judge an incident by what users feel, not by how scary the machine failure sounds.