Severity levels
Bug severity levels are the ordered scale a team uses to rate how much damage a defect does when it occurs, independent of when it will be fixed. The common ladder runs blocker, critical, major, minor and trivial: from a defect that stops testing or use entirely, down to a cosmetic flaw with no effect on function. The level describes the bug, never the schedule.
The longer version
Each level answers one question: what happens to the user, or to the test run, if this defect is left alone. A blocker prevents further testing or use of a whole area, with no workaround: the login form rejects every password. A critical defect breaks a core function or loses data. A major defect breaks a significant function while the rest of the product still works. A minor defect misbehaves in a way the user can step around. A trivial defect is cosmetic: alignment, a typo, a colour off by a shade.
Two other things share the word. Incident management uses severity levels, often written Sev1 to Sev4, to grade a live outage and decide how fast people are paged; that scale rates an event in production, and this page is not about it. Priority is the other neighbour and the more damaging confusion, because it rates when a bug gets fixed rather than what it breaks; the contrast is covered under bug severity vs priority.
Inflation breaks the scale first: reporters learn that only critical bugs get looked at and so file everything as critical, after which the level carries no information and triage has to re-rate every report. A scale with no written definitions breaks it the other way, where one tester's major is another's minor and the meeting argues about the label instead of the bug. A team fixes both by writing one sentence per level, with an example, and by having whoever reproduces the bug set the level rather than whoever reported it.
The level is assigned once, at reproduction, and rarely changes, which lets the rest of the process lean on it. Escalation rules in defect management are written per level: a critical defect that has waited too long is raised, a minor one is not. A build with an open blocker does not ship. A scale that is inflated or undefined breaks every rule that depends on it.
Common questions
What are the five levels of bug severity?
Blocker, critical, major, minor and trivial, in that order. A blocker stops testing or use entirely with no workaround, critical breaks a core function or loses data, major breaks a significant function, minor is a fault the user can work around, and trivial is cosmetic. Some teams merge blocker into critical or add a cosmetic level below trivial; the shape is the same.
Who decides the severity of a bug?
Whoever reproduces the bug, usually a tester or the developer who picks it up, because judging damage needs only the bug in front of you and a written definition of each level. The reporter proposes a level; the reproducer confirms or changes it. Product owners set priority, which is a different question.
How many severity levels should a team use?
Enough that each level changes what happens next and few enough that two testers pick the same one. Four or five is the usual range: with fewer than three every bug is either urgent or ignored, and beyond five nobody can tell adjacent levels apart. Each level needs a written sentence and an example, and the team should be able to name a rule that treats it differently from its neighbours.
What is the difference between a blocker and a critical bug?
A blocker stops testing or use of an area outright and has no workaround, so nothing behind it can be checked until it is fixed. A critical defect breaks a core function or loses data but leaves the rest of the product testable. Teams that merge the two lose the signal that tells a release manager a build cannot even be tested, and that decision is the reason the top level exists.
Can the severity of a bug change after it is filed?
Severity changes when reproduction shows the damage was misjudged: a report filed as minor turns out to corrupt saved data, or a supposed blocker has a workaround the reporter did not find. Once set at reproduction the level stays put, because escalation rules and release gates read it. A level that drifts with the schedule has become priority under another name, and any change is recorded with a reason.
Bug reports that arrive ready to triage.
Collect replay, console, network and browser context with the Free plan.