severity-vs-priority-reference

Overview

Severity and priority are two independent axes, not synonyms. Every triaged defect must have both, and the combination drives scheduling, escalation, and SLA enforcement.

• Severity measures how badly the defect breaks the system / user when triggered. It's an objective property of the defect.
• Priority measures how urgently the fix must land. It's a scheduling decision, influenced by severity but also by business context (customer commitments, regulatory deadlines, release timing).

This skill is a pure reference consumed by defect-report review (which rejects reports that conflate the two), bug-tracker-workflow, and downstream triage tooling. Two companion references extend it: the defect lifecycle (states, allowed / forbidden transitions, tracker vocabulary) in references/lifecycle.md, and the categorisation taxonomies (IEEE 1044, ISTQB CTAL-TA root cause, ODC) in references/taxonomy.md.

When to use

• Reviewing whether a bug report's severity + priority are both set and consistent with the defect description.
• Configuring a tracker's severity / priority field options.
• Coaching a tester / triager on the distinction.
• Checking a state transition against the canonical lifecycle (see references/lifecycle.md).
• Categorising defects for root-cause analysis or process metrics (see references/taxonomy.md).

How to use

1. Read the defect and decide what function it breaks and how badly; pick a severity (S1-S5) from the severity scale anchors.
2. Separately weigh how urgently the fix must land given business context; pick a priority (P1-P5) from the priority scale.
3. Locate the pair in the 5x5 matrix; if the cell is "rare" or "unusual", record an explicit rationale for the combination.
4. Confirm you did not collapse the axes (do not force P1 just because severity is S1) - the two values are independent.
5. Set the tracker's severity and priority fields per the configuring-a-tracker table for your platform.
6. On any later re-triage (state change), keep severity fixed and revisit priority only.
7. Log every priority change with its rationale.

The two axes

Severity scale

Per IEEE 1044-2009 (cite by stable ID; "Standard Classification for Software Anomalies") and ISTQB CTFL syllabus:

Severity	Definition	Example
Critical (S1)	System unusable; data loss; security breach; total loss of function.	Production database is corrupted on user signup.
High (S2)	Major function broken; significant user impact; no easy workaround.	Checkout fails 30 % of the time; no workaround.
Medium (S3)	A function works incorrectly; usable workaround exists; impact bounded.	Date picker shows wrong month abbreviations; users can type the date.
Low (S4)	Cosmetic or minor inconvenience; functionality not impaired.	Footer copyright year not updated.
Trivial (S5)	Spelling, alignment, polish-level issues.	"Settngs" link misspelled in sidebar.

Per ISTQB Glossary "severity" entry (glossary.istqb.org): "the degree of impact that a defect has on the development or operation of a component or system."

Priority scale

Per ISTQB Glossary "priority": "the level of (business) importance assigned to an item, e.g., a defect."

Priority	Definition	Example
Immediate / Blocker (P1)	Stop the release; halt downstream work until fixed.	Production outage; security incident in progress.
High (P2)	Fix in current sprint / before next release.	Customer-reported high-severity bug.
Medium (P3)	Fix when capacity permits in coming sprints.	Medium-severity bug with workaround.
Low (P4)	Backlog; fix opportunistically.	Low-severity, no customer impact.
Deferred (P5)	Won't fix in foreseeable horizon.	Cosmetic in deprecated feature.

Why they must be separate axes

The "combined severity-priority" scale is a common anti-pattern. Each cell in the 5×5 matrix below represents a real combination:

	P1 Immediate	P2 High	P3 Medium	P4 Low	P5 Deferred
S1 Critical	typical	unusual but real	rare	rare	rare
S2 High	typical	typical	typical	rare	rare
S3 Medium	unusual	typical	typical	typical	unusual
S4 Low	unusual	unusual	typical	typical	typical
S5 Trivial	unusual (PR demo!)	rare	unusual	typical	typical

Worked examples that prove the distinction:

• S1 / P5 (Critical / Deferred) - Real defect: a security vulnerability in a legacy system scheduled for deprecation next quarter. Severity is objectively Critical (the bug, if triggered, breaks the system) but priority is Deferred (the whole subsystem is being removed; patching it is wasteful). Action: track + monitor; if exploited before deprecation, re-prioritise.

• S5 / P1 (Trivial / Immediate) - Real defect: a typo on the hero copy of the marketing site that says "Bweekly" instead of "Biweekly," and the CEO just shared the page on LinkedIn. Severity is objectively Trivial (no function broken) but priority is Immediate (PR / brand impact).

• S2 / P3 (High / Medium) - Real defect: checkout intermittently fails for a specific bank's cards (3 % of transactions), workaround = use a different card. Severity is High (major function impaired), priority is Medium (workaround exists + no SLA breach yet).

• S3 / P1 (Medium / Immediate) - Real defect: a regression surfaced 4 hours before a marketing email lands that mentions the feature. Severity is Medium (function works in 90 % of cases) but priority is Immediate (marketing dependency).

Worked example

A CSV export appends one trailing blank column for every user. The export still succeeds and every data value is correct.

1. Severity: the function works and data is intact; the defect is a cosmetic formatting flaw -> S4 Low (functionality not impaired).
2. Priority: a major enterprise customer's ingestion pipeline breaks on the extra column, and their renewal is next week -> P2 High (business commitment, near-term deadline).
3. Matrix check: the pair is S4 / P2 - an "unusual" cell, so log the rationale (customer renewal dependency).
4. Set Jira Severity = Low, Priority = High; do not bump severity to justify the urgency.

Result: the fix is scheduled in the current sprint despite Low severity, driven entirely by the High priority - a clean demonstration that the two axes move independently.

Mapping to lifecycle

Severity and priority don't change with state (typically). However:

• Re-triage at state change: when a deferred defect resurfaces as a production failure, severity is unchanged but priority jumps.
• Severity escalation rules: many trackers auto-bump priority if severity is S1/S2. The platform-workflow skills configure these.
• SLA driven by priority: P1 = e.g., 4-hour response, P2 = 1 business day, P3 = 1 week. SLAs are tracker config, not severity-derived.

See references/lifecycle.md for the state transitions; this skill defines the orthogonal severity / priority fields.

Common confusions

Confusion	Reality
"Severity = customer impact."	Severity = system function impact. Customer impact informs priority. A bug that breaks an unused admin tool is High severity but Low priority.
"Priority always equals severity."	False. The 5×5 matrix has 25 distinct cells, all observed in practice.
"Critical defects always get fixed immediately."	False. A Critical defect in a deprecated subsystem may be Deferred.
"Cosmetic = Low / Trivial severity always Low priority."	False. Marketing context (S5 / P1 example above) can make trivial severity an immediate priority.
"Severity is set by the reporter; priority by the PM."	Convention varies. Best practice: reporter proposes severity; triager confirms severity + assigns priority.

Configuring a tracker

Tracker	Severity field	Priority field
Jira	Custom field (often Severity with select-list)	Built-in Priority field
Linear	Custom field (Severity), no first-class	Built-in Priority field (Urgent / High / Medium / Low / No priority)
GitHub Issues	Labels (severity:critical etc.)	Labels (priority:p1 etc.) + Projects status
Azure DevOps	Built-in Severity (1 - Critical / 2 - High / 3 - Medium / 4 - Low)	Built-in Priority (1 / 2 / 3 / 4)
Bugzilla	Built-in Severity + Priority	Same

bug-tracker-workflow covers the platform-specific configuration for Jira, Linear, GitHub Issues, and Azure DevOps.

Anti-patterns

Anti-pattern	Why it fails	Fix
Single combined field	Loses information; everyone fights about whether to bump severity to escalate or vice versa	Two fields, explicit per defect
Default both fields to "Medium"	Triagers don't engage with the distinction	Require both fields explicitly on transition from New → Open
Auto-equate severity with priority	"S1 must be P1" - disallows S1/P5 deferred legacy bugs	Allow both axes independently
Severity is purely subjective	Variability across reporters; metrics unreliable	Use the IEEE 1044 examples + team-calibrated rubric
Priority changes weekly without reason	Metrics chaos; no accountability	Log priority changes with rationale
Critical defects always escalated to release manager	Floods the escalation channel	Use severity and priority for escalation rules

Limitations

• Subjectivity remains. Severity has objective anchors (S1 = data loss, S5 = cosmetic) but boundary cases (S2 vs S3 with a workaround) require judgment. Reviewer disagreement is normal.
• Priority is a moving target. Business context shifts weekly; a P3 today may be P1 next sprint after a customer escalation.
• Severity classes vary by tracker. Some use 4-point scales (Bugzilla), some 5 (Azure DevOps), some define their own. The team must pick one and document it.
• No automation can fully classify severity. AI classifiers can suggest, but the final call needs a human in the loop.

References

• ISTQB Glossary - "severity" and "priority" entries - glossary.istqb.org.
• IEEE 1044-2009 "Standard Classification for Software Anomalies" - cite by stable ID; canonical severity classification.
• ISO/IEC/IEEE 29119-3:2021 §6 anomaly-report fields - cite by stable ID.
• Atlassian "Configure priority levels" - Jira priority docs.
• Linear "Priority levels" - Linear priority docs.
• GitHub Issues - labels + Projects status documentation.
• Companion references: references/lifecycle.md, references/taxonomy.md, references/tracker-vocabulary-map.md.