Requirements Traceability Matrix
beta · Family governance-docs · Phase undefined · Sizes lean, full · ~5,200 tokens
The standing table that links each requirement to the work that verifies it, read in both directions, so that a requirement with no verification can be seen. A linked test is not a verified requirement, so the matrix defines what its status words mean before it fills a row. Distinct from the traceability field on a single test case and from the dated result of one test cycle. A contested type: keep one only where the team has decided it owes a whole-set answer to which requirement has no verification.
Fast reference for using the requirements-traceability-matrix bundle. For the full reasoning, the
disagreements between sources, and the citations, read
requirements-traceability-matrix_companion.md.
Decide whether to keep a matrix before you fill one in. This type is contested. Named practitioners say it rarely repays its upkeep, this library’s research found no named software team outside regulated or safety-critical work that describes keeping one (high confidence that the search found none, medium that none exists), and no source read shows that keeping one improves outcomes. The case this bundle can make is narrow: where a team has decided it owes a whole-set answer to the question “which requirement has no verification”, a table is one way to give it. Read When NOT to use first. That recommendation is this library’s position, not a measured finding.
When to use
Section titled “When to use”- The team has decided that someone, such as a release reviewer, an auditor or the team itself, is owed a single answer to “which requirement has no verification”, and a requirement with no test must be visible in that answer. A table with one row per requirement shows it, and a test case’s own traceability field cannot, because a requirement with no test has no test to hold the field. That reasoning is this library’s inference, and no source states it.
- The requirement set is bounded: one feature, one release, or one product area. A bounded matrix is this library’s way of applying NASA’s value-based tracing.
- The work is regulated or safety-critical. The sources that say traceability is owed, or that publish a matrix, are regulated practices. Whether any standard requires a matrix was not read.
- No tracker already holds the requirement-to-test links, or you can state in one sentence what this file adds that the tracker does not.
When NOT to use
Section titled “When NOT to use”- The team is unregulated and cannot name a reason. If the Purpose and Boundary section cannot say what you chose to trace, what you chose not to, and when you will stop, do not start. Companion section 6 gives the conditions Ambler names.
- A tracker already holds the links. Name the tracker as the record, and do not keep a copy. A table that restates links held elsewhere creates a second record that can disagree with the first. That reasoning is this library’s own.
- The situation is simple. Ambler says that where people know the system, it is “much easier and cheaper to simply ask them to estimate the change”. In this library’s view, a solo or very small team should treat the lean size as the largest to consider, and no matrix may be the right answer.
- You want NASA’s requirements verification matrix. It is scoped to “shall” statements and records how each is verified. It is a related artifact, and this bundle’s template is not a copy of it.
- You want proof that the matrix improves outcomes. Nothing read supplies it. See companion section 6.
Matrix, or somewhere else?
Section titled “Matrix, or somewhere else?”Several things a reader might expect in a traceability document already have a home elsewhere. The matrix owns the current links, and it owns nothing else. Each line below is this library’s position unless noted.
| You want to record | It belongs in |
|---|---|
| The current link from each requirement to what verifies it, across the whole set | This matrix |
| The link from one test to what it verifies | The test-case bundle’s own traceability field. The matrix reads it as input |
| The dated result of one test cycle | The test-summary-report bundle. A report is a dated result of one cycle, and a matrix row stays open. No source states this boundary directly |
| The decision to change a requirement, and who agreed | The change-log bundle. The change log records the decision, and the matrix records the current links |
| A risk’s likelihood, impact and owner | The risk-register bundle. The full variant’s Risk Control Trace links to it by identifier and copies no score |
| The requirement’s own text | The source document, such as the prd, user-stories or acceptance-criteria bundle. NASA’s handbook says “do not duplicate requirements in the traceability matrix” |
A matrix that starts carrying risk scores, issue status or metric definitions has begun to collapse into a neighbouring governance-docs member, which the family contract warns against.
Pick a variant
Section titled “Pick a variant”- Lean (default): Purpose and Boundary, Trace Conventions, The Matrix, Coverage Gaps, and Upkeep and Review Trigger. A requirement-to-test table read in both directions, the narrowest definition in the sources, with the rules that give its status words a meaning and the owner who keeps it current. Start here unless you owe an answer the lean table cannot give.
- Full: adds Implementation Trace (each requirement to its design reference and its code or task reference), Risk Control Trace (each risk to the requirement that controls it and the test that verifies the control) and Baselines (dated snapshots of what a release or an audit saw). Move to full when the team owes one of three answers: which design element implements this requirement, which control covers this risk, or what an auditor saw at a given release. Without such a question, lean is enough.
Grow lean into full by inserting the three sections after Coverage Gaps and before Upkeep and Review Trigger; the five lean sections keep their name and order. Confidence that this split is right is medium-low: no source read shows a template that carries every link type, so the split is this library’s hypothesis and not a sizing rule any source states.
Three cautions on the full sections:
- Drop Implementation Trace when a tracker already holds the links. Write “N/A” and one line saying which tracker is the record.
- Risk Control Trace has the thinnest support of any section. The only sources are one medical-device vendor and the FDA’s reference to a trace to “risk analysis results”. No general software source read asks for it. Use it only when a risk register already exists, and write “N/A, no risk register exists for this product” otherwise.
- Baselines suits a team with audits or formal releases. A snapshot of an out-of-date table keeps the out-of-date table, so reconcile the table before you snapshot it.
Quality rubric (self-grade)
Section titled “Quality rubric (self-grade)”Score each row 0, 1, or 2. Below 17 out of 24, a reader who opens the matrix to learn which requirement has no verification cannot trust the answer: a requirement may have no row, a status word may mean only that a test exists, or the table may have drifted from its sources with nothing in the file to show it. A lean matrix carries no Implementation Trace, Risk Control Trace or Baselines section to grade, so it is scored, and cleared, against a smaller table; see the scope table below.
| # | Criterion | 0 | 1 | 2 |
|---|---|---|---|---|
| 1 | Reason to keep it | Purpose and Boundary says what the matrix covers but gives no reason the team keeps one | A reason is given, but nothing says what was left untraced, or what ends the matrix | You can point at the sentence giving this team’s reason, the sentence naming what was left untraced, and the stated event that ends it, so a reader could tell from the file alone when to delete it |
| 2 | One named record | Nothing says whether this file or a tracker holds the links | A record is named, but the file does not say what it adds beyond the tracker, or what the tracker holds instead | You can point at the sentence naming where the links are authoritative and, where a tracker holds any of them, the sentence saying what this file adds that the tracker does not |
| 3 | Verified is defined | No definition of verified or covered appears, or the status words are used undefined | A definition exists, but a row could meet it on a linked test alone, with no statement of what the test had to do or what evidence is kept | You can point at the sentence, and say what evidence a verified row rests on beyond a linked test, so two people reading the same row would mark it the same way |
| 4 | Other means marked | A requirement checked by inspection, analysis or demonstration is shown as a gap, or is marked verified with no record of the check | A status for other means exists, but names no record of the check | You can point at a requirement checked by other means, see its status and the record named, or point at the sentence saying no requirement here is checked that way |
| 5 | Every requirement has a row | A requirement you pick from the source document that has no test cannot be found in the table | Rows exist for the requirements you check, but no rule says which statements of the source count as requirements, so you cannot tell whether others are missing | A requirement you pick from the source document that has no test has its own row with an empty Verified by cell, and the scope rule tells you which statements were counted |
| 6 | Links, not copies | Rows copy the requirement’s full text, or the Source cell is blank | Labels are short, but some Source cells point nowhere a reader can open | For any row you can open the Source cell’s target and find the requirement’s full text there, and the row carries only the source’s own identifier and a short label |
| 7 | Status follows definition | A row marked verified has an empty Verified by cell, or names a test that has not run | Verified rows name a test, but the evidence the definition asks for cannot be found from the row | Pick any verified row: you can follow its Verified by cell to a test result or an other-means record, and confirm the condition in Trace Conventions is met, without asking the keeper |
| 8 | Gaps have decisions | Coverage Gaps is blank, or lists gaps with no owner and no decision | Requirements with no verification have owners and decisions, but nobody checked the tests against the requirements | You can point at the line showing the tests were checked against the requirements, naming each test with no requirement or stating none was found and when; every gap has a named owner and a recorded decision |
| 9 | Current against changes | No owner, no update events or no reconciliation date appears, or the section says only “updated as needed” | An owner, events and a date are given, but the date is older than the latest change to the requirements or tests, or the only trigger is a calendar interval | You can point at a named owner, the events that force an update and a date last reconciled, and comparing that date with the newest change in the source documents or test records shows the table was reconciled after it |
| 10 | Implementation references open (full) | Cells say “done” or “in code”, or copy design text or ticket titles | References are present, but some point nowhere a reader can open, or a requirement with no reference is left unmarked | For any row you can open the design reference and the work reference; a requirement with neither is marked as a finding; or the section says N/A and names the tracker that is the record |
| 11 | Risk links by ID (full) | Risk likelihood or impact is copied into the table, or a risk has no register identifier | Register identifiers are used, but a risk with no controlling requirement, or a control with no verifying test, is left unmarked | Every risk row carries the register’s own identifier, the controlling requirement and the verifying test, a missing link is listed as a finding, and no score appears; or the section says N/A because no risk register exists |
| 12 | Dated baselines (full) | A baseline is mentioned with no date and no location | A date and an event are given, but the location or the reviewer is missing, or the copy’s own last-reconciled date is later than its snapshot date or is absent | For each snapshot you can open the copy at the stated location and see its date, the release or review, and who reviewed and approved it, and the copy’s own last-reconciled date is on or before the snapshot date |
Rubric scope by variant
Section titled “Rubric scope by variant”| Variant | Rows scored | Maximum | Threshold |
|---|---|---|---|
| lean | 1-9 | 18 | 13 |
| full | 1-12 | 24 | 17 |
Lean ships no Implementation Trace, Risk Control Trace or Baselines section, so rows 10 through 12 grade content it does not carry; score a lean matrix against rows 1 through 9 only. A lean matrix at 13 out of 18 clears a comparable bar to a full matrix at 17 out of 24.
Named anti-patterns
Section titled “Named anti-patterns”Each of these is a failure mode a source describes or this library infers. None is measured for how common it is.
- Copying requirement text into the matrix. NASA’s handbook says “do not duplicate requirements in the traceability matrix”. In this library’s reasoning, a copy that drifts from the source document turns one requirement into two. Fix: a short label and a link to the source, which owns the text.
- Treating a linked test as a verified requirement. A test case identifier in the Verified by cell shows that a test exists. NASA ties the completion of verification to “documented objective evidence of compliance with requirements or waiver” and to closure of nonconformance reports. A table full of green rows can hide requirements nobody verified. Fix: define verified in Trace Conventions in terms of evidence, and mark a row only when its cell meets that definition.
- Filling rows only for requirements that already have a test. The table then shows the team’s tests and not its requirements, and the requirement nobody verified is the one missing from it. Fix: build the rows from the source document’s requirements, and leave Verified by empty where nothing verifies the requirement.
- A forward-only matrix. Orcanos, a vendor selling a medical-device platform, says “An RTM that only maps forward is incomplete.” That is one vendor’s claim, in an audit frame, and no source read measures how common it is. A row-per-requirement table cannot show a test with no requirement, because that test has no row. Fix: check the tests against the requirements and record each orphan test in Coverage Gaps.
- Flagging a gap and leaving it. NASA’s guidance is that orphans are discussed by the project team and assurance personnel, and that a missing requirement found to be necessary is added. A gap with no owner and no decision is a list of worries. Fix: an owner and one decision per gap, and a one-line statement with a date when the list is empty. An empty list is only as good as the rows and the definition of verified behind it.
- Letting the table go stale. NASA warns that manual methods “do not include automatic updates”, so the project has to be diligent about keeping traces current. No source gives a calendar interval, and none says how fast a hand-kept table goes stale. Fix: a named owner, a list of events that force an update, and a date last reconciled, compared against the last change to the requirements or tests.
- A matrix kept for its own sake. A team that cannot say what it chose to trace, what it chose not to trace and when it will stop is keeping the table because tables are kept, and the file drifts into a second copy of whatever the tracker already says. Fix: write the three sentences in Purpose and Boundary, and if they cannot be written, drop the matrix.
- A requirement without an identifier. A statement with no ID has no row, so nothing can trace to it. That the loss goes unseen is this library’s inference. Fix: assign an identifier before the statement is traced, and record the decision in the table.
- Copying risk scores into Risk Control Trace (full). The score then lives in two places, and the copy goes stale the first time the risk register changes. Fix: reference the risk by its register identifier and leave likelihood, impact and owner in the register.
- Baselining a table that has not been reconciled (full). The snapshot preserves the stale table and gives it the weight of a reviewed record. Fix: reconcile first, record the reconciliation date, and then take the snapshot.
What this guide does not settle
Section titled “What this guide does not settle”Whether a matrix pays for itself outside regulated work, who may mark a row verified, what review interval to use, and what any standard requires are all unsettled by what was read. Companion sections 2, 3 and 6 state them.
No paired skill (yet)
Section titled “No paired skill (yet)”This bundle’s family contract states plainly that no governance skill exists in pm-skills today for any of
its members, so every member adopts pairs_with: [] until one does. Until a governance-side skill exists,
this template is filled by hand.
The artifacts
Section titled “The artifacts”requirements-traceability-matrix_template-lean.md · ~5,200 tokens
---title: "{{product_name}} Requirements Traceability Matrix"product: "{{product_name}}"owner: "{{owner}}"last_reconciled: "{{last_reconciled}}"status: "{{status}}"doc_type: requirements-traceability-matrixsize: leansource_template: requirements-traceability-matrixsource_template_version: 0.1.0---
<!--LEAN REQUIREMENTS TRACEABILITY MATRIX. The smallest matrix that is still a real one: what it coversand which place is the record, the rules that say what a link and a status mean, the table ofrequirements with what verifies each, the gaps in both directions, and the owner with a stated triggerfor updating it. This is what TestDevLab describes, requirement to test, read in both directions. Use itby default. To grow it into the full form (see requirements-traceability-matrix_template-full.md), ADDsections: Implementation Trace, Risk Control Trace and Baselines, after Coverage Gaps and before Upkeepand Review Trigger. Never rename or reorder the sections below, because the full variant is a strictsuperset of this one. The signal to scale up is a question the team owes an answer to that this tablecannot give: which design element implements a requirement, which control covers a risk, or what anauditor saw at a given release. That signal is this library's POSITION, and confidence in the lean andfull split is medium-low.
WHAT A REQUIREMENTS TRACEABILITY MATRIX IS. TestDevLab, a vendor, gives the narrowest definition: "ATraceability Matrix is a document that links and maps project requirements to their corresponding testcases." In this library's reading, it is a standing register, edited as requirements and tests change, not a record filed once.Traceability is a property, and the matrix is one way to hold it: InfoQ reports Brad Appleton writing"Traceability matrix on the other hand is one of the possible implementations of traceability." Thefirst use of the phrase is not established by anything read, and this template offers no origin story.
A LINKED TEST IS NOT A VERIFIED REQUIREMENT. NASA ties completion of verification to "documentedobjective evidence of compliance with requirements or waiver" and to "closure of all discrepancy andnonconformance reports". A test case ID in a Verified by cell shows that a test exists, not that therequirement was verified. Trace Conventions is where the matrix says which of the two its status wordsmean.
THIS TYPE IS CONTESTED, AND NOTHING READ SHOWS THAT KEEPING ONE IMPROVES OUTCOMES. InfoQ's 2008 roundupreports Alistair Cockburn writing "I have, in 30 years in the business, in industry, research and alittle bit in government, NEVER seen a traceability matrix pay off." That is a second-hand report of amailing-list post. This research found no named software team outside regulated or safety-critical workthat describes keeping one (high confidence that the search found none, medium that none exists). Amapping study of 63 papers says "strong empirical evidence from industry is needed to validate theimpact of traceability on maintenance and evolution". Keep a matrix only where the team has decided itowes a whole-set answer to the question of which requirement has no verification, and say so in Purposeand Boundary. Where a tracker already holds the links, that tracker is the record and this file shouldsay so. That recommendation is this library's POSITION. This matrix is also not NASA's requirementsverification matrix, which is a related artifact. No standard was read for this bundle, so nothing heresays what IEEE 829, IEEE 830, ISO/IEC/IEEE 29148, 21 CFR 820 or DO-178C requires.
HOW TO FILL THIS IN1. Read the comment under each heading: WHAT, WHY (with a companion pointer), ASK, GOOD, WEAK, TRAP; the three table sections also carry PRIORITY and ROW HINT.2. Replace each {{placeholder}}. Write the Purpose and Boundary and Trace Conventions sections before you fill a single row, because the rows mean nothing until the status words are defined. Give each requirement a short label and a link to its source document; never copy the requirement's text into the matrix.3. If a section does not apply, write "N/A" and one line of why, rather than deleting it. A reader cannot tell a deleted section from one nobody considered.4. A matrix is never finished, but before you share it: self-grade against requirements-traceability-matrix_guide.md, then DELETE every HTML comment. They are guidance, not content.-->
# {{product_name}} Requirements Traceability Matrix
## Purpose and Boundary
<!-- WHAT A short statement of the product and requirement set this matrix covers, what the team chose not to trace, who keeps it, and whether this file or a named tool is the record. WHY The sources disagree on what the matrix is: a table of requirements and test cases for TestDevLab, a wider chain for Visure, and for Jama, a vendor that sells the live replacement, the wrong form altogether. A reader needs to know which of these the file claims to be, so the section names the one place that is the record. NASA says a matrix "can be manually created and maintained, or may be a by-product of a requirements management tool". This is also where an unregulated team justifies the matrix to itself: what it chose to trace, what it chose not to, and when it will stop. The boundary statement is this library's POSITION, and no source read asks for one. Deep dive: requirements-traceability-matrix_companion.md section 3 (Purpose and Boundary) and section 6. ASK Which product, and which set of requirements, does this matrix cover, and what did the team choose not to trace? Who keeps it? Is this file the record, or does a tracker hold the links, and if so what does this file add? Why is the team keeping a matrix at all, and what will make it stop? GOOD "This matrix covers the offline-sync requirements of Harbor Notes, REQ-20 to REQ-27, and nothing else in the product. It is kept by the sync team's QA lead. This file is the record: the team's tracker holds tickets but no requirement-to-test links. Not traced: attachment sync, which is out of scope this quarter. The team keeps a matrix because a sync defect can lose a user's notes, and will stop when the sync rewrite ships and the tracker holds the links." WEAK "Requirements traceability matrix for the product." (names no requirement set, no record, no reason to keep one and no stop condition, so a reader cannot tell what the table claims to cover or whether it is still wanted) TRAP Keeping the table because tables are kept. A team that cannot say what it chose to trace, what it chose not to trace and when it will stop has no reason to keep a matrix, and the file will drift into a second copy of whatever the tracker already says. -->
**Product and requirement set covered:** {{covered_set}}
**Not traced:** {{not_traced}}
**Kept by:** {{owner}}
**The record:** {{record_of_truth}}
**Why this team keeps a matrix:** {{reason_for_matrix}}
**When it will stop:** {{stop_condition}}
## Trace Conventions
<!-- WHAT The identifier schemes for requirements and tests, the rule for what counts as a requirement, a statement of what counts as verified, how other means of verification are marked, how links that are not one-to-one are written, and the status vocabulary. WHY A table of links is only as meaningful as the rules that decide what a link means. No source read gives a template for this section, so its form is this library's POSITION, and the companion lists the supports. NASA's verification guidance ties completion to "documented objective evidence of compliance with requirements or waiver", together with "closure of all discrepancy and nonconformance reports". In this library's reading, a linked test case is therefore a claim that a test exists, and not a claim that the requirement was verified. The jamb documentation, a practitioner's tool guide, flags requirements "indicating they are verified by other means (inspection, analysis, demonstration)". Without such a status, in this library's reading, a requirement verified by inspection reads as a gap. NASA also says "the relationships identified in the matrix are not required to be one-to-one". Who may mark a row verified is not established by any source read, so a team that wants a rule states it as its own. Deep dive: requirements-traceability-matrix_companion.md section 3 (Trace Conventions). ASK How are requirements and tests numbered, and which statements count as requirements? In one sentence, what must be true before a row is marked verified: a test exists, a test passed, or evidence is attached and no defect against the requirement is open? How does the matrix mark a requirement checked by inspection, analysis or demonstration, and how are many-to-many links written? What does each status word mean, in one sentence each? PRIORITY The definition of verified is written before any status word is used, and it says what evidence a verified row rests on, not only that a test is linked. If the team says covered, define that word here: it can mean a linked test or a verified requirement. ROW HINT A good status row names the status and the condition that earns it, so two people reading the same row would mark it the same way. A weak row names a status with no condition, or defines it as the existence of a test. GOOD Requirement IDs: "REQ-n, assigned in the sync spec and not reused." Test IDs: "T-n, from the test management tool." Scope: "Every numbered sentence in sections 4 and 5 of the sync spec; the introduction and the glossary are not requirements." Verified: "A named test passed on a named build, the evidence is attached to the test record and no defect against the requirement is open." Other means: "A requirement checked by inspection carries the status Other means and names the review record." Links: "One requirement may list several tests and one test may list several requirements; both directions are written out." Status table rows: | No verification | Nothing is linked: no test and no other-means record | | Test linked | A test exists but has not yet run or has not passed | | Verified | The verified condition above is met | | Other means | Checked by inspection, analysis or demonstration, with the record named | WEAK Status table row: | Covered | Done | (does not say what evidence earns it, so a linked test and a verified requirement read the same) TRAP Letting covered mean that a test case ID sits in the Verified by cell. A table full of green rows can then hide requirements that were never verified, and nothing in the file shows it. -->
**Requirement identifiers:** {{requirement_id_scheme}}
**Test identifiers:** {{test_id_scheme}}
**What counts as a requirement:** {{scope_rule}}
**What counts as verified:** {{verified_definition}}
**Verified by other means:** {{other_means_rule}}
**Links that are not one-to-one:** {{cardinality_rule}}
| Status | What it means in this matrix ||---|---|| {{status_name}} | {{status_meaning}} |
## The Matrix
<!-- WHAT One row per requirement, with its identifier, a short label, its source, what verifies it, its status and any notes. WHY This table is the artifact. The other sections explain it, bound it or extend it. The columns draw on Visure's component list, "Requirements ID and Description", "Source of Requirements" and "Associated Test Cases", and on TestDevLab's requirement-to-test pair. Neither the Visure page nor the TestDevLab page shows a full table, so the column set is this library's assembly from partial lists and not any one source's shipped layout. The row holds a short label and not the requirement's full text, because NASA's handbook says "do not duplicate requirements in the traceability matrix". In this library's reasoning, a copy of the text drifts from the original. The source document owns the text, and the matrix owns the links. One matrix or several is a stated choice: NASA says "Traceability matrices can be included in the documents to which they apply", or "they can be combined into a single matrix covering higher level requirements, software requirements, design, code, and verification", and its systems handbook adds "This matrix could be divided into multiple matrices (e.g., one for each requirements document)". This template ships one table for the covered set. The table is read in both directions and is not kept as two: TestDevLab describes both checks on the same table, that "every requirement is covered by at least one test case, and conversely, that every test case is linked to a valid requirement". A table with one row per requirement cannot show a test that has no requirement, because that test has no row, so Coverage Gaps is where that half of the check lives; that reading is this library's inference. Deep dive: requirements-traceability-matrix_companion.md section 3 (The Matrix). ASK Which requirements are in scope, and does every one of them have a row, including those with nothing yet in the Verified by cell? Where does each requirement's source document say it, and does the row link there rather than copy it? Is this one table, or does the team split it by requirements document, and where are the other tables? PRIORITY Every requirement in scope has a row, including the ones with no verification. An empty Verified by cell is the reason the table exists, and a row for a requirement with no verification is a finding and not an error. ROW HINT A good row has a stable identifier, a label short enough to scan, a source a reader can open, every test or other means that verifies it, and a status that follows the definitions in Trace Conventions. A weak row copies the requirement's full text, leaves the source blank, or marks a status the Verified by cell does not support. GOOD | REQ-22 | Edits made offline are synced on reconnect | Sync spec 4.2 | T-31, T-32 | Verified | T-32 passed on build 118 with the log attached; no defect open | WEAK | 22 | sync | | T-31 | Covered | | TRAP Filling in rows only for the requirements that already have a test. The table then shows the team's tests, not its requirements, and the requirement nobody verified is the one that is missing from it. -->
| Requirement ID | Requirement (short label) | Source | Verified by | Status | Notes ||---|---|---|---|---|---|| {{requirement_id}} | {{requirement_label}} | {{requirement_source}} | {{verified_by}} | {{requirement_status}} | {{requirement_notes}} |
## Coverage Gaps
<!-- WHAT The requirements with no verification, the tests with no requirement, and what each gap's owner decided. WHY Both vendor sources name finding gaps as a purpose. TestDevLab lists "Identify Gaps: Highlight any requirements that do not have a corresponding test case." Visure's product description includes the ability to "detect missing links". The backward half matters too: Orcanos, a vendor selling a medical-device quality platform, says "An RTM that only maps forward is incomplete." Its frame is an audit of a medical device, and the claim is one vendor's. A gap that is only flagged has not been dealt with. NASA says orphans "need to be discussed by the project team and assurance personnel", and that "If they are determined to be necessary, any missing source requirements are added to the project." So the section asks what each gap's owner decided: add a test, add a requirement, mark the requirement as verified by other means, or remove the orphan. How a gap is displayed is not shown by any source read, so this section's layout is this library's POSITION. Do not read an empty list as a clean bill of health: it is only as good as the rows and the definition of verified behind it. That caution is this library's inference. Deep dive: requirements-traceability-matrix_companion.md section 3 (Coverage Gaps). ASK Which requirements have no test or other means of verification? Which tests point to no requirement? Who owns each gap, and what did they decide to do about it? PRIORITY Every gap has an owner and a decision. A gap with no owner and no decision is a list of worries. If the list is empty, say so in one line and say when the rows were last checked. ROW HINT A good row names the kind of gap, the identifier, an owner, and a decision that is one of the four above or a reasoned alternative. A weak row says that something is missing and names neither the identifier nor the owner. GOOD | Requirement with no verification | REQ-25 | Sync lead | Add a test: T-40 to be written before the beta | Beta is the next review point | WEAK | gap | sync stuff | team | look into it | | TRAP Flagging a gap and leaving it. A gap list that grows without decisions teaches readers to ignore it, and it makes the matrix look more complete than the verification behind it. -->
| Gap | ID | Owner | Decision | Notes ||---|---|---|---|---|| {{gap_kind}} | {{gap_id}} | {{gap_owner}} | {{gap_decision}} | {{gap_notes}} |
## Upkeep and Review Trigger
<!-- WHAT The events that require an update, the owner who makes it, the points at which the matrix is reviewed, and the date it was last reconciled. WHY A matrix that nobody maintains turns from a record into false assurance. NASA's guidance is "Assign responsibility for creating and maintaining the matrix to a project team member" and "Ensure a review of the matrix at major phases/key reviews of the project.", and it warns that manual methods "do not include automatic updates". The update rule comes from Jama, a vendor: a matrix "should be updated whenever requirements, test cases, or linked lifecycle artifacts change, not just on a calendar schedule." No source read gives a calendar interval, and this template does not invent one. Tools handle staleness with a mechanism, such as the content hash the jamb documentation describes. A hand-kept table has none, so this template's stand-in is an owner, a list of trigger events and a date last reconciled. No source read puts a last-confirmed field on a hand-kept row, so that form is this library's POSITION. Deep dive: requirements-traceability-matrix_companion.md section 3 (Upkeep and Review Trigger). ASK Who owns the upkeep? Which events force an update: a changed requirement, a new or retired test, a change to a linked ticket? At which reviews or releases is the whole table checked? When was it last checked against its sources? GOOD "Owner: the sync team's QA lead. Triggers: a requirement is added, changed or deferred; a test is added, retired or fails after passing; a change request touching sync is approved. Review points: before each beta and release review. Last reconciled: 2026-09-10, against sync spec v6 and the test management tool." WEAK "Updated as needed." (no owner, no trigger and no date, so a reader cannot tell whether the table is current or was current once) TRAP Setting a calendar interval and calling it a rule. Jama ties updates to change and NASA ties reviews to major phases and key reviews, and no source read gives an interval. A date older than the last trigger event is a warning that the table may be stale, however recently the calendar says it was reviewed. -->
**Owner:** {{owner}}
**Events that require an update:** {{update_triggers}}
**Review points:** {{review_points}}
**Date last reconciled:** {{last_reconciled}}requirements-traceability-matrix_template-full.md · ~7,050 tokens
---title: "{{product_name}} Requirements Traceability Matrix"product: "{{product_name}}"owner: "{{owner}}"last_reconciled: "{{last_reconciled}}"status: "{{status}}"doc_type: requirements-traceability-matrixsize: fullsource_template: requirements-traceability-matrixsource_template_version: 0.1.0---
<!--FULL REQUIREMENTS TRACEABILITY MATRIX. Everything in the lean variant, plus three sections for thewider links: Implementation Trace (each requirement to its design reference and its code or taskreference), Risk Control Trace (each risk to the requirement that controls it and the test thatverifies the control), and Baselines (dated snapshots of what a release or an audit saw). Reach for itwhen the team owes an answer the lean table cannot give: which design element implements thisrequirement, which control covers this risk, or what an auditor saw at a given release. Without such aquestion, the lean size is enough. That signal is this library's POSITION. Confidence in the lean andfull split is medium-low, because no source read shows a template that carries every link type, sothe split is this library's hypothesis. Risk Control Trace is the thinnest-sourced section: onemedical-device vendor and the FDA's reference to "risk analysis results".
This is a STRICT SUPERSET of requirements-traceability-matrix_template-lean.md: every lean sectionappears here in the same relative order, under the same names. If you are growing from lean, insertImplementation Trace, Risk Control Trace and Baselines after Coverage Gaps and before Upkeep and ReviewTrigger. Never rename or reorder the lean sections.
WHAT A REQUIREMENTS TRACEABILITY MATRIX IS. TestDevLab, a vendor, gives the narrowest definition: "ATraceability Matrix is a document that links and maps project requirements to their corresponding testcases." In this library's reading, it is a standing register, edited as requirements and tests change, not a record filed once.Traceability is a property, and the matrix is one way to hold it: InfoQ reports Brad Appleton writing"Traceability matrix on the other hand is one of the possible implementations of traceability." Thefirst use of the phrase is not established by anything read, and this template offers no origin story.
A LINKED TEST IS NOT A VERIFIED REQUIREMENT. NASA ties completion of verification to "documentedobjective evidence of compliance with requirements or waiver" and to "closure of all discrepancy andnonconformance reports". A test case ID in a Verified by cell shows that a test exists, not that therequirement was verified. Trace Conventions is where the matrix says which of the two its status wordsmean.
THIS TYPE IS CONTESTED, AND NOTHING READ SHOWS THAT KEEPING ONE IMPROVES OUTCOMES. InfoQ's 2008 roundupreports Alistair Cockburn writing "I have, in 30 years in the business, in industry, research and alittle bit in government, NEVER seen a traceability matrix pay off." That is a second-hand report of amailing-list post. This research found no named software team outside regulated or safety-critical workthat describes keeping one (high confidence that the search found none, medium that none exists). Amapping study of 63 papers says "strong empirical evidence from industry is needed to validate theimpact of traceability on maintenance and evolution". Keep a matrix only where the team has decided itowes a whole-set answer to the question of which requirement has no verification, and say so in Purposeand Boundary. Where a tracker already holds the links, that tracker is the record and this file shouldsay so. That recommendation is this library's POSITION. This matrix is also not NASA's requirementsverification matrix, which is a related artifact. No standard was read for this bundle, so nothing heresays what IEEE 829, IEEE 830, ISO/IEC/IEEE 29148, 21 CFR 820 or DO-178C requires.
HOW TO FILL THIS IN1. Read the comment under each heading: WHAT, WHY (with a companion pointer), ASK, GOOD, WEAK, TRAP; the six table sections also carry PRIORITY and ROW HINT.2. Replace each {{placeholder}}. Write the Purpose and Boundary and Trace Conventions sections before you fill a single row, because the rows mean nothing until the status words are defined. Give each requirement a short label and a link to its source document; never copy the requirement's text into the matrix.3. If a section does not apply, write "N/A" and one line of why, rather than deleting it. A reader cannot tell a deleted section from one nobody considered. Risk Control Trace in particular may be "N/A, no risk register exists for this product".4. A matrix is never finished, but before you share it: self-grade against requirements-traceability-matrix_guide.md, then DELETE every HTML comment. They are guidance, not content.-->
# {{product_name}} Requirements Traceability Matrix
## Purpose and Boundary
<!-- WHAT A short statement of the product and requirement set this matrix covers, what the team chose not to trace, who keeps it, and whether this file or a named tool is the record. WHY The sources disagree on what the matrix is: a table of requirements and test cases for TestDevLab, a wider chain for Visure, and for Jama, a vendor that sells the live replacement, the wrong form altogether. A reader needs to know which of these the file claims to be, so the section names the one place that is the record. NASA says a matrix "can be manually created and maintained, or may be a by-product of a requirements management tool". This is also where an unregulated team justifies the matrix to itself: what it chose to trace, what it chose not to, and when it will stop. The boundary statement is this library's POSITION, and no source read asks for one. Deep dive: requirements-traceability-matrix_companion.md section 3 (Purpose and Boundary) and section 6. ASK Which product, and which set of requirements, does this matrix cover, and what did the team choose not to trace? Who keeps it? Is this file the record, or does a tracker hold the links, and if so what does this file add? Why is the team keeping a matrix at all, and what will make it stop? GOOD "This matrix covers the offline-sync requirements of Harbor Notes, REQ-20 to REQ-27, and nothing else in the product. It is kept by the sync team's QA lead. This file is the record: the team's tracker holds tickets but no requirement-to-test links. Not traced: attachment sync, which is out of scope this quarter. The team keeps a matrix because a sync defect can lose a user's notes, and will stop when the sync rewrite ships and the tracker holds the links." WEAK "Requirements traceability matrix for the product." (names no requirement set, no record, no reason to keep one and no stop condition, so a reader cannot tell what the table claims to cover or whether it is still wanted) TRAP Keeping the table because tables are kept. A team that cannot say what it chose to trace, what it chose not to trace and when it will stop has no reason to keep a matrix, and the file will drift into a second copy of whatever the tracker already says. -->
**Product and requirement set covered:** {{covered_set}}
**Not traced:** {{not_traced}}
**Kept by:** {{owner}}
**The record:** {{record_of_truth}}
**Why this team keeps a matrix:** {{reason_for_matrix}}
**When it will stop:** {{stop_condition}}
## Trace Conventions
<!-- WHAT The identifier schemes for requirements and tests, the rule for what counts as a requirement, a statement of what counts as verified, how other means of verification are marked, how links that are not one-to-one are written, and the status vocabulary. WHY A table of links is only as meaningful as the rules that decide what a link means. No source read gives a template for this section, so its form is this library's POSITION, and the companion lists the supports. NASA's verification guidance ties completion to "documented objective evidence of compliance with requirements or waiver", together with "closure of all discrepancy and nonconformance reports". In this library's reading, a linked test case is therefore a claim that a test exists, and not a claim that the requirement was verified. The jamb documentation, a practitioner's tool guide, flags requirements "indicating they are verified by other means (inspection, analysis, demonstration)". Without such a status, in this library's reading, a requirement verified by inspection reads as a gap. NASA also says "the relationships identified in the matrix are not required to be one-to-one". Who may mark a row verified is not established by any source read, so a team that wants a rule states it as its own. Deep dive: requirements-traceability-matrix_companion.md section 3 (Trace Conventions). ASK How are requirements and tests numbered, and which statements count as requirements? In one sentence, what must be true before a row is marked verified: a test exists, a test passed, or evidence is attached and no defect against the requirement is open? How does the matrix mark a requirement checked by inspection, analysis or demonstration, and how are many-to-many links written? What does each status word mean, in one sentence each? PRIORITY The definition of verified is written before any status word is used, and it says what evidence a verified row rests on, not only that a test is linked. If the team says covered, define that word here: it can mean a linked test or a verified requirement. ROW HINT A good status row names the status and the condition that earns it, so two people reading the same row would mark it the same way. A weak row names a status with no condition, or defines it as the existence of a test. GOOD Requirement IDs: "REQ-n, assigned in the sync spec and not reused." Test IDs: "T-n, from the test management tool." Scope: "Every numbered sentence in sections 4 and 5 of the sync spec; the introduction and the glossary are not requirements." Verified: "A named test passed on a named build, the evidence is attached to the test record and no defect against the requirement is open." Other means: "A requirement checked by inspection carries the status Other means and names the review record." Links: "One requirement may list several tests and one test may list several requirements; both directions are written out." Status table rows: | No verification | Nothing is linked: no test and no other-means record | | Test linked | A test exists but has not yet run or has not passed | | Verified | The verified condition above is met | | Other means | Checked by inspection, analysis or demonstration, with the record named | WEAK Status table row: | Covered | Done | (does not say what evidence earns it, so a linked test and a verified requirement read the same) TRAP Letting covered mean that a test case ID sits in the Verified by cell. A table full of green rows can then hide requirements that were never verified, and nothing in the file shows it. -->
**Requirement identifiers:** {{requirement_id_scheme}}
**Test identifiers:** {{test_id_scheme}}
**What counts as a requirement:** {{scope_rule}}
**What counts as verified:** {{verified_definition}}
**Verified by other means:** {{other_means_rule}}
**Links that are not one-to-one:** {{cardinality_rule}}
| Status | What it means in this matrix ||---|---|| {{status_name}} | {{status_meaning}} |
## The Matrix
<!-- WHAT One row per requirement, with its identifier, a short label, its source, what verifies it, its status and any notes. WHY This table is the artifact. The other sections explain it, bound it or extend it. The columns draw on Visure's component list, "Requirements ID and Description", "Source of Requirements" and "Associated Test Cases", and on TestDevLab's requirement-to-test pair. Neither the Visure page nor the TestDevLab page shows a full table, so the column set is this library's assembly from partial lists and not any one source's shipped layout. The row holds a short label and not the requirement's full text, because NASA's handbook says "do not duplicate requirements in the traceability matrix". In this library's reasoning, a copy of the text drifts from the original. The source document owns the text, and the matrix owns the links. One matrix or several is a stated choice: NASA says "Traceability matrices can be included in the documents to which they apply", or "they can be combined into a single matrix covering higher level requirements, software requirements, design, code, and verification", and its systems handbook adds "This matrix could be divided into multiple matrices (e.g., one for each requirements document)". This template ships one table for the covered set. The table is read in both directions and is not kept as two: TestDevLab describes both checks on the same table, that "every requirement is covered by at least one test case, and conversely, that every test case is linked to a valid requirement". A table with one row per requirement cannot show a test that has no requirement, because that test has no row, so Coverage Gaps is where that half of the check lives; that reading is this library's inference. Deep dive: requirements-traceability-matrix_companion.md section 3 (The Matrix). ASK Which requirements are in scope, and does every one of them have a row, including those with nothing yet in the Verified by cell? Where does each requirement's source document say it, and does the row link there rather than copy it? Is this one table, or does the team split it by requirements document, and where are the other tables? PRIORITY Every requirement in scope has a row, including the ones with no verification. An empty Verified by cell is the reason the table exists, and a row for a requirement with no verification is a finding and not an error. ROW HINT A good row has a stable identifier, a label short enough to scan, a source a reader can open, every test or other means that verifies it, and a status that follows the definitions in Trace Conventions. A weak row copies the requirement's full text, leaves the source blank, or marks a status the Verified by cell does not support. GOOD | REQ-22 | Edits made offline are synced on reconnect | Sync spec 4.2 | T-31, T-32 | Verified | T-32 passed on build 118 with the log attached; no defect open | WEAK | 22 | sync | | T-31 | Covered | | TRAP Filling in rows only for the requirements that already have a test. The table then shows the team's tests, not its requirements, and the requirement nobody verified is the one that is missing from it. -->
| Requirement ID | Requirement (short label) | Source | Verified by | Status | Notes ||---|---|---|---|---|---|| {{requirement_id}} | {{requirement_label}} | {{requirement_source}} | {{verified_by}} | {{requirement_status}} | {{requirement_notes}} |
## Coverage Gaps
<!-- WHAT The requirements with no verification, the tests with no requirement, and what each gap's owner decided. WHY Both vendor sources name finding gaps as a purpose. TestDevLab lists "Identify Gaps: Highlight any requirements that do not have a corresponding test case." Visure's product description includes the ability to "detect missing links". The backward half matters too: Orcanos, a vendor selling a medical-device quality platform, says "An RTM that only maps forward is incomplete." Its frame is an audit of a medical device, and the claim is one vendor's. A gap that is only flagged has not been dealt with. NASA says orphans "need to be discussed by the project team and assurance personnel", and that "If they are determined to be necessary, any missing source requirements are added to the project." So the section asks what each gap's owner decided: add a test, add a requirement, mark the requirement as verified by other means, or remove the orphan. How a gap is displayed is not shown by any source read, so this section's layout is this library's POSITION. Do not read an empty list as a clean bill of health: it is only as good as the rows and the definition of verified behind it. That caution is this library's inference. Deep dive: requirements-traceability-matrix_companion.md section 3 (Coverage Gaps). ASK Which requirements have no test or other means of verification? Which tests point to no requirement? Who owns each gap, and what did they decide to do about it? PRIORITY Every gap has an owner and a decision. A gap with no owner and no decision is a list of worries. If the list is empty, say so in one line and say when the rows were last checked. ROW HINT A good row names the kind of gap, the identifier, an owner, and a decision that is one of the four above or a reasoned alternative. A weak row says that something is missing and names neither the identifier nor the owner. GOOD | Requirement with no verification | REQ-25 | Sync lead | Add a test: T-40 to be written before the beta | Beta is the next review point | WEAK | gap | sync stuff | team | look into it | | TRAP Flagging a gap and leaving it. A gap list that grows without decisions teaches readers to ignore it, and it makes the matrix look more complete than the verification behind it. -->
| Gap | ID | Owner | Decision | Notes ||---|---|---|---|---|| {{gap_kind}} | {{gap_id}} | {{gap_owner}} | {{gap_decision}} | {{gap_notes}} |
## Implementation Trace
<!-- WHAT Each requirement mapped to its design reference and its code or task reference. WHY The wider definitions reach past tests. Visure's component list includes "Development Tasks and Deliverables", and its page defines backward traceability as linking "design elements, code, and test cases back to the original requirements." NASA's guidance allows a single matrix to cover design, code and verification. Visure's page is a vendor page that shows no table, so the column set is this library's assembly. Drop this section when a tracker already holds the links: InfoQ reports that Appleton gets traceability from commits, "When he does a commit he associates the name or id of the story with the commit." A table that restates links the tracker already holds creates a second record that can disagree with the first; that reasoning is this library's own. Deep dive: requirements-traceability-matrix_companion.md section 3 (Implementation Trace). ASK Where does the design for each requirement live? Where does the work live: which ticket, module or deliverable? Does a tracker already hold these links, and if so is this section needed at all? PRIORITY Each cell is a reference to where the design or the work lives, not a copy of its text. A row whose requirement has no design or work reference is a finding, not an error. ROW HINT A good row gives a design reference and a work reference a reader can open. A weak row says "done" or "in code" with no place to look. GOOD | REQ-22 | Sync design note, section 3 | Ticket SYNC-114; module sync-queue | | WEAK | REQ-22 | in the design doc | done | | TRAP Copying ticket titles or design text into the cells. The copy drifts from the original, and the matrix ends up as a second place where the same facts are kept. -->
| Requirement ID | Design reference | Code or task reference | Notes ||---|---|---|---|| {{requirement_id}} | {{design_reference}} | {{work_reference}} | {{implementation_notes}} |
## Risk Control Trace
<!-- WHAT Each risk, the requirement that controls it, and the test that verifies the control. WHY To show whether a control is stated by a requirement and checked by a test. That purpose is this library's reasoning and not a sourced claim. This is the section with the thinnest general support. Orcanos, a vendor selling a medical-device platform, names a failure, "The RTM has no column for risk controls.", and says "The RTM is where risk management and design control intersect." That is one vendor in a medical-device frame, and its claims may not transfer to other software. The FDA's 2002 guidance asks for a trace to "risk analysis results", and the copy read carries a note that 21 CFR part 820 was amended, so the guidance predates the current rule. No general software source read asks for a risk-control trace, so the section is optional by design and sits in the full variant only. The matrix owns no risk scoring: likelihood, impact and owner stay in the risk register, and this section only links a risk to the requirement and test that control it. That line is this library's POSITION. Deep dive: requirements-traceability-matrix_companion.md section 3 (Risk Control Trace) and section 8. ASK Does a risk register exist for this product? Which risks depend on a specific requirement and a specific test? What is each risk's register identifier? PRIORITY Reference each risk by its register identifier and leave the scoring to the register. A risk with no controlling requirement, or a control with no test, is a finding. ROW HINT A good row gives the register identifier, a short label, the requirement that controls the risk and the test that verifies the control. A weak row restates the risk's score or its description in full. GOOD | RK-07 | Offline edits overwritten on reconnect | REQ-22 | T-32 | | WEAK | Data loss | high | we test it | | TRAP Copying the risk's likelihood and impact into this table. The score then lives in two places, and the copy goes stale the first time the register changes. -->
| Risk ID | Risk (short label) | Controlling requirement | Verifying test | Notes ||---|---|---|---|---|| {{risk_id}} | {{risk_label}} | {{controlling_requirement_id}} | {{control_test_id}} | {{risk_notes}} |
## Baselines
<!-- WHAT Dated snapshots of the matrix, recording what a release or an audit saw. WHY A standing register changes, and a reviewer sometimes needs the state at one moment. Orcanos says "At each formal design review gate, baseline the RTM." and "Record which version of the RTM was reviewed and approved." Jama, while arguing that the matrix should be live, describes "a specific versioned artifact that captures those links at a point in time." This library reads the two together as a standing register that is snapshotted for an audit. That reading is this library's POSITION, not something either vendor states. The section is full-only because a team without audits or formal releases has little use for it, and a snapshot of an out-of-date table preserves the out-of-date table; that reasoning is this library's own. Deep dive: requirements-traceability-matrix_companion.md section 3 (Baselines). ASK Which release or review does each snapshot belong to? On what date was it taken? Where is the snapshot file kept, and who reviewed and approved that version? PRIORITY Every snapshot names its date, the release or review it belongs to, and where the copy is kept. A snapshot taken without first reconciling the table records a stale state. ROW HINT A good row gives the date, the event, the matrix version, a location a reader can open and the reviewer. A weak row says "baselined" with no date and no location. GOOD | 2026-09-12 | Beta review | v14 | snapshots/rtm-v14.md in the sync repo | Sync lead and QA lead | WEAK | recently | review | latest | the repo | | TRAP Taking a snapshot of a table that has not been reconciled since the last change. The snapshot then preserves the stale table and gives it the weight of a reviewed record. -->
| Date | Release or review | Matrix version | Where the snapshot is kept | Reviewed and approved by ||---|---|---|---|---|| {{baseline_date}} | {{baseline_event}} | {{baseline_version}} | {{baseline_location}} | {{baseline_approved_by}} |
## Upkeep and Review Trigger
<!-- WHAT The events that require an update, the owner who makes it, the points at which the matrix is reviewed, and the date it was last reconciled. WHY A matrix that nobody maintains turns from a record into false assurance. NASA's guidance is "Assign responsibility for creating and maintaining the matrix to a project team member" and "Ensure a review of the matrix at major phases/key reviews of the project.", and it warns that manual methods "do not include automatic updates". The update rule comes from Jama, a vendor: a matrix "should be updated whenever requirements, test cases, or linked lifecycle artifacts change, not just on a calendar schedule." No source read gives a calendar interval, and this template does not invent one. Tools handle staleness with a mechanism, such as the content hash the jamb documentation describes. A hand-kept table has none, so this template's stand-in is an owner, a list of trigger events and a date last reconciled. No source read puts a last-confirmed field on a hand-kept row, so that form is this library's POSITION. Deep dive: requirements-traceability-matrix_companion.md section 3 (Upkeep and Review Trigger). ASK Who owns the upkeep? Which events force an update: a changed requirement, a new or retired test, a change to a linked ticket? At which reviews or releases is the whole table checked? When was it last checked against its sources? GOOD "Owner: the sync team's QA lead. Triggers: a requirement is added, changed or deferred; a test is added, retired or fails after passing; a change request touching sync is approved. Review points: before each beta and release review. Last reconciled: 2026-09-10, against sync spec v6 and the test management tool." WEAK "Updated as needed." (no owner, no trigger and no date, so a reader cannot tell whether the table is current or was current once) TRAP Setting a calendar interval and calling it a rule. Jama ties updates to change and NASA ties reviews to major phases and key reviews, and no source read gives an interval. A date older than the last trigger event is a warning that the table may be stale, however recently the calendar says it was reviewed. -->
**Owner:** {{owner}}
**Events that require an update:** {{update_triggers}}
**Review points:** {{review_points}}
**Date last reconciled:** {{last_reconciled}}requirements-traceability-matrix_example.md
---title: "Saved Views for Dashboards Requirements Traceability Matrix"product: "Saved Views for Dashboards (Acme Analytics, Reporting Platform Modernization)"owner: "Anjali Rao (QA Lead, Reporting)"created: "2026-07-20"last_reconciled: "2026-07-20"status: "active"related: - "../prd/prd_example.md (Saved Views for Dashboards PRD, version 0.3.0; the requirement set this matrix covers)" - "../test-plan/test-plan_example.md (the test plan whose risk-ranked areas the Verified by column names)" - "../test-summary-report/test-summary-report_example.md (the final 2026-07-17 report whose counts and acceptances every status below rests on)" - "../test-case/test-case_example.md (TC-047, the case that traced to a risk and to no requirement)" - "../bug-report/bug-report_example.md (DEF-2291, closed, found by TC-047 step 4)" - "../risk-register/risk-register_example.md (the register that owns R-05, R-06 and R-02)" - "../acceptance-criteria/acceptance-criteria_example.md (the default-view criteria that DEF-2298 misses)" - "../epic/epic_example.md (the SV-n stories that serve each requirement, which is why Implementation Trace is N/A)" - "../incident-postmortem/incident-postmortem_example.md (the DEF-2291 postmortem, dated 2026-07-16)" - "../sdd/sdd_example.md (the design document that refers to the PRD security NFR)" - "../kpi-dashboard/kpi-dashboard_example.md (the dashboard that carries R-06's 500ms budget)" - "../issue-log/issue-log_example.md (ISS-12, the same view-list problem under Dana Osei)"doc_type: requirements-traceability-matrixsize: fullsource_template: requirements-traceability-matrixsource_template_version: 0.1.0---
> **Worked example.** A filled `requirements-traceability-matrix`, full variant, for the Reporting> Platform Modernization program at the fictional Acme Analytics. It is the same program, and the same> Saved Views feature, that the [`prd`](../prd/prd_example.md), [`test-plan`](../test-plan/test-plan_example.md),> [`test-case`](../test-case/test-case_example.md),> [`test-summary-report`](../test-summary-report/test-summary-report_example.md),> [`bug-report`](../bug-report/bug-report_example.md) and> [`risk-register`](../risk-register/risk-register_example.md) examples already cover.> `requirements-traceability-matrix` joined the `governance-docs` family as its eighth member on 2026-11-05> under> [ADR 0084 (the requirements traceability matrix joins governance-docs as an eighth member)](../../docs/internal/decisions/0084-requirements-traceability-matrix-joins-governance-docs-as-an-eighth-member.md).>> This matrix is dated **2026-07-20**, the first working day after the test summary report closed on> 2026-07-17. Everything it cites existed by then, including the PRD, the epic, the product backlog, the> test plan, TC-047, the bug report, the DEF-2291 postmortem, the risk register and the test summary report.> It does not cite the> `change-log` example of 2026-08-17. That log records that the PRD stayed at version 0.3.0, which is why> this matrix labels the PRD's non-functional rows itself and does not edit the PRD.>> **What this example does and does not show.** It shows the shape of a matrix and the discipline of its> status words. It does not argue that an unregulated team should keep one. This bundle's research found no> named software team outside regulated or safety-critical work that describes keeping its own matrix, and> nothing it read shows that a matrix improves outcomes. The team below makes a narrow, time-limited choice> for one feature. NASA's software handbook describes the same bargain for projects with small budgets:> value-based tracing that focuses the effort on the most important requirements. Read the Purpose and> Boundary section first, and read the stop condition as part of the document.>> Counts, builds and defect states come from the test summary report as it stood at 2026-07-16 17:00 UTC.> Every figure marked "illustrative" is made up for the example. All names, paths and dates not already> fixed by the library's Acme Analytics thread are illustrative too. Acme is not regulated, and nothing in> this example is a regulation or a contract that requires a matrix.
# Saved Views for Dashboards Requirements Traceability Matrix
## Purpose and Boundary
**Product and requirement set covered:** Saved Views for Dashboards, phases 1 and 2. The set is the sixfunctional requirements FR-1 to FR-6 and the four rows of the non-functional table in the[PRD](../prd/prd_example.md), version 0.3.0. That is ten rows, and nothing else in the product.
**Not traced:**
- The dashboard permissions service. Platform owns and tests it, and this matrix traces only this feature's use of it.- Adoption (R-04 on the register). It is measured on the KPI dashboard after launch, and a test cannot verify it.- The legacy report flow, which the standing regression suite covers.- The PRD's goals, UX states, success metrics, instrumentation list and rollout plan. The scope rule below explains why.- Individual cases. Only five case IDs appear below. Every other row names a test plan area, because the run record already lists the cases.
**Kept by:** Anjali Rao, QA Lead for Reporting.
**The record:** This file is the record for one thing: the link from each PRD requirement to the cases orthe test plan area that verify it, and the status of that link. The PRD owns the requirement text. Theproduct backlog owns the work, and each story in the Saved Views epic names the requirement it serves.The sprint 14 run record in the test tool owns the case results. The risk register owns every risk score.This file copies none of those. What it adds is the whole-set view. The team's reasoning is that arequirement with no test has no test record to carry a link to it, so a whole-set table is where thatrequirement shows up.
**Why this team keeps a matrix:** R-05 (a shared view exposing personal data to a viewer withoutentitlement) sits above the program's PII appetite line. It has been with the steering group since2026-07-14, waiting on one decision: fund a stronger entitlement control, or accept the residual formally.Shared-view entitlement was the first-ranked area in the test plan, and TC-047 (a restricted viewer openinga shared view) found DEF-2291 (the aggregate that disclosed data across an entitlement boundary). Yet nodocument pairs the PRD's permissions row with the cases that test it, because that row has no identifier.The steering group will decide on the strength of how well the control is verified. So the team wants onepage that says which requirement states the control and what verifies it. The matrix covers one feature,holds ten rows and is kept by hand, because one person can reconcile ten rows against four documents in asitting.
**When it will stop:** When two things have both happened: the steering group has decided R-05, and thegradual rollout in the PRD has reached 100 percent. At the next reconcile after that, Anjali Rao takes alast baseline, marks this file closed in Upkeep and stops updating it. The exception is a line from PriyaNair, added here with a reason, saying it continues. Any finding the team still needs moves into the testplan for the next release.
## Trace Conventions
**Requirement identifiers:** FR-1 to FR-6 are the PRD's own, unchanged. The PRD's non-functional table hasno identifiers. This matrix labels its four rows NFR-1 to NFR-4 in the PRD's table order: Performance,Permissions, Reliability, Accessibility. The labels exist only in this file. The PRD stays at version0.3.0, and Priya Nair decides whether its next revision adopts them.
**Test identifiers:** TC-nnn, as numbered in the test tool, for a case that is individually tied to arequirement or a risk. The thread holds five of them: TC-046, TC-047, TC-048, TC-052 and TC-053. Every otherrow names the test plan's risk-ranked area, with the planned, executed and passed counts the test summaryreport gave on 2026-07-16 at 17:00 UTC. Those counts are a reading at one moment, and the run record is thelive source.
**What counts as a requirement:** Each row of the PRD's functional requirements table and each row of itsnon-functional requirements table, at version 0.3.0. Goals, non-goals, UX states, success metrics,instrumentation and rollout steps are not counted, because the team did not treat them as statements toverify one by one. Stories and acceptance criteria are not counted either. They refine a requirement, andthey are reached through the requirement they serve.
**What counts as verified:** Every case the test plan planned for the requirement ran and passed. The passwas on the release build, 2.3.2, or on the earlier build 2.3.1 with a named person's acceptance of thatearlier result for 2.3.2 recorded in the test summary report. Every defect raised against the requirementis closed, or is accepted in that report by a named person. A case that exists does not make a rowverified. A case that is blocked, not run or failing leaves the row Partly verified or Test linked.
**Verified by other means:** None of the ten requirements rests on a review or an analysis alone, so theOther means status below is defined and unused. The one review this release held, Sam Okafor's security review of 2026-07-10, gated phase 2 testing. It isnamed in NFR-2's notes and does not stand in for the tests.
**Links that are not one-to-one:** One requirement may name several cases or areas, and one case or areamay serve several requirements. NFR-2 names four cases. The View-list load performance area serves bothFR-2 and NFR-1. A link is read in both directions from the Verified by cells.
| Status | What it means in this matrix ||---|---|| No verification | Nothing is linked: no case, area or other-means record names the requirement. It is a finding and goes to Coverage Gaps. No row holds it at version 1 || Test linked | A case or area is named, but this matrix holds no passing result for it. The case has not run, or its result has not been read into this file || Partly verified | Some planned cases passed and at least one did not. It failed, was blocked or was never run. The row says which, and who accepted what is open || Verified | The definition above is met || Other means | Checked by review, analysis or demonstration, with the record named. No row holds it at version 1 || Deferred | The requirement is not built in this release. The row carries the reason, and the status goes when the requirement is built |
## The Matrix
| Requirement ID | Requirement (short label) | Source | Verified by | Status | Notes ||---|---|---|---|---|---|| FR-1 | Save the current dashboard state as a named view | [PRD](../prd/prd_example.md), functional table | The save cases in the sprint 14 run record. The test plan names no area for them | Test linked | Must. The plan and the report count by area, so no pass count for save can be read from them. Open in Coverage Gaps || FR-2 | List saved views and switch between them in one action | [PRD](../prd/prd_example.md), functional table | The list and switch cases in the run record. Switch timing is also measured in the View-list load performance area | Test linked | Must. Same open item as FR-1. The timing result is carried under NFR-1. DEF-2304 (S2, open) is a finding against R-06's 500ms budget, which no numbered requirement carries, so this matrix does not count it against FR-2 || FR-3 | Set one view as the default for a dashboard | [PRD](../prd/prd_example.md), functional table | Default-view resolution area: 14 planned, 13 executed, 13 passed, 1 blocked. Run on 2.3.1 only | Partly verified | Must. The blocked case needs a mid-session access revocation that Platform has not scheduled. 9 of the 14 come from the [acceptance criteria](../acceptance-criteria/acceptance-criteria_example.md) and 5 from test design. Priya Nair accepted the 2.3.1 result for 2.3.2. DEF-2298 (S3, open) misses the criterion that a view with a deleted filter names it; see NFR-3 || FR-4 | Share a view so other permitted users of the dashboard can select it | [PRD](../prd/prd_example.md), functional table | Shared-view entitlement area: the 12-combination permission matrix, 12 of 12 passed on 2.3.2, including TC-046 (fully entitled recipient) | Verified | Should. The matrix ran in full on 2.3.2 after DEF-2291 (S1, closed). See NFR-2 || FR-5 | Rename and delete views the user owns | [PRD](../prd/prd_example.md), functional table | Rename and delete area: 8 planned, 8 passed. Run on 2.3.1 only | Verified | Should. Priya Nair accepted the 2.3.1 result for 2.3.2 in the report's residual-risk table || FR-6 | Indicate when a shared view's dashboard has changed in a way that affects it | [PRD](../prd/prd_example.md), functional table | None | Deferred | Could. It was not built in this release, so the test plan deferred it with the requirement and there is nothing to test || NFR-1 | Switching to a view renders in under 1s at p95 (illustrative target) | [PRD](../prd/prd_example.md), non-functional table | The view-switch cases in the View-list load performance area, run on the pre-prod replica at 2.3.2. p95 0.74s (illustrative) | Verified | The 500ms view-list budget is not in this PRD row. It belongs to risk R-06, and DEF-2304 (S2, open) is against that budget, not against NFR-1. See Risk Control Trace || NFR-2 | A shared view never exposes data the recipient could not already access | [PRD](../prd/prd_example.md), non-functional table | TC-046, TC-047 and TC-048, the three entitlement partitions of the permission matrix, all passed on 2.3.2 | Verified | ID assigned in this file, see Coverage Gaps. DEF-2291 was found by TC-047 step 4 and closed in 2.3.2. TC-047's rendered-UI step is manual, once per release, and Sam Okafor accepted that in the report. TC-053 was added on 2026-07-15 to guard the row-count badge on the same path, and its own result is in the run record, not here. The 2026-07-10 security review preceded the tests and is not counted as verification || NFR-3 | A view that references a deleted filter loads what it can and flags the rest | [PRD](../prd/prd_example.md), non-functional table | TC-052, in the Stale-field degradation area: 9 executed, 8 passed, 1 failed | Partly verified | TC-052 raised DEF-2277 on 2026-07-08, fixed and re-verified on 2.3.2. The failed case is DEF-2298 (S3, open): the notice gives a count and does not name the filters. Priya Nair accepted it on 2026-07-16 and deferred it || NFR-4 | The Views control is keyboard-operable and screen-reader labeled (WCAG 2.2 AA) | [PRD](../prd/prd_example.md), non-functional table | Accessibility area: 14 planned, 11 executed, 10 passed, 1 failed, 3 not run | Partly verified | NVDA on Windows is complete. The VoiceOver on macOS pass never started and has no owner or date. DEF-2286 (S3, open) was accepted in writing by Priya Nair on 2026-07-16. DEF-2284 was fixed in 2.3.2 |
## Coverage Gaps
On 2026-07-20 the five case IDs this matrix names were checked against the requirements, the backwarddirection. TC-047 was the one whose own record names a risk and no requirement. TC-046, TC-048, TC-052and TC-053 were placed by what each verifies. Cases named only by area were not checked backward, which isthe third row below.
| Gap | ID | Owner | Decision | Notes ||---|---|---|---|---|| Requirement with no identifier, so no row | The PRD's Permissions row, now NFR-2 | Anjali Rao | Assigned NFR-2 in this file and traced TC-046, TC-047 and TC-048 to it. The PRD is unchanged, and Priya Nair decides whether its next revision adopts NFR-1 to NFR-4 | The row is in the PRD at version 0.3.0. The design document refers to it as the PRD security NFR. The bug report and the postmortem do not point at it, and it had no number to point at. SV-9 (an acceptance criterion for what a partially entitled recipient sees inside an aggregate) is still proposed, and should name NFR-2 as its parent when written || Test with no requirement | TC-047 | Anjali Rao | Resolved by the row above. TC-047 now traces to NFR-2 as well as to R-05 | Its record names R-05 and states that no acceptance criterion covers it. It names no requirement || Requirement with no recorded result | FR-1 and FR-2 | Anjali Rao | Read the save and the list-and-switch cases in the run record, copy their IDs and results here, and move each row to Verified or leave it with the defect named. Done before the next reconcile | The test plan scopes save, list and switch but ranks no area for them. This is also the backward check for the cases named only by area || Planned cases never executed | FR-3 (1 blocked case) and NFR-4 (3 VoiceOver cases) | Dana Osei for the access-revocation slot, and Marta Reyes for the VoiceOver reviewer | Both asks are open in the test summary report's Impediments section. This row mirrors their state and holds no separate decision | The rows stay Partly verified until the cases run || Risk with no controlling requirement | R-02 and R-06 | Priya Nair | No requirement added. R-02's control is a Data Eng procedure, verified by the migration area. R-06's 500ms budget is a metric on the KPI dashboard. The test summary report tracks it as DEF-2304 under Marcus Bell, and the issue log tracks it as ISS-12 under Dana Osei, who owns R-06 on the register. A row here would be one more place for the same facts | Revisit when DEF-2304 is closed |
## Implementation Trace
N/A. The product backlog is the record. Each story in the [Saved Views epic](../epic/epic_example.md) namesthe requirement it serves: SV-2 serves FR-1, SV-3 serves FR-2, SV-4 serves FR-3, SV-5 serves FR-4, SV-6serves FR-5 and SV-7 serves FR-6. The design document holds the design. A table here would restate linksthe backlog already holds, and the two could disagree.
## Risk Control Trace
| Risk ID | Risk (short label) | Controlling requirement | Verifying test | Notes ||---|---|---|---|---|| R-05 | Shared view exposes personal data to a viewer without entitlement | NFR-2 | TC-047, with TC-046 and TC-048 | Escalated on the register and with the steering group. The register owns the score. The evidence under NFR-2 is API assertions plus one manual rendered-UI step per release || R-06 | View-list load exceeds its budget as views accumulate | None numbered. The budget is stated in the register and in exit criterion 5, not in the PRD | View-list load performance area. The 150-view case failed (DEF-2304, open) | A finding, recorded in the last Coverage Gaps row || R-02 | Saved-view configs lost in the migration to the new schema | None in the PRD. The control is a Data Eng procedure | Config migration and rollback area: 18 planned, 18 passed. Run on 2.3.1 only | A finding by design, recorded in the same row |
R-01 (vendor licence), R-04 (adoption) and R-07 (the telemetry opportunity) are left out. No row in thecovered set controls them.
## Baselines
| Date | Release or review | Matrix version | Where the snapshot is kept | Reviewed and approved by ||---|---|---|---|---|| 2026-07-20 | Pack sent to the steering group for the R-05 decision | v1 | snapshots/saved-views-rtm-v1.md in the Reporting QA wiki (illustrative path) | Anjali Rao reconciled it, and Marta Reyes reviewed and approved it |
This is the first version and the only snapshot. It was reconciled on the day it was taken, so it does notpredate the table. The next snapshot is taken when the steering group decides R-05, so that the decisionrecords which evidence it saw.
## Upkeep and Review Trigger
**Owner:** Anjali Rao.
**Events that require an update:** The PRD moves past version 0.3.0, or a change request is approved againstit. The 12-combination permission matrix is re-run, or the entitlement area gains a case. DEF-2298, DEF-2286 orDEF-2304 changes state. The register's row for R-05, R-06 or R-02 changes.
**Review points:** Before the steering group decides R-05, and at each step of the PRD's gradual rollout(10, 50 and 100 percent). The whole table is checked against its sources at each one.
**Date last reconciled:** 2026-07-20, against the PRD at version 0.3.0, the test plan approved on2026-07-06, the test summary report final on 2026-07-17, and the register's rows for R-05, R-06 and R-02.The run record in the test tool was not read, which is why FR-1 and FR-2 read Test linked. Those two rowsare the first thing to reconcile next.Provenance
Section titled “Provenance”The reasoning, the history and every source, in the repository:
- Companion - the long-form argument: why these sections, where the sources disagree, and what the bundle refuses to claim
- History - what changed in this bundle, and when
- Research log - every source consulted, with what each one actually supports
- Catalog metadata - the machine-readable record this page is generated from
Catalog record: 8 sections across 1 format(s), methodology generic, typically owned by BA / QA.