Coverage or chains? — which requirement view answers your question¶
Audience: anyone doing design control who has opened the Traceability workspace and found two entries that both seem to be about requirements and evidence.
The workspace carries two requirement views, and they are easy to mistake for one another because both read requirements, both read links, and both report a number over a total. They answer different questions, and the difference is not a matter of detail:
- Requirement coverage returns a verdict per requirement. Is each requirement covered?
- Requirement chains returns a path per stakeholder requirement. Does this need reach evidence, and where does it break?
A project can be entirely green in one and full of gaps in the other, with both answers correct. §4 explains why that is not a contradiction.
1. The one-line difference¶
| Requirement coverage | Requirement chains | |
|---|---|---|
| Question it answers | "Is every requirement covered by something?" | "Does every stakeholder need reach executed evidence?" |
| Unit of the answer | one requirement | one end-to-end path |
| Rows are | every requirement item in the project | the project's stakeholder requirements only |
| How far it looks | one hop, in isolation | every hop, start to end |
| Shape on screen | a flat covered / not-covered list | expandable rows that branch and have unequal depth |
| Headline | "{{count}} of {{total}} requirements covered" | "{{satisfied}} of {{total}} stakeholder requirements satisfied", and "{{count}} fully verified" beside it |
| Take-away artifact | the requirements registry on screen | the flat traceability matrix as a CSV download |
Both are derived reads. Neither stores a verdict of its own, neither can be edited, and neither can be made to look better than the links in the documents make it — you change what they say by changing the documents (see trace links).
2. Requirement coverage — the verdict per requirement¶
Open it when the question is "have we left a requirement standing with nothing pointing at it?"
A requirement counts as covered when it carries at least one inbound Verifies or
Satisfies link. That is the whole rule, and it is deliberately a low bar: it asks whether
anything at all answers the requirement, not whether the answer is finished.
How to read the page:
- The headline and bar — covered over total, for the requirement kinds currently in view.
- Filter by kind — ask the question of one kind at a time (functional, non-functional, stakeholder, safety, whatever this project declares; see key rules). The headline follows the filter, so "3 of 4 covered" under a filter is a statement about those four only.
- The gap list first — the requirements not yet covered, enumerated. This is the working list; it is put above the covered list because it is the part you act on.
- The covered list — each requirement with its evidence counted by class: verified ×N (an executed, released report), planned ×N (a verification designated — a plan, or a report not yet released) and satisfied ×N (answered by other requirements rather than by evidence). The covered verdict does not distinguish them; the chips do, so a requirement covered only by planned evidence is visible as such rather than hidden inside a green number.
- The requirements registry — every defined key with its kind, priority, defining document and version, link counts, and a text excerpt. A key touched by a suspect link is flagged here, with the hint to re-attest it on its own document.
What this view cannot tell you. Because it looks one hop out, a covered requirement may still sit
on a path that goes nowhere. A stakeholder requirement covered by a Satisfies link is covered —
even if the software requirement that satisfies it was never verified by anything. That is not a
defect in the verdict; it is the boundary of the question. The next section is where that boundary is
crossed.
3. Requirement chains — the path, end to end¶
Open it when the question is "can I walk from a stakeholder need to executed evidence, out loud, in front of an auditor?"
Rows are the project's stakeholder requirements. Expanding one shows the software requirements that satisfy it, and each of those shows its verification evidence — the motion an auditor performs. Chains branch and have unequal depth (one need may feed several software requirements, some of which end without a verification hop), which is why this is expandable rows and not a rectangular grid.
Each chain carries roll-up flags, and they are honest and non-suppressible:
- Not satisfied — no software requirement satisfies this need at all.
- Satisfied, unverified — a satisfying software requirement carries no verification evidence. Satisfied and verified are different claims and are never averaged into one.
- Planned only — the chain is satisfied and verification is designated, but every piece of evidence on it is still a plan, not an executed report. A quiet amber state: the honest middle between unverified and clean.
- Suspect on path — some link on the chain is suspect because a document it depends on changed. The path is not trustworthy until it is re-attested on its source document.
The headline reports satisfied of total and fully verified side by side, plus the planned-only and suspect counts where there are any. Read both numbers: a project can satisfy every need and verify none of them, and the first number alone would call that finished.
Download matrix (CSV) exports the flat form — one row per complete path, need key through software key to evidence document and version, with the per-hop suspect state and the roll-up flag. That file is what auditors mean when they ask for "the traceability matrix"; the screen is for working, the CSV is for handing over.
4. Why both answers can be right at once¶
Take one project holding two requirement items: a stakeholder need, and one software requirement that
Satisfies it. The software requirement carries one inbound Verifies link, and it comes from a
verification plan — a document that designates the verification but has not been executed.
- Coverage reports 2 of 2 covered, with no gap list at all. Both requirements carry an inbound link, which is exactly the question it asks. Its evidence chips do say planned ×1 on the software requirement, so the state is on the page — but the verdict is green.
- Chains reports 1 of 1 satisfied, 0 fully verified, and flags the chain planned only.
Nothing in either answer is wrong. Coverage is asking about attachment and chains is asking about completion, and a project between those two states is where most of design control actually lives.
The practical habit: work the coverage gap list to make things attached, then work the chain flags to make them finished. They are two stages of the same job, in that order.
5. When one of the two is not there¶
The Requirement chains tile appears only where the project actually speaks the stakeholder level — where it declares that kind or already holds such items. A supplier project that is handed requirements and verifies against them has no stakeholder lane, and a permanent 0 of 0 there would be a reproach for something it was never asked to do.
Requirement coverage is never hidden this way. A project that writes requirements and never verifies them has a real gap in every lane, so that view stays and stays loud.
6. A third entry that is neither¶
Reference coverage sits in the same workspace and is about documents, not requirement items: it checks the project's own citation rules ("every record cites its governing SOP") and appears only where such a rule was configured. If you are looking for requirement-level answers, it is not the view you want — see the traceability workspace §6.
For what requirement items, keys and kinds actually are, and how a project declares which of them it uses, start at trace items.
Governing decisions: ADR-0077 (requirement items as keyed blocks inside controlled documents, and
§6's item-level coverage read — the per-requirement verdict), ADR-0086 (the end-to-end chain view,
its non-suppressible roll-up flags, and the flat path export), ADR-0080 (the traceability workspace —
why these are two sub-pages of one hub, and the permission that gates them), ADR-0088 (planned vs
verified evidence, derived at the read layer — the class chips and the planned-only flag), ADR-0085
(one trace vocabulary: why verifies and satisfies are trace-link claims and document citations are
a separate axis). Regulatory frame: IEC 62304 §5.1.1 and §5.7 (software development planning and
verification), ISO 13485 §7.3.9 (design and development files), EU MDR Annex I §17 (electronic
programmable systems).