Who holds each function here, and what may it do¶
Audience: quality managers preparing for an audit; anyone who needs to know who is responsible for what in a project.
The authority view answers one question per project: who holds each function here, and what may that function actually do? It is derived — it renders live staffing and live policy data and stores nothing of its own. That is the whole design: it cannot drift out of date, because it is the same data the system enforces on.
1. Where it is¶
Authority is a top-level entry in the toolbar's navigation (in the more menu at narrow widths). It is not an admin surface: any member of a project can open it.
The page reads the project context from the toolbar — the same switcher every other view uses. With All projects selected it shows a project picker rather than an aggregate, because a role's holders and its grants only mean anything inside one project. Its heading names the project (code and name), the owning organization, and as at the moment it was read.
Top right: Download snapshot — the export described in §5.
2. What it shows¶
One block per role activated in the project, in name order:
- Holders — the people staffed in that role here, named as they are named everywhere (their display name, or their email while an invitation is still unbound), each with their account status. A role nobody holds says "nobody" rather than showing an empty line.
- The accountable holder — the one holder, if any, marked accountable for the role's duties, carrying the same star the admin surface uses (see staffing & assignments §3 and §7). It confers no extra rights whatsoever.
- Permits — what holding the role permits. These are labels, not buttons: nothing here is clickable, because nothing here is an action. Hover one and it tells you what that permission actually means — the text comes from the permission's own definition inside the system, so it cannot drift out of step with what is enforced. A few read "RESERVED: granting this changes nothing today", and that is the truth about them rather than an omission: they exist in the catalog but no surface enforces them yet.
- May do — the functional authority: what the role may do in this project's document lifecycle, per document type: Submit for review / Approve / Release / Revise & periodic review / Revoke / Cancel, with the number of approvals required where the policy attaches one. These are the review policies rendered, not a second claim about them.
- Gap flags — see §3.
3. The two gap flags¶
- "Unstaffed — no active holder" — a review policy in this project grants this role authority, and no active person holds it. That authority can be exercised by nobody, and documents relying on it cannot pass their gates. It is the same fact the project's setup check reports, taken from that check's own derivation, so the two surfaces can never tell different stories. This is the loud flag.
- "No accountable holder named" — the role has holders, but none is named accountable, so no individual is answerable for its duties. This one is deliberately quieter: it is an unanswered question, not a hole (see staffing & assignments §6).
A role that is activated, ungranted and unheld carries neither flag: it is simply idle, and flagging it would be noise.
A deactivated holder is shown, with their status — and the role is still counted unstaffed, because only active people staff a role. That is exactly what makes the flags useful the moment someone leaves.
4. "Not disclosed" is different from "nothing granted"¶
The functional-authority section is not readable by everyone. The per-type policy composition ("this type grants Approve to the Quality Reviewer, two approvals") is governed by the review-policy configuration permission wherever it appears, and this view honours that boundary rather than widening it.
So the payload states the fact explicitly instead of leaving it to be inferred from an empty list:
- a reader with the configuration permission sees the May do tables;
- a reader without it sees everything else — roles, holders, the accountable-holder star, permissions, both flags — and is told, in a note at the top of the card, that the grants were omitted: "Functional authority (per-type grants) is not disclosed to your permissions. Roles, holders and permissions below are complete."
An evidence surface must not conflate "nobody is granted anything" with "you may not see who is granted what". Everything else in the view is readable by any member of the project — and deliberately so: a standard that requires authorities to be communicated within the organisation is not satisfied by a system where nobody can see what the roles around them may do.
5. The snapshot, and the evidence pair¶
Download snapshot exports the view as a point-in-time snapshot: a PDF named
authority-snapshot_<PROJECT-CODE>_<date>.pdf, carrying a summary table (Role | Holders | Primary
(accountable) | Gaps) and then one section per role — holders, permissions, and the functional-authority
table where it is disclosed. The file explains in its own header what it is, because an evidence file separated
from the tool that made it must say so.
Three properties make it usable as evidence:
- It is attributed. Exporting is recorded as an audit event, like every other derived export.
- It is the same form as the rest of the evidence. PDF, through the same render path as the dossier and the audit pack — what an auditor receives from this tool always looks like one system.
- It is honest about what it omits. A snapshot pulled by someone without the configuration permission carries a banner saying the functional-authority tables were left out and who can produce a complete one.
The snapshot answers "who holds what, now". It deliberately does not answer "what changed since last quarter" — that question has its own answer on this page, and it does not need two snapshots to compare — §6 does.
The evidence pair. The snapshot is the occupancy half. The other half is the hierarchy chart — a controlled document living in the QMS project, drawn as a draw.io diagram so it renders both on screen and in the PDF export. That chart is deliberately de-personalised: it draws functions and their relations — line reporting, staff positions, functional-authority annotations — and never a person's name. Consequence: staff turnover never touches the chart. It changes only when the organisation changes, which is rare and exactly what a controlled release with review and acknowledgement is for. "Who holds each function" is answered by the snapshot; never by the drawing.
Pair them as a recurring obligation — "review the org chart" — whose completion record carries the freshly exported snapshot as its attachment (see recurring obligations). Releasing the record is the execution; the snapshot is what makes it evidence.
The house cadence is quarterly. Tying it to the management review instead is tempting and too coarse: a management review that happens twice a year leaves up to six months in which the chart and the actual occupancy can diverge with nothing noticing. Quarterly keeps the drift small enough that a review is a glance rather than an investigation, and it gives you four dated occupancy records a year to hang the change list (§6) against. Pick a longer interval only if you can say why.
Who runs it is a role, not a hope. A recurring obligation names the responsible role (or roles) that own its completion — typically the Quality Manager — and the periodic sweep raises the task for that role's holders. That is the whole reason to model this as an obligation rather than a calendar reminder: the review becomes an assigned duty with a completion record, visible in the inbox of whoever holds the role, including after the person holding it changes.
Point the chart at the practice, not at a file. Give the chart document one sentence naming where its occupancy half lives — for example: "Current occupancy is evidenced by the quarterly authority-snapshot records in this scope." Deliberately not a filename: the snapshot is a new record every quarter, so a named file would be stale within three months and the chart would need revising to fix a pointer. The record series is the durable reference.
6. "What changed since…" — the governance change list¶
The snapshot tells you the state now. The question that actually comes up in a review is what changed since last time, and that one is answered here rather than by diffing two files.
Under the occupancy panel, Governance changes in this project takes a window — the last month, the last quarter, or a date you pick — and lists what happened to this project's governance inside it: who was staffed, who was unstaffed, who was made accountable, and which review policies were configured. Each line reads when · what · who did it, followed by what the recorded act actually changed.
It loads when you ask for it, not when the page opens: the audit trail is long and interleaves everything that ever happened here, so reading it back is a cost you should only pay for the question you came with.
Why a change list and not a "diff". A diff needs two stored states, and storing occupancy snapshots to compare would create a second source of truth about who holds what — the exact drift this whole surface exists to avoid. A change list needs only the audit trail, which already exists and is already the record. It is also better evidence: a net diff hides churn (someone assigned and unassigned inside the window looks like nothing happened), while the list shows every act with its actor and its timestamp.
Two boundaries, stated on the page itself.
- It is this project's governance, not everything that affects it. A person deactivated for the whole installation, a role activated centrally, or an edit to what a role permits all change what this project's authority means — and none of them is a change to this project, so none appears here. Widen the window on those with the installation's own trail under Administration › Activity.
- Nothing is replayed. The list shows recorded acts, as recorded. It never reconstructs "the state as of that date" from them, because an audit trail is not an event store and a confidently-wrong history is worse than none. If a true point-in-time state is ever needed, that is the day stored snapshots earn their keep — with a real reason, not as a side effect.
7. What it deliberately is not¶
- Not an org chart. Reporting lines are nowhere in the role system, and no chart concept ever grants, implies or inherits a permission. Reporting is not permission — the role system's answers are audit-proof precisely because authority comes only from an explicit grant.
- Not installation-wide. Every read is per project, which keeps the access gate a real tenant boundary. There is no cross-project roll-up today.
- Not a place to edit anything. It renders; staffing is changed on the user administration surface, authority in the review policies.
One edge worth knowing: a role staffed both at the installation level and inside the project can show two accountable holders in the project's view — the "at most one" bound holds per level, holders resolve downwards, and the view shows every accountable holder it finds rather than inventing a tie-break. It does not arise in a normally provisioned project.
Governing decisions: ADR-0098 (org chart × role system — the de-personalised chart, the derived authority view, the quarterly snapshot record, the primary bit; its snapshot-format and change-list clarifications carry the PDF default and the "changes since…" shape), ADR-0096 (the review policies this view renders as functional authority, and the disclosure boundary it honours), ADR-0089 (recurring obligations — the cadence that turns the snapshot into evidence), ADR-0054 (responsibility is a role, not a person), ADR-0074 (PDF export — why the chart is a draw.io diagram and not Mermaid). Regulatory frame: ISO 13485 §5.5.1 (responsibilities and authorities defined, documented and communicated), §5.5.2 (an individual management representative).