Triage the review queue
This page shows you how to read the Quality Review Queue and decide what each open task needs.
Quality Admin and Corporate Admin. Other administrators do not see this surface.
The queue is at Quality > Review queue. Open tasks land here automatically as people author content and as source documents change.
The five task types
Every task in the queue is one of five types. The filter chips at the top switch between them, and each chip shows a count. One queue rather than five places to look is the point.
| Filter chip | Task type | What it means | What you do |
|---|---|---|---|
| Pending versions | version_approval | An author submitted a new training version. It is pending review and not live. | Compare and approve or reject. See Approve content. |
| Pending documents | document_approval | A controlled document or a translation of one is waiting for a quality decision. | Approve and sign, reject, or make effective, from the card itself. |
| Responses to grade | assessment_grading | Somebody wrote a response to a scenario assessment and it needs a person to judge it. | Read the response, set a level per criterion, record the grade. |
| Outdated sources | source_outdated | A controlled document a training relies on changed, so the training may be out of date. | Open the task and record a resolution. |
| Recert decisions | recert_decision | A re-certification decision is waiting for a human call. | Open the task and record a resolution. |
The All chip shows every open task. Use Refresh to reload the list after you act on a task elsewhere.
The queue is scoped by role. A Quality Admin sees the tasks addressed to the Quality Admin role. A Corporate Admin sees everything. This scoping is enforced in the database, so a task you cannot act on never appears for you.
Read a task card
Each card shows:
- The training name and version (for example
v3.2). - A type badge (Version approval, Source updated, or Recertification decision).
- An AI-generated badge when the pending version was drafted with AI.
- The triggered by document, for an outdated-source task.
- The created timestamp.
Act on a pending version
For a Pending versions task, the card offers:
- Open training to view the material.
- Preview to read the whole version as it will be delivered.
- Compare to see a side-by-side delta between the pending version and the current active version.
- Approve and Reject.
Preview and Compare answer different questions and you normally want both. Compare tells you what changed. Preview tells you what you are approving, which is more than the body text.
The preview opens a wide review workspace. Sources and the signature requirement stay visible in a side column. Use Content to read the learner-facing material and Assessment to inspect the questions and reviewer-only grading detail. This separation keeps a long policy or question set readable without nested scrolling.
- The sources. When the training is anchored to a controlled document, that document is named with its document code, its revision and its lifecycle state. Below it, every source attached to the version is listed, including external references such as the release of a compliance program. Each internal source is a link that opens the exact document revision the training cites, so you can read what the material rests on without leaving the decision. If the list cannot be loaded you are told so and offered a retry, because a version with no sources cannot be approved and an empty list is something you act on.
- The objectives and the opening guide text.
- The content, rendered exactly as the learner reads it. A document with high-fidelity page renders opens on those page renders, the same way it opens for the learner.
- The video, when the version carries one, played the way the learner gets it.
- The assessment, in full. For a scored quiz that means every question, every option with the correct one marked, the pass mark and the number of attempts allowed. For a scenario assessment it means the reflection questions plus the grading criteria.
- Whether finishing the version requires an electronic signature.
The workspace loads in stages. The learning content is fetched when Preview opens. The assessment and its grading data load only when you choose Assessment. Long assessments are divided into pages of ten questions. Hovering a source loads a short rendered preview of that document; click the source to open the exact cited revision.
The correct answers and the grading criteria sit inside a tinted panel carrying a Reviewers only marker. They exist so you can judge the assessment, and the person taking it never sees them. The grading criteria for a scenario are visible to the quality function only, so an administrator outside it sees the questions without the criteria.
The full approval flow, including the change rationale and the source requirement, is on its own page: Approve content.
If you authored the pending version, the Approve and Reject buttons are disabled and the card tells you another Quality Admin must review it. This is segregation of duties, and the server enforces it even if the buttons were somehow bypassed. See Segregation of duties.
Grade a scenario assessment
A Responses to grade task carries somebody's written answer to a scenario. There is no percentage here and no right option they should have picked. You are judging what a person would do.
The card shows three things.
- What they wrote, prompt by prompt.
- Publisher guidance, the risks a competent response notices unprompted and the actions it contains. This is the grading key. The person being assessed never sees it, which is what stops the exercise becoming a reading comprehension test.
- Your judgement, one level per criterion, with the descriptor for the level you pick shown underneath so you can check it says what you mean.
As you set levels, the card tells you whether the result clears the role. Recording the grade is disabled until every criterion has a level, because an unanswered criterion is not the same as a failed one.
If the response is yours, the levels are disabled and the card says so. The server refuses the grade as well, so this holds even if the buttons were bypassed. See Segregation of duties.
How the outcome is decided
Most scenarios require every criterion to reach a stated level. There is no averaging, and that is deliberate. Somebody who spots the anomaly and approves the batch anyway has demonstrated the exact failure the oversight role exists to prevent, so a strong score elsewhere cannot compensate for it.
Your optional comment goes to the person. Where somebody did not clear, it is the most useful part of the record, because it tells them what to work on.
What gets recorded
The levels you set, the outcome, your name and the timestamp. The record is permanent: a regrade adds a new record rather than editing this one, so the history of a competence decision stays intact.
Grading competence is a judgement about a human being, and under the EU AI Act an automated system doing it would itself be high-risk. Better Comply does not offer one. The grader of record is always you.
Resolve a source-outdated or recertification task
For an Outdated sources or Recert decisions task, click Resolve. You choose a resolution and write a reason of at least 10 characters:
| Resolution | When to choose it | Effect |
|---|---|---|
| Cosmetic change, no impact | The document change does not affect the training subject. | The version is marked reviewed and the outdated flag clears. Campaigns can use it again. |
| Re-approved after review | You compared the content against the new document version and the training is still valid. | The outdated flag clears. |
| Re-certify users | The change is significant enough to warrant re-certification. | The flag stays. Your next step is to create a new training version and a campaign targeted at users who already completed it. |
Your decision and your reason are written to the audit log permanently. There is no way to undo a recorded resolution; choose the right one and document why. See Read the audit log.
If you created the training version a task points at, you cannot resolve that task yourself; another Quality Admin must record the decision. This mirrors the approval rule for pending versions and is enforced by the server.