Training record

A training record is a piece of evidence — usually one row per person, per document, per version — that proves a specific individual was trained on a specific revision of a controlled document on a specific date, by a named qualified trainer, was assessed competent by a recorded check, and is currently signed off to execute on it. The record is per-person, not per-class. A roster sign-in sheet from a classroom session is not, on its own, a training record; it is an input to the training record.

The record is the answer to the inspector's question 'show me that the operator who ran this batch was qualified on the current version of this SOP.' An empty answer is a finding. A vague answer ('everyone was trained back in March') is a finding. A spreadsheet that does not reconcile to the QMS document register is a finding. A record that proves training on v3 of an SOP while the operator was running v4 is a finding that often escalates beyond the training matrix into the broader CAPA system.

Every regulated regime carries a personnel-qualification rule. The wording varies; the operational requirement is identical.

The pattern is identical across all of them: identify the training need, deliver the training, assess understanding, record it per-person, retain the record, and refresh when the underlying document changes. The auditor's job is to walk that loop on a sampled SKU and see whether it produces the documented evidence.

A training record that survives an audit contains at minimum the following fields. Missing fields are not blank cells in a spreadsheet — they are findings.

• Person — full name, employee or contractor ID, current role, current job grade.
• Document — title, document number, exact revision identifier or hash.
• Training type — read-and-understand, classroom, on-the-job (OJT), assessed competency, or refresh.
• Trainer — named, qualified to deliver this specific document or topic; trainer's own training record must show the train-the-trainer credential.
• Date trained — the date the training event occurred.
• Method of competency check — quiz score with pass threshold, observed demonstration with assessor sign-off, signed acknowledgement of read-and-understand, or attendance-plus-quiz combination.
• Date competent — when the person is signed off as qualified to perform the activity (may differ from date trained if a probationary or supervised period is required first).
• Next refresh date (if a periodic refresh cadence is required by the underlying document, the change-control classification, or local QMS policy).
• E-signature of the trainee acknowledging completion (Part 11 §11.50 compliant — printed name, timestamp, meaning of signing).
• E-signature of the trainer or assessor (where the training type requires assessor sign-off).
• Link to the source training material version (the slide deck, video, OJT checklist or quiz bank used).
• Audit-trail entry — every training event, refresh, deferral and disqualification is an immutable trail row.

Not all training needs the same depth. A change to the contact details on a phone-list SOP needs an acknowledgement. A new dispense procedure for a high-potency API needs assessed competency under direct observation. Quality systems typically classify training into four levels, with the level chosen on the basis of risk rather than convenience.

1. Acknowledgement (read-and-understand). The operator reads the document and signs a Part-11 acknowledgement that they have read it and understand it applies to them. Appropriate for low-risk procedural changes — phone lists, organisational reshuffles, minor formatting updates that do not change the operation. Single-signature, no assessor.
2. Quiz. The operator reads the document and completes a short knowledge check with a pass threshold (typically 80%). Failed quizzes route to remediation (re-read plus re-test, or escalation to OJT). Appropriate for revisions that materially change the procedure but do not introduce new physical hazards.
3. On-the-job training (OJT). The operator performs the activity under direct supervision of a qualified trainer using a structured checklist; the trainer signs off each checklist item before the operator is signed off as qualified. Appropriate for new operations, new equipment, or revisions affecting safety-critical steps.
4. Assessed competency. A qualified independent assessor observes the operator performing the activity end-to-end and signs off competency against a defined assessment criteria sheet. Appropriate for high-risk operations — sterile manufacturing, high-potency dispense, radiopharmaceutical handling, aseptic process simulation, complex equipment operation.

The choice should be driven by the risk classification of the change (typically captured in the change-control assessment) and by the inherent risk of the underlying activity. An inspector will notice when a high-risk task is signed off on read-and-understand alone, and will treat that as a CAPA-worthy gap in the training-needs analysis.

The single biggest failure mode in training systems is the SOP-revision drift. The SOP moves from v3 to v4. The training matrix still says 'trained on v3.' Operators continue to execute on v4 because the printout at the workstation is now v4. The training record is silently stale, and the inspector — who routinely walks from the workstation printout to the training record — finds the gap in minutes.

The correct pattern: every controlled-document effective-version transition triggers a re-training requirement for everyone in the document's training scope. Until the new version is acknowledged (at the appropriate depth), the person is not qualified to execute on it. The system, not the trainer, must enforce this — because the trainer has no realistic way to chase a hundred operators through a hundred SOP updates per quarter.

Three sub-patterns recur. First, an effective-date in the future: the document is approved today but takes effect in 14 days; the system queues re-training tasks now with a due date of the effective date. Second, an immediate effective-date with a grace period: rare and only used for safety-critical changes; the kiosk shows a banner on the affected task and operators must acknowledge before the next execution. Third, withdrawal of a document: the system disqualifies everyone from executing on the withdrawn document immediately and surfaces any tasks still routing to the old document for re-routing.

The principle V5 Ultimate enforces — and the principle most regulators implicitly expect — is that an operator who is not currently qualified on the current effective version of a required document cannot start a task that references that document. The gate is hard, not advisory. There is no 'override and continue' button for routine production work; the only escape paths are supervisor sign-on with a documented operational reason (e.g. emergency trainer-led OJT in progress), or an explicit deferral by QA with a deviation record.

The hard-gate is more demanding than the advisory model used by most legacy LMS-only systems, where the operator sees a warning and clicks through. The reason for the hard-gate is that the advisory model has a known failure mode: under production pressure the warning becomes wallpaper, click-through becomes habitual, and the training matrix becomes meaningless. The hard-gate forces the conversation to happen at the supervisor or QA level rather than silently at the operator level.

• Day-one enforcement — the gate is on from the first day the document goes live; there is no soft-launch period during which the gate is advisory.
• Per-document, per-version — the gate is keyed on the operator's acknowledgement of the current effective version, not on the operator's acknowledgement of any version of the document.
• Per-task — the gate runs on task start, on resumption after a break, and on any task that references the document indirectly through a parent assembly or sub-recipe.
• Supervisor override produces a deviation — the override path requires a supervisor e-signature, a reason from a controlled list, and writes a deviation record routed to QA for review.
• QA deferral is a controlled record — QA can defer a re-training requirement for a named individual for a defined period with a documented justification; the deferral itself is a controlled record subject to audit.

1. Operator performed a task on SOP v4 with training only on v3. The most common training-related 483 observation. Fix: kiosk hard-gate on current effective version.
2. Training matrix lists 'all personnel' without naming them. The matrix is a policy statement, not evidence. Fix: per-person records with named scope, not role-only or department-only scope.
3. No competency check beyond a signature. The record says 'trained' but the depth was inadequate for the risk. Fix: change-control classification drives the training-depth decision; read-and-understand is not the default.
4. Trainer not qualified to deliver the training. The trainer's own training record does not show a train-the-trainer credential. Fix: trainer qualification is itself a training requirement with its own record.
5. Training records held by HR but not visible to operations or QA. The matrix is on a separate system; reconciliation is manual. Fix: training is part of the QMS, not a separate HR system.
6. Re-training not triggered on minor revision changes that materially affected the procedure. The change was classified as minor; the procedure changed in a way operators did not notice. Fix: change-control includes a 'does this affect training' assessment with QA sign-off.
7. Roster sign-in sheets used as primary evidence. A classroom roster shows attendance, not competency, and does not survive an audit on its own. Fix: classroom attendance is an input to a per-person record that also captures competency verification.
8. Deferred training never resolved. QA defers a re-training task for a specific operator for a defined period; the deferral expires; nobody notices. Fix: deferrals are time-limited records with auto-expiry that re-engages the hard-gate.

• 21 CFR 211.25 (US pharma) — qualification, training continuing basis with sufficient frequency, documented.
• 21 CFR 211.34 (US pharma) — consultants performing duties under Part 211 must have documented education, training and experience.
• 21 CFR 820.25 (US medical device) — procedures for identifying training needs; awareness of defects from improper performance.
• 21 CFR 111.12 / 111.14 (US dietary supplements) — personnel qualified; records retained 1 year past last use of MMR or BPR.
• 21 CFR 117.4 (US food / FSMA) — qualified individual; PCQI training requirements explicitly defined.
• EU GMP Part I Chapter 2 §2.10–2.13 — initial and continuing training; effectiveness periodically assessed.
• EU Annex 1 §2 (sterile manufacturing) — additional training requirements for personnel entering grade A/B cleanrooms; periodic re-qualification cited.
• ISO 9001:2015 §7.2 / ISO 13485:2016 §6.2 — competence with documented evidence and effectiveness assessment.
• ISO 22000:2018 §7.2 — food-safety competence and training documentation.
• 21 CFR Part 11 §11.50 — every training acknowledgement is a signed electronic record subject to the same Part 11 manifestation and binding rules as any other signing.
• ICH Q10 §3.2.3 — pharmaceutical quality system; resources including human resources are part of the management responsibility for the PQS.

• Current-version coverage — share of in-scope personnel with acknowledgement on the current effective version of each in-scope document. World-class: 100% as of the effective date. Below 95% means the kiosk gate or the re-training fan-out is failing.
• Overdue-training rate — share of personnel with at least one in-scope document overdue. World-class: zero on production-critical documents. Trending upward means the fan-out is fast but the operator follow-through is slow.
• Kiosk hard-block rate — share of task starts that were blocked on training. World-class: very low single digits and decreasing — operators are pre-training before the effective date. Spikes after a release indicate the fan-out timing or the operator notification path needs tightening.
• Supervisor-override rate — share of task starts allowed by supervisor override on a documented operational reason. World-class: near zero on routine production. Above 5% means the hard-gate is being routinely bypassed and the training process is failing.
• Competency-depth alignment — share of training events where the depth (read / quiz / OJT / assessed) matches the change-control risk classification. World-class: 100%. Mismatches indicate the classification step is being skipped or under-rated.
• Trainer qualification currency — share of training events delivered by a trainer with a current train-the-trainer credential on the document being trained. World-class: 100%.
• Effectiveness check — share of re-training events that include a post-training effectiveness check (deviation rate, error rate, refresher quiz result trend). World-class: 100% for OJT and assessed; not required for read-and-understand.
• Inspection findings on 211.25 / 820.25 / 111.12 / EU GMP §2.10–2.13 — mature sites: zero. Any finding here is a structural process gap, almost always traceable to one of the eight failure modes above.

1. Document control and training are one feature, not two. Every controlled document carries a training requirement (none / read-and-understand / quiz with pass threshold / OJT / assessed competency) and a training scope (by role, by team, by site, by explicit named-user list).
2. On document approval through the two-person e-sig path, the new version moves to status 'approved — pending effective.' On the effective date, the system auto-creates per-user re-training tasks for everyone in scope with the due date set to the effective date.
3. Operators see overdue and upcoming training on the kiosk home tile, with the document title, the training depth, and the due date. Notification also goes by email and, where configured, by SMS for safety-critical documents.
4. On task start, the kiosk reads the task's required-document set, walks each required document to its current effective version, and checks the operator's acknowledgement against that version. Any gap produces a hard-block with a clear remediation path ('complete training on SOP-DISP-014 Rev 4 before starting this task').
5. Acknowledgements are Part 11 §11.50-compliant — printed name, timestamp, meaning of signing ('trained — read-and-understand acknowledged' / 'training completed — quiz passed' / 'OJT signed off by qualified trainer' / 'assessed competent by named assessor'), and a version hash of the document content.
6. Trainer and assessor qualification are themselves training requirements — the system blocks a trainer from signing off OJT or assessed competency if their train-the-trainer credential is not current on the document being trained.
7. Deferrals are time-limited controlled records — QA can defer a re-training requirement for a named individual for a defined period with a documented justification; deferrals auto-expire and re-engage the hard-gate on expiry.
8. Reports surface current-version coverage, overdue training, kiosk hard-block trend, supervisor-override trend, and competency-depth alignment. The audit-export view produces, for any sampled operator-and-task pair, the full training history with version hashes in a single PDF rendered through the @react-pdf/renderer Worker-safe pipeline into the regulated-reports bucket.
9. Retention is indefinite by default; the per-tenant retention policy can shorten this where local law requires but the platform default exceeds every regulated regime (the longest standing requirement being device lifetime plus 2 years).
10. Mobile-safe ≤390px rendering for the kiosk training tile, the acknowledgement screen, and the operator training history — handheld and tablet operators get the same hard-gate semantics as desktop reviewers, with no horizontal scroll.