You are managing the wrong number
If the matrix you took to the last management review sat at ninety-eight percent complete, and last month's deviations still named "procedure not followed", what did that percentage buy you?
I have watched more than one quality head defend the first number and lose the room on the second. Completion is cheap to report. It is a workload metric: assignments closed, overdue falling, the chart pointing the right way. Competence is a decision you take about a person, a procedure and a date. Mixing them up is how a site pays for an LMS and still staffs the floor on last month's method.
Training management is the process. Training effectiveness is the verification step. This post is the management version: what the green cell costs you when it lies, what to put on the next review agenda, and what to demand of a record — or of a vendor — before you treat the dashboard as a control.
QMS versus eQMS is whether training even sits on the procedure. Document control is the version on the line. Read those if the file is already wrong. Read this if the file looks fine and the floor has not moved.
Completion is a count. Competence is a decision.
A completion record answers: did this person finish this item, and when? That is worth having. It is not the same question as: was this person qualified to run this revision on the day they signed the batch?
Across the implementations I have seen struggle, the habit is the same. Training is owned as a completion problem — HR or a coordinator chasing overdue rows — while QA inherits the deviations that name the procedure. Nobody owns the join. The matrix stays green because chasing signatures is a job you can staff. Changing how a task is done on the line is a job you cannot fake with a chart.
You do not need a new slogan for that. You need to stop promoting the count into a control.
I cannot give you a house ratio of completion-to-recurrence. We do not keep one, and a vendor who quotes someone else's is selling a brochure. The join is cheap to measure on your own site: three months of percent-complete next to three months of deviations that cite the procedure. If they do not move together, the first number is not earning its slot in management review.
What percent-complete cannot answer
| The count | The decision |
|---|---|
| Item closed, signature captured | Person qualified on the revision in force that day |
| Overdue trending down | "Procedure not followed" trending down after the same training |
| Annual GMP module finished | Continuing CGMP familiarity for the work they do, not a calendar tick |
| Retrain box ticked on a CAPA | A check that the behaviour moved, not a second signature so the CAPA can close |
| Matrix cell = completed | Matrix cell points at the individual record (person, item, revision, date, the check you defined) |
The left column is administration. Run it well; it still does not answer the right-hand column. If you only staff the left, you will keep buying tools that make the left faster.
What the signature does not buy you
21 CFR §211.25 is a personnel rule, not an LMS rule. People who manufacture, process, pack or hold a drug product need education, training and experience — in some combination — to do the job they were assigned. Training has to cover the operations they perform and CGMP as it relates to those functions, on a continuing basis, given by qualified people, often enough that they stay familiar with the requirements that apply to them. Supervisors get their own sentence. Headcount has to be adequate. None of that is "percent of modules closed this quarter."
EU GMP is more explicit about the missing half. In Chapter 2 (Personnel, 2014), continuing training is required, "and its practical effectiveness should be periodically assessed." Programmes are approved; records are kept. The text does not prescribe a quiz. It does prescribe that someone periodically looks at whether the training worked. A completion rate is not that look.
For medical devices, ISO 13485 §6.2 (now the QMSR personnel baseline) asks the organisation to evaluate the effectiveness of the actions taken to reach competence, and to keep the records. Same idea, different book. Do not import a device clause into a tablet line and call it done; do not ignore it if you are under the QMSR.
ICH Q10 tells management to provide the resources and training to hit the quality objectives. It does not tell you that a vendor dashboard discharges that. Clause-level reading of Part 11 — when the training record is electronic — sits in Vijay's Part 11 post. Here the point is narrower: the predicate already asked for qualified people. Software that only counts signatures has not bought the predicate.
Retraining as a CAPA action is often a budget leak
"Retrain the operator" is the cheapest sentence in a CAPA. It is also the one I have seen close the most records without touching the recurrence.
Sometimes training is the action: the revision changed a critical step, the person was not on that revision, you train them on it, you check they can do it, you watch whether the same deviation returns. That is a real spend. Sometimes the action is a default because the investigation stopped at "did not follow procedure" and nobody asked why the procedure is unworkable, unsupervised, or missing at the station. Then you have paid for a signature and kept the cause. The training-effectiveness glossary walks that pattern; deviation, CAPA and change is where the chain loses the people-link. I will not re-try the investigation science here. The management question is: how many open CAPAs name training as the action, and how many of those have an effectiveness check that is not another attendance mark?
If you cannot answer that from the CAPA list without a side spreadsheet, you are not managing training effectiveness. You are managing closures.
Four questions for the next management review
Take these as they are. If you cannot answer them in the room, you do not yet have a training KPI.
- For one batch released last month, can we name the procedure revision that governed a critical step and show the training evidence for the people who performed it — person, item, revision, date — without opening a second tracker?
- After the last SOP change that mattered on the floor, what besides a quiz told us the line moved? Who owned that check?
- Of the CAPAs closed in the last two quarters that named retraining as the action, how many have a defined effectiveness check, and how many of those checks are only a second signature?
- Does the matrix know the revision, or only the document ID? If it only knows the ID, a green cell can sit on last year's method.
Question 1 is retrieval cost. Question 2 is whether you defined "worked." Question 3 is whether CAPA is leaking money into theatre. Question 4 is whether the overview is lying about version. None of them is "are we above ninety-five percent complete?"
Download the training matrix template for that overview layer. It is a map, not proof: completing a cell is not the controlled individual record. Point the cell at that record. Do not let the cell become it.
What to demand of the record — and of a vendor
You are not choosing an LMS feature list. You are choosing whether qualification evidence is cheap to produce when the question arrives.
Ask:
- Can assignment come from the change or the CAPA, against the revision that will be in force, rather than from a reminder someone typed later?
- Does the record carry the check you defined — knowledge, observed task, supervised run — or only the signature that the item was opened?
- When the procedure revises, does overdue mean "not yet on this revision," or "the annual module is stale"?
- If we export one person's file for one SOP, do we get person, item, revision, date and the check, or a completion dump we have to narrate?
A demo that only shows a green matrix has shown you the count. How to evaluate an eQMS is the wider bake-off. Put these four on the training row of that scorecard so training does not get scored as "has an LMS."
Electronic training records pick up Part 11 because they are electronic, not because the purchase order said "validated LMS." The 2003 guidance did not retire the predicate. Qualified people are still the predicate.
Who still owns qualification
The quality unit still decides what "qualified" means on this site: which roles, which procedures, which checks, which frequency. Supervisors still have to know who is current before they put someone on a step. A vendor cannot sit in that judgement. ICH Q10's line about management providing training is a resource obligation, not a licence transfer.
What you can buy is a record those judgements sit on: the person and the revision together, the reason the assignment existed, the check you said you would do, still readable when the coordinator who chased the overdue row has left. If the tool cannot do that, keep the matrix as an overview and stop calling it the training system.
What Complere does with a training record
The person and the revision stay together. You can answer "who was qualified on this SOP, this revision, this date" from the record, not from a join between HR and a shared drive.
A signature is not the effectiveness check. You can record the check you defined — and see, in the same place, whether it was done. You still write the procedure that says what "done" means.
Retraining has a reason you can name. Assignment from a change or a CAPA carries the trigger. Closing the CAPA on attendance alone is a choice you can still see.
The overview is not the proof. A matrix can show overdue. The file behind the cell is what you would hand over.
You still decide who is allowed to do the work. Complere is worth the fee only if the evidence of that decision sits on the person and the procedure, not in a dashboard that counts signatures.
