Tempo di lettura: 9 minuti
A guide to assessing real competence after corporate training
Measuring training impact doesn't start by choosing a metric. It starts by designing a check that someone who can't do the task would fail.
When leadership asks whether the training worked, the honest answer in most organizations is that nobody knows. There's data: hours delivered, participants, completion rate, high scores on the satisfaction survey. None of it answers the question.
The usual diagnosis is that metrics are missing. It almost never is. The problem shows up earlier, at the moment someone decides how it will be checked that a person has learned something. If that check is badly designed, it doesn't matter how many indicators you stack on top: they all inherit the same flaw.
We're going to look at why measurement fails before the indicator is chosen, what competence verification in training means and how it differs from measuring consumption, and how to pick the right check for the kind of knowledge you're trying to confirm.
The typical path is familiar. Training gets delivered, a ten-question multiple-choice test fires at the end, who passed is recorded, and that figure feeds the report.
The test works as a gate: almost everyone passes. And they pass because the questions can be answered with what was just read, not with what you need to be able to do. The training could have been excellent or it could have taught nothing, and the test result would look much the same either way.
That's the flaw. An assessment that can't discriminate between someone who can execute the task and someone who can't produces compliance metrics, not competence metrics. And building impact indicators on top of a compliance metric produces reports nobody believes, starting with whoever signs them.
We call Competence Verification the practice of confirming that a person can perform the task the training was meant to teach, using a check designed so that someone who can't perform it won't pass.
The difference between completing and being competent isn't in the percentage the LMS displays. It's in whether the check behind it could tell those two people apart.
It isn't the same as completion rate, which measures consumption, nor satisfaction, which measures perception. It also isn't the same as compliance tracking, which answers a different and perfectly legitimate question: whether everyone has done what they were required to do. We cover that side in our piece on ensuring mandatory training gets completed.
Worth noting that this requirement isn't a methodological preference. ISO 9001 states it explicitly in clause 7.2: taking action to acquire the necessary competence isn't enough, you have to evaluate the effectiveness of the actions taken.¹ An attendance record documents that the action took place, not that it was effective: those are two distinct requirements of the same clause, and only the first is covered by a signature sheet.
And in occupational safety, Spanish law has been asking for "theoretical and practical" training since 1995.² The word practical has been sitting there for thirty years, and it's still frequently discharged with a slide deck and a signature sheet.
The false positive test is the criterion that separates a useful check from a formality, and it consists of a single question:
Could someone pass this check without being able to do the task?
If the answer is yes, the check measures something else: attention, short-term memory, the ability to deduce the right option by elimination, or simple persistence until the system lets them through.
Applying the false positive test to an existing assessment takes a few minutes and consists of answering these four questions before looking at the results.
An assessment that fails three of the four isn't measuring learning. It's generating a record.
The most widespread error is using the same instrument, almost always the multiple-choice test, to check things that have nothing in common. What you need to verify determines how you verify it.
Facts, limits, definitions, what a procedure says. The temperature threshold, the notification deadline, which product doesn't get mixed with which.
Here a test does work, but only if the questions place the fact inside a situation rather than asking for it in the abstract. Not "what is the notification deadline?" but "you get this alert on a Friday afternoon, by when do you have to report it?".
Executing a sequence in the right order, with the intermediate checks. Locking out a machine, closing a till, completing a delivery to protocol.
A test can't verify this, because listing the steps and performing them aren't the same skill. The verification is observed execution, real or simulated, against a short checklist. Two executions are enough; what matters is that they're dated and that the observer notes what went wrong, not just whether it passed.
Deciding well when the situation doesn't map cleanly onto the procedure. It's what separates someone with five years from someone with two, and it's almost never verified.
The instrument here is a case with no single obvious answer, assessed on the reasoning rather than the option ticked. It requires someone with judgment to read the answer, which is why it's the most expensive. It's also the only one that detects whether the training produced anything beyond obedience to the manual.
| What you want to verify | Instrument that works | Instrument that produces false positives | Relative cost |
|---|---|---|---|
| Knowing what (facts, limits, regulation) | Questions placed inside a concrete scenario | Definition test with the material open | Low |
| Knowing how (executing a sequence) | Observed or simulated execution against a checklist | A test that asks you to order the steps | Medium |
| Knowing when and why (judgment) | Ambiguous case assessed on the reasoning | Any closed question | High |
The practical consequence of this table is that not everything needs verifying at the same depth. Verifying judgment across the year's thirty training actions isn't viable. Picking the three where a wrong decision has real consequences and verifying those properly is entirely possible, and it changes the conversation with leadership more than thirty completion rates do. Our piece on which training metrics to report to leadership covers how that translates into the language of the executive committee.
A well-designed check fails people. That's its job, and it's also why many organizations don't want one: a 100% pass rate generates no awkward questions.
The right reading of a failure isn't individual. If one person fails, it's a case. If 40% fail on the same point, what failed is the content: that step of the procedure wasn't explained well, or was explained in a way that doesn't hold up on the job. Verification stops being an exam and becomes the diagnostic that tells you which module to rebuild.
The loop between the verification's error rate and the redesign of the module it flags is what makes measurement useful. Verifying without rebuilding the content the verification points to is accumulating evidence of a problem nobody fixes. Platforms that let you edit one specific module without reproducing the whole lesson make that loop affordable, but the decision to look at the data and act comes before any tool.
Training impact doesn't get discovered by building a more complete dashboard on top of the same data. It gets discovered by changing what you record, and that starts with a check that can tell apart someone who can do the task from someone who can't.
The false positive test applied to the last assessment launched usually delivers the diagnosis in ten minutes. If someone could pass it without being able to do the job, that's the explanation for why the training report convinces nobody.
An organization doesn't know what it has taught. It knows what it can check.
It's confirming that a person can perform the task the training was meant to teach, using a check designed so that someone who can't perform it won't pass. It's distinct from completion rate, which measures consumption, and from satisfaction, which measures perception.
It works for declarative knowledge, and only if the questions place the fact inside a scenario rather than asking for a definition. For verifying the execution of a procedure or the ability to decide in an ambiguous case, it produces false positives.
ISO 9001 clause 7.2 requires evaluating the effectiveness of the actions taken to acquire the necessary competence, not merely having carried them out.¹ In occupational safety, Spanish Law 31/1995 has required "theoretical and practical" training since 1995.²
Not all of them, because it isn't viable. The sensible approach is to verify thoroughly the ones where incorrect execution has real consequences (safety, quality, compliance) and settle for lighter checks on the rest.
Treat it as a content failure, not a people failure. A high error rate concentrated on one step means that point wasn't explained well or doesn't hold up in the real job, and it tells you exactly which module to rebuild.
With a date, the assessor or instrument, the version of the content the person was trained on, and the outcome. Without the version, the record can't tell you what each person learned once the procedure changes. The full field list is in our guide to the multi-standard training record.
¹ ISO 9001:2015, Quality management systems, clause 7.2 Competence - International Organization for Standardization ² Ley 31/1995, de 8 de noviembre, de prevención de Riesgos Laborales, article 19 - BOE
Programma di promozione dell'assunzione a tempo indeterminato di giovani qualificati, nell'ambito del Sistema Nazionale di Garanzia Giovani. Vidext ha ricevuto una sovvenzione di 25.401€, concessa da LABORA (Servizio Valenciano per l'Impiego e la Formazione), per l'assunzione a tempo indeterminato nel 2024 di giovani qualificati iscritti al Sistema Nazionale di Garanzia Giovani, azione suscettibile di cofinanziamento da parte del Fondo Sociale Europeo Plus (FSE+) 2021-2027 o di qualsiasi altro fondo dell'Unione Europea. Expediente ECOGJU/2024/550/46. La presente pubblicazione è effettuata in adempimento degli obblighi di trasparenza stabiliti dalla Legge 19/2013, del 9 dicembre.
