What is the difference between completion and competency in training?

Completion records that a person reached the end of the material — they attended, watched, or clicked through. Competency is the demonstrated ability to perform the task correctly, including under time pressure, in poor conditions, and without supervision. Completion is a fact about the course; competency is a fact about the person. Most platforms measure the first because it is trivially easy to log, and report it in a place where managers read it as the second.

Why is a high training completion rate not enough?

Because completion is insensitive to whether anything was learned. A worker who barely passed on their fourth attempt and one who performed flawlessly the first time produce an identical completion record. If your dashboard shows 98% completion while your near-miss log is climbing, those numbers are not in conflict — they are measuring different things, and only one of them describes risk.

How do you measure training effectiveness properly?

Measure behaviour rather than attendance. Useful signals include: first-attempt success rate, hesitation and self-correction during the task, performance under degraded conditions, decay between the training date and a later re-check, and transfer — whether the behaviour appears on the floor, not just in the assessment. The Kirkpatrick model formalises this as four levels (reaction, learning, behaviour, results); most organisations only ever measure level one.

Can an LMS measure competency?

A traditional LMS is built as a content delivery and record-keeping system, so it measures delivery and record-keeping well. It can host an assessment, but a multiple-choice score mostly measures recall, which correlates weakly with the ability to perform a physical or judgement-heavy task. Measuring competency requires capturing what a person did during a realistic task, not what they selected from four options afterwards.

What is competency-based training?

Competency-based training defines the outcome as demonstrated ability rather than time served or material covered. A learner progresses when they can perform the task to a defined standard, not when they reach the last slide. It changes the unit of record from 'course completed' to 'capability evidenced', which is also what auditors and incident investigators are actually asking about.

Completion vs Competency: What Your LMS Isn't Telling You

Two numbers sit in most safety and training reports, and they rarely get put side by side.

The first is the training completion rate. It is usually excellent — 95%, 98%, sometimes a suspicious 100%. The second is the near-miss log, or the audit findings, or the incident rate. That one is often not excellent at all.

Managers treat this as a puzzle. It is not a puzzle. The two numbers are not in conflict, because they are not measuring the same thing. One measures attendance. The other measures reality.

What completion actually records

Completion answers a narrow question: did this person reach the end of the material?

That is a real fact, and it is worth recording. It is also very nearly the only thing most platforms record, for a simple reason — it is trivially easy to log. A video reaches its final frame. A page's last slide renders. A quiz returns a passing score. Tick.

The problem is what completion is blind to:

The worker who passed on their fourth attempt looks identical to the one who was flawless on their first.
The worker who guessed correctly looks identical to the one who knew.
The worker who could do it in a quiet room in March looks identical to the one who can do it on a loud floor in November.
The worker who was clicking through in another tab looks identical to everyone.

Every one of those pairs is a different level of risk, and completion collapses them into the same green tick.

That is what we mean when we say completion is attendance wearing the costume of a result.

What competency asks instead

Competency asks a harder question: can this person perform the task correctly, under realistic conditions, without someone standing over them?

It is harder because it cannot be inferred from a content record. It has to be observed — which is why every serious safety standard, and every serious apprenticeship, eventually routes through a human being watching someone do the thing. (We wrote about where that line falls legally in does VR safety training satisfy OSHA requirements.)

The gap between the two words is not academic. It shows up as:

The audit finding that contradicts the certificate. The record says trained. The observation says otherwise. The record loses.
The incident involving a trained operator. Which is most of them, because most operators are trained on paper.
The retirement cliff. The person who actually knew how it was done leaves, and it turns out the know-how was never in the LMS at all — it was in their hands.

Why the LMS cannot fix this by trying harder

A traditional LMS is a content delivery and record-keeping system. It is genuinely good at those jobs. Asking it to measure competency is asking it to do something it was not architected for.

You can bolt an assessment onto it, and most organisations do. But a multiple-choice test mostly measures recall, and recall correlates weakly with the ability to perform a physical or judgement-heavy task. Knowing that you should isolate the energy source is a different capability from doing it correctly at 2am on equipment that is not quite like the diagram.

This is also why "we added a quiz" rarely moves the incident numbers. The quiz measures a different variable than the one causing incidents.

Five things worth measuring instead

If you want signal rather than reassurance, these are the measurements that actually discriminate between workers:

1. First-attempt success. Not "did they eventually pass" — did they get it right the first time, without prompting? This single change is the highest-value edit most programs can make to their reporting.

2. Hesitation and self-correction. Where did they pause? What did they start doing and then undo? Hesitation is one of the most honest signals available, because it reveals uncertainty that a final answer hides. A confident wrong action and a hesitant right action are very different states of knowledge.

3. Performance under degraded conditions. Time pressure, noise, an unfamiliar variant of the equipment, a step out of the expected order. Competence that only exists in ideal conditions is not the competence you are buying.

4. Decay. Re-check at 30, 90, 180 days. Almost nobody does this, and it is the measurement most likely to change how you schedule refreshers — because it replaces a guessed interval with an observed one.

5. Transfer. Does the behaviour show up on the floor, or only in the assessment? This is the hardest to instrument and the only one that finally matters. Supervisor spot-observations, tied back to the training record, are the usual practical route.

If you want a formal frame for this, the Kirkpatrick model has described it for decades as four levels — reaction, learning, behaviour, results. The uncomfortable observation is that most organisations measure level one, report it as though it were level four, and are then surprised by the gap.

The honest objection

Measuring competency is more expensive than measuring completion. That is the real reason completion won, and it deserves a straight answer rather than a dismissal.

Observation takes a qualified person's time. Realistic scenarios take money to build. Behavioural data takes somewhere to put it and someone to read it.

The counter-argument is not that competency measurement is cheap. It is that the current arrangement is already expensive — you are paying for training, paying for the incidents the training did not prevent, and receiving a number that cannot tell you which of those two things you bought. The cost is not absent. It is just recorded somewhere other than the training budget.

What we are building

We build industrial XR training, and repeatedly ran into the same wall: clients could tell us exactly how many people completed a module and almost nothing about whether any of them could do the job.

That is why we are building LSM. Where traditional platforms record completion, LSM is a browser-native way to measure competency — capturing what a person actually did during a realistic task, so the record you keep is behavioural evidence rather than a tick.

It is patent-pending, pre-launch, and arriving in 2026 with a limited early-access cohort.

If your completion rate and your incident data are telling you two different stories, that is the conversation worth having — see what LSM is, or tell us what your numbers look like.