The workplace metric that creates false confidence

by • News, Technology, Workplace design

A learning dashboard showing 96 per cent completion feels reassuring. It is precise, comparable and easy to report. But it may be answering the least useful question: who reached the end of the material? Completion can confirm attendance or fulfil an administrative requirement. It cannot, on its own, show what people understood, what they still remember or what they can do differently at work. That distinction matters as organisations ask employees to adapt to new systems, regulation and AI-enabled work.

The OECD’s 2025 Trends in Adult Learning report found that around two in five adults across participating countries engage in formal or non-formal learning each year. Yet much adult training is short, job-related and driven by compliance or safety. A high completion rate in that environment may demonstrate reach, but it can also create confidence without capability evidence.

 

Completion is a receipt, not a result

Completion data is not useless. It helps teams manage access, reminders and mandatory deadlines. The mistake is allowing one convenient metric to stand in for the whole learning outcome. Finishing a module is closer to receiving a receipt than proving that the purchase works.

The gap becomes visible when a worker encounters a situation that does not look exactly like the example in the course. Can they recognise the relevant risk? Can they explain the principle in plain language? Can they choose the right action when two options appear plausible? A completion percentage cannot answer any of these questions.

This is why the CIPD’s current guidance on learning evaluation, impact and transfer starts with the performance gap and the organisational objective. Evaluation becomes useful when it tests whether the intended change occurred, rather than merely counting activity around the intervention.

 

Use a short chain of evidence

A more credible measurement approach does not require an elaborate analytics programme. It requires a short chain of evidence. Keep completion for operational control, then add a low-stakes check that asks people to apply the idea rather than repeat wording. Revisit the most important topics after a delay. Finally, look for a proportionate workplace signal that the behaviour is appearing in practice.

The immediate check should be diagnostic. A scenario, prioritisation task or brief explanation can reveal whether people can distinguish between similar choices. Topic-level results are more useful than one average score because they show where the material, question or explanation needs improvement. A delayed check, even if it is very short, separates momentary familiarity from retention.

Application evidence should be chosen carefully. Depending on the role, it might be a manager observation, a quality review, fewer repeated errors or a better decision in a simulated case. It should not become blanket surveillance or a punitive ranking of individuals. The purpose is to improve the learning system and provide support, not to manufacture a new performance target.

Confidence can add another useful signal when it is paired with a question rather than collected alone. Ask people how certain they are about an answer, then compare that judgement with performance. A wrong answer given with low confidence usually points to a straightforward knowledge gap. A wrong answer given with high confidence is more urgent: the learner may be ready to act on a misunderstanding. That pattern can guide a targeted explanation, practice scenario or manager conversation. Used in aggregate, it can also reveal where course wording creates false reassurance. Confidence should never become a personality score; its value lies in showing where perceived readiness and demonstrated understanding diverge.

 

Report uncertainty instead of hiding it

Responsible reporting also means being honest about what the data cannot establish. A strong knowledge check does not prove long-term behaviour change. A lower error rate after training does not prove that the course caused the improvement; process changes, coaching or staffing may also matter. The CIPD’s evidence-based practice guidance recommends combining evidence from several sources and appraising its quality rather than relying on a single convenient indicator.

The same discipline should shape dashboard design. Show completion, understanding, delayed recall and application as different signals. Break results down by topic, location or role only where the groups are large enough to be fair and privacy-safe. Add the date and sample size. Mark where evidence is missing. A dashboard that displays uncertainty may look less polished, but it gives leaders a more accurate basis for action.

 

A better question for leaders

The practical question is not whether completion rates should disappear. It is whether leaders know what those rates can and cannot tell them. When the stakes are low, completion may be enough. When training is meant to change a decision, reduce risk or build a transferable skill, it is only the first checkpoint.

The most useful learning metric is not the one that produces the cleanest green dashboard. It is the one that reveals what people are ready to do, where confidence exceeds understanding and where the organisation should help next.

Main image: Sedus