Why compliance training fails, and what changes behaviour instead

Ninety-eight per cent completion. Average score 91. Same three issues in the incident log as last year.

If that is your dashboard, the training did not fail at delivery. It failed at the brief. Someone asked for a module about a topic instead of a change in a behaviour, and everything downstream did what it was told.

The mistake at the top

Most compliance training is commissioned to answer a question about liability: can we show we told them? That is a reasonable question, and the regulations do care about it. The WHS Regulations require information, training and instruction that is suitable and adequate for the work and the risks, and readily understandable. The Australian Human Rights Commission's positive duty guidelines expect knowledge to be communicated and expectations to be clear.

But a programme designed only to answer that question optimises for coverage. Coverage means everyone gets the same thing, the thing must be short enough that everyone will finish it, and the assessment must be easy enough that everyone passes. Each of those is individually sensible and collectively fatal. You end up with content that is universal, brief and undemanding, which is a precise description of content that changes nothing.

Five reasons it fails

It targets knowledge when knowledge is not the barrier. Nobody harasses a colleague because they were unaware harassment is wrong. Nobody skips a lockout because they did not know the procedure existed. The gap is almost never information. It is social pressure, time pressure, a belief that the shortcut is what management actually wants, or a well-founded fear that reporting will cost more than staying quiet. A module that explains the rule addresses none of these.

It is written for the audience, not for a moment. "All employees" is not an audience. A supervisor deciding whether to send a crew out short-handed, a call-centre worker on the receiving end of abuse, and a finance officer looking at an invoice that smells wrong have nothing in common. Content pitched at all three teaches all three that it is about somebody else.

It never puts anyone in the hard position. Real compliance failures happen when the right thing is costly. Scenarios where the correct answer is obvious, and where every distractor is transparently the wrong answer, train people to recognise a quiz format. The useful scenario is the one where two defensible options exist and the reader has to reason.

It ends at the module. No manager conversation, no team discussion, no follow-up, no visible consequence. The organisation spends its energy on the artefact and none on the environment the artefact lands in.

It measures the wrong things. Completion is attendance. Post-test score measures short-term recall of content you just presented, which is close to worthless as a predictor of behaviour six months later. Neither number tells you whether anything changed.

What actually moves behaviour

Start with the behaviour, not the topic. Write the brief as a sentence: "When a customer becomes aggressive, we want the worker to disengage and press duress rather than continue to de-escalate alone." Now you have something to design against and something to measure. "Run bullying training" gives you neither.

Find the real barrier. Ask the people doing the work why the behaviour does not happen now. The answer is rarely ignorance. If the barrier is that duress alerts take eleven minutes to get a response, no training will fix it, and running training instead of fixing it is worse than doing nothing because it locates the blame in the worker.

Practise the decision. People need repetitions of the actual judgement, with consequences that play out. Branching scenarios drawn from your own incidents, where a plausible-looking choice leads somewhere bad, do more in eight minutes than an hour of explanation.

Make it specific and short and repeated. One decision, well practised, four times a year beats every topic once a year. Spacing beats duration.

Fund the managers. The strongest single predictor of whether a compliance programme works is what the first-line manager does the following week. If they reinforce it, it holds. If they roll their eyes, it dies, and the workforce learns that the standard is decorative. Manager capability is the highest-return line in most compliance budgets and the one most often cut first.

Fix the system where the system is the problem. The model Code of Practice on psychosocial hazards puts training low in the hierarchy of controls for good reason. Training is what you use when you cannot design the hazard out. If workload is the hazard, change the workload. Training people to cope with a hazard you could have removed is not a control, and a regulator will say so.

Measure something real

Pick behavioural indicators before you build and track them either side:

reporting and near-miss rates (usually you want these to go up first, because silence is not safety)
quality of escalations, judged against your own criteria
time from first report to first response
repeat issues in the same team
whether people who raised something would raise it again

Then be willing to conclude the training did not work. That conclusion is more valuable than another year of 98 per cent completion, because it sends you looking for the control that will.

The uncomfortable part

Some of what gets called a training problem is a leadership problem wearing a training problem's clothes. If a high performer behaves badly and keeps their job, no course in the world will convince anyone that the standard is real. Training can teach the rule, rehearse the decision and lower the cost of speaking up. It cannot make a consequence exist.

If you want training designed against a behaviour rather than a topic, starting from your incident data and measured against something other than completion, that is the work Hungry Minds does.

Common questions

Why does compliance training not change behaviour?

Because most of it targets knowledge when the actual barrier is something else, usually social pressure, unclear process, time pressure or a belief that reporting will backfire. Teaching a rule to someone who already knows it changes nothing.

Are completion rates a useful measure?

They measure whether people opened a module. They tell you nothing about whether behaviour changed. Useful measures are behavioural, such as reporting rates, escalation quality, near-miss reporting and repeat-issue rates.

Does making training longer make it more effective?

No. Length correlates with resentment, not retention. Short, frequent, scenario-based practice targeting one decision beats an annual ninety-minute module in almost every case.

What is the single biggest predictor of whether compliance training works?

What managers do afterwards. If a manager visibly ignores the standard or punishes someone for raising an issue, the training is dead regardless of quality.

How do we know if our training is working?

Pick one or two behaviours before you build, measure them before and after, and be willing to conclude the training was not the right control. Sometimes the answer is to change the process, not the course.