Show me an induction that runs four hours, covers eleven topics, ends with a twenty-question multiple choice and has a 99% pass rate, and I will show you a control that exists for the file.
I am not being cynical about the legal requirement. Section 8 of the OHS Act and the equivalent duties under the Mine Health and Safety Act are clear, and the training obligation is real. But there is a persistent confusion between discharging a duty and changing behaviour, and it produces organisations that are simultaneously fully compliant and genuinely unsafe.
The gap shows up in the near-miss data
The tell is always the same. Incident investigations that repeatedly conclude "failure to follow procedure", corrective actions that repeatedly read "retrain the employee", and a near-miss rate that is either implausibly low or entirely flat.
If your employees keep failing to follow a procedure, the useful question is not who to retrain. It is why the procedure loses to the alternative. Usually one of three things is true: following it takes materially longer and production pressure is real, the procedure was written by someone who has not done the task, or the equipment required to do it correctly is not actually available on the section.
None of those are fixed by another training session. All three are commonly closed out with one.
Competence has a legal definition, and it is not attendance
Under the Construction Regulations and the general machinery regulations, competence is knowledge, training, skill and experience applied to the task. Four elements. An attendance register evidences one and a half of them.
This matters practically because the register is what gets produced to an inspector or in a Section 32 inquiry, and it is thin cover if the person demonstrably could not perform the task. What holds up is a competence file: the training record, a documented practical assessment against defined criteria, a record of supervised experience, and a dated authorisation by a named responsible person for a defined scope of work.
Building those files is administratively heavier. It is also the difference between a defensible position and an exposed one.
Retraining as a corrective action is usually an admission
When I audit corrective action registers, I count what proportion of actions are training-based. Above about a quarter and something is wrong with the investigation method, not the workforce.
Training sits near the bottom of the hierarchy of controls for a reason. If a task can be redesigned so the unsafe option is physically harder — an interlock, a different coupling, a permit that cannot be issued without a lockout confirmation — you have engineered the risk out rather than asking a tired human at the end of a twelve-hour shift to remember a rule.
Where training genuinely is the right control, say so explicitly and specify what capability must exist afterwards. "Retrain on procedure SOP-114" is not an action. "All section artisans to complete practical assessment on isolation and lockout, assessed by the engineering foreman, before returning to unsupervised work" is.
What good looks like on the ground
The best-run operation I visited last year had an unremarkable training programme and an exceptional stop-work culture. A junior operator had halted a crane lift the previous week because the rigging looked wrong. He was right. Two things followed: the lift was corrected, and the shift manager mentioned it at the next three toolbox talks by name.
That costs nothing and it does more than any refresher course. The signal it sends is that stopping work carries no penalty, which is the only condition under which people actually do it.
The inverse is equally clear. Where I find production superintendents who talk about safety in terms of lost time injury rates and bonuses, reporting quietly collapses. Nobody logs a near-miss that threatens a section bonus.
Three things I would change first
- Recode your last two years of incidents by contributing factor rather than injury outcome. The pattern that emerges is normally not the one in the annual report.
- Cap training-only corrective actions. Force the investigation team to justify why an engineering or administrative control is not viable.
- Move at least a third of induction time out of the classroom and onto the plant, assessed practically, in the language the crew actually works in.
Compliance is the floor. Treating it as the ceiling is how organisations end up with a perfect file and an incident nobody saw coming.