An incident has happened, or a serious near miss, involving a vehicle and a person. The investigation is under way. At some point in the next month someone outside the safety team will want to see evidence. This guide is about what that evidence is, and what usually gets offered instead.
Different things, with one common thread.
An HSE inspector works from the duty in the Management of Health and Safety at Work Regulations 1999 to have arrangements for "monitoring and review" of preventive measures, and from HSG136's expectation that workplace transport controls are not just designed but maintained. The question is whether the risk assessment was "suitable and sufficient" and whether the controls it identified were in place and being used. "In place" on paper is not the test; a painted walkway that the night shift crosses diagonally is not a control.
An insurer's risk engineer, especially after a claim, wants to know whether the site can demonstrate that controls were operating at the time, because that bears on liability and on the premium going forward. A training record dated eighteen months ago does not demonstrate that.
Your own board wants to know whether it will happen again. The honest answer depends on whether anyone knows the state of the controls on the shifts nobody inspects.
The common thread: evidence that the control was working, not evidence that it existed.
The question is never whether the control was installed. It is whether it was working on the shift that mattered.
Because most of what a site can produce after an incident is evidence of installation, not operation.
| Usually offered | What it proves | What it does not prove |
|---|---|---|
| Traffic management plan | A design existed | The design is what happens |
| Barrier and signage invoices | Equipment was bought | It is still where it was put |
| Training and induction records | People were told | People do it, including agency staff who started since |
| Risk assessment | Hazards were considered | Controls were reviewed after conditions changed |
| Toolbox talk register | A message was delivered | Behaviour changed and stayed changed |
| Inspection sign-off sheets | Someone walked round | What they looked at, when, and what they found |
The last row is the one with potential. An inspection record that says "yard: no issues" proves very little. One that says "Gate A barrier: in place, latched, 06:40, night shift observed using the walkway" is evidence. The difference is that the second one records the control's state, the time, and the shift, so it can be compared with the next one.
HSG65's Plan, Do, Check, Act model puts this under Check, and splits it into active monitoring, checking controls before anything goes wrong, and reactive monitoring, learning from incidents after. The weeks after an incident are reactive by nature; what an outsider is looking for is the active record that came before, and what the site is looking for is the active record that will come after.
Four things make up that record:
Control state, checked in the field. Not "do we have a barrier?" but "was the barrier in place and functioning on this date, at this time?" ISO 45001 clause 10.2 asks for the effectiveness of corrective action to be reviewed, not just its completion; the same logic applies to every control. HSE's process safety guidance (HSG254) puts it plainly: "all systems deteriorate if unchecked."
Coverage across shifts. Controls fail at night, at shift change, and at peak, which are the times inspections rarely happen. A record that covers only the day shift is a record of the day shift.
Behaviour, not just equipment. A barrier can be in place and ignored. A walkway can be clear and unused. The evidence an inspector finds persuasive is that the control is used: that people walk the route, that drivers stop at the line, that the exclusion zone is empty during discharge.
Before and after. If the incident led to a change, the evidence that the change worked is a comparison: how the spot behaved before, and how it behaves now, on the same shifts, for long enough to be sure. A change checked once, the week after, proves it was installed. A change checked for two months proves it held.
If the incident has already happened, start with what exists.
To an inspector: a site that monitors its controls, finds problems, and fixes them is a site with a functioning management system, which is what the regulations require. Finding a failed barrier on a check and recording it is not an admission; it is evidence the system works.
To an insurer: a demonstrable, dated record of controls being verified is the difference between "we had a policy" and "we can show you what was happening." It is also, over time, the evidence base for a conversation about premium that is otherwise impossible.
To the board: it turns "we have addressed it" into "here is how we know."
That the risk assessment was suitable and sufficient, that the controls it identified were in place and being used, and that arrangements existed to monitor and review them (Management of Health and Safety at Work Regulations 1999, regulation 5; HSG136; HSG65). Records showing controls were checked in the field over time carry more weight than records showing they were installed.
No. A plan is a design. Evidence that the plan is what actually happens, on every shift, is what is being asked for. HSG136 expects controls to be maintained and reviewed, not just designed.
Checking that controls are in place and working before anything goes wrong: inspections, audits, observation of work, checks on procedures. Reactive monitoring is learning from incidents after. HSG65 asks for both.
There is no fixed interval in the guidance; the test is whether checks are frequent enough to catch deterioration before it causes harm, and cover the shifts and conditions where controls are most likely to fail. A check recorded with the control's state is worth more than a frequent check recorded as "no issues."
To an inspector, the opposite. A monitoring system that finds and fixes failures is evidence of a working management system. A system that never finds anything is evidence of nothing.
The control chosen and why, who owns it, when it closed, when it was checked afterwards, and what was found. ISO 45001 clause 10.2 requires the effectiveness of corrective action to be reviewed; the check date is how that review becomes real.
The question after an incident is never whether the control was installed. It is whether it was working at 06:40 on a Tuesday. That is what a camera on the spot answers. Safety Mapping: a free two-week trial on one camera.
Published 27 October 2026 · Last updated 27 October 2026