A forklift came out of an aisle, or reversed from a bay, and a person was closer to it than anyone wants to think about. Nobody is hurt. The first instinct is relief, the second is to move on, and the third, if the site is well run, is to ask what to do about it. This is the sequence, built on HSE's investigation guidance (HSG245) and workplace transport guidance (HSG136).
Make the area safe. If the near miss exposed a hazard that still exists, a missing barrier, a blocked walkway, a blind corner with no mirror, deal with that first. A near miss at 10:00 that nobody acted on is an incident waiting for 10:30.
Get the time to the minute. Ask the people involved while they remember. "About ten" costs half a day of searching footage later; "10:07" costs a minute. Note the camera that covers the spot.
Preserve the footage. Most CCTV systems overwrite in days. Ask whoever controls the cameras to export the clip now, with ten minutes either side. Even if you decide later that the event was minor, the clip is the one piece of evidence that does not depend on anyone's memory.
Check for injury. A near miss can include a stumble, a jump out of the way, a knock against racking. If anyone is hurt, the RIDDOR questions apply.
Record it. Date, time, location, who was involved by role, what happened in two sentences, what was done immediately, and how bad it could have been. A near-miss log with those fields takes two minutes; the record is what makes the next step possible.
Decide the RIDDOR question. A forklift near miss with no injury is not reportable. Two exceptions: if the truck overturned, or a load-bearing part (mast, forks, chain) collapsed or failed, that is a dangerous occurrence under Schedule 2, paragraph 1, reportable within 10 days whether or not anyone was hurt. The dangerous occurrences list has the detail.
Talk to the people involved, without blame. The driver and the pedestrian both know something you do not: why the pedestrian was there, why the driver did not see them, what the normal route is and why this one was different. HSG245 is explicit that investigations exist to find causes, not culprits. If the conversation feels like a disciplinary, you will get the account that protects the person, not the one that explains the event.
If the conversation feels like a disciplinary, you will get the account that protects the person, not the one that explains the event.
Investigate in proportion. HSG245 scales the investigation to the potential outcome, not the actual one. A near miss between a forklift and a pedestrian has a potential outcome of death. That does not mean a week-long inquiry, but it means more than a note in the log: look at the layout, the lighting, the traffic plan, the shift pattern, and whether the same spot has produced near misses before. The pattern is usually in the log already if anyone has kept one.
Separate the immediate cause from the underlying one. The immediate cause is what happened in the last five seconds: the pedestrian stepped out, the driver was looking at the load. The underlying cause is why that was possible: there is no barrier at the aisle end, the walkway runs through the vehicle route, the mirror was never replaced, the agency crew were not shown the route. Fixing the immediate cause produces a toolbox talk. Fixing the underlying cause produces a change.
Choose a control that does not rely on everyone remembering. The hierarchy of controls applies here more than anywhere. Retraining and reminders sit at the bottom; they depend on every person, every shift, including the ones who start next month. Separating the route, closing the aisle end with a barrier, moving the crossing, adding a physical stop at the dock: these keep working when nobody is thinking about them. HSG136's first principle for workplace transport is segregation, and most near misses are an argument for more of it.
Put a date in the diary to check it held. This is the step most sites miss. The barrier goes in, the action closes, and nobody goes back. Controls decay: barriers get moved for a delivery, mirrors get dirty, the new route gets a shortcut. Set a date, two weeks and again at two months, to stand at the spot and see whether the control is in place and whether behaviour has actually changed. Record that check. A closed action is not a controlled risk.
Tell them what happened, what was found, and what has changed, in that order, and close with thanks to whoever reported it. HSG245 and every study of reporting culture agree on the mechanism: people report near misses when they see reports lead to action, and stop when they see them lead to blame or silence. The way this one is handled decides whether you hear about the next one.
Everything above assumes someone saw it and said so. Most forklift near misses do not get that far: they happen on the night shift, or to an agency driver who does not know the form exists, or to someone who decided it was not worth the conversation. The one you are dealing with is real, and it is also a sample of one.
The honest question at the end of the week is not "have we fixed this one?" but "how often does this happen here, and would we know?"
Not if nobody was injured and the truck stayed upright. If the forklift overturned, or a load-bearing part collapsed or failed, it is a dangerous occurrence under RIDDOR Schedule 2 paragraph 1 and must be reported within 10 days regardless of injury. If someone was hurt, the injury rules apply.
Yes, in proportion to its potential outcome, which for a vehicle–pedestrian near miss is serious injury or death. HSG245 recommends identifying immediate, underlying and root causes and acting on the underlying ones.
Not as a reflex. HSG245 distinguishes investigation from blame; most near misses have underlying causes in layout, routes or workload that discipline does not address. Where a clear and deliberate breach is established, that is a separate process, and keeping it separate protects reporting.
Export it immediately; most systems overwrite within days. Keep the clip with the near-miss record for as long as the record is kept, and at least three years if the event is connected to any reportable incident or likely claim.
Whatever removes the interaction rather than reminds people about it: separate routes, physical barriers at aisle ends, relocated crossings, dock-door controls. Training and signage support these; on their own they depend on every person every shift.
By going back and looking, on a date set when the action was closed, and recording what was found. A control that was in place on day one and gone by week three was never a control.
This one got reported because someone saw it. The question for the next 72 hours is how many did not. Safety Mapping: a free two-week trial on one camera.
Published 27 October 2026 · Last updated 27 October 2026