What counts as a near miss
An event that could have caused harm and did not. That's it. Nobody went to hospital, nobody filled in the accident book, and by dinner time half the gang has forgotten it happened. The board that slid off the fourth lift and hit the pavement where the delivery driver had been standing ninety seconds earlier is a near miss. So is the abrasive wheel that shattered and went past somebody's visor rather than into it.
People tie themselves in knots over the vocabulary. Near miss, close call, near hit, undesired circumstance. HSE's incident reporting guidance uses near miss for the event that could have caused injury and didn't, and it's the phrase every operative on site already understands, so use it and stop worrying.
Worth widening the net slightly, though. A near miss doesn't have to be dramatic. The hazard you spotted and made safe before anybody was exposed counts as useful information: the exposed reinforcement bar with no cap, the temporary lighting cable lying in a puddle, the fire door propped open with a bag of sand. Some firms split these into "near miss" and "unsafe condition" and it makes tidy pie charts. On a twelve-man refurb job, one box is plenty.
What it isn't is damage-only. A reversing wagon that clips the site cabin corner and cracks a downpipe is worth recording too, and most decent reporting systems take both, but keep them distinguishable. Property damage tends to get chased because somebody has to pay for it. Near misses only get chased if you decide they matter.
Reading out the last fortnight's near misses and what got fixed is one of the most useful five minutes you can spend, and it slots straight into your toolbox talks without any extra preparation.
Why the firms that record them have fewer accidents
Because near misses are free warnings. The same conditions that dropped that board with nobody underneath will drop the next one with a plasterer underneath, and the only difference between the two events is luck. Luck is not a control measure. If you find out about the failure while it's still cheap, you get to fix it before it invoices you.
This is where the accident triangle usually appears, so let's deal with it honestly. Herbert Heinrich published his ratios of major injuries to minor injuries to no-injury incidents back in the 1930s, based on insurance records that nobody can now check, and safety researchers have spent decades pointing out the problems with both the data and the assumption that the bottom of the pyramid causes the top. Treat the triangle as an interesting bit of history, not as a law of physics. Anybody quoting you a precise ratio is quoting a guess.
The underlying principle survives the criticism perfectly well without any numbers attached. Serious injuries almost never come out of nowhere. Go back through the chain on a bad one and you'll usually find the housekeeping had been poor for a fortnight, the edge protection had been altered twice, the loading bay gate had been tied back for convenience since Easter. All of that was visible beforehand. Somebody probably mentioned it in the van.
There's a second reason, less noble but real. A firm where people bother to report near misses is a firm where people are looking. Reporting is a symptom of attention as much as a cause of safety, and once you have a steady trickle of reports you can see your job the way your operatives see it, which is not the way it looks in the construction phase plan.
On a small job the cheapest workable system is a couple of honest lines in the site diary plus a photo on the phone, dated and kept.
What a usable near-miss report looks like
Two minutes on a phone, standing up, in the rain, with gloves on. That is the whole design brief. If your report takes longer than the time it takes to make a brew, it will get done in the office on Friday from memory, or more likely not at all.
Four things earn their place: what happened, where, when, and what could have happened. Add a photo, because a photo of the board and the gap in the toe board says more than three lines of prose ever will. Name the reporter if they're happy to be named, and allow anonymous if they're not. Everything else, the reference numbers, the risk-matrix score, the drop-down for root cause category, can be added later by whoever reviews it. Don't make the person nearest the hazard do your admin.
The killer detail is the "what could have happened" line, and it's the one most forms leave out. "Board fell from fourth lift" gets filed. "Board fell from fourth lift onto the delivery route, driver had walked through 90 seconds earlier" gets acted on that afternoon. Same event, different urgency, and the urgency is the whole point.
One live objection: not everyone on site has a smartphone they'll use for work, and not everyone reads or writes English comfortably. So keep a paper pad in the cabin and let the supervisor type it up, and accept a verbal report at any time. A near miss told to the foreman at the tea break and written down by him is a proper report. A near miss that nobody mentions because the form is on a laptop in the office is nothing at all.
Near misses involving chemical splashes, dust clouds or a decanted substance in an unlabelled bottle should send you straight back to your COSHH assessments, because the assessment has clearly been overtaken by what's happening on site.
Blame is what kills near-miss reporting
Nobody reports a near miss that makes them look like an idiot, and most near misses make somebody look like an idiot. That's the whole difficulty. The lad who nearly went through an unprotected riser opening was, if we're honest, walking backwards while carrying a length of trunking. He knows what the conversation will be like.
So decide in advance what happens to a reporter, and say it out loud. In most firms the honest position is this: report it and you're fine, including if you caused it. Hide it and you're not. Deliberate reckless behaviour still gets dealt with, and no sensible reading of blame-free means you ignore somebody cutting a handrail out to get a wheelbarrow through. But there's a wide gap between a genuine mistake and a two-fingers-up decision, and if you treat both the same way you'll only ever hear about near misses involving somebody else's subbie.
Watch out for the reporting-target trap as well. Set a target of one near miss per operative per month and you will get exactly that many, most of them saying "cable across walkway", written in the same handwriting on the last day of the month. Counting reports as a KPI turns a diagnostic tool into a paperwork exercise. If you must measure something, measure how many reports got a fix and how quickly.
And keep it away from the disciplinary file physically as well as in principle. If the same person who handles final written warnings also holds the near-miss box, the box stays empty.
If you're building the wider health and safety paperwork out from scratch, the guides hub covers the documents that near-miss trends usually end up amending.
What to do with the pattern once you have one
Individual near misses are worth fixing on the day. The pattern is worth more, and it only shows up if somebody actually reads the pile. Half an hour a month with the last thirty reports in front of you, looking for repeats, is more useful than any amount of reporting software.
Sort them by where and by what, not by who. Three reports in six weeks all involving loads swinging near the same corner is telling you the crane's approach or the delivery sequence is wrong, not that three different slingers were careless. Four dropped-object reports that all happen between 3pm and 4:30pm are telling you something about how the last hour of the shift gets rushed to make the wagon. A cluster on scaffold alterations says your handover and tagging process has a hole in it, and that one is worth taking seriously because unauthorised alterations are what turns a compliant scaffold into a fall from height under the Work at Height Regulations 2005.
Then be honest about which fixes actually remove the hazard. Retraining everyone and issuing a memo is the cheap answer and it's usually the weakest one. Moving the pedestrian route away from the loading bay, changing the delivery time so it isn't happening while the follow-on trades are moving through, getting a proper netted debris chute instead of the drop-and-shout arrangement: those change the physics of the job rather than asking people to be more careful.
Feed the recurring ones straight back into the daily paperwork. If the same near miss keeps happening, the risk assessment for that activity has been overtaken by reality and needs rewriting rather than re-signing. The site records you already keep are the natural home for the follow-up: what was done, on what date, by whom.
Which near misses must be reported to HSE
Some near misses are not just internal learning, they're legally reportable. RIDDOR 2013, Schedule 2 lists dangerous occurrences, and these are reportable by the responsible person (broadly the employer, the self-employed person, or the person in control of the premises) whether or not anybody was hurt. That's the crucial bit people miss: a dangerous occurrence is defined by what happened to the plant or the structure, not by whether there was an injury.
The construction ones you're most likely to meet, all from RIDDOR 2013, Schedule 2, include the collapse, overturning or failure of load-bearing parts of lifting equipment such as cranes, hoists and MEWPs; the complete or partial collapse of scaffolding over five metres high, or any scaffold erected over water where somebody could drown; the unintended collapse of any structure or falsework under construction, or of any wall or floor being used as a workplace; plant or equipment coming into unintended contact with an overhead electric line carrying more than 200 volts; and an electrical short circuit or overload that causes fire or explosion and stops the plant for more than 24 hours. Explosion or fire that stops work for more than 24 hours is on the list too, as is the accidental release of a substance likely to injure health.
Timing and method matter. Under RIDDOR 2013, dangerous occurrences must be reported to HSE without delay, and the practical route for most construction firms is the online form on HSE's website, with the telephone service kept for fatalities and specified injuries. RIDDOR 2013 also requires records of reportable incidents to be kept, so keep a copy of the submitted form rather than trusting that you'll remember what you typed.
Do not read the Schedule 2 list as the boundary of what's worth recording. A slinging failure that drops a bundle of studwork onto a slab from three metres is probably not a Schedule 2 dangerous occurrence, and it is absolutely a near miss you want in writing with a photo. Report the reportable ones because RIDDOR 2013 says you must. Record the rest because they're the ones that tell you where the next accident is coming from.
Making it stick on a small firm's site
Start smaller than you think. One route in, one person who reads them, one place the fixes get written down. A joiner and groundworks outfit with fifteen people does not need a categorised taxonomy of incident types; it needs the supervisor to have somewhere to put the four things he saw this week, and five minutes on Monday to talk about them.
Then handle the awkward middle bit, which is subcontractors. Near misses on your site involving somebody else's operatives are still your problem in practice, because they'll be the same hazards that catch your lads next. Say in the induction that reports are wanted from everybody on site regardless of who pays them, and resist the urge to bill anyone for the fix while you're still learning what the fix is. Under CDM 2015 the principal contractor has to coordinate the work and the site rules, and you can't coordinate what nobody tells you about.
Expect the volume to be lumpy. You'll get nothing for three weeks, then eleven reports in a fortnight after a genuine scare, then nothing again. That's normal and it isn't a sign the system's failing. What would worry me is a site with three hundred operatives and zero near misses for a quarter. Nothing nearly happened on a job that size? Of course it did. It just didn't get written down.
The cheapest version of all this is a couple of lines in the daily record and a photo on the phone. Not elegant. Perfectly good. Better than the immaculate near-miss procedure in the health and safety file that nobody on site has ever opened.
Questions people ask
- Do I have to report a near miss by law?
- Not all of them. There is no general legal duty to tell HSE about every near miss, but the ones that fall under RIDDOR 2013, Schedule 2 as dangerous occurrences must be reported by the responsible person without delay, injury or no injury. Everything else is internal learning, and recording it is good practice under HSE's incident reporting guidance rather than a statutory report.
- What's the difference between a near miss and a dangerous occurrence?
- A near miss is any event that could have caused harm and didn't. A dangerous occurrence is a specific list of events set out in RIDDOR 2013, Schedule 2, such as the failure of load-bearing parts of lifting equipment or the partial collapse of scaffolding over five metres high. Every dangerous occurrence is a near miss if nobody got hurt, but most near misses are not dangerous occurrences.
- Should near misses go in the accident book?
- You can keep them together, but keep them labelled clearly as no-injury events so you don't muddle your injury records. The accident book is where injuries to people go. Many firms run a separate near-miss log precisely so the accident book stays clean for insurers and RIDDOR 2013 record-keeping, and so the near-miss pile can be read as a trend rather than filed one page at a time.
- How do I get people to actually report them?
- Make it take two minutes on a phone, accept anonymous and verbal reports, and never punish somebody for reporting their own mistake. Then show the fix. People report when they've seen a previous report change something, and stop when reports disappear into an inbox. Reporting targets tend to produce fake reports rather than real ones.
- How long should I keep near-miss records?
- RIDDOR 2013 requires records of reportable incidents, including dangerous occurrences, to be kept, so hold those securely for the retention period set out in the regulations and current HSE guidance. For non-reportable near misses there's no set period, but keeping a couple of years' worth is genuinely useful because the patterns only show up over time, and because insurers and clients ask.
The short version
A near miss is an event that could have caused harm and didn't. It's a free warning about the same conditions that cause real injuries, which is why firms that collect them tend to have fewer accidents. The famous accident triangle is Heinrich's 1930s work and its ratios are contested, so take the principle and leave the numbers.
A usable report is blame-free, takes two minutes on a phone, and says what could have happened as well as what did. Read the pile monthly, fix the pattern rather than the person, and check every event against RIDDOR 2013, Schedule 2, because dangerous occurrences must go to HSE whether or not anybody was hurt.