Download the near-miss log (Excel, with printable form) (XLSX · 36 KB)
Every site has a near-miss form. Most of them were written by the last safety manager, ask for too much, and sit in a drawer or a SharePoint folder nobody opens from the floor. The result is a log that records the near misses people had time to write up, which is a small and unrepresentative sample of the ones that happened.
This template is built on three ideas from public guidance, and one from experience.
As little as possible at the point of reporting. HSE's Investigating accidents and incidents (HSG245) is clear that the quality of an investigation depends on information gathered close to the event; the form's job is to capture that, not to complete the investigation.
The report half of the form asks for fourteen things, and nine of them are a date, a dropdown or a tick box. (Printable version, PDF) The free-text fields are two: what happened, in one or two sentences, and what was done on the day. Everything else is a choice from a list the site sets up once: zone, shift, type, immediate cause.
Two fields are deliberately prominent:
Time, to the minute. A near miss reported as "about 2pm" costs half a day when someone looks for the footage. "14:07" costs a minute.
Could it have been worse? Minor, serious, or life-changing: the potential outcome, not the actual one. A near miss is by definition an event with no injury, so actual severity is always zero. Potential severity is what tells you which of this month's forty reports deserves an hour.
Because closing an action is not the same as controlling a risk. Most logs end at "Closed on". This one has a further column: the date someone looked and found the fix still in place.
The reasoning comes from the Check stage of Plan, Do, Check, Act (HSG65) and from ISO 45001 clause 10.2, which asks for the effectiveness of corrective action to be reviewed, not just its completion. A barrier that was reinstated on Tuesday and knocked down again on Friday is closed on the log and open on the yard. The summary sheet counts closed but not checked separately, and that count is the one worth putting in front of the site lead.
A barrier reinstated on Tuesday and knocked down on Friday is closed on the log and open on the yard.
Nothing clever: counts by month, by zone, by type and by potential severity; open follow-ups; closed-but-not-checked; and reports per week over the last eight weeks.
That last one needs a word. A rising count of near-miss reports is usually a sign that people trust the process, not that the site is getting more dangerous, and a falling count is more often a sign of reporting fatigue than of improvement. Read it as a measure of the reporting culture, and read the zone and severity counts as the measure of the site.
Three things, none of them about the form.
Make it one minute. If the form takes longer than the event did, people stop. The printed version fits on one page; the photo-and-send route ("photograph the form and send it to…") is on the footer because for most people that is faster than typing.
Close the loop visibly. Every report gets an owner and a date, and the person who reported it hears what happened. HSG245 puts this plainly: people report when they see reports lead to action. Nothing suppresses reporting faster than silence.
Never attach consequences to reporting. The form asks for role, not name, by default. A site can change that; most should not. The moment a report can be held against the person who made it, the log becomes a record of the near misses nobody minded admitting.
It records what people report. It cannot record the pedestrian who stepped behind the trailer at 05:40 when nobody was watching, or the driver who has reversed out of bay 4 without looking every day for a month without incident. Those are the near misses that precede the reportable one, and a form cannot catch them because no one experiences them as an event.
That is the limit of every log, including this one. It is worth knowing where it is.
Not on its own. RIDDOR 2013 requires reporting of deaths, specified injuries, over-seven-day incapacitation, injuries to non-workers taken to hospital, and listed dangerous occurrences. A near miss with no injury and no dangerous occurrence is not reportable, but should be recorded and investigated in proportion. See Is it RIDDOR reportable?.
Date, time to the minute, location, who reported it, what happened in one or two sentences, the type of event, who and what was involved, how bad it could have been, the immediate cause, what was done on the day, and any evidence such as photos or the camera and time for CCTV. Follow-up owner, due date, action and closure are added afterwards.
RIDDOR requires reportable incident records to be kept for three years. Near misses are not covered by that rule, but keeping them for at least as long is sensible: they are the context an investigation into a later incident will need.
They should at least allow it. Asking for a role rather than a name removes the main reason people do not report. Where a name is needed for follow-up, the site's policy should say who sees it and that reporting carries no consequence.
There is no benchmark figure in the guidance, and comparisons between sites are unreliable because the rate depends mostly on reporting culture. A more useful test is direction and spread: are reports coming from every zone and every shift, or from the people who already report?
Forms catch what people report. What is happening that nobody reports? Safety Mapping: a free two-week trial on one camera.
Published 20 October 2026 · Last updated 20 October 2026