Software can fix real problems with paper: lost forms, no follow-up, and a monthly report built by hand. It can also add new ones: logins nobody remembers, a form that takes longer on a phone than on paper, and a dashboard nobody opens. These are the questions that separate the two, written so they apply to any supplier.
This decides everything else. If a report takes longer than the event did, people stop reporting.
Ask the supplier to hand a phone to someone on your site, at the place where near misses happen, and time a real report. Check it works with gloves, in poor signal, and for people whose first language is not English. A QR code at each zone, a photo instead of typing, and a form short enough to finish in a minute all help. Our free near miss report form shows how few fields you actually need.
Agency staff, contractors and visiting drivers see a lot of near misses, and most will never have an account. Ask whether anyone on site can report, and how.
Then ask about names. Asking for a role rather than a name removes one of the main reasons people do not report. Where a name is needed for follow-up, ask who can see it, and whether that can be limited.
HSE's investigation guide, HSG245, asks investigators to "provide feedback to all parties involved". Silence is what kills reporting: people report twice, hear nothing, and stop.
Ask to see what the reporter receives, and when. A short message saying what was found and what changed is worth more than any dashboard.
Silence is what kills reporting. People report twice, hear nothing, and stop.
Most systems track an action to "closed". Closed is not the same as controlled. A barrier put back on Tuesday can be knocked down again by Friday.
Ask whether an action can stay open until someone has checked, on site, that the fix is still in place, and whether the system reports "closed but not checked" separately. This is the Check in HSE's Plan, Do, Check, Act, and ISO 45001 asks for the effectiveness of corrective actions to be reviewed, not just their completion.
A near miss has no injury, so the actual severity is always zero. What tells you which reports deserve an hour is the potential. HSE says the depth of an investigation should be set by "the potential consequences and the likelihood of the adverse event recurring".
Ask whether the form asks for the worst realistic outcome, and whether the system can sort and filter by it.
Most near misses are not reportable to HSE, but some are: RIDDOR's dangerous occurrences, such as a forklift overturning, must be reported even if nobody was hurt.
Ask whether the system prompts the RIDDOR question at the right moment, records whether and when a report was made, and keeps the records you need. It should help you decide and remember. It does not replace the decision, which is still yours. If in doubt, use our RIDDOR checker.
Reports often contain names, photos and sometimes health information, which is personal data under UK GDPR. Ask where the data is stored, who can see it, how long it is kept, and whether you can export everything, in a usable format, if you leave. A system you cannot leave is a risk in itself.
Every system gets a burst of reports at launch. What matters is month twelve. Ask for a reference at a comparable UK site, and ask them three things: how many reports a month now compared with the first month, how long it takes for a reporter to hear back, and what changed on site because of a report.
Every reporting system, paper or digital, records the near misses that someone noticed, recognised as a near miss, and chose to report. It cannot record:
That is not a reason to skip reporting software. It is a reason to read the numbers carefully. A rising count of reports usually means people trust the process more, not that the site is getting more dangerous. A falling count usually means the opposite.
Software is not the right first step for every site.
Software earns its place when there are several sites, several shifts, or more reports than one person can follow up by hand.
A distribution centre compares two apps. Each supplier gets the same test: two weeks, one zone (the goods-in dock), all three shifts.
| Test | App A | App B |
|---|---|---|
| Time for a night-shift driver to report, on the dock | 3 min 10 s | 55 s |
| Can agency staff report without an account? | No | Yes, by QR code |
| Reporter told what happened next? | Only if the manager remembers | Automatic message when closed |
| Action stays open until checked on site? | No | Yes |
| Export everything to Excel? | On request | Yes, any time |
App A had the better dashboard in the demo. App B got nine reports from agency staff in two weeks, where App A got none. The site chose B.
RoboK makes camera-based safety monitoring for sites with vehicles. It starts from what the cameras see, not from what people report, though it overlaps with near miss software: events seen on camera become records, each with a time, a type and a clip. That is why our own product is not in the eight questions above. For the questions to ask about camera systems, see choosing AI safety monitoring.
A report that takes about a minute on the shop floor; reporting without a login, and without a name if people prefer; feedback to the reporter; actions tracked until the fix is checked on site; a record of how bad it could have been; help with the RIDDOR question; and data you can export if you leave.
Some apps have free versions, often limited by users or features. For a single site with a few reports a month, a free one-page form and a spreadsheet log can work just as well. Ours is at near miss report form.
They should at least allow it. Asking for a role rather than a name removes one of the main reasons people do not report. Where a name is needed for follow-up, the site's policy should say who sees it.
Reporting to HSE under RIDDOR is the responsible person's decision and duty. Most near misses are not reportable; dangerous occurrences are. Software can prompt the question and keep the record, but check how any report to HSE is actually made.
Usually because reporting takes too long, nobody hears back, or people worry about blame. The software is rarely the cause, and rarely the cure, on its own.
Software records the near misses people report. A camera can show you the ones nobody did. Safety Mapping: a free two-week trial on one camera.
By Hao Zheng, CEO at RoboK · Last checked November 2026