Accident investigation: what to do when something goes wrong

When something goes wrong on site, the first few minutes matter more than anything you'll write down afterwards. Get people safe, get help, and protect the scene. Everything else, the statements, the photos, the paperwork, flows from that.

Accident investigation sounds like a job for big firms with a safety department. On a small site it's simpler than that. It's you, a notepad, your phone camera and a bit of method. This guide walks you through the whole thing: what to do in the minutes after an incident, how to gather the facts, how to work out what really went wrong, and how to stop it happening again.

Think of this as the companion to our RIDDOR reporting guide. Reporting is one step in the process, and a small one at that. The real value of an investigation is what you learn and what you change.

Why Investigate Accidents On Site

The Health and Safety Executive puts it plainly. You investigate to find out what went wrong, identify risks you could have avoided, and learn lessons that stop the same thing happening again. Notice what's missing from that list: finding someone to blame.

Done properly, a site accident investigation is one of the cheapest safety tools you have. One afternoon of honest fact-finding can save you a repeat incident, a lost-time injury, or a client who won't have you back.

Learning, Not Blame (Near Misses Count)

If your crew thinks an investigation means someone's getting sacked, you'll get silence. People will suddenly have seen nothing, and you'll learn nothing. Make it clear from day one that the point is to fix the job, not the person.

Near misses count too. A scaffold clip that fell from the third lift and missed someone by a foot is free information. Investigate it the same way, with far less pressure, and you've stopped the next one before it lands.

Secure The Scene And Preserve Evidence

People Before Paperwork: First Aid And Emergency Help

Before anything else, deal with the injured. Get first aid from your appointed first aider where applicable, call an ambulance if it's serious, and send someone to meet the crew at the road or gate, because small sites are often hard to find.

Only when everyone is treated and the area is stable does the investigation start. As you go, note down times in your phone, when it happened, when help arrived, who was called. Times blur quickly.

Should You Stop Work?

It depends on the situation. For anything serious, an injury needing hospital treatment, a structural failure, a dangerous occurrence, stop at least the affected area and anything sharing the same hazard. For a minor cut sorted with a plaster, a full stoppage helps nobody.

A short pause is often worth it even for small incidents. It gives you space to photograph the scene, and it tells the crew you take safety seriously. Where it's safe to carry on elsewhere on site, let the rest of the job run.

Preserving The Scene: Photos, Equipment And Positions

Evidence disappears fast on site. Plant gets moved, tools get tidied, mud gets trampled. Before anything shifts, take photos and video on your phone from several angles, wide and close, including where the person was standing and where the equipment ended up.

• Photos from wide and close angles, including the surroundings
• Positions of plant, tools, materials and barriers, marked with chalk or spray paint if they must be moved
• The state of edge protection, excavations, access routes and any temporary works
• Weather, lighting and ground conditions at the time
• The exact time, date, location and who was on site

What needs recording depends on the site, but the habit is the same everywhere: record before you disturb. If something has to be moved for safety reasons, photograph it first where possible, then note where it was.

Get The Facts (Witness Statements And Records)

Memories start editing themselves within hours. The people who saw what happened will talk it over at lunch, and by the next morning they'll all remember the same version, which may not be what actually happened.

Taking Witness Statements While Memories Are Fresh

Take short statements the same day, while everyone is still on site. Ten minutes each is plenty for most incidents. Ask open questions: what were you doing, what did you see, what did you hear, what happened next. Don't lead them and don't argue.

• Write what the person actually saw and heard, not what they think caused it
• Include their name, role, contact details, and the date and time of the statement
• Let them read it back, then both sign it
• Separate witnesses while you take statements, so accounts stay independent
• If someone is injured, take their account later, when they're able to give it

The injured person's version matters too, but it can wait until they're well enough. Don't chase anyone to an ambulance with a clipboard.

What Records To Keep

Pull together everything that shows the state of the job that day. On a small site that's a short list, and most of it should already exist if your paperwork's in order.

• The accident book entry, made the same day. Employers with 10 or more employees must keep an accident book under social security law, and keeping one regardless helps you spot patterns
• The RAMS and risk assessments in use for that task
• Permits, briefings and sign-in sheets for that day
• Training and competency records, and any card or ticket details where applicable
• Inspection and maintenance records for the plant or tools involved
• Photos, videos, statements and your own notes, kept together in one file

Keep the file even for incidents that turn out minor. If a claim or an HSE query lands two years later, a dated accident book entry and a few photos are worth their weight in gold.

Working Out What Went Wrong (Root Cause)

Most investigations stop at the first answer: he wasn't looking, the ladder slipped, the guard was off. Those are direct causes. The root cause is the thing underneath, the thing that, if you fix it, stops the same incident coming back through a different door.

Root Cause, Not Blame

Reportability under RIDDOR doesn't depend on finding blame, and neither does your investigation. People make mistakes inside the systems you build around them. If a guard came off because it fouled the cut, telling the operative off fixes nothing. Fixing the guard, the disc or the method does.

A Simple Five Whys Approach For Small Sites

The five whys method is exactly what it sounds like: ask why, then ask why again, up to five times, until you hit something you can actually fix. Here's a worked example, a cut hand on an angle grinder.

• Why was he cut? The grinder kicked back because the guard had been removed
• Why was the guard removed? It fouled the cut line on the steel
• Why was that grinder used for the job? It was the only one on site, and the RAMS assumed a smaller tool
• Why didn't the RAMS match the job? It was copied from a previous job and never checked when the scope changed
• Why wasn't it checked? There's no review step when the method or materials change

The root cause sits at the bottom: no review step when the job changes. Fix that and you fix every future version of this incident, not just this one. Five isn't a magic number. Stop when you reach something you can control, and stop earlier if you find yourself inventing reasons.

Corrective Actions And Prevention

An investigation that ends in a file is a waste of everyone's time. HSG245 is blunt about this: the point of identifying risk control measures is to produce an action plan and implement it. So decide your actions, give each one an owner and a date, and tick them off.

Updating Your Risk Assessment And RAMS After An Incident

An incident is proof that something in your paperwork or your practice didn't hold. After any investigation, review your risk assessment and the method statement for that task, and update them with what you learned.

Prefer fixes in this order: remove the hazard altogether, swap the method or material for something safer, add physical controls like guards and barriers, and only then lean on PPE and training. If your fix is "be more careful", it isn't a fix. Write the change into the RAMS, date it and version it, so everyone can see which edition they're working to.

Briefing The Team On What Changed

New controls only work if the people on site know about them. Run a short toolbox talk, walk the crew through what happened without naming and shaming, explain what's changed and why, and collect signatures on the updated RAMS.

Then check back in a week or two. Have the toe boards gone up? Is the smaller grinder actually in the van? Fixes that live only on paper are how the same accident happens twice, and the second one is much harder to explain.

RIDDOR Reporting And Legal Duty

Reporting is one step in the process, not the whole job. The Reporting of Injuries, Diseases and Dangerous Occurrences Regulations 2013 (RIDDOR) set out what must legally go to the HSE, and it's a shorter list than most people think. Many incidents only need recording, not reporting.

The main triggers in brief: death and specified injuries, which include fractures other than to fingers, thumbs or toes, amputations and serious burns; worker injuries causing more than 7 days of incapacitation; injuries to non-workers, such as members of the public, that involve them being taken directly to hospital for treatment; and dangerous occurrences from the listed categories. For the full detail on when and what you must report, read our RIDDOR reporting guide.

Two timing details trip people up. The 7 days exclude the day of the accident itself but include weekends and rest days. And if someone is off for more than 3 days but not more than 7, there's no report to the HSE, but you must record it in the accident book or your own records.

What An Accident Investigation Is Not

An investigation is not a disciplinary hearing. If someone was genuinely reckless or unfit to work, that's an employment matter to handle separately, with its own process. Mixing the two contaminates both, and word gets round. Next incident, nobody saw anything.

It's also not a courtroom, and you're not the judge. You're the person who works out how the job failed and fixes it. That mindset gets you honest statements, real root causes and a safer site. The other mindset gets you a tidy report and a repeat accident.