Accident investigation steps

Accidents at work rarely come out of nowhere. They are usually the result of conditions, decisions and controls lining up in the wrong way. An accident investigation is the process that helps you understand how that happened, so the same conditions don’t exist tomorrow, next week or on another site.

An investigation should not be about assigning blame or completing paperwork. It’s about protecting people, preserving evidence and learning from real work as it is actually carried out. Done poorly, investigations become rushed clean-ups leading to surface-level conclusions that ignore the underlying risks.

This guide sets out a practical, UK-focused approach to accident investigation. It covers what to do immediately after an incident, how to secure the scene, gather evidence, interview witnesses fairly and analyse causes properly. It also explains how to move from findings to corrective actions that actually reduce risk, track them to completion and share learning so mistakes are not repeated elsewhere.

Accident investigation steps checklist

When an incident happens, workers often feel pressure to tidy up, restart work and move on. That instinct can result in lost evidence, false assumptions and guesswork. A simple checklist gives structure at a stressful moment, helping you secure facts and make decisions based on what actually happened rather than what feels convenient.

Here’s a practical checklist you can print, keep in vehicles or include in your incident response plan:

• Make the situation safe and provide first aid.
• Get urgent medical help if needed and contact emergency services where required.
• Secure the area and preserve evidence.
• Notify the right people internally (supervisor, manager, health and safety officer, site lead).
• Decide whether the incident is reportable under RIDDOR and, if so, report within the required timescales.
• Record initial details while memories are fresh (who, what, where, when).
• Take photos and sketches before anything changes.
• Identify witnesses and capture early accounts.
• Collect relevant documents (RAMS, permits, training, inspections, maintenance).
• Choose a proportionate investigation method for the risk.
• Identify immediate causes, underlying causes and root causes.
• Agree on corrective actions that change the system, not just behaviour.
• Assign owners and deadlines, then track actions to closure.
• Share learning with the workforce and contractors.
• Review effectiveness after changes are in place.

Although this looks like a long list, most steps fit naturally into the HSE’s four-step approach to incident investigation. That approach is to gather the facts, analyse what happened, identify suitable controls and act on the findings.

The checklist helps you move through those stages in a calm, structured way: securing the scene and evidence first, understanding causes rather than assumptions, then putting controls in place and checking they actually work. The detail matters most at the start – once evidence is lost, it cannot be recovered.

What to do immediately after an accident

The first actions taken after an accident matter. Handled well, they protect people and preserve the evidence needed to understand what happened. Handled poorly, they can create further risk or erase the facts that explain why the incident occurred.

Start with people. Check for danger, give first aid and get help. If the incident is serious, call 999 and follow your emergency plan. While that happens, make sure someone takes responsibility for the area so the scene doesn’t become crowded with people.

Next, record the basics. Even a quick note on a phone can help later:

• Time and date
• Exact location
• Who was involved
• Brief description of what happened
• Equipment or substances involved

Then, decide what happens to the work itself. In some cases, the task must stop completely so equipment can be isolated and the area secured. In others, it may be safe for unrelated work to continue nearby while the incident area remains controlled. What matters is that this decision is deliberate and documented – based on risk, not on pressure to “get back to normal”.

Finally, think about legal reporting. You don’t need to know everything immediately, but you should flag whether the incident might be reportable under RIDDOR. HSE explains when you must report and highlights that many reportable events must be notified “without delay”, with a written report received within set deadlines. If you need a reliable reference point, use HSE’s guidance on when to make a RIDDOR report.

How to secure the scene safely

Securing the scene is about safety first, then evidence preservation. Never freeze a dangerous area “for investigation” if it puts people at risk. Your first question should always be: what must be done now to prevent further harm?

Here’s a safe approach to follow:

• Stop the activity that caused the incident.
• Isolate energy sources (electrical, mechanical, hydraulic, pneumatic, gravity).
• Stabilise loads and structures where there is any risk of collapse.
• Prevent access using barriers, cones, tape or locked doors.
• Assign one person to control access and log who enters and why.

Once the scene is safe, you can work to preserve evidence.

• If you need to move something for safety, document it first.
• Take wide photos that show the full area, then medium shots, then close-ups.
• Capture lighting, signage, floor condition and weather if relevant.
• If you can, include a reference object for scale.

You can borrow good practice from broader incident scene guidance: keep a clear boundary, log actions and take a careful, holistic view of the scene so you don’t miss details that later become important.

In higher-risk sectors, secure the paperwork too. For example, immediately retain the permit to work, lifting plan, isolation certificate or pre-use check sheet, and prevent them from being altered or completed retrospectively. These records often explain what controls were meant to be in place, and they can disappear or change quickly when people feel under pressure.

Accident investigation roles and responsibilities

Investigations go wrong when nobody is clearly in charge, or when roles overlap. The best approach is to define responsibilities in advance and use them consistently.

Most organisations benefit from three layers:

• The incident controller – this is usually the supervisor or manager on scene. They take control immediately, secure the area and start noting information for the initial accident record. They also decide what work stops and what can safely continue.
• The investigator or investigation lead – this person runs the investigation and writes the report. In a small business, this might be the manager. In a larger organisation, it may be a health and safety professional, an operational manager or a trained investigator. The key requirement is competence and enough independence to investigate fairly, without direct blame pressure.
• Specialists and support – depending on the incident, you might involve maintenance engineers, occupational health, HR, union or safety reps, or technical experts. For complex events, you may need to bring in external specialists.

To keep it fair, separate the following two functions: “care for people” and “finding causes”. The injured person needs support and compassion. At the same time, an investigation needs calm fact-finding. When you separate those goals, you avoid turning early conversations into interrogations.

Also clarify who makes decisions on adjustments and corrective actions. Investigators can recommend controls, but senior managers often own the resources and priorities. The action plan and implementation need management commitment to be effective.

How to collect evidence effectively

Evidence is anything that helps you understand what happened and why. You must act quickly and be methodical, because evidence can degrade fast.

Here are some useful types of evidence in accident investigation:

• Photographs and video, including CCTV where available
• Measurements, sketches and layouts (distances, heights, clearances)
• Physical items (broken parts, PPE, packaging, labels)
• Documents (risk assessments, method statements, permits, training records)
• Maintenance history and inspection records
• Digital data (machine logs, telemetry, access control logs)
• Environmental factors (weather, lighting readings, noise levels)
• Work planning evidence (schedules, shift patterns, workload changes)

Objectivity matters because investigation findings must be based on evidence, not opinion or hindsight. An objective investigation relies on what can be seen, measured or verified, such as physical conditions, documents, records and consistent witness accounts. Conclusions should follow from that evidence, rather than from assumptions about fault or intent. This is what makes an investigation fair, credible and defensible if it’s questioned later on.

These two practical tips help a lot:

• Label and store everything clearly, especially if equipment will be repaired or replaced.
• Keep an “evidence register” so you can show what you collected, when and by whom. That will make your report easier to defend in a review.

Questions to ask witnesses after an accident

Interviews can feel awkward, especially in situations where someone has been injured or emotions are high. However, good interviews are one of the fastest ways to uncover missing details and control weaknesses.

Aim for a calm, respectful tone. Set expectations from the outset. For example:

• You are not looking to blame anyone.
• You want to understand what happened and make work safer.
• You will focus on facts and conditions, and will not make a personal judgement.

Then use open questions that let the witness describe the event in their own words:

• Please talk me through what you saw, from the start.
• Where were you standing, and what could you see clearly?
• What task was being done, and what was the plan?
• What tools, equipment or vehicles were involved?
• Did anything change just before the incident?
• What made the task you were completing easier or harder that day?
• What controls were in place, and did they work as expected?
• What did you do immediately afterwards, and why?
• If you were doing the same task tomorrow, what would you want changed?

After that, ask clarifying questions that fill gaps in the timeline. Keep them neutral, and avoid leading language like, “Why didn’t you…?”

Be aware that different witnesses need different approaches. An injured person may only recall events up to the point of injury and may be distressed or fatigued, so interviews should be short and carefully paced. Where needed, they can be delayed until the individual is medically and emotionally able to take part. Quieter or less confident witnesses may need reassurance that their factual observations are helpful, even if they feel unsure or think what they saw is insignificant.

Finally, document interviews promptly. Write what was said, note the time and place, and ask the witness to confirm accuracy. If you use recordings, consider privacy and data protection requirements.

Accident investigation report template UK

A good report is simple, factual and action-focused. It should stand up to internal review, insurer scrutiny and regulator questions. It doesn’t need to be long to be clear.

Here is a practical UK-style template you can adapt.

1. Summary

• What happened, where and when
• Who was affected
• Actual outcome and worst credible outcome
• Immediate controls put in place

2. Incident classification

• Injury type or event type
• Whether it’s reportable under RIDDOR, and what was reported and when
• Any other notifications (client, principal contractor, insurer)

3. People and roles

• Details about the injured person and their work role
• Supervisor, manager and investigation team
• Witnesses

4. Timeline

• Pre-incident activity
• The incident sequence
• Immediate response

5. Evidence collected

• Photos, CCTV, sketches, diagrams
• Physical evidence
• Documents reviewed
• Measurements and tests

6. Findings

• Immediate causes
• Underlying causes
• Root causes
• Human factors and organisational factors

7. Risk control review

• Controls expected (as per RAMS, permits, procedures)
• Controls present in reality
• Control failures and gaps

8. Corrective actions

• Actions with owners, deadlines, and priority
• Control type (eliminate, substitute, engineer, administrative, PPE)
• Any interim controls

9. Communication and learning

• Who will be briefed
• Toolbox talk topics
• Procedure updates

10. Close-out and verification

• Evidence actions completed
• Follow-up date for effectiveness review
• Sign-off

If you want to strengthen your report further, include a short appendix with photos and a simple diagram of the area. That can prevent misunderstandings later.

How to identify immediate causes

Immediate causes are the direct conditions or actions that led straight to the incident. They are what you can see or confirm happened at the time.

Immediate causes commonly fall into two groups:

Unsafe conditions

• Slippery floor, poor lighting, missing guard, unstable load
• Poor separation between vehicles and pedestrians
• Defective equipment, worn components, missing signage

Unsafe actions

• Bypassing a guard, not using a restraint, entering a restricted zone
• Using the wrong tool, rushing, carrying awkward loads
• Working without a permit, or deviating from the established working method

Identifying these immediate causes matters because they often highlight hazards that need fixing quickly to prevent the accident from happening again. However, stopping at immediate causes only tells you what happened – not why the situation was able to occur in the first place. If you treat the visible trigger as the root of the problem, the same type of incident is likely to happen again under slightly different circumstances.

Root cause analysis methods compared

Root cause analysis (RCA) helps ensure that near misses and incidents don’t happen again. It looks at the system conditions that made the outcome possible in the first place – not the trigger.

There is no single best method for RCA. It depends on the potential severity of the event, how complex the situation is, and how experienced the investigation team is. Many organisations use a simple method first, then add structure if the picture becomes more complicated.

Below are common RCA methods, with guidance on when each works best.

5 whys

• Best for – straightforward incidents with a clear sequence of events
• Strength – quick to use and easy for supervisors and line managers to apply
• Limit – can become subjective if teams make assumptions or stop too early

The 5 whys method works well when the story is largely linear and you want to trace the issue back to a planning, resourcing or management decision.

Example:

• Near miss – forklift nearly hit a pedestrian.
• Why? – the pedestrian stepped into the vehicle’s route.
• Why? – the designated walkway was blocked.
• Why? – pallets were stored in the walkway.
• Why? – overflow storage was not planned.
• Why? – peak delivery volumes were not considered in the layout plan.

This points to a system issue with space planning, not individual behaviour.

Fishbone (Ishikawa) diagram

• Best for – incidents with multiple contributing factors
• Strength – helps teams think broadly across different influence areas
• Limit – identifies contributors, but further analysis is needed to decide which causes are truly “root”

Fishbone diagrams are useful when you suspect several interacting causes. Common categories include people, plant, procedures, environment and management. Each potential factor is tested against evidence rather than assumptions.

A useful rule of thumb is this: if you keep saying “and also…”, a fishbone is usually the better starting point.

Barrier analysis

• Best for – high-potential events, particularly in construction, logistics and process-based environments
• Strength – focuses directly on controls and what failed, was missing or was bypassed
• Limit – requires a clear understanding of what barriers should have been in place

Barrier analysis asks simple but powerful questions: what should have prevented this, and why didn’t it? That makes it especially effective for near misses that could have caused serious harm.

Fault tree analysis

• Best for – complex technical or equipment-related failures
• Strength – maps combinations of failures in a logical, structured way
• Limit – more time-consuming and requires specialist knowledge

This method is less common for routine near misses but valuable where multiple technical failures interact.

Bowtie thinking

• Best for – visualising hazards, controls and escalation factors
• Strength – excellent for communicating, learning and showing how controls fit together
• Limit – relies on good-quality underlying data to be accurate

Bowtie diagrams are often used after analysis to explain risk and controls clearly, rather than as the first investigation step.

Human factors in accident investigations

Broken equipment and rule-breaking are not the only reasons for accidents. Human factors are the conditions that shape how people perform. They include workload, attention, fatigue, design, training, supervision and culture.

Looking for human factors does not involve finding signs of human error and assigning blame.

Instead, they help you ask better questions, such as:

• Was the task designed to be done safely in the time given?
• Were controls easy to use, or were they awkward and time-consuming?
• Did the environment make mistakes more likely (noise, heat, poor visibility)?
• Was the worker new, unfamiliar with the site, or covering a different role?
• Was supervision present and effective at the time?
• Were there conflicting goals, such as speed versus safety?

For example, a worker might bypass a guard because the machine jams repeatedly and production targets are strict. At first glance, it might look like the accident happened because the worker broke the established safety rules – but underlying causes likely include poor maintenance, poor design, weak planning and high pressure.

Also consider “normalisation of deviance”. If a shortcut becomes common, people stop seeing it as risky. Investigations should look for signs of that drift, such as repeated minor near misses, informal workarounds or “everyone does it” language.

Treat human factors carefully, as they can be sensitive. Discussing them respectfully often helps unlock the real story and leads to better prevention.

Corrective actions and prevention measures

Reports should not be written and filed away if corrective actions haven’t yet taken place. The best actions change the system so the same conditions can’t line up again and cause another accident.

Start by separating actions into the following categories:

• Immediate fixes – these are actions that reduce risk right away, such as cleaning a spill, replacing a damaged guard or adding a temporary barrier.
• Longer-term prevention – changes like redesigning a process, changing layouts, improving maintenance planning or updating training help address root causes.

When selecting actions, use the hierarchy of control. Aim to eliminate or reduce risk at source before relying on warnings, reminders or personal protective equipment.

The examples below are system-level corrective actions, not behaviour fixes:

• Engineering controls – fixed guarding, interlocks, local exhaust ventilation or improved extraction
• Physical segregation – redesigned routes and solid barriers to separate vehicles and pedestrians
• Equipment selection – choosing lifting aids that match the load and task, rather than forcing manual handling
• Control checks – pre-use inspections that focus on high-risk failure points
• Work control systems – clearer permits to work that reflect real site hazards and up-to-date drawings
• Task-based training – hands-on training using real equipment and scenarios
• Supervision and resourcing – increased supervision during higher-risk periods, such as shift starts, shutdowns or night work

Each action should include a success measure so you can confirm it works in practice. For example, a redesigned route might be checked through observation results, near-miss trends or speed compliance data.

Choosing good actions is only half the challenge. Tracking them to completion is where many organisations lose momentum. Daily pressures take over, deadlines slip and the same hazards remain.

A simple tracking system prevents this. It should record:

• The action taken in plain language
• Risk level and priority
• A named owner with authority to act
• A due date that reflects the level of risk
• Status (open, in progress, closed)
• Evidence of completion (photo, document update, sign-off)
• An effectiveness check date

Whether you use software or a spreadsheet, actions need regular review. A practical approach is to review progress weekly, escalate overdue actions to senior leadership, close actions only when evidence is provided, and re-open them if effectiveness checks fail.

Finally, communicate progress. When workers see that reports lead to visible change, reporting increases and learning improves. A short “what changed” update – via a toolbox talk, noticeboard or briefing – reinforces that raising concerns leads to safer work.

Accident investigation timeframes in the UK

There is no single legal deadline for completing an internal accident investigation. However, there are very clear reasons to move quickly. Evidence fades. CCTV overwrites. People forget important details. Site conditions change. This is why you should start the investigation immediately and aim to complete it as soon as reasonably possible.

In practice, many organisations set internal timeframes such as:

• Initial report completed within 24 hours
• Scene evidence captured on the same shift
• Witness interviews within 48 hours, where possible
• High-potential incidents investigated with a draft report within 5 working days
• Corrective actions assigned within 7 days
• Effectiveness review within 30 to 90 days, depending on the changes

Alongside internal timeframes, you must also consider statutory reporting under RIDDOR for certain incidents. Most reportable incidents must be reported to HSE within 10 days. An “over-seven-day injury” means a work-related injury that prevents a worker from doing their normal job for more than seven consecutive days (not counting the day of the accident) – and these must be reported within 15 days of the incident. You can check the full reporting requirements in HSE’s guidance on when to report.

Record keeping supports this process and the investigation itself. Accident books and incident records provide an early factual account of what happened, who was involved and when it occurred. These records help confirm reporting decisions, support evidence gathering and create a clear timeline if the incident is reviewed later.

The principle is straightforward: start quickly, document early and stay focused. Even if a full investigation takes time, securing evidence and putting interim controls in place reduces the risk of a repeat while deeper analysis is ongoing.

Common accident investigation mistakes

After an accident, people want reassurance and quick answers. However, the tendency to find a quick solution and move on can hide risks that haven’t been dealt with.

Common mistakes include:

• Treating the investigation as a blame game
• Starting with a conclusion, then selecting evidence to fit it
• Failing to secure the scene, so evidence is lost
• Asking leading questions that push witnesses towards the story you want
• Stopping at “human error” without asking what made the error possible
• Focusing only on what the worker did, not on controls and planning
• Recommending actions that are only reminders or retraining
• Not assigning owners and deadlines for actions
• Closing actions without checking they work in practice
• Failing to share learning, so other teams repeat the same mistakes

You can avoid these mistakes by slowing the process down enough to follow a clear structure. Securing the scene, gathering evidence, analysing causes and then putting effective controls in place keeps the focus on how work is really done – not on who was involved and whether they made a mistake.

When actions are owned, tracked and checked for effectiveness, investigations can start preventing repeat incidents. That consistency is what turns accidents and near misses into genuine learning that guides stronger systems and safer everyday work.