How many steps are there in the accident investigation process? Most structured programs use nine, from securing the scene through to feeding findings back into the safety program. Skip one, and the investigation usually falls apart right where it matters most: root cause.
Most structured accident investigation processes use nine steps: securing the scene, reporting, evidence collection, timeline reconstruction, root cause analysis, corrective action planning, implementation, verification, and feeding findings back into the safety program.
OSHA itself prefers the term incident investigation over accident investigation, since calling something an accident implies it was random and unavoidable. Nearly every workplace incident traces back to a cause that could have been caught. That distinction shapes how we run an investigation: the goal is never to assign blame, it's to find the gap in the system that let the incident happen.
We've written separately about what OSHA expects to see documented in a completed investigation report, covering recordkeeping timelines and the core elements a written report needs. This post is the walkthrough that comes before the paperwork: what actually needs to happen, in order, from the moment an incident occurs through verifying the fix worked.
A well-run investigation also protects you outside the incident itself. Prequalification platforms often ask about your incident investigation process directly, and insurance carriers factor it into experience modification calculations after a claim. A documented, repeatable process is worth more in those conversations than a folder of one-off incident reports that don't share a consistent structure.
Get injured workers care first; nothing else in this process matters until that's underway. Once a medical response is in motion, the supervisor on duty should restrict access to the immediate area, using barrier tape or a physical barrier where the layout allows it, and hold the scene as close to its original condition as safely possible.
This step creates the most friction on an active jobsite, because production pressure pushes toward getting equipment back in service and clearing the area fast. Resist that pressure until the scene has been documented. Evidence that gets moved, cleaned up, or put back into service before anyone photographs or measures it is evidence the investigation can't recover later, and a rushed scene is a common reason root cause analysis stalls out at step five.
Report through your internal process immediately, and separately confirm whether the incident meets OSHA's reporting thresholds; the two obligations run on different clocks and neither one waits for the other. A workplace fatality must be reported within 8 hours, while an in-patient hospitalization, amputation, or loss of an eye must be reported within 24 hours, under OSHA's recordkeeping requirements.
Whoever owns safety at your company should know these time windows without having to look them up under pressure. Missing a notification deadline creates a separate compliance problem on top of whatever caused the incident itself, and it's an easy one to avoid with a clear internal protocol for who calls OSHA and when.
Photograph the scene from multiple angles before anything is disturbed, and don't stop at the point of injury. Environmental conditions matter too: lighting, weather, noise levels, housekeeping, and anything unusual about the layout that day. Collect physical evidence, including any equipment, tools, or PPE involved, and tag it before it goes back into circulation.
Interview witnesses separately and as close to the incident as possible. Memory degrades fast, and witness accounts collected a week later are noticeably less reliable, and more likely to have absorbed details from conversations with coworkers, than ones collected the same day. Interviewing people separately, rather than as a group, keeps one person's account from shaping everyone else's.
Lay out what happened in order, using the evidence and witness statements gathered in the previous step. Build an actual timeline, not just a narrative summary, and mark the specific point where the normal sequence of work turned into the sequence that led to the incident.
Where witness accounts disagree, don't just average them out. A discrepancy is often a clue on its own, pointing to a step that wasn't as standardized as everyone assumed. A clear, evidence-backed timeline is what separates an investigation that identifies a real cause from one that just documents an outcome everyone already knew.
This is the step most investigations get wrong, and it's the one that determines whether the same incident happens again. It's easy, and often misleading, to stop at "worker error" or a missed procedure as the explanation, because that's usually where the visible chain of events ends.
A simple technique for pushing past that point is to ask why repeatedly, sometimes called the 5 Whys: why wasn't the procedure followed, why was that condition allowed to exist, and so on until the answer stops being about the individual and starts being about the system. Was training inadequate? Was the equipment the wrong fit for the task? Was production pressure a factor in the moment? The immediate cause is what happened. The root cause is why the system allowed it to happen, and it's usually a few layers deeper than the first honest answer anyone gives.
Every corrective action needs a named owner and a deadline, not just a description of what should change. Where possible, favor higher-order fixes on OSHA's hierarchy of controls, elimination, substitution, or engineering controls, over administrative controls or PPE alone, since the further down that hierarchy a fix lands, the more it depends on someone doing something correctly every single time.
A plan with no owner is a suggestion, not a plan, and it rarely survives past the meeting where it was written down. Assign each corrective action to a specific person, not a department, and give it a deadline that's realistic enough to hit and specific enough to enforce.
Put the plan into action and tell the affected crew what changed and why, not just that something changed. Workers are more likely to follow a new procedure when they understand the incident that drove it, rather than experiencing it as an arbitrary new rule.
Corrective actions that get implemented quietly, without explanation, tend to get quietly ignored once the memory of the incident fades and the old habit reasserts itself. A short toolbox talk explaining the change costs almost nothing next to the cost of the incident it's meant to prevent from happening again.
Check back after the corrective action has been in place for a reasonable period, not just at the point it's marked closed. A new guard, procedure, or training requirement needs to be observed in actual use, under normal production pressure, before anyone can call it effective.
Verification is where a lot of programs stop short, closing out the action item on paper without confirming the underlying hazard is actually gone. That gap is exactly how the same root cause resurfaces months later in a different form, as the example below shows.
Update your written safety program, training materials, and any related procedures so the lesson from this investigation doesn't stay isolated to one incident report. If the same hazard exists at other locations or on other crews, the fix needs to reach them too, not just the site where the incident happened.
This is also where findings should connect to your broader audit and safety management system, so the same gap doesn't surface again in a different form somewhere else in the operation. An investigation that ends with a closed incident report, but never touches the safety program itself, has fixed a symptom without addressing the system.
We see the same handful of mistakes across client investigations, regardless of industry.
Here’s a perfect example of this kind of situation. A contractor closed out a forklift near-miss investigation by retraining the operator on the immediate cause: not checking the mirror before backing up. Six months later, a nearly identical near-miss happened on a different shift, involving a different operator who had never received that retraining.
The root cause was never addressed: a blind corner in the aisle layout with no mirror or spotter requirement built into the area itself. Retraining one operator fixed the symptom for as long as that operator remembered the correction. It didn't fix the layout, so the same hazard was waiting for the next person who worked that aisle. A verification step at 30 or 60 days, checking whether the blind corner itself had been addressed rather than just confirming the training record, would have caught the gap before the second near miss happened.
Verification gets skipped more than any other step in the process, mostly because it doesn't feel urgent once the corrective action is marked complete and the incident stops being the thing everyone's talking about. But an unverified fix is still an open hazard wearing a closed status. Build verification into your safety calendar the same way you'd schedule a recurring inspection, not as an afterthought that depends on someone remembering.
Most structured programs use nine: securing the scene, reporting, evidence collection, timeline reconstruction, root cause analysis, corrective action planning, implementation, verification, and feeding findings back into the safety program.
A supervisor or safety professional trained in root cause analysis should lead it, ideally with input from employees who work the area day to day. Combining leadership experience with frontline knowledge tends to surface causes that either group would miss alone.
A near miss didn't result in injury, while a recordable incident did. The investigation process itself doesn't change much between the two, since a near miss often has the same root cause as a future injury that simply hasn't happened yet.
As soon as injured workers have received care and the scene is safe to approach, ideally within hours rather than days. Evidence and witness recall both degrade quickly.
Every incident and near miss deserves some level of review, though the depth should scale with severity. A minor first-aid case may need a short internal review, while anything involving a recordable injury or a significant near miss warrants the full nine-step process.
The 5 Whys, repeatedly asking why a condition existed until the answer points to a system issue rather than an individual, is the simplest and most widely used. For more complex incidents with multiple contributing factors, a fishbone (Ishikawa) diagram can help map how several causes intersect at once.
If a recent incident has exposed gaps in how your team investigates, or you're building an investigation process from scratch, a SafetyPro consultant can help you put one in place that holds up under an audit.
Lance Roux, CSP, is the founder and principal consultant at SafetyPro Resources, LLC. He has nearly three decades of experience across construction, oil and gas, manufacturing, chemical processing, refinery, and power generation industries, including conducting incident investigations for Process Safety Management-covered facilities along the Gulf Coast. Lance serves as an expert witness in workplace incident litigation and is the past president of the ASSP Louisiana Chapter. SafetyPro Resources is based in Baton Rouge and serves Louisiana, Texas, Mississippi, and the broader Gulf Coast.
Need help building an incident investigation process that holds up under an OSHA inspection? Talk to a SafetyPro consultant.