How we investigate: the same way, every time

Every case starts the same way, however serious the outcome. A near miss and a lost-time injury go through the same front end, in the same order, with the same questions.

That consistency is what makes the results worth something later. A safety manager at one plant, a supervisor at another, and a director across forty sites follow the same steps, so their findings can be compared. When the same causes keep showing up across cases, they stand out. They can only stand out if every case was looked at the same way.

A method that changes with the person running it, the pressure of the week, or how the story sounds on the first day produces answers you can't line up side by side. A repeatable process is what turns one investigation into something the whole company can learn from.

Every case looks at three things, and always in this order:

1. 1 · PaperWhat was written down: the procedure, the permit, the training record.
2. 2 · PeopleWhat the people who were there say happened.
3. 3 · PlacesWhat the place itself shows: the equipment, the layout, the job as it's really done.

The order is not decoration. By the time anyone asks a question about how a person performed, you already know what the company asked of them on paper. That is what keeps an investigation from turning into a search for someone to blame. Most of the time, the person closest to the incident was doing the job the way it made sense given the conditions and resources available at the time.

You cannot inform yourself. Everything worth knowing lives outside your head: in a file, in a person, or in a place. An answer only counts once the record behind it is attached: the document you read, the conversation you had, what you saw on the floor.

If you can't find out, the honest answer is Unknown, and Unknown is allowed. What isn't allowed is filling the gap with a guess. A finding built on an assumption reads the same as one built on a check, and a year from now nobody can tell them apart.

Whoever gives an answer can defend it later, because it says where it came from.

Interviews ask what happened, where, who was there, and when. They never ask a person to explain or justify what they did: no "why did you do that?", no "why didn't you just…"

If the person could have explained it in the moment, the incident probably wouldn't have happened. Asked early, a demand for reasons puts people on the defensive. It gets you an explanation shaped to sound reasonable instead of a picture of what the work was actually like. Worse, it teaches everyone watching that talking to the safety team is risky, and that cost lasts long after one case closes.

The reasons come out of the analysis, once the context is laid out: what was written, what was said, what the place shows. It's something the investigation finds, never something it demands.

"I know it was wrong, I was just in a hurry." People often say this to end the process and get back to work. It's honest, and it's worth writing down. But it's the beginning of a question, not the answer to one.

A hurry has a source: the pace, the staffing, the way the job is set up. That source is what the investigation follows. A case that stops at the admission fixes nothing, and the next person in the same spot will say the same thing. So an admission is a flag the investigation follows up. It never closes the case or shortens it.

Real incidents rarely have one cause. There are usually several, and stopping at the first one that looks right guarantees you miss the others. So each cause is looked at on its own, with the same questions, rather than one quick pass over the whole event.

And a case is finished when the work is done, not when someone says it's done. PraxysEHS works out what's complete from what has actually been checked. There's no checkbox that marks a step finished, because any checkbox that can be checked to mean "done" will eventually be checked just to get something off the list.

And there's no "Other" category, no "not paying attention," no catch-all to put a cause in when it doesn't fit. When nothing fits, that's worth knowing too, and it's recorded in plain words and looked at, not forced into the nearest box.

Not every incident needs a full investigation. A paper cut from filing a permit shouldn't take the same effort as a serious injury, and a method that treats them the same way teaches people to stop reporting the small ones. The small ones are often where the warning signs are.

So there is one honest way to close a small case early. It has to be earned with checks that were actually made, someone else confirms it, and it's never open when someone was, or could have been, badly hurt.

This isn't a shortcut. It's never granted on anyone's word alone. Nothing is deleted, and the case can be reopened into a full investigation at any time.

A finding tells you what went wrong. Its code tells you what kind of problem it is, and so what kind of fix it needs. Two findings that sound alike can call for very different fixes, and the code is what keeps them apart.

Codes are also how patterns become visible. One case says the same procedure was too hard to follow. Ten cases across three sites saying the same thing is a decision waiting to be made. The codes serve the investigation, never the other way around.

Every fix says what kind of fix it is, strongest first: remove it, replace it, change the work itself, change how the work is organized, train or brief people, and protective equipment last. The strongest fixes make the safe way happen without anyone having to decide to do it. Retraining someone for a problem built into the job is a fix with a shelf life of about six weeks.

So a finding isn't closed by training or protective equipment alone. Those have their place. They just aren't a substitute for changing how the work is set up. Write the fix the problem calls for, not the one that's easiest to schedule, and decide up front how you'll know it worked.

The person who runs an investigation is too close to it to be the one who says it's finished. They know what they meant to check; a fresh reader sees what was actually checked. So a case closes only when someone who didn't work it reads it and signs.

When a company is small enough that there's nobody else, the investigator can sign their own case, but the record says so plainly. Nothing about it is hidden.