Why injury counts are not enough to measure a safety program

A month with zero recordable injuries can describe a strong safety program or a lucky stretch. From the outside, the two look the same. That’s the problem with running a safety program on injury counts alone.

TL;DR: Injury totals are lagging results. They tell you what already went wrong, not whether your program is working. A useful safety dashboard also tracks the work that prevents injuries, including hazard reports, near misses, inspections, training, and whether corrective actions close on time.

I run a construction business. Roofing, sheet metal, crews on decks and edges and ladders every working day. Around here safety is the difference between a person going home at the end of the shift and a person not going home. So I care a lot about whether the number on the report is telling me something true.

A zero is a result, not a system

Here’s the uncomfortable part. A recordable injury already happened. By the time it lands on your dashboard, the hazard existed, the exposure happened, and somebody got hurt. Counting those events is worth doing. It’s still a long way from knowing whether your program works.

Think about what has to go wrong for an injury to occur. A condition gets created. Somebody either notices it or doesn’t. If they notice it, they either say something or they don’t. If they say something, it either gets fixed or it sits. If it sits, exposure continues until either luck runs out or the condition gets corrected.

The injury count measures the last link in that chain and nothing else. Everything before it is where the prevention work lives, and none of it shows up in a recordable total. You can have a bad month with no injuries. You can have a good month with one. One number at the end of a long chain can’t tell them apart.

So I don’t treat a clean month as proof of anything on its own. I’ll take it. I just can’t tell from the number alone whether the system produced it.

What the injury number does and doesn’t tell you

Lagging measures have real uses. Injury and illness rates let you compare periods, meet reporting obligations, and spot patterns that repeat. If the same body part or the same task keeps showing up, that’s information you can act on. I’m not arguing to throw the number away.

The limits are worth being honest about. Small crews and short periods produce small counts, and small counts move around for reasons that have nothing to do with your program. A crew of thirty on a six week project can go the whole job clean and it won’t tell you much about the program that put them there. The sample is too thin to carry the weight leaders put on it.

There’s a second limit that matters more. Injury counts measure failure, so they can be improved by reducing what gets reported instead of reducing what happens. That’s not a hypothetical risk. Any measure that goes down when people stay quiet is a measure that rewards quiet.

So the question I want answered goes past “how many people got hurt.” I want to know whether the work that prevents people from getting hurt is actually happening, and whether I can see it.

The indicators OSHA recommends tracking

There’s published guidance on this. OSHA’s guidance on program evaluation and improvement recommends measurable indicators such as reported hazards, near misses, inspections, management walkthroughs, training, and completion of corrective actions.

Read that list again and notice what it has in common. Every item describes activity that happens before an injury, not after. Hazards reported. Near misses reported. Inspections done. Managers on site walking the work. Training delivered. Corrections closed.

Those are things you can count while the job is still running. That’s the whole point. A hazard report tells you today that somebody saw something and said something. A corrective action that closed on time tells you the organization did something about it. Neither one requires waiting for an injury to generate a data point.

OSHA doesn’t say this next part, I do. What makes those indicators useful to a leader is that they’re countable and they’re current. You don’t have to wait for a bad outcome to know whether the process ran.

Reporting drops when speaking up carries a cost

Leading indicators come with their own trap, and leaders need to walk into it with their eyes open.

Most of the indicators above go up when people speak. Hazard reports go up when a foreman feels comfortable flagging something. Near miss reports go up when a crew member says “that could have been bad” without worrying about what happens next. If speaking up carries a cost, the numbers go down and the hazards don’t.

So a rising hazard report count is not automatically bad news. Early on it may be the best news on the dashboard. It can mean people believe the report goes somewhere. A hazard report count that drops to nothing is the line I would look at hardest, because there are two explanations and only one of them is good.

The way I think about it: if a crew tells me about a problem and the problem gets fixed, I just bought information at a low price. If a crew tells me about a problem and nothing happens, I taught them not to bother next time. That second lesson is expensive and it’s very hard to unteach.

The same logic applies to incentives. If you tie a bonus to a low injury count, you’ve attached money to a number that people can move by staying quiet. I’d rather attach recognition to the work itself. Inspections completed. Hazards found and closed. Training done on schedule.

What a useful safety review looks like

A safety review that only reads out the injury total is a status meeting with a safety label on it. Here are the questions I think are worth asking instead.

• What did we find this period, and who found it? If every hazard report comes from the safety manager, the field isn’t participating yet.
• How many corrective actions are open, and how old are the oldest ones? An open item with no age limit is a decision to accept the risk without saying so out loud.
• Which findings repeat? A hazard that shows up on three jobs is one gap in a process, a spec, or a training module, wearing three different job numbers.
• Where are the walkthroughs happening? If leadership only walks the jobs that are going well, the walk isn’t producing information.
• Who hasn’t been trained yet, and why? Training completion is one of the easiest gaps to see and one of the easiest to leave open.
• What did we change because of what we learned? If nothing changed, you held a meeting and called it a review.

None of these questions require a new system. Most of them require somebody to keep a list and read it out loud on a schedule.

How I would build the dashboard

Practical version. If you’re starting from an injury count and nothing else, you don’t need a platform. You need a handful of numbers you can trust and a rhythm for looking at them.

Start with four counts. Hazards and near misses reported. Inspections and leadership walkthroughs completed against what was planned. Training completed against who is required to have it. Corrective actions closed against corrective actions opened. Four lines. A spreadsheet handles this.

Then add an age column to the corrective actions. The count of open items matters less than how long they’ve been open. Sort oldest first. That sort order is the most honest view of a safety program I know of, because it shows what the organization keeps deciding not to do.

Report by job as well as company wide. Company totals hide the job that is drifting. If one project generates all the findings and another generates none, that gap is worth a conversation with both superintendents.

Set a review rhythm and hold it. Whether that’s weekly, monthly, or per project depends on your size and how fast your work turns over. What matters more than the interval is that it doesn’t get skipped when the schedule gets tight, because tight schedules are exactly when the corners get cut.

Keep the injury count on the page too. I’m not arguing to delete it. Put it next to the leading indicators so you can see them together. Over time the pattern between them is the interesting part.

And close the loop out loud. When a hazard report leads to a fix, say so, and say who reported it. That’s the cheapest thing on this list and it does more for reporting rates than any policy memo.

The takeaway

A low injury count is a result you want. By itself it stops short of proving your program earned it. The measures that tell you whether the program is working are the ones that describe the work being done before anyone gets hurt: hazards surfaced, walks completed, training delivered, corrections closed.

If your safety report has one number on it, you’re reading the scoreboard and skipping the game. Add the lines that show whether the prevention work is happening. Then read the oldest open corrective action out loud, every time, until the list is short.