Safety Indicators Explained: 4 Types Leaders Should Not Mix

A safety indicator is a signal that helps leaders read control, exposure, or change. The trouble starts when one number tries to do several jobs, because then the team cannot tell whether it is looking at past harm, early warning, weak signals, or proof that a control still holds.

This article is for managers who need a clean taxonomy before the next dashboard review. If you already see why metric ownership matters, metric ownership shows how the owner, the threshold, and the field check keep a measure from drifting into theater.

Across 25+ years in executive EHS, Andreza Araujo has seen the same mistake repeat: leaders ask one indicator to prove too much. In Safety Culture: From Theory to Practice, the useful habit is the same as in James Reason's work on latent failure, which is to separate the signal from the story before the story becomes a decision.

Definition

Safety indicators are useful only when each one answers a specific question. A lagging indicator asks what already happened. A leading indicator asks whether control is moving. A precursor indicator asks where a weak signal is forming. A control check asks whether the barrier still exists now.

That distinction matters because a clean dashboard can still be weak if every number claims to show the whole picture. The article on SIF rate, TRIR, and precursors makes the same point from a broader executive angle. Not every measure belongs in the same review slot.

The 4 types

Lagging indicator

A lagging indicator describes what has already landed. It is useful for trend review, regulatory context, and learning after the fact, but it is not an early warning tool. If a team uses it as proof of prevention, the dashboard starts to flatter the past instead of governing the future.

Leading indicator

A leading indicator shows whether the conditions that support control are improving. It can be a review cadence, a verification rate, a closeout time, or another measure that changes before harm appears. It only works when the measure still connects to a real decision in the field.

Precursor indicator

A precursor indicator points to a weak signal that often sits close to a serious event. It is not just activity volume. It should name the scenario, the control that is under strain, and the action that should follow if the signal appears again.

Control check

A control check asks whether the barrier still exists and still works. This is the hardest one to fake, because it demands live verification, not slide language. If the check cannot be tied to field evidence, it is only reporting confidence, not control.

How to differentiate in practice

The easiest way to separate the four types is to ask what decision each one should change. If the answer is "look back," you are probably dealing with a lagging indicator. If the answer is "reassess the routine," you are probably dealing with a leading indicator. If the answer is "watch this weak signal closely," you are likely in precursor territory. If the answer is "prove the barrier is still live," you need a control check.

For a practical companion, the article on building a safety metric dictionary in 30 days helps the team standardize definitions before the dashboard grows any further. That step keeps one site from calling the same thing by four different names.

When to use one vs another

Use lagging indicators when the goal is trend memory. Use leading indicators when the goal is to see whether control is getting stronger. Use precursor indicators when the team needs a sharper signal near serious exposure. Use control checks when someone must prove the barrier still works in the field.

Andreza Araujo's view is simple: if one number has to do all four jobs, the review becomes vague. A safer system gives each number one job, one owner, and one decision. That is how a dashboard stops being decoration and starts becoming management.

If your team still treats all safety indicators as the same thing, the board may feel informed while the field stays under-read.

For the next step, compare this taxonomy with metric ownership, then bring the definitions back to the worksite and test them against the controls people actually use. If you want the broader body of work, Andreza Araujo is the place to start.