Metric Drift Explained: 4 Forms That Make Safety Numbers Lose Their Line

A safety number can stay on the slide while the work underneath changes shape. That is metric drift, and it matters because leaders may keep using the same label after the denominator, source, or review habit has shifted enough to change what the chart means.

Metric drift is the slow separation between a safety indicator and the work it was built to describe. It matters when the chart still looks stable, yet the definition, denominator, source, or decision rule has moved far enough to make the trend less trustworthy than it appears.

Definition

Metric drift is not the same as real operational change. A plant may genuinely improve, worsen, or reset after a shutdown. Drift appears when the number changes for a reason that is hidden inside the metric itself. The label stays in place, but the meaning slides.

Across 25+ years in executive EHS roles, Andreza Araujo has seen this pattern repeat whenever leaders change the rule but keep the chart. In Safety Culture: From Theory to Practice, repeated decisions define culture, so a repeated change in measurement rules is also a culture signal, not just a reporting detail.

4 forms of drift

Definition drift

The name stays the same, but the inclusion rule changes. What counted as a near miss last quarter no longer counts this quarter, so the trend looks cleaner without becoming truer.

Denominator drift

The rate still uses a familiar base, yet the work mix has changed. Hours, tasks, shifts, or exposures no longer represent the same risk profile, which makes the comparison softer than it seems.

Source drift

The number now comes from a different system, owner, or collection habit. A dashboard that once reflected field evidence can slowly become a report about administrative convenience.

Decision drift

The metric once triggered action, but now it only fills a slide. When nobody can name the next step, the number has lost part of its operating role even if the formula never changed.

How to tell drift from a real change

The fastest test is simple. Ask who owns the number, what changed in the rule, and what evidence still links the chart to the field. If the answers are vague, the chart may be more polished than truthful.

James Reason's work on latent failures fits here because weak conditions often move before harm does. A metric can drift long before anyone notices a bad event, which is why the question is not only whether the number changed, but whether the decision behind the number still fits reality.

Metric drift vs metric hygiene

Metric drift is the symptom. Metric hygiene is the discipline that keeps the symptom from growing unnoticed. Hygiene asks whether the definition is stable, the source is named, the denominator still fits the exposure, and the action rule is still alive.

If your team already has a definition pack, the article on metric hygiene is the right companion. If the question is whether the indicator still proves a field condition, the leading-indicator quality audit article is the next step.

Andreza Araujo's practical view is direct. A number that no longer changes decisions is not a strong control signal, even when it looks neat in a dashboard. The organization should keep the measures that still speak clearly and retire the ones that only preserve the illusion of continuity.

What leaders should do first

Start with the three metrics that matter most in executive review. Check the last rule change, the current source, and the denominator against the current work pattern. If any of those three moved without a clear note, treat the trend as provisional until the rule is repaired.

Then turn the review into a habit. One owner, one definition log, one change note, and one field check are enough to stop most drift before it shapes a decision. If a metric still cannot pass that test, it belongs in the margin, not in the meeting.

Learn more about Andreza Araujo's work

FAQ

What is metric drift in safety?
Metric drift is the slow separation between a safety indicator and the work it was meant to describe. The number still exists, but the rule, source, or denominator has moved enough to weaken trust in the trend.

How is metric drift different from poor performance?
Poor performance is a real operational problem. Drift is a measurement problem. The site may be getting better or worse, but the chart no longer proves which one is true until the rule is checked.

What should EHS do first when drift is suspected?
EHS should check the definition log, source system, denominator, and current decision rule. If the chart changed after a rule change, the trend should be labeled carefully until the indicator is rebuilt or re-baselined.