Safety KPIs: 6 Distortions That Make Good Numbers Hide Serious Risk

A safety dashboard can turn green while the worksite remains exposed to a serious event. That happens when leaders read activity as control, frequency as protection, and closure as effectiveness. The numbers may be accurate, yet the conclusion drawn from them is wrong.

The executive question is therefore not whether the KPI improved. It is whether the organization changed the conditions that could produce serious harm. The distinction matters because a low-frequency outcome can coexist with a weak barrier for months or years, while a well-presented dashboard quietly rewards the appearance of control.

Key Takeaways

• A favorable KPI is not the same as a safer operating condition.
• Frequency, completion, reporting, and closure measures need a second layer of evidence.
• Critical-risk review should test whether a control is available, understood, used, and effective where the work occurs.
• James Reason's work on active and latent failures helps leaders connect visible outcomes with hidden organizational conditions.
• The best monthly review produces a decision, an owner, a changed control, and a verification date.

Why good safety numbers can still hide serious risk

Most dashboards combine lagging measures with activity measures. Incident rates, lost-time cases, training completion, inspections, observations, and corrective-action closure each describe one part of the system. Trouble begins when a single measure is asked to prove more than it can show.

Andreza Araujo makes this distinction central to the leadership conversation in Safety Culture: From Theory to Practice. Culture is visible in repeated choices, especially when leaders must decide whether to slow production, fund a redesign, accept a deviation, or challenge a comfortable result. A dashboard has value only when it helps those choices become more reliable.

In a practical review, every important KPI should answer four questions. What does the number represent? What does it leave out? Which decision should it change? What evidence could prove that the interpretation is wrong? Without those questions, the dashboard becomes a polished record of attention rather than a test of protection.

1. Completion is mistaken for effectiveness

Training completion, inspection completion, and action closure are easy to count because they have clear administrative endpoints. A person attended the course, a form was submitted, or an item was marked complete. None of those events proves that the control works under the conditions that matter.

A completed critical-control inspection may still miss an unavailable isolation point, an ambiguous acceptance limit, or a supervisor who has never practiced the response. The activity occurred, but the protective function remains uncertain. When the dashboard reports only completion, it rewards movement in the system without testing the quality of the movement.

The better measure is completion linked to field evidence. That evidence might show that a barrier was tested, a design was changed, a decision route was used, or a worker could explain the operating boundary. The exact evidence depends on the control, although the principle remains stable: activity must be connected to an observable change.

Executives should ask which completed activity would have prevented the exposure if the task had gone wrong yesterday. If no one can answer, the KPI is measuring administrative progress rather than protection.

2. Averages erase high-consequence exposure

An average can be statistically correct and operationally misleading. A site may report a stable incident rate across departments even though one small task carries a credible fatal exposure and receives little attention because its event frequency is low.

Frequency measures remain useful for understanding patterns in recorded events, but they should not be used as a proxy for the safety of every activity. Serious injuries and fatalities require leaders to examine pathways, energy sources, control dependency, and recovery time, even when the historical count is zero.

James Reason's model of organizational accidents is useful because it directs attention toward latent conditions that can sit behind an event. Weak design, unclear authority, production pressure, and degraded supervision may remain invisible in a monthly average until several defenses align badly.

Replace the question “Did the rate fall?” with “Which high-consequence exposures changed, and how do we know?” A dashboard that cannot answer that question gives precision to the wrong level of risk.

3. Reporting volume is treated as proof of trust

More observations, near misses, and concerns can indicate that people are paying attention and that the reporting route feels usable. They can also indicate repeated exposure, low reporting thresholds, duplicate entries, or a backlog that the organization has learned to tolerate.

The number becomes meaningful only when leaders examine what happens after submission. Does the concern reach the right decision owner? Is the immediate exposure protected? Does the response distinguish a design problem from a competence problem? Does the person who raised the concern receive a clear explanation?

Psychological safety supports early reporting, yet reporting alone is not the outcome leaders need. The outcome is earlier, better action. A high volume with no change in control quality can teach people that speaking up produces records but not protection.

Pair reporting volume with response time, recurrence, action quality, and evidence that the operating decision changed. That combination gives leaders a better reading of whether voice is becoming a control input or remaining a data stream.

4. Thresholds create behavior around the metric

Every threshold sends a message about what matters. When leaders attach recognition, escalation, or budget decisions to one narrow number, teams learn to optimize the number even when that optimization weakens the underlying objective.

A zero-injury target can encourage silence when people believe that a report will damage the site score. A closure target can encourage shallow actions when teams discover that “complete” is rewarded before effectiveness is checked. An inspection target can encourage fast, low-value rounds when the conversation never examines what the inspection found.

Andreza Araujo addresses this danger in Muito Além do Zero, her Portuguese title often translated as Far Beyond Zero. The point is not that leaders should ignore outcomes. It is that a number becomes dangerous when it replaces judgment about exposure, barriers, and the quality of decisions.

Before approving a KPI threshold, ask what behavior the target will encourage when the result is difficult to achieve. Then add a countermeasure that protects the real objective, such as independent verification, transparent escalation, or a review of unreported and unresolved conditions.

5. The dashboard has no decision owner

A metric without an owner can circulate for months while everyone agrees that it matters and no one changes the work. The EHS team may prepare the chart, operations may review it, and senior leaders may comment on it, yet the control gap remains between departments.

Ownership is more precise than visibility. The owner needs authority to alter the condition represented by the KPI, access to the necessary resources, and a defined route for escalation when the operating plan cannot support the required control.

A useful dashboard shows the accountable decision beside the number. If the leading indicator falls, who must investigate? If a critical control is unavailable, who can stop or redesign the work? If an action is closed but recurrence continues, who reopens the decision?

When those answers are missing, add them before adding another metric. A smaller dashboard with visible decision rights is more protective than a large dashboard that distributes concern without responsibility.

6. Closure is counted before field verification

Corrective action is often considered complete when a document is revised, a briefing is delivered, or a purchase order is issued. Those steps may be necessary, although they do not establish that the new condition works in the setting where the exposure occurs.

Field verification should be designed around the failure the action is meant to prevent. If the action changes an isolation method, observe an isolation. If it changes a handover, sample a handover under normal shift pressure. If it changes a rescue arrangement, test the route, equipment, communication, and decision authority.

The verification should also have a time horizon. A control that works during the first week may degrade when staffing changes, contractors arrive, or production demand increases. Evidence of durability is part of effectiveness, not an optional extra.

Leaders should be willing to reopen a closed action when the field evidence disagrees with the administrative record. That response protects credibility because it shows that the organization values the condition of the work more than the appearance of a completed backlog.

How leaders should redesign the monthly KPI review

Start with the serious exposure, not the dashboard page. Select the high-consequence pathways that deserve executive attention, then identify the controls that must hold for each pathway. Use the KPI as a signal that directs inquiry rather than as a verdict that ends it.

For each signal, record the operating decision, the accountable owner, the evidence already available, and the evidence still needed. Include one disconfirming question that could challenge the favorable reading. This small discipline reduces confirmation bias because the meeting is designed to search for weakness as well as improvement.

Close the review with a changed condition and a verification date. The change may involve engineering, staffing, authority, sequencing, procurement, or supervision. It should be specific enough that a person outside the meeting can inspect it later and determine whether the decision survived contact with the work.

That is the standard Andreza Araujo applies across safety leadership and culture work. The purpose of measurement is not to make leadership feel informed. It is to help leaders protect people before the organization needs an incident to reveal what the dashboard failed to show.

FAQ

Why can safety KPIs improve while serious risk remains?

Many safety KPIs count recorded events, completed activities, or closed actions. Those measures can improve while a critical control remains weak, unavailable, or unverified. The dashboard becomes more reassuring without proving that the exposure changed.

What is the biggest problem with relying on TRIR alone?

TRIR is a frequency measure, so it can summarize recordable injuries without showing whether a low-frequency, high-consequence exposure is controlled. It is useful as part of a wider review, but it cannot stand in for critical-control assurance or serious-injury-and-fatality prevention.

How should leaders test a leading safety indicator?

Leaders should ask what decision the indicator triggers, what evidence supports the reported value, who owns the response, and whether a field check could disprove the favorable interpretation. If no decision or verification follows, the measure is probably tracking activity rather than control quality.

Does a high near-miss reporting rate prove a strong safety culture?

No. A high reporting rate may show trust and attention, but it may also reflect repeated exposure, easy reporting of low-value observations, or unresolved backlog. The useful question is whether reports change priorities, controls, and operating decisions.

What should an executive safety dashboard include?

It should connect outcome measures with critical exposures, control health, overdue decisions, worker concerns, and evidence that corrective actions changed the work. Each item needs a decision owner and a clear next verification point.

For more practical guidance on safety leadership, culture, and prevention, explore the work of Headline Podcast and Andreza Araujo's books, including Safety Culture: From Theory to Practice.

Conclusion: a safety KPI earns executive trust only when it points to a decision, survives field verification, and demonstrates that serious exposure changed rather than merely becoming easier to report.