Safety Indicators: 5 Myths That Hide Weak Controls

Safety indicators can improve while the exposure behind them stays unchanged. This article tests five common beliefs about safety metrics and gives plant managers a 30-day way to separate activity from control.

The problem is not measurement itself. The problem appears when a green dashboard is treated as evidence that risk has fallen, even though the metric only records what the organization chose to count. A useful indicator should change a decision, reveal a weak barrier, or trigger a timely correction.

Key figure: ISO 45001:2018 gives organizations a management-system structure, but a documented metric is not proof that a critical control worked in the field. A 30-day review should compare the dashboard with field evidence.

1. Myth one is that more safety data automatically means better control

More data improves safety only when the extra information answers a decision that leaders must make. A dashboard with 40 measures can still leave the organization blind if none of them shows whether a serious-risk barrier was available, understood, and used under pressure.

Activity counts are easy to collect because they fit existing systems. The number of observations, meetings, inspections, training hours, and closed actions can be reported every week, while the condition those activities were meant to change remains untested. This is why the first question for any metric is not whether it is leading or lagging, but what action it should cause.

Andreza Araujo makes this distinction central to the difference between compliance and culture. In The Illusion of Compliance, her Portuguese title A Ilusão da Conformidade, the practical test is whether the requirement changes what people can do and what leaders decide. A metric that never changes a decision is reporting effort, not controlling exposure.

Use a three-part test. Identify the risk pathway, name the decision owner, and state the evidence that would change the next action. If those elements are missing, the measure may still be useful for administration, but it should not carry the status of a control indicator.

2. Myth two is that a high completion rate proves a healthy safety system

A completion rate proves that an assigned activity was recorded, not that the intended control held. One hundred percent inspection completion can coexist with repeated critical defects when the inspection is rushed, the finding is downgraded, or the person who can fund the correction is not part of the review.

OSHA's leading-indicators guidance recommends measures that show preventive activity and progress before an injury occurs. That principle does not make completion percentages useless. It means leaders should pair them with evidence that explains whether the activity affected the hazard, the barrier, or the decision condition.

These percentages are examples of dashboard signals, not proof of performance. The metric becomes more credible when the record shows the exception, the owner, the due date, and the field evidence that confirms the risk condition changed.

3. Myth three is that zero recordables means zero serious-risk exposure

A period without a recordable injury can be encouraging, but it cannot establish that serious-risk pathways are controlled. A low injury count may reflect effective prevention, limited exposure, underreporting, or simply a short observation window in which a high-consequence event did not occur.

The distinction matters because recordable outcomes are lagging evidence. They describe harm that reached a person, while a critical-control review asks whether the organization is prepared to prevent or contain a credible high-consequence event before the outcome occurs. The two views belong together, although they answer different questions.

Andreza's book Far Beyond Zero, translated from Muito Além do Zero, challenges the use of zero as a target when people learn that bad news threatens their standing. A stronger dashboard shows whether concerns are reported, escalated, investigated, and converted into better controls, because silence can make a green number less trustworthy.

Review the last 90 days through two lenses. First, examine injury and illness outcomes using the definitions required by the applicable regulator. Second, select the serious-risk scenarios that could produce a fatality or life-changing injury and verify the barriers that should interrupt them. The second review is where a zero-event period becomes informative rather than reassuring.

A green outcome metric should never close a serious-risk review when the barrier evidence is missing or the exposure has changed.

4. Myth four is that every leading indicator should be converted into a target

Targets help when they define a minimum condition that must be protected, but they distort behavior when people can improve the number without improving the work. A target of 20 observations per supervisor may increase observations while reducing the quality of the conversation, the seriousness of the issue, or the chance that a worker will speak honestly.

Leading indicators are better treated as decision thresholds. For example, a site can require that a critical-control verification identifies the barrier, the observed condition, the person who owns the correction, and the escalation route if the gap cannot be closed before work begins. The requirement protects the quality of evidence instead of rewarding volume alone.

In more than 25 years of multinational EHS leadership, Andreza Araujo has built her authority around the connection between culture and operating conditions. Her experience includes a 50% reduction in accident ratio in six months at PepsiCo South America, yet the lesson is not to copy a percentage. The lesson is to connect measurement with a management decision that changes exposure.

When a metric becomes a target, ask whether a person can reach it while leaving the hazard unchanged. If the answer is yes, keep the metric as descriptive information and add a second measure that tests the condition leaders actually need to protect.

5. Myth five is that one safety score can represent the whole system

A single safety score simplifies communication, but it also compresses different kinds of evidence into a number that can hide disagreement. A site may score well on training, audits, and reporting while its permit quality, maintenance backlog, contractor interface, or emergency readiness remains weak.

The score becomes more useful when leaders can open it and see at least three layers. The first layer describes work conditions and exposure. The second shows decisions, ownership, and escalation. The third records outcomes, including injuries, illnesses, and confirmed control failures. A score that cannot be traced to these layers should not be used to rank sites or reward managers.

The ISO 45001:2018 standard frames monitoring, measurement, analysis, and performance evaluation as part of the management system, which supports a balanced reading rather than a single vanity number. The Bureau of Labor Statistics records occupational injury and illness data, yet those records are only one part of a prevention picture that also needs operational evidence.

Use the score as an index page, not as the conclusion. The leadership conversation should move quickly from the number to the weakest barrier, the accountable owner, and the date when evidence will be checked again.

6. What to measure when the dashboard looks green

A green dashboard needs a challenge set that tests whether the number is telling the truth. Start with four questions that can be answered from records and field evidence within 30 days.

• Which serious-risk pathway is most exposed right now, and what evidence supports that judgment?
• Which critical barrier failed, drifted, or became unavailable during the last 30 days?
• Who had authority to correct the condition, and how quickly did escalation reach that person?
• What changed in the work after the corrective action was marked complete?

Then compare the answers with the dashboard. If the score says healthy while the field review shows missing barriers, retain the disagreement instead of averaging it away. The disagreement is a management signal because it identifies where the measurement system is weaker than the operating reality.

James Reason's work on latent conditions helps explain why this review must look beyond the final event. The visible failure may sit in a task, while the conditions that made it likely were created by planning, maintenance, staffing, design, supervision, or competing priorities that the dashboard never displayed.

7. What leaders should change before the next monthly review

Leaders should retire any safety metric that cannot explain what decision it supports. Replace it with a small set of measures whose evidence can be checked by the person who owns the risk, including barrier status, exposure changes, escalation time, and the quality of corrective-action verification.

Set a review rhythm with three moments. During the first week, confirm the serious-risk scenarios and owners. During the second week, sample field evidence and test whether the barrier works under normal variation. During the third week, challenge unresolved gaps with the business leader who controls resources, sequencing, or design. Reserve the fourth week for decisions, not another round of data collection.

Andreza Araujo's Safety Culture: From Theory to Practice is useful here because it treats culture as something visible in decisions, habits, and conditions rather than as a slogan. That approach keeps metrics close to the work and gives leaders a defensible way to act when the number and the field evidence disagree.

Safety indicators are valuable when they expose a decision, not when they decorate a report. A dashboard becomes trustworthy when it can show what changed, who changed it, and what evidence proves that the protection still works.

To deepen this practice, explore Andreza Araujo's books and Safety Culture resources, or use the safety ROI review to test whether prevention changed risk rather than merely increasing activity.