Leading Safety Indicators: 6 Signals That Reward Activity Instead of Control

A safety dashboard can look active while a dangerous exposure remains untouched. Training percentages rise, observations are submitted, corrective actions close, and leaders attend more field visits. The organization feels disciplined because the indicators are moving in the expected direction.

That appearance becomes dangerous when activity is mistaken for control. A leading safety indicator should reveal a condition that changes exposure before harm occurs. If it only proves that a form was completed, a meeting happened, or a record was entered, the measure may reward administration while critical barriers weaken in the field.

Andreza Araujo’s work on safety culture treats this distinction as a leadership responsibility. Her experience includes more than 25 years in global EHS and a 50% accident reduction during her PepsiCo South America tenure over six months. The transferable lesson is not to collect more numbers. It is to connect leadership attention, operating discipline, and field decisions to evidence that can be tested.

Why leading indicators can create false confidence

Leading indicators are often introduced to correct the limits of lagging metrics such as recordable injuries, lost-time cases, or total recordable incident rate. That move is sensible because a serious event is a late signal. By the time the outcome appears, the organization has already lost the opportunity to prevent it.

The problem begins when the replacement measure is easier to count than the condition it is meant to represent. A team can complete a training module without being able to isolate stored energy. A supervisor can perform an observation without correcting a recurring design weakness. An action can be closed in the software while the worker still faces the same exposure on the next shift.

James Reason’s Swiss Cheese Model helps explain the gap. Accidents emerge when weaknesses in several defenses align, so a dashboard should examine whether barriers are present and dependable, not simply whether an administrative event occurred. The most important question is therefore not “Did we do the activity?” but “What changed in the path to harm?”

The six signals below are not arguments for abandoning measurement. They are tests for deciding whether a familiar measure deserves to remain on an executive dashboard.

Signal 1: activity volume rises without exposure context

Observation counts are attractive because they create a visible stream of data. A site can set a monthly target, compare departments, and celebrate a higher submission rate. Yet volume alone says little about whether the observations represent meaningful exposure.

One hundred observations about housekeeping can coexist with no credible review of lifting, isolation, confined-space entry, or vehicle interaction. The total looks healthy because the dashboard treats every observation as equivalent, although the potential consequence and control strength differ sharply.

A better measure identifies the task, exposure, and control that the observation tested. It records whether the worker could explain the control, whether the supervisor had authority to change the condition, and whether the response was verified after the conversation. The count then becomes a sampling input rather than a performance trophy.

When leaders review the number, they should ask which serious exposures it represents and which ones remain invisible. A smaller sample that reaches high-consequence work can be more useful than a large volume of low-risk observations.

Signal 2: training completion stands in for competence

Training completion is necessary for many tasks, but completion proves attendance more easily than competence. A worker may pass through an online module and still be unable to recognize a missing isolation point, select the correct control, or stop the task when the plan no longer matches the field.

This is not a criticism of training records. It is a warning about the decision attached to the percentage. If leaders use completion to authorize work, the measure is being asked to do more than it can support. The dashboard should distinguish exposure-critical learning from general awareness and should show how competence was checked where failure could be severe.

The verification method should fit the task. A practical demonstration may be appropriate for lockout, a field questioning exercise may fit a permit review, and a supervisor observation may test whether a control remains usable during a normal production cycle. The record matters, but the evidence must reach the point of work.

The article on assumptions hidden by safety training completion develops this distinction further. A percentage is a starting point for inquiry, not a conclusion about readiness.

Signal 3: observation volume rises without control change

Behavioral observations can support learning when they identify a decision, a condition, and a conversation that improves the work. They become decorative when the system rewards the number of forms rather than the quality of what follows.

Suppose a supervisor records repeated manual-handling deviations in the same area for three months. If the dashboard celebrates the observation total but does not show redesign, equipment adjustment, workload review, or task-level verification, it is measuring recognition of the problem without measuring the organization’s response to it.

Frank Bird’s accident-ratio work is often used to support attention to precursor events, but precursor data has value only when it informs prevention. A near miss, an unsafe condition, or a behavioral observation should lead to a question about the barrier that allowed the exposure. It should not become a quota that encourages low-value reporting.

Track the proportion of high-consequence observations that produced a control decision, then verify the decision in the field. This keeps the conversation close to risk and makes it harder to substitute activity for prevention.

Signal 4: corrective actions close faster than controls improve

Fast closure is often presented as evidence of responsiveness. It can be useful when the action is simple, the owner has authority, and the evidence shows that the hazard was removed or reduced. It becomes misleading when closure is defined as uploading a document, sending an email, or completing a briefing.

A corrective action should change a condition, decision right, barrier, or verification practice that contributed to the exposure. If the action closes before workers can use the new control, the system has reduced backlog rather than risk. The dashboard then creates pressure to make records disappear instead of making hazards disappear.

Measure closure together with effectiveness. The owner should show what changed, who is exposed, how the control is used under ordinary pressure, and when it will be checked again. A reopened action is not automatically a failure. It can be evidence that verification found a weakness before the weakness became an incident.

Leaders who want a sharper test can compare closure time with the time required for field verification. When those two timelines are identical for every action, the organization may be closing work before it has actually checked the result.

Signal 5: reporting volume is detached from response quality

More reports can mean that people trust the reporting route, or it can mean that the system has created a low-value reporting habit. The number cannot answer that question by itself.

Response quality is visible in what happens after a concern is raised. Was the issue acknowledged? Did the owner have the authority to act? Did the person who raised it receive a safe explanation? Did the organization identify whether the same condition existed elsewhere? If the answer is unclear, a high reporting count may conceal frustration rather than demonstrate culture.

Psychological safety matters here because a worker must be able to question a plan without fearing punishment or humiliation. Amy Edmondson’s concept becomes operational when a concern can move through the system, receive a proportionate response, and return to the workforce as evidence that speaking up changes decisions.

Pair reporting volume with response latency, quality of closure evidence, repeat concerns, and worker confirmation. The goal is not to maximize reports. It is to make credible information travel far enough to change exposure.

Signal 6: leadership presence is measured by attendance

Field visits are valuable when leaders use them to see work, ask precise questions, and remove barriers that local supervisors cannot remove. Attendance alone does not show visible felt leadership.

A leader can attend a walk, inspect a checklist, and leave before hearing disagreement about schedule, equipment, staffing, or a permit condition. The visit then becomes a performance of concern rather than an opportunity to improve the decision environment.

Measure what leadership presence makes possible. Record the critical control discussed, the unresolved obstacle escalated, the decision owner assigned, and the date when the field will be revisited. A leader’s presence should produce a traceable decision or a verified reason why no change was required.

That trace also protects leaders from confusing visibility with influence. When the same exposure appears on multiple visits without a decision, the problem is not a lack of attendance. It is a lack of authority, prioritization, or follow-through.

How to redesign the dashboard around control evidence

Start with the exposures that could produce serious injury or fatality, then name the controls that prevent or limit those outcomes. Do not begin with the measures already available in the software. Begin with the barriers the operation cannot afford to lose.

For each control, define an owner, a usable field test, and a response when the test fails. The indicator should show whether the control exists, whether workers can use it, and whether verification found it effective after conditions changed. If the measure does not support one of those decisions, it may belong in an operational record rather than an executive dashboard.

The dashboard should also show where the evidence is weak. A missing verification is not a reason to hide the measure. It is a management signal that the control cannot yet be trusted. This approach aligns with the critique of zero-accident targets in the three failures that turn safety metrics into underreporting, because a healthy number is not the same as a healthy system.

What leaders should ask before approving a leading indicator

Before adding a measure to a monthly review, ask whether it can change a decision before harm occurs. If the answer is no, the measure may still have administrative value, but it should not be presented as evidence of prevention.

• Which exposure does this indicator represent?
• Which critical control should improve when the number changes?
• What field evidence confirms that the control is usable?
• Who owns the response when the indicator deteriorates?
• How will the organization detect a good number with a weak control?
• When will the result be checked after a change in people, equipment, workload, or schedule?

A safety metric earns trust when it helps a leader see a decision that would otherwise remain hidden. If it only creates a reassuring score, it has become part of the problem.

For leaders building a more disciplined review rhythm, the guide on building a monthly safety decision pack for the C-suite provides a practical companion. The central principle remains simple without being simplistic. Measure the condition of the barrier, then verify that the barrier still works when the operation is under pressure.

Leading indicators should expose control weakness, not decorate the dashboard

Leading safety indicators are valuable when they provide early evidence about exposure, control quality, and decision effectiveness. They become dangerous when they reward activity without testing whether the work is safer.

The six signals in this article point to a common failure. Organizations count what is easy to record, then infer what is difficult to verify. Stronger leadership reverses that habit by starting with serious exposures, defining the controls that matter, and asking for evidence from the field.

That is how a dashboard moves from reporting activity to governing risk. The number still matters, but the decision behind the number matters more.