Leading Safety Indicators: 5 Distortions That Make Activity Look Like Control

A dashboard can show more safety activity while the work becomes less safe. The contradiction appears when leaders count what is easy to submit, close, and display, although the measure says little about whether a serious exposure is controlled at the moment people face it.

That is why leading safety indicators require a stricter test than “does this move before the lagging rate?” A useful indicator must reveal a condition early enough to matter, identify who can respond, and show whether the response changed the work. Without those links, activity becomes a substitute for control.

Andreza Araújo’s critique of compliance theater in The Illusion of Compliance is relevant here. A completed form can demonstrate that a process exists, but it cannot prove that the process protected a person during a changing task. The five distortions below help operations and executive leaders separate evidence from reassurance.

1. Activity counts replace evidence of control

The first distortion is the assumption that more activity means better prevention. A site reports thousands of safety observations, toolbox talks, inspections, or training completions, and the dashboard treats the total as proof that the system is active. The number may be accurate while the conclusion is wrong.

Activity counts describe effort. They do not automatically describe risk, control quality, or worker protection. A monthly increase in observations may reflect a campaign, a new reporting requirement, or a supervisor who is entering low-value items to meet a target. Unless the measure distinguishes a meaningful control signal from routine volume, the organization rewards submission rather than attention.

The practical test is to follow a sample from entry to decision. Did the observation identify an exposure that could produce serious harm? Did a named owner respond within the relevant time window? Did the task, equipment, staffing, or sequence change? If the trail ends at “closed,” the indicator has measured administration, not prevention.

This does not make activity irrelevant. Conversations, inspections, and observations are necessary inputs. They become leading indicators only when their quality and consequence are tested. The dashboard should therefore display the proportion of high-value signals that produced a verified change, rather than celebrating volume on its own.

2. Closure rates hide unresolved exposure

Closure is attractive because it creates a clean visual story. The backlog falls, the red cells turn green, and the meeting moves on. Yet an action can be closed while the hazard remains because the response addressed the wording of the finding instead of the condition that created it.

Consider a recurring issue with vehicle and pedestrian interaction. Repainting a walkway may close the action, but it does not establish that traffic routes, visibility, delivery timing, and driver behavior now prevent the conflict. The indicator is not the percentage of actions closed. It is whether the exposure was reduced and whether the control remains reliable under the conditions that made the problem recur.

James Reason’s analysis of latent failures gives leaders a useful lens. Repeated unsafe conditions often reflect decisions about design, staffing, maintenance, supervision, or production planning. If an action closes at the frontline while those conditions remain untouched, the organization has removed a line from the register without removing the failure path.

Replace a single closure metric with three linked questions. What was corrected? Who verified the correction at the point of work? What will reopen the action if the condition returns? That sequence makes closure evidence-based and gives the executive team visibility into recurring exposure, not only administrative completion.

3. Training completion is mistaken for changed capability

Training completion is one of the most common leading measures because attendance is easy to record. The problem starts when an organization interprets attendance as competence, and competence as reliable execution under pressure. Those are different conditions.

A worker can pass a course and still lack the opportunity, equipment, supervision, or confidence required to apply the method. A supervisor can understand a critical control and still accept a shortcut when the schedule collapses. A completion rate cannot see those gaps unless it is connected to observation of the task and to the decisions made when conditions change.

The stronger measure follows the learning into work. It asks whether the person can recognize the exposure, explain the control, use it at the correct point, and escalate when the control is unavailable. It also asks whether the work system makes the expected behavior possible. If the answer depends on improvisation, another course may be the least useful response.

For leaders, the implication is direct. Keep completion as a readiness measure, but do not place it beside a claim that risk has fallen. Pair it with verified task capability, supervisor response, and evidence from abnormal or non-routine work, where training is most likely to be tested.

4. Reporting volume is treated as psychological safety

Near-miss and concern reports can provide early information, but a higher reporting volume has no single meaning. It may indicate stronger trust, more exposure, a campaign, easier digital access, or a workforce that has learned to report low-consequence items because serious concerns disappear into the system.

Leaders often celebrate the count without examining what happens after a person speaks. That misses the cultural mechanism that gives reporting its value. A report is a leading signal only when the organization responds visibly, protects the reporter from retaliation, and gives the workforce evidence that raising a concern can alter a decision.

A useful dashboard separates report volume from response quality. It tracks time to acknowledgement, time to risk decision, escalation when the first response is insufficient, and recurrence after closure. It also examines who is reporting and who is silent, because a quiet department may have strong control or may have learned that speaking up costs more than staying quiet.

Amy Edmondson’s work on psychological safety helps clarify the distinction. Permission to submit a form is not the same as confidence that bad news can travel upward. The indicator should therefore test the route from voice to action. When that route is weak, publishing a higher reporting number can disguise a deeper failure.

5. Green averages erase critical variation

An average can make an unstable system look controlled. A site reports strong compliance across a month, a region reports a low injury rate, or a contractor score remains above the target. The aggregated result may conceal night-shift gaps, temporary work, new supervisors, seasonal demand, or a small number of high-energy tasks that carry most of the serious-risk exposure.

Risk is not distributed evenly just because the dashboard is. The executive question should be where the indicator is weakest, when it deteriorates, and which work is most exposed when the measure turns green overall. A weighted average can still be misleading if the weighting gives routine tasks more influence than infrequent but severe exposures.

This is the point at which a metric must connect to the risk profile. A site with no recordable injury may still have weak isolation, rescue, lifting, or mobile-equipment controls. The relevant indicator is the verified condition of those barriers during the exposure window, not only the outcome count after the month closes.

Segment the data by task, shift, contractor, location, and abnormal condition when those dimensions change exposure. Then make the segmentation actionable. If a night shift has weaker verification, the response may involve staffing, supervision, handover, or work timing. Showing the pattern without changing the decision only produces a more detailed report.

6. A better indicator changes the meeting

The practical value of a leading indicator appears in the meeting where someone must decide what happens next. If the number is presented, praised, and archived without changing the work plan, resource allocation, or escalation route, it has not led anything.

Executives can test this by asking four questions for every metric. What exposure does it represent? What threshold requires action? Who has authority to act? What changed after the last signal? A metric that cannot answer the fourth question is still an activity report, even if its formula is sophisticated.

The safety dashboard blindness problem is not solved by adding more measures. It is solved when the dashboard becomes a decision interface. The discussion moves from “Are we on target?” to “Which control is weakening, where is it weakening, and what will we change before the exposure becomes normal?”

That shift also improves accountability. The owner is no longer the person who updates a spreadsheet. The owner is the leader who can change the condition represented by the metric and explain why the response was proportionate to the risk.

7. The executive dashboard needs a decision trail

A monthly dashboard should preserve enough context for a leader to see whether the system is learning or merely reporting. That does not require dozens of charts. It requires a small set of measures with a clear relationship to critical exposure, control assurance, and decision rights.

For each indicator, record the signal, the affected exposure, the accountable owner, the threshold, the response, and the verification date. A short note about what changed is more valuable than another decimal place. When the measure deteriorates but no decision follows, that gap should be visible at the next review.

Andreza’s book Far Beyond Zero, the English gloss of her Portuguese title Muito Além do Zero, challenges leaders to look beyond the comfort of a zero count. The same principle applies to leading indicators. A green number is not an achievement if it prevents the organization from seeing what still requires attention.

The dashboard should also show unresolved disagreement. If operations, maintenance, and EHS interpret the same control differently, the conflict is useful information. It identifies a decision that has not been made clearly enough for the field, which is often more important than another completed inspection.

8. The final test is whether exposure changes

Five distortions can be reduced to one final question. After the indicator moved, did the exposure change in a way that can be verified? If the answer is unknown, the organization should not call the measure predictive. It can call it an input, a process measure, or a management activity, but the stronger claim is not supported.

This standard is demanding because prevention is not a single event. Conditions change, controls degrade, and decisions move between teams. A leading indicator must remain connected to the work as it is actually performed, including the moments when a plan no longer matches the field.

Leaders do not need a perfect metric before acting. They need an honest one that exposes uncertainty early, assigns a response, and makes the result visible. That is how a safety dashboard supports care rather than compliance theater.

The most useful leading indicator is therefore not the one that produces the smoothest trend. It is the one that makes a difficult decision arrive early enough to protect people.

What leaders should change this month

Start with the five measures that occupy the most space in the executive review. For each one, identify the exposure it is supposed to reveal, the decision it is supposed to trigger, and the evidence that would prove the decision worked. Remove or relabel measures that cannot meet that test.

Then review one recent green result with the people who perform the work. Ask what the number missed, which conditions were atypical, and what would have caused the metric to turn red sooner. That conversation often finds more useful information than a new scorecard.

Finally, keep the indicator set small enough that leaders can remember the response rules. Measurement is part of safety leadership only when it improves the next decision, not when it makes the last meeting look organized.

Frequently asked questions

Can a lagging indicator also support prevention?

Yes. A lagging measure can reveal a pattern that prompts preventive action, although it arrives after the event or outcome. The distinction is about timing and intended use, not whether a measure is inherently good or bad.

Should every safety activity become a KPI?

No. An activity belongs on the dashboard only when its quality, relevance, and response can be assessed. A long list of activities can hide the few conditions that require leadership attention.

What is the best first step for a weak indicator system?

Choose one critical exposure and trace the current measure from field signal to leadership decision. That small review usually reveals whether the problem is poor data, unclear ownership, slow escalation, or a measure that was never connected to control in the first place.