Safety KPIs vs Bonuses vs Control Checks

Safety KPIs are the measures executives use to judge whether risk is being controlled before people are hurt. The problem is that different KPI families create different behavior, especially when the board compares injury rates, bonus targets, reporting quality, and critical-control verification in the same dashboard.

The LATAM food operation accident-ratio case shows the same principle in practice: the outcome number improved only when leaders reviewed follow-up quality, exposure signals, and field verification beside the score.

Most executive safety dashboards still mix three things as if they were equivalent: injury outcomes, incentive targets, and evidence that controls actually worked in the field. That mix is dangerous because each indicator answers a different management question. ILO reports that nearly 3 million workers die every year from work-related accidents and diseases, and that scale is precisely why a board cannot treat a low injury rate as proof of controlled risk.

The thesis of this comparison is direct. Safety KPIs should not be chosen by what looks cleanest in a monthly deck. They should be chosen by the behavior they create, the quality of evidence they produce, and the risk decisions they make visible before a serious event occurs. Across 25+ years leading EHS at multinationals, Andreza Araujo identifies that the most damaging metric is often not the wrong number, but the right-looking number attached to the wrong incentive.

What criteria should executives use to compare safety KPIs?

Executives should compare safety KPIs across 5 criteria: behavioral incentive, proximity to fatal risk, data integrity, decision speed, and field verifiability. A KPI that performs well on all 5 criteria can guide resource allocation, while a KPI that performs poorly on even 2 criteria can make leadership confident for the wrong reason.

The comparison matters because a board safety pack is not a neutral mirror. It teaches managers what will be rewarded, what can be ignored, and what kind of bad news is welcome. ISO 45001:2018 specifies that an OH&S management system should manage risks and improve OH&S performance, which means performance measurement must connect to operational controls, not only to annual injury arithmetic.

For this article, the scoring uses a 1 to 5 scale for each criterion. A score of 5 means the KPI family gives executives timely, field-checkable evidence for decisions. A score of 1 means it may still be useful for compliance or history, although it should not steer bonuses, capital allocation, or risk appetite alone.

Injury-rate KPIs: when do they still belong in the dashboard?

Injury-rate KPIs still belong in the dashboard when they are treated as lagging evidence, not as the main proof of safety control. TRIR, LTIFR, DART, and severity rate can reveal historical harm, regulatory exposure, and trend movement over 12 months, but they cannot show whether tomorrow's high-energy work is controlled.

The value of injury-rate KPIs is comparability. They help a multi-site operation see whether one plant has a materially different injury profile from another, and they support external reporting where regulation requires it. SIF rate, TRIR, and precursor metrics each answer a different executive question, which is why a single-rate dashboard usually creates false precision.

The weakness is statistical and cultural. A 120-employee plant may go 18 months without a recordable case and still carry uncontrolled confined-space, electrical, or line-of-fire exposure. When that same number becomes a bonus gate, the dashboard starts measuring reporting courage as much as injury experience.

As Andreza Araujo argues in Far Beyond Zero, lagging indicators look in the rearview mirror because they show consequences rather than causes. The book's warning is not that lagging data should disappear, but that a rigid zero target can protect the number while leaving the person exposed.

Bonus-tied KPIs: why can they damage reporting truth?

Bonus-tied KPIs can damage reporting truth because they convert safety information into personal economic risk. When a supervisor's payout depends on 0 recordables, 0 lost-time cases, or a green monthly score, the system quietly asks people to choose between transparency and reward.

This is the most common executive trap. Leaders believe they are showing seriousness by tying compensation to safety, although the design often rewards silence. The more severe the financial consequence, the stronger the temptation to downgrade cases, delay classification, challenge medical treatment, or pressure teams to avoid reporting borderline events.

That does not mean safety should be absent from compensation. It means the compensated element should favor verified management behavior rather than the absence of bad news. A bonus gate based on action closure quality, field verification, and response time to serious precursors is harder to game than a gate based only on injury absence.

5 criteria should be tested before a safety KPI enters compensation: whether it encourages reporting, whether field evidence can verify it, whether managers can influence it ethically, whether it relates to critical risk, and whether it remains useful when incidents increase because reporting improves.

Reporting-quality KPIs: what do they reveal that injury rates miss?

Reporting-quality KPIs reveal whether the organization is willing and able to surface weak signals before harm occurs. A useful reporting metric does not celebrate volume alone; it measures specificity, timeliness, classification quality, response quality, and whether workers receive a response within a defined cycle such as 7 or 14 days.

Near-miss count alone is a poor proxy for learning. One site may report 200 vague observations because the campaign is active, while another reports 35 high-energy precursors that deserve urgent executive attention. The better KPI is quality-weighted reporting, because it distinguishes noise from risk intelligence.

This is where psychological safety enters the metrics discussion without becoming a slogan. When people believe that bad news will be punished or ignored, the dashboard loses its early-warning function. Near-miss quality matters because volume can look like learning while the real exposures remain unnamed.

In more than 250 cultural transformation projects, Andreza Araujo observes that an increase in reports can be a sign of maturity when leadership closes the loop. The board should ask 3 questions every month: what was reported, what changed because of it, and which critical risk remained open after the response.

Critical-control verification: why is it closest to fatal-risk prevention?

Critical-control verification is closest to fatal-risk prevention because it checks whether the controls that prevent fatal or life-altering events are present, effective, and used at the moment of exposure. Unlike injury rates, it measures the barrier before the accident; unlike reporting volume, it tests whether risk control exists in the field.

HSE explains in HSG254 that leading indicators can be too far removed from the critical control measure, which creates a false sense of progress. That warning is practical. Training completion may be 98 percent, although the gas test for confined-space entry was not calibrated, the rescue equipment was incomplete, or the isolation point was wrong.

Control verification also improves executive decision speed. If 12 percent of high-energy work observations show missing or ineffective controls, the board does not need to wait for an injury-rate trend. It can ask for a shutdown rule, capital correction, competency review, or contractor intervention while the risk is still recoverable.

The weakness is that critical-control verification demands operational discipline. It requires a defined list of critical controls, trained verifiers, sampling rules, escalation thresholds, and a way to separate paperwork confirmation from field evidence. Critical control registers help because they make the control owner, failure mode, and verification routine visible.

Decision matrix: which KPI family wins for each context?

The best KPI family depends on the executive decision being made. Injury-rate KPIs are strongest for historical trend and external reporting, reporting-quality KPIs are strongest for cultural truth and weak-signal flow, while critical-control verification is strongest when the decision concerns fatal-risk exposure and immediate resource allocation.

The matrix does not make injury-rate KPIs obsolete. It puts them in their proper role. A board that needs assurance over high-energy work should not rely on a number that improves after nothing happened, especially when the exposure could have been present every day for 30 shifts.

How should a C-level team choose the right mix?

A C-level team should choose a 4-part mix: 1 lagging outcome indicator, 1 reporting-quality indicator, 1 critical-control verification indicator, and 1 leadership response indicator. This mix keeps history, truth, control, and accountability visible without turning the dashboard into a 40-metric archive.

A practical monthly pack can stay compact. Use TRIR or LTIFR for trend, quality-weighted near misses for weak-signal flow, critical-control verification for fatal-risk exposure, and corrective-action effectiveness for management response. A safety metric dictionary prevents drift because every KPI has an owner, formula, inclusion rule, exclusion rule, and escalation threshold.

4 KPI families are enough for most executive safety packs when the board uses them to ask better questions. More metrics often create less accountability because every weak signal can hide behind another green column.

During the PepsiCo South America tenure, where the accident ratio fell 50 percent in six months, Andreza Araujo learned that executives move culture when they ask operationally precise questions. The question is not only whether the rate improved. The better question is which exposure was removed, which control failed verification, and which leader changed a routine because the data made the risk visible.

What traps make safety KPIs look better than risk reality?

Safety KPIs look better than risk reality when the dashboard rewards silence, averages unlike risks, or counts activity without testing control quality. These 3 traps are common in mature companies because they appear sophisticated in reporting while still failing to challenge how work is executed.

• Trap 1: green averages. A regional dashboard can be green while one plant carries repeated high-energy control failures.
• Trap 2: activity inflation. Leaders count inspections, walks, and observations without checking whether they changed a risk condition.
• Trap 3: bonus pressure. A target tied to 0 events can turn a reportable injury, a near miss, or a technical concern into a negotiation.

These traps explain why zero-accident targets deserve executive scrutiny. A clean number can reflect genuine improvement, but it can also reflect fear, classification pressure, or the random absence of harm during a short period.

Each month without redesigning incentive-sensitive safety KPIs allows underreporting habits to harden, while high-energy exposures keep accumulating outside the executive conversation.

Which option should the board choose first?

The board should choose critical-control verification first when fatal-risk exposure is the priority, then pair it with reporting-quality KPIs and a limited injury-rate trend. This sequence gives executives evidence before harm, culture signals before silence, and historical outcomes without letting the past dominate the decision.

For an industrial company with high-energy work, the first 90 days should not begin with a new bonus formula. It should begin with identifying the top 5 critical risks, defining the controls that must never fail, and checking those controls in the field with named owners. Once the evidence base is stable, compensation can include verified behaviors such as timely closure of critical findings and leader response to serious precursors.

OSHA states that leading indicators are proactive and preventive measures that can reveal potential problems in a safety and health program, while lagging indicators measure events that already occurred. The board-level implication is clear. Lagging data confirms part of the story, but preventive governance depends on signals that arrive before injury.

As Andreza Araujo writes in Diagnosing Safety Culture, good indicators do not guarantee good practices. A safety KPI system becomes useful only when it changes what leaders inspect, fund, stop, and reinforce during real work.