Safety KPIs That Tell Leaders What They Need to Know

Begin with the decision

A safety measure earns its place when it supports a decision. Before adding a metric, leaders should ask what action will change when the result moves. A count of training hours may show activity, but it does not show whether people can apply the learning. A percentage of inspections completed may show discipline, but it does not show whether inspectors found weak controls or whether operations corrected them. Good dashboards start with the organization’s material risks and the decisions needed to keep those risks controlled.

Keep outcome measures

Fatalities, serious injuries, recordable cases, occupational illness and lost time are essential facts. They show harm that occurred and must never be hidden by a preference for positive indicators. Their limitation is timing. They arrive after exposure and may remain at zero for long periods even where risk is poorly controlled. Small workforces also produce volatile rates. A single case can move the rate sharply, while a year without injury can reflect good management, low exposure or simple luck.

Measure control health

For each material risk, identify the controls whose failure could lead to serious harm. Measure whether those controls are available, meet their performance requirements and are independently verified. Useful examples include the percentage of high-risk isolations sampled without error, overdue safety-critical maintenance, time spent operating with an impaired barrier, or the quality of confined-space rescue drills. These measures sit closer to the causal pathway than general activity counts and give managers something specific to correct.

Measure exposure and capacity

Control data needs context. Ten verification failures among ten checks is different from ten failures among ten thousand. Track the amount of high-risk work, such as lifting hours, energized interventions, vehicle kilometres or entries into confined spaces. Pair this with capacity measures: supervisor span, competence coverage, maintenance backlog and access to specialist support. A rising workload with falling control capacity is an early warning even before an incident or verification failure appears.

Listen to the workforce

Quantitative metrics cannot explain every weak signal. Add structured information from crews, contractors and safety representatives. Track recurring constraints, unresolved procedure mismatches and occasions when work stopped because conditions were unsafe. Do not turn speaking up into a quota. A sudden fall in reports may indicate improvement, but it may also indicate fear or fatigue. Leaders should test the meaning through field conversations and compare what people say with operational evidence.

Avoid composite-score theatre

A single red, amber or green score is attractive because it simplifies reporting. It can also conceal trade-offs. Strong training completion can mathematically offset a failed critical control, although the two are not equivalent. If a composite is used, serious control failures should never be averaged away. Display the underlying measures, define thresholds in advance and explain data quality. Trends, exceptions and time at risk usually tell a more useful story than a polished overall score.

Build a board-ready set

A practical board view can fit on one page: actual serious harm; high-potential events; exposure to the top material risks; critical-control failures and restoration time; overdue safety-critical maintenance; workforce concerns; regulatory actions; and the quality of completed investigations. Each item needs an owner and an agreed response threshold. The board should spend less time admiring charts and more time asking what changed, where uncertainty remains and which operational decisions follow.

Build measures around a decision

Start with one decision the board or site manager must make. Choose the smallest set of measures that shows outcomes, exposure, control health and capacity relevant to that decision. Define the owner, threshold and response for each measure. Remove indicators that no one uses or that reward reporting behavior unrelated to risk reduction.

Questions for each indicator

What decision does this indicator support? Can a favorable number hide degraded critical controls? Does the measure create pressure to underreport? Who acts when the threshold is crossed? What evidence will show that the response reduced exposure rather than improved the dashboard?

Remove measures that do not change action

A useful KPI changes a decision. If no one can state the decision or response, the measure is probably noise.

Sources & Further Reading

Author/editor: Myaser HSE Hub Editorial

Last reviewed: August 4, 2026

Disclaimer: This article provides general HSE education. Apply applicable law, standards, engineering judgment and competent professional advice to your specific operation.