SHARE Knowledge Hub

Health and Safety Key Performance Indicators

A key performance indicator (KPI) is a selected measure that helps people judge whether important health and safety controls work and what decision to make next. A balanced set combines evidence of critical-control health, worker reports, investigations and actual harm; counts or percentages alone can hide failure. Define each measure, validate its data, set an owner and response threshold, and check whether action improved the work.

Updated 2026-10-19Great Britain and Northern Ireland

Understanding Health and Safety Key Performance Indicators

A health and safety KPI is a deliberately defined measure used to help a named decision-maker determine whether risk is being controlled, whether performance is improving or deteriorating, and what intervention is justified. It is not simply any number that can be collected. A count may be useful operational information without being a KPI; a KPI earns its place only when it leads to a repeatable question and a consequential decision.

Why this matters at work

Link the measure to an actual exposure-control objective. For example, track verified performance of relevant extraction and action on exposure-control failures alongside the completion and follow-up of the appropriate health-surveillance scheme. Surveillance attendance alone does not establish prevention, and reported disease may reflect latency or historical exposure rather than the current month alone.

Event classification should reflect both actual and credible potential consequence. A near miss involving a pedestrian and moving vehicle may have no injury but reveal a failed control with fatal potential. Recording credible potential requires trained, consistent judgement and a review process; it should not be left to a reporting person's self-assessment. Equally, potential severity must not be inflated merely to obtain attention. The evidence and decision rationale should be retained.

How organisations manage this in practice

In the Do stage, management defines controls in practical terms. Examples include a permit-to-work process, a lifting plan review, local exhaust ventilation examination, competence authorisation, guard inspection, fatigue control, contractor prequalification and return-to-work support. The performance question is not whether a policy exists; it is whether the control is available, understood, used, maintained and capable of its intended result.

A stronger design uses several linked measures: critical-control assurance for planned LEV examinations, maintenance and task-use observations; quality measure for whether hoods are positioned and maintained effectively in sampled work; exposure monitoring completion and results against the agreed occupational hygiene plan; and lagging learning from failed capture checks, worker respiratory concerns, sampling excursions and relevant group-level health themes. Every elevated result should trigger review of task, control, sampling method and remedial action rather than being averaged into a quarterly score.

A weekly operational review can examine new events, urgent exceptions, data completeness, overdue actions, work changes and control checks. A monthly site review can test trends, assurance quality, contractor performance, workforce themes and escalation effectiveness. A quarterly leadership review can consider material-risk status, cross-site patterns, resource decisions, independent audit, strategic change and whether KPIs still fit the risk profile. An annual review should refresh the control map, definitions, thresholds and source base.

What makes a measure key?

A measure is key when it relates to a material risk or objective, has a dependable definition, and leads to a decision.

Start with the credible harm and the control that prevents it. Choose a signal near that control, decide what a concerning result looks like, and name the person who can change work. One number rarely proves control: sample records, observe the task and ask workers whether the reported result matches reality. Sources: HSE HSG65: Managing for health and safety, HSE HSG254: Developing process safety indicators.

What is a KPI in plain English?

A key performance indicator is a selected measure used to judge whether important controls and improvement actions are working.

An injury total reports harm after it happened. An isolation verification measure asks whether a critical control was actually used before maintenance. Both matter. The first may miss a low-frequency catastrophic risk; the second can be gamed if the check is superficial. Pair them with observations, defects, worker reports and evidence that corrective action worked. Sources: HSE HSG65: Managing for health and safety, HSE HSG254: Developing process safety indicators.

A metric is a measure, such as a count, rate, percentage or duration; a KPI is selected because it informs an important decision. A target is the intended level. A threshold is the point that triggers review, a hold or escalation. The indicator is useful only when its owner can explain the response and the evidence behind the number.

The HSE guides cited for KPI methods are GB regulator guidance. The methods may help NI workplaces, but NI legal duties and incident reporting must be checked through the separate NI sources below.

Why do KPIs matter for risk management?

They let an organisation detect control drift before people are harmed and test whether its response changes the conditions of work.

For example, rising overdue inspections on safety-critical equipment should lead to a review of the affected equipment, maintenance resource and any required hold. A falling near-miss count after workers were blamed may indicate lost trust, not improved safety. Ask why the trend changed before praising or penalising anyone. Sources: HSE HSG65: Managing for health and safety, HSENI: Co-operation and monitoring safe working.

Which types of indicator belong together?

Combine lagging harm, leading prevention, control-health, exposure, learning and worker-feedback signals around the risk.

For a vehicle interface, injury reports show past harm; segregation inspections, barrier defects, traffic observations and worker reports show whether prevention still works. For hazardous substances, the right measures may cover exposure control, maintenance, valid sampling, surveillance where required and action on findings. Do not treat an activity count as proof of effectiveness. Sources: HSE HSG65: Managing for health and safety, HSE HSG254: Developing process safety indicators.

How is a useful KPI designed?

Define the risk, decision, owner, numerator, denominator, period, source, threshold and validation method before putting a number on a dashboard.

Write inclusion and exclusion rules in a KPI dictionary. Test the calculation against examples, contractor and shift coverage, missing data, duplicate records and changed operations. Ask workers whether the chosen measure tracks the real task. A target should improve a control without discouraging honest reporting. Sources: HSE HSG65: Managing for health and safety, HSENI: Co-operation and monitoring safe working.

Illustrative decision: nine of ten vehicle–pedestrian barriers due for verification are recorded as effective, giving 90%. If the unchecked barrier protects a live crossing, the supervisor does not release that crossing because the percentage looks high. They separate people and vehicles, verify or restore the barrier, record the exception and test whether the dashboard exposed it promptly. The measure must identify which control failed, not just the average. Sources: HSE HSG65, HSE HSG254 process-safety indicator method. For the physical separation decision, see HSE separating pedestrians and vehicles (GB) and HSENI Drive Danger Out (NI practical guidance).

What is a bad KPI compared with a useful KPI?

Useful KPIs show whether important controls are working; bad KPIs often count activity without proving risk reduction.

A bad KPI is not always a useless number. It may still tell part of the story, but it becomes weak when leaders treat it as proof of safety. For example, completed inspections show that checks happened; they do not prove that critical controls were present, defects were risk-ranked, actions were owned or repeat problems were challenged.

A useful KPI is closer to the risk. It asks whether the thing that prevents serious harm was verified, whether workers understood the control in the task, whether corrective actions changed behaviour and whether the organisation learned. This is why good dashboards combine quantity with quality, field verification and evidence of follow-through.

Bad or weak KPIWhy it can misleadMore useful KPI
Number of inspections completedIt counts activity but not whether important controls were effective.Percentage of critical controls verified as effective, with high-risk defects closed and rechecked. Sources: HSE HSG254: Developing process safety indicators.
Zero accidentsIt may reflect luck, under-reporting or low exposure rather than good control.Near-miss reporting quality, repeat-event themes, action closure and control-health trend. Sources: HSE HSG65: Managing for health and safety.
Training completedAttendance does not prove competence or safe behaviour in the task.Competence observed during task, supervisor coaching completed and recheck after change. Sources: HSE HSG65: Managing for health and safety.
Toolbox talks deliveredA talk can be delivered without understanding or behaviour change.Briefing understanding checked, worker questions captured and field observations confirm application. Sources: HSE HSG65: Managing for health and safety.
Actions closed on timeActions may close administratively without fixing the system.Actions closed with evidence, effectiveness review and no repeat finding after sampling. Sources: HSE HSG65: Managing for health and safety.
Audits completedAudit volume does not prove audit quality.Risk-ranked findings, systemic themes, overdue verification and leadership response. Sources: HSE HSG65: Managing for health and safety.

What should a practical dashboard show?

A short dashboard shows the risk, current control-health signal, trend, data quality, threshold, owner, decision and follow-up evidence.

Separate immediate exceptions from monthly trends. A failed critical barrier may need a same-shift hold even when the monthly percentage is green. Show the denominator and sample size beside a percentage; show overdue high-consequence actions separately from routine actions. Include a narrative about changed work or data gaps. Sources: HSE HSG65: Managing for health and safety, HSE HSG254: Developing process safety indicators.

Which calculations are useful?

Rates can compare periods with different exposure, but their definitions and limits must travel with the result.

For an on-time control-check rate, divide checks due in the reporting period that have a valid result completed by their individual deadlines by all checks due in that period, then multiply by 100. Fix the reporting cut-off and retain each due date and actual completion time. If all 49 of 50 due checks had valid results completed on time, the rate is 98%; the remaining check still needs a risk decision. Show late completions and missing results separately. Old overdue checks are outside this period’s on-time numerator; a late completion must never be backdated into success. Injury frequency rates also depend on hours, inclusion rules and a chosen multiplier. A small number of events can swing a rate sharply, so show the count and context too. Sources: HSE HSG65: Managing for health and safety.

Exposure-rate example: in one stated calendar year, one work-related injury causing more than seven days’ inability to do normal duties occurred among the employees and contractors included in 100,000 recorded hours worked. Using a declared multiplier of 1,000,000: 1 ÷ 100,000 × 1,000,000 = 10 qualifying injuries per million hours worked. The injury category, period and employee/contractor scope must match the hours; this is a rate per hours, not per workers, and the multiplier is a reporting choice rather than a universal target.

One event cannot establish a reliable organisation-wide trend. Nor can the worked example’s five deliberately selected high-risk contractors represent the whole organisation. Show counts, hours, sampling method and uncertainty; incompatible scopes or exposure cannot support simple league tables. Act on critical-control failures despite statistical uncertainty.

How can KPI results be calculated without hiding missing evidence?

Show both performance and evidence coverage. Separate passed, failed and missing records; a zero denominator has no percentage, and a combined result should retain the site or shift results needed to make a risk decision.

Decide what question each percentage answers before collecting it. A pass rate, evidence coverage and completion rate can use different denominators even when drawn from the same log. These original teaching calculations make that distinction visible; they are not regulator-set targets. HSE’s monitoring approach connects reports to control effectiveness and action. Sources: HSE: Check performance, HSE HSG65: Managing for health and safety.

Log: 100 checks are due. There are 80 valid completed records: 72 pass and eight fail. The remaining 20 due checks have no valid result. Keep those three groups separate. A missing record does not prove a control failed, but it cannot provide evidence that the control passed.

On a narrow screen, scroll this example table sideways to read all columns.

Worked KPI calculations and evidence gaps
MeasureCalculationMeaning and next decision
Observed pass rate72 passed ÷ 80 valid completed checks × 100 = 90%.This answers how completed checks performed. Display the eight failures and 20 missing records alongside it; 90% is not coverage. Sources: HSE HSG65: Managing for health and safety.
Evidence coverage80 valid completed checks ÷ 100 due × 100 = 80%.Twenty due checks have no valid result. Find out which controls lack evidence and make a risk decision; do not invent passes or failures. Sources: HSE HSG65: Managing for health and safety.
Verified pass evidence against all due72 verified passes ÷ 100 due × 100 = 72%.Call it this explicitly. The other 28 comprise eight known failures and 20 unknown results, which need different follow-up. Sources: HSE HSG65: Managing for health and safety.
No activity due0 completed ÷ 0 due: percentage undefined.Display “not applicable: no checks due”, with the underlying zero counts. Confirm the activity really did not occur before accepting this status. Sources: HSE HSG65: Managing for health and safety.
Combined sites, a separate exampleSite A: 980/1,000 = 98%; site B: 10/20 = 50%. Combined: 990/1,020 ≈ 97.1%.Combine compatible counts, not percentages. Keep the site results visible: the large site dominates the total and the small site still has a serious gap. Sources: HSE HSG65: Managing for health and safety, HSE HSG254: Developing process safety indicators.

The combined-site result is a second example with its own counts. The simple average of 98% and 50% is 74%; that gives the sites equal weight and answers a different question. Use 990 divided by 1,020 only when the underlying check definitions and periods are compatible. Neither aggregation removes the need to investigate site B or any failed critical control.

For zero activity, distinguish “nothing was due” from “the data is missing”. If work occurred but the due count is unknown, show an unavailable result and obtain the missing evidence. Do not label it not applicable or green. Keep original records and a dated correction explaining any duplicate removal or denominator change so the revised result remains traceable. Sources: HSE HSG65: Managing for health and safety.

Management sets the response rule from the actual risk and control standard. HSG254 discusses tolerances and follow-up for process safety; a dashboard tolerance does not grant permission to operate with a failed critical barrier. A trend can go to the routine review while a serious control exception goes to the responsible manager immediately. Sources: HSE HSG254: Developing process safety indicators, HSE: Check performance.

Learning check: explain what 90%, 80% and 72% each mean, identify the eight known failures and the 20 unknown results, and name the person who makes the next work decision. A correct percentage without that explanation is incomplete assurance.

Who checks KPI evidence and acts on it?

The work owner needs to act on control failures, a competent reviewer should challenge the quality of the measure, and management should test whether corrective action worked. Keep these responsibilities clear even in a small organisation.

A supervisor checks the actual guard, isolation or permit boundary and stops unsafe work. A safety or technical adviser reviews definitions, sampling and recurring exceptions. Management reviews the evidence, assigns resources and verifies closure. HSE places monitoring and review within its Plan, Do, Check, Act approach. Source: HSE HSG65 managing for health and safety.

An independent check can provide additional challenge, but a green dashboard cannot overrule a failed critical control. Record who owns the control, who verified the data and who authorised restart after a failed barrier. Source: HSE HSG65 managing for health and safety.

ResponsibilityUseful KPI evidenceDecision or challenge
Work ownerActual guard, isolation, permit and pre-use checks, with exceptions visible.Hold unsafe work and verify correction; an activity count alone does not prove control. Source: HSE HSG65 managing for health and safety.
Competent reviewerRisk-ranked exceptions, sample quality, repeat defects and verified close-out.Challenge weak definitions, optimistic samples and systemic risk; escalate findings. Source: HSE HSG65 managing for health and safety.
Management review or independent checkTest control design, data quality, response and unresolved themes.Decide resources and follow-up; independent review should not take over the work owner’s control decision. Source: HSE HSG65 managing for health and safety.

What can make a KPI misleading?

Missing or duplicate records, inconsistent definitions, changing denominators, biased samples and incentives to hide reports can make a reassuring figure false.

Compare source records with dashboard totals. Sample different sites, shifts and contractors; check whether closed actions changed field conditions. A sudden drop in hazard reports may reflect fear or a broken reporting route. State uncertainty rather than manufacturing a precise trend from incomplete data. Sources: HSE HSG65: Managing for health and safety, HSENI: Co-operation and monitoring safe working.

How can a good-looking result conceal risk?

A 98% contractor induction result can hide a serious understanding gap if attendance is counted as competence.

Suppose 49 of 50 contractors signed in, so the induction KPI reads 98%. A supervisor then samples five high-risk contractors: two cannot explain permit boundaries and one cannot describe the site alarm. The supervisor holds the affected work, briefs the people involved and verifies understanding at the point of work; the one missed induction is traced as well. The 98% headline remains in the record but is no longer described as assurance. Sources: HSE HSG65: Managing for health and safety, HSENI: Co-operation and monitoring safe working.

The contract manager owns a revised measure: high-risk contractors who pass a task and emergency understanding check, divided by high-risk contractors sampled. The sample, exclusions and assessor are recorded. Any failure on a critical permit boundary triggers an immediate hold regardless of the monthly percentage. The manager reviews the induction content, language, supervision and onboarding process; a later field sample closes the action only if work is correct. Sources: HSE HSG65: Managing for health and safety.

Three steps for a health and safety KPI. Name the risk and critical control; define and validate numerator, denominator and field sample; set the hold or escalation threshold and verify corrective action. A 98 percent induction attendance rate is not competence.
A KPI is useful when its measure leads to a risk decision and a field check. Open diagram at full size

How should a KPI follow the control hierarchy?

Measure whether the strongest feasible control works on the real hazard, then make a failed barrier change the work decision. A high completion percentage cannot compensate for an unsafe entry.

For the illustrative contractor permit-boundary example, the hazard is entry to a live high-energy area, not the KPI itself. HSE places elimination and substitution ahead of engineering and administrative controls, with PPE for residual risk. Measure whether the chosen barrier works in the field, and hold work when it fails. Source: HSE managing risk using PPE.

Task-specific HSE control hierarchy decisions
Task and control levelActual control and feasibilityOwner, evidence, failure and stop
Live-area contractor entry: 1 EliminationIn the illustrative Crownway task, use remote inspection or complete fabrication outside the live area where the information and plant design allow; then no contractor crosses the live permit boundary.Task owner records which entries were genuinely removed, with a field sample. Any unplanned entry or missing observation stops the release decision. Sources: HSE HSG254: Developing process safety indicators, HSE managing risk using PPE.
Live-area entry: 2 SubstitutionTrial a lower-energy or off-site method that still proves the required condition; do not call a remote image equivalent to a physical test without technical acceptance.Engineering owner signs the method specification and verifies the evidence. An unproven substitute is rejected. Sources: HSE HSG254: Developing process safety indicators, HSE managing risk using PPE.
Live-area entry: 3 EngineeringBefore unavoidable entry, prove physical isolation and the relevant barrier or interlock at the boundary; a dashboard or induction record cannot create separation.Permit issuer verifies isolation on the plant and records the check. Failed or uncertain barrier means no entry. Sources: HSE HSG254: Developing process safety indicators, HSE managing risk using PPE.
Live-area entry: 4 AdministrativeUse a permit, boundary walk-through and task demonstration, then sample contractor understanding in the field. The KPI records failed checks and overdue fixes, not attendance alone.Supervisor owns the hold and retest; manager trends exceptions and resources corrections. A failed boundary check overrides a 98% attendance result. Sources: HSE HSG254: Developing process safety indicators, HSE HSG65: Managing for health and safety.
Live-area entry: 5 PPESelect any residual protection for the identified exposure and verify compatibility and fit; it does not replace isolation or a correct permit. Do not turn PPE issued into proof that a critical barrier works.Competent selector checks condition and use; failed isolation remains a no-entry decision regardless of PPE. Sources: HSE managing risk using PPE, HSE HSG254: Developing process safety indicators.

For the separate Crownway proof-test example, a failed critical barrier is an immediate hold even if the monthly KPI is green. The owner first asks whether the hazardous task can be removed or changed, then verifies engineered protection before relying on permit conditions or residual PPE. Record the failed test, action owner, independent field recheck and restart decision; do not count administrative closure as restored protection. Sources: HSE HSG254: Developing process safety indicators.

How does useful measurement change work?

A useful KPI leads to a verified decision, while a weak one rewards easy counts or hides exceptions.

Counting 100 inspections says little if none tested a critical control. A better review asks which controls were sampled, what failed, who held or changed the task, whether workers agreed that the fix worked and whether the same defect returned. Leaders should question apparently perfect figures where independent checks or worker reports tell a different story. Sources: HSE HSG65: Managing for health and safety.

What evidence shows KPIs are managed well?

Evidence should show that KPIs are defined, reviewed, challenged and acted on.

Evidence should be read as a chain rather than a filing exercise. A single record rarely proves good management on its own; stronger assurance comes from matching records, observations, interviews, trends and verified action closure.

EvidenceWhat it provesQuality check
KPI dictionaryDefinitions, scope and calculations are consistent.Can every site calculate it the same way? Sources: HSE HSG65: Managing for health and safety.
Risk-profile linkKPIs reflect significant risks.Are top risks visible on the dashboard? Sources: HSE HSG254: Developing process safety indicators.
Dashboard reviewsLeaders review performance regularly.Are decisions, challenges and actions recorded? Sources: HSE HSG65: Managing for health and safety.
Action trackerPoor performance results in owned actions.Are actions risk-ranked and verified? Sources: HSE HSG245: Investigating accidents and incidents.
Audit samplesData is tested for accuracy.Do samples match records and work as done? Sources: HSE: Check performance.
Trend analysisThe business learns from patterns.Are repeat themes escalated? Sources: HSE HSG65: Managing for health and safety.
Worker feedbackIndicators include workforce voice.Are concerns closed with feedback? Sources: HSENI: Co-operation and monitoring safe working.

Who needs to learn and demonstrate KPI decisions?

Competence means being able to define, validate, question and act on the measure appropriate to the role, not merely attending dashboard training.

A worker should know how to report a failed control without being judged by an injury target. Supervisors must distinguish a missing record from a failed barrier and know when to hold work. Analysts need to reconcile source data and challenge denominator changes. Managers and directors must decide resources and verify closure. Assess these decisions using actual examples and retain evidence of performance. Sources: HSE HSG65: Managing for health and safety, HSENI: Co-operation and monitoring safe working.

Health and Safety Key Performance Indicators competence: roles, demonstrations and review records
RoleWhat to learnDemonstrationAssessorRecordReassess when
WorkerControl failure and honest reporting routeExplain and report a failed control even if monthly KPI is greenSupervisorScenario observation and reportProcess or reporting route changes Sources: HSE HSG65: Managing for health and safety, HSENI: Co-operation and monitoring safe working.
SupervisorThreshold, task hold, sample and escalationRespond to failed critical check and trace a missing recordOperations managerObserved decision and field verificationNew risk or repeated defect Sources: HSE HSG254: Developing process safety indicators.
KPI analystDictionary, numerator, denominator, sampling and data ethicsReconcile dashboard with source records and explain uncertaintyAssurance leadTest calculation and data-quality sampleSystem, definition or denominator changes Sources: HSE: Check performance, HSE HSG65: Managing for health and safety.
Manager/directorRisk trade-off, resource decision and closure assuranceChallenge a green dashboard with failed barrier and authorise corrective actionCompetent senior leader or boardDecision log and effectiveness reviewMajor event, drift or strategy change Sources: HSE HSG65: Managing for health and safety, HSE HSG254: Developing process safety indicators.

How should a learner explain performance monitoring?

Link each proposed measure to a risk, a control, the quality of the evidence and the decision that follows.

In an assessment, distinguish lagging harm from leading and control-health evidence, show one numerator and denominator, then explain an exception and the owner’s response. Explain why a zero-injury month cannot prove the system safe. HSG65 frames monitoring and review within Plan–Do–Check–Act. Sources: HSE HSG65: Managing for health and safety.

What can a health and safety management system add?

A management system can use KPIs to test whether important controls work and whether corrective decisions improve the work. Certification alone does not prove legal compliance.

HSE describes ISO 45001 as an optional health and safety management standard. Its guidance asks organisations to tailor arrangements to their size and risk, and says inspectors use a range of evidence and observations. Use defined measures, worker feedback, field verification and recorded management decisions; check any exact standard requirement against authorised text. Source: HSE health and safety management systems.

What warning signs show KPIs are not adding value?

Warning signs appear when numbers are collected but not trusted, understood or acted on.

Warning signs are included because failure usually appears before a serious event. Treat these points as prompts for inspection, audit, supervision and management review rather than as simple compliance statements.

Warning signWhat it suggestsAction
Only lagging indicatorsThe dashboard relies on outcomes or failures without checking preventive activities; a barrier failure or near miss can be a lagging result even when nobody is injured.Add leading and control-health measures. Sources: HSE HSG254: Developing process safety indicators.
Perfect dashboardsUnder-reporting or weak challenge may exist.Audit data quality and reporting culture. Sources: HSE HSG65: Managing for health and safety.
Huge number of measuresSignal is buried in noise.Focus on top risks and decisions. Sources: HSE HSG65: Managing for health and safety.
Actions closed with no evidencePaperwork closure is replacing improvement.Require effectiveness checks. Sources: HSE: Check performance.
No worker voiceDashboard misses work-as-done signals.Add participation and concern close-out measures. Sources: HSENI: Co-operation and monitoring safe working.
Repeated red indicatorsEscalation is weak.Review ownership, resources and risk acceptance. Sources: HSE HSG65: Managing for health and safety.

How can the hierarchy of controls verify actual conveyor isolation behind a safety-performance KPI?

What do the acronyms and key terms mean?

Keep the measure and legal instrument clear whenever results are discussed.

TermMeaningUse
KPIKey performance indicatorSelected measure linked to an important risk decision
LTIFRLost-time injury frequency rateLagging injury rate; definition and multiplier must be stated
RIDDORReporting of Injuries, Diseases and Dangerous Occurrences RegulationsGB 2013 and NI 1997 are different instruments
OH&SOccupational health and safetyHealth and safety of people at work
RAGRed, amber, greenStatus shorthand that needs a transparent threshold and action