10.2 KPIs & Characteristics of Good Metrics
Key Takeaways
- Useful H&S KPIs are specific, measurable, achievable, relevant, and time-bound (SMART-style) and should drive action, not only reports.
- A balanced set covers people, process, plant, and performance — not only injury frequency.
- Pitfalls include vanity metrics, gaming, only lagging measures, and targets that encourage under-reporting.
- KPIs sit primarily in the Check stage of Plan-Do-Check-Act and feed Act (improvement) and re-planning.
- A worked shift from only Accident Frequency Rate (AFR) to a mix of leading indicators shows stronger management control.
10.2 KPIs & Characteristics of Good Metrics
Quick Answer: Good H&S KPIs are SMART (specific, measurable, achievable, relevant, time-bound) and lead to action. Use a balanced set across people, process, plant, and performance. Avoid vanity metrics, gaming, and only lagging injury rates. KPIs support the Check stage of Plan-Do-Check-Act. Prefer a mix of leading indicators over AFR alone.
Knowing active versus reactive is step one. Step two is choosing key performance indicators (KPIs) that busy managers will actually use. A KPI is a small number of measures that tell you whether you are on track against objectives. In health and safety, bad KPIs create theatre; good KPIs create learning and timely control.
What makes a useful H&S KPI
Apply SMART-style tests (wording varies slightly by employer; the idea is consistent):
| Characteristic | Meaning for H&S | Weak example | Stronger example |
|---|---|---|---|
| Specific | Clear what is counted | “Improve safety” | “Close inspection actions within 14 days” |
| Measurable | Data can be collected consistently | “Better culture” (undefined) | “Near-miss reports per month with investigation started within 48 hours” |
| Achievable | Realistic with resources | “Zero every risk forever” | “100% of high-risk tasks covered by in-date risk assessment by Q2” |
| Relevant | Linked to real risk and duties | Counting posters only | % of statutory thorough examinations completed in date |
| Time-bound | Period and deadline clear | “Sometime this year” | “By end of each calendar month” |
Leads to action — the sixth test
A metric that never changes a decision is decoration. Ask:
- Who reviews it?
- What threshold triggers action?
- What do we do when it is red?
- Does the team understand it?
If the board only receives a green traffic light with no narrative, the KPI has failed the action test.
Balanced set: people, process, plant, performance
Relying on one number (usually accident frequency) is fragile. A practical balance for managers:
| Dimension | Focus | Example KPIs |
|---|---|---|
| People | Competence, engagement, behaviour support | % training in date; induction completion; observation of safe behaviours; near-miss participation rate |
| Process | Systems and procedures working | % risk assessments current; PTW audit compliance; investigation quality scores; procedure reviews completed |
| Plant | Workplace and equipment integrity | Inspection completion rate; statutory exam compliance; defects open >30 days; guarding checks |
| Performance | Outcomes and results | Injury/ill-health rates; days lost; RIDDOR events; damage costs; enforcement outcomes |
People + process + plant lean active/leading. Performance often includes reactive/lagging outcomes. Together they answer: Are our people ready? Are our systems alive? Is our plant safe? Are outcomes improving?
Building a small local dashboard (example)
A department manager might track monthly:
- People: % of team with in-date role-critical training
- Process: % of risk assessments within review date
- Plant: Planned workplace inspections completed / planned
- Plant/process: Open high-priority defects aged over 14 days
- Performance: First-aid + lost-time injuries (count and rate)
- People/culture: Near-miss reports with feedback closed to reporter
That is enough for a 15-minute review without drowning in spreadsheets.
Pitfalls — how good intentions go wrong
| Pitfall | What it looks like | Why it hurts |
|---|---|---|
| Vanity metrics | Poster counts, “safety days held,” likes on a safety app with no link to risk | Looks busy; little effect on serious hazards |
| Gaming | Reclassifying injuries, delaying reports, excluding contractors, changing definitions mid-year | Rate improves on paper while risk stays high |
| Only lagging | Dashboard is 100% AFR / LTIs | No early warning; success depends on luck |
| Targets that encourage under-reporting | Bonus solely for “zero accidents”; public shaming of reporters | Silence; hidden harm; fake culture |
| Too many KPIs | 40 measures, none reviewed | Noise; nothing managed |
| Unowned metrics | Data collected by H&S team, ignored by line managers | No operational accountability |
| Activity without quality | 200 toolbox talks of poor content | Tick-box culture |
| Comparing unequal rates | Celebrating lower AFR without normalising hours or risk profile | False comfort |
Exam angle: If a vignette describes a bonus for zero injuries and then falling report rates, identify under-reporting risk and recommend leading indicators + just culture.
Linking KPIs to Plan-Do-Check-Act (Check stage)
HSG65-style Plan-Do-Check-Act (and similar management cycles) places measurement firmly in Check, which then feeds Act and re-planning.
| PDCA stage | H&S activity | Role of KPIs |
|---|---|---|
| Plan | Policy, risk assessments, objectives, resources | Set KPI definitions and targets aligned to significant risks |
| Do | Implement controls, train, supervise, maintain | Generate the activities that leading KPIs count |
| Check | Monitor, inspect, audit, investigate, measure | KPIs live here — compare plan vs reality |
| Act | Review, improve, update assessments and plans | KPI trends drive corrective and preventive action |
Managers should not treat KPIs as an annual report decoration. Monthly Check conversations: What do the numbers say? What did we verify on the ground? What will we change?
Worked example: shifting from only AFR to a mixed set
Starting point (weak)
A warehouse tracks one number: Accident Frequency Rate (AFR) (or similar lost-time frequency). Last year AFR fell. Leadership celebrates. Nobody notices that:
- Temporary agency hours rose (different reporting habits)
- Near-miss forms stopped arriving after a supervisor mocked a reporter
- Three racking inspections were skipped in peak season
- Several risk assessments still refer to old layout before a mezzanine was added
The AFR looked good because harm was not yet realised and some events may not have been recorded.
Improved KPI set (stronger)
| KPI | Type | Target example | Action trigger |
|---|---|---|---|
| AFR / LTI rate | Lagging (performance) | Trend down year-on-year | Any LTI → investigation + leading KPI review |
| Near-miss reports with closed feedback | Leading (people) | Stable or rising quality reports | Drop >30% QoQ → culture check |
| % RA current for warehouse tasks | Leading (process) | ≥95% | Below 90% → freeze non-critical changes until reviewed |
| Planned inspections completed | Leading (plant/process) | 100% of plan | Missed week → escalate to operations manager |
| High-priority defects open >14 days | Leading (plant) | Zero aged critical defects | Daily review until cleared |
| Agency induction completion before first shift | Leading (people) | 100% | Any breach → stop work for individual until inducted |
What changes in behaviour
- Supervisors are scored on inspection completion and defect age, not only “no accidents on my shift.”
- H&S and operations jointly review near-miss quality, not only volume.
- Layout change triggers process KPI (risk assessment currency), preventing silent obsolescence.
- AFR remains on the dashboard as a truth check on outcomes, not the only scoreboard.
One-year narrative for the exam
Before: Single lagging KPI → false comfort.
After: Balanced people/process/plant/performance set → earlier detection of skipped inspections, stale assessments, and reporting fear.
Result concept: Even if AFR is stable for a while, system health is visible and manageable.
Practical rules for writing KPIs (manager toolkit)
- Start from significant risks — not from what is easy to count.
- Prefer rates or percentages where volume fluctuates (hours worked, headcount).
- Define the numerator and denominator in one sentence.
- Set review frequency (weekly operational, monthly management, annual board).
- Pair every lagging KPI with at least one leading KPI in the same risk area.
- Publish definitions so gaming by redefinition is harder.
- Reward learning and closure of actions, not silence.
Manager checklist
- Are our top five KPIs SMART and owned by named managers?
- Do we cover people, process, plant, and performance?
- Could anyone hit a target by hiding bad news?
- Do red KPIs automatically create actions with deadlines?
- Is AFR (or equivalent) supported by leading measures, not worshipped alone?
Choose fewer, better metrics; put them in the Check stage; and use them to change work — that is the Managing Safely standard for performance measurement beyond a single accident rate.
Which set best reflects SMART-style characteristics of a useful H&S KPI?
A balanced H&S KPI set for managers should typically cover:
In Plan-Do-Check-Act thinking, health and safety KPIs are primarily used in which stage?
Which approach best improves on managing only by Accident Frequency Rate (AFR)?