SHARE Knowledge Hub
Safety Triangle, Near Misses and Serious Risk Learning
The safety triangle is a way to prompt learning from near misses, minor injuries and weak conditions before serious harm occurs. Its familiar ratios describe historical datasets, not a law that predicts a fatality after a fixed number of low-severity events. A credible high-potential event deserves urgent control action even if nobody was hurt and the organisation has recorded few incidents.
Understanding Safety Triangle, Near Misses and Serious Risk Learning
The safety triangle is a historical way of depicting many low-consequence events beneath fewer high-consequence events. It can encourage learning from weak signals before someone is badly hurt. It must not be treated as a law of nature, a fixed site-specific ratio or proof that reducing first-aid cases automatically reduces fatalities. The practical decision is to identify credible serious-harm pathways, find their precursors and verify that critical controls work.
Why this matters at work
Record every event with enough detail to assess actual harm and credible potential harm. Prioritise high-potential near misses, dangerous conditions and barrier failures even when nobody was injured. Investigate causes in work design, equipment, supervision and management; control the exposure and verify effectiveness. Keep common injuries and occupational ill health in the programme as real harm, but do not let their aggregate count obscure fatal hazards.
The triangle is a historical model illustrating that learning can occur before serious harm, not a fixed statistical law for every site. Different versions use different categories, eras, datasets and reporting conditions. Do not convert a count of first-aid cases or near misses into a predicted fatality or claim that reducing minor injuries automatically controls major hazards. The practical question is whether credible serious-harm pathways and their precursors are recognised and controlled. Continue preventing common injuries and occupational ill health as real harm, while separately examining high-potential events, dangerous conditions and critical-barrier failures. Record both actual and credible potential consequence and what prevented harm. Use suitable scenario and control evidence to prioritise action rather than a universal pyramid ratio. A low injury count can coexist with a rare but serious uncontrolled hazard. Explain the model's limits when using it in training so it encourages reporting and learning without creating misleading certainty or numerical quotas.
How organisations manage this in practice
Pair outcome data with checks of the barriers needed for each serious scenario. Define what is tested, its performance standard, ownership and failure response, and verify restoration in the field. Review technical, task and organisational contributors rather than closing each defect as an isolated behavioural issue. Record decisions and resource needs and communicate learning to similar areas. Do not use the historical triangle as evidence that improving minor-injury statistics will automatically repair these barriers; the control mechanism needs its own direct evidence.
A credible culture programme starts by removing known serious hazards and demonstrating fair response to concerns. Then improve the management practices that shape work: task design, resourcing, consultation, procedure usability, competence, supervision and learning. Choose a few risk-based priorities with owners, time, resources and evidence of completion. A branding exercise should not precede a decision to fix an obvious barrier.
Commitment is most visible in decisions about people, time, maintenance, equipment and tradeoffs. A director visit or statement has value only when concerns influence resourcing and operating restrictions. HSE leadership guidance emphasises strong, active leadership, worker involvement and assessment and review. A leader should know the organisation risk profile, its critical controls and what happens when they are impaired.
Separate the outcome from credible serious potential
Incident counts and injury outcomes do not measure the strength of a barrier. A heavy falling pallet can have serious potential even where nobody was injured.

What happened?
Describe the facts and immediate outcome without assuming that the absence of injury means the controls worked. Preserve uncertainty where evidence is incomplete.
Credible potential
Consider the actual load, energy, people and route. Distinguish a lightweight object in a controlled zone from a heavy pallet entering an occupied route.
Which barrier failed?
Examine exclusion, load stability, equipment and work planning. A defect found before a lift can require urgent attention despite no incident having occurred.
Make safe and investigate
Protect the affected task, preserve evidence and investigate proportionately to the credible risk. Statutory reporting is a separate check against the facts and current requirements.
Verify the repair
Assign and verify effective corrective action across comparable tasks. Do not close the case solely because the next lift happened without harm.
Historical model limits
Read historical injury triangles as attributed critical-learning material. Their ratios are not universal probabilities, causal laws or a basis for downgrading a high-potential event.
Read the complete controls: ERICPD planning prompts
Discipline verifies and sustains all the controls; it is not a separate substitute for them.
| Prompt or level | Complete task-specific example |
|---|---|
| Eliminate | Remove the need for the suspended-load movement by changing delivery or stock layout where feasible. |
| Reduce | Use a suitable alternative load unit or handling method that prevents the identified instability after an assessed trial. |
| Isolate | Keep people outside the lifting/drop zone and secure the affected equipment and damaged load against reuse. |
| Control | Provide suitable rated lifting equipment, load restraint and stable support; investigate attachment and pallet condition. Reassess the lift plan, load information, competence and communication using evidence from the near miss. |
| PPE | Use task-assessed head, hand and foot protection; PPE cannot protect a person underneath a falling load. |
| Discipline | Verify equipment repairs and the revised load-handling method under controlled conditions before closing the finding. Read the accompanying source table for the framework distinctions. |
Read the complete controls: ESEAP effectiveness order
Compare stronger feasible controls while preserving the actual task, checks and release requirements.
| Prompt or level | Complete task-specific example |
|---|---|
| Elimination | Remove the need for the suspended-load movement by changing delivery or stock layout where feasible. |
| Substitution | No separate measure is specified in this example. |
| Engineering controls | Use a suitable alternative load unit or handling method that prevents the identified instability after an assessed trial. Keep people outside the lifting/drop zone and secure the affected equipment and damaged load against reuse. Provide suitable rated lifting equipment, load restraint and stable support; investigate attachment and pallet condition. |
| Administrative controls | Reassess the lift plan, load information, competence and communication using evidence from the near miss. Verify equipment repairs and the revised load-handling method under controlled conditions before closing the finding. |
| PPE | Use task-assessed head, hand and foot protection; PPE cannot protect a person underneath a falling load. Read the accompanying source table for the framework distinctions. |
No pyramid ratio or incident-count threshold is used to determine safe operation or investigation priority.
Further source background retained from this topic
What is the safety triangle?
The safety triangle is a way to prompt learning from near misses, minor injuries and weak conditions before serious harm occurs. Its familiar ratios describe historical datasets, not a law that predicts a fatality after a fixed number of low-severity events. A credible high-potential event deserves urgent control action even if nobody was hurt and the organisation has recorded few incidents.
This page concerns incident learning and serious-risk precursors, not the fire triangle. Reporting patterns are shaped by exposure and trust; a low count alone cannot prove strong control. Primary source.
The traditional triangle is often explained as a base of unsafe acts, unsafe conditions or near misses that supports a smaller number of minor injuries and an even smaller number of serious injuries or fatalities. The useful learning is that weak controls produce signals before disaster, and organisations should not wait for serious harm before acting.
The model becomes dangerous when it is treated as a fixed ratio or when managers assume that reducing minor injuries will automatically prevent major accidents. A paper cut, a slipped step and a near miss with a reversing vehicle do not carry the same potential. A serious process safety near miss may be more important than a month of minor first-aid cases.
Modern use of the triangle should focus on potential severity and control weakness. Leaders should ask which events could have killed someone, which safeguards failed, which weak signals repeat, and whether actions have changed the system. The absence of near misses or lost-time injuries should never be allowed to quieten concern about critical control weakness.
| Concept | Meaning | Professional caution |
|---|---|---|
| Unsafe condition | A physical, organisational or procedural state that could allow harm. | Do not wait for injury before correcting repeated conditions. |
| Unsafe act | A behaviour that increases risk, often shaped by task design, pressure, competence or supervision. | Do not turn the triangle into a blame counter. |
| Near miss | An event that could have caused harm but did not on this occasion. | Assess potential severity, not just actual outcome. |
| High-potential near miss | A near miss with credible fatality, life-changing injury, major accident or serious loss potential. | Investigate at a level proportionate to what could have happened. |
| Weak signal | Early warning such as repeated defects, bypassed controls, poor reporting or drift from standards. | Amplify weak signals before they become accepted background noise. |
| Lagging indicator | A measure of events that have already happened. | Useful for learning, but too late to be the only measure. |
| Leading indicator | A measure of control presence, use, quality or reliability. | Should test whether serious-risk controls are working. |
Where did the safety triangle come from? Heinrich, Bird and later models
Heinrich’s 1931 triangle, Bird’s 1969 loss-control study and a later pyramid attributed to ConocoPhillips use different categories and evidence. They are useful historical teaching models; none supplies a fixed ratio for predicting the next serious injury or fatality.
The familiar triangle developed from attempts to learn from events beyond the relatively small number of serious injuries. Later versions widened the base to include property damage, near misses or behaviour observations. The diagrams below are original SHARE redraws of those historical models. Their shapes are schematic: band area is not a measured probability, and the three sets of categories are not interchangeable.
Read the diagram as text
Heinrich historical safety pyramid: 1 serious injury, 29 minor injuries and 300 no-injury incidents. Historical ratios, not predictions. Bird historical safety pyramid: 1 serious injury, 10 minor injuries, 30 property-damage incidents and 600 no-loss incidents. Historical ratios, not predictions. Model attributed to ConocoPhillips in 2003: 1 fatality, 30 lost-time injury cases, 300 recordable injuries, 3,000 near misses and 300,000 at-risk behaviours. Attribution is not a verified study claim; historical ratios are not predictions. Compare the categories before comparing the numbers. Heinrich and Bird have a serious-injury apex; the ConocoPhillips-attributed variant has a fatality apex and adds behaviour observations. Equal-height triangles aid comparison only. The original ConocoPhillips report, dataset and method have not been verified in this review.
1931 · Herbert W. Heinrich
Heinrich published Industrial Accident Prevention: A Scientific Approach in 1931. His early industrial-insurance work is associated with the familiar relationship of one major or serious injury, 29 minor injuries and 300 no-injury incidents. The book’s first edition is documented; the original underlying accident records have not been independently reconstructed here.
Read the diagram as text
Heinrich historical safety pyramid: 1 serious injury, 29 minor injuries and 300 no-injury incidents. Historical ratios, not predictions.
Useful lesson: It made prevention easier to explain: investigate warning events and hazardous conditions before someone suffers a serious injury.
Limitation: It does not mean that every 300 near misses will produce one serious injury. Historical no-injury categories are not automatically identical to the near-miss definitions used by a modern organisation.
Historical source basis: Heinrich’s 1931 first edition · WorldCat bibliographic record · Heinrich and Bird ratios · Energy & Utility Skills training appendices · HSG101 historical study.
1969 · Frank E. Bird Jr.
Bird’s work at the Insurance Company of North America widened attention from injuries to loss control. DNV’s historical account describes analysis of 1,753,498 accident reports from 297 companies across 21 industrial groups, producing the widely reproduced 1:10:30:600 relationship. The four levels represent serious injury, minor injuries, property-damage incidents and incidents without injury or damage.
The supplied Bird image uses 9.8 and 30.2. Later institutional training material reports those figures and explains the familiar rounded 10 and 30 used in this redraw. The approximately 600 no-loss incidents came from around 4,000 hours of confidential interviews, rather than a complete census of events.
Read the diagram as text
Bird historical safety pyramid: 1 serious injury, 10 minor injuries, 30 property-damage incidents and 600 no-loss incidents. Historical ratios, not predictions.
Useful lesson: Damage and no-loss events can expose weaknesses in equipment, task planning and management controls that an injury-only review would miss.
Limitation: The study’s reported and interview-based categories should not be treated as a universal causal sequence. A damaged object and a fatal-harm precursor need assessment of their actual hazard mechanisms and safeguards.
Historical source basis: Bird’s study and loss-control work · DNV historical account · Heinrich and Bird ratios · Energy & Utility Skills training appendices.
2003 · A later model attributed to ConocoPhillips Marine
A five-tier pyramid is widely attributed to ConocoPhillips Marine in 2003 and is reproduced in later safety literature, including a 2013 conference paper. This variant places one fatality above 30 lost-time injury cases, 300 recordable injuries, 3,000 near misses and 300,000 at-risk behaviours.
The original company study, dataset and method have not been located or verified in this review. The illustration is therefore labelled as an attributed variant, rather than presented as an independently verified study. “Lost-time injury cases” counts cases, not days lost. “Recordable” is this historical variant’s terminology; it does not define UK RIDDOR reporting requirements.
Read the diagram as text
Model attributed to ConocoPhillips in 2003: 1 fatality, 30 lost-time injury cases, 300 recordable injuries, 3,000 near misses and 300,000 at-risk behaviours. Attribution is not a verified study claim; historical ratios are not predictions.
Useful lesson: It invites attention to conditions and workarounds before an incident, including the equipment, work design and management decisions that shape what people do.
Limitation: It cannot establish that 300,000 behaviours cause one fatality or that collecting a quota of observations will prevent a death. Do not turn its base into a worker-blame score or add the categories into a total when their original definitions are uncertain.
Historical source basis: Later reproduction of the ConocoPhillips attribution · Freibott (2013) · HSE · Behavioural safety approaches and the triangle’s limits.
| Model | Commonly reproduced ratio | Evidence boundary | Limit on use |
|---|---|---|---|
| 1931 · Herbert W. Heinrich | 1 : 29 : 300 | Book and later accounts; original records not reconstructed. | It does not mean that every 300 near misses will produce one serious injury. Historical no-injury categories are not automatically identical to the near-miss definitions used by a modern organisation. |
| 1969 · Frank E. Bird Jr. | 1 : 10 : 30 : 600 | Large reported-accident study; base also uses interviews. | The study’s reported and interview-based categories should not be treated as a universal causal sequence. A damaged object and a fatal-harm precursor need assessment of their actual hazard mechanisms and safeguards. |
| 2003 · A later model attributed to ConocoPhillips Marine | 1 : 30 : 300 : 3,000 : 300,000 | Later published attribution; original company study unverified. | It cannot establish that 300,000 behaviours cause one fatality or that collecting a quota of observations will prevent a death. Do not turn its base into a worker-blame score or add the categories into a total when their original definitions are uncertain. |
UK evidence also shows why one fixed shape is unreliable. HSE’s historical HSG101 analysis discusses different injury and no-injury relationships across organisations and warns about differences in definitions and the limits of generalising from small case studies. HSE’s behavioural-safety guidance also cautions against transferring Heinrich’s triangle across occupational ill health, minor personal injuries and major accidents. Current practice should assess credible potential harm and verify the relevant controls.
Practical UK guidance and historical caution: HSG101 historical study · HSE · Behavioural safety approaches and the triangle’s limits · Primary source · HSG254.
How can harm occur and when must work stop?
Three near misses at a reversing route can be logged as “no injury” while the same unsegregated crossing remains open.
The vehicle is the energy source, pedestrian entry is the exposure route and physical separation is the missing barrier. The supervisor should stop or divert vehicle movement, protect the crossing and investigate why route design allowed repeated conflict. The manager then checks other crossings rather than waiting for the count to rise. HSE
How can the hierarchy of controls prevent another dropped load after a high-potential near miss?
Remove or reduce the hazard before relying on procedures or personal protection.
For safety triangle, assess the actual task and first look for a way to remove the hazard. Where that is not possible, consider a safer substitute, engineered separation and then clear work procedures. PPE protects against remaining exposure; it does not make an uncontrolled task safe. HSE hierarchy of controls
The selected measures need a named owner, a usable stop point and a check that they still work when the task or conditions change. In Northern Ireland, verify the separate legal route and site instructions.
How does the decision transfer between workplaces?
Apply the risk and barrier decision to the actual work and location.
A warehouse reversing near miss may require route redesign. A process plant loss-of-containment precursor may require testing an instrumented or mechanical barrier even with no injury. A healthcare sharps or violence near miss may expose a very different harm path and need a different control owner. Triage each by credible potential consequence and the barrier that failed; do not transfer a fixed pyramid ratio across these settings. HSE
How should safety-triangle information flow through the management system?
Near misses and weak signals should be triaged, investigated, acted on, verified and fed into risk profile and assurance review.
A good reporting system does not simply collect events. It separates immediate response, potential severity, investigation depth, action quality and learning route. An event screened by a competent person as low in both actual and credible potential consequence may be closed locally, but a high-potential event should trigger escalation, competent investigation and senior visibility. HSE.
The learning loop also needs a route back into the risk register, training, procedures, inspections, maintenance plans and board assurance. Otherwise the triangle becomes a graph on a dashboard rather than a control-improvement process.
| Stage | What happens | Quality test |
|---|---|---|
| Stabilise | Immediate controls are restored, area made safe and urgent escalation completed. | Has the immediate risk been made safe before paperwork continues? |
| Report | Workers report near misses, unsafe conditions, defects and weak signals. | Is reporting easy, trusted and free from blame for honest disclosure? |
| Triage | Events are screened for actual harm and credible potential severity. | Would this have been serious if timing, people or energy release had been slightly different? |
| Investigate | Evidence, timeline, controls and human factors are reviewed. | Does the investigation explain why the event made sense at the time? |
| Act | Actions remove or reduce the weak condition, not just remind people. | Will the action prevent or reduce recurrence? |
| Verify | Actions are checked in the workplace after completion. | Has the control actually changed and is it being used? |
| Learn | Lessons update risk assessments, briefings, standards, audits and indicators. | Can similar areas learn before they have the same event? |
| Review | Managers and boards review themes, repeat causes, overdue actions and control confidence. | Is leadership seeing high-potential risk, not only injury counts? |
Immediate emergency response, preventing further harm and safe stopping/isolation come first, with notification in parallel where possible. Only then preserve evidence so far as safely practicable and triage the event. Local closure requires competent screening of both low actual and low credible potential consequence, not merely the absence of injury.
How does a near miss become useful learning?
A near miss becomes useful when the organisation understands the serious-risk pathway and changes the controls that allowed it.
A warehouse has three reports in two months where pedestrians nearly step into a reversing route. A weak response logs all three as low-severity near misses because nobody was injured. A stronger response asks why the same pathway keeps appearing: route layout, lighting, reversing alarms, visiting drivers, cleaning times, supervision and worker confidence to stop vehicle movement. HSE.
The safety triangle works here only if the organisation treats the repeated near miss as a warning about a fatality pathway. The right output is not a bigger report total. It is changed segregation, better route design, stronger visiting driver controls, verified reversing aids and a supervisor check at peak times.
| Issue | Weak response | Stronger response |
|---|---|---|
| Classification | Recorded as three low-harm near misses. | Reclassified as repeated high-potential vehicle-pedestrian events. |
| Investigation | Workers reminded to take care. | Route, lighting, reversing aids, layout and supervision reviewed. |
| Human factors | Pedestrians blamed for not looking. | Time pressure, visibility, visitor unfamiliarity and route design considered. |
| Controls | Additional warning sign added. | Physical segregation, revised traffic flow and delivery briefing introduced. |
| Evidence | Near-miss record closed. | CCTV review, revised map, inspection, briefing and observation records retained. |
| Learning | Local toolbox talk only. | Risk register, transport assessment and proactive checks updated. |
What are the positives and limitations of the safety triangle?
The triangle can encourage reporting and learning before injury. It becomes misleading when a count, ratio or low injury rate is used as proof that serious risks are controlled.
Use the picture to start a conversation about real work, then test what people report against the hazard, exposure and safeguards. More reports can reflect greater trust; fewer reports can reflect lower exposure, changed definitions or silence. Neither direction proves improvement on its own.
| Use | Positive contribution | Limitation or misuse | Better practice |
|---|---|---|---|
| Reporting and worker voice | Makes it easier to explain why near misses and hazardous conditions deserve attention. | Quotas, blame or rewards for low counts can suppress reporting. | Thank people, make reporting simple and show what changed. |
| Learning before serious harm | A no-injury event may reveal a failed or missing safeguard. | Not every minor event shares the causes of a major accident; rare serious hazards may give few warnings. | Prioritise credible potential harm and the control that failed, even after one event. |
| A broader view of loss | Equipment damage and recurring defects can reveal problems missed by injury forms. | Different categories, exposure periods and reporting systems make ratios difficult to compare. | Define the data and compare like hazards, tasks and exposure. |
| A memorable teaching aid | A simple image helps people understand why organisations should act before injury. | Its shape can be mistaken for a sequence, a statistical law or a prediction. | Teach the caveat with the image: historical ratios are not predictions. |
| Better questions about work | Reports can identify awkward tasks, workarounds and resource pressures. | Counting at-risk behaviours alone can blame front-line workers and overlook design or management failures. | Ask why the action made sense in the situation and improve the system. |
| Useful input to a dashboard | Themes and repeat failures can guide investigation and management attention. | A low minor-injury rate does not demonstrate control of a tank overfill, fire or other major hazard. | Use hazard-specific leading and lagging indicators and direct checks of critical safeguards. |
| Evidence of improvement | Verified actions can turn reports into learning across tasks and sites. | Closing a form or lowering a count does not establish that exposure has reduced. | Name an owner, verify the control in use and check whether the problem recurs. |
Use the triangle as a prompt, then act on the actual risk
- Protect people: make the affected area safe and stop or restrict the task where a serious uncontrolled exposure remains.
- Record what happened and what could credibly have happened; distinguish the actual outcome from potential harm.
- Investigate the missing or failed safeguards and the equipment, task and organisational conditions behind them.
- Choose stronger controls, assign an owner and check comparable tasks or locations for the same weakness.
- Verify the change in real work, share the learning and monitor the relevant safeguards rather than aiming for a pyramid ratio.
Practical source basis: Primary source · HSG254 · HSE · Behavioural safety approaches and the triangle’s limits.
Why is the safety triangle not an exact science?
The triangle cannot prove that a certain number of minor events will lead to a fatality; serious-risk control depends on hazards, exposure and barrier health.
A fuel storage depot could report five years with no near misses and no lost-time injuries while still operating tank-filling routines with poorly trained personnel, defective high-high level alarms and a backlog of safety instrumented system maintenance. The injury record looks calm, but the major accident risk is not calm.
This is the flagship learning point: high-potential near misses and weak signals are valuable because they reveal the condition of serious-risk controls. The triangle should push leaders toward critical control verification, not comfort from low event numbers.
| Signal | Wrong conclusion | Better leadership question |
|---|---|---|
| No LTIs for five years | The depot must be safe. | What does current assurance say about tank overfill prevention, alarms and shutdown systems? |
| Few near misses reported | There are few problems. | Do operators and contractors trust reporting, and are weak signals recognised? |
| Routine tank filling continues | The task is normal and familiar. | Are people competent, supervised and working within the safe operating envelope? |
| High-high level alarm defects open | Maintenance will get to them. | Is an overfill prevention barrier impaired and should filling stop or be restricted? |
| SIS proof-test backlog | The backlog is an engineering issue. | What is the risk exposure while safety instrumented functions are overdue? |
| No recent incidents | No escalation needed. | Would a regulator, insurer or board accept the current barrier evidence? |
How should high-potential near misses be triaged?
Triage should ask what could credibly have happened, not just what actually happened.
High-potential triage is one of the most useful upgrades to traditional triangle thinking. It stops serious learning being lost because the actual outcome was lucky. A scaffold board dropped into an empty exclusion zone may be a high-potential dropped-object event; a chemical splash that narrowly misses the eye may reveal a serious exposure pathway. HSE.
Triage should be done quickly by someone competent enough to understand the energy, control failure and credible consequence. The classification can be corrected later as evidence improves, but the initial decision should err toward protecting people and preserving evidence.
| Prompt | Why it matters | Escalation example |
|---|---|---|
| Could timing have changed the outcome? | Many serious events are avoided only because nobody was present at the moment. | Dropped object lands in walkway just after shift change. |
| Was a critical control defeated or absent? | Loss of a critical barrier can matter even without injury. | Permit issued while isolation was incomplete. |
| Was high energy involved? | Vehicle movement, height, pressure, electricity, fire, chemicals and stored energy increase severity. | MEWP contacts structure near live equipment. |
| Has this happened before? | Repeat signals show the system has not learned. | Three similar line-of-fire near misses in one month. |
| Could more people have been affected? | Crowded or public-facing work changes consequence potential. | Gas alarm in public area with unclear response. |
| Is there uncertainty? | Unknown causes require caution and evidence preservation. | Unexplained trip, alarm bypass or release. |
Completed classification must record the object, height/energy and credible harm path; boundary design/maintenance, whether people could credibly enter, preventive and mitigating barriers, uncertainty and recurrence. An effective exclusion zone is evidence that a mitigation functioned, not that no event occurred. Investigation depth follows the credible potential and recurrence evidence, not an unbounded worst imaginable case.
What do learners commonly misunderstand?
The common mistakes are treating the triangle as a law, using it to blame workers, or relying on injury rates instead of control evidence.
For qualification learners, the triangle is best understood as a communication model rather than a scientific guarantee. It can support proactive learning, but it should be combined with incident investigation, human factors, barrier management and performance monitoring.
In professional practice, the triangle should drive better questions. Are we hearing bad news? Do supervisors escalate weak signals? Are we investigating potential severity? Are actions verified? Are boards seeing serious-risk indicators?
| Misunderstanding | Better interpretation | Why it matters |
|---|---|---|
| Low injury rate means low risk. | Low injury rate may hide poor reporting or lucky control failure. | Major accidents can occur after long injury-free periods. |
| More reports mean the site is worse. | More trusted reporting may mean better visibility. | Mature organisations often see more weak signals, not fewer. |
| All near misses are equal. | Potential severity and control criticality matter. | A serious near miss needs deeper learning. |
| Unsafe acts are root causes. | Behaviour is shaped by job, individual and organisational factors. | Blaming behaviour can leave weak systems unchanged. |
| Training fixes repeat events. | Training helps only if task design and controls support the behaviour. | Repeat events need design, supervision and assurance review. |
| The triangle replaces investigation. | It helps prioritise learning; it does not explain causation. | Evidence-based investigation is still needed. |
What should leaders see in a safety-triangle dashboard?
Leaders should see high-potential events, weak signals, repeat causes, action quality and control confidence, not only incident counts.
A board dashboard should not celebrate zero accidents without asking whether the organisation is listening. The best dashboard shows what has been reported, what has been learned, whether serious-risk controls are healthy and what decisions leaders need to make. HSE.
For process, construction or transport risks, the dashboard should separate personal-safety event counts from critical-control indicators. A site can have few minor injuries while relief valves are overdue, scaffold inspections are late or vehicle segregation is failing.
| Measure | What it shows | Leadership question |
|---|---|---|
| High-potential near misses | Serious-risk signals independent of actual injury. | Which controls failed and have they been verified after action? |
| Weak-signal themes | Repeat defects, complaints, bypasses or informal workarounds. | What does this say about system health? |
| Reporting spread | Whether reports come from all teams, shifts, contractors and sites. | Where might silence indicate under-reporting? |
| Action quality | Whether actions remove conditions or simply brief people again. | Are high-risk actions strong enough? |
| Overdue verification | Whether completed actions have been checked in the field. | Who owns overdue assurance? |
| Repeat-event rate | Whether learning is preventing recurrence. | Why are the same causes reappearing? |
| Critical control failures | Whether serious-harm barriers are impaired. | What must stop until controls are restored? |
When should safety-triangle signals trigger escalation?
Escalate when potential severity is high, controls are impaired, weak signals repeat or the organisation may not be learning.
Escalation should be based on credible harm, control weakness and uncertainty. A near miss that looks minor on the injury log may need immediate management action if a person could have been killed, a critical control failed or the same condition exists elsewhere.
Managers should also escalate silence. If a high-risk team never reports near misses, that may mean nothing is happening; it may also mean people do not trust reporting, do not recognise weak signals or believe nothing changes.
| Trigger | Why it matters | Immediate management action |
|---|---|---|
| Credible fatality or life-changing harm | Potential severity is high even if outcome was lucky. | Prioritise rescue, preventing further harm and making safe; then preserve evidence so far as safely practicable and start competent investigation. |
| Critical control failed | The barrier preventing serious harm may be unreliable. | Stop affected activity. Resume only after restoration or demonstrated effective temporary protection is implemented and verified for the credible scenario, within recorded limits, with an accountable owner and expiry/review triggers. Authorisation alone is insufficient. |
| Repeat near misses | Learning is not changing the system. | Run theme review and update risk assessment, supervision and assurance checks. |
| Under-reporting suspected | The organisation may be blind to weak signals. | Speak with workers, reps and contractors; test trust and reporting barriers. |
| Action not verified | Closed actions may not have improved control. | Require field verification before closing high-risk actions. |
| Senior risk appetite affected | Resources, design change or business decisions may be needed. | Escalate through governance with clear decision required. |
Before resumption, verify that temporary controls are implemented and effective for the credible scenario. Record permitted limits, owner and expiry/review trigger. Authorisation alone is not protection; high potential cannot be downgraded because no one was hurt.
Where does safety-triangle learning commonly fail?
Failure usually occurs when reports are counted but not understood, when blame suppresses reporting, or when actions are weak.
A near-miss system can look healthy because it has many entries, colourful charts and monthly targets. It is weak if workers do not trust it, if supervisors downgrade events to avoid attention, if serious potential is hidden in low-severity categories, or if actions are closed without verification.
The most damaging failure is normalisation. When repeated weak signals become familiar, people stop seeing them as warnings. That is why leaders need chronic unease: a disciplined curiosity about what the system is trying to tell them.
| Failure or weak practice | How it develops | What good looks like |
|---|---|---|
| Numbers without learning | Targets focus on how many near misses are reported. | Review quality, themes, potential severity and verified actions. |
| Blame response | Workers are criticised after reporting. | Use just culture and human factors investigation. |
| Severity downgraded | High-potential events are recorded as low risk because no injury occurred. | Use credible worst-case triage. |
| Repeated reminders | Actions say 'take care' or 'brief team' after every event. | Change design, equipment, scheduling, competence or supervision. |
| No extent check | Fix applied only at the exact location. | Check similar tasks, sites, shifts and contractors. |
| Board over-reliance on LTIs | Senior review celebrates low injury rate without critical-control evidence. | Include weak signals and serious-risk indicators. |
| Contractors excluded | Contractor near misses do not enter the system. | Include contractors, visitors and interfaces in reporting and learning. |
What evidence shows safety-triangle learning is working?
Evidence should show that weak signals are reported, prioritised, investigated, acted on, verified and used to improve risk control.
The evidence standard is not a large database. It is traceability from event to learning. A reviewer should be able to select a high-potential near miss and see the report, triage, evidence, causes, actions, verification, communication and risk-profile update.
Mature systems also show the absence of silence. Reporting should come from different teams, shifts, contractors and levels. If only HSE staff report issues, the organisation may not yet have a trusted learning culture.
| Evidence item | What it should show | Assurance question |
|---|---|---|
| Reporting procedure | How people report near misses, weak signals and unsafe conditions. | Is reporting easy and understood by workers and contractors? |
| Triage record | Actual harm, potential severity, immediate controls and investigation level. | Are high-potential events escalated consistently? |
| Investigation record | Evidence, timeline, controls, causes and human factors. | Does it explain system conditions, not only behaviour? |
| Action tracker | Owner, due date, priority, interim controls and verification. | Are actions strong enough and closed with evidence? |
| Trend review | Themes, repeat events, departments, shifts and contractors. | Are repeated signals driving improvement? |
| Board report | High-potential events, weak signals, critical-control failures and decisions. | Are leaders seeing serious risk? |
| Learning communication | Briefings, alerts, procedure updates and training changes. | Did learning reach people doing similar work? |
How should safety-triangle learning be assured?
Assurance should test whether reporting and learning improve controls, not whether the reporting form is complete.
First-line checks should confirm that supervisors discuss near misses, complete immediate actions and brief teams. Second-line assurance should sample high-potential events, action quality and repeat themes. Third-line review should test whether governance can see serious-risk signals and whether the system learns across sites.
The audit question is simple: can the organisation show that the same weak signal is less likely to recur because something meaningful changed?
| Assurance layer | What to check | Example test |
|---|---|---|
| First line | Supervisor review of recent near misses, local actions and briefings. | Ask workers what changed after the last reported near miss. |
| Second line | HSE/manager sample of potential severity, action quality and verification. | Select three high-potential events and test evidence from report to close-out. |
| Third line | Independent audit of reporting culture, trend review and governance. | Check whether serious-risk themes reach board or senior review. |
| Worker voice | Trust, ease of reporting and feedback after reports. | Ask whether people believe reporting leads to improvement. |
| Contractor interface | Whether contractor near misses are captured and learned from. | Sample contractor reports, inductions and action communications. |
| Management review | Whether leaders act on weak signals and resource decisions. | Trace a repeated theme into budget, design or programme decisions. |
Which duties and guidance apply in Great Britain and Northern Ireland?
Use the instrument for the work location and task; regulator guidance and voluntary standards have a different legal status.
Check the current statutory scope and site conditions before using a general lesson as an operational decision. GB duty; NI duty. NI reporting guidance.
| Route | Direct instrument or guidance | Supporting source | Decision supported |
|---|---|---|---|
| GB employer risk duties | Health and Safety at Work etc. Act 1974 | Management of Health and Safety at Work Regulations 1999 | Assess and control foreseeable harm before injury occurs. |
| NI employer risk duties | Health and Safety at Work (Northern Ireland) Order 1978 | Management of Health and Safety at Work Regulations (Northern Ireland) 2000 | Use the separate NI risk-management route. |
| GB incident-investigation guidance | HSE HSG245 incident-investigation workbook | HSE HSG65 managing health and safety | Use potential harm and underlying causes to prioritise learning; guidance is not a separate triangle duty. |
| GB specified dangerous occurrences | HSE RIDDOR dangerous occurrences | HSE HSG245 incident-investigation workbook | Check whether the event meets GB RIDDOR Schedule 2; not every near miss is reportable. |
| NI specified dangerous occurrences | NI RIDDOR 1997 | HSENI report an incident | Check the separate NI reportable-event categories and dutyholder reporting route; not every near miss is reportable. |
How should management check this work?
Use a defined plan, carry out the controls, check performance and correct what does not work.
For safety triangle, assign a competent owner to set the work controls, check their use and review them after change, failure or new evidence. HSE managing health and safety
A management-system label or certificate is not approval of a site limit, procedure or individual work decision. Apply GB and NI duties to the correct location.
What competence does each level need?
Assess the decision a person must make on the real task; attendance alone does not prove competence.
Record who assessed the demonstration, the authorised scope and the change or defect that should trigger reassessment. Users must recognise a failed control even when a training or inspection record exists.
| Role | Learning to test | Demonstration, assessor, record and reassessment |
|---|---|---|
| New worker | Recognise and report a near miss without waiting for injury | Describe a real hazard, immediate safe action and report route; supervisor records induction and checks after a change. |
| Frontline supervisor | Triage credible potential harm and make the task safe | Scenario-based high-potential classification, hold and escalation; manager samples decisions after an event. |
| Investigator | Trace event, exposure, barriers and underlying causes | Complete an investigation with worker evidence and corrective-control test; competent lead reviews after complex events. |
| Operations manager | Fund and verify controls rather than report-count targets | Trace one repeated signal to field change and feedback; senior manager samples action effectiveness each review cycle. |
| Risk or process-safety specialist | Distinguish injury counts from serious-hazard indicators | Demonstrate a barrier-health dashboard and challenge conflicting data; peer check after new major-hazard scenario. |
What do the acronyms and key terms mean?
These terms separate the source, mechanism and decision used in the lesson.
| Term | Full expression | Meaning here |
|---|---|---|
| Near miss | Event that could have caused harm | Assess credible potential consequence as well as actual outcome. |
| High-potential event | Event with credible severe outcome | Triggers urgent barrier and exposure review. |
| Leading indicator | Measure of control activity or health | Examples include safety-critical tests and overdue defect closure. |
| Lagging indicator | Measure after an event | Examples include injuries or loss-of-containment events. |
| RIDDOR | Reporting of Injuries, Diseases and Dangerous Occurrences Regulations | GB statutory reporting route for specified events; NI has a separate route. |
Unlock another resource
You have used this month’s two free downloads. Publish a post in your own words to unlock one more resource, or choose SHARE Individual for 10 downloads per month.
You decide whether to share. If you do, use the relevant platform disclosure tools and do not make claims you do not genuinely hold.