Near miss vs hazard vs incident
Three words used interchangeably in conversation and very differently in a safety program. The distinction is not pedantry — it determines what happens to the report.
The three definitions
Start with what each one is, before worrying about edge cases.
- A hazard is a condition, source or act with the potential to cause harm. It exists. Nothing has happened yet. A missing guardrail is a hazard for as long as it stays missing.
- A near miss is an event that occurred and could have caused harm, but did not. Something interrupted the sequence. The guardrail is missing and somebody stumbled at the edge and caught themselves.
- An incident is an event that did cause harm — injury, illness or damage. Somebody stumbled and fell.
The progression is the useful part: a hazard is a standing condition, a near miss is that condition finding an opportunity, and an incident is the same sequence without the lucky interruption. The same underlying failure can present as all three at different moments.
The same scene, three ways
| Scenario | Classification | Why |
|---|---|---|
| A pallet is stacked above the racking's rated height | Hazard | A condition with potential for harm. No event |
| A box falls from that stack into an empty aisle | Near miss | The event occurred. Nobody was in the aisle |
| A box falls and strikes a picker's shoulder | Incident | Harm occurred, however minor |
| A box falls and destroys stock, nobody nearby | Incident | Damage is harm. It is not a near miss just because nobody was injured |
| An operator routinely climbs the racking to reach top stock | Hazard (unsafe act) | A recurring behavior with potential for harm. Still no event |
| The operator's foot slips on the racking and they catch themselves | Near miss | The fall sequence started and was interrupted |
Two tests that resolve most cases
Applied in order, these settle nearly every argument:
- Did an event occur? If nothing happened and you are describing a state of the world, it is a hazard. If something occurred in time — something moved, failed, was released, or was done — it is an event.
- Did harm result? If yes, at any level, including damage to property, it is an incident. If no, it is a near miss.
The mistake worth guarding against is judging by severity rather than by occurrence. A near miss that could have been fatal is still a near miss; a scratch that required a plaster is still an incident. Potential severity is a separate and equally important field — it tells you how urgently to respond — but it does not change the category.
Where organizations legitimately differ
It would be convenient to say these terms are standardised. They are not, and pretending otherwise causes problems when you compare data across sites or contractors.
The word "incident" is the least stable. Some organizations use it as an umbrella term covering everything reportable, with injuries, near misses and hazards as subtypes underneath. Others use it strictly for events that caused harm, as this guide does. Both usages are common and neither is wrong — but a contractor using the umbrella definition will hand you numbers that mean something different from your own.
Other genuine variations:
- Unsafe acts. Some programs classify these as hazards, others as a separate category, others fold them into near misses. All three are workable.
- Good catch. Often reserved for a deliberate intervention rather than luck, and sometimes tracked separately to reward the behavior.
- Property-damage-only events. Usually incidents, but some programs treat low-value damage as a near miss to keep the incident count focused on injury.
- First aid cases. Incidents in every sensible internal taxonomy — but not OSHA recordable, which is a separate question from what you call them internally.
The practical advice is to write your definitions down, put examples next to them, and use the same taxonomy across every site you compare. A defensible internal definition applied consistently beats a "correct" one applied unevenly.
Why the classification matters
This is not filing for its own sake. The category determines what happens next.
| Hazard | Near miss | Incident | |
|---|---|---|---|
| Central question | Who fixes this, by when? | Why did this nearly happen? | What harm occurred, and why? |
| Typical response | Triage, assign, correct | Investigate, find cause, correct | Full investigation, cause, corrective action |
| Investigation depth | Usually none needed | Proportionate — the event, cause and contributing factors | Deepest — plus injury detail and classification |
| Metric role | Leading | Leading | Lagging |
| OSHA recordkeeping | Not recorded | Not recorded | Recorded if it meets the 1904 criteria |
Two consequences follow. First, misclassifying an incident as a near miss can suppress a case that belongs on your OSHA log — a recordkeeping problem, not just a semantic one. Second, misclassifying hazards as near misses inflates your near-miss numbers with items that were never events, and you lose the ability to see whether actual failure sequences are increasing.
Getting it wrong in practice
Three patterns account for most classification problems:
- The reporter decides the category. Frontline reporters should describe what they saw; classification is a review-time decision by someone who knows the taxonomy. A form that asks for a free description first and a type second gets better data than one that forces the choice up front.
- Severity drives the label. A frightening near miss gets logged as an incident because it felt serious. Keep potential severity as its own field so people have somewhere to put that judgement.
- Nobody revisits the initial call. An investigation sometimes reveals harm that was not apparent at first — a strain reported days later. Categories should be correctable after the fact, and the correction recorded.