Workplace safety analysis divides the causes of incidents into two buckets: unsafe acts and unsafe conditions. An unsafe act is a behavior, something a person does that increases risk. An unsafe condition is a state of the workplace, something about the equipment, environment, or layout that is hazardous on its own. The distinction matters because the two buckets are fixed in different ways: you train and supervise people out of unsafe acts, and you engineer, maintain, and inspect unsafe conditions away. This post explains both, with examples from forklift operations, and how to catch each before it becomes an injury. It sells none of these systems.
What counts as an unsafe act
An unsafe act is any behavior that departs from safe procedure. Common examples in a forklift operation: driving faster than the facility limit, texting while operating, skipping the pre-shift inspection, parking with the forks raised, lifting a load beyond the rated capacity, or riding with a passenger on the forks. The defining feature is choice: the behavior is under the worker’s control, which is why the response is training, supervision, and observation.
What counts as an unsafe condition
An unsafe condition exists in the workplace itself, independent of what anyone does at that moment. Examples: an oil spill on the aisle floor, a forklift with a cracked brake line, blocked emergency exits, pallets stacked past the safe height, poor lighting in a rack aisle, or a damaged load. The defining feature is state: fix the state and the risk goes away, which is why the response is maintenance, housekeeping, and inspection.
How the two combine
Most incidents involve both. A wet floor (unsafe condition) becomes a fall or a skid when someone walks or drives fast (unsafe act). A missing guard on a conveyor (unsafe condition) injures the worker who reaches in to clear a jam (unsafe act). Forklift accidents follow the same pattern: the accident statistics guide on this site shows that struck-by and caught-between events, the two largest patterns in the OSHA severe injury data, nearly always have a behavioral component layered on top of a physical one.
A field test: act or condition?
When a report comes in, the fastest classification is a two-question test. Question one: was a person’s behavior involved? If the answer is no, it is a condition. Question two: could the behavior have been different without changing the environment? If the worker could have followed the procedure and the risk would have disappeared, it is an act. A wet floor with a forklift driving across it at speed is both: the floor is the condition, the speed is the act. The practical rule for the correction is simpler still: if maintenance or engineering removes the risk, fix the condition; if training or supervision removes the risk, correct the act. The inspection checklist catches the conditions, and the observation program catches the acts, which is why both belong in the same program.
How to identify them before someone gets hurt
Conditions are found by inspection. The forklift inspection checklist guide covers the daily pre-shift check that catches defective brakes, horns, lights, and tires. The broader facility inspection catches the aisle, dock, and storage issues. Acts are found by observation: watching operations in real time and correcting behavior in the moment. Observation is the core of behavior-based safety, and it is also where computer vision enters the picture, with AI-based safety systems able to flag risky behavior patterns that human supervisors miss.
The reporting culture that makes it work
Both identification methods depend on reporting. Workers need a channel to flag unsafe conditions without fear of reprisal, and supervisors need a reason to observe rather than punish. The safety rules guide lists the procedures that make up a defensible program. When a near miss is reported, treat it as free data: the near-miss explainer on this site covers why under-reported near misses share the same root causes as the injuries that do happen.
The practical sequence is the same one this site recommends everywhere: score the exposure first with the Risk Scorer, fix the conditions the score points at, then add observation and training for the acts. Compare control families on the systems index, and when you are ready for vendor numbers, send your scope through the quote form. The methodology page explains how the scorer’s inputs are derived.
