Collisions between handling equipment and people account for a disproportionate share of serious injuries in this sector. It is also where monitoring has the strongest justification.
The hierarchy
Monitoring is the last resort, not the first.
Separate them physically. Barriers, segregated walkways, separate doors, one-way routes. Nothing else is as effective.
Separate them in time, where physical separation is impossible: vehicle-free periods, restricted zones during picking.
Improve visibility: mirrors, lighting, sightlines at corners, transparent doors.
Reduce speed by design — layout, surface, gradient — before by rule.
Then detect and warn, which is where monitoring belongs.
A facility deploying proximity detection instead of separation has bought a warning for a hazard it could have removed.
What detection systems do
Proximity warning to the operator, the pedestrian, or both, when they approach.
Zone control, slowing or stopping equipment in defined areas.
Blind spot and corner detection, at fixed points.
Speed limiting by zone, enforced by the machine.
Access interlocks preventing entry while equipment operates.
What to measure
Near-miss events from proximity systems, by location and hour. This is the leading indicator.
Speed exceedances by zone.
Impact events, with location and severity.
Pedestrian incursions into vehicle areas.
Time of day and shift patterns for all of the above, which localise the conditions.
Reading the data as conditions
The distinction that determines whether the system works.
A cluster of near-misses at one corner is a layout finding.
A cluster at shift change is a scheduling finding.
A cluster in one aisle during picking is a separation finding.
A cluster involving one operator may be a training finding — and treating it as a disciplinary one first stops the reporting that produced the finding.
Investigate location and circumstance before person, every time, and say publicly that this is the policy.
The under-reporting risk
If near-miss data produces discipline, near-misses stop being reported.
Which removes the only leading indicator available, leaving injuries as the measure.
Operators will disable sensors, avoid routes that trigger them, or find the detection threshold and work just inside it.
A falling near-miss rate is not necessarily good news, and distinguishing genuine improvement from suppressed reporting requires looking at whether anything changed to cause it.
Alarm fatigue
The most common failure of proximity systems.
Tuned too sensitively, they alarm constantly in a busy aisle, and operators stop reacting.
Tune to the actual risk geometry, which requires observation rather than a default setting.
Measure the alarm rate and treat a high one as a defect.
Differentiate warning levels, so that a genuine imminent collision is distinguishable from routine proximity.
Review the rate quarterly, because layout changes alter it.
What monitoring cannot do
It does not create separation.
It does not slow a machine that is already too close.
It does not compensate for poor sightlines or inadequate aisle width.
It does not work when disabled, which is why the reporting culture matters more than the technology.
Report it as a supplement to separation, never as an alternative, or the physical work will be deferred indefinitely on the grounds that detection is in place.
Tuning the alarm rate
The most common reason proximity systems stop working.
Measure alarms per operator per shift.
A rate above a handful means operators are ignoring them, which is a predictable response rather than a discipline failure.
Tune to the actual geometry: aisle width, sightlines, speed, stopping distance.
Differentiate severity, so an imminent collision sounds different from routine proximity.
Review quarterly, because layout changes alter the rate.
Treat a high rate as a defect in the system, and record it as such.
Reading a near-miss cluster
The order of investigation determines whether the reporting survives.
Location first: a corner, a doorway, a crossing point.
Time second: shift change, break end, despatch compression.
Activity third: which task brings people and equipment together there.
Conditions fourth: lighting, noise, floor, visibility.
Person last, as a training question, and only after the first four have been examined and recorded.
State this order publicly, because it is what convinces people that reporting is safe.