Half a Meter at the Cross-Aisle

Two defects, not one. The score is a symptom; the inputs and the closure are the problem.

Defect 1 - the assessment rated a task, not an exposure. The unit of analysis should be a 3,500 kg vehicle and a pedestrian sharing the same floor space, repeatedly, across two shifts, at separation distances under a meter. That is a permanent-disability-or-fatality consequence. A matrix calibrated so that such a consequence lands at 'medium' is mis-scaled and will mis-rank every traffic hazard in the building. Matrices also multiply likelihood by severity and let a low probability drag a catastrophic severity down to tolerable - the wrong behavior.

Defect 2 - the controls are entirely behavior-dependent. 'Training and signage' requires a pedestrian to look and a driver to sound a horn, perfectly, every time, on a 90-person, two-shift operation, in a blind cross-aisle. They reduce the probability of a human slip but do nothing to the energy or the shared space. They fail silently and are unverifiable after the fact. Controls that depend on continuous perfect human performance should never be the sole barrier against a fatality-class hazard.

Re-rank defensibly. Use a consequence-first scheme: assign severity band, then estimate exposure frequency from the near-miss rate (three events in eight weeks localized to three aisles - that is a high-frequency precursor in a defined zone), then count independent barriers that do not rely on behavior. Alternatively draw a bow-tie with the top event 'pedestrian struck by lift truck' and list the barriers on each side; you will find the threat side has essentially none that are physical.

Result: the hazard is intolerable and must be reduced by separation, not by instruction. Under a reasonable-practicability test, guardrail and gates are cheap relative to the consequence, which makes the current rating indefensible to a consultant or an inspector.

Spend it in hierarchy order, weighted to the zone with the evidence.

Item	Cost	Rationale
Guardrail, bollards and self-closing pedestrian gates at the three blind cross-aisles serving aisles 14–16	~$18,000	Eliminates the interaction; no behavior required; lowest maintenance
Pedestrian detection / blue-spot retrofit and speed limiting at 8 km/h on the six trucks with the highest hours in that zone	~$12,000	Engineered barrier that works when the horn is not sounded
Reconfigure aisles 14–16: one-way flow, 1.2 m marked walkway, clear end-of-aisle storage blocking sightlines, improved lighting	~$6,000	Removes the sightline failure that produced all four events
Rewrite and formally issue the traffic management plan with a management-of-change record, sign placement, and a documented pedestrian route	~$3,000	Converts a 2016 document into an operating control
Reserve	~$6,000	Gate interlock at the highest-volume crossing, or contingency

Why this order: separation removes the exposure independent of human performance, so its risk reduction holds on the night shift, with a new hire, and after the training fades. Detection and speed limiting are second because they act at the moment of failure. Layout and route changes are third because they are partly administrative but durable. Signage and general e-learning are last because they are the least reliable and the least verifiable.

Defer: full 22-truck fleet retrofit (order of $120,000) - build a phased three-year capital plan on the back of the near-miss data; and all-staff e-learning, which consumes budget for the smallest risk reduction.

Interim, this week: temporary barriers and cones at the three crossings, a written rule that no pedestrian enters an active aisle during truck movement, supervisor observation at shift change and peak putaway with documented checks, and a short stand-down with the 12 drivers and 20 pickers in that zone focused on the four actual events, not generic content.

Verification is behavioral and physical, not documentary. Pick metrics that fail visibly:

• Pedestrian–vehicle proximity events in aisles 14–16 measured by four weeks of structured observation at randomized times (include nights and shift change), repeated after installation. Target at least a 50 percent reduction at 90 days.
• Percent of observed truck movements compliant with the 8 km/h limit, from telematics where available.
• Gate condition: percent of spot checks where gates are closed and not propped or tied open. Propped gates are the number-one real-world failure of this control and only unannounced checks find them.
• Near-miss reports geolocated to the zone, and time-to-close for traffic defects.
• Barrier damage reports from the weekly walk.

Report the pre-installation baseline alongside the post number so the comparison is honest, and re-observe at 30, 90, and 180 days because new controls decay.

To the state consultant: be complete and accurate. Present in this order - the hazard, the four events, the original risk assessment, your critique of it, the corrective package with dates and costs, the interim measures, and the verification plan. Say plainly that the prior assessment under-rated the hazard; consultants see that constantly and it builds credibility rather than costing it. Have the traffic management plan, training records, and near-miss log ready and organized. Do not speculate about causes you have not established, and do not remove or revise any document - preservation obligations apply from the moment an inspection is reasonably anticipated. If the consultant identifies an imminent-danger condition you have not addressed, fix or barricade it the same day and document it.