Pallet racking collapse: separating a manager's theory from evidence
At 06:40 on a Tuesday, a two-bay section of adjustable pallet racking collapsed in a distribution warehouse, spilling 14 pallets of canned goods into the main aisle. No one was hurt, but a forklift was struck by a falling pallet and its overhead guard was dented. The truck was being reversed out of the aisle by an agency driver with nine years' experience who had worked on the site for five weeks. A third-party rack inspection eleven months earlier had recorded one upright in that bay as 'damaged - monitor'; there is no record of any repair. Each pallet weighed 1.1 t and the beam level was rated at 1.0 t. Rack protection barriers are fitted at the aisle ends but not at the corner where the truck was turning. The warehouse manager told you at the scene that the driver 'must have been speeding again'. Site CCTV covers the aisle entrance only, and the system overwrites every 21 days. The night shift had handed over 40 minutes earlier.
1. Set out the first 24 hours of your investigation: what you secure or collect, in what order, and why that order matters here.
Read the worked answer
Order is driven by how quickly each type of evidence is lost, not by what seems most interesting.
- Make safe and preserve the scene (hours 0–2). Barricade the aisle, and photograph and sketch the collapse as found before any recovery - upright deformation, beam-end connector damage, baseplate and floor-anchor condition, and the actual position of the truck. Recovery destroys the geometry that tells you whether the frame failed first or was struck first. Tag and quarantine the damaged uprights and beams; do not let them go for scrap.
- Capture perishable records (same day). Export the CCTV now and record the time-sync offset - with a 21-day overwrite you cannot assume it will still exist next week, and the camera only covers the aisle entrance, so it may show approach speed but not the impact. Pull the inspection report, the repair/maintenance log, load signage records, training matrices and any FLT telematics or speed data. Ask for documents in their native state before anyone 'tidies' them.
- Interview witnesses early, separately. The driver first - for facts about load, visibility, aisle width and task, and with an explicit statement that the purpose is learning, not discipline. Then the night-shift picker, the FLT examiner if there is one, and the manager. Interview separately so accounts are not pooled.
- Record the manager's remark as a hypothesis, not a finding. Write it down verbatim and test it against telematics and the physical damage; do not let it steer the evidence collection.
The reasoning: physical condition and CCTV are irreplaceable, memory decays and is contaminated by the first account someone hears, and documents can be altered innocently. Anything collected out of that sequence risks being unusable.
2. The operations director wants the report to conclude 'driver error - speeding'. Using the facts given, explain why that conclusion would be inadequate and how you would build and test alternative causal factors.
Read the worked answer
'Driver error' is a description of the last event in a chain, not an explanation of the collapse - and here several facts point elsewhere.
Why the single-cause conclusion fails
- A truck travelling too fast in an aisle is a normal, foreseeable event; racking is designed and protected on that basis. Speed alone does not bring down a sound frame.
- The bay was loaded at 1.1 t against a 1.0 t beam-level rating. Overloading is a management decision about stock profile and signage, not a driver decision.
- An upright was flagged 'damaged - monitor' eleven months earlier with no repair record. That is a failed inspection-to-repair loop and a latent weakness that predates this driver's employment.
- Protection barriers stop short of the corner where the turn occurs - a layout/engineering gap.
- Handover occurred 40 minutes before the collapse; check whether the bay was loaded or re-configured during handover without a check.
How to test hypotheses rather than assert them List competing explanations - impact, overloading, pre-existing frame damage, incorrect beam pitch, floor/anchor failure, mixed pallet weights - and for each state what evidence would confirm or eliminate it: beam-end connector deformation pattern; whether damage is fresh or has paint/rust history; pallet weight tickets; whether the collapsed upright is the one flagged in the inspection; telematics speed against aisle limit. Where evidence is inconclusive, say so and state confidence.
Framing Report the driver's conduct only in terms of the conditions that shaped it (aisle width, sightlines, time pressure, agency induction). A conclusion of 'speeding' closes the investigation and leaves the overloading, inspection backlog and barrier gap in place - so the same collapse recurs with a different driver.
3. Recommend remedial actions and explain how you would test whether they have actually worked, rather than assuming they have.
Read the worked answer
Actions, ordered by strength of control
- Engineering: replace damaged uprights and record it; extend rack protection to the corner; post load-rating signage at eye level on every bay; verify that the stock profile (1.1 t pallets) matches the beam rating or re-configure the bay.
- System: reinstate a named weekly visual rack inspection with a log, and a closed-loop process for third-party inspection findings, with a defined maximum time to close an 'amber' item.
- Organisational: include agency workers in rack awareness induction, with a competence check rather than a signature; review handover so load changes are recorded.
Testing effectiveness Do not accept 'training delivered' as evidence. Set a baseline now and re-measure:
- Leading indicators: percentage of inspection actions closed within the target time; number of unprotected corners remaining; overspeed events per 100 truck-hours from telematics; percentage of agency staff who pass an observed competence check.
- Lagging indicators: rack damage reports per month; near-miss reports involving racking; structural repairs completed.
- Direct verification: an unannounced walk-round at 90 days checking signage, barriers and the inspection log against the register; a sample of five bays weighed against their rating.
Define success criteria in advance (for example, all amber items closed within 30 days; zero unprotected corners; overspeed events halved) and a review date. If the leading indicators have not moved, the actions were not implemented as intended and the investigation's conclusions need revisiting - this is how you detect the common failure where actions are recorded as complete but change nothing on the floor.