A 96% Audit Score and a Nine-Foot Fall
You are corporate safety manager for a general contractor running four projects with 85 to 240 workers each. Each quarter a superintendent and a safety coordinator walk the site with a 250-line checklist, announced a week ahead, and follow a route the coordinator emails out in advance. Scores have run 88 to 94 percent for two years and the executive dashboard shows an improving trend. Last month a carpenter fell nine feet through a floor opening on Project C and broke a wrist. The opening had been covered, but the electrical crew removed the plywood and did not replace it. The audit three weeks earlier scored that area at 96 percent. Your COO's response is to add 40 more checklist items and require monthly instead of quarterly audits. Superintendents call the audits a paperwork drill and say they know exactly which areas will be walked.
1. Explain how a 96 percent score coexisted with a serious injury. What is the audit actually measuring?
Read the worked answer
A 96 percent score and a fall through an opening are not contradictory - they are what this audit design predicts.
- It measures conditions at one moment. A plywood cover removed by an electrical crew three days after the walk is invisible to the audit and to the score. The hazard was dynamic; the checklist is a snapshot.
- It is announced with a pre-shared route, so the site is staged. That tests preparation, not control.
- The percentage is computed over 250 items of unequal risk. Twelve missing extinguisher tags and one unsecured opening both count as items, so the arithmetic rewards trivia and dilutes the findings that kill people.
- Auditors are known to the superintendents and depend on them operationally; scoring drifts generous.
- It verifies neither behavior nor the system that governs covers: no permit, no attendant, no restoration sign-off, no record of who removed it.
- There is no closure loop, so repeat findings reappear each quarter with no consequence.
The 'improving trend' on the dashboard is therefore a measure of audit behavior rather than of risk. Treat the wrist fracture as a finding about your process, not about one electrician's carelessness.
2. Redesign the inspection and audit program. What changes, and what new measures go on the dashboard?
Read the worked answer
Rebuild on four principles: unannounced, risk-weighted, systemic, and closed-loop.
Sampling: unannounced or at most 24-hour notice; randomly select areas and crews from a list each month so no one can predict the route; vary timing across shift start, after lunch, and occasional night shifts - the times controls lapse.
Content: split into two instruments. One is a short high-consequence checklist - falls, electrical, excavation, cranes, confined space, lockout/tagout, mobile equipment - scored with severity weights so a single unsecured opening cannot be offset by forty compliant items. The other is a task observation of live work with craft-worker interviews ('what do you do when you find an open hole?'), because behavior reveals the system.
Cover control: require a physical control for every opening - permit or tag, barricade, attendant where needed, and restoration sign-off - and audit it every time.
Closure: every finding gets an owner, a due date, and a verification inspection. Report percent closed on time and repeat-finding rate by trade and location.
Calibration: rotate auditors across projects and periodically audit the same area with two teams to check agreement. Publish findings company-wide so an electrical crew's cover habit is corrected everywhere, not only on Project C.
Dashboard: replace the percentage score with findings per inspection, repeat rate, closure time, and worker-reported hazards.
3. How do you respond to the COO's proposal to add 40 items and go monthly?
Read the worked answer
Push back with data, not opinion.
The problem is discrimination, not coverage. Forty more items lower the signal-to-noise ratio further and make the score easier to game; monthly announced audits simply produce twelve prepared snapshots a year. What changes risk is unannounced sampling of high-consequence work plus verified closure of findings.
Proposal to the COO:
- Keep quarterly independent audits but make them unannounced, randomized, and severity-weighted; add weekly supervisor inspections on a short hazard-focused card.
- Add two metrics to the dashboard: percentage of findings verified closed by the due date, and repeat-finding rate - both reported by project and by trade.
- Add the specific corrective action from this injury: a cover-and-opening control procedure with permits, tagging, and a named person accountable for restoration, plus training for every trade that removes covers.
- Pilot the new format on Project C for one quarter, then compare findings per inspection, repeat rate, closure time, and near-miss reports. Bring that comparison back.
Give the superintendents something in exchange: cut the 250-line checklist to roughly 30 risk-ranked items so a walk takes 30 minutes instead of three hours. Their complaint that audits are a paperwork drill is correct - the fix is a shorter, sharper instrument they cannot predict and cannot pass by staging.