JSA Risk Scores Diverge Across Shifts
A general contractor uses job safety analysis worksheets for 14 task categories across two shifts. The risk matrix has likelihood 1–5 and severity 1–5, with risk score = likelihood × severity. Cutoffs are 1–6 low, 8–12 medium, and 15–25 high. Auditors found the same “grinding metal” task scored 4 (low) on day shift and 16 (high) on night shift. Day shift cited “experienced workers”; night shift cited “fatigue and poor lighting.” There are no written definitions for likelihood anchors. Subcontractors submit JSAs in their own formats; 6 of 10 lack controls beyond PPE. The project has had two hand injuries in 90 days, both on night shift. The owner wants a single corporate risk score to compare all tasks.
1. Diagnose why the same task received radically different risk scores and what that means for the matrix’s validity.
Read the worked answer
The matrix is producing inconsistent results because the scoring anchors are undefined and the crews are rating different realities. Day shift appears to rate likelihood based on worker experience; night shift appears to include fatigue and lighting. Neither is necessarily wrong, but the matrix does not tell raters whether to assess the task as designed, as performed, or under worst credible conditions.
A score of 4 versus 16 for the same task means the matrix is not reliable for comparing tasks or shifts yet. The two hand injuries on night shift are a signal that night-shift conditions deserve specific attention, not that the day-shift score should be averaged with the night-shift score.
Before using scores for prioritization, define likelihood and severity anchors in observable terms, specify the exposure scenario being rated, and calibrate raters on common examples.
2. Recommend how to calibrate the risk assessment process and integrate contractor JSAs.
Read the worked answer
Rewrite the matrix with written anchors: what likelihood 1, 3, and 5 mean in terms of exposure frequency and controls; what severity 1, 3, and 5 mean in terms of injury potential. Require raters to state the scenario: normal operation, routine task, and credible upset. Run a calibration workshop using the grinding task and two other examples; have day and night supervisors score independently, then reconcile differences.
For contractor JSAs, issue a minimum template and review criteria: task steps, hazards, initial risk, controls using the hierarchy, residual risk, and responsible person. Reject JSAs whose only controls are PPE; require engineering or administrative controls where feasible. Review contractor JSAs before work starts and audit field use.
Use the scores for prioritization and verification, not as precise measurements. Track whether high-risk tasks receive verified controls and whether injury patterns change.
3. How would you respond to the owner’s request for a single corporate risk score to compare all tasks?
Read the worked answer
Push back on using a single score as an absolute ranking across dissimilar tasks. The matrix is ordinal: a 16 is not “twice as dangerous” as an 8, and scores from different crews, tasks, and assumptions are not directly comparable without calibration.
A better approach is a corporate risk profile: standardized scoring anchors, required scenario description, minimum control expectations by risk band, and review by a competent person. Compare tasks by risk band and by whether controls meet the hierarchy, not by raw numbers alone. If the owner needs enterprise comparison, use a secondary prioritization method that considers exposure, severity potential, regulatory requirements, and control reliability. Keep the raw score as a conversation tool, not a false precision metric.