On this page · 9 sections
A safety audit can produce a precise percentage while leaving important questions unanswered. What was examined? Which evidence supported the ratings? Were any criteria excluded? Does the number describe paperwork, implementation or the performance of controls?
This article uses an illustrative incident scenario and a reconstructed scoring example to examine a proposed safety management system. The narrative does not describe an agency inspection or independently verified field outcomes.
Start with the question the audit is meant to answer
An audit needs a defined scope: the operation, criteria, time period, locations and work conditions examined. A review of written procedures answers a different question from observing whether a protective measure functions during the task. Record those boundaries before calculating a score.
OSHA’s program-evaluation guidance calls for checking whether a program is implemented as intended and effective in controlling identified hazards, with worker involvement. A score may help organize that review, but the underlying findings and evidence are needed to interpret it.
Separate an internal improvement review, a contractual assessment, a regulatory inspection and a recognition-program evaluation. Their criteria, authority and consequences differ. My original exercise borrowed headings from Voluntary Protection Programs materials; that does not make its scoring scheme an official VPP assessment. OSHA’s VPP process includes application review and an agency onsite evaluation.
Check the arithmetic before interpreting the result
The earlier version used three ratings: 1 for “not present,” 2 for “needs improvement” and 3 for “meets requirements.” It reported four element percentages. Rechecking the listed rows reveals a substantial discrepancy in the hazard-prevention-and-control element.
| Element | Listed point total / assumed maximum | Percentage of maximum points |
|---|---|---|
| Management leadership and employee involvement | 15 / 27 | 55.6% |
| Worksite analysis | 17 / 24 | 70.8% |
| Hazard prevention and control | 19 / 30, excluding the two dash-marked rows for this reconstruction only | 63.3%, conditional on that exclusion |
| Safety and health training | 13 / 15 | 86.7% |
The hazard-control section lists ten numerical ratings: 1 + 2 + 1 + 3 + 2 + 3 + 2 + 2 + 1 + 2 = 19. At three possible points per scored row, the assumed maximum is 30. The paper instead reported 7 out of 18 and 38%. Those figures do not follow from the displayed rows. Even 7 divided by 18 would be approximately 38.9%.
The table above uses one decimal place consistently. The other three rows retain the paper’s stated point totals and are now reported to one decimal place. The hazard-control discrepancy changes both the numerator and denominator. These are corrections to the arithmetic record, not validated ratings of a real workplace.
Distinguish not applicable from not assessed
The source table marks Process Safety Management and an occupational-healthcare-program row with dashes. It does not document why those entries were excluded. A dash could mean not applicable, not assessed or insufficient evidence. Those states should not be treated as interchangeable.
For the reconstruction above, I excluded the two dash-marked rows solely to make the numerical assumption visible. That does not establish that the exclusions were justified. Actual applicability needs a recorded basis tied to the operation and the criteria being used. An unexamined requirement remains an assessment gap.
For each excluded or unresolved item, record the reason, supporting information and reviewer. Explain how it affects the denominator and what remains outside the conclusion. A percentage can rise when difficult items disappear from the denominator even though the underlying work has not improved.
A points percentage is not a percentage of compliance
The original scale gives one point to a criterion that is not present. Consequently, ten absent criteria would receive 10 out of 30 points, or 33.3% of the maximum. That nonzero result follows from the scale’s starting value; it is not evidence that one-third of the required protections exist.
The three labels also do not establish equal distances between levels. Moving from “not present” to “needs improvement” may represent a different kind of change from moving to “meets requirements.” Adding the ratings creates a convenient index, but additional justification is needed before treating differences in that index as a measured amount of safety.
State exactly what the number represents: points awarded under a particular rubric and set of inclusion rules. Do not relabel it as the percentage of hazards controlled, probability of avoiding injury or percentage of legal compliance. Those are different quantities that the calculation did not measure.
Keep critical findings visible outside the aggregate
An overall score can combine strong results in one area with a consequential weakness elsewhere. Averaging element percentages gives each element equal influence; combining all item points gives elements with more scored rows greater influence. Neither choice establishes that hazards with serious consequences deserve less attention.
The illustrative scenario describes equipment-defect reports and gaps in maintenance follow-through alongside a comparatively high training score. The useful inquiry is whether reported defects reached someone with authority to respond and whether the equipment’s condition was appropriately resolved. A favourable training subtotal cannot answer that question.
For powered industrial trucks within its scope, 29 CFR 1910.178(q)(1) requires removal from service when a truck is not in safe operating condition and repairs by authorized personnel. An aggregate audit score cannot substitute for the required response to that condition. This article does not make a compliance finding about an actual employer.
Trace each finding from the document to the work
A defensible finding identifies its criterion, evidence, relevant conditions and the gap between expectation and observation. Distinguish what a document says from what people report and what the review directly observed. Different evidence sources can reveal different parts of the same process.
| Evidence | What it can contribute | Question still to examine |
|---|---|---|
| Written defect-reporting procedure | The intended reporting and response process. | Can workers use it, and does the report reach an accountable owner? |
| Completed maintenance record | A record of work reported as completed. | Does appropriate verification support restoration of the required function? |
| Training or attendance record | Evidence of documented participation or completion. | Does the evidence address the relevant task, learning and performance requirements? |
| Worker account or task observation | Information about how the process operates in the conditions examined. | What remains unverified, and do other shifts, tasks or conditions differ? |
Document sample selection and coverage. Reviewing one task, shift or record does not establish conditions everywhere. Record disagreements and missing information rather than converting uncertainty into a favourable score. Use suitable expertise and methods where technical performance needs measurement or inspection.
Turn findings into actions with verification
Give each finding an accountable owner, an appropriate response, resources, a target date and a method for checking the result. Address urgent conditions through the workplace’s prompt response arrangements. Administrative completion of an audit should not delay necessary protection.
OSHA’s hazard-prevention and control guidance includes checking whether controls remain effective. Define that check in relation to the original finding. Revising a procedure or assigning training may be an action, but the verification should examine whether the concern was addressed in the relevant work conditions.
Keep “implemented, awaiting verification” distinct from “verified closed.” If the response falls short, revise or reopen it and record why. The companion Safety Management System Action Plan explains how ownership, resources and management review connect. Employee Safety Suggestions provides a printable action record.
Use comparisons only when their basis is clear
Before comparing scores across time or sites, check whether the criteria, rating definitions, weights, exclusions and evidence coverage are comparable. A higher score may reflect a narrower scope, different interpretation or a changed denominator. A broader review may reveal previously unexamined gaps without proving that the work itself deteriorated.
Retain the rubric version and the evidence supporting each rating. Where reviewers disagree, examine the criteria and records rather than assuming that one total is more accurate. Report consequential findings and unresolved questions alongside the score so that the number supports decisions instead of concealing their basis.
This revision contributes a transparent reconstruction and questions for interpreting audit evidence. It does not validate the original rubric, certify compliance, confer VPP status or predict an injury outcome. The practical value lies in the traceable relationship between a criterion, a finding, the response and verification in actual work.
References and source record
- Occupational Safety and Health Administration. (n.d.). Program evaluation and improvement. Recommended Practices for Safety and Health Programs.
- Occupational Safety and Health Administration. (n.d.). Hazard prevention and control. Recommended Practices for Safety and Health Programs.
- Occupational Safety and Health Administration. (n.d.). All about VPP. Current program overview; the example’s point system is not presented as an official VPP scoring method.
- Occupational Safety and Health Administration. (n.d.). 29 CFR 1910.178: Powered industrial trucks, paragraph (q)(1).