Workplace Safety Audits: A Walkthrough That Finds Real Hazards

A workplace safety audit is a systematic, documented review of whether a facility's safety program actually works: hazards found and controlled, procedures followed, records complete. It differs from an inspection, which checks physical conditions on a given day. Audits examine the system; inspections examine the floor.

Plants that treat the two as the same thing end up with a stack of green checkmarks and the same injuries every year. This guide draws the audit/inspection line, gives an area-by-area walkthrough that finds real hazards instead of confirming the ones you already fixed, and includes a printable checklist you can take to the floor.

What's the difference between a safety audit and a safety inspection?

An inspection asks "is this guard on the machine today?" An audit asks "what is our system for making sure guards stay on machines, and is it working?" Both matter, and they run on different rhythms.

A good audit uses inspection evidence. If three consecutive monthly inspections flagged the same blocked exit, the audit finding isn't "blocked exit", it's "corrective actions don't close."

How do you run a safety audit? The seven-step cycle

1. Scope it. Pick the areas, shifts, and programs the audit covers. Include at least one off-hours shift, the floor at 2 a.m. is a different plant.
2. Gather the paper first. Injury and illness logs, near-miss reports training records, lockout/tagout procedures and annual inspection records, previous audit findings. Gaps in records are findings before you take a step.
3. Walk the floor area by area with the checklist below, and talk to operators. The question "what's the most dangerous thing about this job?" finds hazards no checklist will.
4. Watch work actually being done. Compare what you see against the written procedure and the job safety analysis for that task. Every gap is either a training issue or a procedure that doesn't survive contact with reality.
5. Rate each finding for risk using a severity-by-likelihood matrix, so the two-day fixes and the two-hour fixes get sequenced honestly.
6. Assign corrective actions with owners and dates. A finding without an owner is an observation, not a corrective action.
7. Verify closure and feed the next cycle. Re-check completed actions on the floor, not in the spreadsheet. Carry unresolved items forward visibly.

What should you check, area by area?

These seven areas cover most of what hurts people in a plant. The printable checklist below breaks each into specific line items.

• Housekeeping. Clear aisles, dry floors, stable stacking, no accumulation of combustible dust or scrap. Housekeeping findings predict everything else; a plant that tolerates clutter tolerates worse.
• Electrical. Panel access clear, covers on, no daisy-chained power strips or damaged cords, extension cords not serving as permanent wiring.
• Machine guarding. Guards in place and functional, interlocks not bypassed, point of operation protected, the full review in our machine guarding guide.
• PPE. Right equipment for the hazard assessment, actually worn, in serviceable condition, and stocked, an empty glove bin is a finding.
• Emergency egress. Exits marked, lit, unlocked, and unblocked; extinguishers charged, mounted, inspected; alarm audible over production noise.
• Chemicals. Containers labeled, safety data sheets accessible to the shift that uses them, incompatibles separated, eyewash stations reachable within seconds and flushed on schedule.
• Ergonomics. Repetitive lifts above shoulder or below knee, twisting under load, vibration exposure, workstation heights, the slow injuries that dominate days-away counts.

How do you rank what you find?

Rate every finding on two axes: how bad the credible worst outcome is, and how likely it is at current exposure. The matrix keeps the loud finding from outranking the lethal one.

What do the enforcement numbers say you'll find?

OSHA's citation data is effectively a national audit-findings list, and it's stable year over year:

• Fall protection was OSHA's most-cited standard for the 14th straight year in FY 2024 followed by hazard communication and ladders (OSHA Top 10 Most Frequently Cited Standards).
• Machine guarding (1910.212) drew 1,541 citations in FY 2024, still in the top 10 despite being one of the oldest rules on the books.
• Private industry recorded 2.3 recordable cases per 100 full-time workers in 2024 about 2.5 million cases (BLS), and every one of them was, in principle, an audit finding nobody wrote down in time.

If your audit never finds hazard-communication, guarding, or egress issues, the likelier explanation is the audit, not the plant.

What happens after the audit decides everything

The walkthrough is maybe a third of the work. The value is in what survives it: findings that become corrective actions, actions that close, and closures someone verified on the floor. This is where paper programs die, the binder gets filed, the follow-ups scatter into email, and next year's audit finds last year's findings. Plants that run findings through the same digital capture they use for production checks, photos at the point of finding, owners and due dates assigned on the spot, open items visible on a dashboard, close the loop dramatically faster. That's the connected worker pattern, and it's how Harmony treats safety data generally: captured where the work happens, visible to whoever owns the fix, with follow-ups that don't rely on memory (see how it works). Pair the audit cycle with regular toolbox talks to feed findings back to the crews who reported them, people stop reporting into systems that never answer.