Practical defect classification tips for site audits

Use this thread as a working standard for how we log defects in the field. The goal is simple: every inspector on site should classify the same issue the same way, so the log is usable for close-out, trending, and follow-up.

Reply with a recent defect you logged and how you classified it. If you would classify it differently after reading this, say why.

1. Classify the defect before you write the description

Do not start with a long narrative. Decide four things first, then write:

What failed (object / location / system)
How it failed (missing, damaged, incorrect, expired, blocked, not in use)
Who is exposed (people, plant, environment, process)
How soon it can go wrong (now, this shift, this week, later)

Those four answers drive severity, priority, and the close-out owner. If you cannot answer them, you do not have enough evidence yet. Take another photo, ask the supervisor, or check the last inspection.

2. Use a 4-level severity scale, not a gut feeling

Apply the same scale on every audit. If your site already has a matrix, map it to these four levels so the checklist and the defect log stay aligned.

Critical (stop / isolate now)

Immediate risk of fatality, serious injury, major fire/explosion, or uncontrolled release
Control is missing or defeated, and work is happening (or about to happen) in the danger zone
Examples: live electrical work with no isolation, open excavation with no barrier on a walkway, fire exit locked during occupancy, lifting under a suspended load with people in the drop zone

Action: stop the task or isolate the area, notify the area owner, log the defect, stay until a temporary control is in place.

High (fix this shift / before restart)

Serious injury or significant incident is reasonably foreseeable if the job continues
A required control exists on paper but is not effective in the field
Examples: damaged scaffold board still in use, missing machine guard with the line running, expired gas detector used for confined space entry, blocked spill kit next to a chemical transfer

Action: do not leave it as “for later.” Assign an owner and a same-shift due time.

Medium (plan and close within the agreed window)

Injury or damage is possible, but a working control still reduces the likelihood
Housekeeping, incomplete signage, worn PPE, missing secondary containment with low inventory, incomplete permits that are not currently being worked
Examples: faded traffic markings in a yard with a banksman present, cracked visor still being used with a spare available, temporary cable across a walkway with a cover that is coming loose

Action: log it with a due date that matches site rules (often 7 days). Do not let medium defects pile up until they become high.

Low (record, trend, and correct in routine maintenance)

Non-conformance with little immediate exposure
Cosmetic damage, missing labels on unused equipment, incomplete paperwork that does not affect today’s work
Examples: faded inspection tag on a spare extinguisher in stores, minor rust on a unused handrail, missing name on a completed toolbox-talk sheet

Action: still log it. Low items are how you catch a pattern before it becomes a real incident.

If you are between two levels, go up one level. Under-classifying is how serious issues disappear in a long list of “mediums.”

3. Separate severity from priority

These are not the same thing.

Severity = how bad the outcome could be if the defect stays
Priority / due date = how fast the site can actually close it, given access, parts, and shutdowns

A missing handrail on an active platform is high severity and high priority. The same missing handrail on a platform that is locked out until next month can still be high severity, but the due date can match the restart, as long as the isolation stays in place and is recorded.

Never drop the severity just because the fix will take time. Record the temporary control in the log (barrier, lock, tag, watchman, removal from service).

4. Write defects so a stranger can close them tomorrow

A good log entry answers: where, what, why it matters, what “fixed” looks like.

Weak: “Housekeeping poor in workshop.”

Usable: “Bay 3, west wall: oil-soaked rags in open drum next to hot-work bench. Fire load + ignition source. Remove rags to lidded metal bin and keep 5 m from the bench. Photo attached.”

Use this line pattern:

[Area / asset ID] + [defect] + [exposure] + [required end state] + [temporary control if any]

Include:

Exact location (building, level, grid, equipment tag)
Photo of the defect and, if useful, a photo of the wider context
Asset ID / permit number / last inspection date if known
Who you told on the spot (name and role)

Do not put five issues in one line. Split them. One defect, one owner, one close-out.

5. Classify the type as well as the severity

Type is what you trend. Severity is what you escalate. Log both.

Useful type buckets for site audits:

Missing control (guard, barrier, extinguisher, sign, isolation)
Failed / damaged control (torn, blocked, bypassed, expired)
Wrong control (wrong glove, wrong extinguisher class, wrong LOTO point)
Behaviour / use (control present but not used)
Documentation / competence (permit incomplete, person not authorized)
Condition / housekeeping (slip, trip, accumulation, access)

If the same type keeps appearing in one area, the issue is not the last inspector. It is the system. Raise that in this thread with counts, not opinions.

6. Field workflow that keeps classification consistent

Use the same sequence on every walk:

Scope the area — what work is live, who is in it, what permits are open
Walk the path of the people, not only the equipment (access, egress, muster, lifting radius)
Check the control that should be there, not only the obvious damage
Classify using the four-level scale before you leave the spot
Tell the area owner immediately for Critical and High
Log before you leave the area so details do not get rewritten from memory
Confirm the temporary control is actually in place (photo after, not only before)
Do not close from a verbal “we’ll sort it” — close only when the end state is verified

If two inspectors would argue about the level, write the facts in the log and agree the level here. The discussion belongs in this forum. The field log should stay factual.

7. Common misclassification traps

Calling a bypassed interlock “medium” because nobody got hurt today. That is High or Critical. Outcome luck is not a control.
Logging “PPE non-compliance” as Low when the task is grinding, chemical handling, or work at height. Match the exposure, not the paperwork gap.
Bundling a whole area into one defect. You cannot assign or trend that.
Closing a Critical because a cone was put down. Temporary control reduces immediate risk. It does not close the defect.
Copying last week’s wording. If the photo does not match the text, rewrite the text.
Using “N/A” on a checklist item you did not actually check. If you could not access it, log “not inspected — no access” instead of passing it.

8. What “good” looks like on a close-out

A defect is closed when:

The required end state is in place (not a promise)
A photo or inspection note shows the fix
Any temporary control has been removed or replaced by the permanent one
The owner and verifier are named
Recurrence is considered (same asset, same type, last 30 days)

If it failed again, reopen it or raise a new defect linked to the old one. Do not hide repeats inside a closed item.

---

Your turn. Drop one real example from a recent site audit:

Location and asset
How you classified it (severity + type)
Temporary control you used
What made you hesitate between two levels

We will use the replies to tighten this standard so INSPECTRA inspectors grade the same way on every site.