Sample report

Real findings, from real documents, redacted

Every finding on this page came out of an actual review of an actual submitted document. Client names, site addresses, personnel, contact details, licence and project numbers have been removed. The engineering content — severity, source citation, the finding itself — is unchanged.

Why these are real and not invented

If you are weighing this up, the question you actually have is would it have caught something in my study? A made-up example cannot answer that. So these are redacted real ones — including the ones where the answer was uncomfortable. One study in the set states, in its own words, that its results and labels are not valid. Another reports an arcing current higher than the bolted current it derives from. Both were sealed, issued documents in service.

What a finding contains

Every finding in every report carries the same elements, so a report is auditable rather than a wall of prose.

Checklist ID: The published check that produced it. Findings never appear from nowhere.
Severity: Critical, significant, observation or information — the consequence.
Confidence: High, medium or low — how certain the review is. Set independently of severity.
Category: The checklist group, so related findings read together.
Source: Where in your document it was found: page, sheet, appendix, table column, note letter or step number.
Code reference: Whether a code section is cited, and where that text came from.
Recommended action: What to do. Never a redesign — this review does not produce corrected values.

Two rules that shape every report

A code section is cited only if you supplied its text. If you upload a code document, findings cite it. If you do not, the review still runs but reports findings as general engineering observations and cites no article number. It will not quote a code section from memory. In the test run this meant a handful of legitimate questions were raised and then deliberately dropped rather than answered from recall — and the checks that compare your drawing set against itself carried the report instead.

Absence of a finding is not a clearance. These reviews report what a document shows and what it fails to show. None of them approves anything, and two of them are built so they cannot: the coordination review reaches no selectivity verdict, and the MOP review has no words for “adequate” or “approved” in its vocabulary at all.

Findings by product

18 shown · 14 critical · 4 significant

AI Arc Flash Study Review

Six studies went through this engine in the test run — a bank data centre, a combustion-turbine plant, a wind facility, a hotel tower, an energy centre and a colocation site. They produced 114 findings between them.

Download the full redacted report (PDF) — Data centre arc flash study, 50 pages, 26 findings

AF-ELEC-03

high confidenceCritical

Single electrode configuration applied to every bus in the facility

Every one of the ~90 rows in the incident energy report carries Electrode Config = VCB. The same facility contains 480 V service switchboards, automatic transfer switches, generator skid breakers, UPS modules, static transfer switches, PDUs, 208 V branch panelboards and dry-type transformer secondaries. Those are not one electrode arrangement. The uniformity is the signature of a global software default that was never revisited per bus.

Source: Source: pp. 42–44, Appendix F Incident Energy Report, Electrode Config column
Code ref: not available — reported as a general engineering observation
Action: Assign electrode configuration per bus from actual equipment construction and re-run. This input changes incident energy directly.

One column, ninety rows, one value. It takes a reader about four seconds to check and it changes every number in the study.

AF-RES-01

high confidenceCritical

Arcing fault current reported higher than bolted fault current at a 0.23 kV bus

One 0.23 kV row reports Bus Bolted Fault = 0.496 kA and Bus Arc Fault = 0.538 kA. Arcing current cannot exceed the bolted fault current at the same location. The row that follows reports 1.708 kA bolted / 1.268 kA arcing, which is the expected ordering, so the anomaly is confined rather than systematic. The result derived from the anomalous row is 2.7 cal/cm² with a 31.5 in boundary.

Source: Source: Attachment A — Incident Energy Tabulation, 0.23 kV row
Code ref: not available — reported as a general engineering observation
Action: Re-derive that row. Either the bolted value or the arcing value is wrong, and the incident energy and boundary printed for that bus follow from it.

A physically impossible pair of numbers inside a sealed, issued study — and the label on that equipment was printed from it.

AF-STATE-01

high confidenceCritical

The study states its own results and labels are not valid

Results and Recommendations, verbatim: "the incident energy levels shown in both this report and on the arc flash labels are not valid until the recommended protective device setting changes shown in this report are implemented." The settings report header confirms the case: Scenario: RECOMMENDED. This is a submittal for approval carrying a complete incident energy table and a complete label set, every value of which describes a system state that did not exist when the study was issued and may never have been created.

Source: Source: Results and Recommendations section; Protective Device Setting Report header
Code ref: not available — reported as a general engineering observation
Action: Confirm in writing that the recommended settings were implemented, by whom and on what date, before any label from this study is applied.

The study says this itself, in one sentence, in the middle of a section most readers skim. Six years later the labels are still on the gear.

AF-TIME-03

high confidenceCritical

A 37-second trip time is truncated to a 2-second arc time

The 0.23 kV auxiliary rows report Trip Time = 37.087 s and Trip Time = 6.681 s, with Arc Time set to 2 in both cases. The stated basis is that arcing time is limited to two seconds. A two-second cap is an assumption that a worker can move out of the arc within two seconds. The study applies it without evaluating the specific work position, which in this case is inside a wind turbine nacelle or tower base, where egress is constrained by definition.

Source: Source: §4 Basis of Analysis, item 3; Attachment A, 0.23 kV rows
Code ref: not available — reported as a general engineering observation
Action: Evaluate egress at the specific work positions. If a worker cannot clear the boundary in two seconds, the cap does not apply and the uncapped result governs.

The gap between a 37-second device response and a 2-second reported exposure is the largest single assumption in that study, and it is one line of text.

AF-RES-04

high confidenceCritical

Executive summary undercounts the buses above 40 cal/cm² and omits one by name

The summary names three locations above 40 cal/cm². The evaluation table shows four: 100 cal/cm², 73 cal/cm², 71 cal/cm² and 61 cal/cm². The 71 cal/cm² bus — the third highest in the facility — is not mentioned anywhere in the summary. A reader working from the summary alone would not know it exists. The summary also justifies the named locations as being immediately downstream of the existing transformer; the unnamed one is fed from a different device, so that justification does not obviously extend to it.

Source: Source: §1.3.3 Arc Flash Results, p. 4; Appendix C Arc Flash Evaluation table, rows 1–4, p. 49
Code ref: not available — reported as a general engineering observation
Action: Correct the summary to name all four buses and state the disposition of the omitted one.

This is the kind of gap that survives into an O&M manual and then into a work permit.

AF-GEOM-03

high confidenceCritical

Incident energy is reported without working distance in every results table

None of the arc flash result attachments carries a working distance column. An incident energy without the distance it was calculated at cannot be compared against a PPE arc rating or transferred to a label. The narrative asserts that the labels carry working distance; the calculation output does not.

Source: Source: §10 Attachments 6–9, results table column set
Code ref: not available — reported as a general engineering observation
Action: Add working distance per row to the results tables so each energy value is usable and auditable.

A cal/cm² figure with no distance beside it is not a result. It is a number.

Label & PPE Boundary Review

Two label sets: 99 labels from a data centre and 5 from a PDU lineup. Neither came with the study behind it, which is why the highest-value check in this product — study versus label — could not run on either.

Download the full redacted report (PDF) — 99-label data centre set, 12 findings

LB-PPE-02

high confidenceCritical

42 labels above 40 cal/cm² instruct the worker that appropriate PPE is required

Forty-two of the ninety-nine labels report incident energy above 40 cal/cm², topping out at 131.5 cal/cm². Every one carries the same body text as the low-energy labels: "Appropriate PPE Required." At these energies commercially available protective equipment and normal work practices are not a viable protection strategy. The DANGER signal word is present, which is a partial mitigation, but the instruction line contradicts it: the signal word says stop and the body text says gear up.

Source: Source: 42 of 99 labels; highest at 131.5 cal/cm², 480 VAC, 340 in boundary
Code ref: not available — reported as a general engineering observation
Action: Replace the body text on the above-threshold labels with an instruction appropriate to the hazard, and refer these buses back to the study for hazard-elimination review.

Nearly half the label set, on the highest-energy equipment in the building, tells a worker to select PPE for a condition where PPE selection is not the answer.

LB-CONT-05

high confidenceSignificant

One equipment identifier uses a different site prefix from all 98 others

Ninety-eight labels use one site prefix. One uses a prefix differing by a single character. Its position in the sequence confirms it belongs to the same equipment family. A one-character difference in an equipment tag breaks the link between the label, the study row and the one-line entry for that equipment, and it will not be found by anyone searching for the correct tag.

Source: Source: p. 24, left column, Location field, adjacent to the correctly-prefixed sibling label
Code ref: not available — reported as a general engineering observation
Action: Correct the label and confirm the corresponding study row and one-line tag use the right prefix.

One character, in a set of ninety-nine, found by comparing every identifier against its siblings.

LB-CONS-01

high confidenceSignificant

Each label carries two contradictory PPE instructions

Every label in the set carries both "Appropriate PPE Required" in the header block and "PPE — No Arc-rated PPE Required" in the data block. A worker is told PPE is required and told no PPE is required, in the same eight square inches. The two are reconcilable to someone who knows the first refers to the hazard generally and the second to arc rating specifically, but the labels do not draw that distinction — and the same label specifies a glove class, which is a third thing to reconcile unaided.

Source: Source: all five labels in the set, header block and PPE data line
Code ref: not available — reported as a general engineering observation
Action: Reword so the two statements cannot be read as contradictory — scope the "no arc-rated PPE" statement explicitly to the arc flash hazard.

The underlying numbers were fine. The presentation was the defect, and it was on every label.

AI NEC Pre-Check

Two drawing sets, both reviewed without a code document. That puts this engine in degrade mode, where it will not cite an article it has not been given — see the note below on what that costs and what still works.

Download the full redacted report (PDF) — Issue-for-construction set, 16 sheets, 18 findings

NEC-SET-01

high confidenceCritical

Three sheets disagree on how many grounding conductors a parallel feeder gets

The same feeder is described three ways in one set. A general note reads "provide … a #3AWG ground conductor" (singular). The note beside it reads "each conduit shall have … a dedicated ground conductor" (two). The grounding detail sheet reads "provide one (1) full-size #3 Cu equipment grounding conductor in each of the two parallel conduits" and adds that the EGC shall not be divided across parallel raceways. Two of the three agree; the first, read literally, specifies one. An installer working from the sheet carrying the feeder notes could pull a single EGC in one of two parallel raceways.

Source: Source: sheet E102 general notes F and G; sheet E600 key notes A and B
Code ref: not available — reported as a general engineering observation
Action: Reword the feeder note to match the grounding detail. A one-line edit removes the only genuinely ambiguous instruction in the set.

The design intent was correct and stated correctly on one sheet. The sheet an installer actually works from said something else.

NEC-SET-09

high confidenceSignificant

Every title block field in the package is blank

Customer name, project name, PO number, material number, drawn-by, checked-by, title and drawing number are all empty across the package. Only the revision number and internal sales order numbers are populated. A drawing with no number cannot be referenced or superseded; one with no checker has no reviewer of record. The sheets carry six revisions plus a set of hand markups on top, which makes the absence of drawing control worse rather than merely untidy: there is no way to establish that the sheets in hand are the current ones.

Source: Source: title block on every sheet — all identification fields empty, revision field populated
Code ref: not available — reported as a general engineering observation
Action: Populate the title blocks before these sheets are used for anything. If the package was extracted from a controlled set, transmit the controlled version.

Needs no code text to find, and it is the finding that makes every other finding in that package hard to act on.

NEC-SET-11

high confidenceSignificant

Set issued for construction while its own governing input is deferred

The load calculation sheet gives available fault current at eight buses and marks each equipment adequacy determination PASS. A note on the same sheet states that updated fault current calculations, incident energy and protection settings will be provided prior to construction. The eight PASS determinations are computed from values the sheet says are not final, on a set stamped ISSUE FOR CONSTRUCTION.

Source: Source: sheet E010, General Notes — Available Fault Current, notes 1 and 2, alongside items A–G and the revision block
Code ref: not available — reported as a general engineering observation
Action: Mark the PASS determinations as provisional pending the updated study, and identify who reconciles them and when.

Not necessarily wrong — deferring a study to a construction submittal is common. But nothing on the sheet says the conclusions are provisional.

Coordination & Selectivity Review

Two studies. This engine reaches no selectivity verdict — it flags device pairs whose declared settings need checking against manufacturer curves, and says so in every finding.

Download the full redacted report (PDF) — Coordination study, 52 pages, 16 findings

CO-OVL-03

medium confidenceCritical

Instantaneous enabled on both devices of a series pair at the same pickup

Upstream and downstream devices both have instantaneous enabled at a declared pickup of 8000 A. Available fault current at that bus is 29,154 A — roughly 3.6 times the pickup. A fault anywhere on the load side at a current the bus can supply is above both instantaneous pickups. Instantaneous overlap is the most common cause of lost selectivity and is frequently invisible in a summary table, which is what happened here: the study narrative does not mention it. This pair requires verification against manufacturer curve data.

Source: Source: Appendix B series-breaker TCC settings blocks, p. 28; one-line bus fault current 29,153.67 A
Code ref: not available — reported as a general engineering observation
Action: Verify this pair first against the manufacturer curves at the available fault current.

The study named one coordination concern. This engine flags four more in the same appendix the narrative points the reader to.

CO-RAT-01

high confidenceCritical

Adequacy conclusion covers devices whose ratings were assumed, not observed

The executive summary states the equipment and breakers reviewed are properly rated for the calculated available fault current. The assumptions section states that panels without breaker information were assigned industry-minimum ratings, that unfused disconnects were assumed at a minimum value, and that fused disconnects which "could not be safely reviewed" were assumed to contain Class R fuses at 100 kA. Three categories of device carry an assumed rating, and the study does not identify which entries are observed and which are assumed.

Source: Source: p. 1 Executive Summary; p. 4 Assumptions and Clarifications; Appendix C
Code ref: not available — reported as a general engineering observation
Action: Tag every device whose rating was assumed rather than observed, and bound the adequacy conclusion to verified devices.

An assumed 100 kA on a disconnect nobody could safely open is an assumption in the permissive direction, on exactly the equipment least likely to have been checked.

MOP Review

One draft method of procedure with its drawings. This engine cross-checks the procedure against the drawing and against itself, and asks seven fixed questions on every submission that no document review can answer.

Download the full redacted report (PDF) — Draft MOP, 35 steps, 36 findings

MOP-BO-02

high confidenceCritical

The back-out plan does not reverse most of what the forward sequence does

Comparing every forward state change against the back-out section: fourteen distinct changes are made by the procedure — conductors demolished, a breaker relabelled spare, a panel fabricated and mounted, a fuse block and a new breaker installed, six new conductor runs pulled, fuses replaced. The back-out addresses wiring restoration and one breaker. It does not address the fabricated assembly, the new fuse block, the new breaker, the fuse change or the relabelling. The wiring it does address was demolished rather than disconnected, so restoration requires new conductor the back-out does not call for.

Source: Source: §3 steps 3.01–3.15 (forward state changes) against §5 steps 5.01–5.07 (back-out)
Code ref: not available — reported as a general engineering observation
Action: Rewrite the back-out as a numbered device-level reversal with expected results, covering every forward change. Confirm replacement conductor is on site.

Every MOP has a back-out section. This check reads it against the forward sequence line by line, which is what nobody does at 2 a.m.

MOP-XREF-01

high confidenceCritical

The procedure and the drawing disagree on which section the 4000 A utility breaker is in

The equipment summary places the utility breaker in section A2. The drawing section schedule places the utility incomer in A1 and the generator incomer in A2. A step instructs the crew to disconnect a conductor on the utility breaker, and a later step instructs them to replace it. A crew working from the procedure would go to the section the drawing identifies as the generator incomer. Both sources cannot be right, and this review does not resolve which is.

Source: Source: MOP Notable Equipment Summary; drawing section schedule rows A1 and A2
Code ref: not available — reported as a general engineering observation
Action: Resolve the section reference before this MOP is used. Confirm at the equipment, not from either document.

Two documents, both authoritative-looking, disagreeing about which cubicle a 4000 A breaker is in.

MOP-XREF-08

high confidenceCritical

Four conductors are demolished with no stated function and no impact assessment

A demolition step removes eight designated conductors. Four are re-installed by later steps. Four are cut out and never mentioned again — nothing says what they fed, what loses its supply, or why removing them is correct. Two of the four appear exactly once in the whole document: in the demolition list. The procedure has no impact assessment section, so there is nowhere else this could have been captured.

Source: Source: step 3.02 demolition list, traced against steps 3.03–3.15, §4 and §5
Code ref: not available — reported as a general engineering observation
Action: State what each removed conductor serves and confirm the removal is intended. If any is a live circuit to something not in scope, this is the finding that catches it.

Found by tracing every designator in the demolition list through the rest of the document and noticing which ones never come back.

MOP-STEP-04

high confidenceCritical

The document defines a Safe Stop Point notation and never uses it

The footer legend defines "SSP = Safe Stop Point". No step in the document is marked SSP — the notation is defined and applied nowhere across all 35 steps. The consequence is that the procedure has essentially no hold points: the only decision gate sits inside a test step after all the physical work is complete. There is no gate before the step that demolishes conductors.

Source: Source: footer legend; all 35 steps searched for the SSP marker
Code ref: not available — reported as a general engineering observation
Action: Mark the Safe Stop Points the legend anticipates, at minimum before the demolition step.

The author intended hold points — the legend proves it. They just never got placed.

What is not on this page

The findings that were dropped. A verification pass removes any finding whose citation it cannot confirm in your document, rather than rewording it into something defensible. Three were dropped across the test run. That mechanism is why the ones here can be trusted.
The checks that found nothing. Real reports list them, with a line on why — so you can tell the difference between a check that passed and a check that could not run.
A refusal. One submission in the test run was the wrong kind of document entirely. The engine refused it at intake instead of generating a page of findings about a file that was never a label. A refused submission produces no report and consumes no payment.

These are redacted excerpts from a pre-launch test run, shown to illustrate format and substance. They are not a promise of what any particular document will produce. Arc Flash Files is a product of PEFG, PLLC. Every report is an automated preliminary screening — not a stamped engineering review, not a compliance determination, and not the creation of an engineer-client relationship. All findings must be verified by the Engineer of Record.

See what each product checks Pricing