Shop drawings get rejected far more often for what the package fails to say than for what the engineer disagrees with: a drawing index that does not match the sheets, a dimension that reads two ways, a plate with no grade, a weld symbol with no size. Across three fabrication packages graded by hand against Plumb's 14 review categories, 50 findings came back, nearly all of them documentation a checker could have fixed before the package went out.
Where the 50 findings come from
Three packages, each graded finding by finding: a louver package of seven sheets (28 findings), a structural steel package checked against its project spec (9), and a joist and deck submittal returned by the engineer with redlines (13, a trade reviewer's draft rows, the last of them still being confirmed). Names, locations and dates are left out. The counts are rows in those answer keys: what three graded packages show, not an industry statistic. A few rows could sit in two groups.
1. Document control (13 of 50)
What it looks like: the index REV column blank for every sheet, two sheet titles misspelled. A sheet numbered differently in the index and its title block. Job number and project blank on every sheet. A second page with a sheet number already used and nothing drawn on it. A callout pointing to a sheet not in the set, another with a detail number and no sheet. A "details" sheet listed and blank. On the joist package, the same initials in "detailed by" and "checked by". Eleven of the thirteen were on the louver package's seven sheets.
Why the reviewer bounces it: they cannot tell which revision they hold, cannot find the detail the callout points to, and cannot log it. A blank sheet reads as unfinished.
Before it goes out: the index is read against the sheets, every callout's target is found, the set is flipped for blank pages, every title block is complete, and the checked-by initials are not the detailer's.
2. Dimensions missing or not reconciled (11)
What it looks like: on the louver elevation, the overall height read 121'-0" on the left and 11'-0" on the right. The overall width differed irreconcilably between two sheets. Blade pitch, blade angle and assembly depth were not dimensioned. Two arch radii, nothing saying which governed what. No detail for the arch-to-straight corner joint, the blade-to-frame attachment or the bird screen. On the steel package, the spec called for a fabrication tolerance and the drawings stated none.
Why the reviewer bounces it: the fabricator guesses or writes an RFI, and the reviewer will not guess for them. Two values for one dimension is a stop.
Before it goes out: every chain closes, every view that shows an overall agrees with the others, every curve has a radius and a center, every attachment has a detail, and any tolerance the spec names is on the sheet.
3. Cross-sheet conflicts
What it looks like: the louver elevation called the blades 12 Ga., the detail on the same sheet called them 16 Ga., and the next sheet went back to 12 Ga. Counted strictly that is one row, but the sheet-numbering conflict and the two dimension disagreements above are the same failure: a value changed in one place and not the others.
Why the reviewer bounces it: an engineer who approves the elevation has approved the wrong gauge if the shop builds from the detail.
Before it goes out: each value has one source, and when it changes the set is searched for every place it appears. Cross-sheet consistency is one of Plumb's 14 categories.
4. Finishes, anchors, seals and field conditions not specified (6)
What it looks like: the louver package named its aluminum alloy and no finish, no AAMA class, color or coating system, for an exterior installation. Carbon steel masonry screws through the aluminum frame with no isolation. Anchor length called out; embedment and substrate not. No sealant at the frame-to-masonry joint. No field-verify note for an existing opening. On the steel package, shop primer on nine sheets, a spec requiring surface prep before finishing, and no prep note.
Why the reviewer bounces it: these come back at install, and a reviewer who cannot find a finish cannot approve one. The embedment item was flagged PE-Review-Required: the design basis of an anchor is an engineering question; the finding was only that the drawing is silent. Dissimilar-metal contact is a documentation finding, not a flagged one.
Before it goes out: a finish line for every exposed material, embedment and substrate on every anchor, dissimilar metals isolated, sealant at every exterior perimeter, and a field-verify note at every existing opening.
5. Material and fastener grades (5)
All from the steel package checked against its spec: a splice plate called out A36 where the spec required A572 Grade 50 for plate. Plates and flat bar with shape only and no grade on eight sheets, and no general material note. Cap-plate field bolts called out A307 where the spec required F3125 Grade A325 for concealed structural bolting. The other bolts read A325, a designation ASTM has since folded into F3125; a Minor. Angles in A36 where the spec said nothing about angles, worth a comment rather than a rejection.
Why the reviewer bounces it: a grade that conflicts with a "shall" in the spec is a rejection on its own. A missing grade is a silence the reviewer cannot approve.
Before it goes out: a grade on every BOM line, one general material note, and the spec's materials section read against the BOM once. That comparison is Plumb's Tier 3 — Specification Correspondence Review.
6. Standards without an edition, or the wrong code year (4)
What it looks like: the joist package's notes cited "latest SJI", the AISC specification, the SDI manual and ACI with no edition, and AWS not at all for field and deck welds. The only year on the set was the code year in the title block; the state code in force on the drawing's date is based on a later edition, and nothing showed which governs. The steel package's notes invoked neither AWS D1.1 nor AISC 303; the one reference was the AISC manual, for snug-tight bolting.
Why the reviewer bounces it: a permit reviewer verifies against an edition, and "latest" drifts. The code-year question is PE-Review-Required.
Before it goes out: every standard carries its edition, and the code year matches the contract documents.
7. Weld specification incomplete (3)
What it looks like: louver weld symbols on one detail with no size, type or process, a general note reading "grind visible welds", and no welding standard named. Steel: fillet symbols throughout and no note that welding is by qualified welders to an approved WPS, which the spec required. Joist deck attachment: "weld or equivalent" with no basis for the equivalence.
Why the reviewer bounces it: a symbol without a size is a question to the welder, and the reviewer will not answer it.
Before it goes out: every symbol complete, one general weld note naming the standard, and "or equivalent" replaced or struck.
8. The engineer's redlines not answered (7, all from one package)
What it looks like: the joist and deck submittal came back with an engineer's stamp reading "exceptions noted" and no exceptions listed. The contractor's stamp on the same cover read "reviewed", so the two disagreed. Red markups on three sheets carried no initials, date or tie to the stamp. One moved a bearing-plate setback past the limit the package's own SJI note stated, and nothing reconciled the two. A request to verify the cap-plate hole location was answered "coordinate with the steel shop drawings", not in the package. Two notes putting connection material and bracing outside the fabricator's contract got no response and no supplier.
Why the reviewer bounces it: the next reader cannot tell what was approved, and the shop cannot release.
Before it resubmits: every exception enumerated, every redline attributed and answered on the sheet or by RFI, and both stamps saying the same thing.
What the three packages say
The engineering questions were flagged PE-Review-Required and left for the engineer. The rest were on the sheet, findable by anyone reading the set against itself and the spec, which is what a documentation review does: the sample review shows it on the louver package, how it works shows what comes back, and an Anytime Review is $25 for 10 sheets, then $2.00 (pricing).