Construction Drawing Review Process

There is a moment on most projects when somebody realizes the drawings do not agree with each other. Usually it is a foreman, usually it is at seven in the morning, and usually the disagreement has been sitting in the set for eight months waiting for someone to try to build it.

A door schedule says a frame is 4 inches deep. The wall type it sits in is 6 inches. The detail shows a condition that works for neither. Three consultants each drew something reasonable and nobody reconciled the three. That is not a design failure in any interesting sense. It is a review failure, and review failures are cheap to fix and expensive to skip.

Drawing review is the discipline of reading a document set critically before somebody builds from it. Not scanning it. Not familiarizing yourself with it. Interrogating it, from several directions, looking for specific classes of defect, and logging what you find in a form the design team can act on.

Most contractors do some version of this. Very few do it as a process. What usually happens is that experienced people read the sheets they care about, find the problems they have seen before, mention a few in a meeting, and move on. The findings that matter most, the ones between disciplines and between the drawings and the specifications, are exactly the ones that nobody’s discipline-shaped reading habits will surface.

This article lays out drawing review as a repeatable process: six review lanes, a milestone cadence, a comment lifecycle, and the decision rules for what becomes a comment, what becomes an RFI, and what you absorb. It is written for contractors reviewing a design team’s documents, though the same lanes work for a design firm’s internal quality control.

What Drawing Review Is, and What It Is Not

Construction drawing review is the systematic examination of a drawing set and its associated specifications to identify errors, omissions, internal conflicts, unbuildable conditions, and regulatory exposure, before the documents are used for pricing or construction.

The definition matters because drawing review gets confused with three adjacent activities that share its inputs and answer entirely different questions.

Table 1. Drawing Review Compared With Adjacent Activities

The distinction from clash detection is the one people most often collapse. A federated model will tell you a duct and a beam intersect. It will not tell you that the beam appears at two different elevations on two sheets, that the duct is not in the specification at all, that the detail resolving the condition references a section that does not exist in the set, or that the resulting assembly cannot be installed because there is no access above the ceiling. All four of those are drawing review findings and none of them is a clash.

Core Terminology

Set integrity. Whether the set is complete, current, internally consistent in its references, and correctly indexed. The precondition for everything else.

Comment. A logged finding directed at the design team, with a location, a description, a category, and a severity. The unit of output.

Backcheck. Verification that the design team’s response actually resolved the comment in the reissued documents. The step most often skipped, and the reason comments recur.

Discipline-internal defect. A conflict inside one consultant’s own documents. A schedule disagreeing with a plan, a detail disagreeing with a section.

Cross-discipline defect. A conflict between two consultants’ documents.

Correlation defect. A conflict between the drawings and the specifications, in either direction.

Constructability finding. A condition that is drawn coherently and cannot practically be built, sequenced, or accessed.

Dead reference. A callout pointing to a detail, section, sheet, or specification section that does not exist in the set. Individually trivial, collectively diagnostic.

Comment disposition. The design team’s response: incorporated, incorporated as modified, rejected with reason, or deferred. Anything else is not a disposition.

Objectives of the Review Process

• Find defects while they are still paper. A conflict resolved by a design revision costs a fraction of the same conflict resolved by a change order, and a small fraction of the same conflict resolved by demolition.
• Reduce RFI volume during construction. A large share of early-construction RFIs are questions the documents could have answered, and a meaningful share are questions somebody could have asked during review.
• Establish document reliability before pricing. An estimate built on a set with unresolved internal conflicts carries contingency for the conflicts, or does not, and neither outcome is good.
• Produce an actionable record. A comment log with locations, categories, and dispositions is a working document. A list of concerns raised verbally in a meeting is not.
• Protect the schedule. Constructability and sequencing findings surfaced in design development change the drawings. The same findings surfaced during construction change the schedule.
• Build institutional knowledge. The defect patterns on this laboratory building will repeat on the next one. Captured properly, review findings become the checklist.

Why Formal Review Pays

Industry research on the cost of design errors and omissions consistently lands in an uncomfortable range, generally somewhere between two and six percent of construction cost on commercial building work, with rework attributable to document problems making up a substantial share. You do not have to accept any particular published figure. Run the arithmetic on your own change order log instead: pull the last three completed projects, categorize the change orders by cause, and see what proportion trace to something a careful reader could have found in the documents.

Most contractors who do that exercise find a number they did not expect. The categories that dominate are rarely dramatic. Missing details. Dimensional conflicts. Schedules that disagree with plans. Specifications requiring something the drawings never showed. Conditions that were drawn but could not be built in the sequence the schedule demanded.

There is a second argument that has nothing to do with cost recovery. Review changes your relationship with the design team, and the direction it changes depends entirely on how you do it. A contractor who submits 400 comments, most of them trivial and many of them wrong, trains the architect to discount the next 400. A contractor who submits 90 comments that are specific, correctly located, properly categorized, and genuinely consequential gets treated as a resource. The second contractor’s comments get incorporated. That difference compounds over a project and over a relationship.

Stakeholders and What Each Lane Needs

Drawing review requires several competencies that rarely sit in one person. A reviewer who is excellent at cross-discipline coordination is often mediocre at constructability, because those skills come from different careers.

Table 2. Drawing Review RACI

R responsible, A accountable, C consulted, I informed.

Two assignments deserve defending. The superintendent is accountable for constructability, because constructability is a question about installation sequence, access, tolerance, and what a crew can physically do, and the office answers those questions wrong when it answers them alone. And trade partners are responsible contributors on lanes 3 and 5 wherever your delivery method allows it, because a sheet metal foreman reading a ceiling plan sees things nobody in your office will see.

The Six Review Lanes

The central methodological claim of this article is that no single reading of a document set finds all defect classes, because different classes are visible only from specific directions. Run six independent passes. Findings from all six merge into one comment log.

Table 3. The Six Lanes at a Glance

Lane 1: Set Integrity and Version Control

Run this first, always, and run it before you read a single sheet for content. Everything downstream is invalid if the set is wrong, and reviewing a superseded sheet wastes the reviewer’s most expensive resource, which is attention.

The checks are mechanical:

• Reconcile the sheet index on G-001 against the sheets actually received. Missing sheets and extra sheets both occur, and both are invisible without the reconciliation.
• Record the revision number and date of every sheet. Confirm all sheets belong to the same issue, or document the exceptions deliberately.
• Verify every addendum and bulletin has been incorporated, and identify which sheets and specification sections each one modified.
• Confirm the specification table of contents matches the sections actually provided.
• Test references. Every detail callout, section cut, enlarged plan reference, and specification citation should point to something that exists. Dead references are individually trivial and collectively the best single indicator of set maturity.
• Confirm the drawing list, sheet numbering, and discipline series are internally consistent.

Dead references deserve a note. A set with a handful is normal. A set with sixty is telling you the documents were assembled under time pressure and probably were not internally reviewed, which should change how you allocate reviewer time across the other five lanes.

Lane 2: Discipline-Internal Accuracy

Question: does each consultant’s set agree with itself?

This is the lane where discipline experience pays, and it is best run by someone who has built the work. The recurring defect classes are consistent enough to check for directly.

Schedule-to-plan conflicts. Door schedules, window schedules, room finish schedules, equipment schedules, panel schedules, and fixture schedules all restate information that also appears on plans. Whenever information appears twice, it eventually disagrees. Door schedules are the single most reliable source of findings in a commercial set.

Detail-to-plan conflicts. A wall type detail shows a 6 inch stud; the plan dimension string accommodates 4 inches. A curb detail shows 8 inches; the section shows 6.

Dimensional integrity. Do dimension strings add up to the overall? Do gridlines agree between disciplines and between levels? Do stacked elements align vertically across floor plans?

Tag and callout consistency. Every tagged item should appear in the corresponding schedule, and every scheduled item should appear tagged somewhere.

Internal specification conflicts. Within one section, part 2 products and part 3 execution sometimes contradict. Across sections, the same product gets specified twice with different requirements.

Table 4. Lane 2 Recurring Defects by Discipline

Lane 3: Cross-Discipline Coordination

Question: do the consultants agree with each other?

Two kinds of conflict live here, and most teams check only one.

Spatial conflicts are geometric. Systems competing for the same space, insufficient clearance, penetrations conflicting with structure, equipment not fitting the room drawn for it. If you have a federated model, coordination handles much of this well, and you should use it.

Informational conflicts are not geometric and models do not catch them. One discipline states a requirement the other has not accommodated. The mechanical schedule lists this unit at 4,200 pounds; the structural drawings were designed for this unit at 2,800. The electrical one-line assumes a 480 volt service; the mechanical equipment schedule lists units at 208 volts. The architectural rated wall schedule shows a two-hour assembly; the mechanical drawings show unrated duct penetrating it with no damper.

Table 5. Lane 3 Cross-Discipline Check Pairs

Method note: review each pair as a batch rather than sheet by sheet. Ten findings on the mechanical-to-structural pair reviewed together reveal a pattern, usually that one consultant worked from a superseded background, and identifying the pattern is worth more than the ten individual comments.

Lane 4: Drawing-to-Specification Correlation

Question: do the drawings and the specifications agree, in both directions?

This is the most skipped lane but is also one of the two highest yield lanes. Tiring manual reading makes this lane easy to skip. Adding to the skippable nature of this lane is the fact that this section appears to be absent from every job description.

Four distinct defect classes, and each requires looking a different way.

Specified but not drawn. A requirement exists in the specifications with no graphical representation. Hangers and supports, seismic restraint, grounding and bonding, insulation extent, firestopping locations, and testing requirements all live primarily in specification text. The failure mode is that estimators quantify from drawings, so an obligation with no drawing representation gets no quantity and therefore no dollars.

Drawn but not specified. Something appears on the drawings with no governing specification section. Late additions and details borrowed from previous projects are the usual causes. The item gets installed to whatever standard the installing contractor chooses, and the argument happens at submittal.

Conflicting instruction. The drawing and the specification both address the item and say different things. The drawing shows a mechanical fastener; the specification requires adhesive. Both are coherent, and the installer will pick one.

Missing detail reference. A note or specification paragraph refers to a detail, section, or requirement that does not exist. Related to Lane 1 dead references but consequential rather than clerical, because the referenced content was needed to define the work.

Table 6. Lane 4 High-Yield Correlation Targets

Lane 5: Constructability and Means and Methods

Question: can this be built, in this sequence, with the access available, within achievable tolerances?

Everything in this lane may be drawn correctly and coordinated properly. That is what distinguishes it. A constructability finding is a condition where the documents are internally consistent and the construction is impractical.

Check for:

• Access. Can a person and their tools reach the work? Above hard ceilings, inside chases, behind fixed equipment, within shaft walls.
• Sequence. Does the drawn condition require trade A to finish before trade B, and trade B before trade A? Closed-loop sequences are more common than people expect.
• Installation clearance. Is there room to install, not merely room for the finished assembly? A duct that fits in the plenum by half an inch cannot be lifted into place.
• Tolerance stack-up. Do accumulated tolerances of adjacent trades make the drawn condition unachievable? Concrete tolerance plus steel tolerance plus curtain wall tolerance at a single interface is where this bites.
• Maintenance access. Can equipment be serviced and replaced? Filter access, coil pull space, valve reach, panel working clearance.
• Temporary conditions. Does the design accommodate shoring, crane access, laydown, hoisting paths, and phased occupancy?
• Weather and enclosure sequencing. Does the drawn assembly require conditions the schedule cannot provide?
• Achievable means. Is there a practical method to build the drawn condition with available equipment and crews, or does it assume an approach nobody uses in this market?

Lane 6: Code, Regulatory, and Authority Having Jurisdiction

Question: does this comply, and will this specific jurisdiction accept it?

A contractor is not the code official and should not position itself as one. What a contractor can do, and should, is flag conditions that appear noncompliant and conditions where the jurisdiction’s known interpretation differs from what is drawn. That second category is where local knowledge is worth real money, because AHJ interpretation varies substantially and design teams working outside their home market frequently do not know the local position.

Check life safety plans against the architectural set for egress widths, travel distances, dead ends, rated assembly continuity, and smoke compartmentation. Check accessibility for clearances, reach ranges, slopes, and routes. Check energy compliance for envelope continuity and equipment efficiency against the stated compliance path. Check whether the drawn deferred submittals and special inspections match what the jurisdiction requires. Check that anything requiring an alternative means and methods determination is identified.

The severity here justifies the attention even at moderate frequency. Code findings discovered during inspection can gate occupancy, and occupancy delays carry liquidated damages.

Milestone Cadence

Review is not one event. The lanes get progressively more valuable at different milestones, and running the wrong lane at the wrong time wastes effort.

Table 7. Review Cadence by Design Milestone

The 90 percent construction document review is the one that carries the weight. Before that, the documents are still changing enough that detailed findings get overtaken. After that, changes cost money. If your organization can only staff one thorough review, staff that one, and give it three to four weeks of calendar time.

Addenda deserve their own discipline. A revision issued to fix one problem routinely creates another, because the consultant who revised the mechanical sheet did not tell the structural engineer. Difference every revision against the prior issue and rerun Lane 3 on whatever changed.

The Comment Lifecycle

Findings are worthless until they are logged in a form somebody can act on, and comments are worthless until somebody verifies the response.

Log. Every data entry includes a unique ID as well as a reference to the sheet or specification, the location on the sheet, the description of the condition, the defect category, the severity, the reviewing lane, and the reviewer.

Description of the condition, not a solution. “Door schedule lists frame depth 4 inches at openings 112, 118, and 124; wall type B at these locations is 6 inches per detail 3/A-501” is a comment a consultant can act on. “Doors are wrong” is not. Where you have a preferred resolution, offer it separately as a suggestion, clearly marked, so the consultant retains design authority. That distinction keeps the relationship and, incidentally, removes you from the business of designing.

Prioritize and categorize. Two axes: category tells you who resolves it, severity tells you when.

Table 8. Comment Categories

Table 9. Comment Severity and Required Action

Issue. Submit in the format the design team can consume, on a stated schedule, with severity visible. Batch low-severity items separately from severity 1 and 2 findings. Burying four critical structural comments inside 300 clerical ones is a self-inflicted wound.

Design response. Four legitimate dispositions: incorporated, incorporated as modified, rejected with a stated reason, or deferred with a stated date. “Noted” is not a disposition, and accepting it as one is why the same comment appears in three consecutive reviews.

Backcheck. Verify in the reissued documents that the response actually resolved the condition. This is the step almost everyone skips. Track a resolution rate, meaning comments verified resolved divided by comments issued, and expect the first honest measurement to be lower than you assumed.

Escalate or convert. Comments unresolved as the documents go to bid become one of three things: an RFI, a qualification in your proposal, or a priced contingency. Every unresolved comment must become one of the three. None of them may simply remain open, because an open comment at bid is a risk nobody has priced.

Table 10. Comment Lifecycle Metrics

Required Documentation

Table 11. Documents Required for Effective Review

Technology Integration

Table 12. Review Approach Comparison

The honest read of that table is that no single technology supports all six lanes, and the two lanes that carry the most schedule risk, 5 and 6, are barely supported by any of them. Constructability and code review are judgment work performed by experienced people, and they will remain so.

Where tooling changes the picture is Lane 4 and the informational half of Lane 3. Both require systematic comparison across an entire document set, which is a volume problem rather than a judgment problem, and volume problems are the ones that get solved.

AI-Assisted Opportunities

The structural difficulty in drawing review is that a commercial set is too large to read exhaustively in the time available. A 400 sheet set plus a 1,200 page project manual is beyond what a reviewer can cover, so every team samples. Most sample without saying so, and the sampling is shaped by discipline habit, which means Lane 4 and the informational parts of Lane 3 get thin coverage precisely because they require reading things outside the reviewer’s discipline.

This is where conversational AI over the document set has changed the work. Platforms built for construction documents, iFieldSmart AI among them, index every sheet and every specification section, extract each note with its exact location on the drawing, and let a reviewer interrogate the whole corpus directly. A reviewer asks which specification sections are referenced on drawings but absent from the project manual, or which notes reference details that do not exist in the set, or where mechanical equipment weights appear alongside structural capacity, and gets a sourced answer with links back to the drawing location.

The distinction worth holding onto is that this does not review the documents. It answers questions about them at a speed and completeness a human cannot match, which converts several lanes from sampled to complete.

Table 13. AI Use Case Matrix for Drawing Review

Implementation

• Baseline first. Pull the last three completed projects, categorize every change order and early RFI by cause, and map each to the lane that would have caught it. Two people, about a week. You will learn which lanes your organization is weakest in, and the answer is usually 4 and 5.
• Fix the cadence. Decide which milestones get which lanes, and staff the 90 percent construction document review properly with three to four weeks of calendar time.
• Add Lane 1 as a gate. Nobody reads for content until set integrity is confirmed. Half a day, and it protects everything downstream.
• Add Lane 4 next. It is the highest-yield lane most teams skip. Start with the twelve targets in Table 6 rather than attempting comprehensive correlation.
• Staff Lane 5 with field people and trades. Schedule walkthroughs as meeting invitations, not soliciting feedback meeting invitations.
• Standardize the comment log with categories and severity. A shared spreadsheet is sufficient to start.
• Institute backcheck and measure the resolution rate. This single addition changes design team behavior more than anything else on the list.
• Close the loop. At project completion, map every scope-related change order and every early RFI to a lane, and revise your checklists accordingly.

Table 14. Implementation Roadmap

Best Practices

Table 15. Drawing Review Best Practices

Common Mistakes

Table 16. Common Drawing Review Failures

How the Lanes Perform on Real Projects

Class A office tower, 340,000 square feet, core and shell. Lane 3, mechanical to structural, produced the finding that mattered. Rooftop unit weights in the equipment schedule had increased through design development, and the structural framing at the penthouse had been designed to the earlier weights. Caught at 90 percent construction documents it was a framing revision. Caught after steel fabrication release it is a schedule event with long-lead implications.

Regional hospital expansion, 190 beds, occupied campus. Lane 6 dominated, as it usually does on healthcare work. The set showed a rated corridor assembly with mechanical penetrations lacking dampers on two levels, and the jurisdiction’s interpretation of smoke compartmentation at the connector to the existing building differed from what was drawn. Both were flagged during design development, when they were drawing revisions rather than inspection failures gating occupancy.

Hyperscale data center, 60 MW. Lane 4 justified the whole review. The specifications required seismic restraint and hanger support engineering for mechanical and electrical distribution that appeared nowhere on the drawings, a specified-not-drawn defect carrying delegated engineering, a deferred submittal, and a special inspection. On a project with equipment density that high, the exposure was substantial and it was invisible to anyone reviewing drawings alone.

Advanced manufacturing facility, food grade. Lane 5 with trade participation. The hygienic wall panel system was drawn as a coherent assembly, and the installation sequence it implied required the panels to be installed before overhead process piping and the piping to be supported from structure the panels concealed. A closed sequence loop, invisible in the model, obvious to the panel installer in about four minutes.

University science building, 11 floors. Lane 2, discipline-internal. The door schedule listed 61 openings with hardware sets inconsistent with the rated assemblies shown on the life safety plans, and 14 openings appeared on plans without appearing in the schedule at all. Door schedules are the most productive single Lane 2 check in commercial work and this set was a typical rather than an exceptional example.

Multifamily podium, 240 units. Lane 1 produced the finding. Two unit type plans had been revised by addendum, and the corresponding enlarged plans and interior elevations had not, leaving three versions of the same bathroom in the set. On repetitive residential work a single unresolved unit type conflict multiplies by the unit count, which makes Lane 1 unusually valuable on that project type.

Highway interchange, design-bid-build. Lane 3, civil to structural to electrical. Conduit for future intelligent transportation systems appeared on the signal plans at depths that conflicted with the drainage structure inverts shown on the civil plans. Caught in review it was a coordination revision. Caught during trenching it is a stop-work and a redesign with traffic control implications.

Historic warehouse conversion to hotel, 140 keys. Lane 5, existing conditions and access. The new mechanical distribution was drawn through an existing structural bay whose actual clear dimensions, verified by field measurement during review, were three inches tighter than the record drawings the design team had worked from. Adaptive Reuse places this defect class at the center, hence the need to conduct field verification during the review process and not after.

Frequently Asked Questions

How long does a thorough drawing review take on a commercial project?

For a 90 percent construction document review of a 300 to 500 sheet set with a full project manual, budget three to four weeks of calendar time and roughly 80 to 160 person-hours spread across the six lanes and several reviewers. Lane 1 is half a day. Lane 4 is the most time-consuming at 20 to 40 hours if done properly. Lane 5 is a half-day walkthrough plus preparation. Compressing below about two weeks means sampling, and sampling should be documented rather than implied.

At which milestone does review deliver the most value?

Design development for lanes 3, 5, and 6, because findings there change the design rather than the price. Ninety percent construction documents for the comprehensive review, because it is the last point at which corrections are free. If you can only staff one, staff the 90 percent review. If you can staff two, add design development, not the bid set.

How is drawing review different from clash detection?

Clash detection answers a geometric question: do two modeled systems occupy the same space? Drawing review asks whether the documents are correct, complete, internally consistent, correlated with the specifications, buildable, and compliant. A model shows a duct hitting a beam. It does not show that the beam is at two elevations on two sheets, that the duct is unspecified, that the resolving detail references a nonexistent section, or that there is no access to install it. Roughly one of the six lanes is well served by coordination, and it is the spatial half of one lane.

Who should lead the review?

Preconstruction should own the process and the log. Lane ownership should be distributed by competency: VDC leads coordination, a superintendent leads constructability, estimating contributes to specification correlation, and the project manager owns code and regulatory with a consultant where the project warrants one. A single reviewer running all six lanes will do two of them well.

How many comments should we expect?

At 90 percent construction documents on a commercial set, 40 to 120 comments per 100 sheets is a reasonable range, with 15 to 30 percent at severity 1 or 2. Density below about 30 per 100 sheets usually means shallow review. Density above 200 usually means the log has been padded with clerical findings, which costs you credibility with the design team and buries the findings that matter.

What do we do with comments the design team rejects?

Accept the rejection, record the reason, and decide what it means commercially. A rejected comment on a genuine conflict does not disappear. It becomes an RFI, a qualification in your proposal, or a priced contingency. Where the rejected comment concerns life safety or structural adequacy, escalate it in writing to the owner rather than absorbing it, because those categories carry exposure a contractor should not silently accept.

Should we share review findings with subcontractors?

Yes, selectively and to your benefit. Sharing the relevant findings for a trade’s scope produces more accurate bids and fewer defensive exclusions. On negotiated and CMAR work, involving key trades directly in lanes 3 and 5 during design is one of the highest-value things integrated delivery makes possible, and most teams underuse it.

Does drawing review create liability for the contractor?

This is worth a careful answer and worth confirming with your own counsel, since it varies by jurisdiction and by contract. In general, contractors under most standard forms have a duty to report discovered discrepancies and do not assume design responsibility by reviewing documents. What can create exposure is directing a design solution rather than reporting a condition. Describing the condition and letting the design professional resolve it keeps the line clean, which is a practical reason for the drafting convention, not only a diplomatic one.

What is the highest-yield single check if we only have a day?

Lane 1 in the morning, then the door schedule and the mechanical-to-structural pair in the afternoon. Set integrity protects everything else, door schedules restate information appearing in four other places, and mechanical-to-structural carries the highest-severity coordination findings in a typical commercial building.

How do we review a set delivered only as a model, with no traditional sheets?

The lanes hold, and the methods change. Set integrity becomes model version and federation control. Discipline-internal accuracy becomes schedule and parameter validation against modeled geometry. Coordination becomes model coordination plus the informational checks, which remain non-geometric and still require reading parameters and specifications. Constructability and code review are unchanged. The one lane that becomes harder is Lane 4, since specifications remain documents regardless of how the geometry is delivered.

How should addenda during construction be reviewed?

Difference the revision against the prior issue to identify exactly what changed, then rerun Lane 3 on the changed content only. Revisions issued to fix one problem create new coordination conflicts often enough that this should be routine. It typically takes an hour or two per addendum and prevents the specific failure of a correction introducing an uncoordinated condition nobody rechecked.

How do we know the review process is working?

Three signals in order of speed. Resolution rate at backcheck, available immediately and the only direct measure of effect. Early-construction RFI volume, which drops within the first sixty days of mobilization. And change orders attributable to document defects as a share of contract value, which needs two to four completed projects to trend credibly. All three require the baseline exercise, which is why that comes first.

Expert Recommendations

For the preconstruction manager: make Lane 1 a gate. Nobody reads for content until set integrity is confirmed and dead references are counted. It costs half a day and it tells you how to allocate the rest of the review.

For the VDC or BIM manager: publish your coordination-heavy zone list to the review team before the walkthrough. Corridor plenums, mechanical rooms, shaft walls, and equipment yards concentrate both clashes and constructability findings, and the rest of the team does not know where they are until you tell them.

For the superintendent: lead the constructability walkthrough and ask one question repeatedly. How would your crew actually install this? That question produces findings no office review generates, and it produces them in the hour before lunch.

For the estimator: take specification correlation seriously, because you are the person who will discover a specified-not-drawn obligation the hard way. The twelve targets in Table 6 account for most of the exposure and can be checked in a day.

For the project manager: institute backcheck and report the resolution rate. It is the only metric that measures whether the review changed anything, and reporting it changes design team behavior faster than any amount of meeting discussion.

For the project executive: escalate unresolved severity 1 findings to the owner in writing. Life safety and structural findings that a design team declines to resolve should not be quietly absorbed by a contractor, and the written record matters.

For the owner or owner’s representative: ask your contractor for the comment log and the resolution rate, not the summary. How specific the comments are and how many were verified resolved tells you more about your project’s risk position than any design progress report.

Where This Leaves You

Drawing review is not clever work. It is reading, done deliberately, from six directions, with the findings written down in a form somebody can act on and then verified after they respond. The reason it produces such uneven results across the industry is not skill. It is that five of the six lanes require reading outside your own discipline, and almost nobody’s reading habits do that naturally.

The two lanes that separate strong review teams from average ones are specification correlation and constructability. Both are skipped for the same reason, which is that neither has an obvious owner. Correlation requires someone to read a project manual, and constructability requires someone to leave the office. Assign both explicitly and most of the benefit follows.

Start with the baseline. Map your last three projects’ change orders and early RFIs to the six lanes, and the profile will tell you exactly which lane to staff first. It is almost always Lane 4.