Almost none of these are attempts to get away with something.
That framing matters, because the instinctive reading of a non-compliant submittal is that somebody tried it on. Occasionally true, rarely useful. Most compliance issues arise from a rational process producing a wrong result: a supplier’s standard data sheet does not happen to state the attribute the specification cared about, an estimator bought a configuration that matched the schedule but not paragraph 2.4, a fabricator’s detailer used the assembly they always use because the drawing did not show a different one.
Mechanism change the remedy. An unsolicited proposal is considered silent on a performance criterion when a manufacturer’s literature fails to publish that value in that condition. The solution is a request to the manufacturer, not a rejection note to the subcontractor to up the ante. Should a submitted document identify a specific assembly, but the cited evaluation report pertains to a different but similar assembly, the fix is identification of the specific listing that matches the assembly described in the submitted document, which the submitter may genuinely not know exists.
This article catalogs fourteen compliance issues that recur across commercial projects. For each: what it looks like in the submitted document, why a competent submitter produces it, how to detect it, what it costs if it survives approval, and what the comment should say. It closes with the question that matters most operationally, which is which of these a reviewer catches and which an inspector catches.
Why These Issues Recur
Four structural causes produce most of what follows, and none of them is carelessness.
Manufacturer literature is written for a market, not for your specification. A data sheet states the attributes the manufacturer believes buyers care about, at the conditions the industry commonly uses. Your specification may care about a different attribute at a different condition. The literature is complete and does not contain what you need.
Configuration happens in procurement, not in submittal preparation. By the time a submittal is assembled, the configuration was frequently decided weeks earlier by someone comparing the schedule against a price. Attributes specified in the section but not carried onto the schedule get lost at that step and reappear as submittal non-compliance.
Standards compliance is layered. A section cites a standard, the standard cites others, and compliance means the whole chain. A submitter reading only the section text can comply with everything it says and still not comply with what it invoked.
Nobody’s job description includes reading the specification against the submittal. The subcontractor’s estimator read the section during pricing. The supplier’s application engineer read the schedule. The person assembling the submittal read neither in detail. That gap is where twelve of the fourteen issues below originate.
Table 1. The Fourteen Issues at a Glance
| # | Issue | Detection Difficulty | Severity | Usually Caught By |
|---|---|---|---|---|
| 1 | Silence on a specified criterion | High | Moderate to very high | Nobody, until commissioning |
| 2 | Product family rather than specific model | Low | Moderate | Reviewer |
| 3 | Performance at a non-specified test condition | High | High | Nobody, until performance testing |
| 4 | Standard cited without compliance demonstrated | Moderate | Moderate to high | Reviewer, if attentive |
| 5 | Non-equivalent tested configuration | High | Very high | Inspector |
| 6 | Superseded standard edition or expired listing | Moderate | Moderate to high | Inspector or AHJ |
| 7 | Incomplete component and accessory coverage | Moderate | Moderate | Field, at installation |
| 8 | Non-compliant accessory on a compliant base product | High | Moderate to high | Inspector or commissioning |
| 9 | Catalog literature substituted for project-specific content | Low | Moderate | Reviewer |
| 10 | Unflagged deviation | Moderate | Moderate to high | Nobody, until it matters |
| 11 | Substitution disguised as product data | Low | Moderate to high | Reviewer, if procedurally alert |
| 12 | Missing, expired, or wrongly issued certification | Low | Moderate | Closeout or inspection |
| 13 | Delegated design defects | Moderate | Very high | AHJ or special inspector |
| 14 | Conflict with a previously approved submittal | Moderate | Moderate to high | Field |
The last column is the most useful one in this article. Five of the fourteen are typically caught by somebody other than the reviewer, which means they are caught after the work is fabricated, delivered, or installed. Those five are where the money is.
Objectives
- Recognize each issue by its signature in the submitted document
- Understand the mechanism, so the remedy addresses the cause rather than the symptom
- Detect the five issues that currently reach inspectors instead of reviewers
- Write comments that produce a targeted correction rather than a guess
- Distinguish issues requiring a different product from issues requiring more information
Issue 1: Silence on a Specified Criterion
What it looks like. The submittal is complete, professional, and says nothing whatsoever about a requirement the specification stated. No wrong value to notice. Nothing at all.
Why a competent submitter produces it. The manufacturer’s data sheet does not publish that attribute. The submitter assembled the standard package the supplier provides, which addresses the attributes the market asks about. Nobody compared the package against the section paragraph by paragraph, so the absence was never visible to them either.
How to detect it. Only by working from a decomposed list of specified requirements and recording a determination for each. Recognition review cannot detect an absence, because there is nothing there to recognize. This is the single strongest argument for requirement-driven review.
What it costs. Depends entirely on the criterion. An unaddressed finish thickness is minor. An unaddressed acoustic criterion, filtration efficiency, pressure differential capability, or seismic qualification is a compliance failure discovered at commissioning, at inspection, or after occupancy.
Comment to write. Name the paragraph, name the criterion, and request the value at the specified condition. “Paragraph 2.2.F requires sound power not exceeding 82 dB in octave band 3. Submitted data does not state sound power in any octave band. Provide certified sound power data by octave band.” That gets you data. “Acoustic performance not demonstrated” gets you a resubmittal that guesses.
| KEY TAKEAWAY This is the dominant compliance issue in commercial construction and the one least likely to be found. Every other issue on this list involves something visible that a reviewer could notice. This one involves nothing being there, and the only method that finds it reliably is comparing the submittal against a written list of what was required. |
|---|
Issue 2: Product Family Rather Than Specific Model
What it looks like. The submittal identifies a manufacturer and a series. Data covers a range of sizes or capacities. No specific model number appears, or one appears in a place that does not govern.
Why it happens. The supplier provided series literature, which is what suppliers publish. The submitter forwarded it.
How to detect it. Look for a model designation and a configuration. If performance data appears as a table covering several sizes, the submittal has not stated which one is being furnished.
What it costs. Every determination downstream becomes provisional. It also creates a real dispute risk: an approval of series literature arguably approves whatever the submitter later says was intended.
Comment to write. Request the specific model designation and the configuration, and state that performance data must be specific to the furnished unit rather than to the series.
Issue 3: Performance at a Non-Specified Test Condition
What it looks like. A value that meets or exceeds the specified number, measured under different circumstances. External static pressure at a different airflow. Sound at a different distance or a different metric. Thermal performance at standard rather than project conditions. Air and water performance at the test standard’s minimum pressure rather than the project’s design pressure.
Why it happens. The manufacturer publishes at industry-standard conditions. The specification stated project conditions. The submitter matched the number, which is what a reasonable person does when comparing a specification value against a data sheet value.
How to detect it. Verify the condition alongside every performance value. This has to be a deliberate check, because the number itself looks compliant and the eye goes to the number.
What it costs. High, and it fails late. The product performs exactly as its data says and not as the project requires, so the failure surfaces at performance testing, at commissioning, or in occupant complaints.
Comment to write. State the specified condition explicitly and request data at it. Do not simply say the data is insufficient.
Table 2. Issue 3, Recurring Condition Mismatches
| Attribute | Specification Typically States | Submittal Typically States | Why They Are Not Comparable |
|---|---|---|---|
| Fan performance | Static pressure at the design airflow | Static pressure at a different airflow | Fan curves are non-linear |
| Acoustic performance | Sound power by octave band | Single-number sound pressure at a distance | Different metric entirely |
| Thermal performance | U-factor or R-value at project boundary conditions | Value at standard test conditions | Conditions materially affect result |
| Air infiltration and water penetration | Test at project design pressure | Test at the standard’s minimum pressure | Higher pressure is a harder test |
| Electrical short-circuit rating | Rating at the furnished configuration | Rating at a different bracing or configuration | Rating is configuration-dependent |
| Seismic qualification | Qualification at the specified importance factor and mounting | Qualification at a lower factor or different mounting | Qualification does not transfer |
| Fume hood containment | Containment at the specified face velocity | Containment at a different face velocity | Velocity drives containment |
| Efficiency | Efficiency at the specified rating point | Peak or nominal efficiency | Part-load differs from peak |
Issue 4: Standard Cited Without Compliance Demonstrated
What it looks like. The submittal states that the product complies with a standard. No test report, no certified rating, no listing reference. Just the claim.
Why it happens. The submitter read the specification as requiring compliance and provided a statement of compliance. Where the specification required evidence, the distinction between asserting and demonstrating was not obvious from the section text.
How to detect it. Determine what the specification requires. Many sections require certified test reports or ratings under a recognized program; some require only compliance. Where evidence is required and a claim is offered, the determination is not addressed.
What it costs. Variable. On performance-critical and rated products, substantial, because the claim may be true and unprovable at inspection.
Comment to write. Cite the paragraph requiring the evidence and name the evidence type. Where the specification is genuinely silent on evidence, record that the determination rests on the manufacturer’s certification, which is a different and weaker position worth documenting.
Issue 5: Non-Equivalent Tested Configuration
What it looks like. A test report or listing that is real, current, from a qualified agency, and covers an assembly that differs from the one being furnished. Different framing spacing. Different fastener type or pattern. Different substrate. Different resilient channel. Different sealant.
Why it happens. This is the one where the submitter is genuinely doing their best. Assembly listings are numerous and specific, the differences between them are technical, and finding the listing that matches your detailed assembly exactly requires searching a listing directory with the drawing in hand. A submitter who finds a listing with the right rating reasonably believes they have complied.
How to detect it. Compare the tested assembly description against the detailed assembly, component by component. Framing type and spacing, sheathing, fasteners, insulation, finish layers, and any resilient or isolation component.
What it costs. The most expensive issue on this list, because it is typically caught by an inspector rather than a reviewer, and by then the assembly is installed. On repetitive construction, multiplied by the repetition count.
Comment to write. Identify the specific difference and request a listing matching the detailed assembly, or a design-change request if none exists. Being specific about the difference is what makes this comment actionable, because the submitter frequently cannot see it.
| WARNING This is the finding I would prioritize above all others on rated assemblies. Reviewers verify that a listing exists and that its rating matches the requirement. Very few verify that the tested assembly matches the drawn assembly, and that is exactly the check an inspector performs. Discovering it at inspection on a project with hundreds of repetitions is a serious event. |
|---|
Issue 6: Superseded Standard Edition or Expired Listing
What it looks like. A test report to an earlier edition of a standard than the specification cites. A listing that has been withdrawn or revised. A certification that has lapsed.
Why it happens. Reports and listings persist in supplier files. Standards change and the change is invisible unless someone checks the edition. Certifications expire on cycles nobody tracks.
How to detect it. Check the edition against the specification’s citation. Where the citation is undated, the current edition at bid date generally governs. Check listing and certification dates and validity.
What it costs. Moderate to high, and the failure mode is an authority having jurisdiction rejecting evidence you believed you had. That happens at inspection.
Comment to write. Cite the specified edition, note the submitted edition, and request current evidence. Where a standard changed materially between editions, say what changed if you know, because the submitter may not.
Issue 7: Incomplete Component and Accessory Coverage
What it looks like. The primary item is fully documented. Specified accessories and components are absent from the submittal. Isolators, flexible connections, access doors, trim, closures, controls, sensors, spare parts.
Why it happens. The submittal was assembled around the equipment. Accessories are separate line items, sometimes from separate suppliers, and they get treated as procurement detail rather than as specified scope.
How to detect it. Compare the submittal contents against the specification’s list of what the assembly includes. Mechanical, not judgmental, and it belongs in the contractor’s screen.
What it costs. Usually discovered at installation when the accessory is not in the shipment, which is a delay and a rush order rather than a compliance failure. Occasionally worse, when the missing component was performance-relevant.
Comment to write. List the specific components required by paragraph and not addressed. A list gets a complete resubmittal; “incomplete” gets a partial one.
Issue 8: Non-Compliant Accessory on a Compliant Base Product
What it looks like. The equipment complies. An accessory does not. A compliant damper with a non-rated actuator. A compliant door with non-compliant hardware. A compliant panel with an incompatible sealant. A compliant luminaire with a driver that does not meet the specified dimming protocol.
Why it happens. Base product selection receives attention because it carries the cost. Accessories are selected on availability and price, frequently by someone who never read the section.
How to detect it. Verify accessory compliance separately from base product compliance. This requires reading the accessory data rather than noting its presence, which is the step usually skipped.
What it costs. Moderate to high, and it fails at inspection or commissioning where the accessory is what actually performs. On rated assemblies a non-rated accessory defeats the rating entirely.
Comment to write. Identify the accessory, the requirement it fails, and its data source. Reviewers frequently write about the assembly generally when the problem is one component, which produces a resubmittal that changes the wrong thing.
Issue 9: Catalog Literature Substituted for Project-Specific Content
What it looks like. Where shop drawings were required, generic catalog details. Where project dimensions were required, standard dimensions. Where field verification was required, contract document dimensions carried forward unchanged.
Why it happens. Catalog details are available immediately and project-specific drawings take detailing time. On a compressed schedule the available option gets submitted.
How to detect it. Look for project-specific content: actual dimensions, actual conditions, gridlines and elevations matching the project, adjacent trades’ work shown at interfaces.
What it costs. A review cycle at minimum, and where it is approved, fabrication to standard rather than actual dimensions.
Comment to write. State what project-specific content is required and why the submitted content does not satisfy it. Reference the specification paragraph requiring shop drawings rather than product data.
Issue 10: Unflagged Deviation
What it looks like. A departure from a specified requirement, present in the data, not identified anywhere in the submittal. Most general conditions require the submitter to identify deviations in writing.
Why it happens. Sometimes oversight, when the submitter did not notice the departure either. Sometimes deliberate, on the reasonable calculation that an unflagged deviation may be approved. Both produce the same document.
How to detect it. Requirement-driven comparison. There is no shortcut; an unflagged deviation is only visible against the requirement.
What it costs. High, and asymmetrically. A deviation that reaches the design team unflagged and is approved has arguably been accepted, which weakens your position considerably. Unwinding an approved deviation is a negotiation you enter from behind.
Comment to write. Identify the deviation, note that it was not identified as required by the general conditions, and state the determination. The observation that it was unflagged matters, because it establishes the record.
Issue 11: Substitution Disguised as Product Data
What it looks like. A product that is not a named product and not established as an approved equal, submitted as ordinary product data.
Why it happens. Substitution procedures in Division 01 carry deadlines that have usually passed, required comparative data, and an approval authority. Product data submission is easier and looks the same when it arrives.
How to detect it. Verify the submitted manufacturer against the named manufacturers and against any approved-equal record. Where neither matches, it is a substitution regardless of its label.
What it costs. Moderate to high, and the cost is procedural. If it is reviewed and approved as product data, the specification has effectively been changed without the substitution having been evaluated on cost, performance, warranty, or availability.
Comment to write. Return rather than review. State that the submitted product is not a named product, that the submittal constitutes a substitution request, and that the Division 01 substitution procedure applies. Reviewing it on the merits and then rejecting it is worse than returning it, because it invites an argument about the merits when the issue is procedure.
| FIELD REALITY Most substitution rejections in commercial construction are procedural rather than technical, meaning the request was late or lacked required comparative data. When a substitute is sent via product data, the original submitter can expect an arbitrary rejection that will likely irreparably damage the relationship. The contracting party that sent the data can expect the relationship damage. |
|---|
Issue 12: Missing, Expired, or Wrongly Issued Certification
What it looks like. A certification required by the specification is absent, out of date, issued by an entity other than the one specified, or not project-specific where the specification required it to be.
Why it happens. Certifications are administrative from the submitter’s perspective and are frequently assembled last, from files, by someone who does not know what the specification required.
How to detect it. Check presence, currency, issuer, and scope against the specification. Four quick checks and they are frequently reduced to one, which is presence.
What it costs. Usually a closeout or inspection problem rather than a construction problem. Where the certification conditioned an extended warranty or a licensing requirement, considerably more.
Comment to write. Specify which certification, the currency requirement, the required issuer, and the required scope.
Issue 13: Delegated Design Defects
What it looks like. Four recurring forms. Calculations absent or unsealed. An engineer not licensed in the project jurisdiction. Design criteria that do not match the criteria the contract documents state. A delegation scope that does not match what was delegated.
Why it happens. The delegated engineer works from what the subcontractor provided, which may be the drawings without the structural general notes, or a prior project’s criteria. Jurisdiction licensing is an administrative detail nobody checks until the AHJ does.
How to detect it. Verify four things: seal present, jurisdiction correct, criteria matching the structural general notes and the specification, and scope bounded correctly. That is the entire check, and it explicitly does not include reviewing the engineering analysis, because the delegated engineer is the engineer of record for that element and reviewing the analysis assumes liability for it.
What it costs. Very high, because the failure surfaces at the authority having jurisdiction or at special inspection, both of which sit on the critical path to occupancy.
Comment to write. Identify the specific defect. Criteria mismatches in particular should cite both the criteria used and the criteria required, because the submitter frequently does not know they diverged.
Table 3. Issue 13, Delegated Design Defect Forms
| Defect | Detected By | Consequence | Comment Should State |
|---|---|---|---|
| Calculations absent | Presence check | Cannot verify criteria conformance | Which calculations, per which paragraph |
| Calculations unsealed | Presence check | Not acceptable to AHJ | Seal required; jurisdiction |
| Engineer licensed elsewhere | License verification | Deferred submittal rejected | Jurisdiction requirement |
| Criteria do not match structural notes | Comparison against general notes | Design based on wrong loads | Both the criteria used and required |
| Load combinations differ from specified | Comparison | Under-designed element | The specified combinations |
| Delegation scope exceeds what was delegated | Scope comparison | Engineer designing outside the delegation | The delegated scope boundary |
| Delegation scope narrower than delegated | Scope comparison | Part of the element unengineered | The gap |
| Deferred submittal not identified | Register check | Legal installation gated | The requirement and responsibility |
| Special inspection not identified | Register check | Occupancy gated | The requirement and responsibility |
Issue 14: Conflict With a Previously Approved Submittal
What it looks like. A submittal that complies fully with its own specification section and conflicts with another submittal already approved. Dimensions that do not match. Connection sizes that do not align. Electrical characteristics that disagree. Sequencing that is incompatible.
Why it happens. Each submittal is reviewed against its own section by a reviewer looking at that section. Nobody is looking across the boundary, and the design consultant’s review scope generally does not extend to trade coordination.
How to detect it. The contractor’s screen, comparing against previously approved submittals at interfaces. This is Tier 1 work and it is the check most consistently skipped, which produces the awkward situation where two approved submittals conflict and both parties hold an approval.
What it costs. Moderate to high, discovered in the field. And the position is poor for everyone, because each approval was correctly granted against its own section.
Comment to write. Identify the conflicting approved submittal by number, state the specific conflict, and state which one you propose to govern. This is a coordination comment from the contractor rather than a compliance comment from the design consultant, and getting the routing right matters.
Frequency, Severity, and Who Catches Them
Table 4. Issue Ranking by Expected Cost
| Rank | Issue | Frequency | Severity | Typically Caught By | Why It Ranks Here |
|---|---|---|---|---|---|
| 1 | 5, Non-equivalent tested configuration | Moderate | Very high | Inspector | Caught after installation; multiplied on repetitive work |
| 2 | 1, Silence on a specified criterion | Very high | Variable, up to very high | Commissioning or occupancy | Highest frequency; structurally hard to detect |
| 3 | 13, Delegated design defects | Moderate | Very high | AHJ or special inspector | Gates occupancy |
| 4 | 3, Non-specified test condition | Moderate to high | High | Performance testing | Product performs as stated, not as required |
| 5 | 8, Non-compliant accessory | Moderate | Moderate to high | Inspection or commissioning | Defeats an otherwise compliant assembly |
| 6 | 10, Unflagged deviation | High | Moderate to high | Late, if at all | Approval weakens your position |
| 7 | 14, Conflict with approved submittal | Moderate | Moderate to high | Field | Both parties hold approvals |
| 8 | 6, Superseded edition or expired listing | Moderate | Moderate to high | Inspector or AHJ | Evidence you believed you had |
| 9 | 4, Standard cited without demonstration | High | Moderate | Reviewer, if attentive | Usually catchable |
| 10 | 11, Substitution disguised as product data | Moderate | Moderate to high | Reviewer | Procedural, and catchable |
| 11 | 7, Incomplete component coverage | High | Moderate | Field, at installation | Delay rather than defect |
| 12 | 2, Product family rather than model | High | Moderate | Reviewer | Easily caught |
| 13 | 9, Catalog literature for shop drawings | Moderate | Moderate | Reviewer | Easily caught |
| 14 | 12, Certification defects | High | Moderate | Closeout | Administrative, though occasionally licensing |
The rank emphasizes severity and assessment difficulty, more so than frequency. On the other hand, Issues 2, 9 and 12 occur sufficiently that they get noticed and cost a review cycle. Since Issues 1, 3, 5, 8, and 13 are most likely to be caught downstream, they are highlighted first.
Table 5. Issues by Who Currently Catches Them
| Caught By | Issues | Implication |
|---|---|---|
| Reviewer, reliably | 2, 9, 11, 12 | Already working; no change needed |
| Reviewer, if attentive | 4, 6, 7 | Improvable with a checklist |
| Only requirement-driven review | 1, 3, 10 | Requires a decomposed requirement list |
| Contractor coordination screen | 14 | Requires interface comparison; frequently skipped |
| Inspector or AHJ | 5, 6, 13 | Currently escaping review entirely |
| Commissioning or occupancy | 1, 3, 8 | The most expensive discovery point |
That table is the practical summary of this article. If your review already catches the first row and you want to improve, the highest-return additions are a decomposed requirement list, which addresses issues 1, 3, and 10, and an assembly-equivalence check on rated construction, which addresses issue 5.
Stakeholders
Table 6. Detection Responsibility by Issue
| Issue | Best Positioned to Detect | Why |
|---|---|---|
| 1, Silence | Design consultant with a requirement list; contractor screen for presence | Requires the requirement baseline |
| 2, Product family | Contractor screen | Mechanical identity check |
| 3, Test condition | Design consultant | Requires knowing why the condition was specified |
| 4, Assertion versus demonstration | Design consultant | Requires knowing what the specification demanded as evidence |
| 5, Tested configuration | Design consultant, with the detail in hand | Requires comparing assembly descriptions |
| 6, Edition and currency | Contractor screen | Date checks |
| 7, Component coverage | Contractor screen | Comparison against the specified list |
| 8, Accessory compliance | Design consultant | Requires accessory data review |
| 9, Catalog literature | Contractor screen | Obvious once looked for |
| 10, Unflagged deviation | Both tiers | Requires the requirement baseline |
| 11, Disguised substitution | Contractor screen | Identity against named products |
| 12, Certification defects | Contractor screen | Presence, currency, issuer, scope |
| 13, Delegated design | Design consultant for criteria; contractor for seal and jurisdiction | Split by expertise |
| 14, Conflict with approved submittal | Contractor screen | Only party seeing all approvals |
Eight of the fourteen are best caught by the contractor’s screen, and none of those eight requires engineering judgment. That is the single most useful observation in this article for a contractor, because it means most of this is available without anyone’s cooperation.
Required Documentation
Table 7. Documents Required to Detect These Issues
| Document | Issues | Consequence If Absent |
|---|---|---|
| Decomposed requirement list | 1, 3, 10 | The three hardest issues undetectable |
| Complete specification section | All | No basis for any determination |
| Drawings and detailed assemblies | 5, 14 | Assembly equivalence unverifiable |
| Referenced standards or access | 4, 5, 6 | Standard compliance accepted on assertion |
| Division 01 substitution procedures | 11 | Substitutions reviewed on the merits |
| Structural general notes with design criteria | 13 | Criteria mismatches invisible |
| Register of named and approved-equal products | 11 | Cannot identify a substitution |
| Previously approved submittals at interfaces | 14 | Conflicts between approvals invisible |
| Listing directories for rated assemblies | 5, 6 | Cannot verify or find a matching listing |
| Project deviation register | 10 | Accepted deviations undocumented |
Technology Integration
Table 8. Detection by Method
| Issue | Manual Review | Checklist-Driven Review | AI Extraction and Comparison |
|---|---|---|---|
| 1, Silence | Rarely detected | Detected with a requirement list | Detected systematically |
| 2, Product family | Usually detected | Detected | Detected |
| 3, Test condition | Rarely detected | Detected if the condition is on the list | Detected where conditions are captured |
| 4, Assertion versus demonstration | Sometimes | Detected | Detected |
| 5, Tested configuration | Rarely | Detected if the check is on the list | Partially; assembly comparison is judgment-heavy |
| 6, Edition and currency | Sometimes | Detected | Detected |
| 7, Component coverage | Sometimes | Detected | Detected |
| 8, Accessory compliance | Rarely | Detected if accessories are on the list | Detected where accessory criteria are extracted |
| 9, Catalog literature | Usually detected | Detected | Partially |
| 10, Unflagged deviation | Rarely | Detected with a requirement list | Detected systematically |
| 11, Disguised substitution | Usually detected | Detected | Detected |
| 12, Certification defects | Usually detected | Detected | Detected |
| 13, Delegated design | Sometimes | Detected | Criteria comparison detected; scope judgment human |
| 14, Conflict with approved submittal | Rarely | Detected if the check is on the list | Partially |
The pattern is that a checklist closes most of the gap and automated comparison closes more of it on the issues that are fundamentally about matching values against requirements. Issue 5 resists both, because comparing a tested assembly description against a drawn detail is a technical judgment about equivalence.
AI-Assisted Opportunities
Three issues on this list are detection problems at volume rather than judgment problems, and they happen to include the highest-frequency one.
Issue 1 requires establishing that a value is absent from a document that may run two hundred pages, for each of forty to eighty requirements. Establishing absence means exhausting the document, which is precisely what human review cannot do reliably at speed.
Issue 3 requires capturing the condition alongside every value and comparing both.
Issue 10 requires comparing every submitted attribute against every specified requirement to find departures the submitter did not identify.
Platforms built to interrogate construction documents address all three. iFieldSmart AI, for instance, extracts specification requirements with citations to the source and compares submitted documents against them, flagging missing and non-compliant content. Detecting silence is the capability that matters most here, because it is the failure mode a reviewer structurally cannot catch by reading.
Table 9. AI Use Case Matrix by Issue
| Issue | AI Contribution | Human Judgment Retained |
|---|---|---|
| 1, Silence | Flags every requirement with no corresponding submitted value | Materiality of each absence |
| 2, Product family | Flags absence of a specific model designation | Whether the identification is adequate |
| 3, Test condition | Flags values stated at conditions other than specified | Whether the difference is material |
| 4, Assertion versus demonstration | Flags compliance claims without supporting evidence | Whether the specification required evidence |
| 5, Tested configuration | Retrieves assembly descriptions for comparison | The equivalence determination |
| 6, Edition and currency | Compares cited editions and dates | Whether the edition change is material |
| 7, Component coverage | Compares against the specified component list | Whether an omission is material |
| 8, Accessory compliance | Extracts accessory criteria and compares | Technical adequacy of the accessory |
| 9, Catalog literature | Partially; detects generic content | Whether project-specific content is adequate |
| 10, Unflagged deviation | Flags departures whether or not identified | Classification and acceptability |
| 11, Disguised substitution | Compares against named and approved products | Whether an approved-equal basis exists |
| 12, Certification defects | Checks presence, dates, and issuer against requirements | Scope adequacy |
| 13, Delegated design | Compares criteria used against criteria required | Scope boundary and engineering judgment |
| 14, Conflict with approved submittal | Compares attributes across approved submittals | Which should govern |
| IMPORTANT Automated comparison produces a list of flagged discrepancies, and on a substantive submittal that list will contain many immaterial items alongside the few that matter. The reviewer’s work shifts from finding to triaging, which requires more experience rather than less. A forty-row discrepancy report that nobody triages is worse than a shorter manual review somebody actually reasoned through. |
|---|
Implementation
- Decompose one high-consequence specification section into a requirement list. Issues 1, 3, and 10 are collectively the majority of what goes undetected. These issues are dealt with.
- Add an assembly-equivalence check to every rated assembly submittal. Compare the tested assembly description against the drawn detail component by component. This addresses issue 5, the highest-cost item on the list.
- Institute the contractor screen for the eight issues it can catch without engineering judgment: 2, 6, 7, 9, 11, 12, plus the seal and jurisdiction portions of 13, and the interface comparison in 14.
- Add a condition column wherever you record a performance requirement, so issue 3 becomes visible.
- Verify accessory compliance separately from base product compliance as a standing habit, addressing issue 8.
- Compare every submittal against previously approved submittals at interfaces, addressing issue 14.
- Start a project deviation register and record every accepted deviation, addressing the aftermath of issue 10.
- At project completion, list every compliance failure found after approval and classify it by issue number. The distribution tells you which detection to strengthen.
Table 10. Implementation Roadmap
| Phase | Weeks | Activity | Issues Addressed | Gate |
|---|---|---|---|---|
| Requirement lists | 1 to 4 | Decompose highest-consequence sections, capturing conditions | 1, 3, 10 | Top 10 sections done |
| Assembly equivalence | 2 to 3 | Standing check on all rated assemblies | 5 | Applied to every rated submittal |
| Contractor screen | 3 to 6 | Eight-issue screen instituted | 2, 6, 7, 9, 11, 12, 13 partial, 14 | Nothing forwarded unscreened |
| Accessory discipline | 4 to 6 | Separate accessory verification | 8 | Applied to all assemblies |
| Interface comparison | 4 to 8 | Comparison against approved submittals | 14 | Applied at every interface |
| Deviation register | 5 to 7 | Standing register established | 10 | Every acceptance recorded |
| Measure | 8 to 14 | Classify findings by issue; report monthly | All | Reported to project manager |
| Improve | Ongoing | Post-approval failures classified by issue | All | Two consecutive project reviews |
Best Practices
Table 11. Practices That Catch These Issues
| Practice | Issues Addressed |
|---|---|
| Work from a decomposed requirement list, not from the submittal | 1, 3, 10 |
| Capture the test condition with every performance requirement | 3 |
| Compare the tested assembly against the drawn assembly, component by component | 5 |
| Check edition and currency on every report, listing, and certificate | 6, 12 |
| Require a specific model and configuration, never a series | 2 |
| Determine whether the specification requires evidence or accepts assertion | 4 |
| Verify accessory compliance separately from base product compliance | 8 |
| Compare submittal contents against the specified component list | 7 |
| Verify identity against named and approved-equal products before reviewing | 11 |
| Return substitutions rather than reviewing them on the merits | 11 |
| Compare against previously approved submittals at every interface | 14 |
| Verify delegated design criteria against the structural general notes | 13 |
| Check the engineer’s licensing jurisdiction | 13 |
| Flag deviations the submitter did not identify, and note that they were unflagged | 10 |
| Record every accepted deviation in a standing register | 10 |
| Write comments naming the paragraph, the specified value, and the submitted value | All |
Common Mistakes in Responding to These Issues
Table 12. Responses That Do Not Work
| Response | Why It Fails | What Works |
|---|---|---|
| “Insufficient information provided” | Submitter cannot tell which information | Name the paragraph and the missing value |
| “Does not comply with specification” | Submitter guesses at the requirement | Cite the specified value and the submitted value |
| Rejecting rather than requesting on issue 1 | Not-addressed needs information, not a different product | Distinguish not-addressed from non-compliance |
| Reviewing a substitution on the merits | Invites an argument about merits when the issue is procedure | Return; cite the Division 01 procedure |
| Commenting on the assembly when one accessory fails | Resubmittal changes the wrong thing | Name the accessory and its requirement |
| Assuming the submitter knows why the condition matters | They frequently do not | State the specified condition and request data at it |
| Assuming the submitter can find the matching listing | Listing directories are large and technical | Identify the specific difference you found |
| Treating repeated rejection as submitter persistence | Usually a requirement ambiguity or an unstated standard | Third revision triggers a conversation |
| Approving with a note where the criterion is material | Accepts an unknown | Require the information |
| Recording an accepted deviation only on the returned submittal | Disputed at inspection with no record | Standing deviation register |
How These Issues Appear in Different Sectors
Class A office tower, 340,000 square feet. Issue 3, in its air-and-water form. Curtain wall test reports were to the correct standards at the standards’ minimum test pressures rather than at the project design pressure. Every value complied with the standard cited; none complied at the project’s condition. The submitter had matched standards correctly and conditions not at all.
Regional hospital expansion, 190 beds, occupied campus. Issue 1, on licensing criteria. A submitted air handling unit was silent on final filter efficiency and on the pressure differential capability required for protective environment rooms. The manufacturer’s data sheet did not publish either at the relevant condition, which is why the submittal was silent rather than wrong. On healthcare work, these are licensing conditions.
Hyperscale data center, 60 MW. Issue 3 on electrical distribution. Switchgear short-circuit current rating was stated at a bracing configuration other than the furnished one. The rating was real and did not apply. On equipment with lead times measured in tens of weeks, catching it after fabrication release would have been measured in months.
Advanced manufacturing facility, food grade. Issues 7 and 8 together. Hygienic panel submittals omitted the joint sealant, and when it arrived by supplement, the sealant was not compatible with the facility’s sanitizing chemistry. A component coverage gap followed by an accessory compliance failure, both of which would have surfaced at sanitation validation.
University Science Building, 11 floors. Issue 3 on fume hood containment, at a different face velocity than specified. In a laboratory, that difference is a personnel safety matter, and the submitted value looked compliant because it was measured under easier conditions.
Multifamily podium, 240 units. Issue 5, the expensive one. A floor-ceiling assembly listing used different framing spacing and a different resilient channel than the detailed assembly. Same rating, different assembly. On a project with hundreds of repetitions, discovery at inspection would have been severe, and the submitter had reasonably believed a matching rating meant a matching assembly.
Highway interchange, design-bid-build. Issue 12, in the form that costs money rather than time. Mill certificates were not traceable to delivered heat numbers, which on unit-price work is a condition of acceptance and payment.
Historic warehouse conversion to hotel, 140 keys. Issues 1 and 14. Window assemblies provided thermal data and no acoustic data, and the submitted profile conflicted with the previously approved historic restoration details. The acoustic silence mattered on a hotel adjacent to a rail line, and the profile conflict was two approvals in disagreement.
Frequently Asked Questions
Which of these issues costs the most?
Issue 5, non-equivalent tested configuration, per occurrence, because it is typically caught by an inspector after installation and because it multiplies on repetitive construction. Issue 1, silence on a specified criterion, costs the most in aggregate because it is by far the most frequent and because its severity distribution has a long tail. Issue 13 has the worst schedule profile, because delegated design defects surface at the authority having jurisdiction.
Why is silence so much more common than stated non-compliance?
Because manufacturer literature is written for a market rather than for your specification. A data sheet publishes the attributes buyers commonly ask about at the conditions the industry commonly uses. Where your specification asked for something else, the literature is complete and does not contain it. The submitter forwarded a complete package, and the absence is invisible to everyone who does not have a list of what was required.
How do we catch the issues that currently reach inspectors?
Two specific additions. An assembly-equivalence check on every rated assembly submittal, comparing the tested assembly description against the drawn detail component by component, which addresses issue 5. And a four-point delegated design check covering seal, jurisdiction, criteria, and scope, which addresses issue 13. Both are checks rather than judgments, and both are currently absent from most review processes.
Should the contractor be catching these, or the design team?
Eight of the fourteen are best caught by the contractor’s screen and none of those eight requires engineering judgment: product family, edition and currency, component coverage, catalog literature, disguised substitution, certification defects, the seal and jurisdiction portions of delegated design, and conflicts with previously approved submittals. The remaining six need the design consultant, because they require knowing why a criterion was specified. That split is worth acting on, because it means most of this is within the contractor’s control.
What is the difference between not-addressed and non-compliant in practice?
The remedy. Not-addressed needs information, and the resubmittal should be narrow and fast. Non-compliant needs a different product or configuration, and the resubmittal is substantial. Comments that conflate them produce resubmittals that change the wrong thing, which is a leading cause of resubmittal spirals.
How do we handle a submitter who genuinely cannot obtain the specified data?
It happens, particularly on issue 3 where a manufacturer does not test at the specified condition. The options are to accept data at the nearest available condition with a documented determination of the difference, to require independent testing at the specified condition, or to raise it as a specification issue with the design team. What does not work is repeated rejection, because the submitter cannot produce what does not exist.
Are any of these issues the specification’s fault rather than the submittal’s?
Frequently, and it is worth being honest about it. A section citing ten standards without stating which require test reports produces issue 4. A section stating a performance criterion without a test condition produces issue 3 and cannot be verified. A section requiring “compliance with the referenced standards” produces issue 5. Where the same issue recurs across multiple submittals on one section, suspect the section.
How should repeated occurrences of one issue be handled?
As a pattern rather than as instances. If three submittals on the same section all exhibit issue 1 on the same criterion, the criterion is probably stated in a way submitters do not see, or the data does not exist in the market. Address it once with the design team and the subcontractor rather than three times through rejection notes.
Does approving one of these issues create liability?
It changes positions rather than creating liability cleanly. An approved non-compliance that was visible in the submitted data weakens the reviewer’s position and does not fully protect the submitter, who submitted it. An approved deviation that was flagged and accepted is a documented decision, which is the outcome you want. An approved deviation that was unflagged and unnoticed is the worst case for everyone, which is why issue 10 matters more than its frequency suggests.
Which issue should we address first?
Issue 1, by building requirement lists, because it is the most frequent and the hardest to catch. Issue 5 second, by adding the assembly-equivalence check, because it is the most expensive per occurrence. Both are additions to the process rather than changes to it, and neither requires anyone outside your organization to agree to anything.
How do these differ on renovation work?
Issue 3 and issue 5 both intensify, because “match existing” specifications often reference performance the existing condition never achieved, and because assemblies incorporating existing construction are almost never identical to any tested assembly. On renovation, the honest answer on rated assemblies is frequently that no matching listing exists and an engineering judgment from a qualified party is required, which should be identified during review rather than at inspection.
What is the best single indicator that our review is missing these?
Count not-addressed determinations per Tier A submittal. If the number is near zero, the review is not detecting issue 1, which is statistically improbable rather than good news. A healthy requirement-driven review on a substantive submittal typically finds several criteria unaddressed, and finding none usually means nobody was looking for absence.
Expert Recommendations
Build requirement lists for your highest-consequence sections and capture the test condition with every performance criterion. That single artifact addresses the three issues that currently escape review entirely.
Add an assembly-equivalence check to every rated assembly submittal, and make it explicit: does the tested assembly match the drawn assembly, component by component. It is the highest-cost issue on this list and the one inspectors catch.
Institute the eight-issue contractor screen. None of those eight requires engineering judgment, all are within your control, and together they prevent a substantial share of the rejections that consume full review cycles.
Verify accessory compliance separately, as a habit, on every assembly. A compliant product with a non-compliant accessory is non-compliant, and the accessory is where nobody looks.
Return substitutions rather than reviewing them. Reviewing on the merits and then rejecting invites an argument about merits when the issue is procedure, and it damages a relationship unnecessarily.
Count not-addressed determinations and report the number monthly. It is the clearest available evidence that your review is detecting the dominant failure mode rather than only the visible one.
Classify every post-approval compliance failure by issue number at closeout. The distribution will be specific to your market and your project types, and it will tell you which two checks to strengthen rather than leaving you to guess.
Where the Taxonomy Leaves You
Fourteen issues, and the five that cost the most are the five that a recognition-based review structurally cannot find. That is the whole story of this article. They are not subtle judgments requiring rare expertise. They are absences, condition mismatches, and equivalence questions, and each of them is detectable by a specific check that most review processes do not include.
The framing worth carrying is that almost none of these arise from bad faith. A submitter forwarding a complete manufacturer package that happens not to state your acoustic criterion has done nothing wrong from where they are standing. Which means the remedy is not vigilance about people, it is a small number of checks about documents.
Build one requirement list and add the assembly-equivalence check to your rated assembly reviews. Those two actions address the highest-frequency issue and the highest-cost issue, and both are available on the project you are running now.
