Construction Submittal Review Checklist

A submittal review checklist organized by workflow tells you when to review. A checklist organized by attribute class tells you what to look at, which is the part reviewers actually get wrong.

The distinction matters because submittal review fails by omission rather than by error. Reviewers rarely approve a value they can see is wrong. They approve packages where an entire class of requirement was never examined: the acoustic criterion nobody checked, the accessory nobody counted, the test condition nobody compared, the referenced standard everybody assumed was satisfied. Each of those is an attribute class, and a review that does not walk the classes deliberately will walk whichever ones the reviewer habitually notices.

This checklist is organized that way. Thirteen parts, each covering one class of verification, from identity through disposition quality. Run the relevant parts by depth tier rather than running all thirteen on everything, because a certificate of compliance does not need a dimensional coordination check and a curtain wall submittal needs all of them.

One boundary before you start. This checklist verifies whether the submitted content complies with what was specified. Whether the submittal is on time, who holds it, and what it gates in the procurement chain are separate questions handled by the submittal log, and they are not addressed here. Different instrument, different failure modes.

How to Use This Checklist

Have the requirement list before you start. The single largest determinant of review quality is whether the reviewer is working from a decomposed list of specified requirements or from memory. Parts C through F are close to useless without one.

Assign a depth tier and run the matching parts. Table 1 maps them. Running all thirteen parts on a certificate wastes an hour; running four on a life safety submittal misses the point.

Every check produces a determination, not a checkmark. Four values: complies, does not comply, not addressed, not applicable. The third is the one that catches what recognition review misses.

Record where you found the value. A determination without a location cannot be verified by anyone else and makes the resubmittal backcheck a full re-read.

Table 1. Parts by Depth Tier

Tier A covers life safety, structural, code-consequential, delegated design, and long-lead equipment. Tier B covers substantive equipment and assemblies without regulatory consequence. Tier C covers standard products with few specified criteria. Tier D covers certificates, qualification statements, and informational submittals.

Key Definitions

Requirement. One verifiable criterion extracted from the contract documents.

Determination. The per requirement outcome is: complies, does not comply, not addressed, or not applicable.

Not addressed. The submittal is silent on a specified requirement. Distinct from non-compliance and considerably more common.

Attribute class. A category of thing being verified. Identity, performance, material, standard compliance, component completeness, and so on. The organizing principle of this checklist.

Test condition. The circumstances under which a performance value was measured. A value at the wrong condition is not compliance.

Tested configuration. For assembly-based standards, the specific assembly that was tested. Compliance attaches to that configuration, not to similar ones.

Deviation. A departure from a specified requirement, whether or not the submitter identified it.

Tier 1 and Tier 2. The contractor’s completeness and requirement-presence screen, and the design consultant’s technical conformance review.

Objectives

• Verify every class of requirement rather than the classes the reviewer habitually notices
• Detect silence on specified criteria, which is the dominant failure mode
• Confirm that referenced standards are complied with rather than merely cited
• Produce determinations specific enough to be verified and defended
• Keep the review inside the scope boundary in both directions
• Make resubmittal backchecks fast rather than repetitive

Part A: Review Preparation

Five checks before opening the submittal. They take three minutes and they determine whether the review is capable of finding anything.

Table 2. Review Preparation Checks

Part B: Identity Verification

Establish that the submitted item is the specified item. Every determination downstream is provisional until this is settled.

Table 3. Identity Verification Checks

Check B.6 is a return condition rather than a comment condition. A different product submitted as product data has bypassed the substitution procedure, and reviewing it as product data means an approval effectively changes the specification without the substitution ever having been evaluated.

Part C: Performance Attribute Verification

The class where silence does the most damage. Work the requirement list, not the submittal.

Table 4. Performance Attribute Verification Checks

Check C.16 is the discipline that makes the whole part defensible. A reviewer who knows a product achieves a requirement is not entitled to determine compliance on that basis, because the submittal is the contractual record. If the value is not in the package, the determination is not addressed.

Part D: Material, Physical, and Construction Verification

Table 5. Material and Physical Verification Checks

Check D.13 recurs on assemblies. A compliant sealant and a compliant panel can be materially incompatible, and specifications frequently require compatibility confirmation that submittals omit.

Part E: Referenced Standard Verification

The class that separates a technical review from a documentary one.

Table 6. Referenced Standard Verification Checks

Table 7. Standard Verification by Standard Type

Check E.5 deserves the emphasis. On assembly-based standards, compliance attaches to the tested assembly. A report for a similar assembly is not evidence for the furnished one, and this is the finding most likely to be caught by an inspector rather than by a reviewer.

Part F: Component and Accessory Completeness

Table 8. Component Completeness Checks

Check F.8 catches a specific and common failure: a compliant base product with a non-compliant accessory. The accessory is frequently a smaller line item that receives no scrutiny, and on rated assemblies or performance systems it can defeat the whole assembly.

Part G: Documentation and Evidentiary Completeness

Table 9. Documentation Completeness Checks

Check G.13 is more consequential than it looks. A submittal consisting of unmarked catalog pages covering twelve models has not identified which one is being furnished, and an approval arguably approves whatever the submitter later claims was intended.

Part H: Dimensional and Coordination Verification

Primarily a contractor Tier 1 responsibility, since the design consultant’s review generally does not extend to field dimensions or trade coordination.

Table 10. Dimensional and Coordination Verification Checks

Check H.10 is the one most consistently skipped, and it produces the awkward situation where two approved submittals conflict with each other. Neither party who approved them was looking across the boundary, and the conflict is discovered in the field with both parties holding an approval.

Part I: Code, Regulatory, and Life Safety Verification

Fewer checks, higher consequence. These findings gate occupancy rather than cost money.

Table 11. Code and Regulatory Verification Checks

Check I.12 applies on healthcare, laboratory, food processing, and childcare facilities, where a product attribute can be a licensing condition rather than a performance preference. Filtration efficiency, pressure differential capability, surface cleanability, and gas system certification are the recurring examples.

Part J: Delegated Design Verification

The narrowest scope in this checklist, and getting the narrowness right matters.

Table 12. Delegated Design Verification Checks

Check J.10 is stated as a prohibition deliberately. The delegated engineer is the engineer of record for that element. A reviewer who checks the analysis has stepped across the scope boundary and assumed responsibility that the contract placed elsewhere. Verify criteria conformance, not engineering adequacy.

Part K: Deviation Identification

Table 13. Deviation Identification Checks

Check K.6 catches something a per-requirement method can obscure. Three individually minor deviations can compound into a substantive one, particularly on assemblies where tolerances stack or where several small performance shortfalls accumulate.

Part L: Disposition and Comment Quality

Run before returning the submittal. This part determines whether the review produces action or argument.

Table 14. Disposition and Comment Quality Checks

Part M: Resubmittal Backcheck

A resubmittal is not a new review.

Table 15. Resubmittal Backcheck Checks

Check M.4 catches a real failure. A resubmittal that changes model number invalidates every determination made against the previous model, and reviewing only the commented items means approving a product that was never fully reviewed.

Role-Specific Emphasis

The same checklist reads differently depending on who is running it.

Table 16. Emphasis by Reviewing Role

The Tier 1 row is the one worth internalizing. The contractor’s screen is about requirement presence, completeness, deviation identification, and coordination. It is not a technical conformance review, and a contractor who renders technical determinations has assumed a responsibility the contract placed on the design consultant.

Importance: What These Checks Prevent

Compliance failures found in submittal review cost a comment. The same failures found later cost progressively more: at fabrication, a change order; at delivery, a restocking charge and a delay; at installation, rework; at inspection or commissioning, a schedule event with occupancy exposure; after occupancy, a warranty claim or a defect action.

The categories that dominate are consistent, and every one of them has a check on this list. A performance criterion the submittal never addressed. A test report for a similar rather than identical assembly. A compliant product in a non-compliant configuration. An accessory nobody verified. A deviation nobody flagged.

The second effect is evidentiary. Inspections, commissioning disputes, and defect claims all eventually turn on whether what was installed matched what was specified and who approved it. A review that recorded determinations is evidence. A review that recorded a stamp is not.

Stakeholders

Table 17. Who Runs Which Part

Required Documentation

Table 18. Documents Required to Run This Checklist

Technology Integration

Table 19. Checklist Execution by Method

Parts H and I move least. Coordination is a geometric and sequencing judgment, and code interpretation is a judgment about how a jurisdiction will read a requirement. Both remain human.

AI-Assisted Opportunities

Parts C through G are search and comparison at volume, which is exactly where manual review breaks down. A substantive submittal may run two hundred pages of manufacturer literature, and the reviewer’s task is to locate a specific value for each of forty to eighty requirements, or to establish that it is absent. Establishing absence is the hard part, because it requires exhausting the document rather than finding something in it.

Platforms built to interrogate construction documents address this directly. iFieldSmart AI, for instance, extracts specification requirements with citations to the source and compares submitted documents against those requirements, flagging missing and non-compliant content. For this checklist specifically, the capability that matters is detecting silence, since that is the determination a human reviewer structurally cannot make reliably at speed.

Table 20. Checklist Parts Mapped to AI Capability

Implementation

• Decompose one high-consequence specification section into a requirement list. Without this, Parts C through F cannot run, and this is the step that makes the rest possible.
• Assign depth tiers across the register so the right parts run on the right submittals.
• Adopt the four-determination vocabulary and use “not addressed” explicitly. Vocabulary drives attention.
• Institute the Tier 1 screen with Parts A, B, F, G, H, and K. No engineering judgment required, and it prevents a meaningful share of Tier 2 rejections.
• Standardize the comment format from check L.4 and enforce L.6, the not-addressed versus non-compliance distinction.
• Start the project deviation register at check K.8, and record every accepted deviation there.
• Retain compliance matrices for Tier A and B submittals so backchecks are fast and the compliance record exists.
• At project completion, list every compliance failure discovered after approval and identify which part would have caught it. Revise the checklist.

Table 21. Implementation Roadmap

Best Practices

Table 22. Checklist Discipline Best Practices

Common Mistakes

Table 23. Common Review Failures

Applied Examples

Class A office tower, 340,000 square feet. Check E.5 and the air-and-water row of Table 7. Curtain wall test reports were to the correct standards for a similar system at the standard’s minimum test pressure rather than at the project design pressure. Every value complied with the standard cited and none complied at the project’s condition. Found in review it was a comment; found at mockup testing it is a redesign.

Regional hospital expansion, 190 beds, occupied campus. Checks C.1 and I.12. A submitted air handling unit was silent on final filter efficiency and on the pressure differential capability required for protective environment rooms. Both were not-addressed determinations on licensing criteria, which on healthcare work means occupancy rather than performance.

Hyperscale data center, 60 MW. Check C.2 on electrical distribution. The submitted switchgear stated its short-circuit current rating at a configuration other than the one being furnished. Correct product, wrong condition, and a safety determination on equipment with a lead time measured in tens of weeks.

Advanced manufacturing facility, food grade. Checks F.1 and D.13. Hygienic panel submittals addressed the panel and omitted the joint sealant, which the specification required to be compatible with the facility’s sanitizing chemistry. A component completeness finding that would have surfaced at sanitation validation.

University science building, 11 floors. Check C.2 on fume hood containment. Data was to the correct standard at a different face velocity than specified. In a laboratory that difference is a personnel safety matter, and the submitted number looked compliant.

Multifamily podium, 240 units. Check E.5 on a floor-ceiling assembly. The cited listing used different framing spacing and a different resilient channel than the detailed assembly. On a project with hundreds of repetitions, correction after installation would have been severe.

Highway interchange, design-bid-build. Check E.8. Mill certificates were not traceable to delivered heat numbers, which on unit-price work is a condition of acceptance and payment rather than a schedule matter.

Historic warehouse conversion to hotel, 140 keys. Checks C.1, E.6, and H.11. Window assemblies provided thermal data, omitted acoustic data entirely, and showed a profile differing from the approved historic detail. The acoustic omission was important at the hotel across the rail line, and it was a not-addressed determination as opposed to a visible case of non-compliance.

FAQ’s

How long does each checklist submission take?

By tier. Tier A with all applicable parts runs 45 to 120 minutes. Tier B is 20 to 45. Tier C is 10 to 20. Tier D is about five minutes. The variable that matters most is whether the requirement list exists, which turns Parts C through F from open-ended reading into a finite comparison.

Which parts should we adopt first?

Part A, Part B, and Part K, plus the L.4 comment format. Preparation determines whether the review can find anything, identity determines whether anything else means anything, deviation identification protects you contractually, and the comment format determines whether findings produce action. That combination requires no engineering judgment and is available immediately.

How is this different from a submittal log checklist?

Different question entirely. A log checklist asks whether the register is complete, whether items are scheduled against need dates, and whether anything is stalled. This checklist asks whether the submitted content complies with what was specified. A project can have an exemplary log and superficial reviews, which is a common combination, because the log is visible in meetings and the review quality is not.

Can a contractor render compliance determinations?

A contractor should render determinations on requirement presence, completeness, identity, deviation identification, coordination, and field dimensions. Technical conformance determinations belong to the design consultant of record, and a contractor who renders them has assumed a responsibility the contract assigned elsewhere. Table 16 sets the split.

What if we cannot obtain a referenced standard?

Record the limitation in the determination and in the disposition: compliance with that standard was accepted on the submitter’s certification rather than independently verified. That is a defensible position. Treating the reference as satisfied without saying so is not, and it is the more common practice.

How do we handle a submittal covering many items at once?

Check B.8. Identify each item separately and determine each against its own requirements. Composite submittals are where a single non-compliant item hides, because the review forms an overall impression of a package rather than a determination per item. Where the volume is unmanageable, return it and require separate submittals.

Should reviewers use the manufacturer’s own compliance statement?

It depends on what the specification requires. Where the specification requires certification or test reports, a manufacturer’s statement is an assertion and the determination is not addressed. Where the specification requires compliance without specifying the evidence, an assertion is generally acceptable, and it is worth recording that the determination rests on it. That distinction matters if the product later underperforms.

What is the most commonly missed check on this list?

E.5, the tested configuration against the furnished configuration, particularly on fire-rated assemblies. Reviewers verify that a listing exists and that its rating matches, and do not verify that the tested assembly matches the detailed one. That is the finding an inspector catches, at the worst possible point in the schedule.

How do we prevent this from becoming a bureaucratic exercise?

Tier it honestly and report one metric. Running all thirteen parts on everything is what makes it bureaucratic, and it also makes it unsustainable. Report not-addressed findings per Tier A submittal monthly; that single number demonstrates the method is finding things recognition review would not have, which is what keeps it alive.

Does an approval based on this checklist protect us?

Better than a stamp does, considerably. A documented determination per requirement, with locations recorded, establishes what was reviewed and on what basis. It does not extend the scope boundary, so it offers no more protection on field dimensions or means and methods than any other approval. What it changes is your position where a compliance question later arises, because you can show what was verified.

How should this apply to resubmittals of previously approved products?

Part M, and specifically checks M.3 and M.4. If the model or configuration changed, prior determinations against the earlier product no longer hold and the affected requirements need re-verification. Reviewing only the commented items on a resubmittal that changed the product is how an unreviewed item gets approved.

Which parts matter most on renovation work?

Part H, dimensional and coordination, because existing conditions rather than contract documents govern the fit, and field verification is frequently required and frequently unperformed. Also Part E, because “match existing” specifications often reference standards or performance the existing condition never met, and the submittal has to be judged against the stated requirement rather than the existing condition.

Expert Recommendations

Decompose the specification sections for your five highest-consequence packages before submittals arrive. Everything in Parts C through F depends on that list existing, and building it under review pressure does not happen.

Adopt “not addressed” as a determination and count it. It is the category that catches the dominant failure mode, and counting it per submittal proves the method is working.

Institute the Tier 1 screen with Parts A, B, F, G, H, and K. It needs no engineering judgment, it is entirely within the contractor’s control, and it prevents a meaningful share of Tier 2 rejections that each cost a full review cycle.

Enforce the comment format from check L.4 without exception. Requirement, specified value, submitted value, location. It ends debates and it produces targeted resubmittals rather than guesses.

Start a project deviation register on day one and record every accepted deviation in it. An accepted deviation nobody wrote down becomes an argument at inspection, and the person standing there will not be whoever accepted it.

Make check E.5 a standing habit on every rated assembly. Tested configuration against furnished configuration. It is the finding most likely to be caught by an inspector and least likely to be caught by a reviewer.

Brief your project engineers on the review limitation clause, and on Table 16. Most of them believe an approval protects more than it does and that their own review scope is narrower than it is.

Using This From Here

Submittal review is one of the few construction activities where method reliably beats experience. A reviewer working from a decomposed requirement list and walking attribute classes will outperform a veteran reviewing by recognition, because the veteran’s expertise tells them what a correct submittal looks like and the method tells them what this specification actually required.

The two changes that produce most of the benefit are having the requirement list before you open the package, and recording “not addressed” when the package is silent. Everything else on this list is refinement.

Decompose one section, run Parts A through G on the next substantive submittal that arrives, and count the not-addressed determinations. On most first attempts that number is high enough to settle the question of whether the method is worth the time.