Best Practices for Construction Submittal Reviews

Submittal review can fail in three directions, and almost all published guidance addresses only one of them.

The familiar failure is under-review: the rubber stamp, the four-minute approval, the package nobody compared against the specification. It gets the attention because its consequences are visible and expensive.

The second failure is over-review, and it is barely discussed. A consultant who comments on erection sequence has stepped into the contractor’s responsibility and created a claim that would not otherwise exist. A reviewer who checks a delegated engineer’s analysis has assumed liability the contract placed elsewhere. A review returning forty comments, thirty-two of them preferences, has burned a review cycle and taught the submitter that comments are negotiable. Over-review feels like diligence and behaves like damage.

The third failure is mis-review: substantial effort spent verifying the wrong things. A reviewer who spends forty minutes confirming products they already recognize and never checks the tested assembly configuration has worked hard and found nothing that mattered.

Good submittal review is not maximum review. It’s calibrated review time and calibration happens along four dimensions. First, how deep? Next, how broad? Third, what counts as proof? Lastly, how do you communicate your findings? These four dimensions describe the calibration process. This article organizes practice around those four, plus the organizational and relationship conditions that make calibration possible.

What follows assumes you already have a review process and want it to work better. The mechanics of verification and the specific things to check are separate subjects.

The Three Failure Directions

Table 1. How Submittal Review Fails

Under-review and over-review are not opposite ends of one scale, which is the common misconception. They coexist routinely on the same project and frequently in the same review: a reviewer who comments extensively on the fabrication approach and never checks whether the submitted assembly matches the tested one has done both simultaneously.

Mis-review is the quietest of the three because it produces activity. A file full of marked-up submittals looks like a functioning review process regardless of whether the marks were on the attributes that mattered.

Core Terminology

Calibration. Matching review depth, scope, evidentiary standard, and communication to the consequence of the submittal.

Depth. How thoroughly a submittal is verified. Correctly variable across a register.

Scope boundary. The line between what the reviewer is responsible for evaluating and what remains the contractor’s. Stated in the general conditions.

Evidentiary standard. What the reviewer accepts as proof that a requirement is met. Assertion, certification, test report, or independent verification.

Preference comment. A comment recording what the reviewer would have done differently, where the submittal complies with the documents. The fastest way to have a whole review discounted.

Two-tier review. The contractor’s completeness and requirement-presence screen, then the design consultant’s conformance review.

Review capacity. The finite reviewer hours available in a period. A shared resource whose loading the contractor controls and whose throughput the design team controls.

Why Over-Review Deserves Attention

This is worth developing, because the industry has almost no literature on it and the consequences are real.

It transfers responsibility. Most general conditions state that the design consultant reviews for conformance with the design concept and the contract documents, while the contractor remains responsible for field dimensions, quantities, fabrication means and methods, construction safety, and trade coordination. A reviewer who comments on erection sequence has offered direction in an area the contract assigned to the contractor. If the contractor follows it and something goes wrong, a claim exists that would not have existed had the reviewer stayed inside the boundary.

It creates reliance. Detailed comments on dimensions invite the contractor to rely on the review for dimensional accuracy, which the review limitation clause says they cannot do. That mismatch between what the clause says and what the review behaved like is exactly where disputes originate.

It consumes capacity that consequential submittals needed. A reviewer spending ninety minutes on a standard product with three specified criteria is not reviewing the life safety submittal in the queue behind it.

It erodes comment credibility. This is the most immediately damaging effect. A review returning thirty-two preference comments alongside eight compliance findings teaches the submitter that comments are opinions to be negotiated. On the next submittal, the eight findings that matter get the same treatment as the thirty-two that did not.

It delays procurement for non-compliance-related reasons. A resubmittal required because the reviewer preferred a different arrangement, on a compliant product, costs two review durations and a fabrication delay for no compliance benefit.

Table 2. Over-Review Patterns and What They Cost

Two of those rows deserve nuance rather than prohibition. A reviewer who notices a genuine safety hazard should absolutely raise it, and the way to do it without assuming the responsibility is to raise it separately from the submittal review, in correspondence, as an observation rather than as a review comment. The same applies to constructability concerns. The channel matters as much as the content.

The Calibration Model

Four dimensions, each requiring a deliberate setting rather than a default.

Table 3. The Four Calibration Dimensions

The practices that follow are organized by these four, then by the organizational conditions that make them possible.

Practice Group 1: Depth Calibration

Practice 1.1: Assign depth tiers when the register is built, not when the submittal arrives.

Depth assigned on arrival is depth assigned under time pressure, and time pressure produces shallow reviews on exactly the items that needed deep ones. Assigning at register construction is a considered decision made when nothing is urgent.

Practice 1.2: Tier by consequence, not by value or by size.

A high-value standard product with three specified criteria needs less review than a modest custom assembly with a rated fire performance requirement. The variable is what happens if non-compliance survives approval: cost, schedule, occupancy, licensing, or life safety. Dollar value correlates weakly with that.

Table 4. Depth Tiering by Consequence

Practice 1.3: Make the tier assignment visible on the submittal.

A reviewer who can see the tier reviews to it. A reviewer who cannot will default to their habitual depth, which is uniform, which is the problem.

Practice 1.4: Retain documentation proportional to depth.

Full determination records on the highest tiers, partial on the middle, a dated notation on the lowest. The goal is not universal documentation, it is documentation where a dispute is plausible and consequential.

Practice 1.5: Accept that some submittals receive a light review, and record that it was a decision.

Every project reviews some submittals lightly. The difference between a calibrated process and a careless one is whether that was chosen. A recorded Tier C designation is a defensible decision about depth. An undifferentiated fast review is not.

Practice Group 2: Scope Calibration

Practice 2.1: Read the review limitation clause at project start and brief the team.

Most project engineers and many reviewers believe an approval offers considerably more protection than the general conditions provide. This costs ten minutes and prevents both failure directions: reviewers stay inside the boundary, and contractors stop relying on approvals for things approvals do not cover.

Practice 2.2: Keep compliance comments and constructability observations in separate channels.

Compliance findings belong in the review. Constructability, sequencing, and safety observations belong in correspondence or a meeting. The channel separation is what lets you raise a genuine concern without assuming responsibility for the area it concerns.

Practice 2.3: On delegated design, verify criteria conformance and stop.

Check that the seal is present, the jurisdiction is correct, the criteria match the structural general notes and the specification, and the delegation scope is bounded correctly. Do not check the analysis. The delegated engineer is the engineer of record for that element, and reviewing the calculations for adequacy assumes liability the contract placed elsewhere.

Practice 2.4: Have the contractor own the checks the design consultant’s scope excludes.

Field dimensions, quantities, trade coordination, and conflicts with previously approved submittals sit outside the design consultant’s review scope in most contracts. That means the contractor’s screen is the only review those aspects receive. A contractor who treats their screen as a formality has left an entire category unreviewed.

Practice 2.5: Write comments that describe conditions rather than directing solutions.

“Submitted damper actuator is not rated for the assembly it penetrates per paragraph 2.4.C” describes a condition. “Use a Model X actuator” directs a solution and puts the reviewer in the position of having specified it. The first also preserves the submitter’s ability to find a compliant answer you did not think of.

Table 5. Scope Boundary in Practice

Practice Group 3: Evidence Calibration

Practice 3.1: Determine what the specification requires as evidence before deciding whether the submittal satisfies it.

Specifications vary. Some require compliance, some require certification, some require test reports from independent agencies, some require testing at project conditions. A reviewer who applies a uniform evidentiary standard will over-demand on some submittals and under-demand on others, and both are calibration errors.

Table 6. Evidentiary Standards and When Each Applies

Practice 3.2: Distinguish assertion from demonstration explicitly, and record which you accepted.

Where the specification requires evidence and a claim is offered, the requirement is unmet. Where the specification accepts a claim, accept it and record that the determination rests on the manufacturer’s certification rather than on verified data. That record matters later if the product underperforms.

Practice 3.3: Verify the condition, not just the value.

A performance value is only meaningful at a stated condition, and manufacturers publish at industry-standard conditions rather than project conditions. Verifying the number without the condition is one of the most common ways a review produces false confidence.

Practice 3.4: On assembly-based standards, verify the tested configuration.

Compliance attaches to the tested assembly, not to the rating. A listing with the right rating for a different assembly is not evidence for the furnished one. This is the check most likely to be performed by an inspector and least likely to be performed by a reviewer.

Practice 3.5: Do not substitute reviewer knowledge for submitted data.

A reviewer who knows a product meets a requirement is not entitled to determine compliance on that basis, because the submittal is the contractual record. This is a discipline rather than an insult to expertise: if it is not in the submittal, it is not demonstrated, and the review cannot be defended on the strength of what the reviewer happened to know.

Practice 3.6: Record honestly where a referenced standard could not be consulted.

Standards access is uneven, and many reviewers cannot obtain every cited standard. The defensible response is to record that compliance with that standard was accepted on the submitter’s certification rather than independently verified. The indefensible response, and the common one, is silence that implies verification.

Practice Group 4: Communication Calibration

Practice 4.1: Cite the paragraph, the specified value, the submitted value, and its location in every comment.

Four elements. A comment with all four ends a discussion and produces a targeted resubmittal. A comment with none of them produces a resubmittal that guesses, and then another one.

Table 7. Comment Quality by Element Included

Practice 4.2: Distinguish “not addressed” from “does not comply” in the comment itself.

The remedies differ entirely. Not addressed needs information; non-compliance needs a different product or configuration. A comment that conflates them produces a resubmittal that changes the wrong thing, and this is a leading cause of resubmittal spirals.

Practice 4.3: Eliminate preference comments.

A comment recording what the reviewer would have preferred, on a submittal that complies with the documents, is the single most credibility-destructive thing a review can contain. A set of ten teaches that comments are negotiable. One or two in a set of ten teaches the submitter that comments are negotiable. A compliant submittal that is genuinely undesirable is a design change conversation. The comments should not be used for that purpose.

Practice 4.4: Group or prioritize comments so material findings are visible.

A review with three critical findings and twenty-two minor ones should present the three first and separately. Burying a rated assembly finding at item nineteen is a self-inflicted wound.

Practice 4.5: Name the specific difference on equivalence and configuration findings.

On a non-equivalent tested assembly, state which component differs. The submitter frequently cannot see it, because they believed a matching rating meant a matching assembly. Naming the difference converts a frustrating rejection into a solvable problem.

Practice 4.6: Treat a third revision as a conversation rather than a cycle.

Repeated rejection is almost never submitter persistence. It is a requirement ambiguity, a reviewer applying an unstated standard, an unacknowledged substitution, or a genuine market unavailability. Each has a different remedy and none is a fourth submission. A fifteen-minute call at revision three saves two review durations.

Practice 4.7: Record accepted deviations somewhere other than the returned submittal.

An accepted deviation noted only on a returned submittal is effectively lost. It will be disputed at inspection, at commissioning, or in a warranty claim, and the person defending it will not be whoever accepted it. A standing project deviation register solves this at negligible cost.

Practice Group 5: Organizational Practices

Practice 5.1: Run two tiers deliberately, with different purposes.

The contractor’s screen checks requirement presence, completeness, identity, deviation identification, coordination, and field dimensions. The design consultant’s review checks technical conformance. Collapsing them means either the contractor renders technical determinations they are not responsible for, or the design consultant catches completeness failures that cost a full review cycle.

Practice 5.2: Sequence submissions to protect review capacity.

Specifications require submission within a fixed period after subcontract execution, and most subcontracts execute in a narrow window, so several hundred submittals arrive over a few weeks. Review capacity is fixed and queue depth explodes. Sequencing submissions by need date rather than by contractual deadline flattens the queue without anyone working harder, and it is entirely within the contractor’s control.

Practice 5.3: Give the design team a prioritized queue.

Tell reviewers which items have the least float. Most consultants will accommodate a prioritized queue and cannot triage an undifferentiated one, because they have no basis for it. Withholding priority information and then complaining about cycle time is common and self-defeating.

Practice 5.4: Measure four things and report them monthly.

First-pass approval rate. Average cycle time by discipline against the contractual duration. Not-addressed determinations per high-tier submittal. Contractor return rate at the screen. Those four diagnose most review pathologies, and the third is the one that proves a requirement-driven method is finding what recognition review would miss.

Table 8. Review Health Metrics

Practice 5.5: Route by expertise, not by availability.

Acoustic criteria to whoever understands acoustics. Rated assemblies to whoever can read a listing. Delegated design criteria to the structural engineer. A reviewer working outside their expertise produces a review that looks complete and verifies little.

Practice 5.6: Build requirement lists once per section and reuse them.

Decomposing a specification section into discrete verifiable criteria is the investment that makes calibrated review affordable, and the artifact is reusable across projects sharing a specification base. Building it per submittal is what makes requirement-driven review feel unaffordable.

Practice 5.7: Escalate persistent review capacity problems to the owner, with data.

Design consultants are under contract to the owner. A persistent bottleneck is an owner problem with an owner remedy, and presenting it early with three months of cycle time data works considerably better than presenting it late as part of a claim.

Practice Group 6: Relationship and Market Practices

Practice 6.1: Hold a pre-submittal conference on high-consequence packages.

Thirty minutes with the subcontractor and their supplier before the submittal is prepared, walking the specified criteria that are unusual or project-specific. This is the highest-return half hour in the entire process on envelope, acoustic, and performance-critical scope, because most non-compliance originates in a submitter who did not know a criterion mattered.

Practice 6.2: Tell submitters why an unusual criterion exists.

A subcontractor who understands that an acoustic criterion exists because of an adjacent recording studio will find a compliant product. One who thinks it is arbitrary will argue. This costs a sentence.

Practice 6.3: Distinguish market unavailability from submitter failure.

Sometimes the specified data does not exist because no manufacturer tests at that condition. Repeated rejection cannot produce what does not exist. The options are accepting nearest-available data with a documented determination, requiring independent testing, or raising a specification issue. Recognizing this situation rather than escalating it is a mark of an experienced reviewer.

Practice 6.4: Return substitutions rather than reviewing them on the merits.

Reviewing a disguised substitution and then rejecting it invites an argument about the product when the issue is procedure. Return it, cite the Division 01 substitution procedure, and the conversation stays where it belongs.

Practice 6.5: Recognize when the specification is the problem.

Where the same issue recurs across multiple submittals under one section, suspect the section. A criterion stated without a test condition, a standard cited without stating what evidence is required, or a performance requirement no product in the market meets are specification defects that submitters cannot fix. Raising it once with the design team beats rejecting it four times.

Stakeholders

Table 9. Practice Ownership

Groups 1, 5, and 6 are largely within the contractor’s control, which is roughly sixteen of the thirty practices here. That is worth noting, because review quality is usually discussed as something the design team determines.

Required Documentation

Table 10. Documents Supporting Calibrated Review

Technology Integration

Table 11. Technology Against the Four Dimensions

That table is deliberately unflattering in two columns. Scope calibration is a contractual and professional discipline that no tool addresses, and it is the dimension where miscalibration creates liability. Comment materiality is judgment, and the volume that automated comparison can produce makes materiality discipline more important rather than less.

AI-Assisted Opportunities

The dimension where automation genuinely changes practice is evidence, and specifically the detection of absence. Establishing that a specification-required value does not appear anywhere in a two-hundred-page submittal requires exhausting the document, which is the one thing human review cannot do reliably at speed.

Platforms built to interrogate construction documents address this. iFieldSmart AI, for instance, extracts specification requirements with citations to the source and compares submitted documents against them, flagging missing and non-compliant content. Used well, that changes the reviewer’s role from searching to triaging, which is the higher-value work and the work that requires more experience.

Table 12. AI Use Case Matrix Against Practice Groups

Implementation and Maturity

Calibrated review is not adopted in one step. Expect recognizable stages, and expect the middle one to feel worse than the first because measurement starts producing uncomfortable numbers.

Table 13. Review Maturity Model

Table 14. Implementation Roadmap

Anti-Practices

Table 15. What Not to Do

How Calibration Shows Up on Real Projects

Class A office tower, 340,000 square feet. Practice 6.1 changed the outcome. A pre-submittal conference on curtain wall walked the project design pressures with the fabricator and their testing consultant before the package was assembled. The initial submittal came in with test evidence at project pressures rather than at the standard’s minimums, which is the finding that would otherwise have consumed two review cycles on a lead time that gated the enclosure milestone.

Regional hospital expansion, 190 beds, occupied campus. Practices 1.2 and 3.1. The team tiered by consequence rather than by value, which moved medical gas, damper assemblies, and pressure-critical air handling equipment to the deepest tier despite modest dollar values, and moved several expensive but simple items down. On healthcare work the consequence class is licensing, and dollar value tracks it poorly.

Hyperscale data center, 60 MW. Practices 5.2 and 5.3. Subcontracts executed in a narrow window and several hundred submittals arrived over five weeks, with review durations reaching seven weeks against a fifteen-working-day requirement. The remedy was sequencing by need date and issuing a prioritized queue with float information, and cycle time on the critical items dropped without the consultant adding resources.

Advanced manufacturing facility, food grade. Practice 6.3. The specification required a cleanability performance value at a condition no panel manufacturer tested at. Three rejection cycles had produced nothing, because nothing was available. Recognizing market unavailability rather than escalating produced the actual resolution: nearest-available data with a documented determination, plus a project-specific test on the mockup.

University science building, 11 floors. Practice 4.5. Fume hood submittals were rejected twice with comments stating that containment was not demonstrated. The third comment named the specific difference, which was face velocity, and the compliant data arrived in four days. The submitter had not understood that a containment value at a different velocity was not comparable.

Multifamily podium, 240 units. Practice 3.4, on rated floor-ceiling assemblies. The reviewer compared the cited listing’s assembly description against the drawn detail component by component and found different framing spacing and a different resilient channel. Same rating, different assembly. On hundreds of repetitions, that check was worth more than every other review activity on the project combined.

Highway interchange, design-bid-build. Practices 5.1 and 5.4. The contractor’s screen was instituted with a defined scope covering certification currency, traceability, and completeness, since on unit-price work those conditions payment. The return rate settled around eighteen percent and design team review cycles dropped correspondingly.

Historic warehouse conversion to hotel, 140 keys. Practice 6.5. Three separate submittals under the same section exhibited the same issue, which pointed at the section rather than at the submitters: an acoustic criterion stated without a test condition, unverifiable as written. Raising it once with the design team produced a clarification by addendum that resolved all three and prevented several more.

FAQ’s

Is more review always better?

No, and the assumption that it is produces both over-review and mis-review. Review capacity is finite, so depth spent on a submittal that did not need it is depth unavailable for one that did. Over-review also creates liability by commenting outside the reviewer’s scope, and it erodes comment credibility when preference comments accompany material findings. Calibrated review beats maximum review on both quality and cost.

What is over-review, concretely?

Commenting on fabrication means and methods, erection or installation sequence, construction safety, quantities, or delegated engineering adequacy. Also recording preferences on submittals that comply with the documents. Each takes the reviewer outside the scope the general conditions assign and, where the submitter relies on the comment, creates exposure that would not otherwise exist.

How should a genuine constructability or safety concern be raised, then?

Separately from the review, in correspondence or in a meeting, framed as an observation rather than as a review comment. The concern should absolutely be raised; the channel is what keeps the scope boundary intact. Putting it in the review comment set is what converts an observation into direction.

How do we decide review depth?

It will be the case that if non-compliance continues to flourish after being granted exemption, then certain elements pertaining to safety of life, licensing, occupancy, adequacy and performance of structures and buildings, or the exposure of long-lead procurement, will be of concern. Assign the tier when the register is built and make it visible on the transmittal, because depth assigned on arrival is depth assigned under pressure.

Are preference comments really that damaging?

Yes, disproportionately. A review containing material compliance findings alongside preferences teaches the submitter that comments are opinions to negotiate rather than requirements to satisfy. On the next submittal the material findings receive the same treatment. One or two preferences in a set of ten is enough to produce the effect.

Who should own the contractor’s review?

The project engineer, with a defined scope covering requirement presence, completeness, identity, deviation identification, coordination against previously approved submittals, and field dimension consistency. Not technical conformance, which belongs to the design consultant. A useful indicator of whether it is happening is the return rate; near zero means submissions are being forwarded unscreened.

How do we improve cycle time without adding reviewer resources?

Sequence submissions by need date rather than by contractual deadline, and give the design team a prioritized queue with float information. Both are within the contractor’s control and both address loading rather than capacity. Most cycle time complaints are loading problems dressed as capacity problems, and the contractor holds more of the remedy than the conversation usually acknowledges.

What evidentiary standard should we apply?

Whatever the specification requires, which varies section by section. Some sections require compliance, some certification, some independent test reports, some testing at project conditions. Applying a uniform standard over-demands on some submittals, which costs time and money the contract did not require, and under-demands on others. Read the section before deciding what the submittal owes.

How do we handle a reviewer who applies standards the documents do not state?

Ask them to cite the requirement. Politely and in writing. This situation is more common than people expect and it usually resolves immediately, because the reviewer is applying a standard from practice or from a previous project without having noticed the documents differ. It is also one of the four causes of resubmittal spirals, and naming it is what stops them.

Should we tell subcontractors why an unusual criterion exists?

Yes, and it costs a sentence. A submitter who understands the reason will find a compliant product; one who thinks the criterion is arbitrary will argue or ignore it. On acoustic, envelope, and containment criteria in particular, explaining the reason at a pre-submittal conference prevents more non-compliance than any amount of reviewing.

How do we know whether our review is calibrated?

Four numbers. First-pass approval rate above roughly seventy percent. Cycle time at or under the contractual duration. Several not-addressed determinations per high-tier submittal, since near zero means nobody is checking for silence. And a contractor return rate between ten and twenty-five percent, since near zero means the screen is not happening. Add a fifth if you can measure it: preference comments as a share of total, which should be near zero.

What is the single highest-return practice here?

Pre-submittal conferences on high-consequence packages, at about thirty minutes each. Most non-compliance originates with a submitter who did not know a criterion mattered, and a conversation before the package is assembled prevents it rather than catching it. If you want a second, tier the register by consequence before submittals start arriving.

Expert Recommendations

For the design consultant: return compliance findings and nothing else, and raise constructability and safety observations in separate correspondence. It reads as unhelpful for a few weeks and then produces a review whose every comment gets addressed.

For the project manager: tier the register by consequence at project setup and put the tier on the transmittal. It costs an hour and it changes reviewer behavior more reliably than any instruction.

For the project engineer: hold pre-submittal conferences on your five highest-consequence packages. Half an hour each, before the submittals are prepared, walking the criteria that are unusual or project-specific.

For contracts: brief both the internal team and, where the relationship permits, the design consultants on the review limitation clause at project start. Miscalibration in both directions traces to that clause being unread.

For the structural engineer: verify delegated design criteria against your general notes and stop there. Reviewing the analysis for adequacy assumes liability that the delegation placed with the delegated engineer.

For the project executive: ask for preference comments as a share of total on one review cycle. If the share is material, your reviews are spending credibility they will need for the findings that matter.

For the owner or owner’s representative: ask both your contractor and your design team how review depth is assigned. If the answer is that all submittals receive the same review, your consequential submittals are being reviewed at the average, and the average is not enough for the rated assemblies.

Where Calibration Leaves You

The useful reframe here is that submittal review is a resource allocation problem before it is a technical one. Every project has limited reviewer hours and a list of hundreds of items with differing effects. Plus, the project has a contract that defines the responsibilities of the various parties. If all this is done well, then the effort would be proportional to the consequence, the responsibility would be adhered to, the contract would be fulfilled without going beyond what is required, and reporting the results of the project in a way that prompts the needed corrections to be made rather than stoking the debate.

None of the thirty practices here requires software, additional headcount, or anyone’s permission, and roughly half of them sit entirely within a contractor’s control. What they require is deciding four things deliberately that most projects decide by default.

If you take two things, take the pre-submittal conference on your highest-consequence packages and the elimination of preference comments. The first prevents non-compliance rather than catching it. The second makes every finding you do report land with the weight it deserves.