Types of Constructability Issues in Construction Projects: A Complete Classification

A regional contractor showed me their change order analysis for a completed year. Forty-one million dollars of work, 214 change orders, and a pie chart. The largest slice, 63 percent, was labeled “design issue.”

I asked what they had done with that finding. Nothing, because there was nothing to do with it. “Design issue” is not a category. It is a shrug with a percentage attached. It tells you where the money went and nothing about how to stop it going there again.

The firms that actually reduce this number do one unglamorous thing first. They classify. Not into two buckets but into categories specific enough that each one points at a different fix: a different checklist item, a different review phase, a different reviewer, a different detection method.

This article is that classification. Nine types of constructability issues, what each one looks like on real drawings, what causes it, how it gets detected, and what it costs when it is not. For how reviews are scheduled and governed, see the article on the constructability review process. For the verification prompts themselves, see the constructability review checklist.

Key Definitions: Separating Issues From Errors

Precision here prevents an argument later. Several distinct things get called constructability issues, and only some of them are.

Constructability issue is a condition in the design documents that will make the work difficult, expensive, slow, unsafe, or impossible to build, price, sequence, operate, or maintain as shown. The design may be entirely correct and still contain constructability issues.

Design error is a condition that is wrong: a calculation mistake, a code violation, a structural member undersized for its load. Errors are a design quality problem with professional liability implications. Constructability issues usually are not.

Omission is required information that is absent. It sits between the two: sometimes a document completeness issue, sometimes a standard-of-care question.

Clash is a geometric interference between two modeled elements. A subset of one issue type, not a synonym for the whole field.

That distinction has consequences beyond vocabulary. Presenting results as mistakes incites a defensive reaction from the design team and creates an exposure risk that obstructs progress. Presenting the same observations as conditions of constructability keeps the focus of the discussion on the procedural aspects of buildability, where it belongs, and where it can be addressed most efficiently.

Objectives: What Classification Buys You

• Route findings to the right fix. A missing detail needs a checklist item. A spatial conflict needs earlier coordination. These are different interventions.
• Assign the right detection method. Some types are found by software, some by models, some only by a person who has installed the work.
• Identify the economical discovery point. Sequence issues must be caught at schematic design. Detail omissions can wait until construction documents.
• Measure improvement credibly. “Design issues down 12 percent” means nothing. “Missing-detail findings down 40 percent after adding reference integrity checks” means something.
• Aim checklist development. Category frequency tells you which part of your library is underdeveloped.
• Set severity consistently. Type correlates strongly with impact, which makes triage faster and more defensible.

Why Taxonomy Changes Outcomes

The mechanism is straightforward once you see it. Different issue types are invisible to different detection methods, and a firm that does not classify cannot tell which method it is missing.

Consider two firms with identical field problem rates. Firm A runs excellent model coordination and no checklist review. Firm B runs a strong checklist and no coordination. Both feel like they are doing constructability review. Firm A leaks missing details and specification conflicts constantly. Firm B leaks spatial conflicts in congested zones. Neither one can see its own blind spot, because both measure the same undifferentiated total.

Classification makes the blind spot visible. When 40 percent of your field problems trace to a category your review process has no method for detecting, that is a diagnosis rather than a complaint.

Read the last column carefully. Clash detection is blind to five of nine categories. That is not a criticism of coordination, which does its job well. It is an argument against treating coordination as the constructability review, which is a substitution I see constantly.

Stakeholders: Who Finds What

Issue types map onto reviewers with surprising consistency, which makes review team composition a design decision rather than a staffing convenience.

The practical use of that table is to check your review team against it. If your reviews are staffed by a project engineer and a VDC manager, you have good coverage on four categories and effectively none on sequence, tolerance, operability, or code. That is a predictable pattern of field problems waiting to happen.

The Nine Categories of Constructability Issues

Nine is a working number. Fewer and the categories stop pointing at distinct fixes. More and reviewers stop applying them consistently.

1. Missing or Incomplete Information

Required information is absent. The most common category on nearly every project and the easiest to detect mechanically.

What it looks like: a detail bubble referencing a detail that was never drawn. A wall type on plan that does not appear in the schedule. A finish schedule entry for a room number that no longer exists. Dimension strings that do not close. A specification section calling for a submittal on a system the drawings do not show.

Root cause is almost always issuance mechanics rather than negligence. Sheets get renumbered between issuances, details get relocated, a consultant’s scope changes and their callouts do not get updated. Nobody decides to omit a detail. The set drifts.

This category is where AI-assisted review produces its clearest wins, because reference integrity is pure matching across a large set. A human checking 900 sheets for orphaned callouts will miss some. Software will not.

2. Conflicting Information

Two documents say different things. Both are complete. They disagree.

What it looks like: drawings show a 4 inch cavity, specifications require 5 inch insulation. Architectural plans show a slab elevation the structural drawings contradict. A door schedule specifies hardware inconsistent with the rated assembly on the plan. Two sheets detail the same waterproofing transition differently.

The dangerous property of this category is that each document is internally defensible. A trade prices what they read, installs what they priced, and the conflict surfaces when two trades meet at the interface with incompatible work already in place.

Root cause is usually parallel document production without a reconciliation pass. Drawings and specifications are frequently produced by different people on different schedules.

3. Dimensional and Geometric Conflict

Two elements cannot both occupy the space shown. The classic clash.

What it looks like: a duct main crossing a beam at an elevation that requires a penetration the structural drawings prohibit. A stair landing that does not align with the floor level. A shaft that changes dimension between floors with no transition. Equipment shown in a room where the equipment plus its clearance exceeds the room.

This is the category model coordination handles well and sheet review handles poorly, because plan-only review cannot see vertical conflict. If your review does not include sections and a model, expect to leak this category into the field.

4. Spatial Congestion and Clearance

Everything fits geometrically, and there is no room to install it, work in it, or get a body between components.

What it looks like: a plenum where ductwork, cable tray, sprinkler main, and structure total more than the available depth once hangers and insulation are counted. A pipe chase sized for the pipes but not for the installer’s hands. Electrical working clearance obstructed by other trades’ work. A congested corridor where four trades must all install above the ceiling in sequence.

The distinction from category three matters. A clash detection run on modeled centerlines will pass a plenum that no installer can actually work in. Congestion is about constructability, not geometry, and it is one of the strongest arguments for having someone who has installed ductwork look at the sections.

5. Access, Operability, and Maintainability

The work can be built. It cannot be reached afterward.

What it looks like: a damper above a hard ceiling with no access panel. A valve behind millwork. An air handler with less coil pull space than the coil is long. A rooftop unit with no replacement path once the parapet and screen are complete. A control device in a ceiling the infection control plan will not allow anyone to open.

Root cause is structural in the review process rather than in the design. The person who would notice these conditions is the owner’s facilities manager, and they are almost never invited. Design teams locate devices where the systems require. Contractors review for installation. Nobody reviews for the following twenty years.

6. Sequence, Means, and Methods

Each element is buildable. The order required to build them is not achievable.

What it looks like: equipment that cannot be set after the enclosure is complete, with no rigging opening shown. A detail requiring shoring in a space the detail leaves no room for. Formwork that cannot be stripped because adjacent work encloses it. A phasing plan where the temporary condition has no structural support. Finish work scheduled before wet trades above it.

No automated method finds this category, and that is unlikely to change soon. It requires reasoning about temporary conditions that appear nowhere in the documents, using knowledge of how crews actually work. It is the single strongest argument for having a superintendent in every review.

7. Interface and Tolerance Conflict

Two systems meet, and their tolerances are incompatible or the transition between them is undefined.

What it looks like: a cast-in-place structure with a tolerance of plus or minus three quarters of an inch receiving a curtain wall system that accepts a quarter inch. A storefront meeting a masonry opening with no defined joint. A precast panel joint dimension that assumes structural accuracy the structural specification does not require. A floor finish requiring a flatness the slab specification does not deliver.

Tolerance conflicts are quietly among the most expensive findings because they surface at fabrication or installation, well past the point where anything is cheap. Trade partners find them immediately and are rarely asked in time.

8. Existing Conditions and Site Logistics

The documented assumption about the site or existing structure does not match reality, or the site cannot support the construction the design requires.

What it looks like: a crane with no position that reaches all structural picks. Staging shown on ground needed for phase-one work. An existing utility at an undocumented depth crossing the new footing line. A renovation detail assuming a wall assembly that turns out to be different. Delivery access requiring a turning radius the site does not provide.

Detection requires a site walk and, on renovation work, reality capture. Document review alone cannot find this category, which is why the schematic design review should include somebody who has physically stood on the site.

9. Code, Permit, and Inspection Conflict

The design is buildable and will not pass, or the sequence it requires is not permitted.

What it looks like: a phasing plan assuming a lane closure the agency will not approve. An inspection sequence requiring access to work that will be concealed by the time the inspector arrives. A rated assembly penetrated with no listed firestop configuration available. An egress path that works at completion and is blocked during a construction phase. Temporary conditions that violate occupancy requirements in an operating facility.

This category carries the sharpest consequence, because the remedy is often a stop-work order rather than a change order. On infrastructure and healthcare renovation work it deserves its own dedicated review pass.

Severity Classification

Category tells you what kind of problem it is. Severity tells the design team where to spend attention first. Both belong on every log entry, and severity should be assigned before the log leaves your office.

Three tiers is enough. Five-tier scales look more rigorous and collapse in practice, because the middle three become indistinguishable and reviewers default to the center.

Severity should reflect consequence, not effort to fix. A missing access panel is a two-minute drawing revision and a demolition event if missed, which makes it significant rather than advisory. Reviewers new to severity assignment consistently rate by fix difficulty, which inverts the whole point.

Required Documentation

Technology Integration

Each detection method covers a different part of the taxonomy, which is the practical reason a single tool never suffices.

Line them up and the gap becomes obvious. A firm running clash detection plus manual sheet review, which describes a large share of the industry, has thin coverage on sequence, tolerance, existing conditions, and code. Those are exactly the four categories with the highest cost per occurrence.

AI-Assisted Opportunities by Category

AI does not improve all nine categories equally, and being honest about that distinction is what keeps expectations survivable.

The pattern is consistent. Where the task is exhaustive comparison against known criteria, uplift is high. Where it requires reasoning about conditions that exist only temporarily and appear in no document, uplift is low. Automating categories 1 and 2 is what should be prioritized. They are the most repetitive categories in the majority of logs.

Implementation: Rolling Out a Taxonomy

Phase 3 is the one people skip and the one that determines whether the data is usable. If three reviewers classify the same finding three different ways, your annual report is noise. An hour of calibration typically resolves most disagreements.. Surviving disagreements often show that the definition of a category could use further refinement.

Best Practices

Common Mistakes

Industry Examples Across Project Types

Healthcare: Category Five Dominates and Nobody Reviews It

Hospital plenums are the densest above-ceiling condition in commercial construction: ductwork, medical gas, sprinkler, cable tray, pneumatic tube, structure, hangers, all inside a floor-to-floor dimension that value engineering compressed at schematic design.

Nearly every one of those systems has a device requiring service access, and infection control constraints limit which ceilings can be opened once the facility is operating. Access and maintainability is the dominant category, and it is the category no standard review team is staffed to find. One facilities manager in the design development review changes the outcome more than any software purchase.

Infrastructure: Category Nine Carries the Sharpest Teeth

On phased infrastructure work the binding constraint is usually permit conditions rather than physics. A staging plan requiring a closure duration the agency will not grant is not a constructability inconvenience. It is a project that cannot be built as planned.

The review implication is timing. Code and permit conflicts must be found during schematic design while the phasing concept is still open. Once permits are in progress and the agency has a submitted plan, the leverage is gone.

Industrial: Category Six Is the Whole Game

Process and industrial projects revolve around equipment that is large, heavy, and delivered on a schedule nobody controls. The recurring failure is a rigging path that exists during design and disappears during construction because the enclosure went up first.

No document review finds this. A superintendent looking at the sequence and asking how the equipment gets in finds it in about four minutes. This category is the clearest evidence that constructability review is a staffing decision before it is a technology decision.

Multifamily: Categories One and Two, Multiplied

A recent architectural and interiors review produced fourteen findings across eight coordination categories. Most fell into missing or conflicting information: a transition detailed two ways, a wall type absent from the schedule, a dimension that did not close.

Individually minor. What made them consequential was repetition across unit stacks. The category one and two findings on repetitive work highlight that the same conditions are likely to occur on each repetition of the work, and therefore, the same issues are likely to remain, even though remedial action has been taken. Both of these categories of findings “automate well”, and therefore, multifamily is one of the strongest arguments for AI-assisted document reviews in the industry.

Frequently Asked Questions

What are the main types of constructability issues?

Nine categories cover the field in practice: missing or incomplete information; conflicting information between documents; dimensional and geometric conflict; spatial congestion and clearance; access, operability, and maintainability; sequence, means, and methods; interface and tolerance conflict; existing conditions and site logistics; and code, permit, and inspection conflict. Each has a different detection method and a different economical point of discovery, which is the reason classification is worth doing at all.

What is the difference between a constructability issue and a design error?

A constructability issue refers to a condition that creates a difficulty or impediment in building, pricing, sequencing, operating, or maintaining a facility as described. The design may be completely correct. A design error is a mistake: a math mistake, a code violation, an undersized member. The practical distinction is important. Making something a framing issue creates a professional liability problem and brings about a defensive response. Constructability issues maintain the discussion on buildability.

Which type of constructability issue is most common?

Missing or incomplete information, by a wide margin, on almost every project. Orphaned detail callouts, wall types absent from schedules, dimension strings that do not close, specification sections requiring submittals for systems the drawings do not show. The root cause is usually document drift across issuances rather than negligence: sheets get renumbered, details get relocated, and callouts do not follow. It is also the category that automated review handles best, since reference integrity is pure matching.

Which type is most expensive?

Interface and tolerance conflicts and sequence and means issues cost the most per occurrence, even though both appear less frequently than document completeness problems. Tolerance conflicts surface at fabrication or installation, when material has already been made. Sequence problems surface when the work is in place and the required access no longer exists. Code and permit conflicts can be worse still, because the remedy is sometimes a stop-work order rather than a change order.

Can BIM clash detection find all constructability issues?

No. Clash detection covers geometric interference between modeled elements well and part of spatial congestion. It is blind to missing information, document conflicts, access and maintainability, sequence and means, interface tolerance, existing conditions, and code and permit issues. That is five to seven of nine categories depending on how you count. Coordination is necessary and it is not a constructability review, and treating the two as equivalent is one of the most common and expensive substitutions in current practice.

How should we assign severity to constructability findings?

Use three tiers and rate by consequence, not by how hard the fix is. Critical means work cannot proceed as documented and cost or schedule impact is likely. Significant means an RFI or field conflict will result if unresolved. Advisory means the condition is worth noting but the work is buildable as shown. The common error is rating by fix difficulty, which downgrades things like a missing access panel: two minutes to draw, a demolition event to miss.

Why does classifying issues matter if we are going to fix them anyway?

Because each category points at a different fix, and without classification you cannot tell which fix you need. A firm leaking missing details needs better reference integrity checking. A firm leaking spatial congestion needs earlier section-based review. A firm leaking sequence issues needs a superintendent in the room. All three feel identical when reported as one undifferentiated total, which is how firms end up with a change order analysis whose largest category is “design issue.”

Which issues can AI reliably detect?

Missing and conflicting information most reliably, because both are comparison problems across a document set: orphaned callouts, schedule-to-plan mismatches, drawing-to-specification conflicts. Threshold-based checks on clearance and congestion work well, with a human judging adequacy. Dimensional conflict, tolerance extraction, and code requirement extraction give moderate uplift. Sequence, means and methods, and site logistics give little, because they require reasoning about temporary conditions that appear in no document.

How do we find access and maintainability issues?

Invite the owner’s facilities group to the design development review. This category is systematically under-detected across the industry for a structural reason: design teams locate devices where systems require, contractors review for installation, and nobody reviews for the twenty years after handover. A checklist item requiring an access provision at every concealed device catches much of it. A facilities manager in the room catches the rest, and costs nothing but a calendar invitation.

Should we track constructability issues found during construction?

Yes, and tag them with the same nine categories you use for review findings. That is the only way to measure leakage by type, which is the diagnostic that tells you where your review process is thin. A firm that classifies review findings but not RFIs knows what it caught and has no idea what it missed. Classify change orders too, because frequency and cost point in different directions and you need both.

How many categories should we use?

Nine or fewer. In addition, as more reviewers are added, inter-reviewer agreement drops and the quality of the data degrades to the extent that it becomes unusable for year-over-year comparisons. whatever number you settle on, implement a calibration study: have three reviewers classify the same thirty findings independently and compare. Below about 80 percent agreement, your category definitions need worked examples from your own projects rather than more categories.

Do issue types vary by project type?

Substantially, and it should drive how you staff and sequence reviews. Healthcare skews toward access and maintainability and above-ceiling congestion. Industrial and process work skews toward sequence and rigging. Infrastructure skews toward code and permit conflicts. Multifamily and institutional skew toward missing and conflicting information, amplified by repetition. Knowing your sector’s dominant categories tells you which reviewers to add and which detection methods to invest in first.

Expert Recommendations

• Code three closed projects retrospectively before changing anything. Three weeks of classification produces a diagnosis that no amount of process discussion will.
• Map your detection methods against all nine categories. The uncovered rows are your predictable field problems, and most firms are surprised by which rows are empty.
• Run the calibration exercise. Thirty findings, three reviewers, one hour. Without it your annual category data is noise.
• Tag RFIs and change orders with the same categories. Review findings tell you what you caught. RFIs tell you what you missed, and only the second one improves anything.
• Report cost alongside frequency. The categories that appear least often are usually the ones spending the most money.
• Rate severity by consequence, not by effort. Train this explicitly, because the intuitive error runs the other way.
• Put a superintendent in every phase review. Category six is invisible to every other reviewer and every automated method, and it is expensive.
• Invite the owner’s facilities manager to design development. Highest return per calendar invitation available anywhere in preconstruction.

Conclusion

That 63 percent slice labeled “design issue” was accurate and useless. It described a year of losses without pointing at a single thing the firm could do differently.

Classification changes that. When findings are separated into categories with distinct causes and distinct detection methods, the log stops being a record of problems and becomes a map of process gaps. Missing details point at reference integrity checking. Spatial congestion points at earlier section-based review. Sequence issues point at who is in the room. Access issues point at who is not.

The technology implication follows the taxonomy rather than leading it. Automated document review produces its largest gains on categories one and two, which are also the highest-frequency categories in most logs, so it is a sensible first investment. It produces almost nothing on category six, which means no purchase substitutes for a superintendent reading the sequence. Knowing which is which is the difference between a tool that pays for itself and a tool that generates a longer log.