Quick answer: Published construction rework figures range from 0.38% to over 12% of project cost — but they measure different things. No peer-reviewed study has published a millwork-specific rework rate. The closest published proxy for fabrication-side rework (vendor error/omission in the Buildings category) is approximately 0.1% of construction-phase cost. Design errors cost 15× more in rework than vendor errors in the same dataset.
Depending on who you ask, construction rework costs anywhere from under 1% to over 12% of a project's budget — sometimes both figures appear in the same conversation. Vendor white papers cite "35% of shop drawings rejected on first pass" with no traceable source. Millwork trade forums repeat "rework kills margins" without any number attached to it.
For this article, I commissioned a literature search across CII publications, peer-reviewed journals, NIST reports, and academic thesis databases — strict instruction: only cite sources traceable to a published document with an author, organization, year, and URL. Here is what that search found, what it did not find, and what it means for anyone who fabricates, specifies, or manages millwork.
Rigorous rework cost data exists. It is just not millwork-specific. And the fact that no one has published a quantified millwork rework rate from a large auditable dataset is itself worth understanding.
Why Rework Percentages Vary by an Order of Magnitude
The first thing the literature makes clear is that a raw "rework percentage" is meaningless without three pieces of context: what is included in the scope, which project phase is measured, and what denominator the percentage is expressed against. These three variables account for virtually all the variation between published figures.
Scope: Some studies measure quality deviations across design and construction phases combined — this captures everything from incorrect specifications issued by the architect through field installation errors. Others measure field rework only — work physically done and then redone on site. In-shop fabrication corrections made before delivery are typically not captured in field rework metrics at all, which is particularly relevant for millwork: a cabinet built to wrong dimensions and corrected in the shop before shipment is invisible in any field rework dataset.
Phase and denominator: Total installed project cost (design + procurement + construction + commissioning) is a larger number than construction-phase cost alone, which is larger than contract value. A percentage expressed against total installed cost will always appear smaller than the same dollar amount expressed against construction-phase cost.
What gets counted: Informal rework — a field carpenter who quietly re-cuts a panel without logging it as a rework event, or an installer who adjusts cabinet heights without a formal change order — does not appear in any dataset. Latent defects discovered after handover may or may not be captured depending on how long the tracking window is.
The Construction Industry Institute addressed this measurement problem directly in their Field Rework Index publication: what you measure and how you define it determines what you find. The variation in published figures is a methodological artifact, not a disagreement about facts.
What the Best Available Data Actually Shows
The most directly traceable, peer-reviewed data comes from four sources. Each is presented here with scope, denominator, and limitations stated explicitly.
CII RS10-1 (1989): Quality Deviations Across Design and Construction
The Construction Industry Institute's landmark 1989 publication Costs of Quality Deviations in Design and Construction (RS10-1) analyzed industrial projects and found that the average cost of rework "exceeds 12 percent" of total installed project cost. This is the number most frequently cited — and most frequently misapplied.
Two qualifications matter. First, this is industrial projects (process plants, heavy manufacturing facilities), not commercial buildings or residential construction. Second, it measures quality deviations across design and construction phases — not field rework alone. The same study found that design deviations account for roughly 80% of increased costs from quality deviations, while construction-phase deviations account for about 20%. The implication: the majority of what gets counted as "rework cost" originates upstream, in design documentation errors and omissions, not in fabrication or installation.
Hwang et al. (2009): 359 CII Projects, Buildings Category
A more granular dataset comes from a 2009 peer-reviewed study by Bon-Gang Hwang, Stephen R. Thomas, Carl T. Haas, and Carlos H. Caldas, published in the Journal of Construction Engineering and Management. Using 359 owner-reported projects from the CII database, the study calculated Total Field Rework Factors (TFRF) — defined as total direct cost of field rework divided by total construction-phase cost — by project category.
For the Buildings category specifically: mean TFRF = 0.046, meaning field rework averaged 4.6% of construction-phase cost. The cause breakdown for Buildings is where the data gets interesting for millwork practitioners:
| Rework Source | Mean TFRF (Buildings) | % of Construction-Phase Cost |
|---|---|---|
| Design error / omission | 0.015 | ≈1.5% |
| Owner change | 0.014 | ≈1.4% |
| Constructor error / omission | 0.004 | ≈0.4% |
| Vendor error / omission | 0.001 | ≈0.1% |
| Vendor change | 0.001 | ≈0.1% |
"Vendor error/omission" and "vendor change" are the closest published proxies for fabrication-side rework — the category that would include millwork shops, glazing contractors, and other specialty fabricators. In the Buildings dataset, they represent the smallest measurable contributors to field rework cost.
The important caveat: this is field rework only. An incorrect millwork unit corrected in the shop before delivery does not appear as vendor error/omission in this dataset. The table shows what drives field rework cost — not total quality deviation cost across the supply chain.
Love et al. (2018): 19,605 Rework Events Across 346 Projects
A longitudinal study by Peter Love and colleagues, published in Production Planning & Control (2018) and hosted through Curtin University, provides the most detailed event-level data available. The dataset: 19,605 rework events across 346 construction projects tracked from 2009 to 2015.
Key findings: mean rework cost was 0.39% of contract value across 98 projects with complete cost records. Mean yearly profit was reduced 28% during the analysis period across the full sample. Most striking: 88 events — just 0.45% of all rework events — drove 34% of total rework costs. Rework cost is not evenly distributed; it is highly concentrated in a small number of high-impact events.
In millwork, this distribution matches what practitioners actually experience. Most packages on a project produce zero serious rework — minor field adjustments, scribe cuts, hardware substitutions that get absorbed. Then a systematic error — wrong ceiling height assumed across an entire casework package, or an incorrect wall furring dimension carried through 40 units — generates a single event that consumes a large fraction of the shop's margin. The concentration of rework cost in rare, high-impact events is why systematic errors matter far more than individual unit errors.
ASCE (2026): A Far Lower Field Rework Estimate
A January 2026 write-up published by the American Society of Civil Engineers reports findings from a recent study showing precompletion field rework costs average just 0.38% of contract value — rising to 0.76% when postcompletion corrections are included.
This is not contradictory to the CII and Hwang figures — it reflects a different scope (field rework only, logged formally, precompletion), a different denominator (contract value rather than construction-phase cost), and likely stricter criteria for what is counted as a rework event. It is consistent with Love et al.'s 0.39% figure from the same scope definition.
The reconciliation: CII's "exceeds 12%" counts quality deviations across all project phases against total installed cost. Hwang's 4.6% counts field rework against construction-phase cost. Love and ASCE's 0.38–0.39% counts formally logged field rework against contract value. All three figures can be simultaneously correct. They measure different things.
What Rework Actually Costs: The NIST Framework
Published cost breakdowns for individual rework events are scarce. The most rigorous framework comes from NIST's 2004 report Cost Analysis of Inadequate Interoperability in the U.S. Capital Facilities Industry (GCR 04-867), which estimated the annual cost of interoperability failures across the US construction industry at $15.8 billion and itemized what those costs actually consist of.
The NIST report describes rework cost components as: design and construction rework (direct labor and material), manually re-entering data across disconnected systems, verifying and validating information transferred between parties, scrapped materials, and delay costs (idle resources waiting for corrected information or replacement components).
Applied to millwork, these components translate directly:
- Material waste: Custom millwork components — veneered panels, solid wood face frames, pre-finished carcasses — cannot be returned or restocked. A wrong-dimensioned unit is scrap. Material cost alone on a single custom casework unit can run $200–$800 in substrate, veneer, and hardware.
- Fabrication labor: A standard commercial casework unit takes 8–16 shop hours to produce. Re-making it means doubling that labor with no additional revenue.
- Installation labor: Removal, resequencing, and reinstallation in the field adds mobilization cost. For union labor in commercial construction, this runs $85–$140/hour plus overhead.
- Delay cost: Millwork frequently sits on the critical path for finish work — flooring, ceiling grid, and paint cannot close in around spaces until millwork is set. A two-week delay on a hospital casework package on a $15M project at standard delay cost rates is not a trivial number.
- Administrative overhead: RFIs, revised submittals, change orders, coordination meetings — these are real costs that rarely appear in anyone's rework accounting but consume project management time at $100–$200/hour.
No published study has combined these components into a millwork-specific average. That gap is addressed directly below.
The Shop Drawing Connection: What One Case Study Measured
The only publicly auditable data on shop drawing first-pass outcomes comes from a 2005 master's thesis by James Andrew De Lapp at Texas A&M University: Impacts of CAD on the Submittal Process. It is a single-project case study — not an industry benchmark — but it is traceable, documented, and directly relevant.
Among 34 shop drawing submittals analyzed across a commercial project: 45% were approved without markup, 42% were approved as noted (minor corrections required), 8% required revise-and-resubmit, and 5% were rejected outright. That means 13% of submittals required a second full review cycle.
The more striking finding involves the comparison between hand-produced and CAD-produced shop drawings on the same project. Hand-produced drawings received 155 reviewer notes and 115 corrections. CAD-produced drawings received 8 notes and zero corrections. Occurrences per sheet: 2.73 versus 0.24. De Lapp's conclusion: the shop drawing process is "susceptible to human error in transferring design information and intent" when drawings are produced manually.
The limitation is explicit: this is one project, not a millwork-specific dataset, and it cannot be generalized as an industry benchmark. Vendor claims of "35% rejection rates for shop drawings" appear widely in trade content, but none of the sources reviewed here traced that figure to a published, peer-reviewed dataset. For millwork-specific common shop drawing mistakes that trigger rejections, the practical patterns are documented separately. The broader question of how submittals move through the approval chain — and where delays accumulate — is covered in the millwork submittal process guide.
The Millwork Data Gap: What No One Has Published
This is the part of the research that I think deserves its own section — because the absence of data is information.
In reviewing CII publications, AWI standards and trade bulletins, peer-reviewed construction management journals, and academic thesis databases, I did not find a single publicly available, auditable dataset that reports a millwork-specific or casework-specific rework rate. There is no published figure for "percentage of millwork packages requiring refabrication," "average rework hours per dollar of millwork installed," or "field rework rate for CSI Division 06 architectural woodwork as a distinct trade." The closest published proxies are the "vendor error/omission" and "vendor change" categories in Hwang et al. — but those are multi-trade figures that include glazing, mechanical equipment, and other specialty fabricators alongside millwork.
Similarly, I did not find a published, trade-isolated study quantifying how millwork schedule delays propagate into overall project delay. The NIST report documents delay cost mechanisms generally, but does not isolate millwork as a driver.
This gap exists for a reason. Custom architectural woodwork (NAICS 337212, with US industry sales of $8.9 billion in 2023 according to Kentley Insights) is fragmented across thousands of small shops. Large-sample, cross-project rework data requires either a general contractor willing to share quality management records across many projects, or a trade association commissioning a structured study. Neither has been published in accessible form.
The AWI (Architectural Woodwork Institute) and WI (Woodwork Institute) publish quality standards in detail — but standards documents describe what correct looks like, not how often incorrect occurs or what it costs when it does.
What This Means in Practice
Four conclusions from the data are worth pulling out specifically for millwork fabricators, subcontractors, GCs, and architects.
Design documentation errors cause more rework cost than fabrication errors. In the largest available published dataset, design errors and omissions generate roughly 15 times more field rework cost per project than vendor errors and omissions. This doesn't mean fabrication errors don't happen — it means the single highest-leverage intervention for reducing millwork rework is upstream: resolving dimension conflicts, clarifying material specifications, and coordinating with MEP and structural backgrounds before shop drawings are issued. Our millwork shop drawing services are structured around catching these conflicts before they reach the field.
Rework cost concentrates in a small number of high-impact events. Love et al.'s finding — 0.45% of events driving 34% of costs — means a millwork shop's actual rework exposure is dominated by systematic errors, not random ones. A wrong template applied to an entire floor of hospital casework, or a ceiling height assumption carried through a full hotel guestroom package without field verification, is the event that matters. Individual unit errors are noise; the systematic ones are the risk to manage.
You cannot cite a single "industry rework percentage" responsibly. Every published figure represents a specific scope, phase, denominator, and methodology. A millwork subcontractor whose GC cites "rework costs 12% of project budgets" to justify a back-charge should ask: 12% of what? Measured how? Including which phases? The range from 0.38% to 12% in the same literature is not a contradiction — it is a measurement methodology difference. Understanding the difference matters when evaluating contracts, insurance, and contingency.
The cost of good shop drawings is small relative to the cost of bad ones. If even a conservative 2–3% of construction-phase cost is at risk from rework events on a $500,000 millwork scope, that's $10,000–$15,000 in potential exposure. See our millwork drawing rates — a complete drawing package for a scope that size typically runs a fraction of that exposure figure. The economic logic of investing in thorough shop drawings as a risk management tool holds at almost any plausible rework rate assumption drawn from the literature.
Sources Cited
- Construction Industry Institute. Costs of Quality Deviations in Design and Construction (RS10-1). 1989. construction-institute.org
- Construction Industry Institute. The Field Rework Index: Early Warning for Field Rework and Cost Growth (RS153-1). 2001. construction-institute.org
- Hwang, B., Thomas, S.R., Haas, C.T., Caldas, C.H. "Measuring the Impact of Rework on Construction Cost Performance." Journal of Construction Engineering and Management. 2009.
- Love, P.E.D. et al. "The costs of rework: insights from construction and opportunities for learning." Production Planning & Control. 2018. Curtin University repository
- Gallaher, M.P. et al. (NIST). Cost Analysis of Inadequate Interoperability in the U.S. Capital Facilities Industry (GCR 04-867). 2004. nvlpubs.nist.gov
- American Society of Civil Engineers. "How Much Does Field Rework in Construction Actually Cost?" Civil Engineering Source. January 22, 2026. asce.org
- De Lapp, J.A. Impacts of CAD on the Submittal Process. Master's thesis, Texas A&M University. 2005. Texas A&M OakTrust
- Kentley Insights via MarketResearch.com. "Custom Architectural Woodwork and Millwork Manufacturing" (NAICS 337212). 2024. Industry sales 2023: $8.9 billion.
For scope questions and pricing, see our millwork shop drawing services or review our drawing rates.
Frequently Asked Questions
Reduce Your Rework Exposure Before the First Unit Ships
The research is consistent on one point: design documentation errors and omissions are the largest driver of field rework cost. Thorough shop drawings that surface conflicts before fabrication begins are the most direct way to reduce that exposure.
Get a Free Quote