Construction Rework Statistics: What’s Actually Measured
The industry’s most-quoted rework and RFI figures, each traced to its primary source and graded for what it really is. Thirteen held. Seven did not.
Short answer: Most of the rework and RFI statistics in circulation trace back to six primary documents. Several of the most-quoted figures trace back to nothing at all. This page holds each number against its source and says plainly what it is — something counted, something estimated, something computed from two estimates, or something the cited source does not contain.
Every figure was checked against the primary document, not a secondary citation. Where a source is paywalled or member-gated, the entry says so rather than resting on second-hand attribution. Two of the struck figures were published on this website until September 2026, which is the clearest evidence we have for how easily they spread.
Estimated survey or interview
Derived computed from estimates
Struck source does not contain it
Rework as a share of contract value
The rework tables cited across the industry come from a 2009 study of 359 projects by Hwang, Thomas, Haas and Caldas, reproduced in the Navigant Construction Forum’s August 2012 perspective. Almost every secondary citation credits Navigant. The measurements are Hwang et al.’s.
The single most useful figure for a building project: mean total field rework on buildings runs 4.6% of contract value, and design error and omission alone accounts for 1.5% within that. Across all building types the figures are 5.2% and 1.8%.
| Population | Projects | Design error & omission | Total field rework |
|---|---|---|---|
| Infrastructure | 14 | 0.9% | 5.7% |
| Add-on | 47 | 1.3% | 3.6% |
| Grass roots | 48 | 1.3% | 4.0% |
| Buildings | 32 | 1.5% | 4.6% |
| All projects | 177 | 1.5% | 4.9% |
| Heavy industrial | 102 | 1.6% | — |
| All building types | 179 | 1.8% | 5.2% |
| Modernization | 82 | 1.8% | 6.5% |
| Light industrial | 31 | 3.2% | 9.3% |
Measured. Hwang, Thomas, Haas & Caldas, “Measuring the Impact of Rework on Construction Cost Performance” (2009), 359 projects — tabulated in Navigant Construction Forum, “The Impact of Rework on Construction,” August 2012, pp. 7–9.
Cite the population, not the range. A figure drawn from light industrial does not describe a building, and the population is the first thing dropped in retelling.
Direct and indirect rework cost
Navigant’s synthesis of the prior literature puts the median direct cost of rework at 5.04% of project cost, within a range of 4.03% to 6.05%. Adjusted for indirect costs at the roughly 80% uplift the University of Alberta work supports — a factor of 1.8 — the same range becomes 7.25% to 10.89%, median 9.07%. The uplift is an adjustment, not a second measurement. Quote 5.04% and 9.07% as a pair, or neither.
Design error cost
The strongest peer-reviewed source is Lopez and Love, “Design Error Costs in Construction Projects,” ASCE Journal of Construction Engineering and Management, May 2012. Across 139 projects it reports a mean direct cost of 6.85% of contract value and a mean indirect cost of 7.36%.
Two cautions are almost always dropped. First, these are questionnaire estimates provided by respondents, not costs counted from project records — peer review is not measurement. Second, the dispersion is enormous: the 6.85% mean carries a standard deviation of 12.72% against a maximum of 98%. The mean does not describe a typical project. The commonly-quoted “14.2% total” is the sum of the two figures, computed by others; the paper states them separately.
Requests for information
One report supplies nearly every RFI statistic in circulation: Navigant Construction Forum, Impact & Control of RFIs on Construction Projects, April 2013. Its measured figures rest on 1,362 projects from the Aconex document-management dataset, covering 1,083,807 RFIs on projects started between 2001 and 2012. Its cost figures rest on interviews with three firms. Both halves are quoted with equal confidence, and only one has earned it.
What was measured
- 796 RFIs per project on average, across the 1,362-project dataset.
- 9.9 RFIs per $1 million of construction cost across the 826 projects with cost data.
- 17.2 RFIs per $1 million on projects between $5 million and $50 million.
- 1.1 RFIs per $1 million on projects between $1 billion and $5 billion.
The counterintuitive result is the one most worth citing and the one least often cited: smaller projects generate substantially more RFIs per dollar than large ones — roughly fifteen times the rate at the extremes of the sample. RFI exposure does not scale with contract value.
What was estimated
The $1,080 average cost to review and respond to one RFI is built from four administrative hours at a median $82/hour plus four technical hours at a median $188/hour, both rates already carrying a 2.5 multiplier. The report is explicit about what this is, in a footnote that rarely travels with the number:
“The Navigant Construction Forum interviewed senior staff from three design professional / construction management firms … These costs are not specific to any particular design discipline, industry sector or geographic area but are estimated costs solely for the purposes of this research perspective. Actual costs for a specific project and/or given design or construction management firm may vary widely.”
Three firms is not a sample. Quote $1,080 as an order of magnitude or not at all.
What was derived
The figure of $859,680 as the total cost of RFI review and response on an average project is 796 multiplied by $1,080 — one measured figure times one three-firm estimate. It inherits every limitation of the $1,080 and adds no rigour of its own. It is among the most-repeated RFI numbers in the industry, and it is an arithmetic product rather than a finding.
The 13.2% that is not from this dataset
The widely-cited figure that 13.2% of RFIs are “not justifiable” — meaning the answer already existed in the contract documents, or the RFI questioned means and methods — is real, but it is not from the 1,362-project dataset. Navigant quotes it from a separate study by Hanna, Tadt and Whited of major highway projects. The two are routinely merged into a single “Navigant found…” claim. If you are talking about buildings, you are citing highway data.
The cost of bad data
The Autodesk and FMI study Harnessing the Data Advantage in Construction is cited constantly, almost always with the wrong year and frequently with figures it does not contain. It was published 14 September 2021 and surveys 2020. It is very widely cited as “Autodesk/FMI 2022,” including in our own earlier working files.
What it actually states: bad data may have cost the global construction industry $1.85 trillion in 2020, and decisions made on bad data are estimated to have cost $88.69 billion in rework alone, accounting for 14 percent of all rework performed in 2020. Note the 14 percent — Autodesk’s own marketing pages have published 16 percent, which the study does not support.
The per-contractor figures come from the trade-press write-up rather than the study release: for a contractor generating $1 billion in revenue, losses from bad data could be as high as $165 million, including up to $7.1 million in avoidable rework. The qualifier is load-bearing. Both are ceilings, not central estimates. Written as “roughly $7.1 million,” which is common, the claim changes meaning.
The 52% that means two different things
The figure behind “nearly half of all rework comes from coordination failures” is real, but it belongs to a different study than the one it is usually credited to. It is from PlanGrid and FMI, Construction Disconnected (2018), a survey of roughly 600 construction leaders. Its words: “On average, 52 % of all rework globally is caused by poor data and miscommunication.” Poor data and miscommunication is a broader category than coordination failures, and the figure is global.
The same report reaches a different number for the United States, and the two collide in a way that is easy to miss. Asked what they attribute rework to, 22% of US respondents pointed to poor project data and 26% to poor communication among stakeholders — 48%. The report then states that “the remaining 52% of rework could be caused by a number of other factors including design changes or issues, faulty or delayed materials, unforeseen conditions and more.”
So in the United States, 52% is the share of rework that is not caused by poor data and communication, and the correct figure for the causes people mean is 48%. The two values being identical is a coincidence of the data, which is precisely why the misquote survives. Quote 52% only with the word “globally” attached.
Measured. PlanGrid & FMI, Construction Disconnected (2018), survey of approximately 600 construction leaders worldwide.
Figures that did not survive the trace
Each of the following is in wide circulation. Several carry a citation, which makes them more dangerous than a bare assertion — an attributed figure reads as verified.
“The average rework event costs $50K–$500K (Autodesk/FMI).”
The study states no per-event rework cost of any kind. The figure has acquired a citation it never had a claim to.
“Nearly half of all rework stems from coordination failures (Autodesk/FMI).”
Not in the study. The nearest real figure is the 14% of rework attributable to bad-data decisions, which is a different proposition.
The real figure, and the trap inside it, are set out under The 52% that means two different things.
“An error caught during construction costs roughly ten times more to fix than one caught in design.”
There is no construction measurement behind this multiplier. The 1–10–100 rule is general quality management; Boehm’s Rule of Ten is software engineering; the MacLeamy curve (CURT, 2004) is an explicitly unmeasured concept diagram drawn to illustrate an argument. The concept is defensible and worth attributing. The number is not.
“The average cost of a clash discovered in the field is $1,500.”
Traces only to vendor blogs citing each other. No study, no sample, no method located.
“60–70% of RFIs are preventable with better coordination.”
No source states it. The closest supportable figure is 30–50% of RFIs stemming from errors, omissions or ambiguities in the documents, consistently attributed to the Construction Industry Institute — but consistently at second hand, never verified against a CII primary document, whose research is member-gated.
“Coordination-driven change orders add 3–8% to project cost.”
The best-sourced comparator, design error and omission as a measured cause of field rework, runs 1.5% to 3.2% depending on population. The circulating figure is roughly double that at the bottom and well beyond it at the top.
“Conflicts split 40% specification mismatches, 35% missing coordination, 15% code violations, 10% spatial clashes.”
Published in two places with two different denominators and no underlying catalogue matching either description. A breakdown that survives a change of denominator was never counted.
How to read a construction statistic
Five habits caught every error on this page, and none require access to anything paywalled.
Separate measured from estimated from derived. A count from 1,362 projects and an interview with three firms are both “from the study.” One is evidence. The other is a working assumption the authors labelled as such, in a footnote that rarely travels with the number.
Read the population before the percentage. Rework “runs 9.3%” is true — of 31 light industrial projects. For buildings the same table says 4.6%. The percentage means nothing without the cut it came from.
Watch for derivative drift. Study, to press release, to trade write-up, to vendor blog, to slide. Each hop drops a qualifier. “Up to $7.1 million” becomes “roughly $7.1 million”; 14% becomes 16%; a highway-projects finding becomes a general one. Go to the source document, not to whoever cited it.
Distrust a mean without a spread. A 6.85% mean with a 12.72% standard deviation and a 98% maximum describes a distribution with no typical case. Reporting the mean alone implies a central tendency the data does not have.
Treat an attached citation as a claim, not a verification. The most persistent errors here all carry citations. An unsourced number invites scepticism; a wrongly-sourced one disarms it. If a figure matters to a decision, open the document.
Sources
- Hwang, B., Thomas, S. R., Haas, C. T. & Caldas, C. H. “Measuring the Impact of Rework on Construction Cost Performance” (2009), 359 projects.
- Navigant Construction Forum. The Impact of Rework on Construction. August 2012.
- Navigant Construction Forum. Impact & Control of RFIs on Construction Projects. April 2013.
- Hanna, A., Tadt, E. & Whited, G. “Request for Information: Benchmarks and Metrics for Major Highway Projects.”
- Lopez, R. & Love, P. E. D. “Design Error Costs in Construction Projects.” ASCE Journal of Construction Engineering and Management, May 2012. DOI 10.1061/(ASCE)CO.1943-7862.0000454.
- Autodesk & FMI. Harnessing the Data Advantage in Construction. 14 September 2021.
This page carries no figures of our own. Corrections are welcome and will be dated and shown rather than silently applied — including corrections to this page. Last verified against primary sources: 12 September 2026.