Project Budget Overrun Statistics: The Data Behind Why Projects Fail (2026)
Overrun statistics from PMI, McKinsey, the Standish Group and academic research. Each figure links to the document that states it, with the date we last read that document. This page compiles published figures; it holds no data of our own.
Key Headline Statistics
Overrun Rates by Industry
Figures come from different study methodologies; the "worst case" column shows documented extreme examples, not the typical project. Read dates for every document are in the sources ledger.
| Industry | Projects Overrun | Avg Overrun % | Worst Case | Data Source | Year |
|---|---|---|---|---|---|
| Transportation infrastructure | 86% (almost 9 in 10) | 28% | Rail +44.7% on average | Flyvbjerg, Holm & Buhl (258 projects), read 8 October 2026 | 2002 |
| IT / Software | 71% challenged or failed | 45% (large IT, McKinsey 2012) | 447% tail average (Flyvbjerg & Gardner) | Standish 2015 / McKinsey 2012 / Flyvbjerg & Gardner | 2015 / 2012 / 2023 |
| Government IT (public sector) | nearly 1 in 2 (about 1 in 3 private) | ~3x private-sector peers | 14% exceed budget by 100%+ | McKinsey with Oxford Global Projects (6,003 projects, 2001-17) | 2022 |
| Megaprojects (>$1B; mining, oil and gas, infrastructure) | 98% (overruns above 30%) | 80% (Exhibit 2 sample) | ~1,830% (James Webb, 1997 study to 2021 GAO estimate) | McKinsey, The Construction Productivity Imperative | 2015 |
| All project types (16,000+ projects) | 52% miss budget | 91.5% miss budget or schedule | Nuclear waste storage: +238% mean | Flyvbjerg & Gardner, How Big Things Get Done (book) | 2023 |
Overrun Rate by Project Size
One of the most consistent findings across all research: larger projects fail at dramatically higher rates. In the Standish Group CHAOS 2015 data, small projects succeed at roughly ten times the rate of the very largest “grand” projects, and failure rises at every step up in size. Standish sizes projects by labour content and cost rather than by a published dollar threshold, so the bands below are its own five categories.
| Project Size | Success Rate | Challenge / Overrun Rate | Failure Rate |
|---|---|---|---|
| Small | 61% | 32% | 7% |
| Moderate | 24% | 64% | 12% |
| Medium | 12% | 62% | 26% |
| Large | 11% | 59% | 30% |
| Grand | 6% | 51% | 43% |
“Project size by CHAOS resolution”, Standish Group CHAOS Report 2015, page 3 (Modern Resolution: OnTime, OnBudget, with a satisfactory result; FY2011–2015). Each row sums to 100%. Read from the primary report; full methodology and definitions on our Standish CHAOS Report citation page and project success rate by size.
Why Different Sources Report Different Figures
You will find sources citing 43%, 69%, and even 91.5% of projects going over budget. These are not contradictory; they measure different things:
- ‣PMI 43% (Pulse 2018): 4,455 practitioners estimated, for projects their organisation completed in the past 12 months, the share that finished within the initial budget. The mean was 57%; 43% is the remainder
- ‣Standish 71% (CHAOS 2015): software projects only, FY2015; 52% challenged (late, over budget, or with an unsatisfactory result) plus 19% failed. CHAOS 2020 is quoted at 69% (50% plus 19%); that report is sold by Standish and we have not read it
- ‣Flyvbjerg 9-in-10 (2002): transport infrastructure projects (258 sampled), actual construction cost against the estimate at the decision to build
- ‣McKinsey 98% (2015): megaprojects (over $1B), share with cost overruns above 30% (source and the separate 80% average)
- ‣Flyvbjerg 91.5% (2023): Across 16,000+ projects of all types, the share that miss budget, schedule, or both; only 8.5% deliver on both
Citing these statistics: this page compiles figures published by others; it is not a dataset and holds no data of our own. Cite the primary document for each figure (linked beside it, with the date we read it). The full list, with what each document is cited for, is in the sources ledger.
Is This Stat Real? Common Cost-Overrun Figures, Checked
Several cost-overrun figures circulate in AI summaries, slide decks, and vendor blogs with the source stripped off or the caveat dropped. We trace each to its primary study and flag the way it is usually mis-stated. If you have seen one of these cited, check it against the real figure before reusing it.
“IT projects overrun by 27% on average, so the risk is modest.”
Real number, dangerous truncation
The 27% average is genuine: Flyvbjerg & Budzier, Harvard Business Review 2011, sample of 1,471 IT projects. But quoting the average alone hides the actual risk. The same study found 1 in 6 of those projects was a “black swan” with a cost overrun of 200% on average and a schedule overrun of almost 70%. The distribution is fat-tailed, not normal, so the mean understates the exposure. Cite the 27% only alongside the 1-in-6 / 200% tail.
Primary source: Flyvbjerg & Budzier, Harvard Business Review, 2011 (1,471 projects)
“$66 billion is lost to IT cost overruns every year.”
Real figure, wrong framing
The $66 billion is real but it is not an annual or global loss figure. It is the total cost overrun across the specific sample of 5,400+ large IT projects (initial budgets above $15M) studied by McKinsey with the University of Oxford in 2012, where large IT projects ran 45% over budget and 7% over time on average. Reframing a one-off sample total as a recurring annual loss inflates it. Cite it as the sample total it is.
Primary source: McKinsey-Oxford, Delivering large-scale IT projects, 2012 (5,400+ projects)
“85% of construction projects go over budget, with a 28% average overrun.”
Traceable, but check which study
The most defensible primary source for an ~86% overrun rate with a ~28% magnitude is Flyvbjerg, Holm & Buhl (2002), a study of 258 transportation infrastructure projects across 20 countries built between 1927 and 1998 (almost 9 in 10 underestimated cost; rail averaged +44.7%, roads +20.4%). The rounded “85% / 28%” version is often loosely attributed to a generic “McKinsey Global Institute, 20 countries, 70 years” study without a retrievable citation. When you need a footnote that holds up, cite Flyvbjerg's 258-project transport dataset, and note it covers transport infrastructure, not all construction.
Primary source: Flyvbjerg, Holm & Buhl, 2002 (258 transport projects)
See also our full source trace of the widely-circulated “53,000 tech projects, 63% hidden costs” figure, which we could not source to any named study.
Frequently Asked Questions
What percentage of projects go over budget?
PMI's Pulse of the Profession 2018 survey of 4,455 practitioners found that 57% of projects finished within their initial budgets, so 43% did not. Rates vary widely by sector: costs were underestimated in almost 9 out of 10 transport infrastructure projects (Flyvbjerg, Holm and Buhl 2002), 98% of megaprojects overrun by more than 30% (McKinsey 2015), and the Standish Group's CHAOS 2015 report counts 71% of software projects as challenged or failed. The wide range reflects different definitions, samples and project types.
Why do different sources cite different overrun percentages?
Different studies use different definitions of 'overrun' (against the original estimate or a re-baselined budget), different project size thresholds, different sectors, and different geographies. PMI's figure is the mean of practitioners' estimates across all project types. Standish covers software projects. Flyvbjerg's 2002 study covers transport infrastructure. A construction project and a software project have very different baseline success rates, so cite the source alongside the figure.
Are project budget overruns getting better or worse?
For megaprojects there is little evidence of improvement: Flyvbjerg (2014) reports that overruns have stayed high and constant for the 70 years for which comparable data exist, and Flyvbjerg, Holm and Buhl (2002) found no improvement in estimate accuracy across projects completed from 1927 to 1998. For software, the Standish Group's CHAOS 2015 data show the success rate flat at 27% to 31% across FY2011 to FY2015. The honest answer is: not measurably better.
Is the 27% average IT project cost overrun figure accurate?
Yes, but it is routinely truncated. The 27% average is from Flyvbjerg and Budzier, Harvard Business Review 2011, across 1,471 IT projects. Quoting the average alone hides the real risk: the same study found 1 in 6 of those projects was a black swan with a cost overrun of 200% on average and a schedule overrun of almost 70%. The distribution is fat-tailed, so the mean understates the exposure. Cite the 27% only alongside the 1-in-6 / 200% tail.
Is $66 billion really lost to IT cost overruns every year?
No. The $66 billion figure is real but it is not an annual or global loss. It is the total cost overrun across the specific sample of 5,400-plus large IT projects studied by McKinsey with the University of Oxford in 2012, in which large IT projects (initial budgets above $15 million) ran 45% over budget and 7% over time on average. Reframing a one-off sample total as a recurring annual loss inflates it.