Reference / Trust surface

Methodology

Publisher sources, calculation framework, EVM formula stack, refresh cadence, scope boundaries, limitations, and the corrections process. Every overrun statistic on this site is anchored to a named publisher in the table below.

Data verified May 2026

Sources

The following publishers are the primary sources for the budget-overrun statistics and case-study figures on this site. Each row shows what the source informs, the refresh cadence we monitor it on, and the URL for the publisher's own page.

Publisher	Refresh	What we use it for
PMI Pulse of the Profession	Annual	Cross-industry project performance benchmarking. Source for the 43% baseline overrun figure and the cross-sector budget-discipline trend data. PMI surveys thousands of project professionals each year and publishes the headline figures in the annual Pulse of the Profession report.
Standish Group CHAOS Report	Annual (latest used 2020)	IT project success, challenged, and failed rates. Source for the 66% IT projects failed-or-challenged figure. We use the CHAOS Report alongside the published methodology critique (small / large project bias, the success-criteria definition debate) so the headline number is never quoted in isolation.
McKinsey project-performance research	On publication	Large IT project cost performance: the canonical McKinsey-Oxford study (2012, Bloch / Blumberg / Laartz) found large IT projects ($15M+) run 45% over budget on average, 7% over schedule, and deliver 56% less value than predicted. The 447% worst-case overrun figure for IT projects that exceed 50% overrun also comes from this research line.
Bent Flyvbjerg megaproject research (Oxford SBS)	On publication (multiple through 2024)	Megaproject cost overruns and reference class forecasting evidence base. Source for the 85% construction overrun rate, the 28% average construction overrun, the 98% megaproject overrun rate, the 80% average megaproject overrun, and the underlying RCF methodology. Selection criteria for the Flyvbjerg database are published and cited inline where the figures are used.
KPMG Project Management Survey	Biennial	Cross-industry project management benchmarking. KPMG's published surveys on project performance, governance practices, and benefits-realisation rates corroborate or contrast with PMI and Standish findings. Used alongside other surveys to triangulate cross-industry overrun trends rather than as a single authority.
Capers Jones software measurement data	On publication	Capers Jones's long-running software-engineering measurement work (Applied Software Measurement, Software Engineering Best Practices, and related publications) provides IT project cost, schedule, and quality benchmarks. Used to corroborate Standish and McKinsey IT overrun findings with an independent measurement programme.
US BLS construction-sector data (PPI and ECI)	Monthly (PPI) / Quarterly (ECI)	Construction input price index and employment cost index. Provides macro context for construction-sector supply-chain cost drift that informs the discussion of post-2020 construction overruns. Not the source of overrun rates, but the cost-driver context for them.
US BLS IT and computer-system project employment cost	Annual (OEWS)	IT loaded-rate inputs used for IT-project cost modelling on the calculator. OEWS Statistics on computer and information technology occupations provide median wage data used for the IT loaded-rate range cited in calculator presets.
US GAO project audit reports	On publication	Federal project audit findings. Source for US federal procurement overrun rates, DoD acquisition programme cost growth, and the High Risk List of federal programmes with persistent cost-management problems. Primary source for US federal case-study figures on the /government-projects page.
UK NAO major project audit reports	On publication	UK National Audit Office reports on major government programmes. Primary source for HS2, Crossrail (Elizabeth Line), Universal Credit, Edinburgh Trams, and other UK case-study figures. NAO reports are the closest available to a primary record on these programmes; press summaries are not used in place of the underlying audit.
Industry-standard contract types (FIDIC, NEC, JCT, AIA)	On revision	Public reference material for construction and engineering contract families. Used on /legal-consequences for the allocation-of-overrun-risk discussion across lump-sum, cost-plus, GMP, and target-cost contract forms. Used for general reference, not as legal advice on any specific contract.

In scope

●Published research statistics on project budget overrun rates, with source attribution and year.
●Sector-specific overrun rates (construction, IT, government, infrastructure, healthcare) sourced from the publishers above.
●Earned Value Management formula reference (PMI / AACE standard notation) and worked examples.
●Historic case-study overrun figures verifiable from public records (NAO, GAO, project final cost statements).
●Free EVM calculator with construction, IT, and government presets that runs entirely in the browser.
●Contract-type allocation of overrun risk under industry-standard contract families.
●Insurance product framing (builder's risk, professional indemnity, surety bonds, Lloyd's-market cost overrun cover).

Out of scope

●Private or confidential project cost data from any individual organisation.
●Predictions about whether a specific named project will overrun, by how much, or when.
●Internal cost models or proprietary estimating frameworks belonging to any individual organisation.
●Vendor-by-vendor PM software comparisons (no commercial relationships exist to inform such a comparison fairly).
●Country-specific legal advice on contract disputes, variation claims, or adjudication procedure.
●Real-time inflation, escalation, or commodity-index forecasting (the calculator uses static reference ranges).

Calculation framework

EVM formula stack

PMI / AACE standard notation: PV (Planned Value), EV (Earned Value), AC (Actual Cost). CPI = EV/AC. SPI = EV/PV. Three primary EAC methods: EAC = BAC/CPI (assumes current efficiency continues), EAC = AC + (BAC - EV) (optimistic), and EAC = AC + ETC (bottom-up re-estimate). PMBOK 7 fourth method: EAC = AC + [(BAC - EV) / (CPI x SPI)]. VAC = BAC - EAC. TCPI = (BAC - EV) / (BAC - AC) when staying on BAC, or (BAC - EV) / (EAC - AC) when staying on the new EAC.

Reference class forecasting

Flyvbjerg's evidence-based estimating method. Identify 10 or more comparable completed projects, use their actual cost distribution as the anchor, then adjust upward for project-specific risks. RCF is the single most robust counter to optimism bias and strategic misrepresentation at the estimate stage. Used as the foundation of the /prevention page's first strategy.

Overrun percentage math

Overrun% = (AC - BAC) / BAC x 100, where BAC is the originally approved budget and AC is the final actual cost. Worked example: Sydney Opera House overrun was (AUD $102M - AUD $7M) / AUD $7M x 100 = 1,357%. All historic case-study percentages on /famous-examples follow this formula against the originally approved budget, not against any re-baselined intermediate figure.

Contingency sizing bands

Construction: 5-10% contingency on a well-defined scope; 15-25% on early-design or unique-engineering projects. IT: 15-25% on well-understood systems; 30-50% on novel platforms or system-integration projects. Megaprojects: 30%+ minimum given Flyvbjerg's empirical overrun distribution. These bands are reference points, not prescriptions; the calculator and /prevention page tie each band back to the publisher that informs it.

Historic case-study triangulation

Famous case studies (Sydney Opera House, James Webb, Big Dig, HS2, Crossrail, Berlin Brandenburg, Edinburgh Trams) use primary records where available: NAO reports for UK programmes, GAO reports for US federal programmes, project final cost statements where published, and academic case-study research where the primary records are not public. Press summaries are not used in place of the underlying audit.

CPI thresholds and recovery probability

PMI research on EVM finds that final CPI tends to stay within 0.1 of the CPI measured at 20-30% project completion. A CPI below 0.95 at that gate is at-risk; below 0.90 is serious overrun territory. The /earned-value-management page presents these bands with the source caveat that they are empirical patterns, not deterministic predictions for any single project.

Refresh cadence

Pages are re-verified against the source publishers in the first business week of each month. The verification date is held in a single constant (LAST_VERIFIED_DATE) imported by every page. Footer text, Article schema dateModified, and the visible "data verified" labels all read from that single source. Cosmetic refreshes (bumping the date without a substantive review) are structurally not possible: the constant rolls the entire site forward simultaneously, so the date and the underlying review have to match.

Out-of-cycle refresh triggers (handled within the week of publication):

●New Standish CHAOS Report edition.
●New Flyvbjerg / Oxford megaproject database publication or research paper.
●Major NAO or GAO public-project audit report with material findings on a case study cited on the site.
●PMI Pulse of the Profession annual refresh.
●KPMG Project Management Survey biennial release.

Limitations

●Calculator outputs are working bands, not predictions for any specific project; project-specific risks dominate the residual variance.
●Vendor-published statistics carry methodology critiques (CHAOS sample bias, Flyvbjerg selection criteria) that are cited inline on the relevant page.
●Megaproject figures use Flyvbjerg's published database; the selection criterion is explicit (projects above defined cost thresholds in the relevant sector classification).
●Historic case-study figures shift as final audits land (e.g. HS2 is an in-flight programme; published cost figures lag the latest committed-cost position).

Editorial position

This site is operated by Digital Signet, an independent reference-content studio. Digital Signet does not sell project-management software, does not act as a project-rescue consultancy, does not run a programme-governance practice, and does not accept paid placements from any vendor in the project-management space. See /about for the operator and the wider portfolio context.

Editorial direction is set by Digital Signet's editor. Drafts are produced through Digital Signet's autonomous AI development methodology and reviewed against this methodology page before publication. Where sources disagree the page presents the range with the methodology context, not a single point estimate.

Corrections

For methodology questions, stale figures, or scenarios that do not fit cleanly, email [email protected] with the page URL and the source you would like cited. Substantive corrections are typically actioned within five business days.