IT Project Failure Statistics (2026): What the Data Says About ERP, S/4HANA and AI Programs

Most IT project failure statistics trace back to a few studies, and many get misquoted along the way. This page collects the numbers that hold up when you check them: the overall failure rates, how large overruns get, results specific to ERP and S/4HANA, and the newer data on AI projects. For each figure we give the year, who was surveyed and a link to the source, so you can use it in a steering committee deck.

The short version: about a third of projects fully succeed. Overruns tend to come from a small number of disastrous projects rather than from the average one. And in S/4HANA programs, testing and data migration are named among the phases teams most often underestimate.

Key takeaways

How often IT projects succeed, struggle or fail

The best-known benchmark is the Standish Group's CHAOS research. It sorts projects into "successful" (on time, on budget, with the agreed features and satisfied users), "challenged" (finished, but late, over budget or with reduced scope) and "failed" (canceled or never used). The full 2020 report is paywalled, so the figures below come from a published summary of it.

Source: Standish Group CHAOS 2020, as summarized by BudgetOverrun.com (historical; secondary summary of a paywalled report)

BCG's more recent survey of executives found a similar pattern for large programs. In its 2024 Build for the Future study, BCG asked more than 1,000 C-suite executives in 59 countries about tech programs that consumed more than 3% of the annual tech budget. More than two-thirds of those programs were not expected to finish on time, on budget or within their defined scope.

Source: BCG, "Most Large-Scale Tech Programs Fail," November 2024

• BCG estimates that a failed program can cost a typical large company more than €20 million a year.
• In the same study, more than 55% of respondents said their company doesn't manage the full portfolio of tech projects, including the dependencies between projects in system replacement programs (BCG, 2024).
• In an earlier 2024 BCG survey of C-suite executives, nearly half said more than 30% of their organization's technology development projects were both over budget and late.
• PMI's 2024 Pulse of the Profession reports an average project performance rate of 73.8% across all respondents. This measure is PMI's own and is far more forgiving than Standish's "successful" category, so don't compare the two numbers directly.

Project size and type: the strongest predictors in the Standish data

Standish has long argued that project size predicts outcome better than anything else. Its FY2011–2015 breakdown, which uses the "Modern" definition of success, is historical but still the most detailed public version of this data.

Success rate by project size. Source: Standish Group CHAOS 2015 (FY2011–2015), as tabulated by BudgetOverrun.com (historical)

In the same dataset, 43% of "grand" projects failed outright, compared with 7% of small projects (CHAOS 2015 via BudgetOverrun.com). The type of project also mattered a lot, which is relevant to the clean-core debate in SAP programs:

Success rate by project type, FY2011–2015, Modern definition. Source: Standish Group, CHAOS Report 2015 (historical)

• Government projects had the highest failure rate of any industry in that dataset, at 24% (Standish, 2015).
• Projects built from scratch with traditional methods had the highest "challenged" rate, at 61% (Standish, 2015).

Cost overruns: the average is not the risk

The McKinsey and University of Oxford study of large IT projects (initial budgets above $15 million) is still the most widely cited source on overruns. It dates from 2012 and is historical, but it remains the baseline that newer research builds on.

• On average, large IT projects ran 45% over budget and 7% over time, and delivered 56% less value than predicted (McKinsey, 2012).
• Seventeen percent of IT projects went so badly that they could threaten the company's survival (McKinsey, 2012).
• Each additional year of planned duration raised cost overruns by 15% (McKinsey on Finance, No. 45).
• Together, the projects in the study overran their budgets by $66 billion (McKinsey on Finance, No. 45).

Average cost overrun, projects above $15 million (2010 dollars). Software projects also averaged a 33% schedule overrun. Source: McKinsey–Oxford, McKinsey on Finance No. 45 (historical)

Bent Flyvbjerg and Alexander Budzier's work explains why the average misleads. In their sample of 1,471 IT projects, the average cost overrun was 27%. But one in six projects was a "black swan," overrunning cost by 200% on average and schedule by almost 70% (historical, 2011). A 2026 follow-up in the Project Management Journal compared 23 project types and found IT was the only one whose cost risk is mathematically unbounded:

Source: Flyvbjerg, Budzier, Aaen, Keil and Zottoli, "The Uniqueness of IT Cost Risk," Project Management Journal, 2026

• Projects in that tail overran by 453% on average, and many overran by more (Flyvbjerg et al., 2026).
• The gap between IT's mean and median cost overrun was 73 percentage points, a typical sign of a fat-tailed distribution (Flyvbjerg et al., 2026).
• In a separate sample of 5,392 IT projects, actual cost averaged 1.8 times the estimate. The worst overrun was a small workflow customization budgeted at $1,500 that ended up costing $425,000 (Bent Flyvbjerg, 2023).

Why the numbers differ: McKinsey's 45% covers only projects above $15 million. Flyvbjerg and Budzier's 27% covers a broader mix of project sizes. Both studies show the same thing: a minority of runaway projects accounts for most of the damage.

ERP project outcomes

Gartner's outlook for ERP is the bleakest of the major analysts: more than 70% of recently implemented ERP initiatives will not fully meet their original business case goals, and up to 25% of those will fail catastrophically. Gartner's research note attributes much of this to technology-centric approaches that ignore stakeholder engagement.

Source: Gartner, ERP research overview (prediction; publication date not shown on page)

Panorama Consulting's 2026 ERP Report surveyed 170 organizations between January 2025 and January 2026. Their median annual revenue was $200.5 million, and the median project took 9 months. Panorama's press release says more than a quarter of organizations exceeded their budgets, most often because they needed technology they hadn't planned for. A secondary analysis of the report gives this breakdown:

ERP budget outcomes, 170 organizations, 2025–2026. Source: Panorama 2026 ERP Report, as summarized by XoroSoft (secondary; adds up to 30% over budget, in line with Panorama's "more than a quarter")

• The same secondary summary reports that 22.3% of projects finished late (XoroSoft, citing Panorama 2026).
• Historical comparison: Panorama's 2015 report gave a five-year average ERP project cost of $6.1 million and duration of 15.7 months. About 58% of projects exceeded budget, 65% overran schedule, and 53% realized less than half of the expected benefits (Panorama, 2015). The current sample consists of much smaller, shorter projects, so the drop in overruns is not a like-for-like improvement.

A statistic to stop repeating: The claim that "73% of ERP projects fail, with 215% cost overruns" is widely credited to Panorama. An independent check of Panorama's latest published study found neither figure in it. If either number is in your business case, take it out.

S/4HANA transformations specifically

Horváth's 2025 study is the most detailed recent dataset on S/4HANA outcomes. It surveyed 200 top executives from six countries at SAP-using companies with at least €200 million in annual revenue and at least 200 employees.

• Transformations took 30% longer than planned on average, and more than 60% of companies reported deviations in budget, schedule and quality of results.
• Fewer than one in ten companies that had finished their transformation stayed on schedule (Horváth, April 2025). heise's coverage of the same study puts the figure at ten percent. We cite Horváth's own wording.
• In almost two-thirds of companies, the result fell short of expectations despite the extra time and money (heise, 2025).
• About 78% of respondents said they had packed too many other changes into the transformation (heise, 2025).
• 75% run S/4HANA on cloud hosting (Horváth, 2025).

Share of 200 surveyed organizations. Source: heise, reporting the Horváth study, March 2025

Asked what they would do differently, respondents most often said they would plan more time:

Source: Horváth press release (German), May 27, 2025

Why projects fail: the causes respondents name

Different studies use different wording, but the same few causes keep coming up:

• Scope growth, weak project management and underestimated phases. Horváth's respondents named scope expansion during the project, weak project management, underestimated testing and data migration phases, repeated rework of concepts and processes, and slow decision-making (2025).
• Business–IT misalignment, unrealistic timelines and too few resources. These were the three main causes of delay in BCG's April 2024 survey.
• Weak strategy and stakeholder management. McKinsey found that poor management of strategy and stakeholders caused about half of all cost overruns (McKinsey on Finance No. 45, historical).
• Unplanned technology. Panorama's 2026 report names unexpected additional technology as the leading cause of budget overruns.

Horváth's other 2025 publication describes where S/4HANA customers stand: around 60% are mid-transformation, one in ten are still preparing, and most of those already live are unhappy with the quality of the results. They cite poorly defined processes, poor documentation and a lack of IT support.

Quality, defects and the cost of getting it wrong

There is little public data that isolates testing as the cause of project failure, but the cost of poor software quality is well documented. The Consortium for Information & Software Quality (CISQ) estimates:

• The cost of poor software quality in the US reached $2.41 trillion in 2022 (CISQ, 2022), up from $2.08 trillion in 2020. These are historical figures; CISQ hasn't published a newer estimate that we could find.
• Accumulated software technical debt in the US is about $1.52 trillion, and CISQ calls it the biggest obstacle to changing existing code (CISQ, December 2022).

For SAP programs, the most useful finding here is Horváth's: testing and data migration were among the phases respondents underestimated most often (Horváth, 2025). McKinsey's 2012 study likewise recommended a hard quality gate: run all test cases, fix critical defects and verify the fixes before moving from unit testing to integration testing (McKinsey, 2012).

People and delivery capability

• PMI's 2025 Pulse of the Profession, based on almost 3,000 professionals, found that only 18% of project professionals show high business acumen and 66% show a moderate level.
• In the same report, professionals with high business acumen met business goals, schedules and budgets more often than their peers, and they used nearly three more factors to measure project performance (PMI, 2025).
• Horváth's study director said companies tend to underestimate project complexity and the resources required, and overestimate their own organization's capabilities (heise, 2025).

AI projects: a new failure curve

AI is now part of most ERP roadmaps and delivery toolchains. The early failure data looks a lot like ERP's history, with business value, cost and risk controls named as the main problems.

Sources: Gartner, July 2024; Gartner, June 2025; MIT Project NANDA, reported by Virtualization Review, August 2025. These measure different things and should not be compared directly.

• Gartner gives poor data quality, inadequate risk controls, rising costs and unclear business value as the reasons GenAI projects get abandoned (Gartner, 2024).
• MIT's "GenAI Divide" report links the 95% figure to $30–40 billion in enterprise GenAI spending. It measures business return, not whether the technology was delivered, so treat it as a sign of a value problem rather than a delivery failure rate.
• Gartner estimates that only about 130 of the thousands of vendors marketing agentic AI offer real agentic capabilities (Gartner, 2025).

Gartner poll of 3,412 webinar attendees, January 2025. Source: Gartner, June 2025

What this means for your sign-off

• Plan for the bad tail, not the average. An 18% chance of an overrun averaging 453% (Flyvbjerg et al., 2026) calls for hard go/no-go gates and rehearsed cutovers, not a 10% contingency line. The Go/No-Go Readiness Checklist and Mock Cutover Rehearsal Tracker in our template library are built for this.
• Protect the testing and data migration phases. Horváth's respondents named these as underestimated phases, and 46% would have planned more time in hindsight. Set the test phases' duration from your scope matrix, not from the time left before the go-live date. Our phase-by-phase ECC to S/4HANA testing plan and guide to S/4HANA test data cover how to do that.
• Keep releases small. In the Standish data, small projects succeeded 61% of the time and grand projects 6% of the time. Splitting a program into separately signed-off waves is a risk control.
• Treat added scope as test risk. 78% of Horváth respondents said they put too much into their transformation. Each extra initiative adds integrations to regression testing. Make the test lead a required approver on scope changes.
• Sign off on value as well as defects. Gartner expects more than 70% of ERP initiatives to miss their original business case goals. Add business outcome checks to UAT exit criteria so that sign-off covers whether the business case is still achievable, not only whether the system works.
• Put AI in delivery behind approval gates. With over 40% of agentic AI projects expected to be canceled, review AI-generated test assets before you rely on them. Our AI-DLC guide for ERP teams explains where to place those gates.

Frequently asked questions

What percentage of IT projects fail?

In Standish Group's CHAOS 2020 data, about 19% of projects failed outright, 50% were challenged and 31% succeeded (secondary summary of CHAOS 2020). BCG's 2024 executive survey found that only 30% of companies' large tech programs met their time, budget and scope expectations (BCG, 2024).

What percentage of ERP implementations fail?

It depends on how you define failure. Gartner predicts that more than 70% of recently implemented ERP initiatives will not fully meet their original business case goals, and up to 25% will fail catastrophically (Gartner). On budget alone, Panorama's 2026 study of 170 organizations found more than a quarter went over budget (Panorama, 2026).

How often do S/4HANA migrations go over budget or schedule?

In Horváth's 2025 survey of 200 organizations, about 24% exceeded their budget very significantly and 41% significantly, and transformations took 30% longer than planned on average (heise; Horváth).

How much do large IT projects typically overrun?

McKinsey and Oxford found that IT projects above $15 million ran 45% over budget on average and delivered 56% less value than predicted (McKinsey, 2012, historical). Flyvbjerg's 2026 research shows the bigger risk is the tail: 18.26% of IT projects overrun by more than 50%, and those average 453% (Project Management Journal, 2026).

Is "73% of ERP projects fail" a real statistic?

We couldn't confirm it. An independent check of Panorama's most recent study found neither the 73% failure rate nor the 215% overrun figure that is often credited to it (NAWA, 2026). Panorama's own 2026 data shows more than a quarter of projects over budget.

What share of AI projects fail?

Gartner predicted at least 30% of GenAI projects would be abandoned after proof of concept by the end of 2025, and expects more than 40% of agentic AI projects to be canceled by the end of 2027 (Gartner, 2025). MIT's Project NANDA reported that 95% of organizations saw no measurable return from GenAI (Virtualization Review, 2025).

Sources

1. Study shows: SAP S/4HANA transformations rarely go as planned, Horváth, April 10, 2025
2. Unternehmen unterschätzen Herausforderungen der SAP S/4HANA-Transformation, Horváth, May 27, 2025
3. How to Optimize Your SAP S/4HANA System in and after the Transformation, Horváth, 2025
4. SAP S4/HANA: Changeover takes on average 30 percent longer than planned, heise, March 2025
5. Most Large-Scale Tech Programs Fail—Here's How to Succeed, BCG, November 13, 2024
6. Software Projects Don't Have to Be Late, Costly, and Irrelevant, BCG, April 30, 2024
7. Enterprise Resource Planning to Optimize Operations, Gartner, accessed October 2026
8. What IT Leaders Must Do to Avoid Disappointing ERP Initiatives, Gartner, accessed October 2026
9. Gartner Predicts 30% of Generative AI Projects Will Be Abandoned After Proof of Concept By End of 2025, Gartner, July 29, 2024
10. Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027, Gartner, June 25, 2025
11. MIT Report Finds Most AI Business Investments Fail, Reveals 'GenAI Divide', Virtualization Review, August 19, 2025
12. Panorama Consulting Group Releases Latest Study of ERP Implementation Outcomes, Panorama Consulting Group, March 4, 2026
13. The 2026 ERP Report, Panorama Consulting Group, 2026
14. ERP Implementation Statistics 2026 (secondary summary of Panorama data), XoroSoft, 2026
15. The ERP "Failure Rate" Myth vs. Panorama's Real Data, NAWA, 2026
16. 2015 ERP Report, Panorama Consulting Solutions, 2015 (historical)
17. The Uniqueness of IT Cost Risk: A Cross-Group Comparison of 23 Project Types, Flyvbjerg, Budzier, Aaen, Keil and Zottoli, Project Management Journal, 2026
18. Why Your IT Project May Be Riskier than You Think, Flyvbjerg and Budzier, SSRN / Harvard Business Review, 2011 (historical)
19. The Empirical Reality of IT Project Cost Overruns, Bent Flyvbjerg (LinkedIn), 2023
20. Delivering large-scale IT projects on time, on budget, and on value, McKinsey & Company with the University of Oxford, 2012 (historical)
21. Delivering large-scale IT projects on time, on budget, and on value (McKinsey on Finance No. 45), McKinsey & Company, Winter 2013 (historical)
22. CHAOS Report 2015, The Standish Group, 2015 (historical)
23. Standish CHAOS Report summary (CHAOS 2020 and 2015 tables), BudgetOverrun.com (secondary)
24. Cost of Poor Software Quality in the U.S.: A 2022 Report, CISQ, 2022 (historical)
25. CISQ Publishes the Cost of Poor Software Quality in the US: A 2022 Report, CISQ, December 6, 2022
26. The Future of Project Work: Pulse of the Profession 2024, Project Management Institute, 2024
27. Pulse of the Profession 2025: Boosting Business Acumen, Project Management Institute, 2025

Last updated: October 13, 2026. Statistics are refreshed as new research is published.