Now accepting guest editorial from SAP, Oracle, Workday and ServiceNow practitioners.Pitch a guide
The Sign-Off Table
All statistics

Statistics · Published

Software Quality Cost Statistics (2026)

CISQ puts the US cost of poor software quality at $2.41 trillion, and 60% of organizations in a 2026 survey ship untested code.

Compiled by Matt Angerer
20 sources · 12 min read

Software quality cost figures get quoted constantly and traced rarely. This roundup gathers the ones a test manager, Testing CoE lead or program director can cite in a sign-off conversation, and it flags where the numbers disagree. Most come from 2024 to 2026 research; older studies are kept and marked as historical.

Key takeaways

$2.41T

Estimated annual cost of poor software quality in the US, 2022 (CISQ lower-bound estimate)

Source: CISQ, 2022

60%

Of organizations deploy untested code to production, per a survey of 2,501 IT and QA leaders fielded April 2026

Source: Tricentis, 2026

67%

Of SAP users have fully or partly moved to S/4HANA, per the Precisely and ASUG survey

Source: Precisely and ASUG, 2026

35%

Of organizations feel fully prepared to manage AI agents at scale, per Tricentis's 2026 report

Source: Tricentis, 2026

The macro cost: what poor quality costs the economy

The most-cited top-line figure comes from the Consortium for Information and Software Quality. Its 2022 update estimates that the cost of poor software quality in the US has grown to at least $2.41 trillion, though the increase was smaller than the jump it recorded between earlier editions. The same report puts accumulated software technical debt at about $1.52 trillion.

CISQ's 2020 edition gave a lower total. That report concluded that poor software quality cost the US upwards of $2.08 trillion in 2020, counting operational software failures, poor-quality legacy systems and unsuccessful projects.

CISQ's 2022 report names three drivers: cybercrime losses from existing vulnerabilities, supply chain problems in third-party components, and technical debt, which it calls the biggest obstacle to changing existing code bases.

CISQ's technical guidance also makes a point that matters for ERP teams. Software flaws at the technology and system level account for 8% of total errors, but they consume over half the effort spent on fixing problems and lead to 90% of the most serious production issues. The page is undated, so treat that split as CISQ's standing estimate rather than a 2022 finding.

CISQ 2018 report (rough aggregate estimate)$2.8T
CISQ 2020 report$2.08T
CISQ 2022 report$2.41T

Source: CISQ, 2018 report; CISQ, 2020 and 2022 reports. Bars are scaled to the 2018 figure. The 2018 number is not comparable to later editions (see the conflicts section).

Project outcomes: how often IT and ERP projects miss

The Standish Group's CHAOS series is the longest-running count of project outcomes, but its methods have been contested. The latest edition we could verify is from 2020, which remains the most recent full CHAOS report we know of.

  • CHAOS 2020: 31% of projects were successful, 50% were challenged and 19% failed.
  • By project size: small projects performed far better, at roughly 90% success, while large projects had under 10% success.
  • Historical (1994): Standish reported a 16 percent success rate, another 53 percent of projects challenged and 31 percent failed outright.
Successful (on time, budget and scope)31%
Challenged50%
Failed or canceled19%

Source: Standish Group CHAOS 2020, as summarized by OpenCommons. Secondary summary; we did not retrieve the Standish report itself.

One academic re-analysis disputes how the Standish figures were built. When the authors applied the Standish definitions to initial forecasts, they found success rates of 59 percent and 55 percent in their samples. That is a different population from the CHAOS survey, so it challenges the method rather than the headline.

For ERP programs specifically, Panorama Consulting's 2026 ERP Report is the most recent independent benchmark we found. More than a quarter of the organizations it studied exceeded their project budgets, with additional technology needs cited as the leading cause. Its sample is 170 organizations, with data gathered from January 2025 through January 2026.

Timelines moved in the other direction. The 2024 Panorama report put the median ERP implementation at 15.5 months, based on an August 2023 survey. The 2026 report puts the median project timeline at 9 months. The survey windows differ, so read the change as directional.

Panorama 2024 report (survey Aug 2023)15.5 months
Panorama 2026 report (survey Jan 2025 to Jan 2026)9 months

Source: Panorama Consulting Group, 2024 and 2026 ERP Reports, as summarized by erp-software.org. Bars are scaled to the 2024 figure.

Defect cost: the multiplier and where it comes from

The claim that a defect costs far more to fix late than early is widely repeated, but the multipliers vary by source and most trace back to secondary write-ups of older work.

  • IBM Systems Sciences Institute data, as relayed by one industry write-up: fixing a bug in production costs roughly 100 times as much as fixing it during the requirements phase.
  • A different reading of the same lineage gives approximately 1x in requirements, 6.5x in design, 15x in coding, 40x in testing and 100x in production.
  • A NASA error-cost study is cited for a 29 times multiplier for fixing a requirements error in production versus catching it in planning. We could not retrieve the NASA source itself.
  • Boehm and Basili's 2001 IEEE Computer work is summarized as showing a production defect can cost around 100 times more to fix than one caught during design on large projects.
  • Historical: NIST's 2002 study estimated the annual US cost of inadequate software testing infrastructure at $59.5 billion.

Use these as directional ranges, not benchmarks. The honest number for your program is the one from your own defect-escape log.

Untested code and the losses it causes

Tricentis's Quality Transformation Report is the most recent survey of this question. Its 2026 edition surveyed 2,501 IT and QA leaders across six countries, with fieldwork in April 2026 across the US, UK, Ireland, Germany, Japan and Singapore.

  • Six in 10 organizations still report deploying untested code, consistent with 63% in 2025.
  • The cause has shifted. In 2025, organizations largely attributed untested deployments to accidental quality slips (40%).
  • On cost, one in five organizations (20%) reports losing more than $1 million annually to poor software quality, mainly from security and compliance failures (30%) and technical debt and rework (28%).
  • Nearly half (45%) estimate losses between $500,000 and $1 million.
  • Sector splits: 64% in financial services, 63% in retail and 58% in energy and utilities report deploying untested code.
Losses above $1M per year20%
Losses of $500K to $1M per year45%

Source: Tricentis, 2026 Quality Transformation Report. Bars are on a 100% scale.

60%

Source: Tricentis, 2026

The 2025 edition of the same report gave a different loss figure, which we flag in the conflicts section. It also reported that 66% of global organizations were at significant risk of software outages in the next year, with 63% shipping code without fully testing it.

AI in quality engineering: adoption is wide, scaling is narrow

The World Quality Report 2025-26, published by Capgemini, Sogeti and OpenText, surveyed 2,000 interviews carried out during June and July 2025 by Coleman Parkes Research. Its headline numbers:

  • 89% of responding organizations are piloting or deploying Gen AI-augmented workflows, with 37% in production and 52% in pilot phases.
  • Only 15% have achieved enterprise-scale deployment.
  • The share of non-adopters rose to 11%, up from 4% in 2024.
  • Organizations report an average productivity boost of 19%, but one third have seen minimal gains.
  • 50% report their organizations lack AI/ML expertise, which is unchanged from 2024.
  • For ERP testing, the report excerpt we reviewed says only 6% of enterprises use Gen AI in ERP testing, as analysts and developers juggle broader QE responsibilities. Verify against the full PDF before quoting, since our copy was partial.
Piloting or deploying Gen AI in QE89%
In pilot phase52%
In production37%
Enterprise-scale deployment15%

Source: Capgemini, World Quality Report 2025-26, survey of 2,000 executives, June to July 2025. Bars are on a 100% scale.

Tricentis's 2026 data points to a trust gap in the same area. 93% of C-level leaders feel confident in their testing strategy, while 30% of QA and DevOps leaders do not. Its report also says the trust in AI agents to make release decisions dropped from 48% to 34% year over year.

35%

Source: Tricentis, 2026 Quality Transformation Report

Test data: the constraint most programs report

The World Quality Report found that 60% of organizations struggle with secure, scalable test data, while 58% cite challenges in adopting AI-powered tools. Synthetic data is moving up the list: its use in testing rose from 14% in 2024 to an average of 25% in 2025. That is a two-point series, so treat the direction as clear and the slope as approximate.

S/4HANA migration: where the SAP base stands

The SAP numbers come from user-group and benchmark surveys with different definitions, so they are easiest to use as a range rather than a single figure.

  • Precisely and ASUG report that 67% of respondents are fully or partially migrated, up from 59% in 2025 and 50% in 2024, and another 23% expect to finish within 24 months.
  • The ASUG 2026 Pulse, a separate ASUG instrument, reports that 56% of customers are either live on S/4HANA or have begun their migration, up from 45% in 2024.
  • Budget constraints are now the top challenge facing SAP customers, cited by 61% of respondents.
  • SAPinsider's 2026 benchmark finds 55% of companies report having deployed SAP S/4HANA, but only 34% indicate a complete transition.
  • Gartner's count at the end of 2024 put 39% of SAP ECC customers, about 14,000 of 35,000, as having migrated to S/4HANA. Gartner projected 17,000 holdouts, or nearly half of the ECC customer base, by 2027 at the then-current migration rate.
  • IDC's view is more optimistic about migration, but only slightly. It expects 40% to 45% of ECC users to remain with the older ERP through 2027.

Source: Precisely and ASUG, 2024 to 2026 surveys. Share of respondents fully or partly migrated to S/4HANA.

Deployed S/4HANA (any form)55%
Completed transition34%
Traditional S/4HANA deployment29%
S/4HANA Cloud26%

Source: SAPinsider, ERP Migration and Transformation 2026 benchmark. Bars are on a 100% scale.

Where the sources conflict

These disagreements are the ones most likely to come up in a steering committee. We have not picked a winner in any of them.

  • CISQ 2018 versus 2020. The 2018 report put the cost at approximately $2.8 trillion, and CISQ itself described that figure as a rough estimate that aggregates publicly available source material. The 2020 edition landed at $2.08 trillion. Do not read this as a decline.
  • Tricentis, 2025 versus 2026. Secondary coverage of the 2025 report says 42% of global firms estimated annual losses over $1 million from poor software quality. The 2026 edition reports 20% above $1 million. The question wording or sample may differ; check the full reports before comparing.
  • SAP migration share. Precisely and ASUG show 67% fully or partly migrated, while the ASUG 2026 Pulse shows 56% live or started. The wording differs (fully or partly migrated versus live or begun), so the gap is partly definitional.
  • Gartner's 39%. Gartner's end-2024 figure is reported both as customers who had licensed S/4HANA and as customers who had migrated. Quote it with its definition attached.
  • Defect multipliers. Secondary sources give 100x, 40x, 29x and 15x for production or late-phase fixes. The differences trace to different phases, project types and source studies.

What this means for your sign-off

  • Cite the macro number with its scope. The CISQ total is a national estimate built from public sources. It helps frame a board discussion but will not justify a single program's test budget.
  • Build your own cost-of-escape baseline. Given the spread in defect multipliers, pull defect-escape counts and rework hours from your last two releases. That number will carry more weight with a CFO than any published ratio.
  • Plan for test data as a first-order risk. When 60% of organizations struggle with secure, scalable test data, a migration test plan that assumes the data is ready will slip. Start with the test data guidance for S/4HANA before your first cycle.
  • Put a release gate on untested code. With six in ten organizations deploying untested code, your sign-off should require evidence of coverage for each change, including AI-generated changes. The template library includes an AI agent guardrail register and a review checklist you can adapt.
  • Build the timeline on the range, not the median. A 9-month median is a benchmark, not a plan. Large ECC conversions with heavy custom code sit well outside it, so use the phase-by-phase migration test plan to size your own window.
  • Use budget pressure as an argument for automation. With budget the top challenge for 61% of SAP customers, present test automation as a way to control the loss bands that Tricentis's respondents report, not as an extra cost.

Frequently asked questions

What does poor software quality cost the US economy?

CISQ estimated at least $2.41 trillion in its 2022 report. Treat it as a national, aggregated estimate rather than a per-program figure.

How many organizations knowingly ship untested code?

Six in 10 organizations still report deploying untested code, according to Tricentis's 2026 survey of 2,501 IT and QA leaders.

How much more does a production defect cost to fix?

Sources disagree. One IBM-based summary gives roughly 100 times the cost of fixing it during requirements, while a second reading of the same lineage gives about 40x in testing and 100x in production. Build your own ratio from your defect log.

What share of SAP customers have moved to S/4HANA?

It depends on the definition. Precisely and ASUG found 67% fully or partially migrated, while SAPinsider found 55% deployed and only 34% with a complete transition.

How many organizations use generative AI at enterprise scale in quality engineering?

The World Quality Report 2025-26 found nearly 90% of organizations pursuing generative AI in their quality engineering practices, but only 15% achieving enterprise-scale deployment.

How often do ERP projects exceed budget?

In Panorama's 2026 ERP Report, more than a quarter of organizations exceeded their project budgets. The sample was 170 organizations surveyed between January 2025 and January 2026.

Sources

  1. Cost of Poor Software Quality in the U.S.: A 2022 Report, CISQ, 2022
  2. CISQ Technical Guidance and Report Archive, CISQ, 2020 and 2022 (undated page)
  3. The Cost of Poor Quality Software in the US: A 2018 Report, CISQ, 2018
  4. Tricentis Report: 60% of Global Organizations are Shipping Untested Code, Tricentis, June 2026
  5. 2026 Quality Transformation Report, Tricentis, 2026
  6. Is delivery speed over quality heralding a software crisis?, Digit (coverage of Tricentis 2025 report), 2025
  7. World Quality Report 2025: AI adoption surges in Quality Engineering, Capgemini, Sogeti and OpenText, 2025
  8. World Quality Report 17th Edition 2025-26, Capgemini and Sogeti, 2025
  9. World Quality Report 2025-26, Capgemini, 2025
  10. Panorama Consulting Group Releases Latest Study of ERP Implementation Outcomes, Panorama Consulting Group, March 2026
  11. ERP Statistics 2026, erp-software.org (summarizing Panorama 2024 and 2026 reports), 2026
  12. CHAOS Report on IT Project Outcomes, OpenCommons (summarizing Standish Group), 2020 data
  13. The Rise and Fall of the Chaos Report Figures, academic paper (Vrije Universiteit), date not stated in source
  14. Bug Cost and Escape Rate Report, TestDino (citing IBM Systems Sciences Institute data), 2026
  15. The Cost of Late Testing: Defect Economics for Enterprise Engineering, TotalShiftLeft (citing IBM and NIST), 2026
  16. The 29x Rule: Requirements Errors Cost More Later, Specira (citing NASA error-cost study), 2026
  17. SAP customers say migration is eating their budgets, and AI is next in line, CIO.com (ASUG 2026 Pulse), 2026
  18. Precisely and ASUG survey finds 67% of SAP users have fully or partly migrated to S/4HANA, Compare the Cloud, November 2026
  19. SAPinsider Benchmark Research: ERP Migration and Transformation 2026, SAPinsider, 2026
  20. Nearly half of SAP ECC customers may stick with legacy ERP beyond 2027, CIO.com (citing Gartner and IDC), 2025

Last updated: November 14, 2026. Statistics are refreshed as new research is published.

More statistics

How this page is made: statistics are gathered from published research with AI-assisted search, every figure is linked to its source, and the page is refreshed as new data is published. Found an error? Tell us.