The Billion Dollar Scoping Problem

Development speed is a huge focus in software right now. Everyone talks about how generative AI is compressing timelines, how coding assistants are writing production code, and how prototypes that took weeks now take hours.

This is all great, but speed won’t fix the most expensive problem in software: building the wrong thing.

When scientific organizations reach out after a project has gone sideways, we trace the failure back and it almost always starts in the same place. Unclear, incomplete, or constantly shifting requirements.

The data on scoping failures is overwhelming. Let’s look at the numbers.

The Scale of the Problem

• $2.41 trillion cost of poor software quality in the U.S. in 2022 (CISQ)
• $260 billion wasted annually on failed U.S. software projects from scoping failures alone
• 69% of software projects fail or are “challenged” (Standish Group CHAOS, 2020)
• 45% average budget overrun on large IT projects (McKinsey-Oxford, 5,400 projects)
• 56% less value delivered than predicted on those same projects

BCG’s 2024 study of 1,000+ companies across 59 countries confirmed the trend hasn’t improved. More than two thirds of large-scale technology programs still miss time, budget, or scope targets.

The #1 Cause: Requirements

Across 30 years of failure analysis, scoping and requirements problems have consistently topped every list of project killers.

Key findings:

• 50% of digital transformation failures stem from requirements issues (Info-Tech)
• 54% higher success rate when requirements are grounded in real-world problems (Engprax, 2024)
• 75% of executives felt projects were "doomed from the start” (Geneca)
• 80% of avoidable rework comes from 20% of the defects

When three quarters of the people on a project believe it’s doomed before it begins, the issue runs deeper than skill or motivation. Nobody established a shared understanding of what was being built.

Errors Get More Expensive As Development Progresses 

Barry Boehm’s research established the principle that defect costs escalate exponentially the later they’re discovered:

• 1x to fix a requirements error during requirements
• 5-16x to fix it during design or coding
• 50-100x+ to fix it in testing or production

NASA’s Johnson Space Center validated this across real aerospace projects, finding a median 50.5x cost multiplier from requirements to testing.

Caveat: Modern agile practices have flattened this curve for small projects. Boehm himself revised his estimate to ~5x for non-critical systems. But for large, complex, or regulated projects (the kind scientific organizations build), the exponential escalation holds.

Scope Creep

• 52% of all projects experience scope creep, up from 43% five years prior (PMI)
• 70%+ of IT projects specifically (Standish Group)
• 27% average cost overrun on projects with scope creep

Yet only 44% of organizations use formal change control. Projects without it are 35% more likely to exceed costs or miss deadlines.

In scientific enterprise, scope creep is especially dangerous because it feels justified. Research goals evolve, instruments change, and regulations shift. 

Billion-Dollar Examples

• Canada’s Phoenix Pay System - Budgeted at CA$309M. The actual cost: CA$5.1B and counting. A 16x overrun. Project leaders cut scoping, testing, and staffing to fit budget constraints.
• UK NHS National Programme for IT - Spent £10-12.7B. Contracts awarded before scope was defined. Delivered 2% of promised benefits.
• Healthcare.gov - Ballooned from $93.7M to $1.7B+. Key requirements unknown at contract signing. On launch day only 6 individuals out of 4 million users could register.
• Lidl’s SAP Implementation - Abandoned after 7 years and €500M. The root cause? A fundamental scoping decision at the outset forced SAP to fit existing processes rather than adapting.

Every one of these failures started the same way. Insufficient investment in understanding what was being built.

The ROI of Getting Scoping Right

The return on requirements investment is among the highest in software development:

• 92% cost reduction on very large projects with thorough upfront scoping (Boehm et al., 161 projects)
• 8:1 ROI equivalent
• $100 returned for every $1 invested in prototyping and UX research (Forrester)

Rework consumes 30-50% of development effort, and 80% of rework traces to requirements errors. Fix the source and savings multiply across every phase.

NASA data showed projects spending less than 5% of budget on requirements suffered 80-200% cost overruns, while those allocating 8-14% saw dramatically better results.

Key Takeaways

Biotech, pharma, and scientific R&D organizations build high-stakes software; clinical trial platforms, data pipelines, regulatory systems, AI/ML tools. The cost of getting requirements wrong cascades through every phase.

My recommended approach:

1. Allocate 15-28% of project resources to requirements and scoping.

Most organizations spend less than 10% on the phase that causes 70% of failures.

2. Implement formal change control.

This alone reduces failure likelihood by 35%. Scope changes are inevitable in scientific software, uncontrolled scope changes are a choice.

3. Use prototyping to validate requirements before full development.

A working prototype surfaces misunderstandings in minutes that would otherwise cost months and hundreds of thousands of dollars in production.

4. Get technical leadership involved from day one.

BCG’s data: 154% higher success rates when technology leaders participate from the beginning of strategy development.

5. Use AI to accelerate scoping, not skip it.

Generative AI is most valuable in the requirements phase: rapid prototyping, live stakeholder sessions, requirement validation. Use the speed where it matters most.

Only 31% of software projects succeed. AI-accelerated development won’t change that number if organizations keep underinvesting in the critical phase that determines success everywhere else.

Planning a software initiative and confused where to start? Reach out to us on LinkedIn.

Alan Barber, CEO

Accendero Software

If this newsletter was helpful to you, please consider sharing with others.

References

The data cited in this newsletter is drawn from the following industry studies and publications:

BCG. (2024). Build for the Future: 1,000+ companies across 59 countries on large-scale technology program delivery.

Boehm, B. (1981). Software Engineering Economics. Prentice-Hall. [Original cost-of-change research at TRW and IBM.]

Boehm, B., Valerdi, R., & Honour, E. (2008). “The ROI of Systems Engineering: Some Quantitative Results for Software-Intensive Systems.” Systems Engineering, 11(3), 221–234. [Analysis of 161 projects in the COCOMO II database.]

Bossavit, L. (2012). The Leprechauns of Software Engineering. [Investigation of the provenance of the IBM Systems Sciences Institute study and the 100x figure.]

Capers Jones. Various publications on software quality metrics and defect removal efficiency across thousands of projects.

CISQ – Consortium for Information & Software Quality. (2022). The Cost of Poor Software Quality in the US: A 2022 Report. [$2.41 trillion total cost; $260 billion in failed projects.]

Engprax / J.L. Partners. (2024). Survey of 600 software engineers on Agile project outcomes and requirements grounding.

Forrester Research. ROI estimates for prototyping and UX research ($100 returned per $1 invested).

Geneca. (2011). “Doomed from the Start? Why a Majority of Business and IT Teams Anticipate Their Software Development Projects Will Fail.”

Info-Tech Research Group. Research on digital transformation failure rates and requirements-related causes.

McKinsey & Company / University of Oxford. (2012). “Delivering Large-Scale IT Projects On Time, On Budget, and On Value.” [Study of 5,400 IT projects totaling $56.5 billion.]

NASA Johnson Space Center. (2004). “Error Cost Escalation Through the Project Life Cycle.” NASA/TP-2004-212784. [Three independent methods validating Boehm’s cost-of-change curve on real aerospace data.]

NIST – National Institute of Standards and Technology. (2002). “The Economic Impacts of Inadequate Infrastructure for Software Testing.” [U.S. software bugs cost $59.5 billion annually; $22.2 billion recoverable.]

PMI – Project Management Institute. (2018). Pulse of the Profession. [9.9% of every dollar wasted; 52% scope creep rate; requirements as top-3 failure cause.]

Standish Group. (1994, 2020). CHAOS Reports. [31% project success rate (2020); incomplete requirements as leading cause of cancellation (1994).]

Wellingtone. (2025). The State of Project Management. [41% cite scope creep as top failure reason; 44% use formal change control.]

Case study sources:

Canada Phoenix Pay System: Auditor General of Canada reports (2017–2024); total cost exceeding CA$5.1 billion.

UK NHS NPfIT: House of Commons Public Accounts Committee, 18th Report (2011–12); £10–12.7 billion spent.

Healthcare.gov: U.S. Government Accountability Office (GAO) reports; CMS contract documentation. Estimates range from $1.7B to $2B+.

Lidl SAP: Trade press reporting (2018); approximately €500 million over seven years.