The Genesis of Section 41 and the Internal Use Software Dilemma

The federal research credit originated in 1981. The Tax Reform Act of 1986 narrowed the definition of qualified research and introduced the statutory framework commonly called the four-part test. It also generally excluded software developed primarily for internal use, subject to exceptions and regulations. The accompanying legislative history described the additional innovation, economic risk, and commercial availability requirements later considered in Norwest.

The four-part framework addresses the research expenditure requirement, technological information, a new or improved business component, and a process of experimentation for an eligible purpose. Improvements must concern function, performance, reliability, or quality. Under the statute as amended in 2025, Section 41(d)(1)(A) refers to expenditures that may be treated as expenses under Section 174A. Historical descriptions referring to Section 174 should be read in their tax-year context. A research deduction and a research credit have separate eligibility requirements.

Fact-Finding and Judicial Reasoning in Norwest Corp. v. Commissioner

Norwest was a bank holding company whose subsidiaries provided banking and financial services. A study identified 67 software activities for the credit claim; the parties selected eight representative activities conducted between 1986 and 1991. They agreed on how the sample findings would affect the remaining activities. The court therefore did not individually adjudicate every project in the larger population.

Detailed Review of the Sample Software Projects

The court evaluated the eight activities against four statutory requirements and three additional internal use software requirements. The table summarizes historical outcomes; it does not assign automatic eligibility to comparable software today.

The Success of the Strategic Banking System (SBS)

The SBS customer module was the only sample activity the court found satisfied all seven tests. Norwest worked with Electronic Data Systems Corp. (EDS) on an ambitious system involving substantial integration challenges. The court accepted that the qualifying development involved technical uncertainty, experimentation, innovation, and significant economic risk under the standards it applied.

The decision also addressed payments to EDS as contract research expenses. Norwest’s rights to use the research results and its payment obligations mattered to that analysis. A perpetual license helped establish rights in the results; the court also examined whether payment depended on successful research. This is more precise than describing the issue simply as research funded by an unrelated third party. Contract terms and actual performance remain important when assessing a research credit.

The Failure of Routine Software Development

The remaining seven activities failed the court’s historical qualification analysis. Substantial programming effort, business value, improved efficiency, and extensive testing did not by themselves establish qualified research. Nevertheless, the decision should not be paraphrased as a permanent rule that every upgrade, integration, or use of established technology is ineligible. Current law requires examination of the particular uncertainty, experimental activities, eligible purpose, and applicable exclusions.

The Discovery Test Controversy: From Norwest to McFerrin

Norwest interpreted the discovery requirement to demand more than information new only to the taxpayer. Its approach required an expansion or refinement of knowledge in the relevant technical field. That historical position became a significant point of disagreement in research credit administration.

The Historical Discovery Standard

The phrase “new to the world” is shorthand for this controversy, rather than a separate statutory test. Norwest’s interpretation imposed a demanding standard on software development even where work was useful and technically challenging. United Stationers, Inc. v. United States, 163 F.3d 440 (7th Cir. 1998), likewise applied a restrictive view of discovery and experimentation. The correct starting page for that appellate citation is 440; page 444 is a pinpoint reference.

The Regulatory Reversal and the Elimination of Uncertainty

Treasury issued T.D. 8930 in January 2001, then announced reconsideration and issued proposed changes later that year. Describing this sequence as a blanket suspension of all rules within 30 days is misleading. T.D. 9104, published in 2004, removed the requirement that research expand the common knowledge of skilled professionals in a field.

Under the later regulations, the discovery inquiry concerns uncertainty about developing or improving a business component, including capability, method, or appropriate design. Established scientific or engineering principles can support qualifying research. A process of experimentation and the other Section 41 requirements must still be established.

In United States v. McFerrin, 570 F.3d 672 (5th Cir. 2009), the Fifth Circuit vacated and remanded a judgment that used overly restrictive definitions of discovery and experimentation. The decision also permitted estimation where qualifying expenditures were established. It did not abolish substantiation requirements, guarantee credits for all technical problem-solving, or settle every internal use software issue nationwide.

Evolution of the Internal Use Software Standard: T.D. 9786

The final regulations in T.D. 9786, issued in 2016, clarified internal use software, third-party interaction, dual-function software, and the high threshold of innovation. The classification inquiry depends on the taxpayer’s intended use and the facts at the beginning of development.

General and Administrative Functions

Internal use software generally covers software developed for general and administrative functions supporting the taxpayer’s business. The principal categories are below; their examples are illustrative.

Software developed for sale, lease, or licensing to third parties, or to enable qualifying third-party interaction, is generally outside the internal use category. Customer banking portals and e-commerce functions may fit that description, depending on their intended operation. A customer-facing label alone does not establish the classification of every underlying module. Related-party use and software used by service providers to support the taxpayer’s administration require particular care.

Dual-Function Software and the 25% Safe Harbor

Software serving both administrative and third-party functions is generally presumed internal use. A separately identifiable third-party-only subset can escape that presumption. For qualifying remaining dual-function software, the regulations permit a 25% safe harbor if anticipated third-party use is at least 10%, measured reasonably at development’s outset. The ordinary qualified research requirements still apply; the percentages do not make otherwise ineligible spending creditable.

The Modern High Threshold of Innovation Test

• Innovation: Successful development would produce a substantial, economically significant cost reduction, speed improvement, or other measurable improvement.
• Significant economic risk: Substantial resources are committed amid substantial technical uncertainty about recovering them within a reasonable period. Ordinary budget or market risk is insufficient.
• Commercial availability: Suitable software cannot be acquired and used for the intended purpose without modifications satisfying the innovation and economic risk requirements.

These conditions supplement the ordinary credit requirements where the internal use rules apply. Certain research-use, production-process, and integrated hardware-software situations have specified exceptions. A project need not succeed commercially for its research activities to qualify.

The IRS Audit Landscape and High-Risk Software Categories

The IRS’s software experimentation audit guidelines describe activities that may warrant greater scrutiny. They are examination guidance, not a substitute for the statute, regulations, or controlling decisions. A risk category does not decide a particular claim.

Software Activities That Warrant Particular Scrutiny

The following table groups examples discussed in the source study. It is not a list of 17 statutory exclusions or a claim that all work in these categories fails.

Review the underlying activities and technical evidence. Project names, release numbers, job titles, and expenditure totals do not independently prove eligibility or ineligibility.

Legislative Changes and the Return of Immediate Expensing

The Tax Cuts and Jobs Act changed Section 174 for tax years beginning after December 31, 2021. Domestic research expenditures generally required five-year amortization and foreign research expenditures required 15-year amortization, using a midpoint convention. Software development expenditures were expressly included. This expenditure treatment did not mean every software cost qualified for the Section 41 credit.

The Introduction of Section 174A

Public Law 119-21, commonly called the One Big Beautiful Bill Act and enacted July 4, 2025, introduced Section 174A. It generally permits current deductions for domestic research or experimental expenditures in tax years beginning after December 31, 2024. Taxpayers can instead elect capitalization and amortization under the statutory conditions. Foreign research remains subject to Section 174’s 15-year treatment.

The transition acceleration is not reserved for large businesses. Small-business retroactive relief uses the Section 448(c) gross receipts test for the first tax year beginning after 2024, including aggregation rules and the tax-shelter restriction; the 2025 threshold is $31 million or less, not strictly below $31 million. Revenue Procedure 2025-28 generally required the retroactive election by July 6, 2026, or an earlier applicable refund limitation deadline. That general deadline has passed as of September 2026, so the election should not be presented as indefinitely available.

Section 280C coordinates domestic research deductions with the research credit, including an available reduced-credit election. Immediate expensing can improve the timing of deductions, but it does not convert all domestic software spending into qualified research expenditures or create an automatic refund.

Procedural Rigor: Section G and Documentation Standards

Business Component Disclosure

The December 2025 Form 6765 instructions make Section G optional for tax years beginning before 2026 and generally required thereafter. Exceptions include specified qualified small businesses electing the payroll credit and original-return filers meeting both the $1.5 million QRE and $50 million average gross receipts limits, with applicable group rules.

Required filers generally identify components covering at least 80% of QREs, capped at 50 components, and aggregate the remainder. Information includes component and software classifications and expense details. Column 49(f), concerning information sought, currently applies to amended returns. It is incorrect to describe a universal original-return requirement for a technical uncertainty narrative for every component.

The instructions should be checked for the particular filing year and return type. These disclosures do not replace evidence establishing qualifying activities and their costs.

Jurisprudence on Eligibility and Substantiation: Eustace and Betz

Betz v. Commissioner, T.C. Memo. 2023-84, involved an air pollution control systems business. The taxpayers failed to establish pilot-model treatment for production costs and qualifying services for claimed wages. The court also found a lack of substantial research rights for five projects. The decision supports careful examination of experimental purpose, employee activities, and contract rights; its holdings should not be reduced to a generic failure to maintain project logs.

Eustace v. Commissioner, T.C. Memo. 2001-66, affirmed at 312 F.3d 905 (7th Cir. 2002), is an older software eligibility case, not a recent documentation decision. Its discovery and experimentation reasoning belongs to the earlier legal framework discussed above. It should not be used to impose an obsolete industry-wide novelty test on current claims.

McFerrin permits reasonable estimation where a foundation for qualifying expenditures exists. It does not support unsupported percentages. Contemporaneous technical and financial records are valuable, while credible testimony and other evidence may also be relevant. The inquiry is whether the evidence establishes the required activities and a reliable connection to the claimed expenditures.

Strategic Implications for Future R&D Applications

The Integration Threshold
<