The Statutory Architecture of the Research and Development Tax Credit

Section 41(d) establishes a four-part test. First, research expenditures must satisfy the research-or-experimental expenditure requirement applicable to the tax year. For the historical years discussed in Bayer, this referred to Section 174; following the 2025 amendments, Section 41(d)(1)(A) refers to Section 174A for applicable domestic expenditures. Second, the work must seek technological information grounded in physical or biological science, engineering, or computer science. Third, that information must be intended to help develop a new or improved business component. Fourth, substantially all research activities must constitute elements of a process of experimentation for a qualifying purpose: function, performance, reliability, or quality. The statutory exclusions also apply. Research-expense deductibility alone does not establish credit eligibility.

Treasury Regulation Section 1.41-4(a)(6) sets an 80% threshold, measured by cost or another consistently applied reasonable basis, such as time. The relevant activities must constitute elements of an evaluative process addressing uncertainty about capability, method, or appropriate design. Modeling, simulation, and systematic trial and error may qualify. A formal scientific-method label is insufficient, and a formal hypothesis or failed experiment is not invariably required. The regulation also imposes conditions on the remaining activities.

The Business Component Test as a Structural Mandate

Section 41(d)(2)(B) defines a business component as a product, process, computer software, technique, formula, or invention held for sale, lease, or license, or used in the taxpayer’s trade or business. The qualification tests apply separately to each business component. Taxpayers therefore need evidence connecting claimed expenses to qualifying activities and identifiable components. Bayer does not mandate a particular accounting platform or a separate contemporaneous time entry for every dollar; reliable allocations must nevertheless support both eligibility and amount.

During the years described in the litigation, Bayer operated across healthcare, material science, and crop science. Bayer estimated that more than 100,000 business components were involved. The February 2012 opinion describes the difficulty of retrieving and organizing records across numerous research sites and changing accounting systems. These historical descriptions should not be read as statements of Bayer’s current corporate structure.

Procedural History and Core Disputes in Bayer Corp. v. United States

Bayer filed the action on March 23, 2009. The disputed research credits arose in 1990–2006, while the complaint sought refunds for 1987–1990, 1995, and 2006. The February 2012 opinion describes complete or partial credit disallowances, including previously agreed credits. Bayer tracked research spending through more than 1,300 cost centers at 49 sites. Those cost centers grouped activities, rather than consistently organizing expenditures by individual business component.

The Denial of Statistical Sampling

On February 6, 2012, the district court denied Bayer’s amended motion for a case-management and protective order based on statistical sampling. Bayer sought to limit discovery and trial proof through a sample of cost centers and extrapolation to the broader claim. The scale of its records was a central practical argument, but the proposed approach would restrict the government’s examination of nonsampled expenses.

The court rejected the particular sampling plan before it because Bayer had not established entitlement to the underlying credits and the proposal would restrict the government’s ability to challenge nonsampled items. Bayer’s recordkeeping and business-component identification obligations remained material. The ruling is not a categorical ban on statistical sampling in research-credit matters. Appropriate sampling may be permitted under applicable IRS procedures or litigation arrangements; acceptance of a filed return does not itself approve the sampling methodology.

The Cohan Rule and its Limitations in the R&D Context

Bayer relied on the Cohan principle in support of its sampling request. Under Cohan v. Commissioner, 39 F.2d 540 (2d Cir. 1930), a court may estimate an amount when the evidence establishes a qualifying expenditure and supplies a reasonable basis for estimation. In the research-credit setting, establishing that qualified research occurred is distinct from quantifying qualified expenses. The Bayer court rejected the use of estimation to bypass that threshold and the identification of business components. Estimation is neither automatic nor a substitute for evidence of eligibility.

The Substantial Variance Doctrine and the September 2012 Ruling

The government sought partial summary judgment under the substantial-variance doctrine, which restricts refund litigation based on grounds materially different from those presented administratively. It argued that Bayer’s business-component substantiation differed from its administrative presentation. The issue concerned the permissible scope of the refund suit, not whether every claimed expense satisfied Section 41.

On September 20, 2012, the court denied the government’s motion. In the circumstances of Bayer’s administrative submissions and examination, supplying additional business-component detail did not create a substantially different claim. This procedural ruling permitted the claim to proceed; it did not establish entitlement to the credits or authorize taxpayers generally to cure every deficient refund claim during litigation.

The Shrink-Back Rule: A Second Chance for Qualification

Treasury Regulation Section 1.41-4(b)(2) provides a shrinking-back rule. The tests are applied first at the level of the discrete business component. If that component fails, the analysis moves to its most significant qualifying subset and continues until a qualifying subset is reached or the smallest subset fails. The rule concerns qualification and does not redefine the business component. It may preserve qualifying research within a broader project, but requires evidence supporting the relevant activities and costs.

The rule should not be described as a separate merits holding in Bayer without an identified supporting ruling. The documented 2012 Bayer decisions discussed in this study concern sampling and substantial variance. Little Sandy Coal provides a clearer illustration of the practical limitation: identifying an interesting subsystem is insufficient if the taxpayer cannot establish the applicable research activities and associated expenses. Claimants should preserve evidence at both component and meaningful subset levels while following the regulation’s order of analysis.

Comparative Jurisprudence: Bayer vs. Little Sandy Coal Co. v. Commissioner

The legal landscape has continued to evolve since the Bayer rulings, most notably with the decision in Little Sandy Coal Co. v. Commissioner, which reached the Seventh Circuit Court of Appeals in 2023. While Bayer focused on the procedural burden of identifying business components, Little Sandy Coal centered on the mathematical rigor of the “substantially all” (80%) test.

The Calculation of the Substantially All Threshold

Little Sandy Coal concerned a shipbuilder’s claims involving 11 vessels, with two representative vessels examined in the litigation. The Tax Court denied the credits. On appeal, the treatment of direct support and direct supervision in the substantially-all calculation was one issue, alongside the taxpayer’s proof of experimentation and allocation of employee activity.

The Seventh Circuit affirmed the judgment in March 2023 while disagreeing with a categorical exclusion of direct support and supervision from experimental activities. Such activities may enter the numerator when they themselves constitute elements of a process of experimentation. They are not automatically experimental merely because an employee supports or supervises a research team. The governing ratio can be stated as follows:

Experimental research activities ÷ total research activities ≥ 80%, measured by cost or another consistently applied reasonable basis. Include qualifying direct support and supervision according to the activities actually performed; do not automatically count all management or support time in the numerator.

The decision clarified the analysis without reducing the 80% threshold or establishing that all logistical and management activities qualify. Taxpayers must substantiate the nature of the activities included on both sides of the calculation.

The Persistence of Proof: Shortcut Estimates vs. Principled Records

Little Sandy Coal failed to provide a reliable basis for distinguishing experimentation from other activities. Its unsupported allocations could not establish the substantially-all threshold or support shrinking back to smaller portions. Contemporaneous project records are valuable, but neither that decision nor Bayer creates a universal requirement for a particular time-tracking system. Credible testimony and reasonable reconstruction must be assessed together with the available evidence; unsupported percentages remain vulnerable.

Evolving Substantiation Standards: Betz and the Adaptation Exclusion

Betz v. Commissioner, T.C. Memo. 2023-84, concerned air-pollution-control systems, claimed pilot-model costs, wage substantiation, and funded research. The taxpayers failed to establish qualifying research, including the relevant uncertainty and experimental purpose of the claimed pilot models. Certain projects also failed the substantial-rights requirement. Although adaptation under Section 41(d)(4)(B) is relevant to customer-specific development, Betz should not be reduced to an adaptation-only decision.

A customer-specific project is not automatically excluded merely because it uses existing technology. Treasury Regulation Section 1.41-4(c)(3) requires a factual inquiry. Conversely, routine adaptation does not become qualified research simply because a product is customized or technically complex. The taxpayer must establish the research tests and address the exclusions separately. Uncertainty about appropriate design can qualify; lack of uncertainty about capability or method alone is not necessarily dispositive.

Funded Research and Contractual Risks

Smith v. Commissioner and System Technologies, Inc. v. Commissioner illustrate funded-research disputes under Section 41(d)(4)(H) and Treasury Regulation Section 1.41-4A(d). The analysis considers payment contingency and substantial rights in the research. If payment is not contingent on research success, amounts received can fund the research; if substantial rights are not retained, the research may be wholly excluded. Retaining rights does not by itself make all customer-financed expenses eligible.

The earlier orders denied IRS summary-judgment motions; they should not be presented as final awards of all claimed credits. In System Technologies, the court considered Indiana-law refund remedies when assessing whether ultimate payment depended on successful performance. Smith involved a separate architectural practice and contractual analysis. Subsequently, Smith, T.C. Memo. 2026-50, issued June 16, 2026, addressed the merits: retained rights differed across projects, and where substantial rights were retained, potential credits were limited to eligible expenses exceeding customer payments. These decisions support reviewing payment terms, enforceable remedies, and intellectual-property rights alongside the technical work.

Implications for Future R&D Tax Credit Applications

Taken together, these cases illustrate recurring substantiation problems for pharmaceutical, manufacturing, engineering, and technology businesses. Their practical lessons concern the connection between activities and costs, the reliability of experimental-activity allocations, and the effect of contracts. Each holding must be read in its factual and procedural context.

Connecting Cost Centers to Business Components

Cost-center accounting may support a research-credit claim if the taxpayer can reliably connect it to qualifying activities and identifiable business components. A separate project ledger is one possible approach, but existing financial records, technical records, and supportable allocations may also be useful. The required infrastructure depends on the business and the evidence available; Bayer does not compel every claimant to purchase a new accounting system.

Strategic Use of the Shrink-Back Rule

A claim focused only on an entire vessel, product, or other large component can be difficult to sustain if the substantially-all test fails. Preserve evidence for significant subsets so the shrinking-back rule can be applied in the required sequence. Do not assume that identifying a novel subsystem establishes qualification or permits an unsupported allocation of the whole project’s cost.

Documenting the Process of Experimentation

Newness and complexity alone do not establish qualified research. Useful evidence explains the uncertainty, alternatives, and evaluation undertaken to develop or improve function, performance, reliability, or quality. Depending on the work, records may describe:

• The uncertainty concerning capability, method, or appropriate design at the outset of the relevant research.
• The alternatives identified and the reasons for evaluating them.
• Modeling, simulation, systematic trial and error, testing, or other evaluative work and its results.
• How the findings informed design decisions, including unsuccessful alternatives where applicable. Actual failure is not a prerequisite to the credit.

The Burden of Contemporaneous Records