Little Sandy Coal v. Commissioner Shrinking Back Doctrine

The Seventh Circuit's 2023 decision in Little Sandy Coal Co. v. Commissioner settles a question that had been sitting uncomfortably in R&D tax credit practice for years: what happens when a taxpayer can't prove that substantially all of its research activities meet the process-of-experimentation standard, but insists the credit should survive anyway on a smaller slice of the work. It can survive, through what's called the shrinking-back doctrine, but only if the taxpayer's own records are built to make that fallback possible. Little Sandy Coal shows what happens when they aren't.

Who Little Sandy Coal was and what it claimed

Little Sandy Coal Company, a company. is the parent of Corn Island Shipyard, a shipbuilding subsidiary. (CIS), a shipbuilder operating in southern Indiana. For the tax year ending June 30, 2014, the company claimed research tax credits under a federal tax code provision governing research credits tied to the design and construction of eleven vessels, each one a first-in-class build that CIS had never produced before. The claim covered the usual three buckets: employee wages, contract research expenses, and supply costs. Novelty was the whole premise. These were new hulls, new configurations, new engineering problems, and the company treated that novelty as effectively self-proving on the credit question.

Why novelty of the product cannot substitute for proof of activities under the "substantially all" test

Section 41(d)(1)(C) sets what's known as the substantially all test: at least 80% of a taxpayer's research activities, measured by cost or some other consistently applied and reasonable method, have to qualify as elements of a process of experimentation. Both the Tax Court and the Seventh Circuit anchored their analysis on one word in that sentence: activities. The test asks what proportion of what employees actually did counted as experimentation, not what proportion of the finished product's physical structure changed.

CIS leaned hard on a 90% hull redesign argument, the idea that because the hull represented roughly 90% of a vessel and had been substantially reworked, that percentage should carry the substantially all analysis. The Tax Court rejected this outright, holding that Section 1.41-4(a)(6) "requires that the substantially all test be applied in reference to activities, not physical elements of the business components being developed or improved." A redesigned hull says nothing about how the shipyard's engineers, welders, and drafters spent their time. Mass isn't labor. A vessel can be almost entirely rebuilt in steel and still fail the credit if nobody can show what the workforce was doing hour by hour, distinguishing work that tested hypotheses from work that merely executed a known design.

The all-or-nothing documentation strategy that blocked the shrinking-back safety net

Treasury Regulation Section 1.41-4(b)(2) exists precisely for situations like this. When a taxpayer can't clear the substantially all threshold at the top level of a business component, the shrinking-back rule lets the analysis drill down into progressively smaller subsets, testing each one until it finds a subset that qualifies or bottoms out at the smallest element and still fails. It's a rescue mechanism, not a trap. The rule doesn't punish a taxpayer for not qualifying at the broadest level; it exists to make sure a genuinely qualifying piece of work doesn't get thrown out just because it was bundled into a larger project that, taken as a whole, falls short.

CIS never gave the court anything to drill into. The company defined its business component as the entire vessel rather than as discrete, separately documented subcomponents, things like the stern notch, the towing bridle, the outboard side plate, or the vapor barrier system, each of which might have told a very different story about experimentation and uncertainty. The Tax Court was blunt about the consequence: it "could not apply the shrinking-back rule contained in Section 1.41-4(b)(2) to see if the 'substantially all' test is met for a subcomponent of either vessel," because CIS had committed to an all-or-nothing framing and never allocated costs to anything smaller. The shrink-back provision needs granularity to operate. CIS's cost records didn't have it, and a court can't manufacture subcomponent-level evidence that was never generated.

The Seventh Circuit's correction of the Tax Court in the taxpayer's favor

The Seventh Circuit affirmed the ultimate result against Little Sandy Coal, but it didn't let the Tax Court's reasoning stand untouched. Two aspects of the Tax Court's February 11, 2021 opinion got corrected on appeal.

The first concerned direct supervision and direct support activities. The Tax Court had categorically excluded these from the numerator of the substantially all fraction, reasoning that supervision and support don't themselves "constitute elements of a process of experimentation," and so belong only in the denominator. The Seventh Circuit rejected that categorical rule, finding that the term "elements" in Section 41(d)(1)(C) is broad enough to reach research activities that support experimentation even if they aren't the experimentation itself. The corrected approach: direct support and supervision go in both the numerator and denominator when they genuinely constitute elements of a process of experimentation, and in the denominator alone when they're Section 174 research activities that fall short of that standard. For any taxpayer that can document real engagement by supervisory or support staff in the experimentation process, that's a meaningfully larger numerator than the Tax Court's rule would have allowed.

The second correction dealt with pilot model production. The Tax Court had excluded model production wages from the numerator entirely, drawing on a distinction from Section 41(b)(2)(B)'s treatment of qualified services. The Seventh Circuit rejected that approach, holding that prototype and pilot model production, along with the associated expenses, can properly sit inside the substantially all numerator.

Neither correction changed the outcome for Little Sandy Coal. The documentation gap at the center of the case, the absence of subcomponent-level records, remained regardless of how generously the numerator was defined. But both corrections widen the field for taxpayers who do keep the right records, since more categories of cost and labor are now available to count toward the 80% threshold.

Related cases that confirm the pattern and sharpen the warning

Little Sandy Coal isn't an outlier. Another company, a shipbuilder. v. United States involved another shipbuilder, another multi-vessel claim, and the same reliance on the novelty of the vessels as a stand-in for proof of experimentation. The the court allowed qualified research expenses for some of the vessels at issue, but Little Sandy Coal cites Trinity specifically because both courts rejected the novelty-based shortcut as a matter of law, and because Trinity's taxpayer also couldn't produce the substantiation needed to run a shrink-back analysis on the vessels that didn't qualify.

Betz v. Commissioner tells a nearly identical story outside the shipbuilding context. The taxpayers were shareholders in Catalytic Products International (CPI), an S corporation that designs air pollution control systems, and the 2014 research credit of $501,531 passed through to them on Schedules K-1. The Tax Court disallowed it because CPI couldn't establish that its products were pilot models, couldn't show its employee activities met the Section 174 research standard, and had no reasonable basis for estimating qualified wage expenses. "Merely identifying a project difficulty and the eventual design solution, without bridging the gap with evidence as to what investigative activities were performed, does not satisfy petitioners' burden," the court said, cutting to the structural problem shared with Little Sandy Coal." Knowing the problem and knowing the answer isn't the same as documenting the investigation in between.

Another company, a design firm. v. Commissioner adds a different angle on the same theme. Phoenix cleared certain threshold hurdles under Section 41(d) yet still lost, because the court found its engineering activities fell short of the uncertainty and investigative character that a genuine process of experimentation requires. Passing the early tests buys nothing if the substantially all hurdle catches you next.

None of these courts show hostility toward the research credit itself. What they consistently demand is activity-level evidence rather than project-level assertion, and in every one of these cases, the shrinking-back doctrine existed as a legal option that the taxpayer's own recordkeeping made impossible to use.

Steps taxpayers and their advisors must take to keep the shrinking-back doctrine usable

The Seventh Circuit didn't just rule on Little Sandy Coal's facts; it issued a direct warning to everyone else. Other taxpayers seeking the research credit, the court said, "would be well advised to document research activities for subcomponents if they cannot demonstrate a process of experimentation at the business component level." It's the court telling practitioners how to avoid CIS's fate, not dicta buried in a footnote. It's the court telling practitioners how to avoid CIS's fate.

Business components must be defined at the right altitude from day one. Not "the vessel," but the stern notch, the towing bridle, the outboard side plate, each named, each tracked separately. A component defined too broadly forecloses the very fallback the regulations are designed to offer.

From there, cost and time tracking need to happen contemporaneously, at that same subcomponent level. Employee time records tied to specific design tasks, not rough project-wide estimates reconstructed after the fact. Supply costs and contract research expenses allocated to discrete activities rather than lumped against the whole build. The goal is straightforward: if the top-level claim fails, there needs to be something smaller already broken out for a court, or a tax examiner, to test.

Finally, the documentation has to capture process, not outcome. It's not enough to know that a design problem existed and that CIS eventually solved it. What was the technical uncertainty at the outset of each subcomponent's development? What hypotheses got tested, through what method, engineering calculations, software modeling, physical design iterations? How did the results of one round feed into the next? CIS's design spiral, the iterative loop of revision and testing that shipbuilding of this kind genuinely involves, was almost certainly real. The experimentation happening was never in question. Nothing in the record showed it happening at the level of granularity the law requires to prove it.