If your only evidence of technical uncertainty was written by a consultant years after the project ended, will it hold up if the IRS pushes back? The litigation involved roughly half a million dollars in disputed R&D tax credits, making the case a useful reminder that documentation can be just as important as the technical work itself. Although the case was decided under the law applicable at the time, its discussion of documentation and technical uncertainty continues to offer practical lessons for businesses claiming the R&D tax credit today.
What Happened in Davenport v. Commissioner?
Morris and David Davenport each owned 50 percent of Burly Corporation, which manufactures metal roofing and buildings through its subsidiary, Mueller Supply Company. Between 2000 and 2003, Mueller rolled out a new business system, called the OneWorld Project, built around two purchased platforms:
- J.D. Edwards OneWorld, a commercially available enterprise resource planning suite with modules for finance, manufacturing, distribution, and more
- IBM’s ERP Bridge, data collection software IBM had already built as a base template for its customers, then adapted for Mueller
In 2006, the Davenports hired a tax credit consulting firm to study the project and see if it qualified for the research credit. Based on that study, they filed amended returns claiming credits for 2002 and 2003. The stakes were significant:
- The IRS had already refunded roughly $292,000 to the Davenports for 2003, then argued the refund was issued in error and sued to get it back.
- The IRS had disallowed a separate $196,898 claim for 2002, and the Davenports sued for that refund.
Both cases were combined into one lawsuit. The government moved for summary judgment on both claims, and the court agreed with the government on every point that mattered.
Why the Court Rejected the R&D Tax Credit Claim
The court didn’t need to work through every argument the government raised. One issue, the process of experimentation test, was enough to decide the whole case.
- Testing happened after the system was already built, not before: IBM’s own proposal documents described its ERP Bridge as “the only solution” that could meet Mueller’s needs, and rated the project’s degree of difficulty as close to a standard implementation. Mueller employees later ran what the record calls script testing, essentially placing simulated orders to confirm the system worked and to train staff on how to use it. That’s meaningful, useful work. It just isn’t experimentation in the legal sense. The court emphasized that a qualifying process of experimentation begins with genuine technical uncertainty about achieving the intended result.
- The strongest evidence of “uncertainty” was written well after the fact: The only document in the record that specifically mentioned technical uncertainty was a statement prepared in 2007, roughly four to five years after the work occurred, by the same consulting firm that had structured the amended returns in the first place. It said the company had encountered “numerous technical uncertainties.” The court didn’t find that persuasive. A retrospective description written to support a tax filing isn’t the same as evidence showing real doubt existed when the project actually began.
- The Davenports didn’t back up their own claims with evidence: A few examples from the record:
- Their main citation to deposition testimony was a single incomplete quote, and the very next page of their own exhibit was missing.
- They argued the court should apply a narrower legal standard for “internal use software,” but never actually offered evidence tied to that standard.
- The law includes a fallback option, sometimes called the shrinking-back rule, that lets a taxpayer claim credit for a smaller qualifying piece of a larger project even if the whole project doesn’t qualify. The Davenports never provided cost information that would have let the court apply it, so that option wasn’t available to them either.
Meanwhile, the government backed its position with more than 500 pages of evidence, including vendor proposals, deposition testimony, and the consulting firm’s own study materials.
What This Case Teaches Businesses Today
- Document uncertainty while it’s happening, not years later: A study commissioned specifically to support an amended return, written long after the work is done, carries far less weight than records created during the project itself.
- Watch what your vendors say about their own work: IBM’s proposal calling its solution “the only” option and rating the implementation as low difficulty became evidence against the Davenports. If a vendor’s own sales language describes a project as routine, that language can undercut a later claim that the same work involved real technical uncertainty.
- Configuration and customization are treated differently, and testing timing matters: Confirming that a purchased system works as intended, after it’s been set up, generally isn’t experimentation. If your team is doing that kind of validation testing, it likely needs to be separated out from any activity that happened earlier and did involve real uncertainty.
- Keep the records that let you fall back to a smaller claim: Even a partially qualifying project can sometimes support a credit for its strongest components, but only if you can show which costs belong to which piece of the work.
Turning Documentation Into a Stronger R&D Tax Credit Claim
Each of the issues highlighted by the court points back to the same challenge: documenting technical uncertainty and the work performed while the project is underway. The Davenports may well have had a legitimate claim buried somewhere in the OneWorld Project. The record simply didn’t preserve it in a form a court could rely on. TaxDrone.AI, built by National Tax Group (NTG), is built to capture that kind of evidence while a project is happening, not years afterward when memories have faded and the paperwork has to be reconstructed from vendor proposals and old emails. That includes documenting uncertainty at the point it actually exists, tracking costs at the level of individual project components so a shrinking-back analysis is possible if needed, and flagging language in vendor materials or internal communications that could undercut a claim before it becomes a problem in litigation.
Before Your R&D Study Gets Written After the Fact
A credit claim built on records from years ago is a much harder case to win than one built on documentation from the day the work happened.
Get your free R&D tax credit estimate and find out what your current documentation would actually support.
FAQ
- Does hiring a consultant to study past projects for R&D credits work? It can, but the study is only as strong as the underlying records. A study written years after a project, without contemporaneous evidence of technical uncertainty, is a weaker foundation for a claim than documentation created while the work was happening.
- Does testing a system after it’s installed count toward the R&D credit? Generally not on its own. Testing done to confirm a system works as intended, after it has already been built or configured, is treated differently than testing conducted to resolve genuine uncertainty about whether something can be achieved in the first place.
- What is the shrinking-back rule, and why does it matter? It allows a taxpayer to claim credit for a smaller, genuinely qualifying piece of a larger project, even if the project as a whole doesn’t qualify. Using it requires being able to show which costs are tied to which specific piece of work, which depends entirely on how well the project was documented.
