Why R&D Credit Claims Get Reduced Even When the Work Clearly Qualifies

It happens more often than people expect: a company is doing legitimate qualifying research. The engineers are working on genuinely novel problems. The activities clearly satisfy the four-part test. Then an IRS examination opens and the credit gets reduced significantly, or denied outright.

Not because the work did not qualify. Because the documentation did not prove it.

This is not a theoretical risk. It is a pattern that shows up regularly in R&D credit examinations, and with the IRS raising substantiation standards and Form 6765 now requiring project-level disclosure starting in 2026, the frequency of this problem is likely to increase. The following are the most consistent reasons it happens.

Retrospective Documentation Instead of Contemporaneous Records

The IRS does not audit your engineering work. It audits your records. If those records were assembled after the fact, experienced examiners recognize it, and they push back hard.

Contemporaneous documentation looks like engineering notebooks, design review records from during the project, test protocols and results, software commit histories with technical annotations, architecture decision logs, and meeting records where technical uncertainty and alternative approaches were discussed. These are records that would exist regardless of whether the company was claiming an R&D credit.

Retrospective documentation looks like a tax professional’s notes from an interview with an engineer conducted months or years after the project ended. Interview-based documentation is not inherently invalid, but it is not sufficient on its own, and practitioners working active examinations report that examiners are increasingly skeptical of it as a primary record. We have written in more depth about why audit-ready R&D credits fail, and the shift from contemporaneous to retrospective documentation is consistently the most common reason.

Business Components Described Too Broadly

A business component is any product, process, software, technique, formula, or invention on which qualifying research was conducted. Each component needs to be specifically identified, and the qualifying research needs to be described in the context of that specific component.

When companies aggregate too broadly, treating an entire product line as a single business component, or describing the component simply as “software development” or “new product research” without specifying what was actually being developed, they make it straightforward for an examiner to conclude the claim is not adequately substantiated. Vague descriptions create vague claims, and vague claims are easy to challenge.

The new Form 6765 Section G addresses this directly. For each component representing part of your top 80% of QREs, you need a description that is specific enough to identify what was being researched, not just a label. If you cannot write that description now for your current projects, that is a documentation problem worth addressing before 2026 filing becomes mandatory.

Officer Wages Without Supporting Documentation

Officer wages generate more examination attention than most companies anticipate. Officers are often highly compensated individuals who were genuinely involved in R&D activities. A CTO who oversaw the engineering team working on qualifying research, a VP of Product who directed the technical development priorities, a Chief Scientist who led the laboratory program: all of these people may have real, substantial QREs tied to their compensation.

But officer wages are also among the easiest line items to challenge in an examination. If the claimed R&D percentage for an officer is not backed by contemporaneous records, actual calendar documentation, project records showing involvement, meeting notes that confirm participation in technical decisions, that percentage is vulnerable. It will look like an estimate, because without those records, it is one.

Section E of the new Form 6765 now requires the total officer wages included in QREs to be disclosed as a separate line item. An examiner can see that number before they have reviewed a single other document in the return. The answer is not to exclude officer wages. It is to document them with the same rigor you would apply to any other high-value line item.

Round Allocation Percentages

One pattern that consistently draws examiner attention is allocation percentages that look like estimates rather than measurements. Round numbers, 50%, 75%, 100% for entire departments, are a signal. Percentages that have remained unchanged for several years despite obvious shifts in the work, new projects, reorganizations, changes in headcount, are another one.

Allocation methodology needs to produce percentages that can be traced back to something: actual time tracking data, project records, a statistically valid methodology that was applied consistently. If the percentages look like someone made a judgment call at year-end and picked a number that seemed reasonable, an examiner is going to reach the same conclusion. That creates a credibility problem that extends beyond the specific allocation.

Including Activities That Do Not Qualify

A claim that includes non-qualifying activities does not just create a problem for those specific line items. It creates a credibility problem for the rest of the claim. When an examiner finds activities that clearly do not qualify, the entire study gets scrutinized more carefully. The benefit of the doubt that might have applied to ambiguous items disappears.

The categories that most often create this problem: internal-use software developed for general and administrative functions, research where the technical outcome was already known or the method was already established in the industry, and activities excluded under the funded research rules where a third party bears the financial risk of the research. Getting these qualification determinations right requires genuine technical understanding of the work and legal knowledge of where the lines are. It is not something that should be left to a checklist.

What a Defensible Claim Looks Like

The companies that consistently get through R&D credit examinations without material reductions share several characteristics. Their documentation was created during the project, not assembled for the audit. Their business component descriptions are specific and technically grounded. Their allocation percentages are supported by time tracking or other contemporaneous data. And the qualification determinations were made by people who understood both the technology being developed and the legal framework for what qualifies.

That combination is not complicated to achieve, but it requires building the right habits into the R&D credit process from the start. It is significantly harder to retrofit after the fact, particularly when an examination is already open.

Want to talk through whether your credit study holds up? Reach out. We would love to help you get ahead of it.