Here’s the question I get asked more than any other: “We improve things all the time — does that count as R&D?” The answer depends entirely on whether your activities pass the IRS four-part test under IRC §41(d). Every dollar of R&D tax credit you claim must trace back to activities that satisfy all four parts of this test, applied separately to each business component.
The test isn’t complicated once you understand it. But it trips up businesses in two ways: companies that clearly qualify assume they don’t (because they picture white lab coats), and companies that don’t qualify file claims anyway (because a promoter told them “everyone qualifies”). Both mistakes are expensive.
After 25 years of helping businesses navigate this test across dozens of industries, here’s how I explain it — with real examples that make each part concrete.
Key Takeaways
- All four parts must be met for each business component — passing three out of four means zero credit for that activity
- The “substantially all” threshold for the process of experimentation test is 80% or more of research activities (Treas. Reg. §1.41-4(a)(6))
- Failed projects can absolutely qualify — the IRS cares about how you worked, not whether you succeeded
- You don’t need a lab, a PhD, or a formal R&D department — the IRS evaluates what employees actually do, not their titles (IRS ATG)
What Is the Four-Part Test?
The four-part test is the IRS’s framework under Section 41(d)(1) for determining whether a business activity constitutes “qualified research” eligible for the R&D tax credit. To qualify, your research must satisfy all four tests — and they must be applied separately to each business component, which the IRS defines as any product, process, computer software, technique, formula, or invention (IRS Form 6765 Instructions, 2025).
If your activities don’t pass at the level of the entire product, a “shrinking back” rule lets you apply the test to the most significant subset of elements — and continue narrowing until you find the level at which the tests are met. This means even routine products can contain qualifying research components.
Part 1: The Section 174 Test — Is This Really R&D?
The first test is the broadest gatekeeper. Your research expenditures must qualify as research and experimental costs under Section 174 of the Internal Revenue Code. This means the expenses must be incurred in connection with your trade or business and represent R&D costs in the experimental or laboratory sense (IRS Audit Techniques Guide).
What does “experimental or laboratory sense” actually mean? It means the activity is aimed at discovering information that would eliminate uncertainty about developing or improving a product or process. If the information is already known — if you’re just following established procedures — it doesn’t qualify.
What passes Part 1
Developing a new manufacturing process to reduce defect rates. Building custom software to automate a workflow that no off-the-shelf solution handles. Designing a new structural system for a building type you haven’t worked with before. Formulating a new food product with specific nutritional targets.
What fails Part 1
Quality control testing of finished products. Management studies and efficiency surveys. Market research. Routine data collection. Exploration expenditures for minerals, oil, or gas. These are explicitly excluded by Treas. Reg. §1.174-2(a)(3) and Section 41(d)(4).
Part 2: The Technological in Nature Test — Are You Using Hard Science?
The research must fundamentally rely on principles of the physical or biological sciences, engineering, or computer science. This test eliminates activities based on style, taste, cosmetic, or seasonal design factors — regardless of how creative or innovative they might be.
Here’s where businesses get confused: “technological in nature” doesn’t mean “high-tech.” A construction company engineering a solution to unstable soil conditions is applying geotechnical engineering principles. A food manufacturer testing formulations to achieve a specific shelf life is applying food science. A machine shop developing custom tooling to hold tolerance on a new part geometry is applying mechanical engineering.
A distinction that matters: The IRS has confirmed that a taxpayer may rely on existing technologies and existing principles of science or engineering to satisfy this test. You don’t need to advance the state of the art. You need to apply scientific or engineering principles to solve your specific technical problem. That’s a much lower bar than most businesses realize.
Part 3: The Elimination of Uncertainty Test — Was the Outcome Unknown?
This is the test that separates genuine R&D from routine work. When your team began the research, there must have been uncertainty about at least one of three things: the capability of developing or improving the product, the method of achieving that result, or the appropriate design of the product or process.
The uncertainty must exist at the outset. If your engineers already knew exactly how to achieve the result before starting — if it was just a matter of executing a known solution — it doesn’t qualify. But if they had to figure something out, test an approach they weren’t sure would work, or evaluate whether a design would meet performance requirements — that’s uncertainty.
Critical point: The project doesn’t need to succeed. If your team conducted genuine research to resolve uncertainty and the project failed, those activities still qualify. The IRS explicitly states that the taxpayer isn’t required to succeed in developing a new or improved business component for it to be qualifying research.
Part 4: The Process of Experimentation Test — Did You Systematically Evaluate Alternatives?
The final test requires that substantially all of your research activities constitute elements of a process of experimentation. Under Treas. Reg. §1.41-4(a)(6), “substantially all” means 80% or more of the research activities, measured on a cost or other consistently applied reasonable basis.
A process of experimentation is one designed to evaluate one or more alternatives to achieve a result where the capability, method, or design is uncertain. This can include modeling, simulation, systematic trial and error, prototyping, testing hypotheses, or other methods of evaluating alternatives.
What counts as experimentation
Systematic Trial and Error
Testing multiple approaches, materials, or configurations to determine which achieves the desired performance. A food company testing 15 formulations to hit a target texture and shelf life.
Modeling and Simulation
Using computational tools to evaluate designs before physical testing. An engineering firm running finite element analysis on a structural design to evaluate load-bearing capacity.
Prototyping and Testing
Building and testing physical or functional prototypes. A manufacturer creating three different tooling designs and running test parts on each to compare dimensional accuracy.
Real-World Examples: How the Four-Part Test Applies by Industry
The four-part test works the same way regardless of industry. But it helps to see how it applies to specific business activities. Here are examples from the industries we work with most frequently at Tax Formulations.
From our practice: The industry where we most frequently find unclaimed credits is construction. Most contractors assume their work is “routine” because they build things every day. But the engineering challenges on individual projects — unusual soil conditions, complex structural designs, value engineering exercises, new material applications — often contain substantial qualifying activities. If your engineers are solving problems they haven’t solved before, it’s worth a conversation.
What the IRS Explicitly Excludes
Section 41(d)(4) provides a clear list of activities that never qualify, regardless of how innovative they might seem. Knowing these exclusions is just as important as knowing the four-part test — because claiming excluded activities is the fastest way to get a credit disallowed.
Activities that never qualify:
| ✕ | Research after commercial production — once a product enters production, post-production testing and quality control don’t qualify |
| ✕ | Adaptation to customer requirements — modifying an existing product for a specific customer’s needs without technical uncertainty |
| ✕ | Duplication — reproducing an existing product from physical examination, blueprints, or publicly available specifications |
| ✕ | Surveys, studies, and market research — efficiency surveys, management studies, and routine data collection |
| ✕ | Foreign research — any research conducted outside the United States or its territories |
| ✕ | Social sciences, arts, and humanities — research that doesn’t fundamentally rely on hard science or engineering principles |
An important nuance on “after commercial production”: This exclusion targets routine post-launch production activities, not ongoing innovation. If you’re developing a next-generation version of a product that’s already in production, the new development work can qualify — as long as the new work itself involves technical uncertainty and experimentation. Improving an existing product is explicitly allowed; the exclusion only bars research on the current version after it enters commercial production.
How to Document the Four-Part Test for Each Project
Documentation is where R&D credit claims live or die. Recent court cases — including Harper (9th Cir. 2021), Premier Tech (D. Utah 2022), and Little Sandy Coal (T.C.) — have all reinforced that vague documentation sinks claims regardless of whether activities actually qualified. And with Section G of Form 6765 becoming mandatory for 2026, business component-level documentation isn’t optional anymore.
For each qualifying project, you should be able to articulate four things clearly:
Document #1
The Business Component
What product, process, software, or technique were you trying to develop or improve? Be specific — “improve manufacturing” is too vague. “Develop a new friction stir welding process for 6061-T6 aluminum to reduce cycle time below 30 seconds per joint” is defensible.
Document #2
The Technical Uncertainty
What didn’t you know when you started? What capability, method, or design question needed to be resolved? “We weren’t sure if the tool could hold ±0.002″ tolerance at the required feed rate” is a documentable uncertainty.
Document #3
The Process of Experimentation
What alternatives did you evaluate? What tests did you run? How did you iterate? “We tested three different tool geometries, measured dimensional output on 50 sample parts per geometry, and selected based on Cpk analysis” documents experimentation.
Document #4
Who Did the Work
Name or identify by title the individuals who performed the research, and describe the information each person sought to discover. This is now required for refund claims and will be mandatory under Form 6765 Section G for 2026.
Start doing this now: Even if you’re not ready to file an R&D credit claim today, begin documenting technical challenges on current projects. A simple log — project name, what uncertainty you faced, what alternatives you tested, who was involved — creates the foundation for a future study. The hardest part of an R&D credit engagement isn’t the tax work; it’s reconstructing technical details from memory months or years later.
Not Sure If Your Activities Qualify?
We’ll walk through your projects, apply the four-part test to your specific activities, and give you a straight answer. No cost, no obligation.
Frequently Asked Questions
Does a project need to succeed to qualify?
No. The IRS explicitly states that a taxpayer isn’t required to succeed in developing a new or improved business component. What matters is that you conducted a genuine process of experimentation to resolve technical uncertainty — the outcome is irrelevant to qualification.
Do I need a formal R&D department to claim the credit?
No. The IRS Audit Techniques Guide states that eligibility is based on what employees actually do, not on job titles, org charts, or department names. Engineers, developers, machinists, and production staff frequently perform qualifying research as part of their regular duties.
Can routine product improvements qualify?
It depends. If the improvement involves genuine technical uncertainty and a process of experimentation, it can qualify — even if the underlying product already exists. But purely cosmetic changes, style modifications, or adaptations with no technical challenge don’t meet the test.
What’s the “shrinking back” rule?
If your research doesn’t pass the four-part test at the level of the entire product, the IRS allows you to apply the test to progressively smaller subsets of the product until you find the level at which it’s met. This means a product that doesn’t qualify overall may still contain qualifying research components.
How does the four-part test relate to Form 6765 Section G?
Section G, mandatory for tax years beginning after 2025, requires you to report qualifying activities at the business component level — essentially documenting how each project meets the four-part test. Companies that build this documentation into their process now will be fully prepared when mandatory reporting begins. Read our complete R&D tax credit guide for more on Section G requirements.
What Should You Do Next?
The four-part test is the foundation of every R&D credit claim. If you’ve read this far and recognized your own work in these examples — technical problems you’ve solved, alternatives you’ve tested, designs you’ve iterated — there’s a real possibility you qualify for a credit you’ve never claimed.
At Tax Formulations, we’ve applied this test across every industry we serve, from aerospace and manufacturing to software and food science. If you want a straightforward assessment of your activities, schedule a free consultation and we’ll walk through it together.
Martin Gamez
Founder, Tax Formulations
Martin is a tax credit specialist with over 25 years of experience in federal and state R&D tax credits, cost segregation, and business tax incentives. His background includes tenure at Big Four and Top 10 accounting firms, with clients spanning technology, manufacturing, aerospace, engineering, and life sciences. Read full bio →