The test comes from Section 41(d) and the Treasury regulations under it. You run it on each thing you are building or improving, what the law calls a business component, and all four parts have to be met.
| Part of the test | What it asks | Example that passes |
|---|---|---|
| 1. Permitted purpose | Were you improving a product or process? | Redesigning a part to cut scrap, or testing a ration to raise milk yield. |
| 2. Technological in nature | Did the work rest on hard science? | Using soil chemistry to dial in a fertilizer program. |
| 3. Elimination of uncertainty | Were you unsure it would work at the start? | Not knowing whether new tooling would hold tolerance. |
| 4. Process of experimentation | Did you test alternatives to find out? | Side-by-side test plots, or first-article runs until the part passed. |
The work has to aim at a new or improved business component: better function, performance, reliability, or quality. It does not have to be new to the world. New to you is enough. What fails here is change for its own sake, like a cosmetic restyle that does nothing to how the thing works.
The work has to rely on the principles of a hard science such as engineering, physics, chemistry, biology, or computer science. It does not have to be cutting-edge science, but it cannot rest on taste, marketing instinct, or business judgment alone.
At the start, you did not know whether you could achieve the result, how to achieve it, or which design was the right one. If you already knew the answer going in, there was no uncertainty to eliminate, and that is one of the fastest ways a claim falls apart in an audit.
You worked through the uncertainty by evaluating alternatives. Modeling, simulation, prototyping, systematic trial and error, and testing all count. There is also a useful rule of thumb: a business component generally qualifies when substantially all of the work on it (80 percent or more, usually measured by cost) is this kind of experimentation.
Did your team face a technical problem they were not sure they could solve, and did they have to experiment to solve it? If the answer is yes, you are most of the way to a qualifying activity. And the work still counts when the experiment fails, because the credit rewards the attempt.
No. The credit rewards the attempt to resolve technical uncertainty, not the outcome. A failed prototype or a trial that did not pan out can qualify the same as a success.
No. You apply the test to each business component. Under the substantially-all rule, a component can qualify when about 80 percent of the work on it is experimentation, even if some routine work is mixed in.
No. Improving something that already exists counts, as long as it is new to you and you had to work through real technical uncertainty to do it.
Two things. Either there was no real uncertainty because you already knew the answer, or there is no record of the experimentation you actually did. The first is a design problem; the second is a documentation problem.
This guide is general information, not tax advice. Your situation has its own facts, so talk to a credentialed professional before you act on anything here.