Four part test: United States

Research qualifies only if it passes all four parts of the test in section 41(d). The costs must be eligible to be treated as research or experimental expenditure under section 174. The work must be undertaken to discover information that is technological in nature. That information must be intended to be useful in developing a new or improved business component. And substantially all of the activities must be elements of a process of experimentation. Everything else in the US scheme, from the costs that count to the records that are kept, is organised around this definition.

What it is

The statute defines qualified research by four conjunctive requirements. The unit of analysis is the business component, meaning a product, process, technique, formula, invention or piece of software, not the company as a whole. The test is conjunctive: failing any single part means the activity does not qualify, however impressive it is otherwise.

The regulations add that the section 174 test is satisfied where the information available to the taxpayer does not establish the capability or method for developing or improving the business component, or its appropriate design. That is the uncertainty element the published guide calls "elimination of uncertainty". It is judged at the outset, before the work starts, not after it succeeds.

Who it applies to

Every claimant and every business component. The test does not vary by industry, size or sector, and it applies to each component separately, which is what Section G of Form 6765 now reflects by asking for the information sought to be discovered for each one. Software developed primarily for the taxpayer's own internal use must pass the four-part test and then additional tests; see the internal use software page.

How it works

Permitted purpose. The work must be intended to create or improve the functionality, performance, reliability or quality of a business component. This is the easiest part to pass and the most commonly misread. It does not require novelty to the world; improving your own product counts. But it does require a development objective. Routine production, cosmetic changes and market research sit outside it.

Technological in nature. The work must fundamentally rely on principles of the physical or biological sciences, engineering or computer science. The question to ask is not "did we use technology?", because every business does, but "does resolving our problem depend on hard science?" Configuring purchased software with vendor documentation generally does not. Building or materially re-architecting software generally does.

Elimination of uncertainty. At the outset, you must have been uncertain about capability, method or appropriate design. The key word is outset: uncertainty is judged before the work starts, not after it succeeds. This is where contemporaneous records earn their keep. A design document that states what was unknown in March is worth more than a memo reconstructing it in December.

Process of experimentation. Substantially all of the activity must constitute a process of experimentation: evaluating alternatives through modelling, simulation, systematic trial and error, or other scientific means. The regulations put a number on "substantially all": 80% or more of the research activities, measured on a cost or other consistently applied reasonable basis. They also describe the process: identifying the uncertainty, identifying one or more alternatives intended to eliminate it, and identifying and conducting a process of evaluating the alternatives.

Shrinking back. If the whole business component fails, the test is applied to the most significant subset of its elements, and so on down, until a subset passes or none does. A product that fails as a whole because most of it is routine can still contain a qualifying module.

Worked example

Illustrative example. Carrow Aerospace, a hypothetical company, develops a lightweight bracket for a satellite bus. The brief in January records that no known lattice geometry had been shown to hold the required stiffness at the target mass with the chosen alloy. Engineers model six geometries, print and test four, and record the results. Of 2,000 engineering hours on the bracket, 1,700 are modelling, building and testing alternatives; 300 are drafting the production drawings once the design is fixed.

Had only 1,400 hours (70%) been experimentation, the bracket as a whole would fail part four. The shrinking back rule would then be applied to the lattice design on its own, which would pass, so the claim would cover the lattice work rather than the whole bracket.

Common questions

Does the project have to succeed? No. Failure is often the best evidence that the uncertainty was real.

Is uncertainty about cost or timing enough? No. The uncertainty must concern capability, method or appropriate design.

Can a patent prove the test is met? A patent is evidence of a technological advance but does not by itself prove a process of experimentation, and the absence of a patent proves nothing either way.

What about work done by a contractor? The contractor's activities must themselves pass the test, and the taxpayer must bear the risk and retain substantial rights.

Related pages

• Guide: a plain-English guide to the four-part test
• Eligibility and the statutory exclusions
• Internal use software
• Guide: what is technical uncertainty
• Program: Technical Teams Program