The four-part test is the same in every industry: a business component, technological uncertainty at the outset, a process of experimentation, and reliance on the hard sciences. What differs is where the qualifying work hides and what the usual mistakes look like. This page walks through the main sectors. The industry guides go deeper on each.
Where qualifying work sits by sector
| Sector | Usually qualifies | Usually does not | Watch for |
|---|---|---|---|
| Software | New architectures, algorithms, performance and scalability work with real uncertainty | Configuration, porting with known methods, UI styling, maintenance | Internal use software rules; post-release support |
| Manufacturing and engineering | New or improved products and processes, prototypes, pilot runs to resolve uncertainty | Routine production, quality control, tooling changes with known outcomes | When commercial production begins; supplies consumed in trials |
| Life sciences | Discovery, preclinical and clinical development, formulation, process development | Post-approval marketing studies, routine testing for release | Funded research in sponsored trials; foreign sites |
| Food and agriculture | New formulations, shelf-life and process development, cultivar trials | Taste and seasonal design, packaging aesthetics, surveys | The style, taste and cosmetic exclusion |
| Construction and engineering services | Novel structural, geotechnical or systems designs with technical uncertainty | Standard designs, code compliance, work the client pays for and owns | Funded research in client contracts |
| Energy and cleantech | Materials, cell, storage and grid technology development | Installation and commissioning of proven systems | Grants and government funding |
Software
Software is the largest category of claims and the most examined. The qualifying work is where the team did not know whether an approach would work: a new data structure to hit a latency target, a model that had to be trained to a required accuracy, a distributed system that had to stay consistent under failure. The non-qualifying work is everything that follows known patterns: configuring purchased software with vendor documentation, porting with established tools, cosmetic front-end changes and routine maintenance.
Two rules bite in software. Software developed primarily for the taxpayer's own internal use must pass additional tests before it qualifies, which the internal use software page covers. And the exclusion for research after commercial production begins applies to each release: a new major version with new uncertainty is a new business component, bug fixes to the shipped one are not. Agile teams should treat the sprint record, the ticket history and the commit log as the evidence of experimentation, with tickets tagged to the business component.
Manufacturing and engineering
Manufacturers under-claim more than any sector because the research happens on the shop floor and nobody calls it research. A new product line is obvious. Less obvious are process developments: a new welding parameter set for a new alloy, a fixture that had to be designed and tested to hold a tolerance, a pilot run to establish whether a line could run a new material at rate. Each is a business component if it addresses technological uncertainty through experimentation. Prototype materials and material destroyed in trials are supplies and count as QREs; the machine tools are not.
The line to watch is commercial production. Research on a process can continue after the product ships if the process itself is still being developed, but once the process is running in production, improvements with known outcomes are routine engineering. Pilot models are a specific case: the cost of a pilot model built to resolve uncertainty during development can be a research or experimental expenditure under section 174 (Treasury Regulation 1.174-2(a)(4)). Whether it is also a QRE for the section 41 credit depends on the supplies rules, which exclude depreciable property.
Life sciences
Pharmaceutical, biotechnology and medical device development is close to the paradigm of qualified research, and the difficulty is rarely eligibility of the science. It is the structure of the funding and the location of the work. Clinical trials run through contract research organisations are contract research at 65% where performed in the United States, and excluded where the sites are abroad. Sponsored research, licensing deals and collaboration agreements must be read for who bears the economic risk and who retains substantial rights; the funded research page explains the two tests. Section 174, which requires research costs to be capitalised and amortised rather than deducted, has had a larger cash effect on loss-making life sciences companies than the credit itself, and the two interact; see the section 174 page.
Other sectors
Food and agriculture. Formulation, shelf-life extension, process development and cultivar trials qualify; flavour preference, seasonal variants and packaging design do not, because research related to style, taste, cosmetic or seasonal design factors is excluded by statute.
Construction and engineering services. Qualifying work exists in novel designs and methods, but much of it is performed for a client who pays for it and owns it. Where the client bears the risk and keeps the rights, the work is funded research for the contractor and excluded; the contract decides.
Energy and cleantech. Development qualifies in the ordinary way. Grants and government contracts are funded research to the extent the funder bears the risk, and some federal energy research payments to universities and laboratories earn the credit at 100% for the payer rather than 65%.
Financial services and other service businesses. Claims rest almost entirely on software, and almost entirely on internal use software, so the additional tests apply to most of them.
Illustrative example
Three hypothetical companies, one project each. Tallis Software rebuilds its search index to cut query latency from two seconds to under 200 milliseconds; the team evaluated four indexing strategies against a benchmark. Qualified research, with the ticket history as the evidence. Dunmore Castings develops a casting process for a new aluminium alloy, scrapping 40 trial pours before the porosity target was met. Qualified research, with the scrap as supplies. Pell Nutrition reformulates a cereal bar to taste more like caramel. Not qualified: taste. Pell's parallel project to keep the same bar from going stale for twelve months in a new film is a different matter.
Related pages
Official sources
Educational content only, not tax advice. See the disclaimer.