What is the R&D Tax Credit? A Plain-English Guide

The R&D tax credit is the most misunderstood incentive in the code, and the misunderstanding costs companies real money. Most of the businesses that qualify never claim it, usually because someone told them years ago that it was for laboratories.

It isn't. Here is what it actually is.

A credit, not a deduction

The federal research credit lives in Section 41 of the Internal Revenue Code. It reduces tax owed dollar for dollar, which makes it worth far more than a deduction of the same size — a deduction reduces the income you pay tax on, a credit reduces the tax itself.

For most companies the credit lands somewhere between 6% and 10% of qualified research expenses at the federal level, before any state credit. A company with a million dollars of qualifying engineering wages is usually looking at a five- to six-figure annual credit.

The four-part test

An activity qualifies if it passes all four of these. This is the whole gate — there is no revenue floor, no headcount minimum and no requirement that the work succeeded.

• Permitted purpose — the work aims to create or improve a product, process, software, technique or formulation, in function, performance, reliability or quality.
• Technological in nature — it relies on principles of engineering, computer science, physics, chemistry or biology.
• Elimination of uncertainty — at the outset, you didn't know whether you could achieve the result, or how.
• Process of experimentation — you evaluated alternatives, through modelling, simulation, systematic trial and error, or testing.

Note what the test does not say. It does not say the research has to be new to the world — only new to you. It does not say it has to have worked. A failed project that was genuinely uncertain at the outset is often the cleanest claim in a study.

What actually counts as an expense

Four categories of qualified research expense, in rough order of how much they usually contribute:

• Wages for employees performing, directly supervising or directly supporting qualified research. This is nearly always the largest bucket.
• Supplies consumed in the research — materials used up in prototyping and testing, not capital equipment.
• 65% of contract research paid to a third party, where you retain rights to the results and bear the economic risk.
• Cloud computing costs attributable to development and testing.

Who this is really for

Software companies building a platform. Manufacturers changing a production process. Hardware startups on their fourth prototype. Agencies doing genuine engineering rather than configuration. If a client's engineers spend their week solving problems they did not already know the answer to, there is probably a claim.

Market research, routine data collection, quality control, style changes, and customising an existing product to a customer's specification do not qualify. A good study says so out loud — the credit that survives examination is the one that left the doubtful items out.

Eligibility and the final figure depend on revenue, company age and the nature of the work. Nothing here is a promise of a specific outcome for a specific client.