What Counts as R&D — and How to Prove It

Whether you’re claiming an R&D tax credit in Canada, the UK, the US, France, or Australia, the rules rhyme. The headline rates and forms differ, but the definition of R&D is remarkably consistent — and so is the thing that makes or breaks a claim: evidence.

The common definition

Across regimes, qualifying R&D is work undertaken to resolve scientific or technological uncertainty — a problem whose solution isn’t readily deducible by a competent professional in the field. The work is experimental: you form a hypothesis, test it, and iterate, because the outcome genuinely isn’t known in advance.

That framing rules things in and out fairly predictably:

• Usually qualifies: developing a new material, process, or device where the technical feasibility is uncertain; overcoming a performance limit no known method achieves; integrating technologies in a way that isn’t routine.
• Usually doesn’t: routine engineering, styling, or configuration; market research; work where the only uncertainty is commercial, not technical; applying known methods to a known outcome.

A useful test for the deep-tech case: ask your engineers whether the technology had been proven in practice before you started — not whether the product would sell. The uncertainty that counts is *technical feasibility*, not market success.

What’s typically eligible to claim

Once an activity qualifies, the costs that usually count include:

• Staff time spent on the qualifying work — generally the largest component.
• Contractors and subcontractors doing R&D on your behalf (often at a reduced rate).
• Materials and consumables used up in the research.
• Cloud and compute costs, in a growing number of regimes.

The exact boundaries vary by country, which is why the same project can yield a different claim in different jurisdictions — but the *shape* of eligible cost is consistent.

The part everyone underestimates: evidence

Here’s what separates a clean claim from a risky one. The credit isn’t earned by doing R&D — it’s earned by being able to show you did R&D, with records made as the work happened:

• What technical uncertainty each piece of work was resolving.
• Who worked on it, and for how long.
• When the experimental phase started and ended.
• How the uncertainty was ultimately resolved — or wasn’t.

Tax authorities have tightened scrutiny in nearly every jurisdiction. A claim backed by contemporaneous evidence survives an audit; one reconstructed from memory at year-end — estimated timesheets, after-the-fact narratives — is where claims get reduced, denied, or clawed back.

Why contemporaneous beats reconstructed

The problem with reconstructing evidence is that the most important details are exactly the ones memory loses: which two weeks the team spent stuck on a feasibility problem versus shipping routine features, who actually worked the experimental path, when the uncertainty resolved. Those distinctions decide what’s claimable — and they’re invisible by the time you’re filing.

This evidence usually doesn’t live in your accounting system. It lives in your delivery record — the tickets, branches, and activity of the people doing the work — captured automatically as it happens, not transcribed later. The companies that claim cleanly, in any country, are the ones whose proof was a byproduct of doing the work, not a year-end archaeology project.

That’s the universal requirement behind every program in the R&D tax credits by country guide: the rules differ, the need to prove it doesn’t. For a practical, step-by-step way to build that proof, see how to document R&D for a claim.