How to Document R&D for a Claim: An Evidence Checklist

Every R&D tax credit regime rewards the same thing: proof that you did qualifying R&D. Not a good story at year-end — a contemporaneous record made while the work happened. This is the checklist for building that record, wherever you claim.

Capture it as the work happens — not at year-end

The cardinal rule. The details that decide a claim — which two weeks were spent on a feasibility problem, who worked the experimental path, when the uncertainty resolved — are exactly the ones memory loses. Reconstructing them at filing produces estimates, and estimates are what auditors discount. Records made in the moment are what they accept.

What to capture, per project

For each piece of qualifying work, you want a record that answers five questions:

1. The uncertainty — what scientific or technological unknown were you trying to resolve? State it as a question a competent professional couldn’t just look up. 2. Why it was uncertain — what made the outcome genuinely unknown (no existing method, conflicting approaches, unproven feasibility)? This is the line between technical and market risk. 3. Who did the work, and how much — the people on the experimental work and the time they spent on it, separated from routine, non-qualifying work. 4. What you did — the hypotheses, experiments, iterations, and what each one showed. Failed attempts count; they’re often the clearest evidence of genuine experimentation. 5. When it started and ended — the dates the experimental phase opened and the uncertainty was resolved (or abandoned).

The evidence checklist

• [ ] A short project record stating the technical uncertainty and why it qualified.
• [ ] Time attributed to qualifying work by person — ideally derived from real activity, not estimated percentages.
• [ ] A timeline marking when experimentation began and when uncertainty resolved.
• [ ] Contemporaneous artifacts — design docs, test results, commit/ticket history, lab notebooks, code reviews — that show the work as it happened.
• [ ] Cost records — staff, contractors, materials/consumables, and (where eligible) cloud/compute, tied to the qualifying project.
• [ ] A clear separation of qualifying R&D from routine development, configuration, and commercial work.
• [ ] Sign-off on the qualifying assessment by someone technical, not only finance.

Where this evidence actually lives

Most of this record isn’t in your accounting system — it’s in your delivery tooling. Tickets show what was worked on and when. Branches, commits, and reviews show who did the experimental work and how it progressed. Test history shows when feasibility was proven. Captured automatically as work happens, that activity *is* the contemporaneous evidence — without anyone keeping a separate diary.

That’s the through-line across every regime in the R&D tax credits by country guide and the broader non-dilutive funding toolkit: the claim is only as strong as the evidence, and the best evidence is a byproduct of the work, not a reconstruction of it.