What HMRC actually expects
The compliance guidelines are refreshingly concrete. Keep sufficient records to identify qualifying projects and to identify qualifying costs, ideally from the beginning of the project. Be able to show a project existed: planning material, project charts, drawings, designs, test results, photos of prototypes, minutes, emails. For smaller projects, the records you normally create while running the business are often enough, this is an evidence standard, not a bureaucracy standard.
Two warnings in the same guidance deserve quoting in substance. Without records you may find it harder to show a project existed and that expenditure qualifies. And if HMRC checks the claim, the process is likely to be quicker and simpler with good written records. That pair is the entire business case.
The four files a claim stands on
- The project file. Evidence the project existed and what it wrestled with: plans, designs, test results, failure notes, minutes, the email where the approach changed. This is what makes the technical narrative checkable rather than believable.
- The professional’s opinion. A written statement from the competent professional: the state of the field, the advance sought and why it counts as an advance, with reasoning, kept alongside your own record of the uncertainties faced and why they were not readily resolvable. HMRC recommends it explicitly, and it is your insurance against the professional leaving before the enquiry arrives.
- The time evidence. Whatever your business genuinely produces, timesheets, tickets, sprint records, rotas, calendars, tied to people and project phases, so staff apportionments have an anchor.
- The money trail. Payroll, invoices, contracts for subcontractors and EPWs, grant paperwork, reconciling cleanly from claim categories back to the accounts.
Retrofitting honestly
Real companies arrive at claim time with imperfect records, and the guidance’s “harder, not fatal” framing sets the honest path. Reconstruct from what exists: artefacts, repositories, calendars, memories cross-checked against documents. Present it as reconstruction, and size the claim to what the evidence will carry. What corrodes trust is reconstruction dressed as contemporaneous record, or round numbers defended as data. If the gap is serious, claim the defensible core this year and fix capture going forward; the delta on next year’s claim will outrun anything gained by stretching this one.
Frequently asked questions
Is there a legally required list of R&D records?
No specific list, and HMRC's guidelines acknowledge that for smaller projects the records you normally create in the business are often enough. The obligation that bites is practical: in a check, HMRC may ask for anything reasonably needed to evidence the claim, and you either have it or you do not.
What single document matters most?
The competent professional's written opinion with supporting reasoning. HMRC recommends keeping exactly that, and warns you may struggle to evidence the claim if the professional is unavailable later. People leave; the file stays.
We have almost no records for last year's work. Can we still claim?
Possibly, honestly reconstructed and presented as such: HMRC's guidance notes that without records it is harder, not automatically fatal. But the claim should be sized to what can be evidenced, and the same guidance is the reason to fix capture for the current year today.
Do timesheets have to be formal?
No. Tickets, sprint logs, commit histories, lab books, calendars and project minutes all evidence engagement. What fails is nothing at all, or percentages invented at year end with no contemporaneous anchor.