How to document R&D activities as you go: a calendar for the financial year
Records made at the time are the difference between a claim that is defended in an afternoon and one that is argued about for a year. Here is a documentation routine a busy team can actually keep.
Every scheme covered here phrases the requirement differently, but each asks for the same thing: records made at the time, not afterwards, that show the technical uncertainty, the work done to resolve it and the costs linked to that work. One scheme asks for a hypothesis, experiments, results and conclusions; another for the advance sought, the uncertainties and how they were overcome; another for the objective, the novelty or technical risk and the work done. Read side by side, they are the same request in different words.
The good news is that a single set of records, kept well, will serve any of them. The bad news is that nobody keeps them by accident. This guide is a routine, organised around the financial year, that produces the evidence as a by-product of doing the work.
What this guide covers
- The four moments in the year when a record should be made, and who makes it.
- How to capture evidence from the tools technical teams already use, instead of writing new documents.
- What "made at the time" means, and what it does not mean.
- How the routine maps onto what each of the seven schemes asks for.
Before the work starts: the project note
When a project is approved, someone technical writes half a page answering four questions:
- What are we trying to build or improve, and for whom?
- What do we not know how to do, and why is that not already answered by what is publicly available or known to a competent professional in the field?
- What approaches might we try?
- Who will work on it?
That note is dated and saved where it cannot be quietly edited later. It is the single most valuable document in a claim file, because it shows that the uncertainty existed at the outset rather than being discovered when the claim was being written. If the person writing it cannot name anything the team does not already know how to do, that is useful too: the project is probably not R&D, and finding out in week one costs nothing.
During the work: capture, do not create
Engineers, developers and scientists already produce evidence. The task is to make sure it survives and can be found:
- Version control and issue trackers. Commit messages, pull requests and tickets that describe what was tried and why it failed are records of the work done to resolve the uncertainty. Ask for one extra sentence of "why" in each.
- Test results. Failed tests are more valuable than passing ones. Keep the failures, with the date and the configuration that was tried.
- Design reviews and meeting notes. A dated note that records "we considered A and B and chose B because A could not achieve X" captures the work done to resolve the uncertainty in one line.
- Time. Whether by timesheet, sprint allocation or a monthly estimate signed by the team lead, record who spent time on which project. Do it monthly at the latest; quarterly estimates are already reconstructions, and annual ones are guesses.
At milestones: the checkpoint
At each release, prototype or phase gate, the technical lead spends fifteen minutes updating the project note: what was resolved, what remains uncertain, what changed in the approach. Three or four such entries across a year form a narrative that no one has to write from memory. They also mark where the R&D ended, which matters because most schemes stop counting once the uncertainty is resolved and routine work takes over.
At year end: assemble, do not invent
With the routine in place, year end is administrative. For each candidate project the file contains the opening note, the checkpoint updates, links to the technical evidence, and the time record. The finance team adds cost data reconciled to the ledger and payroll. The description for the claim is drafted from the notes, reviewed by the person who did the work, and dated. The table below is the whole routine on one page.
| When | Who | What is produced | What it shows a reviewer |
|---|---|---|---|
| Project approval | Technical lead | Dated half-page project note: the aim, the unknown, the candidate approaches, the people | The uncertainty existed at the outset and was judged by someone competent in the field |
| Continuously | Engineers, developers, scientists | Tickets, commits, test results and design notes, each with a sentence of why | The work done to resolve the uncertainty, including the attempts that failed |
| Monthly | Team lead, confirmed by finance | Time allocation by person and project, signed | The basis on which staff costs were apportioned to the R&D |
| Each milestone | Technical lead | Checkpoint entry: what was resolved, what is still uncertain, what changed | How the work progressed and where the R&D ended |
| Year end | Finance, with a technical reviewer | Cost schedule reconciled to the ledger; description drafted from the notes and dated | Costs linked to the work; the description agrees with the records |
What "made at the time" does not mean
It does not mean perfect. It means the records were made at or near the time and were not created to support the claim. A dated engineering wiki page with typos is better evidence than a polished document written the week before filing, because the first shows what the team thought while the outcome was still unknown. The most common sign of a reconstructed record is that it describes the solution rather than the uncertainty.
Who owns it
Documentation fails when it belongs to everyone. Name one person per project who is responsible for the note and the checkpoints, and one person in finance who is responsible for time and cost. Put both names in the project kickoff, and put the monthly time sign-off and the milestone checkpoints in the calendar.
Illustrative example
Illustrative example. Halvard Instruments Ltd, a hypothetical company, runs three development projects in a year with twelve engineers. At approval each technical lead writes the half-page note; one project is dropped at this stage because the lead cannot name anything the team does not already know how to do. Through the year the team leads sign a monthly allocation showing that six engineers spent, on average, half their time on the uncertain work. Four checkpoint entries per project record what was resolved and where routine work took over. At year end finance takes total staff costs of 1,200,000 for the twelve engineers, applies the signed allocations to arrive at eligible staff costs of 300,000 (six engineers at half time), adds 50,000 of consumables tied to the test programme, and drafts the description from the notes in two days. The benefit is calculated at that country's rate on eligible costs of 350,000. When a reviewer later asks how the half-time figure was reached, the answer is twelve signed monthly sheets, not a recollection.
Country differences
The routine above is the same everywhere. What differs is the form the records are eventually reported on and the words each scheme uses. Each country page holds the detail.
- United States. Keep records that link each business component to the four-part test and to the wages, supplies and contract research costs you claim. For tax years beginning after 2025, Section G of Form 6765 asks for information about each business component, with some exceptions: Section G requires, for each business component, its name, type, the information sought to be discovered, and the qualified research expenses allocated to it, and it is optional for a qualified small business making the payroll tax election and for a taxpayer with total QREs of 1.5 million US dollars or less and gross receipts of 50 million US dollars or less. Organising records by business component from the start saves rework. See Documentation: United States and Form 6765.
- United Kingdom. Every claim needs an additional information form that describes the R&D projects and breaks down the qualifying costs. Keep project records that show the advance sought, the uncertainties and how the work resolved them; the form requires, for each project described, the main field of science or technology, the baseline level of science or technology, the advance sought, the scientific or technological uncertainties, and how the project sought to overcome them. These map directly onto the project note. See Additional information form: United Kingdom.
- Ireland. Keep records made at the time that show the scientific or technological uncertainty, the work carried out to resolve it and the expenditure incurred. Revenue expects contemporaneous documentation for the science test, including evidence of the uncertainty, the advancement sought and the systematic investigation, and for the accounting test, including timesheets or equivalent records, invoices and a reconciliation of the claim to the accounts. See Documentation: Ireland.
- Canada. Form T661 asks you to describe the work and calculate your expenditure. Keep records made as the work happens, such as hypotheses, test results and time records, that show the systematic investigation. Part 2 of the form asks for each project the scientific or technological uncertainties attempted to be overcome, the work performed to overcome them, and the scientific or technological advancements achieved or attempted. See Form T661: Canada.
- Australia. Keep records that show the hypothesis, the experiments, the results and the conclusions for each registered activity, and that link expenditure to those activities. The department asks registrants to describe each core R&D activity by its hypothesis, experiment, observation and evaluation, and logical conclusions; the checkpoint entries provide them. See Documentation: Australia and Registration with AusIndustry.
- New Zealand. Approval comes before the claim. Keep records that show each core activity, the uncertainty it addresses and the systematic approach taken, and link expenditure to it. The general approval application asks for each core activity the scientific or technological uncertainty, the new knowledge or improved process, service or good sought, the systematic approach and the start and end dates, so the project note is written before the year starts rather than after it ends. See General approval: New Zealand.
- Singapore. Keep documents that show each project's objective, its novelty or technical risk, the work done and the qualifying expenditure. IRAS can ask for them when it reviews a claim, and expects staff costs claimed for R&D to be supported by records of the time spent on R&D, such as timesheets or project allocation records. See Documentation: Singapore.
Related pages
- Contemporaneous documentation: what reviewers actually look for
- Writing a technical narrative that survives review
- Technical Teams Program, Module 3: Recording as you go
- In-House Finance Team Program, Module 3: Staff time apportionment
- R&D Tax Fundamentals, Module 5: Documentation basics
- Documentation and compliance hub
Educational material, not advice. Record-keeping requirements depend on the scheme and the facts.
Sources
- Instructions for Form 6765Internal Revenue Service
- Submit detailed information before you claim R&D tax reliefGOV.UK
- Tax and Duty Manual Part 29-02-03Revenue Commissioners
- T4088 Guide to Form T661Canada Revenue Agency
- Keeping records for the R&D Tax Incentivebusiness.gov.au
- Research and development tax incentiveInland Revenue
Educational content only, not tax advice. Rules change and eligibility depends on your circumstances, so check the official guidance or speak to a qualified advisor before you claim. See the disclaimer.