A Nine-Part Framework for R&D Project Records

The R&D compliance environment has changed. For claims made on or after 8 August 2023, the required additional information must be submitted on or before the date the claim is made (see CIRD182000). On 23 January 2025, HMRC updated Part 5 of its Guidelines for Compliance 3 (GfC3), setting out its own recommended approach to claims and record keeping. One of the most important practical factors in whether a claim survives HMRC scrutiny is the quality of the contemporaneous record the competent professional kept while the work was being done.

Most companies do not keep records in a form HMRC recognises. Slack messages, sprint tickets, and design documents contain most of the evidence, but they are scattered across systems and written in a language HMRC does not read. HMRC itself is clear that companies do not need to create unnecessary records or spend money on expensive systems. What is needed is a lightweight project record kept alongside the work, structured around the questions HMRC will ask. This blog sets out Lexmore’s nine-part framework for that record, aligned with HMRC’s GfC3 Parts 3 and 5 and the wider guidance in the DSIT Guidelines and the CIRD manual.

Who this applies to

• Competent professionals leading the technical work on an R&D project (CTOs, heads of engineering, senior scientists, lead developers).
• Finance directors and R&D managers responsible for evidencing claims to their accountants or advisers.
• Founder-operators of technical SMEs where the CEO is also the competent professional.

HMRC’s recommended approach: plan ahead, or look back

GfC3 Part 5 makes clear that the strongest position is to identify potential R&D before or during the work. HMRC’s stated view is that claims are more likely to be correct when the company is aware, at the time of the work, that it may qualify. The recommended actions for planning ahead:

1. Consider whether the work you are planning may need new knowledge or techniques in a qualifying field of science or technology.
2. Write down the advances you are seeking, and the uncertainties you need to resolve. HMRC is explicit that the plan should be in proportion to the size and cost of the project; a page of bullet points can be enough for a smaller project.
3. Get the written opinion of a competent professional in the field, including their qualifications and experience. HMRC is clear that a simple assertion that the project qualifies as an advance is not enough.

GfC3 Part 5 also recognises the reality that a project can be identified after it has begun or finished. This is explicitly acceptable. A competent professional can give an opinion on work that is ongoing or already completed. Where a company operates in an industry in which scientific or technological advances are likely, HMRC recommends thinking now about how you will identify qualifying activity and qualifying costs at a later date.

What the competent professional’s opinion needs to cover

Per GfC3 Part 3 and CIRD81300, a competent professional is knowledgeable about the relevant scientific or technological principles, aware of the current state of knowledge in the field as a whole, and has accumulated experience and a successful record in the field. General project management or software delivery experience is not enough, and a director’s title on its own does not qualify.

The written opinion needs to set out relevant facts and reasoning specifically enough that both the company and HMRC can understand:

• the extent of the work, and where it began and ended
• the field and sub-field of science or technology
• the advance being sought, and how it advances the current state of knowledge in the field
• the uncertainties that needed to be overcome

The nine-part framework

This is Lexmore’s nine-part framework, drawn from HMRC’s GfC3 Parts 3 and 5 and the record-keeping guidance in CIRD80560. Kept as a single document per sub-project, updated as the work progresses, and proportionate to the size of the project.

1. Project identification

Sub-project name, internal reference, start date, current status, accounting period(s) the work falls within, and a two-sentence business context (what the deliverable is and where it sits in the product or process).

2. Competent professional(s) and their opinion

Named individual(s) with basis of competence: qualifications, years in the field, relevant prior projects. Include the competent professional’s written opinion on the project, covering the four points GfC3 Part 5 asks for (extent of work, field and sub-field, advance sought, and uncertainties to be overcome).

3. The advance sought

One or two sentences describing the intended advance in overall scientific or technological knowledge or capability, naming the field and sub-field. Not a description of the commercial deliverable. The advance must be to the field, not merely to the company’s own product.

4. The baseline

The existing state of knowledge or capability in the field before the project started. Cite what a competent professional in the field would reasonably have known: published techniques, existing tools, standard industry approaches, documentation or case studies the company has reviewed. If the answer is “we searched X, Y and Z and found no existing solution”, record that search.

5. The uncertainty log

A dated log of each scientific or technological uncertainty as it arose. For each entry: the date, a description of the uncertainty, why it was not readily deducible by a competent professional (routine adaptation, application of known techniques, and reference to publicly available solutions do not count as uncertainty), the initial hypothesis, and the plan for resolving it.

6. The iteration log

A dated log of the systematic investigation. For each attempt: what was tried, why, the result (including negative results and failed approaches), what was learned, and what was tried next. A record that shows only successful attempts provides weaker evidence of systematic investigation than one that includes what did not work and how the approach evolved in response.

7. Resolution and boundary

The date and description of the point at which the uncertainty was resolved, or the point at which the project was abandoned as scientifically or technologically unfeasible. This defines the end of qualifying R&D activity. Work after this point is normally routine development, deployment, or commercialisation and does not qualify.

8. Staff time apportionment and claim methodology

For each staff member contributing, keep a log of their time spent on qualifying R&D activity during the accounting period (not the wider project). Record the claim methodology, any sampling and techniques used, as well as the method used to arrive at any apportioned figures.

9. Cross-reference to supporting evidence

Pointers to the underlying evidence that supports each section. GfC3 Part 5 lists the kinds of documents HMRC may ask to see in a compliance check: project charts, drawings, designs, test results, photographs of prototypes (particularly where modifications show how development problems were resolved), minutes of meetings, email exchanges, and any other relevant records the company holds. The record acts as an index into your existing systems (Jira, Git, Confluence, lab notebooks, shared drives).

Common gaps we see

• No named competent professional, or a bare assertion. GfC3 Part 5 is explicit that a simple assertion that a project qualifies as an advance is not enough. The opinion needs reasoning, and the person’s qualifications and experience need to be recorded.
• Advance described as a commercial deliverable. “A faster checkout” is not an advance in science or technology. “A method for X in the presence of Y that existing approaches cannot handle” is.
• No baseline. Without a documented baseline of existing knowledge in the field, HMRC cannot see what the advance is against. One of the most common weaknesses we encounter in the AIF forms we review.
• Uncertainties described only in outcome. “We solved X” is not evidence of resolving uncertainty. A dated log of the uncertainty as it arose, and why it was not readily deducible, is.

Thin records versus strong records

Thin record (typical)	Strong record (defensible)
Retrospective narrative written up at year-end from memory.	Contemporaneous entries dated as the work happened, updated fortnightly.
Competent professional not named; opinion is a bare assertion.	Named individual with qualifications and experience, plus a written opinion covering extent, field, advance, and uncertainties.
Advance described in commercial terms (“better product”, “faster system”).	Field and sub-field named, advance stated in field-relevant terms, with a baseline it is measured against.
Uncertainties described only in outcome (“we solved X”).	Uncertainties logged at the point they arose, with the reason they were not readily deducible.
Iteration log absent or shows only successful attempts.	Iteration log includes failed attempts, negative results, and the reasoning that followed.
Staff time recorded as a flat percentage with no supporting log; no claim methodology documented.	Time logged against qualifying R&D activity, with the claim methodology and any sampling and techniques documented.

Worked example: a short extract

An extract from a fictional record for a manufacturing process control sub-project, illustrating the intended level of detail:

Sub-project: Closed-loop moisture control for continuous extrusion line. Accounting period 1 April 2025 to 31 March 2026.

Competent professional: J. Patel, MEng Chemical Engineering (Imperial 2011), 12 years in polymer processing, prior project on inline melt-flow control for XYZ Ltd (2020).

Written opinion (extract): The project seeks an advance in the field of polymer processing, sub-field inline process control for continuous extrusion, by developing a closed-loop moisture control method for polyamide at throughputs above 400 kg/hr and process temperatures above 260°C. Existing published techniques rely on offline sampling and manual correction; no commercially available sensor system provides reliable closed-loop control in this temperature and throughput envelope. The uncertainty is whether inline optical sensors can be stabilised sufficiently under sustained thermal load to close the control loop. Qualifying work began 1 May 2025 with baseline literature review and ends at either resolution of that uncertainty or a formal decision that closed-loop control is not achievable with the current sensor generation.

Baseline: Reviewed 2019-2024 literature (SPE, Polymer Engineering & Science, Elsevier database search May 2025); manufacturer specifications from three sensor suppliers (documented in project folder); consultation with a competent professional at [supplier] confirming no off-the-shelf closed-loop solution above 240°C.

Uncertainty log entry, 15 June 2025: NIR sensor drift under sustained thermal load exceeds manufacturer’s stated tolerance by a factor of ~4 in early trials. Not readily deducible: sensor is being operated within its rated temperature range; the drift appears to be a function of exposure duration not peak temperature, which is not documented in the sensor specification or in published literature. Initial hypothesis: thermal cycling of the sapphire window is inducing measurement bias. Plan: instrument a second sensor with independent thermal monitoring and compare drift profiles over 72 hours.

Iteration log entry, 3 July 2025: Attempted mitigation via additional thermal shielding of sensor housing. Result: 30% reduction in drift but still outside acceptable tolerance for control loop stability. Cause: shielding also reduced optical path stability. Next: consider active thermal management of sensor housing rather than passive shielding.

Practical mechanics and cadence

• One document per sub-project, in the same system the team already uses (Confluence, Notion, SharePoint, a shared Word document). The tool matters less than the discipline.
• Version-controlled, so HMRC can see the record was built contemporaneously rather than reconstructed. Editing history in Confluence, Notion, or Google Docs is normally enough.
• Ownership. The technical content should be prepared or approved by the competent professional. Finance teams and advisers can help collect, organise, and reconcile the supporting evidence.
• Weekly cadence: add new entries to the uncertainty log on the day an uncertainty surfaces (reconstruction after the fact tends to lose the “why not readily deducible” detail); add new entries to the iteration log for each attempt.
• Fortnightly review by the competent professional for accuracy and completeness (30-45 minutes).
• End of accounting period: update staff time apportionments, document the claim methodology, and refresh cross-references to evidence.
• Proportionate to the project. GfC3 Part 5 is explicit that HMRC does not expect unnecessary records or expensive systems. A page of bullet points can be enough for a small project; a large project is expected to already have detailed records generated by the work itself.

How Lexmore can help

Lexmore offers a free R&D eligibility appraisal that includes a review of your existing project records against the CIRD80560 record-keeping guidance and the GfC3 Part 5 recommended approach. We identify what is strong, what is thin, and what to fix before the accounting period closes. HMRC compliance protection and a fee guarantee are included as standard for Lexmore Tax Advisory clients.

Lexmore’s View

Contemporaneous records are one of the most important practical factors in whether a claim survives HMRC scrutiny. HMRC has now told companies, in its own words, what it expects. GfC3 Part 5 is not new law and it is not mandatory, but it is HMRC’s explicit recommendation, and following it puts a claim in a materially stronger position than not.

The framework above will not turn ineligible work into eligible work. What it will do is give an eligible project the documentary base a claim needs to stand up to the level of enquiry HMRC now brings.

Log the uncertainty. Log the iteration. Log the person.

References

• Guidelines for Compliance 3 (GfC3), Part 5: Recommended approach to claims and record keeping (updated 23 January 2025) | gov.uk
• Guidelines for Compliance 3 (GfC3), Part 3: Importance of a competent professional | gov.uk
• CIRD80560: Records HMRC may expect to see | HMRC internal manuals
• CIRD81300: Competent professional | HMRC internal manuals
• CIRD182000: Additional Information Form requirements | HMRC internal manuals
• Meaning of Research and Development for tax purposes (DSIT Guidelines, 7 March 2023) | gov.uk