The narrative flags
- Severe: no uncertainty you can say out loud. If nobody can state, in one sentence, what was not known to be feasible or how to achieve it, there may be no claim. The definition turns on it.
- Severe: the baseline is missing. No account of what the field could already do means “readily deducible” cannot be answered, and the narrative is assertion.
- Amber: the narrative describes the product, not the problem. Features, benefits and effort are HMRC’s own examples of what does not qualify; uncertainty and method are what does.
- Amber: no boundary drawn. The claim reads as the whole commercial project, with no start at the uncertainty and no end at its resolution. HMRC’s examples cut at exactly those points.
- Amber: the competent professional has not read it. A narrative unreviewed by someone meeting GfC3’s definition, current in the field as a whole, experienced, with a track record, is unanchored.
The cost flags
- Severe: apportionments with no evidence. Round percentages across the payroll that no timesheet, ticket or calendar supports. This is one of the most common errors HMRC finds in checks.
- Severe: excluded costs inside. Rent, rates, capital spend, patents, production: the exclusion list is published, and their presence signals nobody checked.
- Amber: the 65% rules unapplied. Unconnected subcontractor and agency-worker payments claimed in full.
- Amber: overseas spend claimed as if pre-2024. Contractor work done abroad, or workers outside UK PAYE, without the narrow exception being genuinely met.
- Amber: numbers that do not reconcile. Claim categories that cannot be traced to the accounts and payroll within an afternoon.
The process flags
- Severe: the notification question unasked. First-time or lapsed claimant, and nobody has checked the six-month window. This one invalidates claims outright.
- Amber: the AIF as an afterthought. Drafted in a rush after the return is ready, or worse, the return filed first, which HMRC states leads to rejection.
- Amber: the named officer has not read the claim. The senior contact on the AIF stands behind it; a signature without reading is a personal risk taken unknowingly.
The adviser flags
- Severe: the claim value was promised before the work. Nobody can know it, and a claim built backwards from a promise inherits the promise’s dishonesty.
- Amber: you have never heard “no” or “that part does not qualify”. An adviser who has qualified everything you ever showed them is not applying the test; they are applying a fee model. What good looks like is on choosing an R&D adviser.
Frequently asked questions
How many flags are too many?
One severe flag, no articulable uncertainty, costs that cannot be evidenced, a narrative nobody technical has read, is enough to stop and fix before filing. Several amber ones compound: each is survivable alone, together they draw the pattern enquiries are selected on.
Who should run this checklist?
Two people, deliberately: the competent professional for the technical flags, and whoever owns the numbers for the cost and process flags. The failure mode is one person marking their own homework across both.
We found flags in a claim we already filed. What now?
Grade them honestly. Presentation weaknesses can wait for the next claim; substantive errors, unsupportable costs, work that does not meet the definition, belong on the amendment page's machinery, where moving unprompted is what keeps penalties at the bottom of the ranges.
Is this list what HMRC actually looks for?
It is built from HMRC's published guidelines, its worked examples and its stated compliance concerns, plus enquiry experience. HMRC's selection criteria are its own, but every flag here maps to something in its public material.