Building Your CIR Approval File: What the Ministry Actually Looks At

An application for approval under France’s research tax credit (crédit d’impôt recherche, CIR) is not a form. It is assessed on the merits by a scientific expert in the field, who may request further information, before the minister responsible for research decides. That difference in nature explains most refusals: solid, technically excellent companies fail because they wrote a sales document where a scientific one was expected.

Here are the four blocks that get examined, and what separates an application that gets through from one that stays on the desk.

Block 1 — Human resources

This is the heart of the file, and by far the most decisive part. The ministry wants to identify the people who actually carry out the research work: headcount assigned to R&D, degrees, qualifications, detailed CVs, length of service in the organisation.

Three pitfalls come up again and again.

The phantom team. A file announcing “twelve R&D engineers” without naming anyone, with no CVs or degrees, does not convince. The assessor is looking for trajectories: initial training, specialism, any publications, years of experience on the subject.

The drift towards delivery profiles. A team made up entirely of application technicians or production developers, with no identified scientific leadership, starts with a serious handicap. This does not mean a PhD is compulsory: it means you must show who carries the experimental approach.

Forgetting the technical directors. In small organisations the founder is often the principal researcher. Their CV belongs in the file just as much as the employees’ — and it is frequently the most convincing document in it.

Block 2 — Material resources

Premises, laboratories, test benches, measurement equipment, computing resources, prototyping or simulation environments. The underlying question is constant: does this organisation have the means to do, in-house, what it claims to do for its clients?

For a software research company, “material resources” means test environments, datasets, computing capacity, performance measurement instruments. Describe them as precisely as a laboratory would describe its benches. The apparent absence of heavy equipment is not disqualifying; the absence of description is.

One often-neglected point: shared or rented resources (technical platforms, computing time, access to university equipment) are worth mentioning and documenting. They count, provided they are real and traceable.

Block 3 — The research operations described

This is where it is won or lost. You have to present representative work and, for each piece, follow the same structure:

1. The state of the art at the outset: what was publicly known how to do when the project began?
2. The scientific or technical obstacle: what could that state of the art not resolve?
3. The experimental approach: what hypotheses, what trials, what iterations, what intermediate results — including failures, which are a signal of genuine research.
4. The results and their scope: what is now possible, and what remains open.

The reference criteria are those of the OECD Frascati Manual, to which the French tax authorities constantly refer: novelty, creativity, uncertainty, a systematic character, transferability or reproducibility of results.

The fatal framing mistake is to confuse technical difficulty with scientific uncertainty. A long, complex and costly development may involve no uncertainty within the meaning of the scheme, if the solution was known and only had to be applied to a particular context. Conversely, a short project may hide a real obstacle. It is the nature of the problem that counts, not the effort expended. Reframe each operation by starting with the sentence: “at the outset, nobody knew whether…”. If that sentence cannot be finished honestly, the operation is probably not research — and the innovation tax credit route deserves consideration, as set out in our comparison of CIR and CII approval.

Block 4 — Dissemination

Scientific publications, patents filed, conference papers, collaborations with academic laboratories, PhDs supervised under industrial research agreements. These are not required as such, but they are the most convincing proof that a research activity really exists — because they have been submitted to the judgement of others.

A company that has never published is not doomed. It simply has to compensate with the precision of the operations described and the quality of the profiles presented. Conversely, an organisation piling up publications but describing its third-party work poorly can disappoint: approval concerns the capacity to do research for clients, not scientific renown in general.

What research says about outsourcing R&D

An approval file describes a particular activity: doing research on someone else’s behalf. The management literature shows that this relationship is not neutral for the client. Christoph Grimpe and Ulrich Kaiser published a study in the Journal of Management Studies in 2010, on panel data of innovating firms, showing an inverted U-shaped relationship between R&D outsourcing and innovation performance: the gains from acquiring external knowledge run, beyond a certain point, into a dilution of internal resources and a loss of integrative capability — a relationship that is more favourable the more substantial the firm’s internal R&D.

That lesson translates directly into an approval file: show that you are an integrable research partner, not a supplier of hours. Describe the interfaces with the client’s teams, the intermediate deliverables, the transfers of competence. That is also what the assessor is trying to grasp when reading your operations.

A pre-filing checklist

Block	Expected documents	Sign of weakness
Human resources	R&D headcount, degrees, qualifications, CVs, length of service	No names, no degrees, no identified scientific leadership
Material resources	Premises, equipment, instruments, computing resources, shared platforms	Generic description, disconnected from the operations described
Research operations	State of the art, obstacle, experimental approach, results	A sales narrative of successful projects, with no stated uncertainty
Dissemination	Publications, patents, conferences, collaborations, PhDs	Nothing, with no compensation through the quality of the operations
Overall coherence	The people cited do work on the operations described	Team and projects presented in two parallel universes

The last line is the one most often neglected. A credible file reads as a whole: the people in block 1 must appear in the operations of block 3, and the resources of block 2 must be the ones those operations used.

After filing

Assessment takes several months. The expert may request further information: answer quickly and fully, reusing their wording rather than pointing to a bulky attachment. The decision is then notified by the minister, for a duration that varies with activity and seniority in the scheme — three years as a rule, one year with no research operation in the previous twelve months, up to five years for long-standing holders.

Two habits are worth adopting for what follows. Keep the file you filed: it will be the basis for renewal, in the 15 August to 30 November window detailed in our article on the approval timetable. And separate now, in your management accounts, research carried out on your own account from research carried out for third parties: the distinction, made as you go, beats reconstructing it in front of an auditor.

Take action

Take your three most recent operations and write four paragraphs for each: state of the art, obstacle, experimental approach, results. If the exercise holds up, so will your file — and you will already have written the hardest block. If it resists, that is valuable information: it will save you a pointless filing and a lost year. The overall mechanism is explained in our article on subcontracted research and approval, and the detailed scheme status, with the free ebook, is on our page dedicated to CIR-CII approval.