Positioning · How claims work · SR&ED
La RS&DE, c'est 85 % technique et 15 % comptable — pourquoi un ingénieur devrait rédiger votre demande
Une demande RS&DE n'est pas un formulaire fiscal avec une histoire greffée dessus. Le plus difficile, c'est le récit technique : expliquer ce qui était réellement incertain et comment vous avez travaillé à le résoudre, dans un langage qu'un examinateur de l'ARC reconnaît. Voici pourquoi c'est un travail d'ingénierie, pas de comptabilité.
Note : les articles de l'Insider ne sont disponibles qu'en anglais pour le moment. La version française du journal suivra.
There's a line I've used since my earliest days doing this work: SR&ED is only 15% accounting and 85% technical. The numbers — the labour, the materials, the overhead — are the easy part. They're real, but they're arithmetic. The hard part, the part that wins or loses a claim, is the technical narrative: explaining what was genuinely uncertain about your work, and how you went about resolving it, in a way a CRA reviewer recognizes.
That's an engineering problem. Not an accounting one.
Three ways claims get written — and why the source matters
Accounting-led (the bolt-on). Most SR&ED claims in this country are prepared by accounting and tax firms where SR&ED is one line item among many. They lead with the financials and treat the technical write-up as a final documentation step — something to be filled in once the numbers are settled. The trouble is that the technical story is the substance of the claim, not a formality. When the narrative is thin, that's exactly what a reviewer notices.
Software auto-generation (the export). A newer crop of platforms promises to generate your claim by scraping your Jira, Asana, or Azure DevOps tickets and assembling a narrative automatically. It's fast. But a claim built from a ticket export tends to read like a ticket export — generic, boilerplate, and disconnected from the genuine uncertainty that makes the work eligible. The CRA already triages claims with an automated risk-assessment tool before a human ever opens them — its own privacy impact assessment describes the preliminary accept-or-review determination as automated, with no assessor input (CRA — SR&ED Program privacy impact assessment). What that tool does and doesn't weigh isn't published, and I'd be wary of anyone who tells you they know. What I'll say from my own files is narrower: a narrative that reads like it was assembled rather than written tends to generate questions, because it doesn't answer the ones a reviewer actually has.
Engineer-led (the way I do it). I come at a claim technical-first. Before I think about a single dollar, I want to understand what you were trying to do, what wasn't known when you started, what you tried, and what didn't work. As I put it to clients: my job is identifying those technical challenges and articulating them in a way that CRA will recognize. The financials follow the story — not the other way around.
Why being honest about failure makes a claim stronger
This is the part that surprises people. A good SR&ED narrative doesn't hide the dead ends — it leans on them. Technological uncertainty is what makes the work necessary — the CRA's test is whether you were trying to achieve an advancement the existing knowledge base couldn't deliver, and whether you went after it as a systematic investigation. Nothing demonstrates genuine uncertainty like an approach you tried in good faith that didn't pan out. Work that didn't succeed still qualifies1, and a claim that honestly documents those dead ends is more defensible, not less, because it reads like real development instead of a tidy success story written after the fact.
A machine scraping your closed tickets won't tell that story. It doesn't know which failure mattered. An engineer who's sat with your team does.
What this means when you're choosing who writes your claim
The CRA isn't an adversary here — it accepts about 96% of the claims it processes — 90% of them exactly as filed, another 6% after modifications2 (CRA Annual program statistics). But "accepted" and "accepted for the full, defensible amount you earned" are not the same thing. The difference, more often than not, is whether the technical narrative actually captured the work.
So when you're deciding who writes your claim, ask a simple question: who is holding the pen on the technical story? If the answer is "an accountant, as a last step" or "a tool, from our ticket export," that's worth a second thought.
When you work with me, you talk to me — an engineer, not a salesperson — on every discovery call. Not a rotating bench, not a junior, not a platform. And when the CRA has questions, we're the ones sitting in the hot seat answering them.
Sources
- Work that didn't succeed still qualifies — provided it was a genuine systematic investigation; learning that an idea is not a solution counts as new knowledge — CRA — What work is eligible for SR&ED ("You do not have to achieve your goal in order to gain new knowledge")
- The CRA accepts about 96% of SR&ED claims, with 90% accepted exactly as filed — CRA — SR&ED annual program statistics (FY2025–26: 23,677 claims processed; 90% accepted as filed, 6% accepted after modifications, 4% denied)
Note: the "SR&ED is 85% technical / 15% accounting" framing, and the observation that assembled-feeling narratives tend to draw questions, are the author's professional characterizations from practice — not figures drawn from a published source. The CRA's automated preliminary risk assessment is documented in its published privacy impact assessment; claims can also be selected for review for other reasons, including random sampling.
Want to see what an engineer-led claim looks like for your work? Book a 30-minute call with Amin. No prep, no obligation — just a conversation about what you've been building.