Why CRA denies SR&ED claims: 6 narrative mistakes

Most SR&ED claims that get denied at review were eligible work to begin with. The team ran real experimental development. The technical narrative just described it in a way the reviewer could not accept, and the claim failed on the write-up rather than the work. That gap is the one part of a claim a consultant fully controls, so it is worth getting right.

It matters more this year. Bill C-15 received Royal Assent on March 26, 2026, and the enhanced 35% investment tax credit now runs to a $6 million expenditure limit instead of $3 million, reaches eligible Canadian public corporations, and again covers capital costs incurred after December 15, 2024 (CRA, SR&ED news and updates). More dollars ride on each claim, which means more claims worth a closer look. A narrative that would have squeaked through on a small refund is a bigger target once the refund doubles.

Here are the six narrative mistakes that most often sink an otherwise eligible claim, and how to write around each.

At a glance

• A review checks two things: whether the work meets the SR&ED definition (the "Why" and the "How") and whether the expenditures are allowable (CRA, what work is eligible).
• The service standard for a refundable claim selected for review is 180 calendar days from a complete claim, which CRA aims to meet 90% of the time (CRA review guide, updated April 20, 2026).
• The enhanced 35% ITC limit rose from $3M to $6M and the taxable-capital phase-out band widened to $15M to $75M, for taxation years beginning on or after December 16, 2024. Separately, capital expenditures are eligible again for property acquired on or after that same date (CRA updates).
• A failed experiment does not weaken a claim. Ruling an idea out counts as new knowledge (CRA guidelines).

Why eligible SR&ED work still gets denied at review

A reviewer is not grading your product. They are checking whether the narrative and your contemporaneous evidence show two things: that the work aimed at a scientific or technological advancement against a real uncertainty (the "Why"), and that it was a systematic investigation with a hypothesis tested by experiment or analysis (the "How") (CRA, what work is eligible). Miss either one on paper and the eligibility fails, no matter how hard the work was.

Six narrative mistakes that fail a CRA review

Each of these is a way of describing eligible work so that it stops looking eligible. Fixing them is editing, not new science.

1. You described the product, not the uncertainty

Line 242 asks what technological uncertainty you faced, and the most common answer reads like a feature list. "We built a real-time pipeline that processes 50,000 events per second" tells the reviewer what the product does, not what you did not know how to do. The uncertainty is the reason the work was not routine: the point where standard practice and publicly available knowledge ran out and you could not predict whether your approach would hold. Write that instead. An illustrative rewrite: "Existing back-pressure techniques degraded non-linearly above roughly 30,000 events per second, and available literature did not tell us whether batching by tenant would keep latency under 200ms without dropping events."

2. There is no hypothesis, so the work does not look systematic

The "How" requirement wants a systematic investigation: define the problem, advance a hypothesis, test it by experiment or analysis, and draw logical conclusions (CRA guidelines). Narratives that jump from "the problem was hard" to "we solved it" skip the middle and read as ordinary engineering. Line 244 should carry the hypothesis, the test, and the result, even briefly. Say what you thought might work, what you tried, and what the result told you. If you cannot phrase the work that way, that is worth knowing before the reviewer does.

3. You claimed a business problem as a technological uncertainty

"We needed the feature by Q3" and "the client wanted lower cloud costs" are business pressures, not technological uncertainties. Reviewers are trained to separate the two, and a narrative built on deadlines and cost targets invites the finding that the work was commercial development. Ground the uncertainty in the technology: which property of the system, at the level of science or engineering, could not be resolved with existing knowledge. Business context can frame the project. It cannot be the uncertainty.

4. You hid the experiments that failed

Consultants sometimes polish a narrative into a straight march to success, worried that dead ends look like waste. They do the opposite. CRA guidance is explicit that if your work showed an idea was not a solution for your situation, that can be new knowledge (CRA guidelines). A path with ruled-out approaches reads as a genuine systematic investigation. A flawless one reads as implementation you already knew would work. Keep the failed hypotheses in.

5. You wrote in marketing language instead of technical specifics

"Innovative," "cutting-edge," and "state-of-the-art" carry no information a reviewer can test. They also read as the opposite of engineering notes written at the time, which raises doubt about whether the work happened as described. Swap adjectives for measurements. Instead of "a sophisticated new algorithm," write what it did and against what baseline: "cut false positives from 12% to 4% on the held-out set, where the prior method had plateaued." Specifics do the work that superlatives cannot.

6. The narrative has no contemporaneous evidence behind it

A narrative is a claim, and the review tests it against records made while the work happened. If the write-up describes experiments that leave no trace in commits, test logs, lab notebooks, or dated design notes, the reviewer has nothing to corroborate and the claim weakens. Write the narrative from the evidence, not from memory, and make sure each experiment you describe maps to something dated. If the record does not exist, do not narrate it.

How to pressure-test a narrative before you file

Read the draft the way the reviewer will: 14 months later, with none of your context, looking only for the "Why" and the "How." For each project, check that Line 242 names a technological uncertainty rather than a product goal, that Line 244 shows at least one loop of hypothesis, test, and result, and that every experiment you mention has dated evidence behind it. A section that survives that read will survive the review. If it leans on adjectives, deadlines, or a suspiciously clean story, fix it now, while it costs an edit rather than a refund.

Frequently asked questions

Why do SR&ED claims get rejected at review?

Most rejections are not about the work being ineligible. They come from narratives that describe the product instead of the technological uncertainty, skip the hypothesis-and-test structure, or lack contemporaneous evidence to back what the write-up claims.

What is CRA looking for in the SR&ED technical narrative?

Two things: a scientific or technological uncertainty that could not be resolved with available knowledge (the "Why"), and a systematic investigation that formed and tested a hypothesis by experiment or analysis (the "How"). Both need to be visible in the narrative and supported by evidence.

Does a failed experiment hurt an SR&ED claim?

No. CRA guidance treats a ruled-out idea as new knowledge, and you do not have to reach your goal to have done SR&ED. A narrative that includes failed hypotheses often looks more systematic, not less.

How long does a CRA SR&ED review take?

The service standard for a refundable claim selected for review is 180 calendar days from the date CRA receives the complete claim, which it aims to meet 90% of the time. There is no service standard for non-refundable claims.

Do the 2026 SR&ED changes affect what the narrative must prove?

No. Bill C-15 raised the enhanced expenditure limit and restored capital costs, but the eligibility test is unchanged. The "Why" and the "How" still decide the claim. The changes raise the dollars at stake, not the bar for the narrative.

Where this leaves your practice

The eligibility rules did not move this year. The money did. That makes narrative discipline the cheapest risk reduction on the board. Draft to the "Why" and the "How," keep the failed hypotheses in, and tie every experiment to dated evidence. Whatever tooling you use for it, SREDlog builds the narrative straight from the evidence so the sentence and the record that supports it never drift apart, though the habit matters more than the tool. Write the claim you would want to defend 14 months from now, because that is when the reviewer reads it.