The SR&ED consulting industry runs on templates written during the Harper era.

Same narrative structures. Same Form T661 language. Same advice about what to claim and how to describe it. Find-and-replace the company name, swap in a few project details, send the invoice. This has been the business model for twenty years, and it worked because the program was more generous, CRA reviewers were less trained, and the body of Tax Court decisions defining “technological uncertainty” was a fraction of what exists today.

None of those conditions still hold.

What Actually Changed

Three shifts matter. They’re not subtle, and they’ve been compounding since the 2012 program overhaul [1].

RTAs got better at their jobs. Research and Technology Advisors used to be the quiet step in the process. Now they’re the step that kills claims. CRA has invested in training, in cross-referencing claims within industries, in pattern recognition. An RTA reviewing your claim has likely reviewed dozens of others in your sector this year. They know what template language looks like. They know when a narrative was written by someone who never spoke to the engineers.

The friendly desk review where a polished writeup sailed through? That still happens. Just less often, and never for the claims that actually need scrutiny.

Case law drew sharper lines. Two decades of Tax Court decisions have turned vague program concepts into specific legal tests. “Technological uncertainty” in 2010 was something you could argue creatively. In 2026, there’s a body of precedent that tells you exactly where the line sits, and RTAs know that precedent as well as any tax lawyer.

CRA learned to spot retroactive documentation. They’ve seen every template. Every cookie-cutter format. Every suspiciously polished narrative that was clearly written eight months after the work happened. They know what it looks like when someone reverse-engineers a claim from a project summary, and they know how to test for it.

The Five Ways Your Advisor Is Failing You

Not all of these apply to every firm. But if you recognize more than one, you have a process problem.

Boilerplate Narratives

This is the one that should scare you most.

Open a claim prepared by a traditional SR&ED consultant. Read the project description. Now ask yourself: could this description apply to any company doing similar work? If the answer is yes, that’s exactly what an RTA will think too.

Template language doesn’t just fail to differentiate your work. It actively signals that the person who wrote the claim didn’t understand the technical specifics. And the RTA’s next thought is predictable: if the writer didn’t understand the work, maybe the work didn’t actually meet the threshold.

RTAs review claims across your sector all year long. They recognize recycled phrasing the way you’d recognize a stock photo on a competitor’s website. When your claim reads like the last twelve they reviewed, you’ve already lost credibility before the first question gets asked.

”Just Tell Us What You Did”

The most common piece of bad advice in this industry: describe your projects and we’ll write them up as SR&ED.

Here’s the problem. What your team did and what qualifies as SR&ED are different questions. CRA doesn’t care about your project timeline or your sprint retrospectives. They care about where you hit genuine technological uncertainty, what hypothesis you formed, how you systematically tested it, and what the investigation revealed.

Most consultants produce project descriptions with SR&ED vocabulary stapled on top. When an RTA peels back the language and asks pointed questions, the underlying project description has no answers because it was never built to have them.

The Year-End Fire Drill

This is probably the single most expensive mistake, and it persists because it’s convenient for consultants, not because it serves you.

Ask an engineer what hypothesis they tested on a specific feature eight months ago. Ask them what parameters they varied. Ask what the results told them. They can’t answer with the specificity CRA requires. The work was real. The uncertainty was genuine. But nobody captured it when it mattered, and memory is not documentation.

Evidence degrades fast. By month eight, you’re reconstructing a claim from Jira tickets and vague recollections. CRA can tell the difference between contemporaneous documentation and a year-end reconstruction exercise. One supports a claim. The other undermines it.

Copy-Paste From Last Year

Claims that look substantially similar across years raise an obvious question: is this genuinely new work, or the same ongoing engineering repackaged with new dates?

Every claim year has its own technical context, its own uncertainties. If the writeup doesn’t reflect that, you’ve answered CRA’s question for them.

The “Claim Big, Negotiate Down” Strategy

Cast a wide net, claim aggressively, settle somewhere in the middle. The logic sounds reasonable until you realize the environment changed.

Overclaiming in 2026 doesn’t result in a polite reduction. It triggers deeper scrutiny of the entire claim. It can flag your company for more frequent reviews in subsequent years. And it erodes credibility with RTAs who have long memories and full access to your claim history.

This strategy treats CRA as an adversary to be gamed. CRA has adapted.

The Real Cost

A denied claim is the obvious damage. But it’s not the expensive part.

The expensive part is the legitimate qualifying work your team does every year that never gets captured because your advisor doesn’t know how to identify it. Software companies are particularly exposed here. The line between routine development and SR&ED-eligible work is technical and specific, and most advisors draw it in the wrong place.

Then there’s the credibility account you’re building with CRA. Every claim that doesn’t hold up is a withdrawal. Every year-end reconstruction that looks like every other reconstruction is a withdrawal. RTAs remember patterns across years. Your claim history is a relationship, whether you manage it that way or not.

And most companies have no way to evaluate whether their advisor is any good. If the claim gets approved, the consultant looks brilliant. If it gets denied, they blame CRA. If it was never reviewed at all, everyone assumes things went fine.

CRA only reviews roughly 10% of claims [2]. An unreviewed approval tells you nothing about whether your process would survive actual scrutiny.

The advisory industry has little reason to fix this. Contingency-fee models reward volume. Template processes keep costs low. And because most clients don’t understand the program well enough to ask hard questions, the cycle continues.

What Billions in Credits Buy

The SR&ED program returns more than $3 billion to Canadian companies every year [2]. It ranks among the most generous R&D incentive programs in the G7 [3]. The gap between what companies could legitimately claim and what they actually receive is real, and it’s mostly a function of how claims are prepared, not whether the work qualifies.

Companies doing genuine technical work are either leaving money uncaptured or building claims that can’t survive serious review. Often both.

The program has changed. CRA has changed. The question worth asking is whether your process has kept up.

If you’re not sure, take the readiness check and find out.

References

[1] Government of Canada, “Budget 2012: Scientific Research and Experimental Development Program.” [Online]. Available: https://www.canada.ca/en/revenue-agency/programs/about-canada-revenue-agency-cra/federal-government-budgets/archived-budget-2012/archived-budget-2012-scientific-research-experimental-development-program.html

[2] Canada Revenue Agency, “SR&ED Tax Incentive Program: Annual Program Statistics.” [Online]. Available: https://www.canada.ca/en/revenue-agency/services/scientific-research-experimental-development-tax-incentive-program/annual-program-statistics.html

[3] OECD, “R&D Tax Incentive Indicators,” INNOTAX Portal, 2024. [Online]. Available: https://stip.oecd.org/innotax/