SR&ED for Canadian Corporations: Turn Innovation Costs Into a 35% Refund

Most Canadian companies that qualify for SR&ED never claim it. They picture people in lab coats, or they assume the paperwork costs more than the cheque. Both assumptions are wrong, and both leave real money with the government.

SR&ED stands for Scientific Research and Experimental Development. It is the largest source of federal support for business innovation in Canada. If your company is a Canadian-controlled private corporation, or CCPC, it can get 35% of its eligible innovation costs back as a refundable tax credit.

Refundable is the word that carries the weight. The government pays you even in a year you owe no tax and turned no profit. For a team spending its way through a hard technical problem, that is a cheque, not a deduction.

The rate is not the hard part. The hard part is that a claim is only as strong as the story you can tell about the work, and that story has a specific shape.

What counts as SR&ED

The program does not fund good ideas or new products. It funds the resolution of technological uncertainty. That means a competent engineer, holding the standard tools of the field, could not tell you in advance whether something would work or how to build it.

That rules a lot of things out. Wiring together documented APIs is not eligible, however useful the result. Neither is a redesign, a migration, or a feature that was hard only because it was large. What qualifies is the part where you hit a wall the field had no known answer for, and you ran experiments to get past it.

Every eligible project tells the same four-part story. Capture the four parts as the work happens and the claim almost assembles itself. Reconstruct them a year later from memory and you lose most of what you were owed.

1. The baseline

Start with what was already known. Before you began, what could the field do, and what did your own systems already do? This is the state of the art plus your own starting point. It exists to prove one thing: the answer you needed was not already sitting in a manual or a forum post.

A good baseline is specific. "Existing open-source rerankers topped out around 60ms per query on our hardware, and none handled our document sizes without truncation" is a baseline. "We wanted better search" is not.

2. The innovation step

Name the uncertainty. What did you set out to do that you could not confirm was possible? This is your technological hypothesis: the gap between the baseline and where you needed to be, stated as a question the field had not answered for you.

The Canada Revenue Agency, which reviews claims, applies one test. Would a knowledgeable person in your field have had to experiment to find out? If the path was obvious to a competent practitioner, it is engineering. If it forced you to try, measure, and adjust, you are in scope.

3. The documented failures

Companies skip this part, and it is the part that carries the claim. You prove SR&ED with the record of systematic investigation: the approaches you tried, the ones that failed, what each result told you, and what you changed next. A failure is not a hole in the file. It is your evidence that the uncertainty was real.

Keep the record as you go. Dated commits, experiment logs, benchmark runs, the Slack thread where you ruled out an approach and said why. A reviewer wants to see a loop running across the project: hypothesis, test, result, revised hypothesis. A tidy summary written afterward that jumps straight from problem to solution reads as *weaker*, because it suggests there was never any uncertainty to resolve.

4. The outcome

Close the loop. What did you learn, and did the uncertainty get resolved? The outcome does not have to be a success. A project that missed its target still qualifies if the work advanced your understanding. You now know something about your systems that you did not know before. Sometimes that something is "this approach cannot be made to work," and it still counts.

Tie the outcome back to the baseline and the innovation step. The four parts then form one argument: what was known, what could not be known without trying, how you tried, and what you found.

The money

For a CCPC, the federal credit is 35% of your qualified SR&ED expenditures, refundable, on the first $3 million of eligible spend each year. Qualified expenditures are not your raw costs. The program adds a proxy overhead of 55% on top of eligible salaries, counts 80% of arm's-length contractor payments, and includes materials consumed in the work. So a dollar of SR&ED salary is worth more than a dollar once the calculation runs.

Then the provinces stack on top. In Ontario, a CCPC can add the Ontario Innovation Tax Credit at 8%, also refundable, and the Ontario Research and Development Tax Credit at 3.5%. Combined, a well-documented Ontario claim can return well over 40 cents on every dollar of eligible salary.

Here is what that looks like on your own numbers:

Treat the result as a planning figure. A real filing nets the provincial credits against the federal base, and it phases the 35% rate down for larger or non-CCPC companies. The precise number is your accountant's to certify. The order of magnitude, though, is usually a surprise in the right direction.

Where XY Space fits

Good SR&ED claims are rare for one reason. The four-part story has to be built into how the work happens, and most teams only think about it at year end, once the record has gone cold.

We build custom AI systems, and that work is often SR&ED by its nature. The hard parts are exactly the technological uncertainties the program exists to fund. So we run our engagements to leave the trail behind them. We capture the baseline before we start, name the uncertainty in the plan, log the failed experiments instead of deleting them, and tie the outcome back to the question. The credit becomes a byproduct of building the thing properly, not a spring scramble.

If your company is spending real money to get past hard technical problems, some of that spend is very likely eligible right now. Talk to XY Space about the work you are planning, and we will help you structure it so the innovation and the refund come out of the same project.