SR&ED Tax Credits in Canada: What Qualifies, What It Pays and How to Claim
SR&ED is the largest federal support programme for research and development in Canada, and for a Canadian-controlled private corporation it is one of the very few tax provisions that produces a cheque rather than a reduction. That makes it worth understanding properly, including the parts that get claims reduced.
Claims are rarely denied on arithmetic. They are denied because the work as described does not meet the definition of SR&ED, or because nothing written at the time supports what the claim says happened.
What actually qualifies
The test is not whether the work was hard, expensive, novel to your company, or commercially valuable. The CRA’s guidance on eligible work asks three questions:
Why was the work done? There must be a scientific or technological uncertainty: it must have been unknown whether a given result or objective could be achieved, because the available scientific or technological knowledge was insufficient. Not unknown to you. Insufficient in the field.
How was the work done? It must be a systematic investigation or search in a field of science or technology by experiment or analysis. That means defining the problem, forming a hypothesis, testing it, and drawing conclusions from the results. Trial and error with no recorded hypothesis is not a systematic investigation, whatever it cost.
What was the result? The purpose must be a scientific or technological advancement: the generation of new knowledge that moves the understanding of the science or technology forward, whether or not the project succeeded. Failed projects qualify. Successful projects that advanced nothing technological do not.
The commonest disqualifier in Ottawa’s software and hardware community is routine engineering: applying known techniques competently to a new problem. Integrating two documented APIs is not SR&ED, and neither is adopting an AI model, which is the most expensive current version of this misunderstanding and the reason I worked through it in hosted against open-weight AI. Building a scheduler whose required throughput cannot be achieved with any documented approach, and recording the attempts, may well be.
What the credit is worth
| Claimant | Rate | Refundable? |
|---|---|---|
| CCPC, up to the expenditure limit | 35% enhanced | Yes, fully refundable on current expenditures |
| CCPC, above the expenditure limit | 15% basic | A qualifying corporation can obtain a 40% refund of the credit |
| Other corporations | 15% basic | Generally non-refundable, applied against tax payable |
| Individuals and trusts | 15% basic | 40% refundable |
| Eligible Canadian public corporations | Enhanced rate now available | Per the recent legislative changes |
The expenditure limit on which the enhanced 35% rate is earned, and the taxable capital range over which it phases out, both changed for tax years beginning after 15 December 2024, and changed again in the most recent budget cycle. Do not rely on a figure quoted anywhere, including here. Read the current numbers on the CRA’s investment tax credit page and the SR&ED news and updates page before you model anything.
Two other recent changes matter. Eligibility for the enhanced rate was extended to certain Canadian public corporations, and capital expenditures became eligible again for property acquired after 15 December 2024, having been excluded for over a decade. If you shelved a capital-heavy R&D plan because the equipment did not qualify, that reasoning is out of date.
Associated corporations share one expenditure limit. This is where groups lose money without noticing: three related companies each assuming they have a full limit will discover they have a third each. The association rules are set out in associated corporations, and they are the same rules that carve up the small business deduction.
Provincial credits stack on top of the federal ones. Ontario operates its own R&D credits, and their interaction with the federal claim (provincial assistance reduces the federal pool) needs to be modelled together rather than separately.
What expenditures you can claim
Salary and wages of employees directly engaged in the work, materials consumed or transformed, and contract payments are the core. Two rules shape the number more than people expect.
Arm’s-length SR&ED contracts count at 80%. Where you contract another party to perform SR&ED on your behalf, 80% of the expenditure is a qualified expenditure for investment tax credit purposes. Budget the credit on 80%, not 100%.
Government assistance reduces the pool. Grants, forgivable loans and other assistance reduce qualified expenditures dollar for dollar on the project they relate to, so the credit is calculated on what is left. A grant and a credit are not additive, and a claim built without netting assistance is a claim that will be reduced.
Then choose a method for overhead:
| Proxy method | Traditional method | |
|---|---|---|
| How overhead is claimed | A notional prescribed proxy amount of 55% of the salaries of employees directly engaged | Actual overhead, itemised and traced to the work |
| Record burden | Low | High: every allocation must be supported |
| Best when | Overhead is modest relative to salaries | Facilities, utilities and support costs are large |
| Chosen | Year by year, on the T661 | Year by year |
| Can you change your mind? | No. The choice is irrevocable for the year once filed | No |
The CRA compares the two in its Traditional and Proxy Methods Policy. Most small claimants use the proxy method because it is defensible with far less evidence. Run both once before deciding by habit.
The deadline is hard
Your SR&ED reporting deadline is twelve months after the filing due date of the income tax return for the year. For a corporation that works out to eighteen months after the end of the tax year; for individuals it is roughly seventeen and a half months. The rule is in the SR&ED Filing Requirements Policy.
Miss it and the claim is gone. This is not a deadline that taxpayer relief or a late-filed amendment fixes, and it is the most common way a genuinely eligible claim is lost. File Form T661 with the return where you can, rather than treating the extra twelve months as planning room.
Is this project claimable?
│
├─ Was there a point where you did not know whether the objective
│ could be achieved with existing knowledge in the field?
│ └─ No ──► Routine engineering or development. Not SR&ED.
│
├─ Did you form and test hypotheses, and record the results?
│ ├─ No ───► The work may have been eligible; the evidence is not.
│ │ Fix the process now for the current year.
│ └─ Yes
│ │
│ ├─ Were the costs incurred in Canada, by you or on your behalf?
│ │ └─ No ──► Foreign-performed work is largely outside the
│ │ programme. Check before assuming.
│ │
│ ├─ Did grants or other assistance fund it?
│ │ └─ Yes ─► Net the assistance off the pool first.
│ │
│ └─ Are you still inside 18 months of the year end?
│ ├─ Yes ─► Claimable. Prepare T661.
│ └─ No ──► The reporting deadline has passed for that year.
│ Focus on the current year.
Documentation, which is where claims are actually won
The CRA’s research and technology advisors assess the work, not the write-up. A persuasive claim looks like a record of an investigation that happened, not a narrative composed eighteen months later.
What holds up: dated design notes and hypotheses, test plans and their results including the failures, version control history and commit messages that reference the problem being solved, timesheets or a defensible allocation of each person’s hours to each project, records of prototypes and scrapped materials, and minutes where technical direction changed and why.
What does not hold up: a marketing description of the finished product, an after-the-fact summary written by a consultant who was not there, and a whole development team’s payroll allocated to SR&ED because the team worked on the product.
Where claims are reduced
The four patterns I see repeatedly:
- Describing the product instead of the uncertainty. The claim explains what was built and why it is commercially valuable. The advisor is looking for what was not knowable and how it was resolved.
- Claiming the whole team. Support, QA, documentation, project management and commercial testing are not automatically directly engaged. Some support work qualifies; most administration does not.
- Ignoring the 80% rule and assistance netting, then budgeting a refund that never arrives.
- Structural surprises: losing CCPC status through a foreign investor, or discovering that associated companies share one limit. These are worth checking before a financing round, alongside the other consequences set out in incorporating in Ontario.
Record retention runs six years from the end of the tax year the records relate to, and for SR&ED the technical records matter as much as the financial ones. See record retention in Canada.
A reasonable way to run it
Decide at the start of a project, not the end, whether it is a candidate. Give it a project code and have people book time to it. Keep a running technical log, even a short one. Review the portfolio at your fiscal year end while people still remember what happened. Nothing in that list is burdensome, and it converts a reconstructed claim into a documented one.
If you are unsure whether your development work meets the eligibility test, or you want a claim reviewed before the eighteen-month window closes, send me the project and the year end and we can assess what is genuinely claimable.
Sources & references
Written by