What counts as eligible SR&ED for software companies? A founder's read.

The Scientific Research and Experimental Development (SR&ED)program is the largest source of research and development funding in Canada, and software companies are among the biggest claimants. For a qualifying Canadian- controlled private corporation, it can return 35 percent of eligible spending as a refundable credit, which for an early-stage company is real cash back, not just a tax reduction. The hard part is knowing what counts. Here is how to think about your own work.

What SR&ED is, and what it pays

SR&ED rewards work that advances technology by resolving genuine uncertainty, not work that simply applies what is already known. For a qualifying CCPC, the enhanced credit is 35 percent and refundable, meaning you can receive it as cash even with no tax payable. As of the 2026 changes, that enhanced rate applies to up to CAD 6 million of qualifying expenditure a year. Eligible costs are mostly the wages of the people doing the work, plus certain materials and, now, cloud-computing costs.

The test that actually decides eligibility

Strip away the jargon and SR&ED comes down to three questions:

1. Was there technological uncertainty? Could a competent professional in your field have known the outcome in advance, using standard practice? If yes, it is routine work. If the answer was genuinely unknown, you may have SR&ED.
2. Did you run a systematic investigation? Did you form a hypothesis, build, test, measure, and iterate? SR&ED is experimentation, not trial and error.
3. Were you seeking a technological advancement? Were you trying to learn something new about how to build the thing, even if the project ultimately failed?

The work qualifies on the strength of the uncertainty and the method, not on whether the product succeeded or made money.

What qualifies in software

The eligible part of a software project is usually narrower than founders hope, but real:

• Developing a novel algorithm where the approach was not known in advance.
• Solving genuine scaling or performance problems that standard architecture could not handle.
• Integrating systems in a way that required real experimentation, not just configuration.
• Building capabilities where existing tools and documented methods did not get you there.

The pattern: you hit a wall that competent practice could not get past, and you had to experiment to find out whether and how it could be solved.

What does not

The common disqualifiers, and the ones the CRA looks for:

• Routine development: building features with known methods, however hard the work.
• Standard configuration, integration of well-documented APIs, or normal debugging.
• Work whose only uncertainty was about the market or the schedule, not the technology.
• User interface and experience work, unless it required genuine technical research.

Difficult is not the same as uncertain. A big, demanding build using known techniques is not SR&ED. A smaller piece where the outcome was genuinely unknown can be.

What the 2026 changes added

The 2026 reforms (Bill C-15, Royal Assent March 26, 2026) made the program meaningfully better for software companies:

These are recent and the claim forms reflecting them arrived in 2026, so confirm the current details against canada.ca before you file.

The mistake that costs founders the most

The biggest money left behind is not from claiming too little. It is from books that cannot support a claim. If engineering wages and cloud costs sit in generic buckets with no record of which projects and which people did the eligible work, you either underclaim or spend a fortune reconstructing it at year-end. Track the eligible work as you go: which staff, which projects, and the technical uncertainty you were chasing. A claim built from contemporaneous records survives review. One reconstructed in hindsight rarely does, which is the subject of our companion guide.