SR&ED Eligibility Criteria
Not all R&D qualifies for SR&ED tax credits. CRA applies three specific criteria, and understanding them precisely is the foundation of a defensible claim.
The three eligibility criteria
For work to qualify as SR&ED, it must satisfy all three of the following criteria simultaneously. Missing even one disqualifies the work.
1. Technological uncertainty
The work must address a technological uncertainty that could not be resolved through standard practice, routine engineering, or publicly available knowledge. The uncertainty must be about whether or how a particular technological result can be achieved: not about business outcomes, timelines, or costs.
2. Systematic investigation
The work must involve a systematic investigation through experiment or analysis. This means formulating hypotheses, designing and conducting tests, analyzing results, and drawing conclusions. Ad-hoc trial and error without structured analysis is not sufficient.
3. Technological advancement
The work must be undertaken to achieve a technological advancement: generating new knowledge or capabilities that extend the existing technological base. This includes knowledge gained from unsuccessful experiments.
CRA's five questions
CRA uses five questions to evaluate the eligibility of each SR&ED project. These same questions structure the project description in Part 2 of Form T661:
- What scientific or technological uncertainties did you attempt to overcome?Clearly state what was unknown or unachievable using existing knowledge. Avoid describing business requirements: focus on the technical barrier.
- What hypotheses did you formulate?What possible solutions or approaches did you propose to address the uncertainty? Even informal hypotheses count: "we believed that approach X might resolve the issue because of Y".
- What procedures or experiments did you adopt?Describe the systematic work performed to test your hypotheses. Include the methodology, tools used, and parameters tested. In software, this often means describing the code changes, prototypes, benchmarks, or architectural experiments.
- What progress did you achieve?What results came from the experiments? Include both positive results (it worked) and negative results (it didn't work, and here's what we learned). Quantitative results (performance metrics, accuracy numbers) are more convincing than qualitative statements.
- What conclusions did you draw?What new knowledge was gained? Was the uncertainty resolved? If not, what was learned that informs future investigation?
How SRED AI helps
SRED AI structures its T661 drafts around these five questions automatically. By analyzing your pull request history, it identifies the uncertainty (what problem was being solved), the investigation (what approaches were tried), and the conclusions (what the final implementation reveals about the outcome): giving you a first draft that speaks CRA's language.
Eligible vs. ineligible work
Understanding the boundary between eligible and ineligible work is critical for building an accurate claim.
Eligible (SR&ED):
- Attempting to develop a new technology, process, or product where the approach or outcome is uncertain
- Improving existing technology in ways that go beyond standard practice and require experimentation
- Work that results in failure: unsuccessful experiments still qualify if the investigation was systematic
- Developing tools or techniques necessary to support the core SR&ED activity
Ineligible:
- Routine engineering, design, or development using established techniques
- Market research, sales promotion, or quality control for commercial production
- Social science or humanities research
- Style changes or cosmetic modifications
- Prospecting, exploring, or drilling for natural resources
- Commercial production of a new or improved product
- Data collection or routine testing unless directly tied to SR&ED hypothesis testing
Supporting vs. SR&ED work
CRA distinguishes between directly undertaken SR&ED and support work. Both can be claimed, but they are treated differently:
Directly undertaken SR&ED is the core investigative work: the experiments, prototyping, and analysis aimed at resolving the technological uncertainty.
Support work includes activities that are commensurate with and directly in support of the SR&ED. This includes:
- Engineering activities necessary for SR&ED (designing test fixtures, building prototypes)
- Data collection required for the SR&ED investigation
- Testing directly related to the SR&ED objectives
- Computer programming necessary for the SR&ED (building tools to run experiments, not the end product itself unless the product is the SR&ED)
- Psychological research where it is directly required for the SR&ED
Support work must be proportional to the SR&ED it supports. CRA will scrutinize claims where support work significantly exceeds the core investigative work.
Common misconceptions about eligibility
- "It was hard, so it must be SR&ED". Difficulty alone does not make work eligible. Building a complex system using known techniques is skilled engineering, not SR&ED. The work must involve genuine technological uncertainty.
- "We used new technology, so it's SR&ED". Using a new framework, language, or tool is not SR&ED. Adopting existing technology (even if it's new to your team) is standard practice. SR&ED requires that you create or advance technology, not just adopt it.
- "Only PhDs do SR&ED". There is no education requirement. A junior developer solving a genuine technological uncertainty through systematic investigation is doing SR&ED. What matters is the nature of the work, not the credentials of the person.
- "Failed projects can't be claimed". They absolutely can. CRA explicitly recognizes that SR&ED includes work where the uncertainty was not resolved. The attempt and the knowledge gained are what matter, not the commercial outcome.
- "SR&ED only applies to new products". Improving existing products, processes, or systems can qualify if the improvement requires resolving technological uncertainties. Incremental R&D is eligible.
Software-specific eligibility considerations
Software development has unique eligibility challenges. CRA's guidance notes that software work qualifies when it involves:
- Developing new algorithms or data structures where existing approaches are insufficient for the specific problem constraints
- Solving systems-level challenges (performance, scalability, reliability) where standard architectural patterns fail to meet requirements
- Creating novel integrations between systems where no established approach exists and the interaction behaviour is unpredictable
- Advancing the state of practice in areas like machine learning, natural language processing, or computer vision for specific domain applications
Software work that does not typically qualify includes: using frameworks and libraries as intended, building applications with known architecture patterns, configuring or customizing existing software, writing tests for existing functionality, and UI development using standard design patterns.
How SRED AI helps
SRED AI helps you separate eligible work from routine development by analyzing the nature of your code changes. Pull requests that show experimentation (multiple approaches tried, performance comparisons, abandoned branches) are flagged as potential SR&ED candidates. This gives you a data-driven starting point for identifying eligible projects.
Stop reconstructing what your team built
SRED AI extracts technological uncertainties and experiments from your GitHub pull requests, giving you a T661-ready first draft in minutes instead of weeks.