Part 1 · SR&ED basics
What counts (and what doesn't)
The CRA looks for three things in the work. If all three are there, it can count, even when the experiment failed.
The three tests
- Technological uncertainty. You could not know how to reach the goal (or whether you could) using what is already publicly known or what your team already knew.
- Systematic investigation. You worked through it in a planned way: an idea to test, experiments or trials, and results you measured.
- Technological advancement. You learned something new about the technology, even if the answer was "this does not work".
Failure still counts
An experiment that proves an approach will not work is still an advancement. Keep a note of it. Failed attempts are often the best evidence in a claim.
Software examples
Usually counts
- A routing engine that must re-plan 10,000 stops in under a second when no known method is fast enough.
- A model that must hit an accuracy target nobody has published for your kind of data.
- Making two systems sync without data loss when the usual approaches fail at your scale.
Usually does not count on its own
- Routine bug fixes and maintenance.
- UI design, styling and content changes.
- Setting up a known library, framework or cloud service.
- Market research, sales work or quality control.
Routine work can still count when it is directly needed to run an experiment, like building a test harness to measure the new approach.
How the app checks this
A Future assessment reads your repo and project description, then ranks the SR&ED angles it finds. Each angle says which test it meets and why. A Past assessment reads the commits in a date range and flags the work that plausibly qualifies.
See an assessment
Walk through a sample project that already has one.
Sources
This guide helps you prepare. It is not tax advice. Have your SR&ED preparer review your claim before you file.