Weekly SR&ED Reflections: Turn Team Updates into Claim-Ready Evidence

A 15-minute Friday note can turn sprint work into claim-ready evidence when it happens.

These “Weekly Reflections” capture the uncertainty, trials, and learning your team actually did while it’s still fresh in their minds.

That takes a lot of pressure off everyone when it comes time to submit R&D Tax Credit Claims like SR&ED and IRAP. 

Takeaways

• Contemporaneous > reconstructed. In Canada, the Canada Revenue Agency (CRA) looks for records created during the work, not cleanups at year-end.

• Pattern of proof beats any single artifact. Pair weekly notes with tickets, commits, test data, and timesheets. 

• Make it light and repeatable. 15 minutes every Friday, one owner, same template, auto-linked artifacts.

• Finance can trace costs. Weekly notes help reconcile hours, payroll, and project codes to T661.

Why weekly SR&ED reflections work

People forget fast. Once a problem is solved, it stops feeling “uncertain.” Weekly reflections freeze the uncertainty and the experiments that resolved it.

That’s precisely what CRA’s technical reviewers try to confirm, and it’s much easier to show when notes are timestamped and cross-linked to the work you already do.

How:

• Freeze uncertainty while it’s fresh.

• Show a systematic investigation (problem > test > result > next).

• Time-stamp and cross-link to the work you already do.

The 15-minute Friday ritual

Who: 

Tech Lead (owner), involved engineers, and a Finance/Ops observer when possible.

When: 

Fridays, the last 15 minutes of the sprint week.

*Where: 

Jira/Linear ticket + Confluence/Notion page (or your tool of choice).

* Google Forms also works well with a template that has multiple-choice questions with some open-ended text boxes.

How:

1. Open last week’s note. Duplicate the template below.

2. Capture uncertainty & trials. One-line bullets. No essays.

3. Link artifacts. Tickets, PRs, benchmark outputs, diagrams, logs.

4. Note outcomes and next hypotheses. “We tried X to solve Y; result was Z; next we’ll test A.”

5. Tag it. Tag type (e.g., “​​SR&ED”), project code, and cost center for traceability.

6. Export monthly. Roll up four weekly notes into a single “Monthly Log” and file with Finance. 

This cadence aligns with CRA’s guidance to keep technical and financial documentation that supports the claim and speeds review.

Copy-paste template (use weekly)

Week of: 

YYYY-MM-DD

Project / Code: 

{e.g., “Telemetry Pipeline v2 – PRJ-1427”}

Advancement target (short): 

What capability/limit are we pushing?

1. Problem / uncertainty observed

• {What didn’t work / unknown constraint}

• Tip: Think back on the past week. What were some of the things you weren't sure about and had to investigate?

2. Hypothesis & approach

• {What we tried and why}

• Tip: Looking at this week, which tickets have you worked on that were related to spikes or investigations (record notes in the tickets if you haven't done so).

3. Experiment(s) run

• {Method, config, environment; link datasets, scripts, notebooks}

• Tip: As part of working on the spikes and investigations, did you need to setup a novel solution? Something that makes you proud?

4. Result & interpretation

• {Pass/fail, metrics, logs; what we learned}

• Tip: What did we learn from the spikes? Which succeeded and which failed? What were the learnings?

5. Next step

• {New hypothesis or alternate path}

• Tip: What would be the next steps that would move the project forward?

6. Artifacts (links)

• Tickets: {Jira/Linear}

• Code: {PR/commit SHA}

• Benchmarks: {dashboard/result files}

• Diagrams/Designs: {doc link}

• Tip: These should be part of the tickets. For example, Jira and Linear have an integration with GIT to show the PRs on the tickets.

7. Time & people

• {Names, hours; map to cost center / payroll code}

This structure mirrors what CRA reviewers look for (systematic investigation, uncertainty, and evidence) and helps Finance reconcile salary/wage allocations for T661.

Examples

Software (ML inference latency)

• Uncertainty: Model spikes to p95>180ms on GPU driver 555.xx in k8s.

• Hypothesis: Kernel launch overhead from mixed precision; try pinned memory + driver downgrade.

• Experiment: Run A/B on 20k requests; collect Nsight traces; PR #1842.

• Result: p95 182ms to 138ms; regression persists under burst load.

• Next: Test CUDA graph capture + batch size 2 with token-bucket.

• Artifacts: JIRA ML-421, PR #1842, Nsight run 2025-10-10.

• Time: 18h (Ana 10h, Raj 8h) to CC-R&D-ML.

Hardware (BLE beacon reliability)

• Uncertainty: Packet loss with metal obstructions at 10m.

• Hypothesis: Antenna detuning; test ground plane redesign + channel hop.

• Experiment: 5 layouts, 3 channels, 100 trials each; logs + photos.

• Result: Loss 11% to 3% on Layout C; battery draw +1.2%.

• Next: Validate in production enclosure.

• Artifacts: ENG-987 tasks, CAD v3.2, test logs, photos.

• Time: 22h (Mia 12h, Ken 10h) to CC-R&D-HW.

For Finance

How this maps to T661 and payroll:

• People & Time: Reconcile to salary or wages for SR&ED-eligible work. 

• Problem / Uncertainty: Supports the systematic investigation described in the T661 project narrative.

• Artifacts (tickets, PRs, metrics): Supporting documents that substantiate the claim during review.

• Retention: Keep records generally six years from the end of the last tax year they relate to.

Common pitfalls and easy fixes

Pitfall: 

Writing glossy “project summaries” at year-end.

Fix: 

Keep rough, timestamped weekly bullets with links. CRA values contemporaneous notes over polished rewrites.

Pitfall: 

Tickets describe features, not problems.

Fix: 

Frame tickets around uncertainty and tests; link PRs and metrics dashboards.

Pitfall: 

Timesheets don’t tie to the work.

Fix: 

Reference ticket IDs/PRs on time entries; Finance spot-checks weekly.

Pitfall: 

Using non-quantifiable language such as “fast” , “secure”, “reliable”.

Fix: 

Use specifics like “Loads within X seconds”, “Adheres to SOC2 <control name>”, “Expected uptime %: 99.99”.

Governance

• Owner: Tech Lead (rotating “documentation steward” each sprint).

• Review: 5 minutes in Monday stand-up to confirm last Friday’s note exists.

• Monthly packaging: Ops/Finance exports the four notes to a single “SR&ED Monthly Log,” attaches payroll summary, and files in your SR&ED folder.

Tools

Use what you already have.

• Jira/Linear for issue context

• GitHub/GitLab/Bitbucket for commits/PRs

• Confluence/Notion for the weekly note and monthly roll-up

• Dashboards (Grafana, Datadog, custom) for test/benchmark evidence

No new systems are required. The goal is to connect the evidence you are already creating.

Closing Thoughts

Weekly reflections can help make claiming R&D tax credits like SR&ED and IRAP less of pain than waiting until year end.

Fifteen minutes on Friday locks in uncertainty, tests, and proof. Finance can also trace hours to T661 without chasing people.

Start with one team this week, export a monthly log, and scale from there.

If you have any questions or comments, reach out any time.

Sources

1. Get ready to claim (SR&ED) (CRA’s eligibility + documentation reminders, 2023-10-19).

2. Claim Review Manual for Research & Technology Advisors (CRA’s 2025-01-29 & 2025-05-29 updates).

3. T4088: Guide to Form T661 (Appendix: documentation & evidence).

4. Salary or Wages Policy (how CRA looks at wages/time).

5. After you claim (CRA’s processing & retention).