SR&ED Claims, written by an engineer.

SR&ED is a federal program that pays Canadian companies back part of what they spent on experimental work, in cash, even in a year with no profit. I tell you whether your company qualifies, for free, before you pay anything.

Is this SR&ED?

Three projects, each written twice. One version is a Claim and one is not. Pick before you read the answer.

Software. Illustrative, not a Client.

The query that stopped scaling

A Toronto SaaS company. Its customer dashboards slowed to 40 seconds once the largest account passed 50 million events.

Which version is SR&ED?

Version ANot SR&ED

The team added indexes, moved to a bigger database instance and cached the common queries, following the vendor documentation. It took six hard weeks, but at every step a competent engineer knew what would work.

Version BSR&ED

Every customer defines its own event fields, and they change daily. The published fixes (precomputed views, a columnar store) assume fields that stay put, and both broke on the first schema change. Nobody knew whether sub-second queries were possible under those conditions, so the team tested a rollup keyed on schema version, measured it, and changed course twice.

Decided by: Technological uncertainty. Hard, slow and expensive is not the test. The question is whether a competent professional could have known the answer from published knowledge. For indexes and caching, they could.

Software, machine learning. Illustrative, not a Client.

The model that would not generalize

A Waterloo startup flags defects in photos taken on customers' phones. Accuracy collapsed on phone models it had never seen.

Which version is SR&ED?

Version ASR&ED

The team wrote a hypothesis before each change, held out whole phone models as test sets, logged every run with its result, and dropped each idea the data did not support. Three hypotheses failed before one held, and each failure is on record with its date.

Version BNot SR&ED

The team swapped models and settings until the numbers looked good again. Nobody wrote down what they expected from each change or what happened, so when the model worked, nobody could say why.

Decided by: Systematic investigation. The problem is the same in both. What differs is the method: hypothesis, test, measurement, adjustment, with a record made at the time. Trying things until something works, with no record of why, is weak.

Software, data. Illustrative, not a Client.

The pipeline built abroad

A Montreal fintech needs every transaction reconciled within 200 milliseconds, with events arriving out of order. The known designs trade speed for correctness, and it needs both.

Which version is SR&ED?

Version ASR&ED

The same experiments, run by the company's own engineers in Montreal, who designed each test, ran it and analyzed the results.

Version BNot SR&ED

The design is genuinely uncertain and the experiments are well recorded, but they were run by a contracted team outside Canada. The Montreal staff wrote the requirements and reviewed the results.

Decided by: Work performed in Canada. The experimental work has to be done in Canada. Work done abroad does not count, even when the problem is real and the records are good.

Five more, in hardware, manufacturing, food and agriculture

dates and notes are illustrative

SR&ED pays for the loop, not the result.

01Hypothesis02Test03Measure04Analyse05Adjustpass 1, Feb 9it did not hold.pass 2, Mar 3changed one variable.pass 3, Apr 21measured, recorded.T661direct deposit stub, illustrativeRECEIVER GENERAL FOR CANADAPAY[your company]refundable credit, one year$[amount]
  1. 01Hypothesis
  2. 02Testpass 1, Feb 9: it did not hold.
  3. 03Measurepass 2, Mar 3: changed one variable.
  4. 04Analysepass 3, Apr 21: measured, recorded.
  5. 05Adjust
  6. ↻and again, from 01
direct deposit stub, illustrativeRECEIVER GENERAL FOR CANADAPAY[your company]$[amount]refundable credit, one year

A hypothesis, a test, a measurement, an adjustment, again. If your team did that on a problem nobody had published an answer to, and kept a record, it probably counts.

How it works

  1. 1You fill in a short form.
  2. 2A 30 minute qualifying call.
  3. 3Your Verdict in writing within five business days. Free, and it may say no.
  4. 4If it says yes: engagement letter, $500 deposit, and I write the Narrative and the Fill sheet.
  5. 5Your own accountant files it with your tax return. I never touch your return.

The full procedure, and what I need from you

How long you have

A Claim rides on the tax return for the year the work was done, and it has to reach the CRA within 18 months after that year ends. Miss it and the year is gone.

Illustrative example

Every eligible dollar builds the pool. About 40 cents of it comes back.

Eligible costs, one fiscal year
Eligible salary$60,000
Overhead proxy (55% of salary)$33,000
Materials$4,000
Canadian contractors (80% of $15,000)$12,000
Pool$109,000

Salary, materials and contractors for the experimental work build the pool. For a Canadian-controlled private corporation in Ontario, roughly 40 cents of each pool dollar comes back as cash, even with no tax owing.

What it costs

SCOPEFEE
One project, up to 3 technical staff$3,000
Up to 3 projects, or up to 10 technical staff$5,500
LargerQuoted
Repeat year, same ClientHalf

Flat fee, priced on the work, never on the size of your Claim. The CRA asks every Claim to declare its billing arrangement (T661, line 950). Ours is code 4, flat fee.

If the CRA reviews a Claim I prepared, I handle the response. The first 10 hours are included.

If I start and conclude your Claim should not be filed, your deposit comes back and you owe nothing.

Year one full price, year two half, then a handover package so you can file on your own.

Deposit, payment terms and what is included

Who I turn away

If your expected credit is under about $15,000, no consultant is good value, including me. Use the CRA's freeself-assessment tool, or a $299 self-serve package. I will tell you this on the call too.

No repository or issue tracker? It can still work, but it takes longer and costs a band more.

How I work

For Clients with a code repository and an issue tracker, I rebuild the experimental record from dated commits and tickets, rather than from memory in a meeting. That needs read access to your code, so confidentiality is in the engagement letter before I see anything.

The engineer who did the R&D writes the Narrative.

I built the platform, wrote the Narrative and filed my own company's Claim. The CRA assessed it as filed without selecting it for review, and the refund arrived in September 2026.

SR&ED Lab, founderSigned and dated, September 2026

Ask for a Verdict

Five questions, one of them about what you built. Read the four tests on the next page first: if one of them fails, there is no Claim this year and you will save us both a call.