Pilot Proposal · 2026

Know a cow is
getting sick — before
she shows it.

An early illness detection pilot built around your herd,
your data, and your existing equipment.

The Problem

By the time you see it,
the damage is done.

$350
average cost of a single mastitis case
30%
of cows affected at least once per year
3–5
days of detectable signals before visible symptoms

On a 300-cow operation, that's roughly 90 cases a year — around $31,500 walking out the door. Catch half of them a day earlier and you're talking real money back in your pocket.

The Approach

Cameras and data you
already have. No new gadgets.

What we use
Cameras in the barnWatch gait, posture, time at the feeder, body condition — 24/7, no animal contact
Your milking dataYield, conductivity, milking time — your AMS already collects this every day
Your health recordsTwo years of vet logs tells the algorithm what sick looks like on your specific herd
Why this works better
More signals, fewer sensorsCombine three data streams and the model gets confident without needing any wearable on the animal
Passive — zero extra labourCameras run overnight. No handling. No daily checks.
Gets smarter over timeEvery case — caught or missed — makes the model better for your herd specifically
What We're Proposing

A working pilot on
your herd. 90 days.

1
Connect to your existing milking data Your AMS already tracks yield, conductivity, and milking time. We tap into that — no new hardware.
2
Mount a few cameras in the barn Overhead or side-angle — we handle placement. Watch the whole herd, not just cows you're already worried about.
3
Train the model on your herd's history We use your health records to teach the algorithm what "before sick" looks like on your farm specifically.
4
Daily alert: which cows to check Simple list every morning. No dashboards. No new apps to learn.
The Deal

What you bring.
What you get back.

Your side
30–50 cowsEnough to train a real model. You decide which ones.
Health recordsExisting vet logs, treatment history — gives us a head start.
Access to your milking dataWhat the machine already collects. No extra work on your end.
Honest feedbackTell us when it's right. Tell us when it's wrong.
Our side
Cameras sourced and installedWe handle placement. No margin for us on hardware — cost only.
All the softwareData pipeline, model, alerting — we build it, you use it.
Your data stays yoursWe don't sell it. We don't share it. You can walk away with it.
SR&ED documentationWe handle the paperwork. You capture the tax credit.
The Money Side

The government will pay
for most of this.

43%
combined refund — 35% federal + 8% Ontario — on eligible R&D labour
$215K
back on $500K of eligible spend — real money, not a deduction

SR&ED (Scientific Research & Experimental Development) is a federal + Ontario tax credit for companies doing genuine R&D. Building a predictive illness algorithm trained on live farm data is textbook eligible. Prior failed claims almost always fail on paperwork, not concept — the work qualifies, the documentation didn't. We handle that from day one.

Built in Ontario

Camera-based. No wearables.
No animal handling.

Our approach
Cameras watch — cows don't noticeBarn-mounted. Passive. No boluses to swallow, no tags to clip, no ear readers to calibrate.
Data fusion beats single sensorsThree streams (video + milk data + health records) combined outperform any single device — published research backs this.
University of Guelph OVC involvementWorking with OVC researchers means the science is credible and the methodology holds up — including for your SR&ED claim.
Funding stack available
SR&ED — 43% backFederal + Ontario combined. We document it properly.
IRAP (NRC)Free advisory + potential labour funding for the software development side.
Mitacs Accelerate$7,500 from you unlocks $15,000 per OVC intern working on the project. Effectively 3:1 leverage on research labour.
Success Criteria

We agree on what "working"
means before we start.

Catch 60%+ of mastitis cases before clinical signs Alert fires at least 24 hours before the cow is visibly sick.
Less than 2 false alarms per cow per month Alert fatigue kills every monitoring system. We track this carefully.
Works without adding to your daily labour If it takes more than 5 minutes a day, we haven't solved the problem.
If it doesn't work, we say so clearly We're doing research. Null results are valid. No spin.
Next Step

One conversation.
Three questions.

1
What milking system are you on? Determines what data you already have and what we need to add.
2
Do you have historical health records? Even two years of treatment logs gives us a massive head start.
3
Which problem costs you the most sleep? Mastitis, metritis, fresh cow health — we start where the pain is worst.
← → to navigate