Operations
Your records already hold patterns. We build models that learn from them and give you a clear number, a fair range around it, and the reasons it moved. Then we put it inside the app where you already work.
See it running: the PIRX app we designed and built for pirx.ca.
The shift
You have years of records. Orders, jobs, visits, readings. Somewhere in there is the answer to a question you keep asking. What happens next? How much? How soon? A spreadsheet can add it up. It can’t learn from it.
Many AI tools hand you a number with no reason. That is hard to trust, and harder to explain to a customer. We build models that learn from your own data, show their working, and say how sure they are. When the data is thin, they say so.
The outcome
One clear answer to the question you ask most, worked out from your own records.
Every number comes with a range. When the data is thin or messy, the range gets wider instead of the number pretending.
The number is split into the parts that moved it, so anyone can see why it changed.
One odd day does not swing the result. The number moves when something real has changed.
Where it earns its keep
Turn a customer’s own history into a clear estimate they can plan around, the way PIRX gives a runner a Projected Time.
Spot the pattern in past jobs or orders, so you can plan people and stock before the busy weeks.
Flag the records that look off, with the reason, so a person checks those first.
The model is part of the product, not a report on the side. Your team and your customers see the number, the range and the reasons where they already work. We build the app around it too. See how we build apps
The operating model
From first conversation to a system that runs itself.
We map how the work gets done today, including the calls only a person should make, and agree on it before anything is built.
Built to run where your data lives and connected to the tools you already use, so nothing has to move.
Tested on your real work from last week before any customer sees it.
We launch with close monitoring, keep it running, and you can change the rules at any time. It’s yours to keep.
The decision
A spreadsheet, a data hire, and a model built into your app answer different questions.
| Off the shelf | Hire in-house | Abracadabra | |
|---|---|---|---|
| What you get | A total or a trend line someone has to read. | A report, when the person has time. | A number, a range and the reasons, inside the app you use. |
| When the data is thin | It shows a number anyway. | It depends on the person. | The range gets wider and the app says the data is thin. |
| Who can check it | Whoever wrote the formula. | The person who built the model. | Anyone. Each part of the number is shown. |
| When you don’t need us | If a simple average answers the question, keep the spreadsheet. | Hire a data person when you need new research every week, not one product. | Our advice: start simple. Machine learning earns its place when you have history to learn from and a question you ask every day. |
Named work
We designed and built the PIRX app and its machine learning. A runner connects the watch they already wear. PIRX gives a Projected Time for a race, a Supported Range around it, and the seconds each of five Structural Drivers adds: Aerobic Base, Threshold Density, Speed Exposure, Running Economy and Load Consistency. Personal models, including gradient boosting, learn each runner’s own patterns as their history grows. They are only used when they beat the app’s safe rules.
Running in production today.
Priced to the project. Tell us what you need and we’ll quote it.
Questions, answered
faq-title
Enough history to learn from. We look at what you have first. If it is too thin, we tell you, and we start with simple rules instead.
It works only from your data. When the data is thin or odd, the range gets wider and the app says so.
Yes. Each number is split into the parts that moved it, in plain words.
A new model is tested against safe rules first. If it does worse, or gives an odd answer, the app goes back to the rules.
You do. The data stays yours, and so does what we build on it.
Priced to the project. Tell us what you need and we’ll quote it.
Tell us the question you wish your data could answer, and we’ll show you what a model could do with it.