Deterministic, not predictive: why shop limits shouldn't be guessed
More of the parameters in a product are now set by software. A configurator lets a customer pick a width. A parametric model generates every variant from a few inputs. AI tools can propose sizes and options faster than anyone can check them by hand.
None of these tools knows your router unless someone has told it. A design tool can suggest a panel width without any record of your bed size, your sheet stock or your smallest cutter. Those limits live in the shop.
So as software sets more of the parameters, a question becomes more important: where do the limits on those parameters come from?
Two ways to answer
Predict them. Estimate what the shop can probably make, from past jobs, typical values or a model of the machine. A design tool that assumes a standard sheet size is doing this. The answer is usually close, and it can cover cases nobody has written down.
Record them. Write down what the shop can actually make: this bed, this sheet, this thickness, this cutter. Apply those values directly. The answer covers only what has been recorded, and for those cases it’s exact.
For deciding what reaches a CNC router, the second approach has properties the first can’t offer.
What deterministic means here
A deterministic system gives the same answer to the same question every time. Choose 18 mm or ¾-inch ply, a 2440 × 1220 mm or 4 × 8 ft sheet and a 6 mm or ¼-inch cutter, and you get the same limits today, tomorrow and on the next job.
That matters in a shop for three reasons.
You can check it. If a size is flagged, you can ask why and get a definite answer: it breaks the recorded bed size, or the recorded sheet width. A prediction can be explained in general terms, but not traced to a single fact you can verify.
You can fix it. If a limit is wrong, you change the record. Every product that uses it follows. There’s no retraining and no waiting to see whether the estimate improves.
You can rely on it. The same choices never return different limits on different days. A configuration that was within limits last month still is, unless someone changed a recorded limit, and then you know which one.
What it doesn’t do
Recording limits has an obvious gap: it knows only what has been recorded. A limit nobody wrote down isn’t checked. A new material that hasn’t been measured, or a constraint the shop has always handled by eye, can still reach the machine.
A predictive approach might catch some of those cases. It might also miss them, or flag problems that don’t exist, and you can’t always tell which.
The practical answer is to be explicit about the gap rather than hide it. Recorded limits are checked exactly. Everything else still gets checked, by your team, before cutting. That’s the same check you do now, with fewer out-of-limit parts reaching it.
Where the software sits
The distinction is less about which tool proposes a size and more about what it’s checked against.
A person, a configurator or an AI tool can all set parameters. If each design is checked against the shop’s recorded limits, it doesn’t matter which of them chose it: anything outside those limits is flagged, with the reason. The tool doesn’t need to know your router. The records do.
This only applies to designs from connected products. Geometry drawn another way isn’t checked.
Where Datafurn fits
Datafurn is deterministic, not predictive. Your machine, material and tooling limits, held as data, and each variant checked against them at the design stage. The same design always gets the same result, and every flag traces back to a recorded limit. Datafurn doesn’t use AI to decide them.
We’re recruiting 3–5 manufacturers cutting sheet materials on 3-axis flatbed routers for our Stage 1 pilot. If you use Grasshopper, we connect Datafurn to one of your existing definitions. If you don’t, we build a model for one product family. Either way, we do the setup.