Digital model · Manual in · Manual out
You almost certainly have one already.
A digital model is a replica that does not know the real thing changed. Nothing flows into it on its own, and nothing flows out. That is not a criticism. It is a specification, and for a great many decisions it is the right one.
Recognise it
Every one of these is a digital model
Vendors will call several of them digital twins. They are not. The test is not how detailed or how clever the thing is. The test is whether data reaches it without a human carrying it.
A CAD file or floor plan
It describes the room. It has never once noticed that the room changed.
A Matterport or laser scan
A photograph you can walk through. Accurate on the day it was captured, and only that day.
An offline simulator
It can be extraordinarily sophisticated and still be a model. Intelligence is not the test.
A forecast in a spreadsheet or a web app
If you type the inputs, it is a model with a predictive layer. Useful. Not a twin.
A BIM asset register
A catalogue of what should be there, which quietly diverges from what is there.
What a model genuinely gives you
- Measure a space without visiting it, and share it with people who cannot.
- Test a protocol, a layout, or a robot path before anything is built or moved.
- Give a facility one documented description instead of six conflicting ones.
- Serve as the foundation every higher level is built on. Nothing is wasted.
What it structurally cannot do
- Notice anything. Move a bench, swap an instrument, or recalibrate a sensor, and the model is silently wrong until a person re-enters it.
- Tell you it has gone stale. A model does not know how old it is.
- Act on what it predicts, however good the prediction is.
- Be trusted for a decision that depends on the state of the facility right now.
The failure mode is silence.
A model does not break loudly. It drifts. Someone moves a freezer, someone swaps an objective lens, someone recalibrates a probe, and the model keeps answering questions with total confidence and quiet inaccuracy. Every model has a shelf life, and nothing in the model tells you when it expired.
What a good one costs
Levels 1 and 2 are digital models
We do not pretend otherwise. What separates ours from a 3D scan is that it is annotated, asset-aware, and built to be connected later, so the money is not spent twice.
Geometric replica
A geometry-accurate 3D twin of one room or system, built from a scan, CAD, or point cloud. Walk it, measure it, share it. No behavior and no live data yet, but a scientifically annotated foundation the higher levels build on.
Instrumented asset twin
Every instrument, material, and sensor mapped with specifications and metadata into a queryable asset twin. It still does not run experiments, but it becomes the single documented source of truth for the facility.
Closing the gap
A model becomes a shadow when you connect it
One change turns a model into a digital shadow: data starts flowing in on its own. The model stops being a snapshot and starts being a mirror. That is the next rung, and for most laboratories it is the last one they need.
Questions
Common questions
What is a digital model?
A digital representation of a physical object or facility with no automated data exchange in either direction. A person creates it, and a person updates it. CAD files, 3D scans, BIM models, and offline simulators are all digital models. Under the standard classification (Kritzinger et al., 2018), that manual data flow is what separates a model from a digital shadow or a digital twin.
Is a digital model useless, then?
No, and treating it that way is a common and expensive mistake. Most twin projects should start here. A model is enough whenever the decision you need to make does not depend on the live state of the facility: planning a build-out, testing a protocol before the bench, training staff, or documenting what you own. Buy the shadow when the staleness starts costing you.
Is my predictive web application a digital model?
If a person supplies its inputs, yes. Prediction is a separate axis from connectivity. A forecasting app that a researcher feeds by hand is a digital model with a predictive layer, no matter how good the model underneath is. It becomes a digital shadow the moment it starts ingesting live data on its own.
How do I turn a digital model into a digital twin?
In two steps, not one. First connect it: automated data flow from the physical facility into the model, which makes it a digital shadow. Then close the loop: let the model act on the facility, which makes it a digital twin. The second step is a governance decision as much as an engineering one, and plenty of labs correctly stop at the first.
