Digital model, digital shadow and digital twin describe three levels of connection between a physical system and its digital copy. The distinction comes from manufacturing research and is now widely used: it turns on which way data flows automatically, and it is a useful test for anything sold as a twin.
What is the difference between a digital twin and a digital shadow?
A digital shadow updates automatically from the physical system but does not act back on it; a digital twin does both. In a shadow, sensor data flows into the model and keeps it current, and people read the result and decide. In a twin, the model's output also flows back automatically, adjusting a setpoint, rescheduling a pump or triggering an action, within limits someone has approved.
- Digital model: no automatic data flow either way. Someone exports data in and reads results out. Most hydraulic models used for studies are here.
- Digital shadow: data flows automatically from the system to the model. The model mirrors the network in near real time, and people act on it.
- Digital twin: data flows automatically both ways. The model mirrors the network and its results change how the network is operated.
What this looks like on a water network
A calibrated EPANET model used for a master plan is a digital model. The same model fed every fifteen minutes by SCADA flows, pressures and tank levels, showing operators where the network is now and flagging departures from expected pressure, is a digital shadow. When that model also schedules pumps or adjusts pressure reducing valves directly, inside agreed limits, it is a digital twin.
Why most water "digital twins" are shadows
Most systems sold as digital twins in water stop at the shadow stage, and for good reason. Letting a model act on a live network needs reliable telemetry, a model the operators trust, fail safe limits and a clear answer to who is accountable when it acts wrongly. Many utilities rightly keep a person between the model and the valve. A good shadow is a valuable system; it should simply be called what it is, so the budget and the expectations match.
Which one do you need?
Start from the decision. If the decision is taken once, a digital model. If people take it repeatedly and need to see the network as it is now, a shadow. If the decision is frequent, well understood and safe to automate within limits, such as pump scheduling against tariffs, a twin. Moving a level up adds cost in data quality and governance, not only in software.
The model at the centre is the same at every level, which is why it is worth calibrating properly first. The step from model to twin is set out in digital twin vs simulation model, and where an AI agent fits on top in AI agents vs digital twins.
Smart Bhujal builds the calibrated models these systems depend on and the operational layer around them, delivered as a water decision support system that is honest about which of the three it is.
