The Way the Problem Is Usually Framed
Industrial digital twin programmes typically present their central challenge as model accuracy: how closely does the virtual model represent the physical system? This framing leads to investment in higher-fidelity simulation platforms, more sensor points, faster data acquisition rates, and more sophisticated physics models. These investments are often made before the basic operational question is answered: what decision does this twin support?
The model accuracy problem is real but secondary. A digital twin with 94% model accuracy that informs a decision made every four hours has more operational value than a 99.8% accurate model that no one queries because the interface requires a specialist to operate.
Where Programmes Stall
Across fourteen industrial digital twin implementations observed between 2023 and 2026, stalls occurred at three points. The first is the data integration layer — where the expected sensor data quality from installed instrumentation does not support the model update frequency specified in the programme design. The second is the decision workflow gap — where the twin's output is not connected to any operational procedure, so it becomes a monitoring tool rather than a decision support tool. The third is the maintenance responsibility gap — where no one owns the model update process when physical changes are made to the asset.
The Data Integration Layer
Digital twin programmes in brownfield industrial environments inherit the measurement infrastructure that existed before the twin was conceived. That infrastructure was designed for control and alarm purposes — not for the higher-frequency, higher-accuracy data acquisition that physics-based digital twin models require.
The practical consequence is that many industrial digital twins run in a degraded state from commissioning: some model parameters are estimated rather than measured because the required sensors are not installed, some data feeds are discontinuous because legacy communication protocols do not support the required polling rates, and some physical states are inferred from proxy measurements because direct measurement is not feasible with existing instrumentation.
None of this is fatal to operational usefulness — but it must be explicitly modelled as uncertainty in the twin's outputs. When it is not, the twin produces outputs that appear precise but are not, which erodes operational trust when discrepancies are discovered.
The Decision Workflow Gap
The second stall point is more fundamental. A digital twin is a decision support tool. If the operational team does not have a defined decision that the twin supports — a specific question it answers, a specific operational choice it informs — then the twin becomes a monitoring dashboard, which is a different and significantly less valuable thing.
The decision workflow gap occurs because digital twin programmes are typically led by engineering teams who design the model, not by operational teams who make decisions. The handoff between the engineering design phase and operational use requires explicit definition of use cases, decision criteria, and the procedure changes needed to integrate twin outputs into operational workflows. This handoff is rarely managed as a project deliverable.
The Model Maintenance Problem
Physical industrial assets change. Equipment is replaced, operating parameters are modified, physical configurations are altered. A digital twin that is not updated to reflect these changes degrades in accuracy over time and becomes a source of misinformation rather than operational intelligence.
Model maintenance requires a defined change management process: a trigger for updating the twin (a modification to the physical asset), a defined responsible party, a validation procedure, and documentation of the change. Programmes that do not establish this process during commissioning discover within 12-18 months that the twin and the physical asset have diverged to the point where the twin's outputs are no longer trusted by operational teams.
A More Useful Framing
The programmes that achieve sustained operational value share a common design approach: they start from a specific operational decision, work backwards to define the model requirements needed to support that decision, and then design the data acquisition infrastructure required to keep the model current. This produces a smaller, less technically impressive twin than the full-system physics model — but one that is actually used by the operational team, every day, to make decisions it would otherwise make with less information.