As digital twins become more detailed and persuasive, Satish Swaroop, digital twin and energy intelligence practitioner, asks a simple question: can the people using them tell what has actually been measured, what has been inferred, and what is still unknown?
Digital twins are becoming easier to admire. A building can now be rendered floor by floor, overlaid with energy data, coloured by temperature, animated by plant status and linked to forecasts. The result can look authoritative. Sometimes it is. Sometimes the picture is more certain than the evidence underneath it.
That distinction matters because buildings rarely arrive with perfect data. A relatively modern office may have good main-meter data, useful BMS points and patchy sub-metering. An older hotel may have years of utility bills, some plant measurements and very little that can be trusted at room level. Yet both can be represented beautifully on screen.
Imagine a floor where the electricity meter tells us the total consumption accurately, but there is no dependable meter for each room or zone. It is perfectly possible to allocate that total across the floor and create a heat map. The map may even be useful. The question is whether the user can see that the room values were estimated rather than measured.
If not, a visual aid quietly becomes a statement of fact.
Not all numbers carry the same weight
One practical discipline is to distinguish between different kinds of evidence. Some information is measured directly from a meter, sensor or BMS point. Some is derived from known inputs. Some is modelled because direct measurement is not available. Other information may be representative, included to help explain a building or scenario rather than to describe a live condition. And sometimes the honest answer is simply that the information is unavailable.
These distinctions sound obvious when written down, but they are easy to lose in a dashboard. A value shown as 17.42 kWh feels precise. If it came from a calibrated meter, that precision may be justified. If it came from an allocation based on floor area and assumed occupancy, the calculation can still be useful, but the extra decimal places do not make the underlying evidence stronger.
Precision and certainty are not the same thing.
Why this matters to the person running the building
The issue is not academic. Facilities teams make decisions under time pressure. An energy manager who sees an unusual plant load may decide to investigate immediately. A sustainability lead may use a trend to prioritise a retrofit. A hotel engineer may want to know whether one zone is genuinely consuming more energy than another. The confidence placed in the information changes the decision that follows.
A measured anomaly and a modelled anomaly do not carry the same evidential weight. The first may justify an operational response. The second may justify a check, a temporary hypothesis or a request for better evidence. Both can belong in a digital twin, but they should not be presented as if they are interchangeable.
This is also where visual design becomes part of engineering judgement. A simple label, legend, tooltip or confidence note can tell the user a great deal. It can show where the twin is grounded in direct evidence, where it is calculating from known inputs, and where it is filling a gap with a model. That transparency is often more useful than another layer of graphical polish.
A digital twin does not need to pretend it knows everything
There is an understandable temptation to make a digital twin look complete. Blank areas feel unfinished. An unavailable value can seem like a product weakness. In practice, the opposite can be true.
A system that says 'we do not know this yet' is giving the operator valuable information. It identifies a gap that may deserve better metering, a data-quality check or simply a more cautious decision. It also prevents a modelled assumption from acquiring the authority of a measurement merely because it appears in the same interface.
The most useful digital twins are not necessarily those with the greatest quantity of data. They are those that make the relationship between data, assumption and decision easy to understand.
That principle has become increasingly important in my own work on operational energy digital twins, including the development of ENERGE TWIN. Building data is rarely uniform, and useful systems have to work with that reality rather than hide it. In practice, the value often comes from helping a user understand what evidence supports a conclusion before asking them to act on it.
That does not require exposing technical machinery to every user. A facilities manager should not need to study a data lineage diagram before checking a floor. The responsibility sits with the design of the system: make strong evidence clear, make assumptions visible when they matter, and do not turn missing information into false certainty.
The next measure of digital twin maturity
For several years, much of the conversation around digital twins has focused on how much of the physical world can be reproduced digitally. That remains important, but the next stage of maturity may be judged differently.
Can the twin explain the strength of the evidence behind what it shows? Can a user tell the difference between an observation and an inference? Does the system make uncertainty easier to manage rather than easier to overlook?
A digital twin should be able to support decisions without creating the impression that every part of the building is equally well understood. In some places it may have excellent evidence. In others it may have a reasonable model. Elsewhere it may have almost nothing at all.
Knowing those boundaries does not make the twin less intelligent. It makes it more trustworthy.
The most credible digital twin is not the one that appears to know everything. It is the one that can distinguish, clearly and usefully, between what it knows, what it has worked out, what it has modelled and what it does not yet know.
"Precision and certainty are not the same thing."