If you've been anywhere near the smart-building or IoT world this year, you've been sold a "digital twin." It's the phrase of the moment — every vendor has one, every conference talk gestures at one, and it arrives wrapped in enough futuristic gloss that it's genuinely hard to tell whether there's a real idea underneath or just a very well-funded fog. Having spent the year actually building toward one for real buildings, I want to do the un-glamorous thing and separate the two, because there is real substance here, and it's being buried under exactly the kind of hype that makes serious people roll their eyes and dismiss the whole thing.

Let me be a hype-checker in good faith: I'll defend the real idea and name the nonsense, because both are necessary if the concept is going to survive its own marketing.

The real idea, stated plainly

Strip away the gloss and a digital twin is a genuinely useful concept: a live, digital model of a physical thing, kept current by that thing's real data, that you can query and reason about as a stand-in for the real one.

For a building, that means a digital representation that knows the building's structure — its floors, rooms, systems, and how they relate — and is continuously fed by the building's actual sensor data, so the model reflects not just how the building is built but how it's behaving, right now. And crucially, because it's a model, you can ask it questions and even simulate against it in ways you can't easily do to the physical building.

That's not vapour. That's a real and valuable thing, and I've felt the value of it directly:

  • It turns scattered data into a coherent whole. Without a twin, your sensor data is a pile of disconnected readings — this device says this, that one says that. The twin gives it structure: this reading belongs to this room on this floor served by this system. It's the difference between a heap of facts and a model you can actually reason about.
  • It lets you ask "what if" without touching the building. Want to know what happens if you change how a floor is used, or how a system is run? A good twin lets you explore that against the model rather than experimenting on the live building and its actual occupants. That's genuinely powerful.
  • It's a shared reference everyone can point at. The twin becomes the common object that the facilities team, the analysts, and the planners can all refer to — a single agreed representation of "the building," which, as ever, is worth a great deal precisely because agreement is hard.

So yes — under the hype, there's a real, substantial, useful idea. I'm not here to dismiss it. I'm here to protect it from its own marketing.

The hype, and why it's harmful

Now the fog, because it's thick and it does real damage. The hype around digital twins takes a sound, bounded, useful concept and inflates it into a magical everything-machine, and that inflation isn't harmless — it sets expectations that the real thing can't meet, which then discredits the real thing.

  • "A perfect virtual replica of your building." No. A twin is a model, and every model is a deliberate simplification that captures what matters for a purpose and ignores the rest. Selling it as a perfect replica promises a fidelity that doesn't exist and isn't even desirable — a model that modelled everything would be as unmanageable as the real building. The value is in the simplification, and the hype denies exactly that.
  • "It'll optimise everything automatically." The twin is a representation. It doesn't magically make decisions; it's a tool that helps people (and sometimes algorithms) make better ones. The leap from "live model" to "autonomous optimising intelligence" is where the marketing quietly skips over all the actual hard work of deciding what to do with the model once you have it.
  • "Buy this product and you have a digital twin." This is the one that costs people real money. A twin isn't a product you purchase; it's a thing you build out of your building's structure and its live data, requiring exactly the unglamorous data-platform work — ingestion, modelling, integration — that the shiny demo skips. Vendors selling a "twin in a box" are selling the easy 10% and quietly leaving you the hard 90%.

The harm is specific: when the concept is oversold this hard, the inevitable disappointment gets attached to the idea rather than to the overselling. People try to buy the magic, get the reality, feel cheated, and conclude "digital twins are hype" — throwing out a genuinely good concept because it was marketed dishonestly.

A digital twin is a live, queryable model of a physical thing, built from its real data. That's real and useful. Everything past that sentence — perfect replica, autonomous optimisation, twin-in-a-box — is the fog, and the fog is what will make people give up on the substance.

So, substance or hype? Both — and it matters which you buy

My honest verdict after a year of building toward one: the substance is real and worth pursuing, and the hype is real and worth ignoring, and telling them apart is the whole skill. If you approach a digital twin as "a structured, live model of my building that I build from my data and use to reason and simulate," you're chasing something genuine that will pay off. If you approach it as "a magical replica I can buy that optimises everything on its own," you're chasing fog, and you'll spend real money learning it was fog.

The tooling to do the real version is maturing — the platform pieces to build and maintain a twin are getting genuinely usable — which is exactly why it matters to be clear-eyed now, before the disappointment cycle sets in. The concept deserves to survive its own hype. It'll only do that if enough of us insist, loudly, on the modest true version over the magical false one — and are willing to be the person in the meeting who says "that part's real, and that part's a slide." I'll keep being that person, because the real idea is good enough to be worth defending from the people selling it badly.