I've spent a good while now building the platform that lets a building watch itself — sensors, streams, storage, dashboards, all the machinery I've written about piece by piece. But the machinery was always the means, not the end. The end was learning things about the building, and now that the platform has been running long enough to have watched a building live its life for a while, I want to write down what it actually taught me. Because the most interesting findings weren't technical at all. They were about the stubborn gap between how a building is designed to be used and how people actually use it.

Here are the field notes — the things the data said that no one expected it to.

The building on paper is a fiction

The single biggest theme, running through nearly everything, is this: the building as designed and the building as lived are two different buildings, and only one of them is real. Every space is designed with an intended use, a planned capacity, an assumption about how people will move through it and occupy it. And the data, patiently, over months, shows you how thoroughly those assumptions get ignored by the actual humans involved.

Rooms designed for one purpose get used for another. Spaces planned for many people sit mostly empty while a smaller, less obvious space is constantly overcrowded. Movement flows along paths the designers didn't intend and avoids ones they did. None of this is anyone doing anything wrong — it's just people, doing what makes sense to them in the moment, in aggregate, over time. The design was a hypothesis about human behaviour. The data is the result of the experiment, and the hypothesis is usually part-wrong in ways nobody would have guessed.

The specific surprises

Some concrete examples of what the data surfaced, anonymised and generalised:

  • The popular space nobody planned for. There's almost always a spot that becomes a hub for reasons that have nothing to do with its intended function — it's near something, it feels right, it caught on — and it ends up carrying far more use than the spaces actually designed to be hubs. The data finds these, and once you see one, you realise the building has a social life the floor plan knew nothing about.
  • The expensive emptiness. Spaces provisioned and conditioned — heated, cooled, lit, cleaned — for a level of use they never actually see. From the data's point of view, this is pure waste: real resources spent maintaining spaces for occupancy that isn't happening. But you'd never know without measuring, because an empty room and a used room look identical if no one's counting.
  • The rhythms nobody articulated. Buildings have patterns — daily, weekly, seasonal — that everyone half-knows and no one has ever actually described. The data makes them explicit: this is genuinely when the building fills and empties, this is the real shape of a week here. And the explicit version is always subtly different from the assumed one, in ways that matter for how you'd run the place.

The uncomfortable part: watching people

I can't write honest field notes without the part that sat least comfortably: all of this is, in the end, data about people. Occupancy, movement, presence — however aggregated and anonymised, the raw material is human beings going about their day, and a platform that can see how a building is lived in can, by the very same means, see how the people in it behave.

That's a responsibility I felt more keenly the longer the platform ran. There's a real and important line between understanding a building — how spaces are used, where energy is wasted — and surveilling the individuals in it, and it's a line that's technically easy to cross and ethically essential not to. The discipline is to design deliberately for the building-level insight and against the individual-level intrusion: aggregating early, keeping the resolution no finer than the legitimate question actually requires, and being honest that "we could measure that" is never the same as "we should."

I raise it here, in the field notes rather than in a footnote, because it isn't a footnote to the work — it's part of the work. A smart building that helps you run the place better while quietly eroding the privacy of everyone in it has failed, even if every dashboard is green and every metric improved. The seeing that the data enables is genuinely valuable, and it arrives fused to an obligation to be careful about what you choose to see, and whom. I'd rather build something that sees a little less and respects the people in it than something that sees everything and treats them as instruments. That choice is made in the design, early, or it isn't really made at all.

Why this is the actual point

It would be easy to read all this as "the sensors caught people being inefficient," but that's not the lesson, and it's a slightly mean misreading. The lesson is that a building is a living system shaped by the people in it, and for the entire history of buildings we've had almost no way to actually see how they're lived in. We designed on assumption, built, and then never checked the assumption against reality — not out of negligence, but because the reality was invisible. You genuinely could not know how a building was used, at the level of detail that would let you do anything about it.

That's what the smart-building data actually changes. Not "control the building" — that's the part the hype fixates on. The quieter, more profound thing is seeing the building honestly, for the first time: closing the loop between how we imagine spaces will be used and how they actually are, so that the next decision — about this building or the next one — can be informed by what's real rather than what was assumed.

For all of history we've designed buildings on a hypothesis about how people would use them, built them, and never checked. The genuinely new thing isn't a building that controls itself. It's a building you can finally see being lived in — and the seeing is humbling, because the assumptions are always partly wrong.

From surprise to decision

A finding that just sits in a dashboard is a curiosity; a finding that changes a decision is the point. So let me give the shape of how one actually travelled, because the journey from "huh, that's odd" to "we did something about it" is the part that matters and the part that's hardest.

The data showed a space being conditioned — heated, lit, serviced — far more than its actual use justified. On its own, that's just a number on a screen. What made it matter was walking it to the people who could act: showing them, plainly, "here is what you're spending to maintain this space, and here is how little it's actually used." Not as an accusation — as a picture of a reality they genuinely couldn't see before, because nobody had ever been able to measure occupancy against provisioning at this resolution.

And here's the part that surprised me: the number alone didn't move anyone. What moved them was the number translated into their terms — into cost, into comfort, into the trade-offs they actually cared about. The raw finding was true and inert. The finding, communicated in the language of the person who could act on it, changed what they did. Which is, once again, the whole thesis of how I think about this work: the data surfaced the reality, but it took the human act of translation to turn the reality into a decision.

That gap — between a true finding and an acted-upon one — is where most data value quietly dies. The building's data was full of honest, useful surprises. Only the ones I managed to communicate into someone's actual decisions ever became worth anything. The rest are still sitting in storage, true and unread.

What it changed in how I think

Two things have stuck from watching a building through its own data. The first is a deep respect for the gap between the designed and the lived, in everything, not just buildings — the reminder that any system involving people will be used in ways its designers didn't intend, and that measuring the actual behaviour beats reasoning about the intended behaviour every single time. I catch myself applying that now far outside buildings: to software, to processes, to anything where we assume how people will behave and rarely check.

The second is a wariness about jumping straight to control and optimisation — the exciting, automatable end of smart buildings — before you've done the humble work of just seeing clearly what's happening. The data's first and most valuable gift isn't a lever to pull. It's an honest picture of a reality that was invisible before, and that picture is worth sitting with before you rush to act on it. Most of the value I found in a year of watching a building wasn't in anything the building did automatically. It was in what the building finally let us understand — about itself, and about the difference between the plan and the people.