I spent a long time building systems to understand how busy buildings are used — the flows, the crowds, the rhythms of a space full of people. And then, over the course of a single strange week in March, every building I'd been measuring simply emptied. The occupancy platform I'd built to make sense of a busy building suddenly found itself watching an empty one, and the experience of that — both eerie and, unexpectedly, useful — is worth writing down, because it taught me things about my own systems that a normal year never would have.

Watching a building go dark

The first thing was purely, strangely emotional, and I didn't expect data to make me feel it. Occupancy data, when a building is alive, is a busy, noisy, comforting thing — a constant hum of people arriving, moving, gathering, leaving. Watching that hum fall to nothing, over days, on the same dashboards I'd built to celebrate a building's liveliness, was genuinely unsettling. The graphs I'd designed to show a place full of life were now drawing, in clean confident lines, its complete emptiness. A flatline where there used to be a pulse.

There's something about seeing a dramatic human event rendered in the flat, honest language of your own monitoring that lands differently than the news does. The building wasn't editorialising. It was just reporting, faithfully, that the people were gone — and the faithfulness was the haunting part.

The quiet value of a baseline

Once the initial strangeness passed, the analyst in me noticed something genuinely valuable, and it's a lesson I'll carry well past this: the emptiness was only legible because I had the fullness to compare it against.

An empty building, measured by a system with no history, is just... empty. Not very informative. But an empty building measured against months of what "normal" looked like is enormously informative — because now you can see exactly how far from normal you are, in which spaces, at what times, with what precision. The baseline I'd been quietly accumulating all that time, without thinking of it as valuable in itself, turned out to be the thing that made the anomaly meaningful. You cannot measure a departure from normal without a definition of normal, and mine had been building itself, patiently, the whole time.

That reframed how I think about historical data generally. We collect it to answer the questions we have now. Its deeper value is that it lets us understand the questions we don't yet have — the sudden anomaly, the unprecedented event — because the past is the only yardstick against which the strange present can be measured. I never designed my occupancy history to help navigate a pandemic. It's doing exactly that, purely because it existed.

From understanding to navigating a return

And that's the turn this work has taken, which I never could have predicted. The platform I built to understand a busy building is becoming a tool to navigate a safe return to one — and the shift in purpose is teaching me how much latent capability sits in a system built for a different question.

As people begin, tentatively, to think about returning, a whole set of new questions arrives that occupancy data is suited to answer:

  • How full is too full, now? The same density that was a sign of a vibrant space is now a risk to be managed. The platform that counted heads to understand popularity can count them to understand distancing — how many people are in a space, whether it's within a safe limit, when it's crowding.
  • Which spaces can safely reopen, and to what level? With per-space historical patterns, you can reason about which areas naturally stay sparse and which pack tight, and phase a return accordingly — not on guesswork, but on the measured reality of how each space actually fills.
  • Is the plan working? As people return under new rules, the data becomes the feedback loop: are the distancing measures actually holding, or is that corridor bottlenecking anyway? The building can tell you whether your policy is surviving contact with human behaviour.

None of this was in the original design. I built a system to answer "how is this building used?" and it turns out to also answer "how do we bring people back to it safely?" — because underneath, both are the same question about where people are and how densely, just pointed at a different purpose.

Historical data's real gift isn't answering the question you collected it for. It's being there, ready, for the unprecedented question you couldn't have known to ask — because the only way to understand an anomaly is against a baseline you had the foresight, or the luck, to have been keeping all along.

What this period is teaching me about my own systems

Two lessons are sticking, and they'll outlast the pandemic. The first is a new respect for baselines — for the quiet, unglamorous accumulation of "normal" that feels like it's doing nothing until the day something abnormal happens and it becomes the most valuable data you own. I'll never again think of historical collection as merely feeding today's dashboards; it's insurance against the questions the future hasn't asked yet.

The second is humility about how little we know about the eventual uses of what we build. I built an occupancy platform for one clear purpose and it has quietly become something else entirely, useful in a scenario I never imagined, because the underlying capability — seeing where people are, over time — turned out to be more general than the specific job I built it for. There's a lesson there about building systems that capture fundamental, reusable truths rather than narrowly-scoped answers: the general capability survives a change of question that the narrow one wouldn't.

I don't know what the return looks like, or when, or whether the buildings fill back up to what they were. But I know that the strange experience of watching them empty, on systems I built to watch them full, has taught me more about the real value of what I've been doing than a year of everything working normally ever could. The building went quiet. And even its silence, measured against everything it used to say, turned out to be worth listening to.