Every sensor pays its way
If a device doesn't solve a real problem, it shouldn't be on the wall.
Procurement for building IoT is broken. Hardware vendors sell volume; consultants want a new dashboard. The instinct is a sensor on every desk, a widget on every door, a monitor in every corridor — blind to what all of it costs to keep running.
The rule that fixes it is simple: every device has to pay its way. It solves a direct, measurable problem, or it doesn't go on the wall. Thousands of sensors bought to fill a screen aren't a solution — they're a multi-year maintenance liability.
The cheap-sensor trap.
Look at the cheap air-quality kit flooding the market. Plenty of devices that claim to measure CO₂ don't — they estimate it from a volatile-compound sensor (eCO₂). Those readings drift with humidity and temperature, wander within months, and are useless without constant calibration against a proper NDIR reference.
Don't judge on the sticker price. Cost the whole ten years. Cheap sensors need yearly battery swaps, they drift, and they need recalibrating — by year three the labour of maintaining a fleet of them has eaten the saving. Buy a few good reference devices built for a decade, then use software to infer the rest.
Virtual sensors.
You don't need a high-end sensor in every room. A few accurate references can calibrate or infer readings across many cheaper points. A washroom is the classic case: instead of a privacy-invasive counter in every cubicle, put one accurate counter at the entrance and simple magnetic switches on the stall doors. Knowing there are four stalls, the system correlates the two streams and gets accurate usage without bloating the budget.
Cost it against the outcome.
Cost the project against the outcome, not the chart. If the goal is keeping a washroom stocked, you don't need exact usage curves — you need to know when to send someone before a passenger finds a mess, and to raise that task automatically. Sparse data is good enough.
Meeting rooms are the same. A good CO₂ sensor reliably shows whether a room is in use. CO₂ lags a few minutes behind people arriving — fine if the outcome is releasing ghost bookings or dropping the heating setback. You don't double the budget on real-time desk sensors unless an automation genuinely needs the second-by-second signal.
The point.
Efficient hardware beats more hardware. Make every sensor work twice, infer the rest in software, and keep a lean footprint that still proves compliance. Every box on the wall earns its place.