IoT & Sensors

A schedule is a guess written in the past

Fixed cleaning frequencies are built from an average: this area was busy at these hours, on average, last quarter. It’s the best anyone can do without measurement, and it is wrong most of the time in both directions.

Wrong in one direction costs money — a team services an area nobody used. Wrong in the other costs reputation — an area that took triple its usual traffic sits unserviced because its slot isn’t until later.

Both failures come from the same source: nobody knows what the building is doing right now.

What the sensors measure

Occupancy and people counting — real traffic, which is what should drive frequency. Supply levels — soap, paper towel and toilet paper reported before they run out. Odor — a condition no schedule can anticipate and no walkthrough catches until it’s established. Leaks — water where water shouldn’t be, which is a maintenance issue found by cleaning sensors more often than you’d expect.

Everything reports to a gateway that aggregates the site and alerts the assigned team through CleanVision. That last step is what makes the layer worth installing — a sensor whose reading nobody acts on is a device on a wall.

Privacy is a design constraint, not a policy

These sensors count and measure. They do not see and they do not hear.

There is no camera in the sensing layer, and on the tablets used for feedback the camera and microphone are disabled outright rather than left available and unused. That distinction matters: a disabled capability cannot be re-enabled by a configuration change made later by someone with a different priority.

Where a camera does exist — on the Titan One station — it is a separate, explicitly agreed deployment with written boundaries, described under camera cleaning services. We keep the two conversations apart on purpose, because conflating them is how monitoring quietly expands.

Thresholds per area

A threshold copied across a site is nearly useless. Traffic that’s unremarkable in a building’s main restroom would be extraordinary on a quiet administrative floor.

Levels are set per area and adjusted as the building’s use changes. When a value crosses one, it becomes an alert with an owner rather than a data point in a chart waiting to be noticed.

What it doesn’t do

It doesn’t clean anything, and it doesn’t replace the supervisor. A sensor can report that an area has taken unusual traffic; it cannot know that a tenant is running a three-day event and the pattern is expected.

What it removes is the delay between a condition appearing and someone knowing about it. In a model built on acting before complaints, that delay was the whole problem.