We planned the Beacon rollout for six weeks. The plan was clean. The building was not.
The floor plan is a lie
Our model said one Beacon every twenty meters would cover a floor. It would have, in a simulation with no walls, no racks, and no people. The real floor had steel shelving that ate the signal and a stairwell that turned into a dead zone the size of a conference room. We walked the space with a phone and a notebook — the old way, because the model was wrong — and found eleven gaps the math never saw. You cannot model a building you haven’t stood in.
Batteries die on a schedule you didn’t draw
We rated the Beacon cells at two years and filed the replacement ticket in a drawer. Month eighteen, the first ones went dark, and not evenly. The cold-storage unit’s beacons died first — low temperature slows the chemistry, then drops it off a cliff. The ones by the loading dock died from vibration, not age. By month twenty we had a steady trickle of silent units and a very real “where’s the locator” complaint. We swap now on a staggered calendar keyed to where each Beacon actually lives, not the datasheet’s best day.
People will find the loophole
This one embarrassed us. We tracked shared tools with beacons, sure the data would show who was using what. It showed something else. Someone had taped foil over the Beacon on the tool they wanted to keep, so the system thought it never left the cage. We built “anomaly detection” for a week before a tech walked over and peeled the foil off. The Beacon was fine. The human was creative. No algorithm patches that.
The gateways you forgot to power
We counted beacons. We forgot the gateways that listen. Half our “Beacon dead” tickets were actually a gateway that lost its Ethernet run, or a switch someone rebooted into the wrong VLAN. The locator looked broken. The radio was fine. We monitor the gateways like they’re the product now, because in a Beacon system, they are.
The part that actually worked
Not everything failed. The dashboard that turned “14 beacons silent this week” into a red dot on a floor plan — that got opened every morning. The teams that used it stopped losing things. The lesson wasn’t the tech. It was that a Beacon is dumb until someone reads what it says, and most of our failures were ours.
If we did it again
We’d stand in the building before we opened the spreadsheet. We’d swap batteries on the building’s clock, not the vendor’s. And we’d assume the users will beat the system, because they will. A Beacon deployment isn’t a product you ship. It’s a habit you have to keep.
