In the summer we let a supplier trial an indoor drone at our largest depot. It flies the aisles after the last shift, reads the label on every pallet and photographs every location. In one night it covered about twenty two thousand locations. Our two cycle counters cover about six hundred a day, which means the whole building is counted roughly every seven weeks.
On its first night the drone reported fourteen hundred discrepancies.
Our stock system treats a discrepancy as a rare and serious event. A counter raises it, a supervisor approves the adjustment, and anything worth more than five hundred pounds goes to finance. That process was designed for about twenty a day and works well at that rate. At fourteen hundred it simply stopped. The drone's findings sat in the supplier's portal, because there was nobody with time to approve them into ours.
Most of them were not lost stock. Pallets put away one slot to the left. Labels too creased to read. Pallets the system expected in a location that were on a forklift at the time. A few hundred were genuine, including two pallets of frozen goods recorded in a location that had been empty for a month.
The stock accuracy figure changed too. Measured by our cycle counts it had been ninety nine point two percent for years. Measured by the drone it was ninety four. Nothing in the warehouse had become worse. We could simply see more of it.
I expect this to be the shape of the next ten years in operations. Counting will be done by drones, by cameras on forklifts and by tags on the pallets themselves, continuously and almost for free. Our controls were built around observations that were scarce and expensive, and they assume a person looked at each one. They will need to sort what the sensors find into what can be corrected automatically, with a record kept, and what genuinely needs a person. And our auditors will need to agree what a count is when one happens every night.
Someone will also have to explain to the board that the number got worse because we started looking properly.
When measurement becomes cheap, the hard work moves to deciding which of its findings deserve a human.
– Serguey Shinder
Top comments (0)