A line rated at 12,000 units an hour that stops for forty minutes has not made 8,000 units. At a euro of margin each that is a figure anyone can put in a report, and it is the figure that gets put in the report. It is also the only part of the cost that stops when the line restarts.A line rated at 12,000 units an hour that stops for forty minutes has not made 8,000 units. At a euro of margin each that is a figure anyone can put in a report, and it is the figure that gets put in the report. It is also the only part of the cost that stops when the line restarts.

Restart is not instant. A filler that has been down for forty minutes comes back with product in it that has to be scrapped, a temperature that has to come back to setpoint, and a first pallet that quality will not release. On the food lines we work on, the tail after a stop routinely runs longer than the stop itself. Nobody measures the tail, because the machine is showing green.Restart is not instant. A filler that has been down for forty minutes comes back with product in it that has to be scrapped, a temperature that has to come back to setpoint, and a first pallet that quality will not release. On the food lines we work on, the tail after a stop routinely runs longer than the stop itself. Nobody measures the tail, because the machine is showing green.

Somebody upstream is now blocked and somebody downstream is now starved. If your buffers are honest, that is minutes. If your buffers are the reason the line looked balanced on paper, it is the whole shift. Then there is the labour that was paid for the forty minutes, the changeover that has now been pushed into overtime, and the order that ships a day late to a customer who counts days.Somebody upstream is now blocked and somebody downstream is now starved. If your buffers are honest, that is minutes. If your buffers are the reason the line looked balanced on paper, it is the whole shift. Then there is the labour that was paid for the forty minutes, the changeover that has now been pushed into overtime, and the order that ships a day late to a customer who counts days.

In fifteen years of downtime records the pattern barely moves. A small number of causes produce most of the lost time, and they are almost never the dramatic ones. It is the same sensor, the same jam at the same transfer, the same changeover that takes twenty minutes longer than the standard says. They survive because each one is small enough to clear and get on with, and nobody adds them up.In fifteen years of downtime records the pattern barely moves. A small number of causes produce most of the lost time, and they are almost never the dramatic ones. It is the same sensor, the same jam at the same transfer, the same changeover that takes twenty minutes longer than the standard says. They survive because each one is small enough to clear and get on with, and nobody adds them up.

Capture the reason at the machine, on a button, chosen by the operator who cleared the fault. Not typed into a spreadsheet at the end of a shift from memory. Then read the top three every week and fix one. That is unglamorous and it works: the plants that do it stop arguing about whether OEE is accurate and start arguing about which of the three to fix first, which is the argument worth having.Capture the reason at the machine, on a button, chosen by the operator who cleared the fault. Not typed into a spreadsheet at the end of a shift from memory. Then read the top three every week and fix one. That is unglamorous and it works: the plants that do it stop arguing about whether OEE is accurate and start arguing about which of the three to fix first, which is the argument worth having.