A tea and spice packer had three identical lines. Two performed. One quietly ran at roughly half their OEE, and nobody could say why. The downtime data resolved the mystery to a single recurring failure code, and the 61 hours a month it was costing.
The plant ran three packing lines built to the same specification, doing the same job. Two of them performed as expected. The third consistently produced less, and had done so for long enough that its lower output had become the number everyone quietly planned around. Ask why, and you got theories: an older machine, a difficult product, a tougher shift. None of them were measured.
Because the three lines were judged against a shared target, the shortfall on the weak line was absorbed into the site total. On paper the plant looked broadly fine. On the floor, one line was leaving a shift's worth of product on the table every few days.
Augos monitored all three lines and, crucially, broke each line's downtime out by failure code rather than reporting a single availability figure. The gap stopped being a mystery almost immediately. The two healthy lines lost time to the normal spread of changeovers and minor stops. The third lost the same normal time plus a large, repeating block attributed to one specific failure code.
A single recurring stoppage was costing the line roughly 61 hours of run-time a month. It never looked like a breakdown because each occurrence was short, but it happened relentlessly, all shift, every shift.
Once the cause had a name and an hour figure attached to it, it became an engineering task rather than a debate. Recovering those 61 hours a month represented roughly a doubling of the line's available run-time, closing most of the gap to its two sisters without any change to the product or the crew.
The broader lesson held for the whole site: identical machines do not guarantee identical output, and a shared target will hide the one that is struggling. Only per-line, per-cause data makes the laggard, and the reason for it, impossible to miss.
The three lines shared a single throughput target, and the underperformer's shortfall was absorbed into the site total. Without per-line, per-cause data it read as normal variation.
A recurring stoppage that tripped the line repeatedly through each shift. Individually each stop was short, so it never registered as a breakdown, but it dominated the line's lost hours.
The 61 hours a month recovered from that one cause represented roughly a doubling of the line's available run-time. That was the headroom the data pointed to once the cause was addressed.
Book a demo with one of our manufacturing experts.