The factory that wasn't broken

On paper, the numbers were worrying. One site reported an OEE of just 16.46%. For any production team that raises immediate concern: lost output, a plant that's struggling, something on the floor that needs urgent fixing. If the data was right, someone had a serious problem. There was only one issue, the story didn't quite add up.

A closer look showed two things that didn't fit. First, over 243 hours were recorded as "Power Off" downtime. Some stops ran past four hours, and many fell on night shifts the line was never scheduled to run. Was the line down, or simply not planned to run?

Second, on separate runs the line ran at speed, yet quality showed 0%, as if every unit was rejected. The rejects entered at the terminal didn't match what the line had made.

Rather than jump into mechanical fixes, the focus went to the measurement first. Long "Power Off" periods were reclassified as unpaid time. Quality inputs were reviewed against actual production reality. Job-level data was used to separate true losses from logging errors. As the noise disappeared, the signal became clear. Measured correctly, the site's OEE was about 44.70%, not 16.46%. Nothing on the floor had changed; the record now matched what the line had been doing all along.

With the distortion gone, the actual constraint stood out: label-feed delays, responsible for dozens of stoppages and nearly 40 hours of lost time a month. For the first time the team could target a real, measurable bottleneck instead of a misleading headline number.

The lesson generalises. In many factories the instinct is to fix the process, the machine, or the people. But before fixing performance, make sure you're seeing it clearly, because sometimes it isn't the factory that needs fixing. It's the measurement.