01
What OEE actually measures
OEE multiplies three ratios. Availability is the share of planned time the machine actually ran. Performance is how fast it ran against its ideal cycle time. Quality is the share of parts that were good first time. A machine at 90% availability, 90% performance and 95% quality has an OEE of about 77%.
The figure itself matters less than the three losses behind it. OEE is useful because it shows whether a machine is losing time to stoppages, to slow running or to scrap, and those need very different fixes.
02
Why manual OEE sheets mislead
Most plants already try to track downtime on paper or in a spreadsheet filled in at the end of the shift. Long breakdowns get recorded. The two-minute stops, the jams cleared by hand and the slow running after a changeover usually do not, and together they are often the biggest loss on the line.
Hand-filled sheets also tend to show the number people expect. Automatic capture removes the argument: the machine says when it stopped and for how long, and people only add why.
03
A realistic first 30 days
Week one: pick the bottleneck machine, the one that limits the whole line's output, and agree its ideal cycle time and a short list of stop reasons with production. Week two: connect it. Run, stop and part-count signals usually come from the existing PLC, or from a small IIoT gateway with a few sensors when the machine has no usable controller.
Weeks three and four: let it run, have operators tag stops on a simple screen, and review the downtime pareto with the team every few days. By the end of the month the plant knows its real top three losses, often not the ones it assumed.
04
What to do with the numbers
Attack the biggest loss first and leave the rest. If changeovers dominate, work on changeover. If micro-stops dominate, look at the jam points and the feeding. Re-measure after each change, so the improvement is proven rather than assumed.
Only once the first machine has shown results is it worth rolling OEE out to the rest of the line. A plant that starts with every machine at once usually ends up with lots of dashboards and few decisions.
Written by the ASKworX engineering team