The problem
Order records showed when jobs started and finished, but not what happened in between: waiting for the next job or material, short stops, and alarms nobody owned all blurred together.
What they did
External sensors were mounted on each machine with no PLC access needed, about 17 minutes per machine on average, fitted during normal changeovers. After 2-3 weeks of baseline data, the team made three changes: prepare the next job before the current one ends (with idle alerts), add a shift-start readiness check, and assign an owner to each recurring alarm type.
Reported results
The vendor reports idle time down about 37%, microstops down about 22%, downtime down about 18%, and roughly 4,000 extra productive machine-hours a year across the fleet (annualized). Third shift had the most idle time and interruptions. The reported OEE jump (21.5% to 54.9%) doesn't line up with the much smaller utilization change (66.4% to 69.8%), and the vendor notes it didn't publish the inputs, so treat that number with caution.
Our take
Same lesson as most monitoring wins: the sensors only showed where time went; the gains came from boring process fixes (staging the next job, shift handovers, alarm ownership). The 'install without touching the control' approach is what makes this interesting for shops with older machines. It's a German vendor and an unnamed customer, so we can't verify it, and US availability is unclear.