OEE, briefly
Overall Equipment Effectiveness = Availability × Performance × Quality. It is a simple product. In a container glass plant, the devil hides in every factor.
The three traps
Trap 1: Planned downtime gymnastics
Plants regularly exclude "planned" downtime from Availability — changeovers, scheduled maintenance, mould changes. Sometimes reasonable, often not. A planned 18-hour changeover is still 18 hours your plant is not making product. Excluding it flatters the number and hides the largest accessible opportunity.
Our rule: report OEE with and without planned downtime. Both are honest; showing only one is rarely honest.
Trap 2: Quality at the wrong gate
Where you count quality matters. Hot-end counts flatter the number; cold-end counts are closer to reality; pallet-level counts are closest to what the customer sees. Plants often pick the friendliest gate by default.
Our rule: report three quality numbers — hot, cold, pallet — and target the cold-end one. The delta between them is itself a diagnostic.
Trap 3: Performance against a soft benchmark
Performance = actual run rate / maximum run rate. "Maximum" is often set at what the line is currently capable of, not what it should be. This locks in underperformance.
Our rule: the performance denominator is the design rate or the best recent sustained rate, whichever is higher. Not the average of last quarter.
The decomposition we use
- Availability: scheduled production time minus all downtime, planned and unplanned. Report both.
- Performance: actual run rate vs design rate.
- Quality: cold-end pass rate, with hot and pallet as supporting.
What OEE will not tell you
OEE is a rollup. It is great at the plant level, useful at the line level, and misleading at the section level. For section-level work, use defect pareto by mode, first-hour yield, and swab cycle adherence — not OEE.
How to report it
- One number for the plant, decomposed into A × P × Q.
- Trend (4-week rolling average) more than spot value.
- Variance to plan, with the top 3 contributors to variance in plain English.
What to do first if you are starting from scratch
Measure Availability properly — including planned downtime, even if you report both. Most plants we walk into have a 15–25 point overstatement hiding in this factor alone. Fix it, and the rest of the conversation becomes real.