Key points
- OEE multiplies availability, performance and quality. It tells you how much of your planned time produced good product.
- Most plants already have enough data to find their biggest loss. The problem is usually inconsistent downtime reasons, not missing sensors.
- Spend 30 days on one line: define planned time, record stops with a short reason list, add counts and scrap, then rank losses and fix the top one.
What OEE actually measures
Overall equipment effectiveness (OEE) compares what a line produced with what it could have produced in the time it was scheduled to run. It's the product of three ratios:
- Availability: running time divided by planned production time. Breakdowns, changeovers and waiting for material reduce it.
- Performance: actual output divided by the output the line would make at its ideal speed during the running time. Minor stops and slow running reduce it.
- Quality: good units divided by total units. Scrap, rework and start-up rejects reduce it.
A worked example
A line is planned to run for an 8-hour shift (480 minutes) and loses 60 minutes to stops, so it runs for 420 minutes. Its ideal rate is 60 units a minute, so it could have made 25,200 units in that time. It actually made 21,000, of which 20,400 were good.
| Factor | Calculation | Result |
|---|---|---|
| Availability | 420 ÷ 480 | 87.5% |
| Performance | 21,000 ÷ 25,200 | 83.3% |
| Quality | 20,400 ÷ 21,000 | 97.1% |
| OEE | 0.875 × 0.833 × 0.971 | 70.8% |
The headline number matters less than the breakdown. In this example, performance loses more output than downtime does, which is common and easy to miss when a plant only tracks stops.
The six big losses
A useful way to sort what you find is the classic list of six big losses used in total productive maintenance:
| OEE factor | Loss | Typical examples |
|---|---|---|
| Availability | Equipment failures | Breakdowns, jams that need maintenance |
| Availability | Setups and adjustments | Changeovers, warm-up, first-article checks |
| Performance | Idling and minor stops | Short stops under a few minutes, sensor trips, blocked or starved conveyors |
| Performance | Reduced speed | Running below rated speed to avoid problems |
| Quality | Start-up rejects | Scrap while the line stabilizes after a start or changeover |
| Quality | Production rejects | Scrap and rework during steady running |
A 30-day plan for one line
Week 1: define planned time and a short reason list
Pick one line that matters. Agree on what counts as planned production time, for example shift length minus scheduled breaks and planned maintenance. Then write a list of no more than 10 to 12 downtime reasons that operators can choose from quickly. A long list produces inconsistent data.
Week 2: record every stop, count and reject
Record each stop above a threshold you choose, such as 2 minutes, with start time, duration and reason. Record total and good counts per shift. If you can only do this on paper, do it on paper.
Week 3: rank the losses
Add up lost minutes by reason and sort from largest to smallest. Convert performance and quality losses into equivalent minutes so everything can be compared. Then look for patterns: time of day, shift, product, or the minutes after a changeover.
Week 4: fix the top loss and measure again
Choose the single biggest loss, assign an owner and make one change. Keep recording. A focused fix on the top loss usually does more than small improvements spread across ten.
Where AI helps, and where it doesn't
AI is useful for the parts people find tedious: grouping hundreds of free-text downtime comments into a short list, spotting that jams cluster after changeovers, comparing shifts fairly and writing a plain summary for the morning meeting. It doesn't replace the operator who knows why the guide rail drifts, and it shouldn't be changing machine settings on its own.
Common mistakes
- Comparing OEE between plants or lines that define planned time differently
- Using an ideal speed that's really the speed the line usually runs at, which hides performance loss
- Tracking only major stops and ignoring minor ones
- Collecting data for months without acting on the top loss