The most common OEE errors are a soft ideal rate, rework counted as good, real losses hidden as excluded time, good units used in performance, and a plain average across machines. On one example shift they move OEE from the true 73.5% to anywhere between 70.6% and 84.0%. Reconcile losses to 100% to catch them.
One shift, five different answers
Most OEE errors are not arithmetic. They are rules applied wrongly. To see how much they matter, use one shift and change one thing at a time.
Example: a line is scheduled for 480 minutes and stops for 60, so it runs 420 minutes. The ideal rate is 40 units per minute. It makes 14,700 units, of which 14,112 are good (380 scrapped, 208 reworked). Correct OEE: 0.875 × 0.875 × 0.96 = 73.5%. Check: 14,112 ÷ 40 ÷ 480 = 73.5%.
The ten mistakes
| # | Mistake | Fix |
|---|---|---|
| 1 | Ideal rate set to last month's average | Use the accredited rate |
| 2 | Rework counted as good | Good means right the first time |
| 3 | Real losses hidden in excluded time | Exclude only planned time |
| 4 | Good units used in performance | Use total units for performance |
| 5 | Plain average across machines | Weight by scheduled time |
| 6 | Performance left above 100% | Cap it, then check the rate |
| 7 | Stops written from memory | Log them when they happen |
| 8 | An "other" reason code | Categorise every minute |
| 9 | Losses never reconciled to 100% | OEE + losses = 100% |
| 10 | Early finish counted as a loss | Time freed is excluded time |
The rules on this page follow Overall Equipment Effectiveness (R. C. Hansen, 2001). Where a point is our own advice, it says so.
1. A soft ideal rate
If the ideal rate is what the line did last month, performance reads near 100% and the speed loss disappears. In the example, a rate of 35 per minute turns 73.5% into 84.0%. Use the best accredited rate for the size and format. See ideal cycle time.
2. Rework counted as good
A unit that needed rework was not good the first time. If you count the 208 reworked units as good, good units become 14,320, quality reads 97.4% and OEE reads 74.6% instead of 73.5% (14,320 ÷ 40 ÷ 480). Hansen treats the original time as lost, and has three rules for where the rework itself is counted. See how to calculate OEE quality.
3. Real losses hidden as excluded time
Excluded time is time you planned not to produce: shutdowns planned at least a week ahead, headroom, no product scheduled, an experiment whose product will not be sold. A "tidy-up" that happens every shift without a plan is a loss.
In the example, say 30 of the 60 stop minutes are that tidy-up. Moving them to excluded time cuts the schedule to 450 minutes and lifts OEE to 352.8 ÷ 450 = 78.4%. Nothing improved on the line.
4. Good units in the performance factor
Performance uses all units made. If you use the 14,112 good ones, performance reads 84.0% instead of 87.5%, and quality has already counted the same bad units. OEE falls to 0.875 × 0.84 × 0.96 = 70.6%. See how to calculate OEE performance.
5 and 6. Averages and caps
A plain average of machine OEE values ignores how long each machine ran. Multiplying averaged factors is worse. See how to average OEE across machines.
If performance is over 100%, the ideal rate is too low or the counter is wrong. Hansen caps the speed factor at 1.0 and notes any overspeed made to catch up. On a machine that is not the bottleneck, the ideal rate is the rate needed to feed the bottleneck.
7. Stops written from memory
A stop log filled in at shift end is short. People forget the small ones. The missing minutes stay inside run time, and show up as low performance, so the loss is there but with no cause.
Hansen's checks help: compare the reported uptime with hour meters or cycle counts, and compare the transferred count with the sum of crew reports. Where the two disagree, trust the record that cannot forget. See OEE data collection methods.
Our tip: when you first log stops properly, expect OEE to drop for a few weeks. The loss was always there. Tell people before it happens, so the lower number is read as better data and not as a worse line.
8. An "other" code
Hansen's rule is to categorise every minute and never to use "miscellaneous" or "other". An "other" bucket grows until it is the biggest loss, and it cannot be fixed because it has no cause. See the OEE loss tree for a category set that covers every minute.
9. No reconciliation
OEE plus all the loss percentages must equal 100% of scheduled time. In the example: 352.8 good minutes, 14.7 of bad units, 52.5 of slow running and 60 of stops add to 480.
If your sum is 470 or 495, a loss is missing or counted twice. This check takes one minute and catches most of the errors above.
10. Finishing early counted as a loss
Hansen's principle is that good performance must never hurt OEE. If a run finishes early, the time freed is excluded time, not downtime. Otherwise a crew that works well gets a worse number.
A five-minute audit
- Method 3 against Method 1. Good units ÷ ideal rate ÷ scheduled time must equal availability × performance × quality.
- Losses to 100%. Add every loss in minutes. The total plus good output time equals scheduled time.
- Output check. Does OEE move with what you shipped this week?
- Rules check. Is the ideal rate accredited, is excluded time planned and are rework units out of good units?
When the three methods disagree, Hansen's advice is to assume the lowest value is correct. To go deeper into the basics, read how to calculate OEE and availability, performance and quality.
Most errors start with stops written down by memory. Fabrico takes stops live from the machine, and operators tag the cause at the line, so the minutes in your OEE come from the clock and not from the shift report. See how Fabrico logs stops.
What people ask next
How do I know if my OEE is inflated?
Test it against output. Hansen's case study found online OEE tracked actual output with a correlation above 0.9. If OEE rises and output does not, check the ideal rate, the excluded time and the good-unit count first.
Which of the three OEE methods should I trust?
Hansen says the three methods must agree. If they do not, take the lowest value as correct: the others have missed a loss. Method 3, good units ÷ ideal rate ÷ scheduled time, is the simplest cross-check. See how to calculate OEE.
Is it a mistake to exclude lunch and breaks?
Not if they were planned as not for production, for example planned lunches with no relief crew. It is a mistake to exclude time that just happened to be quiet. See planned vs unplanned downtime.
Why is my OEE lower after I started logging stops?
Because you started finding them. Stops that nobody wrote down were hiding in performance, or were never seen. The number is now closer to the truth. See micro-stops and how to track downtime.
Can a calculator make these mistakes for me?
It can calculate correctly from bad inputs. A calculator such as the OEE calculator only checks the arithmetic. The inputs, meaning the ideal rate, the good count and the scheduled time, still need the rules on this page.