We found no standard that sets one number: you choose a threshold, and stops shorter than it fall into performance as minor stops, while longer ones go into availability as downtime. Moving it shifts the loss between those two factors. OEE stays the same, as long as performance is calculated from the unit count.
A setting that looks like it matters
People argue about this for weeks: should a stop count as downtime after 2 minutes, 5 minutes or 10? Here is the surprise, on one example shift.
Example: a labeller has 480 scheduled minutes, an ideal rate of 60 units per minute, and makes 20,160 units with 19,800 good. Its stops that shift were:
- 1 stop of 40 minutes
- 1 stop of 15 minutes
- 3 stops of 6 minutes (18 minutes)
- 8 stops of 3 minutes (24 minutes)
- 40 stops of 30 seconds (20 minutes)
That is 53 stops and 117 minutes. Only the threshold changes below, and every unit and every stop stays as it was.
| Threshold | Stops / min logged | A | P | OEE |
|---|---|---|---|---|
| Every stop | 53 / 117 | 75.6% | 92.6% | 68.8% |
| 1 min | 13 / 97 | 79.8% | 87.7% | 68.8% |
| 5 min | 5 / 73 | 84.8% | 82.6% | 68.8% |
| 10 min | 2 / 55 | 88.5% | 79.1% | 68.8% |
| 30 min | 1 / 40 | 91.7% | 76.4% | 68.8% |
Why OEE cannot move
OEE = good units ÷ ideal rate ÷ scheduled time. Here that is 19,800 ÷ 60 ÷ 480 = 330 ÷ 480 = 68.75%, and the threshold is not in that sum.
Run time sits on top of availability and on the bottom of performance, so it cancels out. The proof is in time-based vs count-based OEE.
So the threshold never changes how good a shift was. It only decides which factor carries the loss, and who is asked to explain it.
When the reported OEE does move
The sum above assumes performance comes from the unit count. Two report setups break that, and in both the fault is in the report.
- Performance from running speed. If the system reads speed only while the line runs, stops under the threshold land in neither factor. At 5 minutes that gives 84.8% × 92.6% × 98.2% = 77.1%, and 44 minutes of short stops have vanished.
- A capped factor on a soft rate. Say the rate on file is 45 per minute, so raw performance is over 100% in every row. With the cap on, OEE reads 74.3% when every stop is logged and 83.3% at 5 minutes.
Test your own report: change the threshold on last week's data. If OEE moves by more than rounding, look at how performance is calculated.
What the threshold really changes
If OEE does not change, why care? Because three other things do.
- Logging work. On this shift, operators give a reason for 53 stops, 5 stops or 1 stop.
- Causes you can see. At 30 minutes, the 15-minute stop has no reason code, and neither do the three 6-minute stops.
- Who owns the loss. Availability tends to go to maintenance, while performance often goes to operations. The threshold can move a fight from one meeting to another.
The difference in kind between the two is in minor stops vs breakdowns, and the loss class is in the six big losses.
What number should you pick?
This site defines a micro-stop as usually under five minutes, and that is a sensible place to start. Your line may need another number, and these steps help you find it.
Use a machine signal or a person with a stopwatch for one week. Count stops and minutes in each length band.
For each option, how many reason codes per shift? On this shift: 53, 13, 5, 2 or 1.
Ask what they can log during a shift and still run the line. An unused rule gives no data at all.
Same number across a plant. If a line needs a different one, say so on its reports.
With machine signals, you can set the threshold low, as the short stops are captured whether anyone logs them or not. With paper, a low threshold fails, as covered in operator downtime logging.
Our advice for paper is a threshold where the line gets about 5-15 reason codes per shift. That is our rule of thumb, so test it with your own crew.
What to write in the rule
A number alone starts arguments later, so put these five items on one page and give a copy to each shift lead.
- The number. For example, a stop of 5 minutes or longer gets a reason code.
- Start and end. The clock runs from the last unit before the stop to the first unit after it.
- The merge rule. How short a restart must be before two stops count as one.
- Where the rest goes. Shorter stops stay inside run time and show up in performance.
- The date. When the rule started, so a step in the trend can be explained.
Keep the short stops visible anyway
A stop under the threshold still happened. In this example, the 40 stops of 30 seconds took 20 minutes, and they are probably one cause repeating.
So record how often the line stops as well as how long. Mean time between stops is the measure for this, and it falls straight out of the count.
Group the short stops by station, not by operator memory. The fixes are in how to reduce micro-stops, and the sensing options are in machine state detection.
The traps
Comparing lines with different thresholds
Line 1 has a 1-minute rule and reads A = 79.8%, while Line 2 has a 10-minute rule and reads A = 88.5%. In this example they are the same shift, so the 8.7-point gap is purely a setting.
Changing it in the middle of a trend
Lower the threshold and availability drops the same day, with no change on the floor. Note the date on the chart, and tell people first.
Long stops broken in two
A machine that restarts for a few seconds and stops again can turn one 12-minute stop into two stops under 10 minutes. Our advice: merge stops that are separated by very short restarts, for example under 30 seconds.
Using it to manage the score
Raising the threshold makes availability look better, and that may please a team measured on it. Measure teams on OEE or on the full loss list, so nobody gains by moving the setting.
What stays the same no matter what
- OEE itself and the good count, when performance comes from the count.
- Net operating time, because it comes from the count and the ideal rate.
- TEEP and all output figures.
Try it yourself by moving stops between the two factors in the OEE calculator. For the way to code what you log, see downtime reason codes.
With short stops, the hard part is the cause and not the threshold. Fabrico takes stops live from the machine, operators tag the cause at the line, and where cameras are fitted it shows video of each stop, streamed on request next to its machine data. See how Fabrico shows short stops.
What people ask next
How long is a minor stop?
This site uses usually under five minutes, often under one, cleared by the operator with no repair. Your threshold can differ, as long as you apply it the same way across lines. See micro-stops.
Does a stop threshold change OEE?
No, if performance comes from the unit count. OEE is good units ÷ ideal rate ÷ scheduled time, and the threshold is not in that sum, so it only moves the loss between availability and performance. If your OEE moves anyway, the report is using running speed or a capped factor.
Is a 5-minute threshold standard?
We found no standard that sets one figure. Five minutes is a usable place to start, but the right number depends on how you capture stops and how many reason codes operators can really give.
Should short stops count in availability?
They can, if you can capture them all. With machine signals, some plants log every stop, so availability carries the full loss and performance shows only pure speed. See PLC signals for OEE.
How do I find the short stops if they are not logged?
Compare run time with the ideal time for the units made, since the gap holds slow running and short stops. Then watch the line, or add a sensor, and see conveyor short stops for a common source.
Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.
Related guides: Speed Loss Calculation in OEE: Step by Step