Line OEE = good units out of the last machine ÷ the bottleneck's ideal rate ÷ scheduled time. Machine OEE is the same calculation for one machine, on its own count and rate. Do not average or multiply machine numbers to get a line number. Use the line figure for results and the machine figures to find causes.
Two numbers, two jobs
Machine OEE tells you how one machine ran. Line OEE tells you what the whole line delivered. They answer different questions, and one cannot be built from the other by simple maths.
The reason is flow. In a line, every unit goes through every machine in turn. The output of the line is limited by its slowest step, and a stop at one machine stops others through waiting or blocking.
This is Method 3 from Overall Equipment Effectiveness (R. C. Hansen, 2001), applied at line level. You need only three things: scheduled time, good units and the ideal rate. For a line, the units are counted at the end and the rate is the bottleneck's.
Worked example: a three-machine line
Example: a filler, a labeller and a packer run in series for a 480-minute shift. The ideal rates are filler 100 per minute (the bottleneck), labeller 125 and packer 110. The line delivers 33,600 good units.
| Machine | Own ideal rate | Good units ÷ rate | OEE on own rate |
|---|---|---|---|
| Filler | 100 per min | 33,600 ÷ 100 = 336 min | 70.0% |
| Labeller | 125 per min | 33,600 ÷ 125 = 268.8 min | 56.0% |
| Packer | 110 per min | 33,600 ÷ 110 = 305.5 min | 63.6% |
| Line | 100 per min (bottleneck) | 33,600 ÷ 100 = 336 min | 70.0% |
Every machine made the same 33,600 good units, because that is all the line passed on. The labeller and packer show low OEE only because their own rates are higher than the pace of the line. They did nothing wrong.
The average is 63.2% and the product is 0.70 × 0.56 × 0.636 = 24.9%. Both are wrong for the same reason: the machines are not independent.
The rate rule for machines inside a line
Hansen's fix is to give a machine that is not the bottleneck an ideal rate equal to the rate needed to feed the bottleneck, and to cap its speed factor at 100%. Overspeed is allowed only to make up scheduled output, and it should be noted so the stock can be reconciled.
Apply that to the example. The labeller and packer both get an ideal rate of 100 per minute. Their OEE is now 70.0%, the same as the line.
Their lost time is real, but it has a different name. Suppose the labeller ran for 336 minutes and waited 144 minutes for product from the filler. In Hansen's categories the waiting is a stop induced from outside the machine, caused by upstream. The root cause sits at the filler.
Using machine OEE to find causes
Line OEE tells you the result. To act, break the bottleneck's losses into categories. Example: the filler's 480 minutes split as follows.
Now there is something to do: the stops are two thirds of the loss. See the OEE loss tree for how to sort them, and how to track downtime to capture them.
Buffers change the picture
If a buffer sits between two machines, the one behind can keep running for a while after the one before stops. The line then loses less than the sum of machine stops. This is why you cannot add machine stop times to get line stop time.
It also means a short stop at a non-bottleneck may cost the line nothing, and a short stop at the bottleneck always costs. Our advice: log stops everywhere, but rank them by what they cost the line. Hansen's test is the same: if this step improves, will the bottom line be significantly affected?
A last point on blame. A machine OEE chart on a wall can make the people at the labeller look bad when the cause is two machines upstream. Show the waiting time as its own loss, with the machine it waited for, so each crew sees what is theirs.
How to set it up on the floor
- Fix the counting point. Count good units after the last place where rejects leave the line. If rejects are pulled out in the middle of the line, the end count is what matters.
- Name the bottleneck for each product and write down its accredited rate. This is the line's ideal rate for that product.
- Use one scheduled time for the line. It is the time the line was planned to run, with the same excluded-time rules at every machine.
- Log stops at every machine, including waiting for upstream and downstream, so you can see where a wait started.
- Report the line figure each shift, and the bottleneck's loss breakdown beside it.
If several products run in one shift, compute the line OEE per run and combine the runs weighted by scheduled time. Hansen notes that the slowest step can differ between products, so the line's ideal rate can change from one run to the next.
What to track, and when
| Question | Use |
|---|---|
| What did the line deliver this shift? | Line OEE, at the end of the line |
| Where is the line losing the minutes? | Bottleneck machine OEE and its loss breakdown |
| Why is the labeller idle? | Waiting categories at the labeller, then the upstream cause |
| How is the plant doing overall? | Scheduled-time weighted average, with bottlenecks shown separately |
For packaging lines in particular, see packaging line troubleshooting. For the maths of combining numbers, read how to average OEE across machines. For the basics of the formula, go to how to calculate OEE.
You need the line result and the machine detail side by side. Fabrico shows OEE per line and shift, with stops tagged by the operator at the line, so a line loss can be traced to the machine and the cause. See OEE per line in Fabrico.
What people ask next
Is line OEE the average of the machine OEEs?
No. A simple average ignores which machine sets the pace and how long each ran. Measure the line at its output, against the bottleneck rate. If you must combine numbers, weight by scheduled time. See how to average OEE across machines.
Is line OEE the product of the machine OEEs?
No. Multiplying gives a number far too low, because the machines share the same stops and waits. In the example above the product is 24.9% while the line really ran at 70.0%.
Which machine do I measure first?
The bottleneck. Hansen's advice is to apply OEE to bottlenecks first, then to critical or high-cost areas, and to ask whether an improvement there will move the bottom line. See bottleneck analysis.
What ideal rate do I use for a machine that is not the bottleneck?
The rate needed to feed the bottleneck, with the speed factor capped at 100%. That keeps machines from reporting a low OEE just because they are faster than the line. See ideal cycle time.
Can the bottleneck change?
Yes. It can move between products, formats and even shifts. Re-check which step is slowest for each product, and use that rate as the line's ideal. The line balancing calculator helps you compare step rates.
Related guides: OEE for Parallel Machines on One Process Step · V-Graph: Setting Speeds Around the Key Machine