A good machine is not merely one that does not break down, one that runs the moment you switch it on. It must be a machine that, once started, works at full effectiveness — running at its full rated capability. But if a machine is available all the time and runs at full power, and the pieces it produces have no quality, then none of it is of any use.
The quality of what comes out is therefore a further factor in judging a machine — and, importantly, a good machine must be safe to work with.
OEE is well known as the number that states the performance of a plant whose production process rests on its machines. It is also the number used to measure the success of a plant running TPM — or, to put it another way, TPM is carried out in order to raise OEE.
Availability
Total Time means the time we have the machine in the plant. That does not mean we must plan to use the machine for all of it. We will need stoppages for daily maintenance, stoppages for briefing meetings, stoppages for the plant’s own activities such as 5S. All the stoppages we intend are called Planned Shutdown. So the time over which we want the machine to be available is not the total time.
Loading Time means the time planned for production: total time less planned shutdown. It is this loading time over which we want the machine to run without interruption.
ExampleA machine has a total time of 48 hours a week. Over one week it has 6 hours of planned shutdown and loses 3 hours to breakdown. Find its availability for the week.
- Loading time = total time − planned shutdown = 48 − 6 = 42 hours
- Operating time = loading time − downtime losses = 42 − 3 = 39 hours
- Availability = operating time ÷ loading time = 39 ÷ 42 = 92.85%
Performance efficiency
Operating time will not equal loading time once losses stop the machine. But the losses that can occur do not end there. There are also losses that rob the machine of power, cutting an already reduced operating time further still — what remains is called net operating time.
Performance efficiency sometimes cannot be calculated directly, because some of the losses that rob the machine of power cannot be timed — a voltage dip, a machine running unevenly, a stumble or a minor stoppage. The standard time per piece solves this for us: given a standard time, we know how many pieces the operating time should have produced, and how many it actually produced.
ExampleAfter deducting planned shutdown and the losses that stop the machine, a machine is left with only 50 real hours a week. Over that week the losses from the machine running below power total 8 hours. Find its performance efficiency for the week.
- Net operating time = operating time − power-loss time = 50 − 8 = 42 hours
- Performance efficiency = net operating time ÷ operating time = 42 ÷ 50 = 84%
ExampleA machine has a standard time of 0.036 hours per piece. In one day it has 6 hours of operating time and produces 140 pieces. Find its performance efficiency.
- Net operating time = pieces produced × standard time per piece = 140 × 0.036 = 5.04 hours
- Performance efficiency = net operating time ÷ operating time = 5.04 ÷ 6 = 84%
Quality rate
Net operating time does not always create value throughout — meaning the production of good, quality pieces — because part of it is spent producing defects, what we call the loss from producing defective work.
The quality rate sometimes cannot be found from that equation either, because of the difficulty of timing what is lost to producing bad work. But we can look at the loss in the form it takes: the pieces that are scrap and the pieces that must go back for rework.
ExampleA machine has 40 hours a week of running with no losses at all while working — that is, net operating time of 40 hours. But the periods in which pieces came out defective or had to go back for rework total about 2 hours. Find its quality rate.
- Valuable operating time = net operating time − defect-production loss = 40 − 2 = 38 hours
- Quality rate = valuable operating time ÷ net operating time = 38 ÷ 40 = 95%
A second example: a machine produces 300 pieces in a day. Of those 300, 45 are defective beyond repair and 15 can be sent back for rework. Its quality rate for that day is (300 − (45 + 15)) ÷ 300 = 80%.
Arriving at OEE
In what has gone before we have discussed availability, performance efficiency and the quality rate, which amounts to having built the components of OEE in advance. So at this point it only remains to put them together.
| Process | Availability | Performance efficiency | Quality rate | OEE |
|---|---|---|---|---|
| A | 100% | 50% | 100% | 50% |
| B | 90% | 90% | 90% | 72.9% |
| C | 70% | 85% | 99% | 58% |
From the example it can be seen that process A has availability of 100% and a quality rate of 100%, but once performance efficiency of only 50% is taken into account the OEE is left at just 50%. From this case we can read that process A has no problem with breakdowns or stoppages of any kind, and no quality problem either — but the process is running very slowly, at only 50% of standard capacity.
Process B looks as though its OEE ought to come out high, since all three factors are in the high range. But the OEE that actually comes out is 72.9% — because the machine still cannot run throughout, losing 10% to stoppages; it still does not run at full power, short by another 10%; and it produces defects in a quantity as high as 10%.
Process C, although it has no quality problem, breaks down often and still does not run at full power. Yet it is worth noticing that its OEE is still higher than process A’s. This is because it is the lowest variable that drags the OEE down.
So in improving OEE one should improve the lowest variable first, because it has the greatest effect in raising the figure — and because it is easier than pushing a variable that is already high higher still.
Losses from the machine
Loss is what makes OEE low, and reducing loss is the only way to raise OEE. To do that we must know how to separate losses into groups — so that they can be carried into the OEE calculation, and so that improvement afterwards lands on the right point.
Shutdown Loss is any event which, once it occurs, stops the machine — a belt snapping, a die change, an accident. Losses in this group take a long time to put right once they happen, and this is the group that makes availability low.
Capacity Loss is any event which, once it occurs, makes the machine produce more slowly without the machine having broken down — a minor stoppage waiting for an operator, waiting for material, the machine losing speed, or the start-up period in which it has not yet reached full speed. This group is what makes performance efficiency low.
Yield Loss is any event which, once it occurs, costs production time with quality as its cause. Whether the piece cannot be used at all, or can be used but must go back for rework, it counts as loss. This is the group that, once it occurs, brings the quality rate down.
From what has gone before, try to think through your own plant: which events there count as losses — a power cut, defective work, a leaking boiler, a broken bearing, rework, machine set-up, and so on. Then try to work out which loss group each of the events you meet routinely falls into.
The six big losses
The six big losses are met routinely in almost every plant. But that does not mean every plant has only these six. There may be more or fewer depending on the establishment. The six big losses are therefore not a ready-made formula for finding losses, but a guide for applying the idea. The best method is to find our own losses and then divide them into groups.
| Loss group | The loss |
|---|---|
| Shutdown Losses | 1. Breakdown · 2. Set-up and changeover |
| Capacity Losses | 3. Minor stoppages · 4. Speed loss · 5. Start-up |
| Yield Losses | 6. Defects and rework |
If we cannot find the losses, we cannot find where to improve. But finding the losses without sorting them into groups means the improvement does not land on the right point. So the best method is to find all of the losses, and to be able to group them — without being bound to the six big losses.
How to apply it on the floor
- Agree the definition of planned production time with finance before measuring anything
- Automate capture on the constraint asset first, not across the whole plant
- Separate minor stops from breakdowns — they are fixed in completely different ways
- Convert each loss into money per month, then rank
- Review the top loss daily at the line, not monthly in a meeting room
The mistakes we see most
- Excluding changeover or ramp-up from the base, which raises the number without improving anything
- Tying OEE to a bonus, which produces overproduction to push the figure up
- Capturing downtime by hand and concluding there is no minor-stop problem
- Comparing OEE across lines whose work is not remotely comparable
Common questions
What is a good OEE?
The world-class reference is 85%, but your own trend matters more — and so does everyone trusting the same measurement. An honest 55% is more useful than a dressed-up 80%.
Do we need to buy a system to measure OEE?
No — start with paper at the machine. It will under-record minor stops, so automate the constraint asset when you are ready.