What is OEE used for?
The simple answer is “Improvement”. OEE is a measure of improvement and is used as part of the improvement cycle. Unfortunately, there is a lot of talk about 85% ‘World Class Standard’, an arbitrary target found in the original TPM literature. This target is not only out of date (Nissan in Sunderland is running weld lines at 92-93% OEE), but it gives the wrong message. A customer has no interest in your OEE; that is an internal measure that relates to its efficiency and costs. The customer is much more interested in a measure like On Time In Full (OTIF), ie did I receive my order? Running a manufacturing business with an arbitrary measure of efficiency instead of a measure of customer satisfaction is a recipe for disaster. The best use of an OEE target like 85% is to recognize that if you’re reaching that level and the customer still isn’t getting their orders on time, you may have capacity constraints.
OEE doesn’t tell us if we have a problem, the customer does. What OEE does is help us analyze the problem and make improvements. That’s why Toyota uses it as a spot measure on a particular machine where there is a capacity or quality issue. Calculating the OEE of anything but a discrete machine or automated line is pointless; we have much better measures of the efficiency of a factory or department as a whole.
OEE grew out of the need for improvement groups to have a way to measure and analyze equipment issues as part of their Define, Measure, Analyze, Improve and Control cycle. OEE defines the expected performance of a machine, measures it, and provides a loss framework for analysis, which leads to improvement. It can then be used as a follow-up measure to see if the improvement is sustained, that is, if the control is sufficient.
What does OEE measure?
In its simplest form, OEE measures the availability, performance, and output quality of a machine.
A machine is available if it is ready to produce, instead of being broken down or having to make some changes or adjustments. The availability definition allows for planned maintenance, when the machine is not intended to be available for production, but does not account for changes etc. No machine with changes can be 100% available. The reason for taking such a hard line is that changes are a huge loss to both efficiency and flexibility, so the OEE analysis focuses attention on this by not making allowances for changes.
Throughput efficiency measures the output over the available time compared to a standard. There may be debate here about what the standard output should be. A good rule of thumb is to perform the performance calculation based on the best known performance. This can be higher or lower than the design speed. My argument is that if a machine has never reached its design performance, there is no use measuring it. On the other hand, if you have consistently exceeded design specs, you can (and have seen) 140% performance numbers, which can mask poor availability. This is always to remember that one of the purposes of OEE is to help you know if you have the capacity to meet customer demand.
Output quality is a measure of first time: what percentage of the output was correct the first time, without rework. FTT measures are always the best quality measures. The problem with OEE is that sometimes quality feedback is not immediate. At FMCG business, a customer complaint may be received three months or more after production. In these cases, it is better not to include quality in the OEE calculation and to use a more customer-focused measure of quality: number of complaints, etc. If there’s no way we can use the quality component of OEE in a real-time improvement cycle, then there’s no point in measuring it.
loss analysis
The next level of analysis is the seven (or six or eight or sixteen) losses. Within OEE, we generally talk about seven losses, although TPM loss structures are known to define 23 total losses.
Availability losses are mainly breakdowns and changes. The changes can be separated into tool changes, material changes, and reduced performance at startup, but they are essentially the same problem. A more detailed analysis reveals that failures have two fundamental types, those due to deterioration due to inadequate maintenance and those due to the inherent characteristics of the machine.
This gives us three basic answers to availability problems: improve changes through SMED, improve basic maintenance, and improve machine features. Depending on the Pareto loss analysis, we may need to act on one, two, or all three.
Losses in performance are generally divided into loss of speed and minor shutdowns: does the machine run slowly or does it stop and start? The definition of a minor stoppage is also open to debate: originally it was less than ten minutes, then five minutes, then three minutes. The pragmatic approach is to say that if you can measure the amount of time lost by a downtime, it’s a breakdown, not a minor downtime. If you can only record the number of stops, then they are minor stops.
There is some practical use for the minor speed/stall distinction: if a machine is running slowly we can always speed it up, whereas if it stalls we should look at the physical mechanism and try to eliminate the cause of the stalls (my favorite example is where we find that the root cause was when metal washers were loaded into a hopper with a metal scoop, which damaged some, which then jammed the feed – a plastic scoop was the fix!).
However, we can also make a useful distinction between performance losses due to deterioration or contamination and those caused by the inherent characteristics of the machine. As with breakdowns, this gives us two approaches to improvement: better maintenance or equipment redesign.
gets better
The only reason to measure and analyze something is to improve it. If we are not going to use the entire improvement cycle, it makes no sense to measure the OEE. It doesn’t tell us anything we don’t already know. On a raw level, all OEE tells you is how much you made compared to what you wanted to do, and any measure of schedule compliance would already tell you that. Averaging OEEs for entire plants or time periods simply hides problems – OEE is a specific measure to use for specific improvement projects.
The biggest misuse of OEE is using it to compare different processes, plants, or machines. OEE is not a useful executive KPI. It’s not even a very useful operational measure. It is an improvement measure, for people who want to improve the performance of their equipment.
How to massage your OEE
1) When the machine breaks down, please record it in planned maintenance
2) Make changes during scheduled maintenance or on weekends, otherwise 24/7
3) Use an easy performance standard
4) Measure the best machine and quote that figure
5) Set arbitrary goals and achieve them through the above
Using the above strategy, you should be able to report decent OEEs and even earn some money if the pay is related to OEE performance. What this will not do, however, is improve your ability to meet customer demand.
How to improve performance
1) Measure against customer demand (OTIF or similar)
2) Measure OEE on restrictions or problematic equipment
3) Set realistic performance standards
4) Analyze losses to identify improvement problems
5) Use the entire improvement cycle