Metrics are important to measure, track, and compare your progress. Most metrics such as time (“how fast?”), item count (“how many?”), or mass (“how much?”) are very well defined and grounded in the basic laws of physics. But how to measure more abstract phenomena? Over the years, industry (and an industry around it) has defined a series of metrics, some more useful than others, and constantly subject to changes based on whatever ideology is currently in fashion.
Overall Equipment Efficiency (OEE) expresses the ratio between actual output and the maximum output that is theoretically possible:
OEE = Actual Output / Theoretical Maximum Output
Let’s say a copy machine can copy 1000 pages an hour. It is available for eight hours a day and is used to print 2000 pages every workday. It’s OEE is therefore 2000/8000 or 25%. This number might be a good indicator of your employees’ productivity, yielding the same unit-less quantity for every machine. Let’s say you own another copy machine, which can print 500 pages an hour. It is also available for eight hours a day, but only used to print 1000 pages every workday. It’s OEE is 1000/4000, that is also 25%. One day, the fast copy machine only prints 800 pages, reducing its OEE to 800/8000 or 10%. Measuring OEE instead of just the actual page count allows you to skip the question “What is it’s theoretical maximum output?”, to see right away that something must have changed.
This simple example highlights some key properties of the OEE. First, the number in itself is absolutely meaningless. There is nothing anyone can say about your business, knowing that you are using your copy machine at 25% OEE. While it gives you a pretty good indication what happens at an average work day, it clearly is not the right number to track your employee’s productivity. They can just copy more, also known as overproduction. Instead, the OEE can tell you when things are different. For example, a lower OEE could mean that there is less business that day or that some employees did not show up for work.
Second, OEE gives you an idea about capital efficiency. A copy machine that makes copies faster is usually more expensive. A 25% OEE suggests that you are probably operating a too expensive model that is not fully used. This is important as the machine has probably be priced assuming an OEE much closer to 100%. Yet, a faster machine also reduces waiting, which might be well worth it. Similarly, a very high OEE, for example 90%, tells you that the copy machine has become a bottleneck, and it is likely that employees are queueing up to use it.
These examples show that OEE is an important quantity that allows you to access key performance indicators of your business at a glance, but should not be used out of context.
OEE in the Line
The different properties of OEE are also illustrated in the manufacturing line above. Station A is producing work pieces at a rate that is higher than Station B can process it. Station B is working at maximum capacity, but outputs at a much lower rate than Station C can handle, leading to over production, excess inventory and waiting. You can see here that high or low OEEs are neither good or bad, nor have to do exclusively with the machine itself. Instead, given the WIP that is building up at Station B, an OEE of 60% is already too high. Also, the low OEE of Station C is neither the machine’s or the operator’s fault, but solely due to Station B not dimensioned correctly.
Implementing a pull system would allow solving the over production and inventory problem, but not the waiting at Station C. Specifically, Station B is the “bottleneck” and should determine the pace at which Station A is processing goods. Assuming that Station A is producing four work pieces when Station B produces one, would reduce the OEE of Station A to 15%. We can therefore see that the actual numbers do not matter as much as how much work-in-process (WIP) we are storing in the line. In fact, lowering the OEE helps us to reduce WIP and thereby lead-time.
In order to increase the OEE of Station C, we will need to either replace Station B with a faster model or replicate it so that the output of Station A can be processed in parallel. This quickly brings up the question of what the goal actually is, that is how many of Station C’s products are sold on a given day and how the operator of Station C is using their time when not operating the machine. Optimization should therefore never be motivated by increasing or decreasing OEE, but by working backwards from the intended customer value. Then, OEE can be very helpful to spot where a problem comes from.
Measuring OEE in terms of actual and theoretical throughput is by far the easiest and most intuitive use of the metric. So far, we have exclusively considered speed, for example the number of pages a copy machine can print per hour. The OEE is often broken down further into performance, availability, and quality.
- Performance: A metric that allows expressing whatever the machine does at a scale from 0 (zero performance) to 100% (optimal performance). A typical example of performance is speed, that is how fast the machine can process a certain quantity. This is easy to measure for a copy machine, but more complicated for a machine that does different things, such as a CNC mill. Alternatives are actual run time, amount of throughput, or even scrap. What to measure here very much depends on the application and use case and very much depends on why you are actually interested in OEE.
- Availability: The ratio of time that the machine was actually available vs. the total time it should have been available. Availability is usually measured based on the actual shifts and breaks, but can also be measured with respect to a 24/7 workday. This might make sense for machines that can be started before a break or at the end of the work day. Availability might also include (or exclude) time for scheduled maintenance. How to define availability is again dependent on what you actually want to measure and what you want OEE to indicate.
- Quality: The ratio of parts that pass quality inspection vs. the total number of parts produced. For example, if three out of a hundred parts are defect, the quality ratio is 97%. Although this sounds simple, determining quality at run-time might not always be feasible with defects only appearing further down the line when the faulty part is put to the test.
The OEE is hence defined as
As such, it provides a tool to further get down to the root cause of why the OEE of a machine has suddenly changed. If demand has remained constant, a drop of OEE might have been due to increased downtime or due to a decrease in quality. Being able to track these events independently, for example using a combination of machine tracking (availability) and manual data entry (quality), allows a line supervisor or operator to identify and fix a problem quickly.
While combining what are essentially three values into one makes it much easier to understand what it is going on, it is tempting to game OEE by trading performance, availability and quality. For example, it might be possible to increase throughput of a machine at cost of reduced quality. For example, running the machine not at 75%, but at 100% speed might reduce quality from 97% to 93%. The OEE then increases from 73% to 93%. How great this is really depends on what the actual cost of the additional defects are, and how the increased OEE affects overall throughput of your line.
Overall Equipment Efficiency (OEE) captures the relationship of what could be done and what was actually done. OEE is therefore a great way to see what is going on at a glance. While it is easy to measure what actually was done, it is difficult to determine what the theoretical optimum is, and there is no single way to do so. It is therefore not possible to compare OEE numbers across machines or plants. As OEEs also strongly depend on what happens before and after a specific station, OEEs should never be looked at in isolation. For all these reasons, OEE in itself is not a quantity that should be used as the basis for optimization or evaluation of plants, machines, or people.