Line Balancing Part 3 – OEE Caveats

When balancing a line, it is important to distinguish between idealized times without losses, and times that include all types of losses like breakdowns or missing material. The ratio between the ideal time and the real time is the OEE. This post looks at some of the problems that can happen with line balancing if an OEE is used incorrectly or differently, and is the third post on this series of line balancing. Once we have determined what OEE to use, we will look at how to use the OEE in line balancing in the next post.

What Are Our Losses – The OEE

All the times determined above are without losses (i.e., all the times for the individual tasks are assumed to be in perfect conditions without any breakdowns, missing materials, defects, or other problems). Yet the customer takt you want to meet includes losses. Hence, if your customer takt is 40 seconds between parts, and you design the system to have a cycle time of 40 seconds between parts, even the tiniest breakdown will put you behind and make you slower than the customer takt. Again, for the difference of a takt with losses and a cycle time without losses, see On the Different Ways to Measure Production Speed.

The ratio of the cycle time to the customer takt is, of course, the overall equipment efficiency, or OEE. For more details on how to measure the OEE, see my series of posts starting with What is OEE? – Definition of OEE. Using the OEE, you can transfer a customer takt time into a target cycle time, or alternatively a cycle time in a target line or process takt. Both work. I usually prefer to determine a target cycle time, since the people I work with on the shop floor are more likely to think in cycle times than in takt times.

Watch Out for How the OEE is Calculated!

Unfortunately, the OEE is often heavily fudged. Often, managers desire the OEE to be as high as possible to look good. I have even seen OEEs above 100%, meaning that the process can produce more than the theoretical maximum. Of course, this is not possible but merely the result of number fudging.

If you use the OEE to calculate the target cycle times, they would have to be reasonably correct. You have to be especially careful if some time blocks are excluded from the OEE. For example, many companies do not include planned maintenance and other planned stops in the OEE. Often, this is also done in a blanket approach.

Let’s take an example. Assume you need to produce 480 parts per day, and you have an 8-hour work day. This gives you a takt time of exactly 60 seconds. If your measured OEE is 80%, then you would need a cycle time of 60 seconds · 80% or 48 seconds. Hence, you design your system to have a cycle time of one part every 48 seconds, so that including losses you get a takt time of one part every 60 seconds.

Lo and behold, when calculating the OEE, one hour per day for maintenance was automatically removed. The one hour per day was not defined as a loss in the OEE, but completely taken out of the equation! Hence, the time basis for the OEE was not 80% for 8 hours per day, but 80% for 7 hours. The cycle time of 48 seconds still turns into a takt time of 60 seconds. However, instead of 480 parts during an 8-hour day, you get only 420 parts during a 7-hour day! Now your calculation is off! Instead of a takt time of ${\frac{28800s}{480 parts}}$ = 60 seconds, you now have a takt time of  ${\frac{25200s}{480 parts}}$ = 52.2 seconds. Now you either need a cycle time of 80% · 52.2 seconds, or 42 seconds instead of 48 seconds, or one hour of overtime every day to make ends meet.

Unless, of course, the one-hour maintenance is truly and entirely cheated. If for the OEE calculation, the one hour of pretend maintenance was removed but worked it anyway, then you have the rare instance where two errors cancel each other out. But I would not bet on it.

Overall, check if the OEE is calculated reasonably. Look especially at the total time including all losses. If you expect 8 hours per day but get only 7, then you either would have to assume a customer takt that has only 7 hours per day available, or recalculate the OEE with an additional hour of losses to get an 8-hour day.

Watch Out for Interdependence of Processes

The OEE of an individual process and the OEE of an entire system are rarely the same. A problem in one process can also stop another process through blocking and starving the other process. Hence, if you have three processes with an OEE of 80%, then the OEE of your entire system will likely be less than 80%.

In the worst case, if a problem anywhere in the system will immediately stop the entire system, then the overall system OEE will decrease dramatically. As shown in the image below, if all your processes have an OEE of 80%, then with 10 processes your system OEE will drop to 10%. With 20 processes, there will be only 1% of productive use left.

Luckily, in reality it is not that bad. Processes are decoupled through buffers. Additionally, if you measure the OEE on an existing line, it should already include the downtime due to starving (lack of material) or blocking.

You mostly have to keep an eye out for this if you buy a new machine and the manufacturer promises you an uptime of xy%. By this, of course, he means an uptime assuming that there is no lack of material, blocking, a missing operator, or whatever else the machine tool maker considers not his problem. In this case, stay conservative and assume that the OEE will not be as good as the supplier promises (and it rarely will be).

I Don’t Know My OEE!

Getting the OEE is actually not too difficult. You do not need to determine the losses in detail; you merely need the ratio of what you produced to what you could have done under ideal circumstances. For more details, see my series of posts starting with What is OEE? – Definition of OEE.

Of course, this does not help if you do not yet have a system to measure. If you create a new production system, you simply don’t know your OEE (yet). The solution is easy: You need an expert estimate, also known as a wild guess by someone who has at least some familiarity with the system. Look at other similar systems in your company. Your new system will probably behave similarly. Or you could pay me a lot of money and I could tell you for flow shops it is around 80% for manual processes and around 60% to 80% for machines; for job shops it is around 50% to 70% for manual processes and 40% to 70% for machines – terms and conditions may apply.

Summary

You always have to be careful when using an OEE. Depending on how the OEE is calculated, it can mean something completely different from what you think it does. In the next post we will actually use the OEE. Until then, stay tuned and organize your industry!

Line Balancing Series Overview

1. Data Overview: List of products, list of tasks, and customer takt
2. Duration of Tasks: How to get the duration of tasks, especially if they differ among products
3. OEE Caveats: Potential problems when using the OEE to transform takt times to cycle times
4. OEE Usage and Flexibility: Once you have the OEE, how do you use it? Also a bit on flexibility.
5. Balancing using Paper: Finally, actual line balancing using paper
6. Tips and Tricks for Balancing: Some Pro-Tips, and also a bit about line balancing using computers

4 thoughts on “Line Balancing Part 3 – OEE Caveats”

1. Harvey Brandt says:

Knowing the OEE is important for scheduling to meet customer TAKT time. If the OEE can be increased, more work can be scheduled at the work center or machine.

Knowing what the bottlenecks to a higher OEE is useful to enable continuous improvement. I have found it productive for an operator to log down time, analyze the data for trends, and conduct continuous improvement exercises to reduce this down time. Reducing bottlenecks, and creating more up time gives better OEE, and more output from the work center or machine.

2. Radovan says:

Decent post, nice way to connect OEE to Cycle time, I had a lot of troubles explaining to some people that if you improve OEE by 50%, your cycle time didn’t automatically got worse by 50% 🙂
However, if you talk about line balancing, you should maybe explain more on how to properly use buffers and decoupling, since their effective use might change 20% OEE plant with 4 connected machines into 80% OEE plant. Also a side note.. you can’t measure OEE on manual processes, it is not Equipment anymore 🙂 Proper name would be O People E..

3. Hi Radovan, thanks for the input. Basically, I agree with you. I talked extensively about buffering and OEE (as well as OPE) in other posts. (See e.g. my list of all posts for OEE or my glossary). Since this line balancing is already a six post series I did not want to make it even bigger. Also, to my knowledge OPE is usually changed from “overall PEOPLE effectiveness” to the more politically correct and less union-disturbing “overall PROCESS effectiveness”.