Article :: Not All That Can Be Counted Counts

Not All That Can Be Counted Counts

Measuring performance of equipment and labour is very important but having said that, you have to put some thought into the measurements. Just measuring everything that you have on the shop floor might give you a feeling of being on "top of everything" and having your finger on the pulse of the shop, but that may not be true. In fact, it can often be counter productive.

Recently I attended a seminar where the speaker presented a software package to track "Overall Labour Efficiency" and "Overall Equipment Efficiency". The software would track the running time of each equipment through sensors, and using other information like the number of widgets produced in a given time, etc, would then calculate the "real efficiency". Same thing for labour efficiency. To say the least, I was shocked that in spite of all the literature on lean manufacturing and constraint management, people are still listening to this kind of stuff with a straight face. If this was not such a dangerous idea, I would have laughed out loud right there.

Let me explain. To make things as simple as possible, let's say you have a process where the product is worked upon at station A, B, C, and D. Each of these stations has one person working on one piece of equipment that adds value to the job and moves it on to the next station. What the software was designed to measure is the efficiency of each worker and each piece of equipment at each station, in terms of the number of pieces worked upon in a given period, the amount of time that the equipment and worker worked as a percentage of the number of hours that were put in by the worker and the equipment, etc. You get the drift.

However, here is the problem. In most shops, except sections of some assembly lines, station A, B, C and D will have different capacities and therefore different outputs. So what's wrong with that? The problem is that since they have different outputs, one of them has a lesser capacity than all the other three. Depending on what work is being done on what job, one of them will be a bottle neck and the other three will be either racing ahead or held back by the bottle neck. So what?

Well, let's say the bottle neck is station B. The overall labour efficiency measurement will show that the station before the bottle neck (i.e. A) is more efficient than the bottle neck and the rest of them are inefficient in spite of the person and the machine trying their very best. The bottle neck will not be able to supply them enough parts to work on so they will look inefficient. And the station before the bottle neck, A, will be racing ahead producing more parts than B, the bottleneck, can handle. So what?

There are two problems in the above scenario. The first is that the decision maker in his or her nice comfortable chair is looking at the beautiful charts, probably a few days later, produced by the software and is concluding that they need to buy a faster machines for stations C and D or those workers need more training. The chart shows that A and B are working at their full capacity and C and D are slacking off. However, there is another bigger problem. Station A is producing work in process inventory which means creating stacks of parts (actually cash) in front of station A and you will probably need to deploy a fork lift to move them around. And station A will probably receive a bonus for doing this. And what about the accountants looking at returns on investment in each equipment? Doesn't give them any workable information either.

As you can see from this simple scenario, measuring every person and every piece of equipment all the time leads to erroneous decision making and providing the wrong incentives to workers. Instead of measuring local efficiencies by measuring the efficiencies of each worker and each piece of equipment, we must focus on global efficiencies i.e. the efficiency of the whole process. Anything else will be a waste of money and time. Also, what is the use of spending money on systems and then spending time analyzing the outputs from those systems only to come up with questionable or misleading inferences?

The real problem is the perception that if everybody on the floor works efficiently the plant is efficient. I wish this was true. As shown in the simple example above, non-bottlenecks working efficiently can actually make the plant less productive. So the first step is to look at the whole plant, from order entry to shipping and make the entire process work in lock step, instead of making bits and pieces of the plant efficient. In other words, measure plant wide performance and not each part. It's the difference between looking at the thousands of different parts of a car separately to understand how it works, instead of looking at how they work together to move the car.

In lean manufacturing terms, there is the concept of takt time which relates to being in lock step with the customer's demands. If you consider that each subsequent process is the customer of the previous process, then all the process have to be following takt time too i.e. be in lock step with each other as opposed to focusing on individual efficiencies.

There is another issue that has not been given enough thought. Measurements are a form of communication so one must understand what message a particular measurement is sending to the shop floor. If you are measuring the performance of a machine, the message is loud and clear that you want the machine to produce as many units as possible. For a non-bottleneck, this could be a disaster.

In conclusion, I would definitely recommend that you measure performance but think through what you are doing and why. Think how you are going to use the measurement and what kind of message you are conveying to the shop floor personnel with those measurements. Consider who else is going to be looking at the results and what might they infer from them. And remember, you do not have to measure everything that moves on the shop floor and create a data hay stack that makes little sense to anybody. Like Einstein said, not all that can be counted counts.