# On the Different Ways to Measure Production Speed | Christoph Roser

“There are many different ways to measure manufacturing speeds. Depending if you need the losses included or not, if you want parts per time or its inverse or only a time, single processes or entire systems, actual or current values, you may have a completely different number. This post will help you to sort out what is what…”

The main conclusion from this post is that, when discussing production speed, you should define your terms if you want to avoid confusion.

It is a useful and well researched article, but there are a few points I would make differently:

1. OEE is a ratio, not a difference.
2. “Time” can mean duration or timestamp. I would use both terms to avoid ambiguity..
3. I would discuss the practical implications of using time per part versus parts per unit time. When you say you produce 1 part every minute, it usually means that one part is completed every 60 seconds exactly. When you say you make 60 units/hour, on the other hand, you make no difference between completing one every minute and having a batch of 60 coming out every hour.
4. Averages are additive; medians or maxima are not. The average of a sum is the sum of the averages, but this is not true of the other statistics. This is of vital importance when adding up operation times in a process, and a key reason why ERP systems plan based on absurdly long lead times.
5. I prefer to define a parameters by its intended meaning, and the formula then becomes the way you estimate it. The intent of the takt time of a line is to be the interval between the completion timestamps of two consecutive units. The formula then shows how you estimate it.
6. Little’s Law applies to steady-state averages, and it needs to be said. It does not apply, for example, during ramp-up.

See on Scoop.itlean manufacturing

## 4 comments on “On the Different Ways to Measure Production Speed | Christoph Roser”

1. Many thanks for the “useful and well researched” 🙂 . I concur with your comments, and have updated my post to include points 1, 2, 4, and 6. Thanks for the feedback!

2. M. FAZLULLA

There is a much better way to measure production speed. it’s called “Manufacturing Velocity”. It’s a measure of how fast the materials under process are moving towards a despatchable state.
It’s obtained by the equation: Manufacturing Velocity = ∑ cycle time plus setup time divided by Lead time. It’s the summation of all the process cycle times(Floor to Floor time) and Setup times divided by the Lead time. The maximum manufacturing speed is numeric 1, obtained when the Lead time is slashed to match with the numerator in the equation. This ideal state is never obtained due to waste in the system. Typically the velocity in most “excellent” manufacturing systems excluding continuous process industries like refineries is 0.01.
The only job of a manufacturing executive and the CEO is to get rid of the waste in the system to always try to hit the Bulls eye the magic numeric 1 ; the highest manufacturing velocity

• Michel Baudin

A similar metric is used within the semiconductor industry, when manager say they are running at “6 times theoretical” cycle time. There are different interpretations of the theoretical cycle time. To me, it is the shortest time is could conceivably take to run one unit — a silicon wafer — through all the steps of its process. It is the result of a thought experiment: imagine a plant empty of WIP, but with all the equipment ready and set up,
with all the materials available, and with perfectly skilled people available. You then run one silicon wafer through it, and it never waits anywhere, is never inspected, and never needs rework. On the other hand, if it takes 16 hours to process a load of 200 wafers in a diffusion furnace, physics dictates that it still takes 16 hours for just one. The theoretical cycle time does not include setups, because, under the specified conditions, nothing prevents you from having the machines ready when the part arrives.

The time it actually takes in a running plant to go through the entire process is a quantity you can measure or infer, and, as a multiplier of the theoretical cycle time, it measures the impact of all the imperfections of the process. This metric is easily understood in semiconductor wafer processing, because you are running one part through 500+ fabrication steps. It is not quite as clear, for example, in assembly, and I have not heard of it being used in other industries.