I am not familiar with this item, but here is what I am thinking.
Precise optimization of a process or system is not desireable. At some point, the return on the preventive measures taken to control an item will eventually be too costly to be practical. When one does not consider this, maximization can occur and costs begin to rise. Needless to say, these costs go somewhere, some of which or all of which lands in the consumers pocket.
Like Kevin, I have only a vague idea about Taguchi's work. What little I did know has faded into the dim past but, from what I remember, it is not simple and there is no simple way to deal with it. You may have seen the discussions about 6 Sigma and noted how, out of ten people you can get ten different ideas about what it is supposed to be. Now Taguchi, I would say, was a very clever and experienced person who knew just about everything that had been written about variables. Some of it is sure to be pretty difficult for the layman but, Taguchi, having absorbed all that stuff, rolled it over in his mind for years, and teased it and worried it until he came up with a new way of looking at it. If we can't agree about 6 Sigma, what chance have we with Taguchi unless we begin with an understanding of the theory and thought process that led him to his conclusions?
I suggest you start by reading anything you can get hold of that relates to Taguchi's work. Then, having identified the areas where your knowledge isn't sufficient to take you on, go back and pick up the knowledge you need.
The worst thing you can do is drive on with a smattering of knowledge and expect to achieve anything useful. Most people can draw wrong conclusions from a histogram.
rgds, John C
Regarding the use of the Taguchi Quality Loss Function good references are:
1) Quality Engineering using robust design,
author: Madhav S. Phadke, Publisher: Prentice Hall, ISBN 0-13-745167-9. Especially Chapter 2
2) Introduction to Quality Engineering,
author: Genichi Taguchi, Pub: Asian Productivity Organisation, ISBN 92-833-1084-5
The Quality Loss function can be used both internally, i.e. to provide motivation for own improvements, and externally, i.e. to motivate improvements to satisfy and keep customers. In the first case one has good knowledge of own manufacturing costs and can readily use the function. In the case of the benefit to an end user one can try to identify benefits from own knowledge of end use if this is sufficient or by asking the customer or by performing a bench marking to see where own competitors stand. It is clear that this is a delicate problem as an over ambitous attempt to staisfy a customers demands can result in own bankrupcy!
Moving to an aspect of the Loss Function I would like to get an answer to. This quantity can be used to motivate investments to improve quality as a cost per unit for poor quality can be calculated. In these days of adherence to 6sigma philosophy this idea should go home with upper management who generally only priority productivity gains when investing. Does anybody know/have references to such a use of this function?
Hallo again Geoff,
Had to start doing some homework on the Quality loss function.
I became aquainted with this topic when I read D. J Wheeler's book "Beyond Capability Confusion: The average Cost-of-use", SPC Press, ISBN 0-945320-51-5.
Essentially Wheeler's message is that this quantity has nothing to do with Conformance to Requirements/Zero Defects/Cost of Quality/Six-sigma/Capability targets/etc. What he is advocating is that this quantity should be minimised by assuring that all processes are on target and have minimum standard deviation. As such any programme that aims at this will reduce ACU, but one formally does not have to adopt such a programme to use ACU. Wheeler recommends the use of standard SPC practices: i.e. process in control, correctly centered and with the correct tolerances. See pages 79 to 106 in the book.
The problems I have run into when trying to convince our economists is that they cannot accept a coupling between quality and production costs. Instead I have been asked to try to estimate and base a calculation on quality control costs.