Destroying Capital

Bill Parr

This is a fairly simple note. The purpose is to consider a fairly common example by means of an example followed by some general discussion.
 

Example of destroying capital

Consider an engineer in a semiconductor fab who manages, for instance, a diffusion tube which is in high demand. Due to problems with controlling the variation down the tube, the engineer does a fairly common thing:

1) Passive data collection (perhaps 10 shots in the tube, with measurement of oxide thickness at 5 positions per wafer on a wafer at each of several positions in the tube)

2) Analysis of the data to determine a range (subset) of the positions in the tube over which the variation is noticeably less than the variation all the way down the tube

3) Changing the process instructions to only allow use of the chosen subset (fraction) of the tube for the given process.

For instance, a tube which might have held 5 boats is then restricted to only using the three most central positions for boats.

What is the result? Capacity for that tube is 60% (100*(3/5)) of what it used to be, though the variation is reduced.

What was the choice? A choice was made to forego working on the tube to reduce the variation (leveling temperature down the tube, management of flow of gases, studying spacing of the boats and wafers, etc.). The result is that the price paid for the time saved in not determining how to improve the tube uniformity is a longterm, ongoing sacrifice in capacity at the tube.

Now, to make our discussion simple, assume the tube is one of a kind - and that it is a capacity constraint (bottleneck). In such a case, without convening a management group, possibly without checking with marketing or production control, factory capacity has been reduced by 40%.

General discussion

In my experience, this sort of action is fairly common. It often goes unnoticed because the tube is one of quite a few, and hence the percent reduction in capacity is not as severe as in my numerical example above.

What is the real phenomenon - it is one of optimizing a very local measurement (time for the individual engineer to get their work done over the time when work to improve the tube's uniformity could have been done) at the expense of a global measurement of great financial significant (capacity of the factory and hence ability to service customer orders).

A modest proposal

What can be done? A simple change would be to adopt a practice that any change of the above kind is regarded as not a simple engineer change order, but as a matter of potentially serious capital consequences. To "restrict to the sweet spot" would require first an assessment of how close the tool in question was to being a bottleneck. If it were close, then the rather serious financial consequences of restricting capacity at that process point would have to be weighed inn the decision on whether to restrict capacity or to invest further effort in determining how to improve tool uniformity.

What do you think? In my experience, this sort of action (restricting capacity with no serious review of financial consequences) is fairly common. Have you seen such occurrences? Share your experiences or opinions with me at bill@billparr.org.

March 7, 2002