Expert's Panel

A collection of timely and important topics posed to our panel of experts.

Back to Expert's Panel

    How do you evaluate cost of ownership data from multiple sources? Is there a polite way to get people to cough up real data? Why don't people use the SEMI Cost of Ownership formula?

  1. The SEMI COO isn't used because it is too complicated. Big companies have their own models-they are kind of like the SEMI-Sematech model and you're not going to get access to that so don't even try. I don't like to use the Sematech model at all because any savvy customer knows there are 25 things inside that you can manipulate to make yourself look good. So I like to boil it down to four or five things-throughput number, a price amortized over say 5 years, 4 shifts operations (8,000) hours per year, a consumables cost, and then one other thing which might be labor, cleanroom space, etc. If I can sell those things in terms of my being better, I'll be the vendor of choice. The Sematch COO model is too co
  2. I don't like to use yield in a quantitative argument. I love to use yield in a qualitative argument. There are 300 steps in making of a device and the likelihood that you could have a clearly defined advantage is pretty small. You can't defend your credibility. You deal with it qualitatively-you make a generalized claim but stay out of the quantitative claims on yield. In the Sematech model yield is the bottom line devisor-it affects almost everything in a very big way and you have almost no control.
  3. Put the new tool in your own demo lab. The customer is quite happy to come and bring a dozen cassettes of wafers and a few of his own engineers. Run it while you are there. That is the best possible situation because it is under your control and it is also under their control. The data that comes off is on their wafers and with their people-therefore it is customer data. So if it is good, it's useful and you have it. If it is bad, you don't want to use it. Avoid a Beta site.