I have been using DSD's and CD's for quite some time now. What I am finding is that I rarely (well, never, really) use the vintage Factorial Designs, Plackett-Burman, etc. Though I still like the Box-Behnken Designs. My question is: Am I missing something, or are others finding that the "old" designes aren't being used much any more?
I might add that the videos on the JMP site regarding DOE by Peter Goos are very excellent!
DSDs are the first thing I think of for DOE these days, but there are still situations where I go back to classical designs. In particular, if the number of factors is small or most of the factors are qualitative, I think I can get mileage out of full 2-level factorials (with 2-3 factors), and a 2^5-1 is still a good design, especially if there isn't a goal of building an equation that might have quadratic terms. A couple of other reasons to sometimes consider classical designs are a) they are very easy to explain to novices, and b) the analysis is a little more straightforward and easier to understand IMHO.
As a DOE practitioner for over 30 years, certainly DSD's are exceedingly attractive in that it opened the door to push the number of factors capable of being investigated to much larger numbers. As you know "sooner or later it all comes down to money" and if there was any key learning from my experience is that I wish I could have always added more factors to investigate in a multivatiate fashion. However as you know experiments sometimes require a constrained region or multiple level categorical factors and that is where the flexibility of JMP's custom designer shines. The classical designs are very good designs and thus the category "classical desgns" and in some instances they are a design of choice using the custom designer. However, the custom designer can be used to optimize the goal of the experiment by choosing the design criteria, namely, d-optimal, i-optimal or alias-optimal to produce the design that meets ones critria. That flexibility along with the new compare design platforms one can readily evaluate all of the potential design candidates and choose the one that suits the problem rather than forcing one's problem to fit into a design.