DOE

["Alla Linetsky" <alla@concreteelegance.ca>]

I had a sheepish chuckle reading Mark Bailey's post regarding Design of
Experiments (DOE), in Hillside Yellers thread. I, too, once was a
DOE-thumper, consulting in Statistics, failing to understand why so few
people were interested in experimenting effectively. 

But now I'm on the other side of the table. I have my own company,
developing high performance concrete and financing my research by making
concrete countertops and other functional art. So am I using DOE to continue
to improve my concrete? Nope - it's too time-consuming, and it doesn't yield
sufficient benefits for the effort and cost expended.

Major breakthroughs in research are still the result of accidents. Several
times, I was carefully measuring out one ingredient or another, and my hand
shook, or the bag ripped, and a much greater quantity fell into the mixer.
Usually, such an accident has either no effect at all or a detrimental
effect on the different performance measures of concrete. But twice, the
improvement was so dramatic, that I estimated the increased quantity and
changed my mix recipe to reflect it.

Other times, I'd find out about a new ingredient, get a sample to play with,
and often it would result in an improvement. DOE would be of no help in
either situation.

DOE is great for optimizing a process that is already working well. But most
people look for breakthroughs, much in the same way that they keep buying
lottery tickets.

Alla Linetsky, P. Eng.
Concrete Elegance Inc.
416-567-5529
www.concreteelegance.ca