Young, J.C. (1989, ASQC) Institute for Improvement in Quality and Productivity, University of Waterloo, Waterloo, Ontario, Canada
The advantages of carefully planned systematically designed experiments over traditional one-at-a-time studies have been well established. One large corporation reports saving over $60,000,000.00 in a one and a half year program of experimentation. Their average payback on an experiment was 3.5 months!
Most of the publicity surrounding this topic has concentrated on the efficiency of designed experiments or on the technical details of their analysis. The analysis, in fact, usually relatively straightforward if adequate attention has been paid to the initial design, planning and management of the experiment. However, the usefulness of the results is very sensitive to the correct choice of experimental factors and their levels and to the manner in which the experiment is carried out. Since a lot of information is being obtained from relatively little data, good data is essential. There are, therefore, important points to be made regarding the effective implementation of this powerful tool for quality improvement.
Methods such as blocking, replication and randomization are also important. Although not always feasible in the industrial context, these methods can considerably increase the efficiency of designed experiments as well as pave the way to more reliable conclusions.
Design of experiments (DOE),Statistics