This paper describes a sequential method which may be used to determine with efficiency the important variables and interactions in an experiment. The method is useful in certain cases involving up to 100,000 or more variables. Mathematical derivations are given and tables are included showing the optimum designs and the total numbers of tests or samples required.