Factor screening is an important beginning to an experimental program for characterizing and then improving the performance of a product or process. The idea behind factor screening (or more simply, screening) is to identify the few factors that have the most substantial effects on the response of interest. Screening depends on the assumption that the Pareto principle applies. That is, most of the variation in any system is generally due to just a few driving factors.
Screening designs are popular, in part, because they generally require comparatively few runs especially considering the many factors being investigated. Traditional screening designs usually consider factors at two-levels (Low and High) to economize on the required number of runs.
One consequence of running the most economical of the standard screening designs is that main effects might be confounded with two-factor interactions. Also, two-factor interactions might be confounded with each other. If a two-factor interaction effect is substantial, then practitioners who use such designs must perform additional runs at a later time to resolve the remaining ambiguities.
For quantitative factors, engineers and scientists often prefer designs that have three levels (Low, Middle, and High) for each factor; two levels are not sufficient to detect nonlinearity, which is common in physical systems.
Definitive screening designs address both of the above concerns. They have three levels for every quantitative factor, so they can detect and identify any factor causing a strong nonlinear effect. Also, main effects are independent of two-factor interactions and two-factor interactions are not confounded with each other. This allows users to avoid the need for follow-up runs to resolve model ambiguity in many cases.
Note that definitive screening designs are available only for continuous factors and categorical factors with two levels.
Read about the many other benefits of definitive screening in Definitive Screening Platform Overview.