In process qualification protocols, the issue of whether a process has been 'optimized' is sometimes encountered.
What does 'optimized' mean?
It might mean finding the process settings that result in the best location of the process mean.
It might mean finding the process settings that result in the lowest process variation.
It might mean finding the process settings where the change in the process output with respect to changes in the process inputs is minimized (i.e. flat response; smallest slope or first derivative of the response).
Sometimes known as Robust Design.
It might mean some combination of the above, or perhaps something else depending on the goal of the study and the needs of the process outcomes.
EXAMPLE:
If a process has three inputs, a protocol (e.g. an "OQ" or Operational Qualification protocol) that studies only "Low-Low-Low", "High-High-High", and perhaps "Mid-Mid-Mid" can NOT make the following claims in the absence of earlier DOE studies:
It cannot claim to have found an optimum set of process inputs, as no interaction terms were studied.
It cannot claim to have studied the process 'worst case' conditions, as no interaction terms were studied.
An interaction term that is significant in the process model might mean that some other combination of Low-High process inputs is the 'worst case' condition. For example, Low-Low-High.
For those process qualification protocols where "all Low" and "all High" levels of process inputs are studied, the statement of conclusions should use a term other than "optimized", such as "nominal".
"The OQ tested Nominal, Low, and High parameter settings to allow the development of a nominal process and nominal process window."
(Note that this 'nominal' process window may not include the most challenging conditions for the process.)