The optimization strategies are of two main types:

Global procedures, sample an entire portion of the response surface,
while
the Pattern Search technique, a local search procedure, tries to find an optimum by starting at a particular point on the surface.

Thus, the global techniques are less likely to be trapped by local optima (i.e., small valleys on the response surface) than the Pattern Search technique.

The Pattern Search algorithm should only be used to refine parameter values.
The range to be searched for each parameter should be relatively small when using the Pattern Search technique.
The global algorithms should be used if a larger range of possible values needs to be examined.
However, the specified range should be as restrictive as possible so that only the portion of the response surface that is most likely to contain the optimum is searched.
The efficiency of the technique is particularly sensitive to the size of the response surface.