There may occur situations where, due to improper planning or other issues, the original design matrix contains forbidden or unreachable combinations of the factor settings. A computer aided design (for example a D-optimal design) can be used to 'replace' those runs from the original design that were unattainable. The runs from the original design that are attainable are labeled as 'inclusion' runs and will be included in the final computer aided design. 

Given a pre-specified model then the computer aided design can generate the additional attainable runs that are necessary, on top of the inclusion runs, in order to estimate the model of interest. As a result, the computer aided design is just 'replacing' those runs in the original design that were unattainable with a new set of runs that are attainable and still allows the experimenter to get information regarding the factors from the experiment. 

The properties of this final design will probably not be as optimal as the original design and there may exist some correlation among the estimates. However, instead of not being able to use any of the data to analyze, generating the replacement runs from a computer aided design, a D-optimal design for example, allows one to have the option of being able to analyze the data. In this same way, computer aided designs can be used to augment a classical design with treatment combinations that will break alias chains among the terms in the model or permit the estimation of curvilinear effects.