If an experiment is being designed for a small number of factors, a complete factorial design can be used. Ideally the design should be replicated or, if all factors are numerical, additional experimental runs might be conducted at the centre point of the design in order to estimate the error variance.
However the number of runs of the experiment required for a full factorial design increases exponentionally as the number of factors increase. If there are many factors, a complete factorial is prohibitively expensive, so alternative experimental designs should be used that require far fewer runs than a complete factorial.