When a point is lost from a complete 2nfactorial the variance efficiencies of the estimates of the effects may be reduced by as much as 50%. The loss is not so great if the interactions with (n− 1) factors are ignored. This phenomenon is also investigated for some fractional factorials. If in a minimal resolution IV design a pair of complementary points is lost the resulting design has resolution II. It is sometimes possible for the experimenter to save his experiment by running the missing points later in a separate block together with duplications of some of the other points; this situation is discussed in a later section.