The value of multiple testing in cost variance investigation

Cost variance investigation models were development to help managers decide whether and when a reported cost variance should be investigated, given the costs and benefits involved. Typically, the operating system can operate in either one of two possible states-in control or out of control, the state being characterized by a known distribution of the operating cost. The purpose of the investigation is find the fault in the operating system, if one exists, in order to minimize operating costs. Previous models made use of statistical decision theory in order to decide whether to investigate and how much to spend on the investigation. This study focuses on the strategy of the investigation rather than whether or not to investigate. By allowing a multi-test investigation it is shown that, in general, an optimal investigation will consist of more than one test. Moreover, the study identifies cases where an optimal investigation will make use of all tests available. Situation where a multi-test investigation is appropriate are described.