Abstract
Management accountants have used traditional variance investigation models-rule of thumb, x charts, and cumulative sum charts, etc.—for a long time to determine whether a process should be investigated or not. But the traditional model is weak in that it does not incorporate the expected cost and benefit from a variance investigation. The objective of this paper is to demonstrate a procedure regarding how the sophisticated variance investigation model can be applied to the real world. By doing so, it is hoped that sophisticated techniques will be used more widely. A single period Bayesian model was applied to a medium sized firm for the proportional variance investigation. The implementation cost of this model is cheap and can be applied to variance investigation with little hardship.
For the moment, it is a little early to argue that this model would bring more benefits than other traditional models. Although this model accurately and promptly signaled out of control processes for the first month's processes, at least a few years' cost associated with using a single period model should be compared with the cost of using the current simple rule for variance investigation.