Receiver operating characteristic analysis for forecasting obsolescence risk

In today's world of rapidly advancing technologies, the impacts and costs of change on products and systems are dramatic. As new products are introduced into the market, manufacturers cease production of older products. These products that are no longer procurable have become obsolete. Planning ahead for when products become obsolete requires accurate and dependable forecasting methods as well as methods to assess costs associated with obsolescence. Currently, obsolescence risk forecasting models assign statuses of procurable or discontinued to parts to denote the part's availability. As with any forecast, some parts will be mislabeled, i.e., a procurable part may be labeled as discontinued or vice versa. Typically, it is assumed the cost of a type I and type II errors are equal. However, a procurable part labeled as discontinued may trigger a premature redesign and potentially worse, a discontinued part labeled as procurable could cause a starved manufacturing line or even production shutdowns. The objective of this paper is to present a new method that has been developed to avoid significant errors in cost calculations and resultant decisions. The method makes use of Receiver Operating Characteristic (ROC) analysis that makes it possible to account for asymmetric costs when parts are misclassified. A case study of the digital camera market is presented to demonstrate the method. When a new optimal risk cut-off threshold is calculated using ROC analysis, an average reduction of 26.4% in cost per prediction in the camera market is demonstrated.