The New Product Problem: An Approach for Investigating Product Failures

This paper discusses a recently developed variable selection procedure suitable when the available data have discrete components. In the context of the new product problem, a modification to the basic methodology is proposed with a view of identifying product profiles closely associated with success or early failure. A comparative analysis is also undertaken in which the new method is contrasted with the more widely used linear discriminant approach. The results indicate that the proposed methodology stands up well to the more common approach in terms of both practical and classification efficacy.