AN EXPERT SYSTEM THAT WAS Richard G. Vedder Center for Quality and Productivity, University of North Texas, P.O. Box 305249, Denton Texas, 76203-5249 vedder@unt.edu Thomas P. Van Dyke University of Nevada, Las Vegas, 4505 Maryland Parkway, Las Vegas, Nevada 89154-6009 vandyke@ccmail.nevada.edu Victor R. Prybutok Center for Quality and Productivity, University of North Texas, P.O. Box 305249, Denton Texas, 76203-5249 prybutok@cobaf.unt.edu ABSTRACT Management at Mary Kay Cosmetics supported development of an expert system (ES) to assist with the problems inherent with package design. Although the ES performed well and was clearly beneficial, use of the ES discontinued after a short time and the package design process returned to pre-ES practices. Within the context of this case this paper describes from a managerial perspective how and why this valuable tool fell into disuse. INTRODUCTION A decision team composed of managers and staff experts at Mary Kay Cosmetics developed packaging for new and revised products as well as set their prices. However, this decision-making process could result in costly errors; for example, product-container incompatibility can cause product discoloration or failure of the package’s integrity. Consequently the group became interested in pursuing development of an expert system (ES) to assist with the problem. The ES performed so well that the group ceased to meet for purposes of making the decisions but instead meet to provide a check and approval on the expert system's solution. Nevertheless, despite the clear benefits associated with using an ES approach, the tool fell into disuse and the behavior of the group returned to the pre- ES process. This case study highlights both practices for success and issues that can cause failure when using a sophisticated information technology such as expert systems. DOMAIN PROBLEM AND ES DEVELOPMENT Mary Kay Cosmetics decided to investigate an expert systems approach for selecting cosmetic packages and setting product price. The company decided to explore ES for both reasons of curiosity and of a desire to reduce the likelihood of errors in the packaging decision process. The selection of packaging materials for cosmetic products poses a major challenge for the cosmetic industry. The selection criteria are based on marketing requirements and chemical composition of the product. Marketing requirements include type of dispenser and clarity of the package. The chemical requirements include the composition and characteristics of both the product and the package material. The challenge stems from reconciling these two different sets of requirements given a set of cost constraints. The major sources of concern are product-package interaction (ex., contamination of the product by the material of its package), product degradation over time (ex., color change), or package failure (ex., migration of the product through the package). Addressing these problems requires evaluation of varied decision factors such as alcohol content, solubility, oxygen or light sensitivity, fill temperature, pH, water vapor transmission rates, and the technical costs of manufacture. A packaging failure can be very significant: product recall costs, costs associated with redesign and production of new packaging, lost sales, lost customer good will, and so forth [1]. Prior to the use of the expert system, the packaging decision effort was a loosely structured process. A cross-functional work team representing marketing, product formulation, product safety, package engineering, research and development, purchasing, and manufacturing interests made the packaging decisions. No formal procedures existed for either conducting the meetings, or sequencing the decisions, or selecting package materials. Technical and cost reasons encouraged the marketing team charged with developing the new product to give their requirements to the product formulator and the package design engineer at the same