A merchandise investment planning system using knowledge-based and portfolio theory approaches

Charles Vincent Trappey, Purdue University

Abstract

This dissertation addresses the need to study new methods required to automate and improve retail buying. Two methods are developed and tested. The first method is a knowledge base system which preserves the buyer's knowledge of merchandise risk and investment returns. The second method optimizes the buyer's percentage investment decision using a model derived from Markowitz's portfolio theory. Both methods are computer based techniques which can be integrated into a decision support system for retail buying. In order to verify the economic and decision making benefits of the two methods, two tests were performed with data obtained from a large Mid-west retailer. The first test focuses on the problem of storing and accessing the expertise of a retail buyer. A group of 19 retail buyers estimated the same collection of merchandise, but with poorer results than the students. The statistical test demonstrated that the students using the knowledge base system were significantly better than the professional buyers in evaluating the investment returns of the merchandise. The second test focuses on the problem of optimizing the buyer's investment in different classifications of merchandise. The objective is to maximize the investment returns for a portfolio of 22 merchandise classifications while simultaneously minimizing the investment risk. A retrospective test was used to compare the net profit results of the optimized portfolio to the net profit results of an expert buyer. The 22 net profit values derived by the model were not statistically different from the net profit values of the buyer. The model generated an assortment of men's sportswear that was similar to the assortment selected by a successful buyer. The theoretical implications of the research is that retail buying is more of a science than an art. A methodology is presented for developing and testing a knowledge base system for retail buying. The results of the study demonstrate the feasibility of storing and retrieving knowledge about buying, a first step towards building a more comprehensive understanding of the science of retail buying. A methodology for optimizing the buying decision is also presented. (Abstract shortened by UMI.)

Degree

Ph.D.

Advisors

Feinberg, Purdue University.

Subject Area

Marketing|Management|Finance|Artificial intelligence

COinS