Home | Business News | Browse by Publication | I | International Journal of Market Research

Genetic Algorithms for product design: how well do they really work?

Article Excerpt
Recently, Balakrishnan and Jacob (1996) have proposed the use of Genetic Algorithms (GA) to solve the problem of identifying an optimal single new product using conjoint data. Here we extend and evaluate the GA approach with regard to the more general problem of product line design. We consider profit contribution as a firm's economic criterion to evaluate product design decisions and illustrate how the genetic operators work to find the product line with maximum profit contribution. In a Monte Carlo simulation, we assess the performance of the GA methodology in comparison to Green and Krieger's (1985) greedy heuristic.

Introduction

Designing and pricing new products (1) is one of the most critical activities for a firm. In this context, conjoint analysis (Green & Rao 1971) has been used extensively in marketing research to measure the impact of product attributes on consumer preferences for products. The high commercial usage of conjoint analysis for new product/concept identification has been documented in the surveys of Cattin and Wittink (1982) and Wittink and Cattin (1989) for the US market, and by Wittink et al. (1994) for the European market. Theoretical details of the conjoint methodology can be found, for example, in Green and Srinivasan (1978, 1990).

Economically, besides short-term objectives such as market share, a new product or product line should be profitable to a firm. Here, a number of conjoint-based models with focus on profit-oriented product (line) design have been proposed (for example, Green & Krieger 1985, 1992; Dobson & Kalish 1988, 1993; McBride & Zufryden 1988; Kohli & Sukumar 1990; Nair et al. 1995; Chen & Hausman 2000). Since attributes used in conjoint studies are typically discrete, searching for optimal product designs/lines is of combinatorial nature and known to become computationally intractable with increasing problem sizes. Therefore, mainly heuristic solution procedures have been presented in this area (for example, see Kaul & Rao 1995).

In 1996 Balakrishnan and Jacob proposed the use of Genetic Algorithms (GA) as a powerful tool to identify an optimal single new product using conjoint data. We here pick up Balakrishnan and Jacob's idea and generalise the GA approach with regard to the problem of optimal product line design. In the next section we briefly review the basic GA process and adopt the principles of the GA methodology to our problem. We illustrate how the genetic operators work and provide an example of a GA application to find a product line with maximum profit contribution. In the third section we assess the performance of the proposed GA in comparison to Green and Krieger's (1985) greedy heuristic by means of a Monte Carlo simulation. The last section summarises the most important contents of the paper with the focus on practical applications. For readers interested in technical details, the Appendix provides a formal description of the product line design problem.

Application of Genetic Algorithms to product line design

Genetic Algorithms (GA), first proposed by Holland (1975), are based on the principle of natural selection which has become known as 'survival of the fittest'. GAs usually work on a population of binary strings where particular string positions or substrings correspond to the decision variables of an underlying problem. The term fitness, standing in its original sense for the ability to survive, is adopted for string evaluation and represents the value of the function to be optimised at the candidate solution encoded by a string.

To apply the GA approach to the problem of optimal product line design, we define each substring to represent one of K conjoint attributes. Each substring is made up of [L.sub.k] binary string positions which correspond to the feasible levels of attribute k (k = 1, ..., K). Thus, a '1' in a substring indicates the presence and a zero the absence of an attribute level. As we consider a (new) product to...

View this article FREE - Now for a Limited Time, try Goliath Business News
Free for 3 Days!