The Complexity of Competitive Marketing Strategies

Genetic algorithms (GAs) have been used extensively in engineering and computer science to optimize specific functions, especially those which exhibit non-convexities and so are not amenable to calculus-based methods of optimization. A parallel use of GAs has been to solve algorithmic problems. A third domain in which GAs have been used is that of searching for mappings which optimize a repeated procedure, which also reveals their complexity. An offshoot of this has been their use in what has been called co-evolution of mappings. This paper reports results from a project in which GAs have been used to, first, to derive mappings which may explain the behavior of brand managers in an oligopolistic retail market for coffee, second, to attempt to improve on the historical profits of these brand managers, pitted in weekly competition with each other, vying for sales and profits with their different brands of ground, sealed coffee on the supermarket shelves, and, third, to reveal how the artificial agents’ performance is positively related to their complexity. As well as advancing the practice of GAs, with separate populations competing, the work also advances our understanding of modeling players in repeated oligopolistic interactions, or games.