Dynamic Release Management: A Market Intensity Approach

Abstract

We consider a release manager who sequentially releases new versions of her product, and ask how she should account for the overall level of competitive intensity in her industry. We model the problem within a combinatorial framework in which new releases consist of discrete features. Motivated by a simple mixed-integer non-linear program (MINLP) that formed the basis of a collaboration with industry sponsors, we propose an approach in which the manager tunes her discount factor to capture the aggregate impact of market intensity. This technique relies on a multiplicative compound Poisson model of market intensity, which we justify by showing how it emerges naturally in a multi-firm equilibrium. This intensity equilibrium facilitates operational analysis, as it captures the relationships among firms at a statistical level, as opposed to a release-by-release strategic level. We also examine how ignoring the existence and/or stochasticity of the outside market can lead to suboptimal outcomes. To facilitate implementation, we provide a numerical procedure for calibrating the market intensity model based on the stylized multi-firm game-theoretic framework.

Keywords: product development, release management, semi-Markov decision processes, approximate dynamic programming