Applying Supervised and Reinforcement Learning to Design Product Portfolios in Accordance with Corporate Goals

Abstract

Faced with rapidly changing technologies, diminishing product life cycles and heightened global competition, product portfolio managers across all industries encounter increasing challenges within portfolio design processes. Aligning the product portfolio with corporate strategies is central to sustain long-term company success. However, regarding volatile environments, this is becoming increasingly challenging. In the last decades, product portfolio decisions were based on subjective experience, but this is no longer sufficient. Nowadays, as portfolio complexity grows constantly, data-based decision support procedures are needed to enable effective decisions in product portfolio management. Regarding the field of portfolio management, only little research has been conducted on the usage of data-based analytical methods. Additionally, the alignment of product portfolios with corporate strategies is still largely unexplored and, in this context, the application of analytical methods has been largely omitted until now. This paper proposes a methodology that uses neural networks with supervised learning to model correlations among product portfolio control parameters and corporate goal indicators. Based on this, reinforcement learning is applied to derive goal-conform recommendations for product portfolio managers. For both supervised and reinforcement learning, the presented methodology includes generic steps for implementation. Moreover, for both machine learning methods, requirements regarding necessary product portfolio data are elaborated. The methodology is validated using a case study.