Multivariate Profile Monitoring Method

Abstract

Several authors refer to product portfolio management as an essential process because it may be used as a corporate management tool. However, the product portfolio management methods which are often adopted have limitations that prevent its use in practice, mainly due to the high dimensionality of selecting an optimal portfolio. Moreover, the large amount of available data is a relevant issue for practical applications. Thus, the contribution of this article is to propose a method for the product life cycle to monitor time-series behaviour patterns. The goal is to identify changes that may indicate that the product portfolio needs to be revised. The proposed method uses a multivariate regression model to relate financial variables associated with the products portfolio, the performance of products against competition, and even macroeconomic data. The objective is, through profile monitoring, to identify the specific time for the product portfolio review decision-making. We adopted three tools to develop a method – principal component analysis, multivariate regression model, and profile monitoring with Hotelling T 2 Control chart. A Monte Carlo simulation validated the approach. The results showed false alarm rate and average time to signal to be similar to previous studies. Finally, the application of the model is illustrated in a real case, using data provided by a company’s portfolio of agricultural equipment.