Article Preview
TopIntroduction
Product development is an essential activity for enterprise growth and prosperity, and the exploration of success factors is recognized as an important business issue (Zirger & Maidique, 1990). In product development, there are two important indicators: KGIs (Key Goal Indicators) and KPIs (Key Performance Indicators). A KGI is a final goal to achieve in product development, such as sales or market share. A KPI is a leading indicator, which measures how well a process for achieving KGI is going, such as procurement status of resources or strength of management force. For example, the success rate of patent registration is one of the important KPIs for KGIs such as sales because patents often make products extremely competitive. But how patents will influence sales of the product often differs in attributes of enterprises such as business type or annual sales. Therefore, enterprises need to understand relationships among KPIs/KGIs, taking into account the differences in their attributes.
In this study, the authors propose methods to resolve this issue by leveraging a Bayesian network and Latent Dirichlet Allocation (LDA). The proposed methods are able to give graphical insight of relationships among KPIs/KGIs and probabilistic inference of how KPIs influence KGIs, considering the differences in attributes of enterprises. In the example above, the proposed methods can quantitatively infer how the sales will change according to business type and annual sales, if enterprises work to improve the patent registration ratio. In the main sections that follow the Background section of this article, the details of the proposed methods are described. The outline of the main sections is shown in Figure 1.
Figure 1. Outline of the main sections
TopFundamental Concepts
This section outlines a Bayesian network and LDA used as the basis, and then the concept, of this study.