The structure of cross-functional process model collections: An empirical investigation

Abstract

Firms organize work across functional line-of-businesses to deliver the most important outcomes that are needed for clients and other key stakeholders. These cross-functional operations are represented as business processes and are known to be the most critical ones in a firm. However, little is known about business-relevant families of processes because most available BPM literature deals with examples including only a couple of processes or very small collections for very specialized and narrow areas of the organization. The modeling and understanding of cross-functional processes in an industry segment is then a very important task. The representation of these collections involves a network of N core activities that a typical organization carries out in its individual functions as well as the coordination of selected subsets of these activities into a family of P cross-functional processes. In this paper, we present a study of the structure of these cross-functional process models. We show some properties of these collections through extensive experimentation with well-established families of processes for Property and Casualty, Health Care, and Life and Pension harvested in the Insurance Industry for more than 15 years. For the first time in the literature, the structure of a process model collection is studied and shown. The statistical laws followed by these collections of actual and real process models are shown in three major insurance industry segments. One remarkable result shows that the frequency of reused activities follows long-tail distributions. Some statistical tests are applied to determine the specific type of this and related distributions of N activities and to verify whether their tails follow a power-law or other forms. We conjecture that all process model collections capturing all key operations done in an organization (both cross-functional or within a significant Line of Business) exhibit this long-tail behavior. We also investigate the effect of different partitioning criteria by grouping the N activities into K components by following two criteria. In one of them the goal is to achieve a balance between the cost of interactions needed by the P processes across such components and the complexity of managing O(N/K) activities in each component. The other criterion delves into graph community formation by suitable edge criticality properties. This chapter of our work offers a preliminary investigation into Simon's principle of organizing complexity as a set of loosely-coupled interactions among a smaller number of coarser grain components in an enterprise.