Enterprise Organization and Communication Network Hideyuki Mizuta IBM Tokyo Research Laboratory 1623-14, Shimotsuruma, Yamato-shi Kanagawa-ken 242-8502, Japan E-mail: e28193@jp.ibm.com Fusashi Nakamura IBM Tokyo Research Laboratory 1623-14, Shimotsuruma, Yamato-shi Kanagawa-ken 242-8502, Japan E-mail: nakamuf@jp.ibm.com Abstract In this paper, we consider enterprise organization structures and communication networks obtained from email transaction data. Along with recent progress and popularization of Information Technology, social sciences have been experiencing great advances in methodology. It has become possible for researchers to utilize huge computational social data. However, there have been only conceptual studies in business school and very little quantitative analysis for enterprise organizations. We evaluate the strategic organization change with graph/network analysis of the communication network. keywords: Enterprise, Organization, Communication, Network
1 Introduction Collective human behavior causes complexity and difficulty in enterprise management. Observing an enterprise from the view point of such human connection, we notice that there are multi-layers of networks co-existing (Fig. 1). These networks are mutually related and important for business and personal activities in the enterprise. Communication Network Formal Organization Figure 1: Formal Organization and Communication Network To perform business activities strategically with strong leadership, usually there is a personnel organization in an enterprise with tree structure starting from the top which consists of board of directors and CEO, going through several level of managers, and arriving at leaves of employees. Such an organization structure can also be observed in Government and Army. In addition to the strict chain of control, a lot of modeling and managing methods of business processes such as SCM (Supply Chain Management) and CRM (Customer Relationship Management) are studied for efficient business. Similarly, organization studies are motivated to achieve business goals efficiently with strategic construction and transformation of organization considering the management policy of CEO, enterprise culture and business environments. On the other hand, the value creation via flexible collaboration across organization becomes important in services business and R&D where it is difficult to apply such a modeling with typical processes. When we consider about collaboration in general, the personal connection via human-tohuman communication, say Social Network, becomes a typical topic to be discussed. Today, there is a confusion about the term Social Network. Some people use this for tools or web sites to formulate personal communities by introducing friends to friends. In Social Science, particularly, in Computational Organization Theory, it is used to investigate communities from the view point of network. In addition, the agent-based approach has been applied to computational organization models (for example, PCANS [1] and VDT (Virtual Design Team) [2]). And also, along with the recent findings [3] on widely observed characteristics among natural or artificial networks, we often hear other expressions such as small world network [4], scale-free network, and evolving network. In this paper, we simply use the notation communication network. 2 Communication Network In this section, we d like to clarify communication network which we use for the analysis. Today, there are many types of communication method. For business collaboration, we can use various level of technology from traditional face-to-face meeting and telephone to computational email, instant messaging and groupware (Fig. 2). It is still difficult to gather all these activities. But it is expected that in the near future almost all remote communication will be transferred on the IT network such as IP Phone or Web conference. In this paper, we consider the transaction obtained from email send/receive log files together with the enterprise organization as the communication network for analysis. For the traditional social science, researchers usually investigate various communication among examinees by direct interviews
email face-to-face meeting instant messaging phone groupware Communication Figure 2: Various Communication Channels or questionnaires. However such a human survey is so troublesome and time-consuming that the amount of available data is restricted. By using email transaction log files which is automatically stored on the server when people send or receive an email, we can utilize enormous communication data easily. In Econophysics, for example, researchers have revealed new characteristics of the financial market with such automated enormous data. Similarly, it is expected that huge transaction data can expand our knowledge on the communication network. In reality, the power distribution is observed both in the financial market and the Web. We introduce a survey example of the communication network in an enterprise organization before and after strategic organization change. In our analysis, it is also important to consider the location in the strategic enterprise structure where senders or receivers of email correspond. We have analyzed huge communication from email transaction log data as described above. From a viewpoint of privacy, we transformed and aggregated the transaction data as the relationship between department for a month so that a particular employee cannot be traced. Furthermore, we did not investigate subjects and contents of email in this survey. Fig. 3 shows the data flow of personnel data and email transaction for the analysis tool in the next section. HR DB Personnel data Node definition Department Manager XML Organization Structure Analysis Tool Log file email transaction email Addresslist email data Process on RDB unitlink Figure 3: Data Flow for Communication Network
3 Analysis Tools We are developing a tool with Java to analyze the communication network over the enterprise organization. In this section, we briefly introduce a prototype of the tool (Fig. 4). To perform the analysis in this paper, we consider simple indices for graph/network. The developed tool preprocesses input data at first. Then, it analyzes networks with implemented algorithms for network indices. Finally it visualizes the result with corresponding GUI for each index. To analyze the preprocessed data with other algorithms which have not been implemented yet, the tool can also output the preprocessed data into a file. We have developed this tool with object oriented model so that new indices and algorithms can be added easily. In the program, each analysis class and GUI class are implemented by using fundamental interface classes. As the preprocess, the tool aggregates email transaction data between lowest-level departments (units) described in the previous section to link information between strategic groups (nodes) which we want to investigate for consulting services. In addition, the link information which include amount of email between nodes are normalized by a number of corresponding employees in a node, or discretized with a specified threshold with the user s preferences. Figure 4: COA Analyzer GUI : Organization Tree and Network Indices Currently, this tool evaluates Input/Output Degree, Distance, and Closeness which is obtained as a inverse of the sum of distances to all nodes for each node. With an adjacent matrix A = (a ij ) where a ij is the communication frequency between node n i and node n j and a distance matrix D = (d ij ) where d ij is the geodesic distance between n i and n j, Degrees and Freeman s Closeness are written as follows: InDegree(n i ) = a ji, j OutDegree(n i ) = j a ij, Closeness(n i ) = 1 j d, ij We also consider Power index and Faction (subgroup) with UCINET [5]. For 3D visualization of hierarchical organization and communication network, we have combined Ito s visualization tool [6] into our tool (Fig. 5). 4 Example In this section, we introduce an example of case study for organization change and communication network though we cannot reveal all details. We investigated a part of a company with consultants (IBM Business Consulting Services KK) using these data and tools described in the previous section. Especially, we directed our attention to the recent strategic organization change for strong matrix structure. We made an analysis with a node definition for this direction and observed changes appeared in the communication network.
Figure 5: Visualization In this survey, we used email transaction log files and a part of personnel management information to investigate organization structure for input data and to find corresponding department with each email address. Before the organization change, the matrix structure was not balanced and one direction (product axis) was dominant, which caused customer facing various separated products without a unified service. The focused organization change is intended to strengthen another direction (sector axis) to balance the matrix structure and to provide one face services to customer in each sector. Hence, we began the analysis by investigating changes in Degrees of the communication network along these two directions. Summarized and scaled results are shown in Fig. 6. Figure 6: Degree Change in Product and Sector Direction We can observe a significant increase of the communication along sector from the graph. On the other hand, this change does not cause a undesirable side effect on the communication along product. Next, we analyzed changes in Total Closeness (Fig. 7). Together with the previous results which indicate both communication in the matrix organization with sector and product axes become strong and balanced, we can consider that the whole organization network becomes more condensed one.
Figure 7: Total Closeness Change Finally, we show a number of nodes in subgroups divided by UCINET as Faction for each sector (Fig. 8). It can be considered that nodes in a same subgroup communicate frequently. Before the organization change, a subgroup contains nodes of various sectors. However, after the organization change, a subgroup contains nodes of one sector or a few sectors. This result agrees with the degrees along sector increase. Furthermore, it shows that increased communication is strongly correlated with strategically determined sectors. 5 Concluding Remarks In this survey, we observed the effects of an organization change by the top-down strategic decision over the communication network with simple indices. Traditional studies on enterprise organization such as matrix structure are restricted within descriptive intuition usually performed as case studies at Business School, and lack quantitative analysis. With our survey though only with simple indices at now, effects of the organization change can be analyzed by using enormous email transaction data which can be collected automatically. More business-related analysis with complex indices remains as future works. In addition, the distribution of Degree of the communication network surveyed in this paper are shown in Fig. 9 sorted by order of degree from the largest node to 100th node using a log-log scale. This result indicates a power law distribution and that the communication network over enterprise organization also has a characteristic of scale-free as many other natural networks though this network is constructed on a strategically determined organization structure. Future analysis on such characteristics of communication network itself and the influence upon business activities is important. References [1] Carley, K. and Krackhardt, D. A PCANS Model of Structure in Organizations., Proceedings of the 1998 International Symposium on Command and Control Research and Technology. June, Monterey, CA, 1998. [2] Levitt, R. E., Computational Modeling of Organizations Comes of Age., Computational & Mathematical Organization Theory. 10, pp.127 145, 2004. [3] Barabasi, A. -L., Linked, The New Science of Networks. Perseus, Cambridge, 2002 [4] Watts, D. J., Small Worlds - The Dynamics of Networks between Order and Randomness, Princeton University Press, Princeton, NJ, 1999. [5] UCINET, http://www.analytictech.com/
Sector Before After Figure 8: Subgroup Change [6] Itoh T., Yamaguchi Y., Ikehata Y., and Kajinaga Y., Hierarchical Data Visualization Using a Fast Rectangle-Packing Algorithm, IEEE Transactions on Visualization and Computer Graphics.
Order Figure 9: Power Law in Degree Distribution