Predicting and mitigating organizational risks in fast-paced projects
When we started this research in 1987, our research hypothesis was that we could build theory and tools to enable project managers to design their organizations in the same way as engineers design bridges. That is, our theory and tools would enable project managers to build computer models—“virtual prototypes”—of their project work processes and organizations, and then use the computer to generate predictions of the performance of the project organization executing the given set of project tasks. Armed with this kind of analysis tool, a project manager could systematically diagnose schedule, cost, and quality risks associated with the planned configuration of the project; and the project manager could then “flight simulate” the project to explore the impact on project performance of a series of managerial interventions aimed at eliminating or mitigating these risks.
Current Research Status
By the middle of 1996, we had developed a prototype of our “Virtual Design Team” (VDT) system and demonstrated that it could accurately predict schedule, cost, and quality risks for ultrafast-track projects aimed at: rapidly rebuilding earthquake-damaged power plants; installing North Sea oil production facilities inside very short weather windows; and shrinking development time from five years to one year for the commercial satellite version of a military launch vehicle. Based on the success of the prototype system, Stanford University encouraged the researchers to formed Vité Corporation to commercialize the results of the research to date. Subsequently, Vité and Stanford users have applied the VDT methods in many scores of complex, fast-paced projects (Kunz et al 1998).
The information processing view of organizations, first conceived by March and Simon (1958), and introduced to managers by Jay Galbraith (1974), proposed that knowledge workers process information until they encounter “exceptions”—situations in which the information required to execute a nonroutine task exceeds the information available to the person performing the task. They then refer exceptions upward in the formal hierarchy to find someone who can provide the needed information to resolve the exception. In this perspective on organizations, which underlies much of organizational contingency theory (Burton and Obel 1998), the supervisory hierarchy is the primary resource available to workers for resolving their exceptions.
Galbraith (1974) helped to formalize the notion of matrix organizations in which multiple hierarchies overlap, so that many workers have two or more supervisors—for example, a project supervisor and a functional supervisor—to whom they can refer different kinds of exceptions. However, Galbraith still viewed these supervisory hierarchies as the primary knowledge management device for handling exceptions.
Galbraith’s early work focused on both the information processing limitations (the “bounded rationality”) of workers and their supervisors, as well as the information communication limitations of earlier low bandwidth communication technologies such as memos and textual computer printouts. He asserted that bottlenecked supervisors and clogged information channels were the major limitations on the effectiveness of fast moving project teams, and proposed two kinds of generic strategies for addressing the information overload problem: reducing information processing demand and increasing information processing capacity.
Information processing demand on organizations is thus likely to continue to increase rather than decrease for the foreseeable future.
Galbraith’s second strategy proposes that organizations find ways to increase their information processing capacities. To increase organizational information processing capacity, he recommended that organizations: 1) use enhanced communication technologies (hardware and software) to augment vertical communication; and 2) deploy matrix organizations with formalized multidimensional hierarchies and project-based teams to facilitate lateral communication.
Exhibit 1. VDT/SimVision Model of a Project Work Process and Organization: This model shows the milestones (hexagon), tasks (rectangles), actors (human-like icons), and dependencies (connecting lines) for the preconstruction activities in developing a new Biotech facility. Note the relatively small number of milestones and tasks. By analyzing this model, the client determined that it would have to simplify the facility design to complete the project on schedule. (Graphics courtesy of Vité Corporation, www.vita.com)
Exhibit 2. Exective Dashboard: We build a number of versions (“cases”) of the baseline SimVision model. This chart compares the respective completion date, cost, and risk of the different cases. Note that no case simultaneously meets the explicit duration, cost, and risk objectives of the project manager.
Empirically and in theoretical models, exceptions that arise during project execution often result in significant amounts of additional (unplanned) communication and coordination between members of a project team. Recent VDT micro contingency theory of organization behavior, implemented as computational VDT models of organizations, assumes that project-based exception handling has largely been limited to traditional project teams where the majority of the productive project work occurs asynchronously (i.e., in a distributed, offline) mode (Jin and Levitt 1996). VDT models have been able to predict when the emergent communication and coordination load associated with exception handling exceeds the processing capacity of affected members in significant real projects, and both prediction and empirical observation confirm that the impacts can include major schedule delays, quality issues and/or cost overruns.
The Internet has enabled and significantly accelerated both of these strategies for enhancing the information communication channel capacity. Thus, channel capacity is much less of a limitation for many firms that it was even five years ago. However, the hierarchy frequently no longer possesses the requisite knowledge—even though it may have the raw information processing and communication capacity—to address complex technical exceptions encountered by many of today’s most critical knowledge workers.
Modern organizations now emphasize the social network of the actor. For example, to “network” is now a commonly used verb. The Internet has emerged in only the past few years of as an important vehicle for social network communication. However, classical organization theory dating back to Max Weber and Henri Fayol views the supervisory hierarchy not only as the primary means for communicating and resolving exceptions, but as the only legitimate means for sharing knowledge in an organization. Social networks and the Internet simply do not exist in classical or even recent theoretical frameworks of organization behavior.
This presentation describes generalization of our micro contingency theory of organizations to include informal handling of exceptions within a knowledge network. The exception handling process involves identifying and retrieving relevant expertise from members of the informal social organizational network (Lambert 2001). VDT-Knowledge Sharing (VDT-KS) extends the VDT framework to model and simulate “Connective Knowledge Management” whereby organizations facilitate direct knowledge sharing between networked individuals by publishing internet-based knowledge directories and instituting incentives for experts to share their knowledge with others. This paper presents the VDT framework and shows how VDT-KS extends the VDT micro contingency model to represent and analyze effects of use of a connective knowledge network across the Internet.
The Virtual Design Team (VDT) Modeling and Simulation Framework
We have developed the VDT micro contingency theory and system over the last twelve years. The organizational simulation framework has gone through extensive development and validation, and is now in use in a number of Fortune 100 companies. In the remainder of this paper, we discuss the key concepts and limitations of VDT, and show how our research group at Stanford University is currently extending it to model Internet-based knowledge sharing.
Implementation of VDT Concepts
In Jay Galbraith’s (1974) information processing view of organizations, the details of tasks are abstracted away, and work is viewed simply as a volume of information to be processed by an organization consisting of individuals or subteams with specified information processing and communication capacity. Galbraith’s theory provided the kinds of qualitative predictions and recommendations listed above. VDT research operationalized and quantified Galbraith’s theory at the level of individual tasks and project participants.
Exhibit 3. VDT Prediction: Cost breakdown of tasks in a project. The VDT/SimVision simulation model quantitatively predicts the amount of direct work, rework, coordination, and time spent waiting for an executive decision.
VDT models direct work as a quantity of information to be processed, specified in person-hours or person-days. It operationalizes the notion of exceptions and their resolution as packets of information passing through communication tool channels into the in-boxes of organizational participants. Participants stochastically select one of several items to attend to from their in-boxes. We conducted ethnographic research in organizations to quantify the information processing volumes for typical exception handling and communication tasks, and we gathered data on organizational productivity rates and error rates to quantify the effect of participant skills and experience on information processing speed and error rates in task execution. Exhibit 1 shows a typical VDT model, showing that the VDT model integrates the classic organization chart with the critical path method work process. Exhibit 2 shows an “executive dashboard” that summarizes the duration, cost, and risk of a set of relevant cases, showing that the normal method of using the application is to develop different and quantitatively compare the performance of different management alternatives. Exhibit 3 shows the prediction by the VDT simulation of the distribution between direct, coordination, rework, and waiting for executive resolution of a conflict situation. The manager can appropriately set, track, and manage goals for each kind of effort.
The Virtual Design Team—Knowledge-Sharing
The knowledge management solutions that companies are attempting to develop fall into two broad categories. The first set of solutions attempts to capture, formalize, and package expert knowledge in a form that can subsequently be accessed asynchronously by those who need it, without the need for them to consult the expert directly. The second wave of knowledge management solutions recognizes that it is very difficult to capture sufficient context around knowledge for it to be confidently reused by others. These solutions therefore attempt to help workers who encounter exceptions to identify and link up with appropriately skilled and experienced human experts who can provide contextually embedded knowledge through direct communications. Fulk et al (1996) have termed these two approaches “Communal” versus “Collective” knowledge sharing solutions, respectively.
Modeling Connective Knowledge Management Solutions
Wegner (1987) describes transactive memory as a system for encoding, storing, and retrieving information within a group. Our observations of technical project teams suggest that team members routinely employ elements of “transactive memory” for locating and coordinating expertise to handle certain types of exceptions that arise during project execution. Within a transactive memory system, group members engage in three key processes: 1) update their perceptions about the expertise of other members (i.e., who knows what), 2) allocate new information to other members believed to possess related expertise, and 3) retrieve information from other members most likely to possess relevant expertise. The various communications and transactions that occur between group members facilitate these encoding, storage, and retrieval processes.
Exhibit 4. Connective Knowledge Network Extensions to VDT: In this extended version of VDT, individual actors have responsibility for activities, but they also have transactive memory and actor communication technology so that they are not limited to using the actor reporting hierarchy for resolving exceptions. They can seek advice from anyone that they think might be able to help them, and to whom they are connected via one or more communication technologies (including face-to-face communication with collocated peers). The emergent exception-handling network has attributes that, in turn, affect individual question asking and answering behavior. The ongoing co-evolution of individual knowledge and the knowledge network affects both individual and team performance metrics.
Our goal is to draw upon transactive memory as our framework for representing informal exception handling in cross-functional project teams. In the extended VDT model, we take the current state of a project team’s transactive memory as a new input to the model. When an actor encounters an exception, rather than automatically referring the exception up the hierarchy, the actor considers the expertise requirements of the current activity, searches its internal directory of who knows what, and invokes transactive retrieval as a mechanism for obtaining relevant expertise to handle the current exception.
As Figure 4 shows, these transactive memory micro-behaviors, when aggregated across all actors, result in an emergent network structure, which represents the operational (i.e., the “in-use”) exception-handling network. This exception-handling network differs from the default exception handling hierarchy contained in the current VDT model in that the properties of the operational exception-handling network do not remain fixed over the duration of the project, but rather vary with attributes of the set of “active” tasks and the distribution of skills among actors.
We represent actors’ transactive memories as imposing certain constraints on actors’ actions to resolve exceptions using informal exception handling mechanisms. In the extended model, the emergent information-processing load associated with informal exception handling, coupled with certain network-based constraints on actor actions, work together to produce a dynamic load distribution across members of the project team. At the micro level we describe this load distribution in terms of actor centrality within the operational exception-handling network, and at the macro level in terms of network centralization and density. Using measures of individual and project performance contained in the current VDT model (i.e., actor backlog and communication risk, respectively) our goal is to examine the relationship between properties of the operational exception-handling network and individual and project-level performance.
Our early observations of project teams employing real-time concurrent design processes (Oxnevad 2000) suggest that processes involving real-time coordination of expertise can reduce the total rework, communication, and coordination load associated with exception handling. Our simulation models predict the extent to which these real-time processes achieve greater efficiency by: 1) forcing the team network through concurrent design working sessions, 2) decomposing tasks into smaller subunits, and 3) imposing a process in which the expertise of team members is explicitly coordinated. Using this real-time process, teams are able to recognize certain technical constraints and task conflicts early in the design process, and resolve constraints and conflicts quickly and efficiently. The impacts include: 1) reduced total amount of communication and coordination load associated with exception handling, 2) reduced delays associated with asynchronous exception handling (by shortening the length of time between recognizing an exception and handling it), and 3) reduced likelihood of downstream exceptions in the post-design phase.
Value of the Research to the Profession
This research has contributed to operationalizing and extending the qualitative organization theories of researchers such as Jay Galbraith. The research has also contributed to the overall body of social science theory by building new kinds of computational bridges between cognitive and social psychological theories at the micro-organization level, and organization theories based in sociology and economics that could never be rigorously linked before. For the profession, this research has provided new kinds of modeling and simulation tools that are being used by managers on some of the most complex projects worldwide, to gain new kinds of insights about organizational risks that could impact completion dates, operating quality, and investment returns for their fast-track projects.
Burton, Richard, and Borge Obel. (1998). Strategic Organizational Diagnosis and Design, 2nd Edition. Kluwer Academic Publishers.
Fulk, J., A. Flanagin, M. Kalman, P. Monge, and T. Ryan. (1996). Connective and communal public goods in interactive communication systems. Communication Theory 6: 60–87.
Galbraith, Jay, and Edward Lawler, III (Contributor). (1996). Organizing for the Future: The New Logic for Managing Complex Organizations. Jossey-Bass Management Series.
Galbraith, Jay. (1974, May). Organization design: An information processing view. Interfaces 4: 28–36.
Jin, Yan, and Raymond E. Levitt. (1996, Fall). The virtual design team: A computational model of project organizations. Journal of Computational and Mathematical Organization Theory 2 (3): 171–195.
Kunz, John C., Tore R. Christiansen, Geoff P. Cohen, Yan Jin, and Raymond E. Levitt. (1998, November). The virtual design team: A computational simulation model of project organizations. Communications of the Association for Computing Machinery (CACM) 41 (11): 84–91.
Lambert, M. H. (2001). Hierarchies and transactive memory systems: Crafting a model of flexible exception handling. Paper presented at Computational Analysis of Social and Organizational Systems (CASOS 2001), July 5–8 in Pittsburgh, PA, Carnegie Mellon University.
Levitt, R. E., G. P. Cohen, J. C. Kunz, C. I. Nass, T. Christiansen, and Y. Jin. (1994). The “virtual design team”: Simulating how organization structure and information processing tools affect team performance. In K. M. Carley and M. J. Prietula (Eds.), Computational Organization Theory. Hillsdale, NJ: Lawrence Erlbaum Associates.
March, James, and Herbert Simon (with Harold Guetzkow). (1958). Organizations. John Wiley.
Oxnevad, K. I. (2000). The NPDT—The next generation concurrent design approach. Proc. of the 2nd European Systems Engineering Conference (EuSEC 2000) in Munich, Germany, September 13–15.
Wegner, D. M. (1987). Transactive memory: A contemporary analysis of the group mind. In B. Mullen and G. R. Goethals (Eds.), Theories of Group Behavior (pp. 185–208). New York: Springer-Verlag.
Proceedings of PMI Research Conference 2002