Measurement and control of business processes

It is widely accepted that any well-designed organizational process include s a control mechanism through which management decides which aspects of the performance of the process are to be measured and how these measurements a re to be used to change the level of resources utilized in the process. Lit tle is known, however, about the best ways to design such a control mechani sm Far typical business processes. Our goal in this research is to identify control mechanisms for business processes that are effective in different types of environments. In this article we present a system dynamics model o f a typical service-sector business process, such as is used in processing administrative paperwork in insurance, banking, and so on. These processes are subject to random, time-varying, and non-postponable demands for servic e. They are also subject to randomness in processing times, as well as dela ys in the observation of system performance and in the execution of control actions, We assume management has the dual objectives of maximizing profit s (revenues on completed work less the costs of labor employed) and keeping cycle times below a predetermined ceiling. In order to achieve these objec tives it observes the state of the process and adjusts its labor force acco rdingly. Management must chose which of several aspects of process performa nce to measure (cycle time, backlog, or demand) and the parameters governin g the control process. Our analysis highlights the interactions among the d emand environment faced by the process [e.g., random or seasonal], the cont rol signal chosen [e.g., cycle time of backlog], and the type of control us ed [e.g., proportional or differential]. Our results suggest that, regardle ss of the demand environment, a control process based on system backlog is generally more robust than the alternatives in the sense that adequate perf ormance is achieved over a broader range of control parameters. We also fin d that, in most cases, proportional control by itself is inadequate to prov ide effective performance and that differential control is a necessary adju nct. We conclude the article with a discussion of the managerial implicatio ns of this research. Copyright (C) 2001 John Wiley & Sons, Ltd.