Shared capacity resource reallocation in a decentralized service system

Short‐term capacity sharing on an occasional basis is a production planning strategy that is particularly attractive to managers of service organizations that employ relatively standardized resources within a highly unpredictable business environment. These resources may or may not be scarce and/or mobile, e.g., medical diagnostic equipment versus rental vehicles. This scheme implies cooperation in setting up the necessary administrative procedures. Beyond that, however, the participants are free to initiate resource transfers under almost purely competitive conditions, i.e., without the intervention of a higher level authority. An example is an industry‐wide computer system, operated since the late seventies by the American Trucking Association, which enables common carriers to reduce empty or partially loaded mileage. This paper is an initial investigation into the factors which affect the efficiency and effectiveness of decentralized resource reallocation schemes. Two groups of experimental factors are examined. The first group is comprised of features of the decison‐making processes of the independent managers while the second refers to characteristics of the organizational and operational arrangements. The model is a significant abstraction of the problem. For instance, all managers use the same process which is based upon simple feedback control upon the utilization factor at each service organization. However, they do not share information nor do they interact with one another. Instead they deal with a central depot which acts as a neutral “traffic policeman” of sorts. Requests for resources wait at this depot until available resources arrive from other organizations. Service at the depot is FCFS while transportation times between the eight service organizations and the depot are fixed and the same for all. Demand (and service) at each organization observes the same stationary Poisson (and Erlang‐k, respectively) distribution. Furthermore, the total number of resources remains fixed. Simulation was the research method employed. Sixty‐four simulation runs (i.e., experiments) were conducted. This number was a one‐half factorial design of seven experimental factors each at two levels. The first group consisted of two factors measuring the sensitivity of the upper and lower control limits and two factors measuring the dependency of the decision‐making process upon the status of the depot before as well as after a resource request was issued. The second group consisted of factors representing the number of organizations, the transportation time interval, and the homogeneity of the demand distributions at the organizations. Three classes of operational criteria were used: customer service (three criteria), resource utilization (seven criteria), and resource request service at the depot (six criteria). All the criteria refer to total, not individual organization performance. Analysis of variance for each performance criterion was conducted through the use of an additive linear regression model. Coefficients for two‐ and three‐way factor interactions were ignored. In general it was found that the generic decision‐making process was too simple and rigid to have a noticeable effect upon performance (particularly customer service). Still, the results suggest that the design of the decision‐making process is more important than the design of the organizational structure. Since the long‐term behavior of the depot can not be known beforehand, adaptive control limits appear to be a sensible approach. It was also found that some information about the status of the depot should be incorporated in the process. In terms of structure it was determined that a relatively small, compact, and balanced configuration has a favorable impact upon performance.