Extended Abstract:

Time is one of the more salient constraints on managerial behavior, although surprisingly, not much has been written about it in the theoretical management literature (save for a few exceptions such as Mintzberg's 1973 study). Time as a scarce resource has been examined by economists (e.g., Ghez & Becker, 1975). Their analysis, however, treated the allocation of time as a rational decision. Management researchers, on the other hand, are cognizant of the notion of bounded rationality, which is interpreted as an inherent constraint on one's ability to process information. Bounded rationality coupled with time constraints may lead to serious difficulties, especially if one considers managerial decision making in high velocity environments (Eisenhardt, 1989).

Time constraints coupled with the difficulty of attending to many tasks simultaneously, led researchers to describe managerial work as "putting out fires". The idea is that they allocate their attention only to those projects or issues that they perceive as pressing problems (Radner, 1976). Managers may put out fires by dealing with the biggest problems first until those situations are under control and then attending to the next "fire". The "putting out fires" metaphor invokes also the idea that managers may search for problems that need attention rather than attend to projects before they become problems. Radner [1976] used also the notion of a thermostat to describe how problems capture the attention of managers. Both notions, putting out fires and the thermostat model suggest that managers are focusing on issues and problems in an attempt to meet some minimum acceptable levels of performance. This idea draws on the notion of satisficing (March & Simon, 1993; Simon, 1955) which is tantamount to attempting to reach a certain level of performance that is deemed sufficient or "good enough".

In this paper we deal with the problem of attention and effort allocation by a manager who wishes to pursue an important "long term" goal such as developing a rather innovative R&D project at a high level of complexity. The manager would like to devote all his/her attention to this particular project yet has to monitor work on other "short term" projects in a "maintenance" mode. That is, the manager needs to make sure that work on these short term projects continues and doesn't fall below a particular level. If the short term projects are taken care of, that is, performance on these short term projects is above a certain level, the manager can allocate his/her attention to the long term project. We model the performance of the short and long term projects as stochastic processes. Based on the model we derive implications for effective attention and effort allocation procedures. Thus, our model relates to the two aspects raised above, namely, time constraints on the one hand, and the importance of target levels of performance on the other.

To deal effectively with multiple tasks and goals, managers need to develop ways of setting priorities by some methods. Examples of such priority setting mechanisms abound in the study of queueing systems. Managers need to focus their attention on a subset of all problems and issues they face if they wish to make real progress on some of them. A naive model of managerial behavior may assume that managers are capable of monitoring many projects simultaneously. This idea of managers as parallel distributed information processing systems is appealing but may be limited in its applicability to simpler cases. As Simon [1992] noted, the question whether information processing is done in a serial manner or through a parallel distributed system depends on the complexity level of the problem being processed. If "higher" level non-automatic functioning is required, attention may need to be focused on some small part of the problem and this may lead to a serial form of processing, and hence the need for priority setting. Operating within time constraints further aggravates the problem, that is, the inability to monitor simultaneously many projects requires priority setting. Further, working on the "long term" project of a high level of complexity, mandates a serial processing model, at least when work on this project is concerned.

In this paper we present a simple model that follows the above description to analyze the effect of interruptions on managerial performance. The model yields some predictions that are examined using simulations. The thrust of the findings is that alternative rules lead to differential patterns of interruption. Assuming that the ability to schedule uninterrupted periods of working on the long term goal is desirable, we show that certain rules lead to this pattern more effectively than others.