-4.1- THE GENERAL MODEL USED BY EXPERIENCED MODELERS
Both the newness of CBM in organizations and the lack of objective research in the field make a synthesis difficult. Since the problems mentioned are interdisciplinary, few researchers are capable of investigating the problems. The following proposal is a synthesis of the work of practitioners and pioneers in the field, and is intended to represent a general design rather than a particular corporate model. This generalized model represents essentially only large scale organizations, but with the decreasing cost of generating such models, the use of models can be expected to increase in smaller firms.
Figure 4-1 indicates the flow of the model generally found in organizations. The trial balance calculation is used to produce the financial statements, but they are also used as input to the reports that include variables which are not included in the general accounting model. The top level simulation module is maintained independently of the general accounting system to avoid conflicts with the reporting system. This is in recognition of the limitations of historically-based accounting data.
Figure 4-2 indicates how the corporate state, using the general model, is calculated. The simple, one-directional process of the accounting model treats the organization as a homogeneous system, with cash flows as the input and output of subsystems, and the subsystems are structurally linked together. This emphasis on structure makes the accounting based model a poor decision model for the matrix organizational structure with multiple goals.
Simon introduced the concept of the three hierarchical levels of decision making: strategic, tactical and operational. Every manager spends some time in each of these types of decisions, but top management spends the majority of its time in decisions at the strategic level, while first level managers spend most of their time in operational detail. The question is how to coordinate the decisions made at various levels which are made on the basis of bounded rationality (Simon's term for the limited knowledge of the decision maker). Figure 4-3 shows two variations of the hierarchical decision model, the interactive programming approach and the state space approach.
Interactive programming methods have demonstrated an ability to build consensus when used in group decision making, while the state space approach is more quantitative. Figure 4-4 provides a flow of decision making activities through three levels within an organization. However, this approach does not explicitly indicate actual strategic choices being made at a level below top management. The more realistic view is that strategic decisions are made at each level and some are not communicated upward. This filtering of information is necessary due to the limitations of the ability of individuals to handle details. The most important feature of this approach to strategic decision making is that communication of preferences should occur in both directions. Since hierarchical organizations are pyramid in shape, some filtering of detail isessential to prevent the drowning of high level managers in a sea of detail.
Since strategic, tactical, and operational decisions must be made at every level, varying in proportion with the position, it may help to look at the phases of decision making, rather than focusing on the position of the decision maker within the management hierarchy. Figure 4-5 shows the phases of decision making according to Simon. The greater the contact with the environment, the more extended the intelligence activity in searching the environment; but the essential characteristics of the decision process remain regardless of the management level of the decision maker. The problem remains of how to communicate and coordinate guidelines and alternatives across the boundary levels of the organization.
Figure 4-6 attempts to provide some additional detail concerning what should occur within each decision level of the hierarchy. The key feature here is the use of state space analysis with Simon's decision taxonomy. The output, unlike Simon's simple decision structure, is to order the alternatives, using the alternative of highest preference to set the goal variables for the next level below in the hierarchy. The feature vector at each level of decision would probably contain no more than 10 variables. (Although there is some evidence [John Hayes 1962] that decision quality rapidly decreases when more than 4 variables are used, the maximum number should be set somewhat higher.)
-4.2- BUILDING ON THE BASIC MODEL BASED ON THE TRENDS
It is possible to trace the trends which have shaped the growth of OEM. Eased on this historical evidence, and assuming that these trends will continue in the same direction we can anticipate the characteristics of the OEM system which will serve the needs of organizations in the future. The following section reviews the major trends in OEM and the factors which determine the rate of change in CBM.
The major trends in CBM are:
1. an exponential growth in the number of corporations that use computer based models.
2. acceptance of modeling as a standard tool by managers.
3. preponderance of modelers with more general backgrounds (not computer scientists)
4. the decline of the use of minicomputers (and increase in mainframes and microcomputers).
5. increasing use of modeling for less than 5 year planning and a decline in the use of models in production and marketing.
6. power, flexibility, and ease of use being the most valued attributes of a model and mathematical complexity being given the lowest value.
7. a trend toward subsidiary standardization and integration with a centralized model.
If these trends continue, the use of OEM in corporations will both continue to expand into new corporations and be increasingly accepted in the corporations in which it is employed. Modeling can also be expected to become more standardized and more easily integrated with other models. One negative possibility is that OEM may be relegated to narrower functional areas and shorter planning periods unless the trends are changed.
The factors affecting the development of CBM are:
1. government regulations
2. technological advance of the microprocessor
3. increasing sophistication of managers who used the model
4. managerial fashion
5. declining costs of hardware and software
The relative importance of each one of these factors is rooted in the culture in which the organization resides. These factors shape the development of OEM and so they may modify the rate at which these trends develop in different cultures. Experience indicates that other countries follow the lead of the U. S. in developing organizational structures and managerial trends, and so the use of CBM by foreign MNC's can be expected to mirror the development in the U.S..
The technological base in organizations is changing rapidly and innovation is becoming critical to the success of previously regulated industries (e.g., banking). The change in the technology and the rate of innovation should be followed by a change in organizational structure, according to Mintzberg [40]. The configuration of the organization also needs an information system that is adapted to the new organization. The trend in data processing appears to be towards distributed processing systems with nodes of increasing processing capability. Ultimately, the cluster of nodes will make up the system rather than act as preprocessors for the system. In such a computer configuration, the node could be used as a personalized decision support system.
Innovation in the base technology of a business requires that the new skills be imported, which means either higher costs to the organization if purchased outside or a higher personnel turnover rate. In a matrix structured organization, the informal relationships developed over time are necessary byproducts of the increased flexibility. A high rate of employee turnover would impede the development of those informal relationships. The ideal solution would be to find cost effective means of importing the expertise necessary to the stable processes of the firm, changing the expertise without the high costs of retraining, and keeping the level of employee turnover low.
Ginzberg [23] studied the effect of user expectations on the success of MIS implementations because several studies indicated that the user expectations in the definition state of development were strong predictors of final user satisfaction with the MIS. He feels that there is a gap between the ability to develop new managerial technology and the ability to get managers to use it. It appears that effective use of new managerial technology depends on the degree to which this gap can be narrowed. The gap can be narrowed either by tailoring the technology as closely as possible to how the manager currently operates or by somehow inducing the manager to accept the new complexity.
A personal DSS would provide the flexibility for tailoring the decision support to the information needs and the technical sophistication of the individual manager without compromising the entire MIS system. It is self- defeating to implement a decision support system that has as its base the lowest common denominator of managerial complexity and no capacity for expansion as the manager's knowledge increases, or more (or less) knowledgeable managers take his place. A personal decision support system could be tailored to the individual manager and follow him as he moves to different positions within the company.
The personal decision support system (PDSS) must be capable of interfacing with a more generalized corporate model so that the information available within the station can be used at higher levels of decision making as well as accessed by other individuals whose decisions may depend on the information contained in the PDSS. This requires a common language base between the central (OEM) model and the PDSS.
The creation of a PDSS that utilizes both an expert system and natural language facilities would reduce the gap between the specific skills of the manager and the skills required by the organization. A manager who develops the general skill of using an expert system could be given an expert system on any area that was useful to the organization. A natural language interface would reduce the need for the manager to learn the particular jargon on any given area of expertise. The performance of the manager could be based on the CSF's developed for the position. If a manager left the position, the developed workstation could be retained so that the breaking in time for newly promoted managers could be reduced. This approach also would naturally lend itself to a modular design that can be integrated for larger scale planning.