2.1- SURVEYS OF CORPORATE MODEL USERS
2.1.1- Trends in Corporate Models
The history of computer based modeling (OEM) is short. There is little formal research on the effects of OEM and most insights into the process depend upon surveys, case studies and anecdotes. Drawing on these, one can derive useful insight into CBM.
A survey by Grinyer and Wooler [24] indicated that none of the UK companies surveyed had developed OEM's before 1963, and that by 1974 only 8% of the "Times 1000" had developed models. Gershefski [22] indicates a similarly slow development in the U. S.. In 1969, only 63 out of 1900 (approximately 3.3%) U. S. companies used CBM. A survey completed in 1981 [33] of the financial officers of the 600 largest U. S. companies indicated that 85% of the corporations surveyed used CBM. In a UK survey completed in 1983 (this time a stratified sample of the Times 1000) by Grinyer [25], almost all large companies had adopted some form of OEM.
The exponential rate of adoption of the OEM has been accompanied by a change in the attitude of management towards OEM. Their opinion of OEM had changed from viewing OEM as a novelty to seeing it as a standard tool.
The backgrounds of the modelers had changed from modeling specialists to staff members as the methods became more standardized. In the UK, where accountants traditionally perform the roles of providing legal advice on taxation, accounting and finance, accountants had extended their roles in the organization to include the role of modeler. Over half of the model developers were accountants (who include finance directors in the UK and who also provide legal advice to the corporation).
The most radical change in hardware has been the reduction of use of minicomputers in service bureaus. The dominant hardware used in OEM become mainframes or microcomputers. The low cost and high data security are the appeal of the microcomputers. The popularity of the spreadsheet has contributed to the rapid adoption of OEM's by small and medium-sized businesses. But there is difficulty in transferring the data and/or model to a mainframe for further processing if the company's needs exceed the capacity of the microcomputer. Service bureaus have been reluctant to modify their software for the microcomputer as it was in direct competition with their own services. Recently the service bureaus have recognized that the trend away from service bureaus is not likely to be reversed and have brought out modeling products for the microcomputer. This, however, still does not resolve the problem of transferring the data and/or the model to mainframes.
The features reported as having high value to modelers, according to Grinyer's survey [25] were (in descending order):
1. ability to change format on reports
2. ability to consolidate routines
3. ability to do conditional branching
4. ability to do loop functions
5. ability to suppress the generation of reports
6. automatic renumbering
These categories closely corresponded to the frequency of usage of the facilities and indicate that flexibility and ease of programming were more highly valued than speed, size or accuracy. The least frequently used and least valued feature of the modeling system was linear programming. The most :important features in a modeling package were ease of use and power of the package (with cost running further behind). (Note, the most recommended package was FCS with Visicalc running a poor second.)
The most dramatic change between the results of Grinyer's survey in 1973 and his survey in 1982 were that the two areas of increasing usage are financial planning up to 1 year and project evaluation. The use of the OEM in other areas declined substantially, particularly as an aid in production and marketing decisions. This might indicate a limitation in the current state of the art of OEM or it may indicate that there is resistance on the part of the managers. In either case, the OEM is becoming more closely identified with short range planning and control.
Meyer [38] identifies three time horizons in OEM: one to three months, one year, and over 5 years. Holloway [28] also suggests a three tier system, but with (1 yr., 1-5 yrs, and >5 yrs. Klein reports only two categories, one year budgets and 5 year plans. One recent large scale implementation of a corporate model did not use a planning horizon of greater than 5 years [511]. The trend towards shorter and more narrowly defined models appears to be confirmed by these authors.
Changes in the logic of the models were made, according to the surveys, by the users of the models. Users consisted primarily of accountants, planners, and computer personnel; the most rapid change between the two surveys in users was an increase in accountants and computer personnel users. The trend is towards subsidiary standardization and integration with the centralized corporate models. The findings of Grinyer are in general confirmed in the study by MacGregor [35] and the latter provides the additional insight that the statistical analysis features of OEM is seldom used because the users either lacked the necessary knowledge of statistics or were suspicious that the statistics were manipulated by other departments.
2.1.2- Factors Affecting the Development of CBM
Following the large number of failures in the financial market and bankruptcies, the federal government began creating reporting requirements to protect investors and depositors. Legitimate differences between industries made reporting in detail necessary. New forms of wage taxation (e.g., social security and workmen's compensation) required more detailed accounting of wages.
Following WW-II a series of laws protecting the consumer, the environment, the wage earner, and other special interest groups placed additional demands on organizations to maintain increasingly complex sets of records to show their compliance with those laws. In the U. S. the government reporting requirements increased in volume and complexity in proportion to the capacity of the computer to meet those demands as Klein [33] suggests. According to Klein's survey, few of the models still in use were over 10 years old.
The specific causes for initially adopting models, as reported by the respondents to his survey, varied with the industry. The utility companies had to request increases from the utility commissions more frequently as inflation became more entrenched. The price controls of 1974 appeared to have the greatest influence on the industrial firms, as they used the models to support requests for price increases. Transportation companies, which have not faced specific regulatory problems, are dubious of the benefits of modeling. In this initial phase, the largest firms, which could afford the initial costs of computerization, attempted to meet the existing demands of government regulators more cost effectively by developing their own models.
As the government regulators became aware that the largest corporations had demonstrated the capability for supplying more detailed information regarding the cost and pricing structure, government regulators drafted regulations requiring similar reporting from all companies. Since computer costs were rapidly dropping, CBM in the majority of cases became cost-effective for supplying the information, and the models were readily adopted. However, many of the early OEM projects were shelved after completion indicating that the requirements of government regulations alone were not sufficient to motivate firms to continue OEM projects. This indicates that other motivating factors may have been involved.
The exponential growth of OEM appears to be multifaceted. Grinyer states: "Models became cheaper, more quickly produced, more flexible, more easily understood by managers and consequently more readily accepted, as the lessons of the failure of some of the early, large-scale, general-purpose models became assimilated." [10, ch.23]
Grinyer also suggests that fashion may have been one of the driving forces. Keen and Morton said that simulation as a technique has been oversold because it is easily understood, and hence, readily adopted by managers, even where inappropriate. This paints a rather passive and reactive style of management.
2.1.3- The Qualitative Side of CBM
The motivating factors behind the adoption and spread of CBM include the external environment (government regulations), the technology (the development of the microprocessor), the internal environment (the level of technical sophistication of managers), and individual differences (managers attempting to appear progressive by being fashionable). The value managers place on employing 0MB is more consistent than their reasons for adopting the OEM approach. Klein's survey [33] indicates that the user/managers believe that the value of the model lies in the effect it has on the quality of the judgments based on the models. It may seem a contradiction that a quantitative method yields a qualitative benefit. The qualitative benefit is derived from qualitative aspects of the model, i.e., it increases communication. The message is clear: for a model to be considered useful, it must first lead to higher-quality judgments, as perceived by the decision maker. Identifying the perception of higher quality judgments is not a simple matter. The perception of quality is based, in part, on the economic climate. In periods of expansion, CBM is used to explore methods of increasing revenue, while in periods of contraction, CBM is called upon to provide cost control. Since the former is directed outside the organization, it is more global in aspect, while the latter is more application specific. Unless some continuing effort is made to integrate both aspects of OEM, the model deteriorates into an extension of the chart of accounts and becomes useless for planning, or becomes so detached from the accounting data that the model is no longer firmly grounded in reality.
Another way in which corporate models expand is through the creation of ;he ad hoc models. These models are created to deal with special problems. Their specialized nature is reflected in the very specialized information which comes from applying this type of model. The ad hoc models are not created with an intent of integrating with either long or short term models of the firm and are difficult to combine either with each other or with a larger model of the firm.
The integrated model is needed because, in the long run, both cost control and global planning are needed to manage the firm. The problem of satisfying the requirement that CBM lead to higher quality judgments is essentially a question of how to retain the level of integration (or increase it) while satisficing the demands of management over the short run. A carefully chosen series of ad hoc models may provide the bridge to integrated CBM. The ad hoc models, in essence, could become the personalized decision support system of each manager with the general goal of increasing the quality of the judgments made by the manager but would be created and maintained with the explicit intent of providing information in a form which can be easily integrated with other models.
Another factor to consider is that the cyclical shifts between the short and long range perspectives of the model reduce the quality of judgments based or. the model. Top management should be educated about the tradeoffs between the short run and long run CBM modifications and the importance of maintaining an integrated model. This may require that CBM groups be funded as R&D costs, as fixed costs over the short run, rather than as a staff function which can be used as a means of "putting out" decision making "fires." This may be most critical in companies which must include physical flows in the model because the cost of data collection is naturally higher [24, ch.10, sec.5] and some of the necessary data is not ordinarily collected for accounting purposes. The extra cost of increased data collection and maintaining a stable integrated model must be weighted against the value of the higher quality judgments made by managers during periods of contraction and expansion.
2.2- FACTORS RELATED TO THE SUCCESS OF THE CBM
Figure 2-1 shows the factors related to the success of the corporate modeling effort suggested by several authors. The table ranges from the most specific factors identified by Alter to the most general categories of factors identified by Ginzberg. The implementation problems may be detected early and be used to correct deficiencies in the model before they cause the modeling effort to fail. Ginzberg suggests that the monitoring of user expectations can be used as an indicator. The most critical overall factor involved in the success of the CBM system is the support of top management. Top management's support may be necessary to provide some motivation for all parties in the modeling effort to make compromises in the implementation of the model. Corporate signaling through changes in the organizational structure has been used to indicate the support of top management. By having committees reporting directly to the CEO or raising the person in charge of the modeling effort to the office of vice president, the CEO can signal the value he places on cooperation in the modeling effort.
Meyer [38] suggests five principles of successful modeling:
1. simplicity of mathematics 2. awareness of source of data 3. use of the modular approach 4. establishment of credibility with management 5. establishment of the goals of the model.
This approach attempts to set guidelines for model building which attempts to reduce some of the risks identified in Figure 2-1.
The common ingredient of each of the risk factors is the disparity between the technical expertise and the manager's ability to understand and use the new technology. The availability of the technical expertise to the manager on a timely basis appears to be a special problem. Integrating an expert system and a natural language interface into the model could be used to keep both the technical expertise available to the manager and at a level of technical sophistication which is comfortable for the manager.
2.3- LIMITATIONS OF CBM
Alter concludes, on the basis of interviews with implementers, users and maintainers of DSS systems, that interactive online capability can have a substantive impact on group decision making, but perhaps not for individual managers. The decision as to what type of CBM should be done and who should be included should be based on the cost effectiveness of the combination of available people, data, models and technical tools. The overriding concerns in the specification of the appropriate DDS are the power of the system to answer the most critical questions, the accessibility of the answers on a consistent and timely basis, and the flexibility of the system to adapt to changing demands.
Another limitation of OEM is the limited types of answers it can provide. Alter notes the discrepancy between the ideal DSS and what is currently achievable. The real questions a decision maker would like to ask are:
1. What really did happen in the past?
2. What is happening right now?
3. What will happen in the future?
4. What can I do to shape the future to my desires?
but limitations in methods and knowledge of the underlying variables make it impossible to answer the questions directly. The system designer focuses on the following types of questions that can be answered:
1. factual, noninferential
2. factual, inferential
3. causal, inferential
4. predictive
The validity of the answers to these types of questions is what becomes the focus of the decision maker. Since the validity of any model is dynamic both in terms of the constructs used and the content of the measures used to approximate the construct, the decision maker can never be entirely eliminated from the process.