The control function of management can be a critical determinant of organizational success. Most authors discuss control only through feedback and adjustment processes. This article takes a broader perspective on control and discusses the following questions: What is good control? Why are controls needed? How can good control be achieved? If multiple control strategies are feasible, how should the choice among them be made?

After strategies are set and plans are made, management's primary task is to take steps to ensure that these plans are carried out, or, if conditions warrant, that the plans are modified. This is the critical control function of management. And since management involves directing the activities of others, a major part of the control function is making sure other people do what should be done.

The management literature is filled with advice on how to achieve better control. This advice usually includes a description of some type of measurement and feedback process:

  • The basic control process, wherever it is found and whatever it is found and whatever it controls, involves three steps: (1) establishing standards. (2) measuring performance against these standards. and (3) correcting deviations from standards and plans.1
  • A good management control system stimulates action by spotting the significant variations from the original plan and highlighting them for the people who can set things right.2
  • Controls need to focus on results.3

This focus on measurement and feedback, however, can be seriously misleading. In many circumstances, a control system built around measurement and feedback is not feasible. And even when feasibility is not a limitation, use of a feedback-oriented control system is often an inferior solution. Yet, good controls can be established and maintained using other techniques.

What is needed is a broader perspective on control as a management function: this article addresses such a perspective. The first part summarizes the general control problem by discussing the underlying reasons for implementing controls and by describing what can realistically be achieved. In the second part, the various types of controls available are identified. The last part discusses why the appropriate choice of controls is and should be different in different settings.

Why Are Controls Needed?

If all personnel always did what was best for the organization, control — and even management — would not be needed. But, obviously individuals are sometimes unable or unwilling to act in the organization's best interest, and a set of controls must be implemented to guard against undesirable behavior and to encourage desirable actions.

One important class of problems against which control systems guard may be called personal limitations. People do not always understand what is expected of them nor how they can best perform their jobs, as they may lack some requisite ability, training, or information. In addition, human beings have a number of innate perceptual and cognitive biases, such as an inability to process new information optimally or to make consistent decisions, and these biases can reduce organizational effectiveness.4 Some of these personal limitations are correctable or avoidable, but for others, controls are required to guard against their deleterious effects.

Even if employees are properly equipped to perform a job well, some choose not to do so, because individual goals and organizational goals may not coincide perfectly. In other words, there is a lack of goal congruence. Steps must often be taken either to increase goal congruence or to prevent employees from acting in their own interest where goal incongruence exists.

If nothing is done to protect the organization against the possible occurrence of undesirable behavior or the omission of desirable behavior caused by these personal limitations and motivational problems, severe repercussions may result. At a minimum, inadequate control can result in lower performance or higher risk of poor performance. At the extreme, if performance is not controlled on one or more critical performance dimensions, the outcome could be organizational failure.

What Is Good Control?

Perfect control, meaning complete assurance that actual accomplishment will proceed according to plan, is never possible because of the likely occurrence of unforeseen events. However, good control should mean that an informed person could be reasonably confident that no major unpleasant surprises will occur. A high probability of forthcoming poor performance, despite a reasonable operating plan, sometimes is given the label “out of control.”

Some important characteristics of this desirable state of good control should be highlighted. First, control is future-oriented: the goal is to have no unpleasant surprises in the future. The past is not relevant except as a guide to the future, Second, control is multidimensional, and good control cannot be established over an activity with multiple objectives unless performance on all significant dimensions has been considered. Thus, for example, control of a production department cannot be considered good unless all the major performance dimensions, including quality, efficiency, and asset management, are well controlled. Third, the assessment of whether good performance assurance has been achieved is difficult and subjective. An informed expert might judge that the control system in place is adequate because no major bad surprises are likely, but this judgment is subject to error because adequacy must be measured against a future that can be very difficult to assess. Fourth, better control is not always economically desirable. Like any other economic good, the control tools are costly and should be implemented only if the expected benefits exceed the costs.

How Can Good Control Be Achieved?

Good control can be achieved by avoiding some behavioral problems and/or by implementing one or more types of control to protect against the remaining problems. The following sections discuss the major control options.

Control-Problem Avoidance

In most situations, managers can avoid some control problems by allowing no opportunities for improper behavior. One possibility is automation. Computers and other means of automation reduce the organization's exposure to control problems because they can be set to perform appropriately (that is, as the organization desires), and they will perform more consistently than do human beings. Consequently, control is improved.

Another avoidance possibility is centralization, such as that which takes place with very critical decisions at most organization levels. If a manager makes all the decisions in certain areas, those areas cease to be control problems in a managerial sense because no other persons are involved.

A third avoidance possibility is risk-sharing with an outside body, such as an insurance company. Many companies bond employees in sensitive positions, and in so doing, they reduce the probability that the employees' behavior will cause significant harm to the firm.

Finally, some control problems can and should be avoided by elimination of a business or an operation entirely. Managers without the means to control certain activities, perhaps because they do not understand the processes well, can eliminate the associated control problems by turning over their potential profits and the associated risk to a third party, for example, by subcontracting or divesting.

If management cannot, or chooses not to avoid the control problems caused by relying on other individuals, they must address the problems by implementing one or more control tactics. The large number of tactics that are available to help achieve good control can be classified usefully into three main categories, according to the object of control; that is, whether control is exercised over specific actions, results, or personnel. Table 1 shows many common controls classified according to their control object; these controls are described in the following sections.

Control of Specific Actions

One type of control, specific-action control, attempts to ensure that individuals perform (or do not perform) certain actions that are known to be desirable (or undesirable). Management can limit the incidence of some types of obviously undesirable activity by using behavioral constraints that render the occurrence impossible, or at least unlikely. These constraints include physical devices, such as locks and key-personnel identification systems, and administrative constraints, such as segregation of duties, which make it very difficult for one person to carry out an improper act.

A second type of specific action control is action accountability — a type of feedback control system by which employees are held accountable for their actions. The implementation of action-accountability control systems requires: (1) defining the limits of acceptable behavior, as is done in procedures manuals; (2) tracking the behaviors that employees are actually engaged in; and (3) rewarding or punishing deviations from the defined limits. Although action-accountability systems involve the tracking and reporting of actual behaviors, their objective is to motivate employees to behave appropriately in the future. These systems are effective only if employees understand what is required of them, and they feel that their individual actions will be noticed and rewarded or punished in some significant way.

A third type of specific-action control is preaction review. This involves observing the work of others before the activity is complete, for example, through direct supervision, formal planning reviews, and approvals on proposals for expenditures. Reviews can provide effective control in several ways by: correcting potentially harmful behavior before the full damaging effects are felt; or influencing behavior just by the threat of an impending review, such as causing extra care in the preparation of an expenditure proposal. One advantage of reviews is that they can be used even when it is not possible to define exactly what is expected prior to the review.

Control of Results

Control can also be accomplished by focusing on results: this type of control comes in only one basic form, results accountability, which involves holding employees responsible for certain results. Use of results-accountability control systems requires: (1) defining the dimensions along which results are desired, such as efficiency, quality, and service; (2) measuring performance on these dimensions; and (3) providing rewards (punishments) to encourage (discourage) behavior that will lead (not lead) to those results. As with action-accountability systems, results-accountability systems are future-oriented; they attempt to motivate people to behave appropriately. But they are effective only if employees feel that their individual efforts will be noticed and rewarded in some significant way.

Control of Personnel

A third type of control can be called personnel control because it emphasizes a reliance on the personnel involved to do what is best for the organization, and it provides assistance for them as necessary. Personnel controls can be very effective by themselves in some situations, such as in a small family business or in a professional partnership, because the underlying causes of the needs for controls (personal limitations and lack of goal congruence) are minimal. However, even when control problems are present, they can be reduced to some extent by: (1) upgrading the capabilities of personnel in key positions, such as tightening hiring policies, implementing training programs, or improving job assignments; (2) improving communications to help individuals know and understand their roles better and how they can best coordinate their efforts with those of other groups in the organization; and (3) encouraging peer (or subordinate) control by establishing cohesive work groups with shared goals.

Feasibility Constraints on the Choice of Controls

The design of a control system often depends partly on the feasibility of the various types of controls: not all of these tools can be used in every situation. Personnel controls are the most adaptable to a broad range of situations. To some extent, all organizations rely on their employees to guide and motivate themselves, and this self-control can be increased with some care in hiring, screening, and training. Even in a prison, where administrators are faced with a sharp lack of goal congruence and where few control options are available other than physical constraints, inmates are screened so that dangerous ones are not assigned to high-risk positions, such as in a machine shop.

Most situations, however, require reinforcing personnel controls by placing controls over specific actions, results, or a combination of the two. This is where feasibility becomes a limiting factor.

For control over specific actions, management must have some knowledge of which actions are desirable. While it may be easy to define precisely the required behavior on a production line, the definition of preferred behavior for a research engineer cannot be as precise. Being able to keep track of specific actions is also necessary to enforce actions accountability; however, this is usually not a limiting factor, except in rare situations such as a remote outpost, because actions can be observed directly or assessed indirectly through action reports, such as hours worked, sales calls made, or procedural violations.

For control over results, the most serious constraint is the ability to measure the desired results effectively. (Management usually knows what results are desirable.) Ideally, measurements should: (1) assess the correct performance areas — the ones for which results are truly desired; (2) be precise — not determined by only crude estimations; (3) be timely and (4) be objective — not subject to manipulation. While perfect measures are rarely available, reasonable surrogates can often be found or developed. For example, “complaints received” might be a good (negative) indicator of the performance of hotel staff personnel along the customer-service dimension. Significant difficulty in achieving any of these four measurement qualities, however, can lead to failure of a results-oriented control system.

Figure 1 shows how the two factors most limiting control feasibility — knowledge of desirable actions and the ability to measure results on the important performance dimensions — can influence the choice of controls used.5 The most difficult control situation, shown in box 4 of Figure 1, is one in which the desirable actions are not known and the important result areas cannot be measured well. Only personnel controls (or problem avoidance) are available options. In a research laboratory, for example, success might be difficult to assess for years, yet prescription of specific actions could be counterproductive. Fortunately, in this specific setting, control is not a serious problem because research scientists tend to be professional — well trained and responsible to the standards of their profession. They tend to control themselves, and consequently, control of research laboratories tend to be dominated by controls over personnel.

In box 3 of Figure 1, where knowledge of desirable specific actions is poor but good results measurements are available, control is best accomplished by controlling results. Movie production is a good example. It is probably impossible to dictate what a movie director should do or even to observe his or her behavior and predict whether the finished product will be good. It is, however, a relatively easy task to measure the economic performance of the movie and the artistic merit, if that is a concern. In this situation, the best control system would seem to be a results-accountability system that defines to the director the results expected, holds him or her responsible for achieving them, and provides some reinforcement in the form of compensation and/or recognition.

For similar reasons, results controls tend to be dominant at most upper-management levels. It is usually not possible to prescribe and keep track of the specific actions each manager should be performing, but it is relatively easy to define the results desired, in terms similar to those desired by shareholders.

Specific-action controls should dominate where there is knowledge about which actions are desirable but where results measurement is impossible or difficult, as indicated in box 2 of Figure 1. Consider, for example, control over a real-estate development business where large capital investment decisions are made frequently. Results of these decisions are difficult to measure in a timely, accurate fashion because of their long-term nature; they tend to be inseparable from the results of other actions and are confounded by changes in the environment. However, the techniques of investment analysis are well developed (e.g., net present value analysis with tests of the sensitivity of assumptions), and control may be accomplished by formally reviewing the techniques used and the assumptions made.

How to Choose among the Feasible Options

Often managers cannot rely completely on the people involved in a given area and cannot employ one or more of the avoidance strategies mentioned earlier. When this is the case, the best situation is one in which either specific-action or results controls, or both, can be chosen, as is shown in box 1 of Figure 1. In general, the choice of one or more tools should involve consideration of: (1) the total need for control; (2) the amount of control that can be designed into each of the control devices; and (3) the costs of each, both in terms of money spent and unintended behavioral effects, if any. These decision parameters will be described more fully.

Need for Controls

The need for controls over any particular behavior or operation within an organization depends very simply on the impact of that area on overall organizational performance. Thus, more control should be exercised over a strategically important behavior rather than over a minor one, regardless of how easy it is to control each. For example, controlling the new-product-development activity is far more important in many companies than making sure that the production of existing products is accomplished as efficiently as possible. Consequently, more resources should be devoted to controlling the new-product activity, even though it is a far more difficult area to control.

Amount of Control Provided by Feasible Options

The amount of control provided by each of the control tools depends both on their design and on how well they fit the situation in which they are used. Personnel controls should usually provide some degree of control. But although they may be totally effective in some situations, such as in a small business, they provide little or no warning of failure. They can break down very quickly if demands, opportunities, or needs change.

Specific-action and results controls can provide widely varying amounts of control. In general, reasonably certain (or tight) control requires: (1) detailed specific-action of what is expected of each individual; (2) prevention of undesired actions, or effective and frequent monitoring of actions or results; and (3) administration of penalties or rewards that are significant to the individuals involved.

For example, with specific-action-accountability systems, the amount of control can be affected by changing one or more of the elements of the system. First, tighter control can be effected by making the definitions of acceptability more specific. This might take the form of work rules (e.g., no smoking) or specific policies (e.g., a purchasing policy to secure three competing bids before releasing the purchase order), as opposed to general guidelines or vague codes of conduct (e.g., act professionally). Second, control can be made tighter by improving the effectiveness of the action-tracking system. Personnel who are certain that their actions will be noticed relatively quickly will be affected more strongly by an action-accountability system than will those who feel that the chance of their being observed is small. Thus, constant direct supervision should provide tighter control than would an audit sampling of a small number of action reports some time later. Third, control can be made tighter by making the rewards or punishments more significant to the individuals involved. In general, this impact should vary directly with the size of the reward (or the severity of the punishment), although different individuals may react differently to identical rewards or punishments.

Results-accountability systems can be varied along similar lines. Expected performance can be defined broadly, for instance, with a goal for annual net income. Alternatively, expected performance can be defined in more detailed form by prescribing goals for specific result areas (for example, sales growth, efficiency, quality) and by using line items with short time horizons (e.g., month or quarter). Control is tighter when the performance dimensions for which results are desired are defined explicitly and, of course, correctly: this type of control is particularly effective if well-established results standards are available, perhaps in the form of output from an engineering study, industry survey, or historical analysis. Results-accountability control can also be tightened by improving the measurement of results. This can be accomplished by making the measures more precise. more timely, and/or less subject to manipulation.

In addition, reviews can be used to provide either tight or loose assurance. Tight assurance is more likely if the reviews are detailed, comprehensive, and frequent.

Of course, managers do not have to rely exclusively on a single type of control in a control system. Use of more than one type of control — in effect. overlapping controls — will often provide reinforcement. For example, most organizations rely on selecting good people. establishing some set procedures. implementing some accountability for results, and reviewing some key decisions before they are made.

Costs: Outlay and Behavioral

The cost of a control depends on two factors: the incremental dollar cost of the tool and the cost of any unintended behavioral effects. The actual dollar cost of a control might be considerably less than it first appears because some devices that provide control may already be in place for other reasons. For example. a budgeting process for a small firm does not have to justify its' cost on the basis of control reasons alone. Creditors probably already require pro forma financial statements. so the incremental cost might involve only additional detail (e.g.. down to the operations level) and involvement of a greater number of participants.

The costs of any unintended negative effects must also be considered. and these can be very significant. It is beyond the scope of this article to provide an exhaustive enumeration of the many negative side effects possible. Indeed, they come in many different forms, but it is nevertheless useful to mention a few examples.

A common problem with specific-action controls is that they cause operating delays. These can be relatively minor. such as delays caused by limiting access to a stockroom. but they can also be major. For example, after the executives of Harley-Davidson Motor Company bought the firm from AMF, Inc., they found that they were able to implement a rebate program in ten days. rather than the six to eight weeks it would have taken with all the reviews required in the multilayered AMF organization.6 Obviously, where timely action is important, delays caused by control processes can be very harmful.

Another problem with specific-action controls is that they can cause rigid. bureaucratic behavior. Individuals who become accustomed to following a set routine are not as apt to sense a changing environment, nor are they likely to search for better ways of doing the tasks at hand in a stable environment.

Results controls can create severe, unintended negative effects when all the measurement criteria are not met satisfactorily. Perhaps the most serious common problem is a failure to define the results areas correctly. This causes “goal displacement,” a situation where individuals are encouraged to generate the wrong results — in response to the goals defined in the control system — rather than those results truly needed by the organization. For example, a department store introduced an incentive compensation plan to pay employees on the basis of sales volume. The immediate impact was indeed an increase in sales volume, but the increase was accomplished in ways that were inconsistent with long-term organizational goals. The employees competed among themselves for customers and neglected important but unmeasured and unrewarded activities such as stocking and merchandising.7 Another common example of goal displacement is caused by the practice of rewarding managers on the oft criticized return-on-investment criterion.8

Data distortion is another dangerous potential side effect of results controls. If the measurement methods are not objective, then the employees whose performances are being measured might falsify the data or change the measurement methods, and, in so doing, undermine the whole organization's information system.

Many of the ramifications of these unintended effects of control systems are not well understood, and their costs are very difficult to quantify. However, consideration of these effects is an important control-system design factor: they cannot be ignored.

Where Does Feedback Fit In?

Because feedback does not appear prominently in the preceding discussion, it is useful for clarification purposes to consider where feedback fits in, Control is necessarily future-oriented, as past performance cannot be changed, but analysis of results and feedback of variances can often provide a particularly strong addition to a control system. A prerequisite, of course, is the ability to measure results, so feedback can only be useful in the situations presented in boxes 1 and 3 of Figure 1.

There are three reasons why feedback of past results is an important part of many control systems. First, feedback is necessary as reinforcement for a results-accountability system. Even if the feedback is not used to make input adjustments, it signals that results are being monitored. This can heighten employee awareness of what is expected of them and should help stimulate better performance.

Second, in repetitive situations, measurement of results can provide indications of failure in time to make useful interventions. This is shown in the simple feedback control model presented in Figure 2. When the results achieved are not satisfactory, the inputs, which include the specific actions and types of persons involved, can be changed to provide different results. Obviously, these input adjustments are more likely to improve results when there is a good understanding of how inputs relate to results; otherwise, the interventions are essentially experiments.

Third, analysis of how the results vary with different combinations of inputs might improve understanding of how the inputs relate to results. This process is depicted in loop A of Figure 3, a slightly more complicated feedback control model. As this input/results understanding improves, it provides the opportunity to shift the control system from a results-oriented to a specific-action-oriented focus. If managers discover that certain specific actions produce consistently superior results, then it might be beneficial to inform employees of the specific actions that are expected of them, for example, by publishing these desired actions in a procedures manual. The greater the knowledge about how actions bring about results, the greater the possibilities of using a tight, specific-action-oriented control system.

Note that these latter two reasons for analyzing feedback — for making interventions and for learning — are only useful in situations that at least partially repeat themselves. If a situation is truly a one-time occurrence, such as a major divestiture or a unique capital investment, management has little use for feedback information. In these cases, by the time the results are available, it is too late to intervene, and a greater understanding of how results are related to inputs is not immediately useful.

There are other circumstances where feedback need not, and perhaps should not, be a part of a good control system. In many cases, although feedback control systems are not really feasible, they are used anyway. This occurs because of the consistent tendency “to concentrate on matters that are concrete and quantifiable, rather than intangible concepts,” which may be equally or more important.9 Invariably, this will lead to dysfunctional effects, as will all other failures to satisfy the measurement criteria or to define results appropriately.

Cost considerations also commonly lead to decisions not to include feedback in a control system. The design, implementation, and maintenance of results-tracking information systems can often be very expensive. Thus, it is not feasible to have feedback as part of every control system, nor is it necessarily desirable even when feasibility constraints are not present.

The Design Process

As discussed at the beginning of this article, management control is a problem of human behavior. The challenge is to have each individual acting properly as often as possible. Thus, it seems logical to start the control-system design process by considering the personnel component of the organization by itself. In some situations, well-trained, highly motivated personnel can be expected, with a high degree of certainty, to perform their jobs satisfactorily without any additional control steps being taken. A confident reliance on personnel controls is a very desirable situation because additional controls cost money and may have undesirable side effects.

If, however, management determines that personnel controls should be supplemented, the first step should be to examine the feasibility of the various control options. To do this, management must assess two factors: how much is known about which specific actions are desirable, and how well measurement can be accomplished in the important performance areas. This feasibility test might immediately determine whether the controls that can be added should be oriented toward specific actions or results. Control can be made tighter by strengthening the controls in place, along the lines discussed earlier, or by implementing overlapping controls, such as controls over results and specific actions.

In most cases, management has some, but less than complete, knowledge of which specific actions are desirable and some, but not perfect, ability to measure the important result areas. This situation usually calls for implementation of both specific-action and results controls, with feedback loops to improve understanding of the relevant processes.

An Example: Control of a Sales Force

The above observations about control can be illustrated by describing how control of a sales force might work. Generally, personnel controls are some part of every sales force control system. Consider, for example, this statement by a sales and marketing consultant:

I think I can tell a good salesman just by being around him. If the guy is experienced, confident, well-prepared, speaks well, maintains control of situations, and seems to have his time planned. I assume I have a good salesman.

If a sales manager feels confident about all of the salespeople employed, he or she might wish to allow personnel controls to dominate the control system. This is likely, for example, in a small business with a sales force comprised solely of relatives and close friends. But most sales managers are not willing to rely exclusively on hiring and training good people.

What controls should be added? The answer, of course, depends on the type of sales involved. In a single-product, high-volume operation, the volume of sales generated is probably a good simple factor on which to base a results-oriented control system. It provides a reasonable, although not perfect, surrogate for long-range profitability, and the measurements are very inexpensive because the data are already gathered as a necessary input to the financial reporting system. The results-accountability system can then be completed by providing reinforcement in the form of sales commissions. This simple solution will also work where multiple products with varying profitabilities are involved, if the commission schedules are varied so that rewards are assigned in proportion to the profitability of the sales generated.

Consider, however, a situation where salespeople sell large-scale construction equipment and where sales come in very large but infrequent chunks. A commission-type, results-accountability system is still feasible. Measurement of results is not difficult and can be accurate to the penny. The amount of control provided, however, is not high because the measurements fail on the timeliness dimension. Because sales are infrequent, zero sales is not an unusual situation in any given month. Therefore, a salesperson could be drawing advances on hypothetical future commissions for many months without performing any of the desired promotional activities.

Two solutions are possible. One is to augment the commission system with some specific-action controls, such as activity reports. Some activities are probably known to be desirable, such as the number of hours worked and the quantity of calls made. If the product mix and market environment are fairly stable, then requiring and monitoring activity reports is not as costly as it might seem, because it could provide an important side benefit — an activity-oriented data base. The patterns in this data base can be analyzed and compared with results over time to add to knowledge about which activities yield the best results.

An alternate solution is to improve the results-accountability system. It might be possible to define some factors that are strong predictors of sales success, such as customer satisfaction with the salesperson or customer familiarity with the company's products. Measurement of these intangibles, of course, would have to be done by surveying customers. Even though these measures do not directly assess the desired result area (long-range profitability), and measurement is imprecise, they could provide a better focus for a results-oriented control system than a sales-generated measure because of the improvement in timeliness. Over time, it is likely that the choice of measures and measurement methodologies could be improved. The advantage of this results-oriented solution over an action-oriented system is that it is more flexible and less constraining to the salespeople; they can continue to use styles best suited to their personalities.

Conclusions

This article has taken a new look at the most basic organizational control problem — how to get employees to live up to the plans that have been established. In the course of discussion, the following major points were made:

  1. Management control is a behavioral problem. The various control tools are only effective to the extent that they influence behavior in desirable directions.
  2. Good control can often be achieved in several different ways. In some circumstances, the control problems can be avoided, for example, by centralizing or automating certain decisions. If problems cannot be avoided, one or more types of controls are usually desirable or necessary. The options can be classified according to the object of control, labeled in this article as specific actions, results, and personnel.
  3. Not all types of controls are feasible in all situations. Figure 4 presents the questions to ask when assessing the feasibility of control types. If none of the controls is feasible, the probability of undesirable results occurring is high.
  4. Control can be strengthened either by employing a tighter version of a single type of control or by implementing more than one type of control. However, tighter control is not always desirable because of additional system costs and the potential of undesirable side effects, such as destruction of morale, reduction of initiative, or displacement of employee focus toward measurable result areas only. Some of the qualities, benefits, and costs of each of the major control types are listed in Table 2.
  5. The basic management control problems and alternatives are the same in all functional areas and at all levels in the organization, from the lowest supervisory levels to the very top levels of management. The best solutions, however, vary between situations.

An understanding of control can be an important input into many management decisions. For example, control problems should be considered in making some types of investments. An investment in an operation in which control is very difficult — such as a highly specialized and technical area where control must depend heavily on personnel controls — is, by definition, risky. Thus, investments in such areas should promise high returns to compensate for this risk.

Similarly, control considerations should affect the design of the other parts of the management system. Consider, for example, the organizational structure. If independent areas of responsibility cannot be carved out as part of the organizational structure, results-accountability control systems will not work well because employees will not feel that their individual actions have a notice able effect on results. (It should be noted that many of the prescriptions calling for “responsibility accounting” only provide the illusion of results independence because of the many allocations of the costs and/or benefits of shared resources.) If independent areas of authority are not established, specific-action-accountability control systems cannot work. This principle underlies the internal control principle of “separation of duties.” In addition, if tighter reviews of specific actions are necessary for adequate performance assurance, it is likely that the supervisory spans of control will have to be reduced, Similar observations can be made about other management functions, but they are beyond the scope of this article.

This article has attempted to provide a new look at this basic, but often overlooked, management problem. The control area is decidedly complex, and there is much that is not known about how controls work and how employees respond to different types of controls. For example, it would be worthwhile to know more about how controls can be designed to maximize the amount of control provided while minimizing the cost in the form of employee feelings of lost autonomy. However, an increased awareness of the control problem, of what can be accomplished, and of the options available should provide a new perspective that will suggest ways to improve control systems and overall organizational performance.

References

1. See H. Koontz, C. O'Donnell, and H. Weihrich. Management, 7th ed. (New York: McGraw-Hill, 1980). p. 722.

2. See W. D. Brinckloe and M. T. Coughlin, Managing Organizations (Encino, CA: Glencoe Press. 1977). p. 298.

3. See P. F. Drucker. Management: Tasks, Responsibilities, Practices (New York: Harper & Row. 1974). p. 497.

4. A recent summary of many of the findings in this area (illustrating such cognitive limitations as conservative revision of prior subjective probabilities when new information is provided. and the use of simplifying decision-making heuristics when faced with complex problems) is provided by W. F. Wright, “Cognitive Information Processing Biases: Implications for Producers and Users of Financial Information,” Decision Sciences (April 1980): 284–298.

5. A similar scheme is presented in W. G. Ouchi, “A Conceptual Framework for the Design of Organizational Control Mechanisms,” Management Science (September 1979): 833–848.

6. See H. Klein, “At Harley-Davidson, Life without AMF Is Upbeat but Full of Financial Problems,” Wall Street Journal. 13 April 1982, p. 37.

7. See N. Babchukand W. J. Goode. “Work Incentives in a Self-Determined Group,” American Sociological Review (1951): 679–687.

8. For a summary of criticisms of return-an-investment (ROI) measures of performance, see J. Dearden, “The Case against ROI Control,” Harvard Business Review, May–June 1969, pp. 124–135.

9. See D. Mitchell. Control without Bureaucracy (London: McGraw-Hill Book Company Limited, 1979), p. 6.

Acknowledgments

The author wishes to acknowledge Robert N. Anthony, Peter Brownell, and Martha S. Hayes for their helpful comments.