Designing the Superstructure: Unit Size in Organization

The second basic issue in the design of the superstructure concerns how large each unit or work group should be. How many positions should be contained in the first-level grouping, and how many units in each suc- cessively higher-order unit? This question of unit size can be rephrased in two important ways: How many people should report to each manager? That is, what should be the manager’s span of control? And what shape should the superstructure be: tall, with small units and narrow spans of control, or wide, with large units and wide spans of control?

On this point, the traditional literature was firm: “No supervisor can supervise directly the work of more than five or, at the most, six subordi- nates whose work interlocks,” said Colonel Lydal Urwick unequivocally (1956, p. 41). But subsequent investigation has made this statement seem rather quaint. One study (Holden et al., 1968:95) reported an average span of control of ten for corporate chief executive officers, with a range from one to fourteen. Woodward (1965) found an average of six for the chief execu- tives of the industrial firms she studied, but that measure climbed above twelve in five of the “successful” firms. For the first-line supervisors in the firms in mass production, the average span of control was close to fifty, and it ranged into the nineties in some cases. Worthy reported that the merchandising vice-president of Sears, Roebuck and Co. had forty-four senior executives reporting to him; for the typical store manager, the figure was “forty-odd” department managers (1959:109). And Pfiffner and Sher- wood (1960) noted the extreme example of “the Bank of America, which has over 600 branches throughout California, each of which reports di- rectly to corporate headquarters at San Francisco. There is no intervening area structure with directive powers over the branch offices” (p. 161). In some of these cases, notably the Bank of America and perhaps also Sears, Roebuck, Urwick’s qualification about interlocking work may apply. But certainly not in all.

About the concept of span of control, Pfiffner and Sherwood have commented:

Much blood has been let to reduce the executive’s span with inconsequential results to administrative performance. Yet span of control sails merrily on. There is much written about it. Most consultants tab this as an essential in reform proposals. Students sweat over its definition, mainly because they assume the concept should be more complicated than it really is. Thus, re- gardless of what its merits may be, span of control is so entrenched in the administrative culture that it must be accorded a prominent place in any book on organization, (pp. 155-56)

There is no doubt that the concept merits a prominent place in this book. But there is reason to doubt Pfiffner and Sherwood’s suggestion that it is a simple one. Who should be counted as a subordinate? For example, what about the assistant to, or those whose work is reviewed by the manager even though they do not formally report to him? What about the nonsupervisory aspects of the manager’s job—collecting information, de- veloping liaison contacts, and so on? Does a narrow span of control neces- sarily mean close “control,” as the traditional literature suggested, or might it instead imply that the manager is busy doing these other things? What about the influence of the coordinating mechanisms other than direct supervision on the size of the work unit?

What all this suggests is that the issue is not a simple one and the focus on control is misplaced. Control—that is, direct supervision—is only one factor among many in deciding how many positions to group into one unit, or how many units to group in one larger unit, in both cases under a single manager. Hence, we prefer the term unit size to “span of control.”

1. Unit size in relation to the coordinating mechanisms

Much of the confusion in this area seems to stem from considering unit size only with respect to the coordinating mechanism of direct supervision, not of standardization or mutual adjustment. The traditional management theorists set the tone by implying that control and coordination could be achieved only by direct supervision. What else would have prompted Ur- wick to insist on his “five, or at the most, six” formula?

As has been pointed out repeatedly since the start of our discussion, the five coordinating mechanisms are to some extent substitutable. For example, the manager’s job can be “institutionalized” by standardization; and mutual adjustment within the work group can be used in place of direct supervision from above. We would, of course, expect such replace- ment of direct supervision by another coordinating mechanism to affect significantly the size of a unit. Thus, we should be able to explain varia- tions in unit size largely in terms of the mechanisms used to coordinate work.

We can summarize our conclusions in terms of two basic proposi- tions, one dealing with standardization, the other with mutual adjustment. First, compared with direct supervision, the greater the use of standard- ization for coordination, the larger the size of the work unit. It stands to reason that the more coordination in a unit is achieved through the sys- tems of standardization designed by the technostructure, the less time its manager need spend on the direct supervision of each employee, and so the greater the number of employees that can report to him. With this conclusion, we can rather easily explain Woodward’s finding about the very high spans of control encountered in the mass-production firms. Bear in mind two points about her findings. First, the very wide spans of control were found at the first level of supervision—namely, in those units con- taining the operators themselves. Second, the largest operating units— with an average of almost fifty employees—were found in the mass-pro- duction firms. Those in unit (custom) and in process production had units averaging less than twenty-five and fifteen operators, respectively. Indeed, they had virtually no units even as large as the average for the mass pro- ducers. Now, when we combine this with Woodward’s findings that the mass-production firms were the only bureaucratic ones, the other two being structured organically, we see an evident relationship. Unit size was largest where the work was the most standardized—in the operating cores of the most bureaucratic organizations.

So far, we have discussed only the standardization of work processes. However, our first proposition is not restricted to any special kind of stan- dardization. In other words, standardization of skills and of outputs should also lead to larger unit size. In the case of skills, it stands to reason that the more highly trained the employees, the less closely they need be supervised, and so the larger their work units can be. We see this most clearly in general hospitals and universities. At the time of this writing, sixty of my colleagues and I work in a single unit, which runs smoothly under a single dean with no department heads.

Similarly, we would expect that the more standardized the outputs, the larger can be the size of the work unit. Thus, although the Bank of America justified its span of control of 600 on the basis of encouraging the initiative of its branch managers, we would be on safe ground in assuming that this enormous span of control would simply be impossible without a very tight system of performance (output) control, not to mention the use of all kinds of rules and regulations and of training and indoctrination programs for the branch managers. Similarly, those who shop at Sears well know how standardized that operation is. As Moore, referring implicitly to the role of indoctrination, commented, “Sears can decentralize [that is, release the store managers from close supervision]; everyone thinks alike anyway” (quoted in Wilensky, 1967:60). Chains of banks and retail stores frequently exhibit very wide spans of control precisely because each outlet is a carbon copy of all the others, thereby facilitating standardization.

Thus, we cannot conclude that being a member of a large unit auto- matically frees the individual from close control. Control from the boss, perhaps, but not necessarily from the systems of the technostructure—or even from the person’s earlier training and indoctrination. In fact, the most tightly controlled members of organizations are typically those in the largest units—the operators doing unskilled work in highly bureaucratic operating cores. Even their managers feel the same control. I once spoke to eighty branch managers of large Canadian banking firms on the nature of managerial work; the ensuing discussion period was dominated by one issue—their extreme frustration in being unable to act as full-fledged man- agers, because of the rules imposed on their branches by the corporate technostructures.

Our second proposition is as follows: Compared with standardiza- tion and often even direct supervision, the greater the reliance on mutual adjustment (owing to interdependencies among complex tasks), the smaller the size of the work unit. A relationship between complex interde- pendent tasks and small unit size can be explained in two ways. The obvious one is that, all coordinating mechanisms (especially standardiza- tion) remaining equal, the more interdependent the tasks (complex or not) in a unit, the greater will be the need for contact between the manager and the employees to coordinate their work. Ostensibly, the manager will have to monitor and supervise the unit’s activities more closely and to be more readily available for consultation and advice. Therefore, the manager re- quires a small span of control. This suggests yet another angle on the Sears and Bank of America stories—namely, the absence of interdependence. Geographically dispersed retail branches, each serving its own customers, are neither reciprocally nor sequentially interdependent; far more of them can, therefore, be supervised than, say, the sequentially interdependent departments of a factory. That is why Urwick qualified his principle of span of control with the word “interlocks.”

But there is a second, more subtle explanation for the hypothesized relationship between complex interdependent tasks and small unit size. These kinds of tasks are difficult to supervise, so instead of an increase in direct supervision, they give rise to an increase in mutual adjustment. The employees themselves must communicate on a face-to-face basis to coordi- nate their work. But for such communication to function effectively, the work unit must be small, small enough to encourage convenient, frequent, and informal interaction among all its members. Thus, one study indicated that beyond ten members, groups tend to fragment into cliques—that is, smaller groups—and another found that five to seven members was opti- mal for consensus. Now, organizations, being what they are, designate a leader—a “manager”—for each of their units, no matter how small, even when that person acts as little more than the unit’s official spokesperson. And so, when the span of control of units doing interdependent complex tasks is measured, lo and behold, it turns out to be small.

Let us reflect on this conclusion for a moment. On the surface, it is counterintuitive, since it could be restated as follows: the less the reliance on direct supervision (in favor of mutual adjustment), the narrower the manager’s span of control. The confusion, of course, lies with the term used, for here, span of control has nothing to do with “control”; it is merely an indication of the need to maintain a small face-to-face work group to encourage mutual adjustment when the work is complex and interdependent. In other words, although the restatement of the proposi- tion may be technically correct, it is misleading to use terms like “direct supervision” and “span of control.” We are better off to conclude that, because of the need for “mutual adjustment,”  “unit size” must be small.

This point suggests two lessons. First, in the area of structure (I am tempted to say management in general), things are not necessarily what they seem. We cannot rely on the pleasant conceptualizations of the arm- chair; we have to go out and research phenomena directly. Careful obser- vation produces its own share of surprises. Second, we had better choose our terms (like “control”) very carefully, and be quite sure of what we are measuring when we do empirical research.

One final point should be mentioned. Much of the evidence showing that complex interdependent tasks lead to small unit size comes from stud- ies of professional groups. But how can we reconcile this finding with that of the first proposition—namely, that professionalism (that is, standardiza- tion of skills) leads to a large unit sifce? The answer lies in interdependence: Professional work is always complex (as we define it), but it is not always interdependent. There are, in effect, two kinds of professional work— independent and interdependent—requiring two very different structural forms. In one case, the standardization of skills handles most of the inter- dependencies, so there is little need for mutual adjustment and the profes- sionals can work independently, in large units. This is the situation we find in most accounting firms and educational systems, where individual pro- fessionals serve their own clients. In the other case, interdependencies remain that cannot be handled by the standardization of skills, so there must be considerable mutual adjustment. The professionals must work cooperatively in small, informal units. This happens, for example, in re- search laboratories and think-tank consulting firms.

Thus, looking at unit size in terms of all the coordinating mechanisms helps to sweep away some of the confusion. Before we conclude this dis- cussion, however, we should mention some of the findings of other re- search—notably on tall versus flat structures, often carried out in the social psychological laboratory—because that has suggested some other factors that effect unit size. In particular, tall structures (with small units at each level, giving rise to many levels, or a “tall” hierarchy) have been shown to serve better the individual’s need for security, since a manager is always readily available, although they can frustrate the needs for autonomy and self-actualization. Indeed, top managers seem to be more satisfied in tall structures—it is they, after all, who do the controlling—whereas lower-level managers have reported themselves in some studies as happier in flat ones (with large units and few levels in the hierarchy), where they have more freedom from their own managers. Thus, both Worthy and Pfiffner and Sherwood explain the large unit sizes in Sears and the Bank of America by this factor. As the latter note about the span of control of over 600:

When officers of the bank are questioned about this seemingly unorthodox setup, their response is that they do not want to risk setting up an echelon that would take authority away from the branch managers. They want them to be self-reliant local businessmen with a maximum opportunity for making decisions on their own. (p. 161)

Studies of tall versus flat structures have also found that tall struc- tures interrupt the vertically upward flow of information more frequently, which can lead to greater distortion; and flat ones can require more discus- sion and consultation to get decisions made. Finally, studies have shown that the tall structure (or small-sized units), rather than encouraging closer supervision, may free the manager from the need to spend time on super- vision, allowing him to get on with other duties (such as making decisions and interacting with outsiders).

To conclude our general discussion, we have seen that unit size is driven up by (1) standardization of all three types, (2) similarity in the tasks performed in a given unit, (3) the employees’ needs for autonomy and self-actualization, and (4) the need to reduce distortion in the flow of information up the hierarchy; and it is driven down by (1) the need for close direct supervision, (2) the need for mutual adjustment among com- plex interdependent tasks, (3) the extent to which the manager of a unit has nonsupervisory duties to perform, and (4) the need for members of the unit to have frequent access to the manager for consultation or advice, perhaps because of security needs.

2. Unit size by part of the organization

How does unit size vary from one part of the organization to another? Generalizations are somewhat risky here, since, as we have seen, unit size is influenced by many factors. Nevertheless, some general comments are warranted.

It is in the operating core that we would expect to find the largest units, since this part of the organization tends to rely most extensively on standardization for coordination, especially standardization of work processes.

Managerial work is generally complex, so we might expect the size of units in the administrative structure to depend heavily on the interdepen- dence encountered at a given level of the hierarchy. As we saw earlier in this chapter, market grouping is often selected because it contains the work-flow interdependences within each unit (and because the process interdependences are secondary), whereas functional grouping often does not, requiring either that a higher-level manager coordinate the work flow across different units or that the managers or members of each of the units in question do so themselves through mutual adjustment. In either event, the result is the same: only a few functional units can be grouped into a higher-order unit, whereas, typically, many more market-based units can be so grouped. A great many autonomous divisions can report to one company president, as can a great many schools to one superintendent; in contrast, the president of an integrated manufacturing firm or the manager of a television station can supervise only a few interdependent functional departments. (It will be recalled that both Sears stores and Bank of America branches are market-based units.) And since organizations vary the bases for grouping used at different levels in the administrative hierarchy, we would not expect the middle line of the large organization to be uniformly tall or flat, but rather to exhibit a wavy shape, flat where grouping is based on markets, tall where it is based on function.

Earlier we noted that as we move up the hierarchy, managerial deci- sion making becomes more complex, less amenable to regulation. There- fore, holding interdependence constant, we would expect a greater need for mutual adjustment at the higher levels, with a resulting decrease in unit size. So the overall managerial hierarchy should look like a cone—albeit a wavy one—with progressively steepening sides. Thus, holding all else constant, we should expect the chief executive officer to have the nar- rowest average span of control in the organization. What may not, howev- er, remain constant is the basis for grouping. As noted earlier, the market basis is often used toward the top of the middle line. Where it is so used, and the people reporting to the chief executive themselves supervise func- tional units, we would expect his span of control to be wider than theirs.

Another factor that confounds the span of control for the managers of the middle line is their relationship with the staff units. Coordination of line and staff activities typically requires mutual adjustment—that is, flexi- ble communication outside the chain of authority. This, of course, takes a good deal of the line manager’s time, leaving less for direct supervision. So we would expect that where there is much line/staff interdependence, spans of control in the middle line should be narrower. Organizations with great proliferations of technocratic and support staff units should have rather small units in the middle line.

This leads us to an interesting conclusion about highly bureaucratic organizations, heavily dependent on technocratic staff groups to formalize the operating work: although the spans of control of the first-line super- visors should be high because of the extensive standardization in the oper- ating core, that of the managers higher up should be small because of the need for mutual adjustment with the staff members. In fact, this is exactly what comes out of the Woodward study. Mass-production firms, which she found to have bureaucratic structures, followed this pattern. In con- trast, firms in process industries, with organic structures and more exten- sive staff units, exhibited very narrow spans of control for both first-line supervisors and managers in the center of the middle line.

Finally, what about the size of the staff units themselves? How many staff members can a staff manager supervise? In those support units that do relatively unskilled work—the cafeteria and mailroom, for example— the structure would tend to be bureaucratic and the units therefore large. But what of the other units in the technostructure and support staff? The factors we discussed earlier indicate small size for most of the profes- sional-type staff units. The work within these units is complex and, being of a project nature, typically creates interdependences among the profes- sionals. In other words, these staff members are professionals of the sec- ond type discussed earlier—namely, those who must function in small interdependent units rather than as independent individuals attached to larger units. Furthermore, the managers of technocratic units must spend a good deal of their time “selling” the proposals of their units in the middle line. Likewise, the support specialists do not work in a vacuum but serve the rest of the organization, and so their managers must spend a good deal of time in liaison with it. In both cases, this reduces the number of people the staff managers can supervise, and so shrinks the average size of staff units.

To conclude, in general we would expect the operating core of the organization to assume a flat shape, the middle line to appear as a cone with progressively steepening sides, and the technostructure and more professional support units to be tall in shape. That is, in fact, the design of our logo, as a quick glance back at Figure 1-2 will illustrate.

Source: Mintzberg Henry (1992), Structure in Fives: Designing Effective Organizations, Pearson; 1st edition.

Leave a Reply

Your email address will not be published. Required fields are marked *