So far, we have emphasized that a state of routine operation in an organization is in many ways self-sustaining. Judging by the preced ing sections, an organization might be expected to encounter diffi culty in departing from its prevailing routines, but it should have no trouble in conforming to them. Although this generalization is more than half of the story and is a basic assumption of our evolutionary models, it is subject to important qualification. Just keeping an ex isting routine running smoothly can be difficult. When this is the case, the routine (in its smoothly functioning version) takes on the quality of a norm or target, and managers concern themselves with trying to deal with actual or threatened disruptions of the routine. That is, they try to keep the routine under control.
The preceding sections do suggest that there is typically going to be some difficulty encountered in deliberately creating a complex new routine where none existed before. Organization members have to learn the system of coordinating messages . They may have to add new skills to their individual repertoires, and they need to achieve a first reconciliation of their expectations regarding the distribution of costs and benefits in the situation. In such a context- for example, the initial operation of a new plant- the eventual achievement of a state of routine operation also serves as a target for managerial effort, much as it does in the context of control of an existing routine. Be cause there are important parallels between these “routine as target” situations, we discuss them together here. But there are also impor tant differences, relating to the definiteness of the target presented and the adequacy of the available information as to how it may be at tained. With regard to these dimensions of difference, there is a con tinuum of situations ranging from the edge of full routine -“getting this production line working well, like it was yesterday”- to the edge of major innovation -” opening a plant to build small com puters similar to those just introduced by our rival, only better and cheaper. II In the formal models of the following chapters, this con tinuum gets represented by distinct categories and sharp discontin uities. Here we admit that everything is a matter of degree-and examine some of the variables that distinguish the II degrees” of dif ferent cases.
An organization is not a perpetual motion machine; it is an open system that survives through some form of exchange with its envi ronment. Even its most durable machines and oldest hands undergo change with the passage of time and through the organizational process itself, and ultimately are replaced. On a much shorter time scale, current inputs of various kinds flow in, and outputs flow out. The organization’s routine, considered as an abstract “way of doing things,” is an order that can persist only if it is imposed on a contin ually changing set of specific resources. Some part of this task of imposing the routine’s order on new resources is itself handled rou tinely; another part is dealt with by ad hoc problem-solving efforts. Either the routinized or the ad hoc part of the task may fail to be ac complished if the environment does not cooperate-for example, if it fails to yield, on the usual terms, the resources that are required.
A major part of the control problem is related, directly or indi rectly, to the fact that productive inputs are heterogeneous. The firm itself creates distinctions among inputs in the course of “imposing the routine’s order” upon them; it buys a standard type of machine in the market and bolts it to the floor in a particular location in the shop, and it hires a machinist and familiarizes him with the particular capabilities and layout of its equipment and the tasks that are typi cally performed. Further differentiation occurs incidental to the input’s cumulative experience with the idiosyncratic environment of the firm; the machine suffers particular wear patterns and the ma chinist particular p atterns of frustration with his supervisor. But of course the firm also confronts the fact that different units of the “same” input may have distinctive characteristics when they are of fered to the firm for purchase, and that the entire distribution of characteristics displayed by different units offered concurrently may itself be changing over time. This prepurchase heterogeneity in the market complicates the problem of postpurchase modification, s ince the same treatment applied to different units will not necessarily pro duce the same result. Finally, because machines and workers may p ass through the market again after a stay in a firm, the modifica tions resulting from experience in firms contribute to heterogeneity in the market.
The problem posed for the firm is somehow to acquire inputs with the particular characteristics required for the smooth functioning of its routines, in the face of the fact that such inputs may not be avail able on the market at all, or, if available, may not be readily distin guishable from other inputs whose characteristics make them less ef fective or positively dangerous. Since this problem cannot be solved totally and consistently I a corollary task is to limit the damage asso ciated with imperfections in the solution to the primary problem.
The general tactics applied in dealing with these matters are much the same regardless of the class of inputs considered. A basic tactic is to select from the alternatives available from the supply side of the input market those particular inputs that are compatible with the routine. This process is complicated and imperfect if input character istics are difficult and costly to ascertain, and is further complicated by tension with the cost-control problem, arising from the fact that the range of alternatives available is affected by the price offered. There is then an effort to modify acquired inputs so that they meet the requirements of the routine- to dilute, grind, trim, or sort the raw material to a uniform standard, to teach the clerk the filing system
and the portion of the organizational dialect relevant to its use, to bolt down and adjust the new machine, or to instruct the new execu tive in the rudiments of the technology he is now managing. Of course, if too big a mistake has been made at the selecting stage, ade quate modification may be impossible. The central damage-limiting tactic is to monitor the organizational process to detect the shirking or slow worker, the embezzler, the purchased component that fails too often, the paint that does not adhere, and so forth- and, having de tected them, to reinvoke the “modify” tactic or to “select” anew from the market. Some of these problems are of course difficult to detect, particularly the ones that actively seek to avoid detection. As a last resort it may be possible to adapt the routine itself so that it either is more tolerant of heterogeneity or so that it can respond routinely to information on varying input characteristics with compensatory ad j ustments elsewhere. The latter presumes, of course, that available information permits a sorting of inputs into categories of adequate homogeneity .
The first three of these tactics are routinely pursued by various functional subunits within virtually all large organizations. The “se lecting” function described is what purchasing and personnel de partments do. Some “modifying” is also done by the personnel department and by trainers, supervisors, and co-workers, or, for non human inputs, by engineers or production workers. “Monitoring” is done by line supervisors, but is also an aspect of financial control and of quality control. However, the fact that such routinized arrange ments exist does not assure that they are comprehensive or fully effi cacious . Some input selection problems arise too infreq uently to be dealt with routinely: major purchases of durable equipment and re cruitment of high-level executives cannot be entirely routine matters themselves and may be the occasion of major discontinuities in the functioning of the organization as a whole. And if the arrays of alter-natives that input markets present to the firm change rapidly enough in adverse directions, existing routines for dealing with input heter ogeneity are likely to be overwhelmed. Then the organization will either have to adapt its routines or see them go seriously o ut of con trol. Finally, the less that is known about what input characteristics are relevant and th e more difficult it is to detect the relevant charac teristics, the more likely it is that the only symptoms of adverse change in inp ut characteristics will be inexplicable difficulties in car rying out the routine.
As the examples above indicate, the consequences of control lapses are diverse and variable. The plant may have to shut down for a few hours or days while the mess is straightened out. A bad batch may have to be thrown away. Perhaps the customers will get an infe rior product; with luck they won’t even notice, but there is the possi bility of getting hit with a big product liability suit. Or perhaps the stockholders collectively will ju st be a bit poorer, to the tune of what ever the embezzler got away with .
The sorts of consequences that are of particular interest here are those that relate to organizational memory and the long-run continu ity of routine. Control lapses may be the cause or effect of memory lapses. We have, for example, emphasized that the memories of indi-. vidual organization members are a primary repository of the opera tional knowledge of the organization. Some part of the information thus stored may be readily replaced if the particular member storing it leaves the firm; the former employee may have been the o nly one who knew how to run a particular machine, but it may be easy to hire a replacement who knows how to run it. Or it may be that the k nowledge of the employee who has departed is fully subsumed in the knowledge of his supervisor, who remains. But in some cases the memory of a single organization member may be the sole storage point of knowledge that is both idiosyncratic and of great importance to the organization. The knowledge may be tacit-say, an intuitive grasp of the priority structure of the competing demands on the employee’s time that are signaled by incoming messages. It may be articulable but not written down-the first names, marital status, and preferred recreations of the important customers in the region, or the action that is called for when a particular machine starts to vi brate too much.
The loss of an employee with such important idiosyncratic knowl edge poses a major threat to the continuity of routine- indeed, if the departure is unanticipated, continuity is necessarily broken. The new person hired to fill the role may eventually restore a semblance of the old routine, but only by picking up the knowledge more or less from scratch, guided by whatever clues his predecessor left lying about and by the indications provided by those in adj acent roles, within or outside the organization. However, those in adj acent posi tions may be taking the opportunity to attempt to redefine his organizational role in their own interest, so their advice is not fully trustworthy. For this reason, and because the new role occupant may himself be different in significant and durable ways from his prede cessor, and also as the result of other contingencies affecting the role-learning process, it is highly unlikely that a near replica of the predecessor’s role performance will result. In short, the organiza tional routine will mutate.
Mutations, of course, are not always deleterious. To put it another way, maintenance of prevailing routine is often an operational target, but it is not an ultimate objective . Modifications of routine that involve improvements in role performance are presumably wel come. However, in functioning complex systems with many highly differentiated and tightly interdependent parts, it is highly unlikely that undirected change in a single part will have beneficial effects on the system; this, of course, is the basis for the biological proposition that mutations tend to be deleterious on the average. An organiza tion member trying to do a better job can presumably accomplish something more than “undirected change,” but changes that seem like obvious improvements viewed from a particular role can easily have adverse effects elsewhere in the system. With the aid of a com prehensive understanding of the system as a whole, beneficial directed change in a part might reliably be accomplished. But since nobody in a complex organization actually has that sort of compre hensive understan’ding, it is clear a fortiori that a new employee does not have it.
It is not surprising, therefore, that the control processes of (surviving) organizations tend to resist mutations, even ones that present themselves as desirable innovations. For the particular mu tagenic event of loss of a member with a unique knowledge store, the form of the resistance obviously depends on whether the departure is anticipated or not. On the assumption that it is not, control efforts will focus on the selection of a suitably malleable successor who will at least try to respond to the routinized demands placed on the role. The efforts of the veterans to instruct the recruit in the requirements of his role will be colored by their concern to achieve a new truce at least as favorable as the old one; as a result, those efforts will tend to disabuse that successor of “na’ive” aspirations toward innovative change. ”’hen the departure is anticipated, on the other hand, the incumbent is likely to be enlisted in an effort to train one or more possible successors. How well this goes depends on, among other things, the degree to which the knowledge involved is tacit, the de-gree to which experience during the training period is representative of the full job, and-importantly- whether the incumbent really wants to succeed in imparting the knowledge to his successor.
Although the question of whether the organization can maintain continuity of routine is posed particularly clearly by the example of turnover in a key role, all organizational problems of “keeping things under control” pose that question in some degree. Time and envi ronmental changes buffet the organization with potentially mu tagenic events, against which its control systems struggle. In the long run, the most important threats to the maintenance of a successful routine may be the insidious ones, the changes that either escape the control system’s notice entirely or else are susceptible to “sympto matic relief” that leaves adverse underlying trends uncorrected. If, for example, the organization fails to maintain an adequate general level of pay relative to alternatives in the market, it may happen that the quality and motivation of its personnel gradually decline, perhaps with adverse consequences for the quality of its product or service that develop a little too slowly to be detected and linked to the p ay problem. Against the simpler and more visible problems, on the other hand, the routinized control system may be deployed so mas sively that it has the collateral effect of impeding adaptation when adaptation is actually necessary. The fact that organizations need to have routinized forms of resistance to unwanted change in routines thus becomes yet another reason why organizational behavior is so strongly channeled by prevailing routine.
The axiom of additivity is fundamental in orthodox production theory. It implies, among other things, that any feasible pattern of productive activity can be faultlessly replicated: an exact doubling of output per unit time is accomplished by an exact doubling of input. In concrete terms, the claim advanced in this proposition is captured by the image of a plant on a particular site producing a particular out put mix in a particular way; on an identical site elsewhere, an iden tical plant is constructed and produces the identical output mix in the identical way. Or, as F. H. Hahn put it, “If two identical entrepre neurs set up two identical plants, with an identical labor force, to produce an identical commodity x, then the two together will pro duce twice the amount of x that could be produced by one alone” (Hahn, 1949, p. 135).
So stated, the proposition seems to have the compelling quality of the answer to a very elementary arithmetic problem. Presumably, the posit of identical entrepreneurs is supposed to entail an identity of productive technique, and the identical plants are not just identical in themselves, but situated in identical environments . After suitable amplification of this sort, the claim may be regarded as a simple tau tology or perhaps as an assertion of the universal validity of physical law.
The question is whether the proposition says anything that is helpful in interpreting economic reality. For it to do so, the terms “identical entrepreneurs,” “identical plants,” and “identical labor force” must have empirical counterparts at least in the sense that they describe limiting cases that are often approached in real situa tions. In the context of orthodox thought, the idea that these connec tions to reality exist is supported by: (1) a habit of taking the idea of homogeneous input categories seriously, so that the “identical labor force” assumption is not blatantly contrafactual; (2) a propensity to think of individual entrepreneurs as the repositories of productive knowledge, so that positing “identical entrepreneurs” assumes iden tity of productive knowle dge; and (3) ·a tendency to regard produc tive knowledge as articulable and free of idiosyncratic elements, so that the supposition of “identical entrepreneurs” does not relate to an exceedingly remote happenstance.
In our evolutionary models, we make the same assumption that perfect replication is possible, with a similar image in mind of a sec ond plant identical to the first and employing identical routines. 10 However, our interpretation of the assumption is quite different from the orthodox one, and our commitment to it considerably less deep. A basic conceptual distinction is that we think of replication as being a costly, time-consuming process of copying an existing pattern of productive activity. Though in our modeling we abstract from the costs and make the simplest assumption about the time required, this is still a very different concept from the orthodox one, which is concerned entirely with the structure of ex ante possibilities. To put it another way, our assumption relates to what can be accomplished starting from the status quo of a functioning routine, whereas the long-run orthodox theory to which the additivity axiom relates has no notion of a status quo at all. Further, we regard the feasibility of close (let alone perfect) replication as being quite problematic- more problematic than the feasibility of continuation through time of the existing routine, which is itself no foregone conclusion, as the above discussion points out. As an initial perspective on the problem, we would not recommend the Hahn tautology, but the following ac count from Polanyi: “The attempt to analyze scientifically the estab lished industrial arts has everywhere led to similar results. Indeed, even in modern industries the indefinable knowledge is still an es sential part of technology. I have myself watched in Hungary a new, imported machine for blowing electric lamp bulbs, the exact counter part of which was operating successfully in Germany, failing for a whole year to produce a single flawless b ulb” (Polanyi, 1964, p. 52) . The point emphasized by evolutionary theory is that a firm with an established routine possesses resources on which it can draw very helpfully in the difficult task of attempting to apply that routine on a larger scale. Because the creation of productive organizations is not a matter of implementing fully explicit blueprints by purchasing homogeneous inputs on anonymous markets, a firm that is already successful in a given activity is a particularly good candidate for being successful with new capacity of the same sort. The replication assumption in evolutionary models is intended primarily to reflect the advantages that favor the going concern attempting to do more of the same, as contrasted with the difficulties that it would encounter in doing something else or that others would encounter in trying to copy its success.
To understand the nature of these advantages, it is helpful first of all to consider the similarities between replication and control, and the deeper connections to the problem of organizational memory. In replicating an existing routine, the firm seeks to impose that rou tine’s order on an entire new set of specific inputs. That task is a magnified version of one for which the firm already possesses routin ized arrangements. For example, its existing personnel and training operations have the capability to “select and modify” the sorts of employees the routine requires . By diverting these existing capabili ties at least in part to the tasks associated with·the new facility, it can avoid difficulties that would be very likely to arise if the manning of that new facility were accomplished by an equally new and inexperi enced personnel operation. The new plant will ultimately need its own personnel department (at least if “replication” is taken literally), but the new production system does not have to be hampered by the early mistakes of a new personnel department that may be learning to operate in a novel labor market environment. And a functioning p roduction system that is effective enough to detect mistakes by the new personnel department can then help that department to learn its job.
More generally, the existing routine serves as a template for the new one. The use of the template makes possible a relatively precise copying of a functioning system that is far too large and complex to be comprehended by a single person. It is not necessary for there to be a central file that contains an articulate account of how the whole thing is done. Rather, for each organizational role that is a unique storage point for important and idiosyncratic organizational knowl edge, it is necessary that the individual who will occupy that role in the new plant acquire the knowledge required for its performance. This may be accomplished by having that individual observe or be actively trained by the incumbent of that role in the old system, or by transferring the incumbent to the new system and leaving his trained successor in the old one. The collection of new role occupants thus created will make a coordinated, routinely functioning productive organization of the new facility, because the roles were coordinated in the old one- provided that the copying of the individual roles is accurate enough.
Of course, the process described will in general impose some costs in terms of the functioning of the old plant. It is unlikely that there will be enough slack resources available for training new personnel or for actually performing, temporarily, some functions in the new plant. For the replication story to make economic sense, the benefits obtained must exceed or be expected to exceed these costs. This issue is basically one of investment analysis. If the old plant is enjoying a tempora ry period of high prosperity, to be followed by normal or low profits, the opportunity costs of replication may indeed be exces sive Y The knowledge transfer must make it possible to capture a flow of rents in the new plant that lasts long enough to compensate, in present value terms, the loss of rents in the old plant. The likeli hood of this sort of pattern is obviously enhanced to the extent that a large knowledge transfer can be carried out with only small sacrifices in the old plant. Here it is relevant that the costs of a small number of anticipated d epartures or absences from key positions in the old plant are likely to be small, since such isolated gaps pose just the sort of problem that the control system routinely handles. On the other hand, the value of only a few people who know what they are doing may be enormous in providing the basic matrix of the routine in the new plant. That is, there are likely to be diminishing returns to expe rienced personnel, in terms of learning costs saved, in both plants. The transfer of a small number of experienced personnel from the old , predominantly experienced plant to the new, predominantly inexperienced one saves a lot of learning costs in the latter and incurs only small ones in the former. Finally, because of imbalances arising from indivisibilities or for other reasons, there may be some resources in the old plant that are actually idle and can be costlessly applied to the replication effort or transferred to the new plant.
There are some potential obstacles to replication that may be diffi cult to overcome even at very high cost. Some employees at the old plant may be exercising complex skills with large tacit components, acquired through years of experience in the fi rm. Others may have skills of lesser complexity and tacitness, but be very poor at teaching those skills to someone else – doing and teaching are, after all, dif ferent. Some members may for various reasons be unwilling. to co operate in the process of transferring their segment of the memory contents to someone else; they may, for example, be unwilling to disclose how easy their job really is, or the extent of the shortcuts they take in doing it. 12 Finally, personal relationships may be an im portant factor, particularly in the structure and stability of the truce that the existing routine represents . The personnel department is not likely to be up to the challenge of locating a suitably matched set of new role occupants who can be relied upon to maintain the same sort of truce . For these reasons and more, the template provided by the existing routine may not yield a good copy. There will be some muta tion of the routine as it is transferred to the new plant.
Of course, perfect replication is no more of an ultimate objective than perfect control . What matters is not that the plant be the same, but that it work with overall efficiency comparable to the old one.
If an existing routine is a success, replication of that success is likely to be desired. In particular, in the models to follow, the organization in question is a business firm for which success is roughly measured by profits, and replication of productive routines is motivated by a desire to replicate the profit flows that those routines make possible. There are symmetric questions to be addressed if the existing routine is a failure-that is, unprofitable . But while the questions are at least roughly symmetric, the answers are not. Because of their obvious importance to our models of economic selection, we digress briefly to consider them.
One important asymmetry between replication and contraction is that while the former is typically an optional response to success, the latter is typically a mandatory response to failure. As usual, the situa tion is clearest in the case of business firms, though there are analo gous problems in other sorts of organizations. If the revenues derived from the sale of the routine’s outputs fail to cover the costs of the routine’s inputs, then- barring governmental bail-outs, philan thropically inclined investors, and similarly unlikely contingen cies -it will ultimately become impossible to acquire the inputs to continue the routine on the existing scale and something will have to happen.
Under this pressure, a business firm may be expected to initiate some sort of search for a new routine that would be viable in the pre vailing environment. The analysis of this sort of search runs roughly parallel to the analysis of imitation and innovation that will concern us later in this chapter, with the proviso that the initiation of the search under conditions of adversity has implications for the quan tity and quality of the resources that may be devoted to it. But if the search is successful in the limited sense that the firm begins to at tempt to carry out a new routine, then the old routine is no longer the target and has fallen victim to the condition of adversity. The firm it self may live on, at least temporarily.
Although some sort of search response to adversi ty is probably typical, it may happen that the organization remains firmly com mitted to its existing ways of doing things-a course of action that can be rationalized as an attempt to last out a period of adversity that is perceived or hoped to be temporary. In this case, the only iisearch” that goes on is for the resources to continue to finance the existing routine. A likely occasion for such attempts to fall short is when it comes time to repl ace a large, indivisible item of durable equipment. If unable to carry out such a replacement, the firm may simply shrink and carry on roughly as before, but on a diminished scale. This sort of response (in addition to the search response) is envisaged in the formal models that follow . After a series of scale reductions of this sort, the firm and its routine may ultimately disappear entirely.
In reality, a great many factors are involved in determining the consequences of sustained adversity- for example, the degree of owner versus management control, merger opportunities, tax and bankruptcy law considerations, the liquidity or illiquidity of the firm’s assets, and the state of the firm’s balance sheet when adversity began. It is beyond the scope of our present discussion to sort out these factors and relate them to the likely persistence or change of routines. One point perhaps is worth noting here: a firm without a viable routine is a firm without a viable truce in intraorganizational conflict. That consideration I by its elLl affords abundant reason to doubt that firms behave in adversity ias if” they were under the rational control of a single actor.
As a final example of a routine serving as a targetl let us consider the case in which the target is a routine of some other firm . The interest in this sort of situation arisesl of coursel because it often happens that a firm observes that some other firm is doing things that it would like to be able to do-specifically, making more money by producing a b etter product or producing a standard product more cheaply. The envious firm then attempts to duplicate this imperfectly observed success. We will consider here only the case in which the imitatee is not cooperating with the imitation effortl and will assume that non-cooperation implies, at a minimum I that the imitator s personnel cannot directly observe what goes on in the imitatee’s plant.
What distinguishes this situation from replication is the fact that the target routine is not in any substantial sense available as a tem plate. When problems arise in the copy, it is not possible to resolve them by closer scrutiny of the original. This implies that the copy is, at bestl likely to constitute a substantial mutation of the originaC em bodying different responses to a large number of the specific chal lenges posed by the overall production problem. However, the imi tator is not directly concerned with creating a good likeness, but with achi eving an economic success-preferrablyI an economic success at least equal to that of the original. Differences of detail that are eco nomically of no great consequence are perfectly acceptable.
By this economically relevant criterion, the prospects for suc cessful imitation vary dramatically from one situation to another. At one extreme, the production in question may be a novel combination of highly standardized technological elements. If so, close scrutiny of the product its elf-“reverse engineering”- may permit the identifi cation of those elements and the nature of their combination, and this may suffice for an economically successful imitation. Indeed, even vague rumors about the nature of the product may suffice, perhaps permitting the copy to hit the market almost as soon as the original. At the other extreme, the target routine may involve so much idiosyncratic and “impacted” tacit knowledge that even suc-cessful replication is highly problematic, let alone imitation from a distance.
In the wide range of intermediate cases, the imitator’s basic tactic is to follow the example of a replicator wherever possible (and not too expensive), and to fill in the remaining gaps by independent effort. One important application of this tactic is to try to hire away from the imitatee those employees that the imitatee would reasonably want to transfer to a new plant in an attempt to replicate the existing one. Another is to obtain, by whatever means may be available, indirect clues to the nature of the target routine.
An imitator working with an extremely sparse set of clues about the details of the imitatee’s performance might as well adopt the more prestigious title of “innovator,” since most of the problem is really being solved independently. However, the knowledge that a problem has a solution does provide an incentive for persistence in efforts that might otherwise be abandoned.
Source: Nelson Richard R., Winter Sidney G. (1985), An Evolutionary Theory of Economic Change, Belknap Press: An Imprint of Harvard University Press.