What to Read Next
Just as total quality management owes much to tools like statistical process control and the “house of quality,” business process reengineering can benefit from tools to supplement and focus managerial intuition.1 Unfortunately, current tools for managing change don’t do the job.2
Effective change management depends on recognizing complements among technology, practice, and strategy. Interactions play a critical role in affecting outcomes, a role that leads to new analysis and theory.3 In developing a theory of complements, Milgrom and Roberts showed mathematically how interactions can make it impossible to successfully implement a new complex system in a fully decentralized fashion.4 Instead, managers must plan a strategy that coordinates the interactions among all the components of a business system. In particular, because information technology and new organizational paradigms eliminate time, space, and inventory buffers, operations may become more tightly coupled. These linkages further aggravate change management problems and process interactions.5
In this article, we introduce a new tool, the “matrix of change,” that can help managers anticipate the complex interrelationships surrounding change. Specifically, the tool helps manage concerns about feasibility (stability of a new system of practices), sequence (which practices to change first), location (greenfield or brown-field sites), pace (fast or slow), and stakeholder interests (sources of value added). The matrix of change was inspired by formal analyses of Milgrom and Roberts and also draws on the established design principles of Hauser and Clausing.6 The implementation steps may already be familiar to anyone acquainted with quality function deployment (QFD) or the house of quality. The resulting support for process design, analogous to product design, becomes formal and systematic but remains managerially relevant and intuitively accessible.
The Landscape of Change
An old proverb states that “you can’t cross a chasm in two steps.” The same wisdom applies to many organizational change efforts. Advances in information technology (IT) and rising competition have led to new modes of organizing work. Many of these new organizational forms are complete departures from past practice instead of incremental improvements. The resulting gains for companies can be substantial. Hallmark, for instance, discarded sequential product development in favor of cross-functional teams and reduced new-product introduction time for one greeting card by 75 percent. After reorganizing, Bell Atlantic cut service order rework and saved $1 million annually, while simultaneously improving product quality.7
Frequently, however, business process reengineering efforts run into serious difficulties. Up to 70 percent of such projects fail to reach their intended goals, and a program that seeks to become a house of quality more often becomes a house of cards.8 Because success often depends on coordinating the right technology, the right product mix, and dozens of the right strategic and structural issues all at once, near misses can leave a firm worse off than if it had never attempted the change. While several studies have documented the importance of coordinated change, managers continue to have difficulty achieving it.9 Often, the problem is not that the proposed system is unworkable but that the transition proves more difficult than people had anticipated.10 Too often, managers proceed in a hit-or-miss fashion, implementing the most visible bits and pieces of a complex new system, while missing the hidden but critical interconnections.
The path to change has several stumbling blocks. Some companies cannot adapt or miss new opportunities, leaving them vulnerable to start-ups.11 Sometimes companies acquire technology without modifying their human resource practices, mistakenly assuming “technological determinism” — that technology’s effects are independent of the organizational structure in which it is embedded. In the 1980s, for example, General Motors spent roughly $650 million on technology at one plant without updating its labor management practices. As it turned out, the technology upgrade provided no significant productivity or quality improvements.12 Similarly, Jaikumar found that U.S. companies adopting flexible technology often fail to achieve the same gains that comparable Japanese businesses do because they do not alter related operating procedures.13 Recently, Suarez et al. found that the most flexible plants in their sample of printed circuit board manufacturers were those with the right combination of human resource practices, supplier relations, and product design — not necessarily those with the most advanced technology.14 Econometric research also suggests that while IT investments can lead to high productivity, complementary organizational changes are at least as important.15
How the Matrix of Change Works
The matrix of change presents a way to capture connections between practices. It graphically displays both reinforcing and interfering organizational activities. Armed with this knowledge, a change agent can use intuitive principles to seek points of leverage and design a smoother transition. Once managers in a company chart the broad outlines of the new system and the transition path, they can once again decentralize authority for local implementation and optimization.
The matrix highlights interactions and complementary practices. Critical complements include, for example, the use of flexible machinery, short production runs, and low inventories.16 Emphasizing any one practice increases returns to its complementary practices. Likewise, deemphasizing any one of them reduces returns to its operating dependents. In our example of a medical products producer, more flexible machinery drew value from and added value to shorter production runs. Trouble starts when managers fail to identify negative feedback systems that push business units back toward old ways of doing business or when they miss synergy that would strengthen the new and better ways they wish to establish.
Ironically, the bottom-up, continuous improvement principles associated with TQM can also be counterproductive. It may be that no single isolated change can improve a process, but that a coordinated change can. Incremental change can sometimes be more painful than radical change. For example, twenty-five years ago, the Swedish government decided to shift vehicles from driving on the left side of the road to the right side. The scope of the change was enormous. When faced with dramatic change, affected parties often plead for time to adapt. But imagine the consequences of asking trucks to drive on the right-hand side during the first month of the transition and then cars in the second month. Some transitions are smoothest when all involved change their behavior quickly at the same time. Although empowering workers and decentralizing decisions are popular, these practices can also lead to trouble. What if each driver independently determined the best side of the road for driving? As it turned out, Sweden made the change quickly during the least trafficked nighttime hours.
Using the matrix of change involves four steps. First, managers determine which business practices matter most for their business objectives. Second, the matrix highlights interactions among these practices and possible transition difficulties from one set of practices to another. Third, it encourages various stakeholders to provide feedback on proposed changes. Finally, it reveals process interactions that can provide guidelines for the pace, sequence, feasibility, and location of change. Managers used these procedures successfully to reengineer a large medical products company, which we call “MacroMed.”17 We present steps from its implementation experience to illustrate the matrix of change process.
The Matrix of Change at MacroMed
In the early 1970s, MacroMed had an almost 100 percent market share for Betaplex, a sterile adhesive compound mass-produced in its North American facility. Between 1989 and 1991, however, the market share for Betaplex fell nine percentage points to about 48 percent, the fastest rate of decline in the previous sixteen years.18 Competition in the form of private-label and new Japanese products was proving more cost effective and responsive to consumer demand.19 Senior management at MacroMed became increasingly alarmed.
To make matters worse, rising materials costs exerted upward cost pressure on Betaplex, resulting in an 18 percent price hike during the same period. Although Betaplex enjoyed excellent brand-name recognition and a modest quality premium, the accelerating loss of market share was a spur for action.
MacroMed faced critical problems in its need for greater market responsiveness and modern manufacturing methods. It produced five varieties of Betaplex but had not invested in new equipment for years. Set-up times for change-overs averaged almost ninety minutes, and certain designated equipment could not switch product types at all. When demand for certain products was low and other products moved briskly, facilities utilization was very poor. MacroMed’s union contract also enforced rigid, narrowly defined job categories, contributing to inflexibility.
In response, senior managers decided to design new manufacturing equipment and to stop using the relatively inflexible equipment available on the open market. They understood too that simply changing the technology without also rethinking their work organization, market strategy, supplier relations, and other aspects of their business would not lead to success. Accordingly, they wrote an explicit vision statement that outlined new policies and procedures in each area. In realizing process interdependence, they were already ahead of many other firms.
Unfortunately, however, MacroMed’s early experience with the new system was not good. Despite a considerable investment in new capital and explicit calls for new approaches to work, productivity did not significantly improve and, by some measures, actually worsened. Clearly, workers were not using the new equipment at its potential; moreover, there was grumbling about poor management and leadership.
In an effort to coax more efficiency from the equipment, MacroMed’s managers put significant effort into formal modeling of equipment change-over times, capacity requirements, and optimal queuing strategies. Factory visits, however, revealed that the real problem had more to do with an intrinsically difficult organizational transition than suboptimal machine scheduling or the actions of particular individuals. Despite instructions to the contrary, workers continued to use new equipment much as they had used the old, thus wasting its flexibility. Although MacroMed had eliminated piece-rate incentives, workers let large work-in-process and finished-goods inventories build up rather than allow more downtime. Their mental models still led them to keep the machines running at maximum capacity with minimal change-overs. Similarly, line managers were reluctant to cede real authority to the operators. While they spoke of teamwork, empowering the workforce, and maintaining open, trusting communication, some managers suggested privately that operators did not really understand what was happening. There were also mismatches in the skill sets of some operators, who lacked any desire to assume decision-making responsibility, just as there were mismatches in contracts with suppliers and numerous other aspects of the work.
These complex interactions became apparent with hindsight, but most were not explicitly considered in advance. Furthermore, it was unclear how to correct the problems, given the significant investments that the company had already made and the loss of forward momentum that these difficulties were causing. We developed the matrix of change to help organize and sort through these issues.20 The matrix’s development involved academic researchers, senior managers, and operators from the shop floor.
Building the Matrix
The matrix of change system consists of three matrices and a set of stakeholder evaluations. The matrices represent (1) the current organizational practices, (2) the target practices, and (3) a transitional state that bridges these two. The stakeholder evaluations give people in the firm an opportunity to state the importance of the practices to their jobs. Matrix construction proceeds in four steps.
Step 1: Identify Critical Processes
Managers should first list their existing goals, business practices, and ways of creating value for consumers and then break current practices into constituent processes, i.e., into statements of how the work is accomplished. A process is “a structured, measured set of activities designed to produce a specified output, . . . a specific ordering of work activities across time and place, with a beginning, an end, and clearly identified inputs and outputs.”21 A second list describes new or target practices.
Identifying the most important processes can be quite difficult, but certain guidelines can help. In many change management projects, the most important success criterion is “to start with the end in mind,” that is, identify the purpose or business objective of change, whether it is organizational learning, market share, flexibility, customer satisfaction, or something else. Since MacroMed already enjoyed high quality and brand-name recognition, senior managers settled on decreased costs and increased flexibility as their principal goals — improvements that would permit MacroMed to lower retail prices and to pursue niche market margins.
Another guideline is to choose redesign team members for both their knowledge of functions essential to business objectives and their subsequent ability to secure support from these functions. One organizational change effort, for example, sought to cut ninety days from a corporate supply chain.22 The change effort involved only order fulfillment staff, yet close examination revealed that total cycle time consumed seventy-five days of manufacturing lead time, eighty-five days of customer acceptance lead time, and twenty-two days of order fulfillment time. If the design team had eliminated 100 percent of the order fulfillment time, it still would have fallen 76 percent short of its ninety-day goal.
At MacroMed, senior managers assembled a SWAT team from a cross-section of the workforce consisting of managers, design engineers, and union workers across several different functions. The team began by enumerating specific aspects of its existing hierarchical production techniques and forming its vision of a new organization based on the perceived benefits of a flatter, more flexible production line. Then, from general statements of practice, it defined subtasks or constituent practices (see Table 1 for an example).
The steps of any practice can be broken down further and further, if this is helpful. The existing practice of “designated equipment,” for example, can be further broken down to “sterilizing,” “manufacturing,” and “packaging.” To keep explanations simple, we use only two levels in this article. Practices can also be grouped into categories by function (e.g., marketing, human resources, and manufacturing) as well as by strategic initiative (e.g., elimination of non-value-adding costs and speed). MacroMed preferred the second classification, as Table 1 shows.
Step 2: Identify System Interactions
After describing existing practices, the team created a horizontal triangular matrix to identify complementary and competing practices (see Figure 1). Complementary practices reinforce one another, whereas competing practices work at cross-purposes. Doing more of one complement increases returns to the other. For example, in the existing system, narrow job functions made tasks easy to specify and increased MacroMed’s ability to offer piece-rate pay tied to hourly output. These practices were reinforcing. On the other hand, doing less of a competing practice increases returns to the other. A flatter managerial hierarchy, for example, would shift some strategic decisions to workers; this, in turn, would decrease MacroMed’s ability to offer piece-rate pay tied to hourly output. These new and old practices were interfering. Most existing frameworks do not capture interdependencies or practice interference, while interference matrices make these interactions explicit.23
A grid connects each process in an interference matrix. At the junction of each grid, plus signs (+) designate complementary practices, and minus signs (–), competing processes. Thus, since “designated equipment” complements “narrow job functions,” their intersection on the grid is assigned a plus sign. Reading along the same diagonal, “designated equipment” also complements “large inventories” and “piece-rate pay.” A plus sign does not indicate that an interaction is “good,” only that it is reinforcing. In the absence of evidence to support either reinforcement or interference, the space at the junction is left blank. The horizontal triangular matrix for a subset of MacroMed’s existing practices appears in the left half of Figure 1.24
An analogous process develops a vertical triangular matrix for target practices. In the horizontal matrix, no competing practices were found; this system is coherent as a stable unit. In contrast, the vertical triangular matrix has two competing practices. “Line rationalization,” which reduces product variety, works in opposition to “flexible equipment,” which encourages greater variety.
The plus or minus values for each cell are derived in a number of ways. Often, once the practices are classified, the values become self-evident. In other cases, formal models and theory provide guidance. Theories of ownership, for example, suggest that decentralizing data management can boost quality levels in systems that users control themselves,25 and operations management models suggest that processing tasks in parallel adds more value when inputs have higher variance.26 In some cases, empirical data suggest the existence of complementary or substitution effects, which formal statistical analysis can identify.27 Surveying key personnel is also an effective way to gain insight into both perceived and real interactions. MacroMed used all these approaches.
Step 3: Identify Transition Interactions
Next, the team constructed the transition matrix, a square matrix combining the horizontal and vertical matrices that helps determine the degree of difficulty in shifting from existing to target practices. The transition matrix shows the interactions involved in moving from existing to target practices. In contrast, simply using a clean slate tells a team nothing about the difficulty of a transition.28 Using a “blank sheet of paper” for design can also require a “blank check” for implementation.29
A subset of the transition matrix that MacroMed used illustrates important interactions between existing and target practices; most of them are opposing (see Figure 2). “Narrow job functions” interferes with “flexible equipment,” and “several management layers” interferes with “greater responsibility.”
Certain practices complement one another. “Line rationalization” complements “designated equipment” by reducing uncertainty about scheduling. Similarly, “line rationalization” complements “narrow job functions.”
Step 4: Survey Stakeholders
Now the team needed to determine how various stakeholders felt about retaining existing practices and implementing target practices. Just as listening to the “voice of the customer” is essential to building a better product, listening to the “voice of the stakeholder” is essential to building a better process. At MacroMed, several different groups had an opportunity to indicate how important each new practice was to their job performance. Each surveyed employee used a simple five-point Likert scale anchored at zero. A value of “+2” meant that a practice was very important, and a value of “+1” that a practice was important but not essential, while a value of “–2” indicated a strong desire to change or reject business as usual. A value of “0,” which can be omitted, represented indifference. (Figure 3 shows a brief, completed example.)
The respondent in this example felt strongly that, among existing practices, “narrow job functions” should be discontinued (–2), inventory levels should not be as large (–1), and “piece-rate pay” is somewhat important (+1).30 Regarding target practices, the respondent felt positively about most practices and indifferent about one.
Although these examples use a relative Likert scale, several variations are possible. As shown, they measure internal business value from the perspective of a single stakeholder. A firm might also wish to consider a “balanced scorecard” involving other stakeholders and perspectives, including financial indicators, customer preferences, and innovation requirements.31 Thus a company might evaluate the axis for flexible equipment from the additional perspectives of improving customer product offerings and of reducing financial costs. If it requires multiple indicators, it can add more columns. Ideally, a given metric will have quantifiable units such as accounting profits or the number of product configurations offered to the customer. If multiple measures are used, comparisons across practices must use the same units, such as dollars or soft dollar estimates. Combining Figures 1 through 3 creates the matrix of change (see Figure 4).
Counting cross-connections is one way to measure coupling strength or interdependence within blocks of practices. A block is a cluster of reinforcing practices. For example, within the existing practices of Figure 4, “designated equipment,” “narrow job functions,” and “piece-rate pay” represent a complementary block. Counting along the upper and lower diagonals of the horizontal matrix, each has three complementary practices. Similarly, “large WIP” has two complements: along the upper diagonal, it reinforces “designated equipment” and along the lower diagonal, it reinforces “piece-rate pay.” Finally, “several management layers” has only one complement. In this sense, “several management layers” is the least tightly coupled practice within this block. In contrast, the target state for this example has only small blocks in the vertical triangular matrix that are independent and easily separable.32 The large block of existing practices that involve “designated equipment” in Figure 4 illustrates several principles that we discuss next.
Interpreting and Using the Matrix
The matrix of change is useful for addressing the following types of questions:
- Feasibility. Does the target set of practices constitute a coherent, stable system? Are the current practices coherent and stable? Is the transition likely to be difficult?
- Sequence of execution. Where should change begin? How does the sequence of change affect success? Are there reasonable stopping points?
- Location. Are we better off instituting the new system in a greenfield site? Or can we reorganize the existing location at a reasonable cost?
- Pace and nature of change. Should the change be slow or fast, incremental or radical? Which blocks of practices, if any, must be changed at the same time?
- Stakeholder evaluations. Have we considered the insights from all stakeholders? Have we overlooked any important practices or interactions? What are the greatest sources of value?
Each major area in the matrix of change serves various roles and addresses different aspects of these five issues. Taken together, they offer useful guidelines on where, when, and how fast to implement change. (Figure 5 indicates the purpose of the various features.) Interpreting the information captured in the matrix of change motivates the principles that follow.
Feasibility: Coherence and Stability
The sign, strength, and density of interactions are important for determining practice coherence and stability. A system of practices with numerous reinforcing relationships is coherent and therefore inherently stable, whereas one with numerous competing relationships is inherently unstable. The existing system at MacroMed was, in fact, quite stable. This was hardly surprising because the system had been in place for decades, and practices had co-evolved. Fine-tuning a traditional approach over a period of years tends to eliminate conflicting practices.
The desired state (shown in Figure 4) is also fairly stable but has a single competing relationship.33 This implies that it may require more effort to keep the parts working together. The company may also need to evolve new, noncompeting processes or to propose alternatives that are at least neutral. If a proposed state has too many negative relationships, the project will be unstable and must be reevaluated. If a target state has few relationships, whether reinforcing or interfering, it will be neither likely to collapse nor tightly bound together. Thus the tight coupling of the existing system indicated that it was more inherently stable than the target system. Tight or loose coupling also predicts the level of coordination necessary to effect change. Loosely coupled practices require less coordination.
Critically, the transitional state for MacroMed was dominated by interfering relationships, indicating a high degree of instability. This offers a fundamental explanation for the difficulty often found in business process reengineering: when faced with new practices that conflict with current operations, well-intentioned local managers seeking to optimize their piece of the system may consciously or unconsciously undermine change by pushing the system back toward its initially stable state. From a local perspective, an on-site manager may apply old rules precisely when the system of work shifts and the existing regime starts to break down. Resistance appears sensible and even efficient, but, from a global perspective, structural change becomes almost impossible.
Sequence of Execution: Where to Start and When to Stop
The most easily eliminated practices are those that oppose other existing practices. Unfortunately, eliminating them first can also be dangerous because it may render the remaining system even more entrenched and difficult to change. Since stable systems generally have few opposing practices, another alternative is to start removing practices that have no interactions with other practices. On the target side, the easiest new practices to implement are those that complement existing ways of doing business. This can build a bridge from one system to the next, particularly when a practice has numerous complements in the new state. A company should avoid it, however, if new practices strengthen old habits in ways that make dismantling the old regime even harder.
A company must handle practices that support a large number of other practices with great care. It can insert such “linchpin” practices to help lock several new practices in place or remove them to unlock several old practices. At MacroMed, the use of designated equipment acts as a linchpin practice; it has numerous reinforcing interactions, as Figure 4 illustrates.
Designated equipment — inflexible, high-volume machinery — is one linchpin practice that facilitates narrow job functions and pay schedules that are tied to the amounts produced. Removing inflexible equipment aids the simultaneous removal of the entire block in the horizontal triangular matrix. In the ideal case, completely independent blocks may be identified and removed separately, but in this case, the block’s components have less effect on the number of management layers, which might be changed later.
Therefore, as long as the old designated equipment remains in place, the company will have more difficulty expanding job responsibilities, lowering inventory levels, and removing piece-rate pay. For similar reasons, introducing new technology is often used intentionally as a catalyst to facilitate change management. Installing new equipment can signal an irreversible commitment to a new way of doing business and can initiate a cascade of complementary changes in work practices, as workers are forced to adapt. At MacroMed, one manager described the dramatic unveiling of the new technology:
“In phase two, we took down the walls that had surrounded the new equipment and assembled the new machines right on the manufacturing floor in their final location. The workers saw the new technology growing right around them. Because of this, people knew it was real and didn’t want to be left out.”
Although the new technology helped achieve buy-in from the workforce, it was not enough to overcome the ingrained routines of the factory without a lot of additional change management. Ironically, the very flexibility of the new technologies made it too easy to continue with the comfortable old routines. When flexible technology meets an inflexible work-force, often the machines, not the people, are forced to adapt.34
The larger the blocks of reinforcing practices, the more difficult they are to change. The hardest changes involve the installation of new practices that oppose the greatest number of existing practices. In fact, large new blocks may be impossible to install before the opposing practices are removed. One strategy is to dismantle the competing practices beforehand. Another alternative is to lay a foundation of complementary new practices before making the attempted change. Having support in place helps keep employees from reverting to old habits.
The presence of large blocks also suggests that change should pause or stop only after the entire block has been removed. Reducing the pressure to change when an interlocking block is only partially dislodged can allow old practices to roll back into place, thus undoing work and wasting resources. Also, the complete sets of existing and target practices never need to occur at the same time. Managers can use the transition matrix to choose which blocks of practices to move, and in which order, to avoid the worst interactions.
Location: Greenfield and Brownfield Sites
The greater the number of interfering relationships in the transition matrix, the more disruptive the proposed changes will be. Increasing interference indicates a greater need for isolation. Sometimes a fledgling change project must be shielded from bad habits. Natural tendencies toward local optimization will push the system toward an initially stable state as long as opposing practices remain. More disruptive changes make existing or brownfield sites less attractive. In fact, new or greenfield sites are much more popular for introducing new systems, even when they require abandoning years of organizational learning.
Choosing a greenfield site is an issue not just of location but also of attitude. Radical change is “frame breaking” in the sense that it requires changes in mental models of operation.35 Mental models involve goals and values, system boundaries, causal structure, and relevant time horizons.36 A transition matrix with more densely interfering relationships can therefore indicate a greater need for changing mental models. For particularly radical or frame-breaking change, a company may need an outside change agent to help people see processes differently. The company may also need to replace managers because they are too closely tied to former ways of doing business. Replacing senior managers can often lead to more rapid change, while keeping them can sometimes smooth the transition.37 And, if any group is made particularly worse off by the change — in influence, responsibilities, and so on — the company should address this issue early on because members will tend to revert to their former roles.38
Pace and Nature of Change: Fast or Slow, Incremental or Radical
For purposes of implementation planning, it is worth distinguishing between the pace (gradual or rapid) and the nature (incremental or radical) of change.39 Occasionally, radical change may best be spread over several episodic steps, especially if resources are locked in place and initial conditions resist change. A single-step discontinuity may prove too disruptive, too expensive, or too confusing. Yet, as the chasm-crossing proverb suggests, there are other occasions when change is an all-or-nothing proposition. A halfway solution may lead to wasted resources, organizational exposure, or even failure.
Three factors help to determine the appropriate pace: task interdependence, organizational receptiveness to change, and external pressure. The first, task interdependence, concerns how modular and how serial the essential steps are. It measures the divisibility of organizational processes.40 Grouping tasks into blocks reduces the scope of change and the coordination problem that must be managed at any given instant. The pace of change within blocks must be rapid; the pace of change between blocks may be slow. Thus the speed of removing parallel components of an interdependent block may be more important than the serial speed of the whole change process. At MacroMed, the existing block of practices associated with designated equipment is interdependent, whereas the target block associated with low JIT inventory is independent (see Figure 4). The transition matrix, by showing interference, also suggests how radical a change must be.
The second factor, an organization’s culture, determines its receptiveness to change. In a large chemical products company, the information systems group was accustomed to experimentation and risk taking, which greatly facilitated an episodic approach.41 The advantage of a supportive culture and episodic change is that they permit phased adaptation to unfamiliar practices. Particularly if change needs to migrate through several parts of an organization, episodic change can promote experimentation and learning so that late adopters can access the know-how and know-why of the early adopters without repeating their mistakes.42 Experimentation, however, is unlikely if the culture punishes failed experiments. At MacroMed, the culture was not receptive initially to the kind of change that managers sought, but this resistance had some advantages. As one senior manager noted after the first wave of change:
“The fact that the first effort took place in one of [MacroMed]’s oldest unionized plants made the challenges surrounding the change effort all the greater; however, that also made the success all the more marketable in [MacroMed]’s other locations.”
External pressure is the third factor. Low pressure provides slack time for adaptation, but a hostile environment may preclude episodic change. With extreme external pressure, concern for survival and the absence of slack resources may force a rapid pace, which, in turn, interacts with the organization’s culture. If there is a history of opposition to change or a pattern of unsustained or regressive change, then transition times should be minimized. As Gallivan et al. note: “Under these conditions, managers’ intentions for rapid implementation would seem appropriate given that the opportunity to change anything later may be lost as enthusiasm wanes, skepticism grows, resistance accumulates, resources are reallocated, and champions are reassigned.”43
Stakeholder Evaluations: Strategic Coherence and Value Added
Stakeholder evaluations make preferences and expectations explicit. Evaluations help managers anticipate responses to change by providing data on sources of support for, indifference to, or hostility toward proposed changes. If employees give an existing practice low marks, they are likely to support a change. Conversely, if they do not support a change, they will likely give an existing practice high marks. They may require new incentives to support new proposals.
Whereas the transition matrix measures the alignment of practices, the evaluations measure the alignment of incentives. Negative values in the target ratings section indicate a need to cooperate and better align incentives, to increase the pace and avoid drawing out resistance, or to isolate factions whose interests oppose the change initiative.
Highly variable stakeholder evaluations indicate different priorities and a fragmented strategic vision. If evaluations were uniform across employee populations, then stakeholders within the company would jointly focus on tackling the most important issues first. With different priorities, however, stakeholders will tend to work at cross-purposes during implementation. When these differences occur, senior management may wish to establish a more uniform strategic vision early in the change process.
Stakeholder evaluations reveal an organization’s receptiveness to change. In their case study of a large chemical company, Gallivan et al. found that a tradition of open experimentation, a willingness to invest in technology without immediate payoff, and a philosophy of empowerment and learning all created norms that facilitated change.44 These factors influence the stakeholders’ willingness to cope with, participate in, and accept responsibility for change.
The very act of asking workers for their values —and taking them seriously — can positively affect the change process by giving employees a sense of ownership and responsibility. At MacroMed, the workers’ attitudes changed noticeably. According to a leader of the change management team:
“They played the role of ‘final customer.’ They decided where engineering and operations resources should be focused. They also made supplier decisions and traveled together to supplier sites. . . . There is true measurable value in soliciting and developing ownership at the worker level, at the early stages of change.”
Determining the Net Value Added
Once a company has addressed key differences in stakeholder evaluations, a simple mechanism can indicate which changes ultimately add the most value. The formula target value – existing value gives an approximate net value that will be gained by changing practices. This assumes that all units are the same and that practices with no counterpart are paired with a value of zero. For example, “line rationalization,” a target practice with no existing counterpart, has a net value of 1 – 0 = 1. To visualize the net value added, sort practices in a “tornado plot,” which connects importance ratings across categories so that the spread is monotonically decreasing (see Figure 6). Net values can also be negative, as in the case of a stakeholder who feels he or she can earn more through piece-rate than flat pay.
Net value added provides a useful complement to the matrix of change but can be misleading if used in isolation. Principles of net value suggest which changes are important, but principles of coherence suggest which sequence to adopt. Many consulting companies pursue a strategy of trying to get the big payoff items first, but this can be counterproductive45; the matrix shows why. A “greedy” algorithm, which sorts changes based solely on the best value, will miss the possible cost reductions brought about by setting up complements.46 For example, the net values in Figure 6 appear to suggest first giving workers more responsibility, then reducing inventories, and so on, proceeding through the list in a sequence that gains the next best value at each step. This process might stop before implementing the last step, which appears to add negative value. The matrices in Figure 4, however, show that some practices are reinforcing. For instance, at MacroMed, narrowly defined job responsibilities complement the use of designated equipment — the least-cost path involves changing them together rather than two steps apart. Although net value methods suggest changing to just-in-time inventories before moving to flexible equipment, this could be counterproductive and lead to stockouts or increased worker frustration and resistance.
Although cutting layers of management may not create as much value as improving the product offering through flexible equipment, multiple layers of management may complement narrow job descriptions (see Figure 4). It may not be possible to have workers assume greater responsibility when they face oversight at multiple levels. In this case, moving to flatter management constitutes a “stepping stone” —a practice that smoothes the adoption of other practices.47
The sequence of changing practices affects not only how soon a company may realize any given payoff but also the cost and feasibility of changing other practices. Thus there are “alpha” and “beta” benefits to the order of adoption. Alpha benefits represent immediate returns, while beta benefits represent subsequent gains achieved through “setting up complementarities in the adoption of future [practices].”48 Beta benefits also accrue from “learning by doing other things.” In the process of learning to operate with fewer layers of management, an organization may also learn to distribute responsibility.
The greatest benefit from the matrix of change may be that it forces management to make explicit the practices and interactions implicit in the current, target, and transition systems. Recognizing and defining the nature of the problem can be 80 percent of the battle, but without a tool for clearly sorting out interactions among practices, much of change management is relegated to intuition and politics. Once a company identifies the elements of the matrix of change, the most effective strategy may become self-evident.
The Problem of Prediction in Complex Systems
Of course, there will be unforeseen contingencies. No matter how much effort managers invest in the matrix at the outset, companies have myriad, unarticulated rules, procedures, technologies, and cultural mores that they can never completely catalogue or even recognize. The most detailed list will inevitably overlook certain unstated assumptions inherent in any system of work, and managers will sometimes need to improvise.49 Despite the data the matrix provides, the result of any elicitation or mapping process is “a set of causal attributions, initial hypotheses about the structure of the system which must then be tested.”50 The matrix of change helps managers identify important assumptions implicit in their work organization, but they must remember that key components of any system may remain unmodeled, allowing unexpected barriers to surface in the midst of the change process.
The matrix of change can offer two forms of assistance, if not complete assurance, in dealing with complex systems. The first is that a company can revisit the matrix design process as often as necessary. Each design phase can represent a time slice or window onto current and possible organizational configurations (see Figure 7).
Matrix chaining can highlight important stepping stones or phases that a company can skip entirely. In a transition from craft production to mass production to modern manufacturing, for example, a company might omit the intermediate step or use it as a bridge, depending on the ease of the associated transitions. Refining the matrix for more levels of detail or for more time periods can help managers develop organizational metaprinciples, such as the observation that a pair of process dependencies frequently recur. This pair might then be merged or replaced wholesale. Having identified certain dependencies and interactions, systematic process behaviors might be amenable to software simulation to aid in prediction. For example, a feedback model of TQM practices at Analog Devices, reflecting actual events, indicated that significant quality and productivity improvements did not translate into higher financial performance.51 The software offered controlled, repeatable simulations that helped identify a confluence of interacting business practices and environmental factors causing the trouble. The matrix can provide important inputs to such a system.
A second form of assistance is in reshaping mental models. At MacroMed, workers had the unarticulated goal of running the machines at all times to increase productivity, which led them to resist product line changes and inadvertently defeat the value of flexible equipment. When they learned of this belief, MacroMed managers devised compatible incentives and achieved a more coherent work system. To the extent that the matrix can help shift mental models even partly from implicit to explicit parameters, it can improve chances for success. It may never be possible to “manage the magic” of a perfectly functioning system, but managers need simple ways to initiate debate on critical changes. The matrix helps initiate that inquiry, helps identify multiple interactions, and helps uncover at least some of the hidden assumptions.
Lessons Learned at MacroMed
At MacroMed, we administered questionnaires based on the matrix of change to multiple groups within the company. We included managers, engineers, and hourly employees in both the company at large and in a special pilot project designed to test the proposed changes.
Employee input was highly informative. For instance, managers within the pilot group saw positive reinforcement between flexible machinery and line rationalization, an interaction that is inconsistent with the literature. Line rationalization, a top-down optimization process, tends to reduce the scope for on-the-spot decision making and reconfiguration of flexible machinery.
Matrix data also showed a certain degree of ambivalence on the part of workers toward management. Most workers were not receptive to having managers work side-by-side with them on the line. Rather than viewing managers as partners, they felt this practice would discourage them from contributing new ideas or expanding their roles. At the same time, a surprisingly large subset of workers expressed no desire to become empowered with responsibility for programming the equipment. They preferred their traditional roles, which required little on-the-job thinking, allowing them to daydream and chat with coworkers while doing the parts of the job that required physical work.
Several assumptions about the transition surfaced during data gathering, but others did not surface until the new system was implemented. For instance, although the company eliminated piece-rate pay and established an explicit goal of reducing WIP inventory, most workers continued to behave as if the paramount performance indicator was eliminating machine down-time. As a result, they avoided change-overs and kept the flexible machines running on the same product line almost as much as they had with designated equipment. Although this no longer increased the profitability of the factory or their individual pay, this and many other heuristics were too ingrained to overturn easily.
Overall, the high density of interfering relationships in the transition matrix highlighted the need for a greenfield site. Since a completely new site was too expensive, MacroMed chose a modified green-field approach. It isolated one portion of the factory with a temporary new wall, then carefully selected workers and managers for the “SWAT” team. Implementation proceeded in two phases. During the first phase, the SWAT team debugged the most difficult technologies and procedures. It reorganized the line and reduced obsolete finished-goods inventories to zero. During subsequent phases, the SWAT team helped disseminate flexible manufacturing practices throughout the rest of the factory and to other plants. Team members served as trainers and troubleshooters for the second phase of implementation. One MacroMed manager commented:
“All members of the SWAT team, both union and salaried, shared equally in the responsibility. . . . Because they were given as much responsibility as the salaried employees, the union workers on the team cared greatly about the outcome of the project, and they were a positive influence on bringing the other union workers in the plant over to the new way of doing things.”
Paralleling the episodic introduction of radical change, this modified greenfield approach greatly simplified MacroMed’s transition.52 Workers reported that job security was critical, indicating that a sudden transition from a piece-rate incentive scheme to different incentives based on meeting new goals — such as contributing new ideas, lowering inventory, and accepting greater responsibility — could be too disruptive.
MacroMed chose to offer flat-rate compensation initially, while workers adjusted to their new roles and began to more fully understand the risks, and management gained a better understanding of the behaviors it wished to promote. Wage guarantees thus lowered worker resistance to the new changes. As the head manager of the SWAT team observed, “Equipment issues are easy; the people issues are the tough part.” Other factors that eased the transition included the use of contract employees and hand-selected union workers receptive to change.
In applying the matrix of change, MacroMed also discovered conflicts in different employees’ machine set-up procedures. This revealed a way to reorganize process change-overs and resulted in a 67 percent reduction in set-up times as well as a dramatic reduction in variability. Other successes included a fourfold increase in throughput and a cut in waste costs of 65 percent. Purely through attrition and early retirement, MacroMed reduced the number of line employees by 33 percent and staff employees by almost 40 percent. MacroMed stopped its decline in market share and improved flexibility and response time to the point where products could move from concept to store shelf in just ninety-nine days, an impossible time frame before restructuring. The changes were so successful that management ordered the windows in that part of the factory devoted to the new approach painted black to prevent visitors and competitors from seeing the organizational and technical innovations.
The findings from our study of MacroMed and similar companies underscore that successful change depends on leveraging complementary practices and on redesigning contingent business processes. Managing and coordinating increasingly complex systems, however, requires increasingly sophisticated tools.
Flexibility in manufacturing relies not only on powerful new information technologies, as is commonly emphasized, but also on mutually reinforcing practices. Cross-training, incentives, inventory policies, decision-making structures, and open-door communication, among other practices, must function as a coherent, stable system. Not only is it difficult to isolate a single practice and graft it onto another work organization to achieve the same effect, but many subtle interactions also go unnoticed until it is too late. Juxtaposed against this argument, the high failure rate reported for business process reengineering seems less mysterious: without proper tools, most BPR efforts are unlikely to have accounted for the complexities, the finely balanced complements, and the time delays of a stable, coherent system.
By systematizing change management, the matrix of change can help select those practices most likely to contribute to business goals. It detects complementary and interfering practices and presents an overview of an interlocking organizational system. It helps capture the alignment of incentives, showing which practices are the greatest sources of value to stakeholders. Then, after identifying interactions, it suggests guidelines for judging a proposed system’s feasibility and coherence, its sequence of execution, and its relative pace of change. By focusing on the difficulty of a transition, the matrix of change also suggests how disruptive or radical the change may be and thus gives an indication of the need for a green-field location. From this overview, management can learn where it has the greatest opportunity to implement change and which changes are most important.
Each element of the matrix of change proceeds from the fairly intuitive concepts of reinforcement and interference. MacroMed used these steps to develop a deeper understanding both of how its existing practices reinforced one another and of how practices it wished to introduce meshed with what it had done in the past. Since the list of practices can be disaggregated to an arbitrary level, an organization can apply the matrix to its whole structure, departments, and shop floor.
It is also possible to proceed in the other direction and consider aggregation through the entire value chain, including suppliers, inbound logistics, outbound logistics, buyers, and even competitors. From this perspective, a company can use the matrix as a measure of environmental fit, answering questions about how well current or proposed practices work with or against the environment. If environmental factors oppose one another in a triangular matrix, they indicate instabilities that might provide new opportunities. Unstable environments require more flexible —possibly networked — organizations with a premium on innovation. Emerging markets often fit this profile. If environmental factors reinforce one another, they indicate stable systems that are unlikely to change or that might change all at once; a change in regulation, for example, can shock the system. A stable environment requires a structured organization with a premium on efficiency. Commodity markets, for example, tend toward stable, cost-based competition.53
Since the change process is likely to unfold over time, a company can revisit the matrix as necessary to gauge progress. It can capture new relationships as managers develop a deeper understanding of their situation. Managing change without regard to context or interconnections misses what is most important. The true value of the matrix is to optimize steps not just in isolation but as parts of an integrated system with a more cohesive fit.
In adopting change, businesses typically look for cost savings, a first-order effect. Savings then free up resources that can be invested into other areas — a second-order or substitution effect that also needs to influence business decisions. Third-order effects, however, are those most often missed. They represent whole new structures or systems for organizing work.54 Flexible machinery adds little value in environments accustomed to rigidly hierarchical procedures, large inventories, and little autonomy. But in conjunction with shorter production runs and just-in-time deliveries, its effects on competitors can be devastating. Financial analysis alone frequently does not uncover third-order effects because it can overlook complements in strategies and structures and unanticipated interference from incompatible practices. The matrix of change can identify complementary structures and give change agents an intuitively appealing tool for managing them.
This appendix explores one important business process reengineering effort — the transition from a hierarchical to a network organization. Hierarchies tend to be vertically integrated, mass production organizations that seek scale economies through long production runs of commodity products and that reduce risk by owning the assets they use.a Network organizations, in contrast, tend to be partnerships that exploit strategic opportunities by rapidly and flexibly adjusting their outputs to niche markets (seeking scope economies through complementary, possibly intangible assets) and that reduce risk through equity arrangements, repeated cooperation, and trust.b Members exercise joint control over assets rather than taking direction from an executive body. Hierarchical and network practices may be internally consistent but juxtaposed against one another; they typically compete.
Milgrom and Roberts discuss the differences between lean manufacturing and mass production.c In a survey of attributes, Alstyne also provides a direct comparison of hierarchies and networks. (Table A lists elements from these two sources and illustrates the strong differences between these methods for organizing work.)
Internally, these systems appear to be coherent. Long mass production runs accompany large inventories, just as highly customized products accompany flexible machinery. It is unlikely, however, that mixing elements of these independently coherent systems will result in another coherent system. In fact, placing several organizational attributes from Table A into the matrix framework illustrates the difficulty of a potential business process reengineering effort (see Figure A). While the endpoints are stable and cohesive, the transition from a hierarchical to a network organization appears unstable. Given the number of competing practices in the transition region, it is not surprising that, for this type of change, reengineering projects that implement only a handful of features have difficulty reaching their goals.
Although advanced information technology is typically associated with modern manufacturing more than with traditional mass production, it can complement practices in both systems. In the network organization, if IT is used for coordination and decision support, it can complement cross-functional teams and flatter management, as in the vertical interference matrix. By providing everyone with the same data, however, it can also undermine authority. In the hierarchical organization, if IT is used for monitoring and automation, it can complement narrow job descriptions and rank-based authority, as in the transition matrix. These varied effects of IT have been observed in the literature. If an organizational feature has multiple attributes, a helpful rule is to split it into discrete practices, such as disaggregating IT into monitoring, decision support, and automation.
1. J. Hauser and D. Clausing, “The House of Quality,” Harvard Business Review, volume 66, May–June 1988, pp. 63–73.
2. T. Davenport and D. Stoddard, “Reengineering: Business Change of Mythic Proportions?” MIS Quarterly, volume 18, June 1994, pp. 121–127.
3. A. Barua, S.H.S. Lee, and A.B. Whinston, “The Calculus of Reengineering” (Austin, Texas: University of Texas, Department of Management Science, 1995).
4. P. Milgrom and J. Roberts, “Complementarities and Fit: Strategy, Structure, and Organizational Change in Manufacturing” Journal of Accounting and Economics, volume 19, 1993, p. 179; and
P. Milgrom and J. Roberts, “The Economics of Modern Manufacturing: Technology, Strategy, and Organization,” American Economic Review, volume 80, number 3, 1990, pp. 511–528.
5. J.F. Rockart and J.E. Short, “The Networked Organization and the Management of Interdependence,” in M. Scott Morton, ed., The Corporation of the 1990s: Information Technology and Organizational Transformation (Oxford: Oxford University Press, 1991), pp. 189–216.
6. Milgrom and Roberts (1990, 1993); and
Hauser and Clausing (1988).
7. M. Hammer and J. Champy, Reengineering the Corporation (New York: HarperCollins, 1993).
8. B.J. Bashein, M.L. Markus, and P. Riley, “Preconditions for BPR Success,” Information Systems Management, volume 11, Spring 1994, pp. 7–13; and
Hammer and Champy (1993).
9. J. Krafcik and J. MacDuffie, “Explaining High-Performance Manufacturing: The International Automotive Assembly Plant Study” (Acapulco, Mexico: International Motor Vehicle Policy Forum, 1989); and
R. Parthasarthy and S.P. Sethi, “Relating Strategy and Structure to Flexible Automation: A Test of Fit and Performance Implications,” Strategic Managment Journal, volume 14, October 1993, pp. 529–549.
10. J. Champy, Reengineering Management (New York: HarperBusiness, 1995); and
W.J. Orlikowski and J.D. Hofman, “An Improvisational Model for Change Management: The Case of Groupware Technologies,” Sloan Management Review, volume 38, Winter 1997, pp. 11–21.
11. R. Henderson and K. Clark, “Architectural Innovation: The Reconfiguration of Existing Product Technologies and the Failure of Established Firms,” Administrative Science Quarterly, volume 35, March 1990, pp. 9–30.
12. P. Osterman, “Impact of IT on Jobs and Skills,” in Scott Morton (1991), pp. 220–243.
13. R. Jaikumar, “Postindustrial Manufacturing,” Harvard Business Review, volume 64, November–December 1986, pp. 69–76.
14. F.F. Suarez, M.A. Cusumano, and C.H. Fine, “An Empirical Study of Flexible Manufacturing,” Sloan Management Review, volume 37, Fall 1995, pp. 25–32.
15. E. Brynjolfsson and L. Hitt, “Paradox Lost? Firm-Level Evidence of the Returns to Information Systems Spending,” Management Science, volume 42, April 1996, pp. 541–558; and
E. Brynjolfsson and L. Hitt, “Information Technology as a Factor of Production: The Role of Differences among Firms,” Economics of Innovation & New Technology, volume 3, January 1995, pp. 183–199.
16. L. Dudley and P. Lasserre, “Information as a Substitute for Inventories,” European Economic Review, volume 31, January 1989, pp. 1–21; and
Milgrom and Roberts (1990, 1993).
17. A.B. Austin, “Management and Scheduling Aspects of Increasing Flexibility in Manufacturing” (Cambridge, Massachusetts: MIT Sloan School of Management, master’s thesis, 1993).
18. Data supplied by MacroMed based on Nielsen data.
19. Austin (1993).
20. The matrix process has evolved since its inception as a research and consulting project originating in the Leaders for Manufacturing program at the MIT Sloan School of Management.
21. T. Davenport, Process Innovation (Boston: Harvard Business School Press, 1993).
22. J. Sterman, “Learning in and about Complex Systems,” System Dynamics Review, volume 10, number 3, 1994, pp. 291–330.
23. T.H. Davenport and J.E. Short, “The New Industrial Engineering: Information Technology and Business Process Redesign,” Sloan Management Review, volume 31, Summer 1990, pp. 11–27; and
J.M. Harrison and C.H. Loch, “Operations Management and Reengineering” (Stanford, California: Stanford Graduate School of Business, working paper, 1995).
24. Only pairwise interactions are identified. In principle, more complex interactions may be important; for instance, two practices may be complements only in the presence of a third practice. Typically, framing the question in terms of a reference set of practices will resolve such potential ambiguities without resorting to a matrix with more dimensions.
25. M.V. Alstyne, E. Brynjolfsson, and S. Madnick, “Why Not One Big Database? Principles for Data Ownership,” Decision Support Systems, volume 15, December 1995, pp. 267–284.
26. Harrison and Loch (1995).
27. Specification of such models must be done with some care since the clustering of variables will depend on the source of heterogeneity in the environment that gives rise to sample variation. See:
B. Holmstrom and P. Milgrom, “The Firm as an Incentive System,” American Economic Review, volume 84, September 1994, pp. 972–991.
28. Harrison and Loch (1995).
29. Davenport and Short (1990).
30. “Importance” can be usefully interpreted as “benefit-cost” of “the practice in isolation.” We revisit this interpretation in the section on determining net value added.
31. R.S. Kaplan and D.P. Norton, “The Balanced Scorecard —Measures That Drive Performance,” Harvard Business Review, volume 70, January–February 1992, pp. 71–79.
32. For a large system of interconnections, a graph-spanning algorithm can identify independent blocks of practices as well as connection counts within blocks.
33. An early analysis of the complete set of fifteen new practices discovered several other competing relationships. Most were associated with the practice of “line rationalization,” which seemed appealing when proposed in isolation (who could be opposed to “rationalization”?), but which conflicted with the principles of worker empowerment and flexibility embodied in many of the other practices.
34. Firms adopting the popular SAP software package report that it is inflexible in how it handles many basic business functions. Typically, this forces them to change their business practices to conform to the software’s requirements. Some customers view this as a feature, not a bug, because it compels recalcitrant managers to discard their old practices.
35. M. Tushman, W. Newman, and E. Romanelli, “Convergence and Upheaval: Managing the Unsteady Pace of Organizational Evolution,” in M. Tushman and W. Moore, eds., Readings in the Management of Innovation (New York: HarperBusiness, 1988), pp. 705–717.
36. Sterman (1994).
37. M. Tushman and E. Romanelli, “Organizational Evolution: A Metamorphosis Model of Convergence and Reorientation,” in Research in Organizational Behavior, volume 7 (Greenwich, Connecticut: JAI Press, 1985), pp. 171–222.
38. In principle, if the new system really does create more value, then it should be possible to compensate the losers so that everyone is better off. In practice, it may be hard to determine which grievances are legitimate, so some claims will need to go uncompensated. Nonetheless, if a majority of stakeholders appear to be worse off under a new system, this is a warning sign that the change is merely reallocating benefits and not creating much new value.
Milgrom and Roberts (1990, 1993); and
D.M. Rousseau, “Managing the Change to an Automated Office: Lessons from Five Case Studies,” Office: Technology and People, volume 4, January 1989, pp. 31–52.
39. M.J. Gallivan, D. Hofman, and W. Orlikowski, “Implementing Radical Change: Gradual versus Rapid Pace” (Vancouver, British Columbia: Association for Computing Machinery, 1994), pp. 325–339.
40. D. Leonard-Barton, “Implementation as Mutual Adaptation of Technology and Organization,” Research Policy, volume 17, 1988, pp. 251–267; and
D. Leonard-Barton, “Implementation Characteristics of Organizational Innovations: Limits and Opportunities for Management Strategies,” Communications Research, volume 15, number 5, 1988, pp. 603–631.
41. Gallivan et al. (1994).
42. Leonard-Barton (1988).
43. Gallivan et al. (1994), p. 336.
45. J. Sterman, N. Repenning, and F. Kofman, “Unanticipated Side Effects of Successful Quality Programs: Exploring a Paradox of Organizational Improvement,” Management Science, April 1997, forthcoming.
46. D. Croson, “Towards a Set of Information-Economy Management Principles in Improving Allocation Efficiency” (Cambridge, Massachusetts: Harvard University, Department of Economics, unpublished Ph.D dissertation, 1996).
47. R. Nolan and D. Croson, Creative Destruction (Boston: Harvard Business School Press, 1995).
48. Croson (1996), p. 32; and
Croson and Nolan (1995); and
N. Rosenberg, Inside the Black Box: Technology and Economics (Cambridge: Cambridge University Press, 1982).
49. Orlikowski and Hofman (1997).
50. Sterman (1994), p. 321.
51. Sterman et al. (1996).
52. Gallivan et al. (1994).
53. M.V. Alstyne, “The State of Network Organization: A Survey in Three Frameworks,” Journal of Organizational Computing, forthcoming;
C.C. Snow, R.E. Miles, and H.J. Coleman, “Managing 21st Century Network Organizations,” Organizational Dynamics, volume 20, Winter 1992, pp. 5–20.
54. T.W. Malone, J. Yates, and R.I. Benjamin, “Electronic Markets and Electronic Hierarchies,” Communications of the ACM, volume 30, number 6, 1987, pp. 484–497.
a. T.W. Malone, J. Yates, and R.I. Benjamin, “Electronic Markets and Electronic Hierarchies,” Communications of the ACM, volume 30, number 6, 1987, pp. 484–497; and
O.E. Williamson, Markets and Hierarchies (New York: North-Holland, 1975).
b. M.V. Alstyne, “The State of Network Organization: A Survey in Three Frameworks,” Journal of Organizational Computing, volume 7, forthcoming; and
W.W. Powell, “Neither Market nor Hierarchy: Network Forms of Organization,” Research in Organizational Behavior, volume 12 (Greenwich, Connecticut: JAI Press, 1990), pp. 295–336.
c. P. Milgrom and J. Roberts, “Complementarities and Fit: Strategy, Structure, and Organizational Change in Manufacturing,” Journal of Accounting and Economics, volume 19, 1993, p. 179;
P. Milgrom and J. Roberts, “The Economics of Modern Manufacturing: Technology, Strategy, and Organization,” American Economic Review, volume 80, number 3, 1990, pp. 511–528.