Among strategy researchers, Jeffrey Dyer has done one of the cleanest jobs of putting isolating mechanisms under the microscope. Jeff started his investigations of Japanese automobile makers when he was a doctoral student at UCLA, working under the direction of Bill Ouchi. Since then he has crafted a wonderful comparison—he has looked at U.S. companies that make automobile components for both Toyota and American auto firms. For instance, a U.S. company making carburetors might have two different manufacturing “cells,” one making parts for General Motors and another making parts for Toyota. Dyer measured the quality and labor productivity in both cells and tracked their changes over time.
In the same company, in the same plant, making the same type of product, the cell supplying Toyota improved quality and labor productivity faster than the cell supplying the U.S. auto company. Dyer attributes the difference to Toyota’s greater attention to learning and improvement. “But,” he asks, “why doesn’t the supplier replicate what they have learned from Toyota in their operations devoted to other customers?”1
In interviews, Dyer found that most of the plant managers wanted to transfer best practices from the Toyota cell to other manufacturing cells, but had been unable to do so. For example, one plant manager reported that defect rates in the U.S. automaker’s cell were hard to reduce because the parts knocked against one another in the large shipping container. But he couldn’t use a smaller container because the container size was, for this automaker, a worldwide standard. The overarching problem was that the U.S. automakers’ production methods emphasized automation and large lots, large shipping containers, and truckload shipments. Toyota, on the other hand, had less automation, smaller containers, and shipped more evenly. While the U.S. system tried to exploit economies of scale and automation, it was much harder to adjust. While the Toyota system was harder to manage and potentially more fragile, it was also easier to adjust and change as knowledge accumulated.
Jeffrey Dyer’s study showed that the auto suppliers were blocked by the automakers’ manufacturing processes. But this just pushes the question up to a higher level: What keeps U.S. automakers from adopting the Toyota system? After all, General Motors has joint-ventured with Toyota and seen its operations up close.
There are a number of practical management and engineering problems a company faces in adopting a “best practice” like the Toyota system of manufacturing. But each could be surmounted with sufficient effort. To the extent that the skills and knowledge in Toyota can be codified, written down and understood, they can eventually be copied by another. Beyond that, through apprenticeship and training, much of what a person knows, but cannot codify or write down, can nonetheless be transmitted to others. Fighter pilots cannot tell you all that they know about flying and fighting, but the skills can nevertheless be taught and passed on.
What cannot be moved from one organization to another or from one nation to another, is what I have come to call distributed knowledge. Distributed knowledge exists in a group of people, each contributing a small piece of the whole. The essential economic element of distributed knowledge is that no one person is essential—there is significant redundancy and each can be replaced by a newcomer who can learn their role. The essential social element is that each person learns much of his or her role by interacting with the people around them, taking in information, norms of behavior, and bases for judgment.
Inside the firm, distributed knowledge is often called know-how, or culture, or good management. Visit several firms in an industry and each has a distinctive character—what my colleague Sumantra Ghoshal used to call “the smell of the place.” For good or for ill, these human systems are very difficult to engineer, to change, and to replicate.
Among schools of business, INSEAD (France) is known for its executive development programs. During the three years I served on the INSEAD faculty, I marveled at how the school marketed, planned, and executed these programs. Each faculty member had two or three “chunks” of material they were practiced at presenting to executives and most knew about each others’ “chunks.” The marketing staff understood the intellectual elements of the programs as well as knowing a lot about customers and potential customers. The school maintained a computerized contact file that recorded each faculty or staff contact with an executive and recorded their comments and interests. Several thousand executives were tracked in the system. Specialist program coordinators spoke several languages and took care of all of the materials, logistics, travel, dining, and accommodations issues. The administration gave early program planning responsibility to young faculty members. They quickly learned that one could not easily recruit faculty teachers for a program unless you were also available for their programs. The senior faculty and dean expected each faculty member to do some executive teaching. Unlike many U.S. schools, where such activity is restricted to a few, their attitude was clear—if you were a business school faculty member and didn’t have at least one three-hour chunk of material of interest to executives, there was something very wrong. The whole thing worked with admirable smoothness, bringing in funds for the school and exposing faculty to a wide variety of executives and their viewpoints.
On returning to UCLA I made a nuisance of myself by lecturing colleagues and administrators about how our own executive programs could benefit from INSEAD’s methods. But I made little progress because I had no realistic action plan. Each element of INSEAD’s system worked because other elements worked—there was no obvious place to crack the system open and start.
The INSEAD system was distributed knowledge. Each faculty member, administrator, and program manager learned how to perform their role by participating in the system. Each performance depended upon the implicitly coordinated performances of others. No one knew everything and there was no “handbook.” Knowing the basic rules and overall architecture of the system, as I did, was simply not enough to replicate it. It was like knowing how to drive a car, but asking my colleagues to build a car from scratch. INSEAD does not need patents or secrecy to protect its skills at executive education from imitators—distributed knowledge does the job.
Distributed knowledge within the firm resists replication by other firms. Even more importantly, distributed knowledge outside the firm can also be a critical source of advantage. Outside the firm, distributed knowledge exists as reputation, brand image, learning, and relationships, where these apply to the ultimate buyers was well as to suppliers and retailers. Like employees inside the firm, people outside the firm hold knowledge and opinions about how to conduct business, choose products, and consume. And this distributed knowledge, the combined cognitive states and consequent behavior of hundreds, thousands, or millions of people, is a critical resource for the business. If it could be simply created by advertising or marketing efforts, it could be replicated by others. But a great deal of this distributed knowledge is socially very complex. Coke’s advertisements have influence, but they also reach a public that has learned from each other the unspoken rules for thinking about and consuming Coca-Cola. It is this distributed yet coordinated knowledge that is the company’s essential strategic asset
Examples of distributed knowledge abound, once your eyes are tuned to see it. Put on a SCUBA tank, drop fifty feet below the ocean’s surface and watch a school of small silver fish. The group is a rough sphere and sunlight glints off each fish’s scales as they swim in one direction together. Then, like a marching squad responding to “By the left flank, turn,” each fish suddenly turns 90 degrees and darts in the new direction, still in perfect formation. There is a lead fish whose turn initiates the school’s turn, but most of the fish do not see the leader. Instead they watch the fish around them and respond to their movements. The turn, initiated by the leader, is carried out by social coordination. And once they are moving in the new direction, a different fish assumes the role of leader. If the members of the school are inattentive to their neighbors, they will not only miss the turn, they will also provide inaccurate signals to the other fish around them.
Other examples from the natural world are swarms of birds, bats, and bees. The video clip below shows about eighty-thousand European starlings flying in a controlled mass over the Piazza Cinquecento in Rome. Each bird takes it cue from four or five birds nearby, allowing the whole mass to maintain coherence (direct link to video).
In human systems, the coordination goes far beyond physical movements to include coordinated intelligence. The distributed knowledge and routines underlying this coordination can be a strong source of advantage because they resists imitation and because the individuals involved are each replaceable. Distributed knowledge vested in a group whose members could not be replaced would be equally valuable, but the value would eventually accrue to members of the group rather than to the firm.
As a concept, distributed knowledge has the pleasing quality that it is a unified theory explaining the advantages gained from knowledge within the firm as well as knowledge distributed among suppliers or customers outside of the firm. In each case, the value derives from coordinated cognitive states and learned behavior that can be each passed on to new group members. Distributed knowledge can be preserved and expanded but not easily replicated.
Some economists believe that organization is a matter of incentives. Give people the right rewards, they tell us, and people will perform. But exemplary organizational performance is more than the sum of individual performances, it requires flexible social coordination. The members of an organization must respond quickly to the subtle signals of their peers and continually adjust their behavior. And the knowledge of the organization—the rules about how to skillfully play each role within it—is distributed amongst its members. Here, at the mysterious heart of profit, we find knowledge and skill that are not knowable by any one mind. Call it distributed knowledge or call it culture, it remains the ultimate source of persistent advantage.
- Dyer, Jeffrey H. Collaborative Advantage. New York: Oxford University Press, 2000: 81. [↩]