The Emerging Theory of Manufacturing

We cannot build it yet. But already we can specify the “postmodern” factory of 1999. Its essence will not be mechanical, though there will be plenty of machines. Its essence will be conceptual—the product of four principles and practices that together constitute a new approach to manufacturing. Each of these concepts is being developed separately, by different people with different starting points and different agendas. Each concept has its own objectives and its own kinds of impact. Statistical Quality Control is changing the social organization of the factory. The new manufacturing accounting lets us make production decisions as business decisions. The “flotilla,” or module, organization of the manufacturing process promises to combine the advantages of standardization and flexibility. Finally, the systems approach embeds the physical process of making things, that is, manufacturing, in the economic process of business, that is, the business of creating value. As these four concepts develop, they are transforming how we think about manufacturing and how we manage it. Most manufacturing people in the United States now know we need a new theory of manufacturing. We know that patching up old theories has not worked and that further patching will only push us further behind. Together these concepts give us the foundation for the new theory we so badly need. The most widely publicized of these concepts, Statistical Quality Control (SQC), is actually not new at all. It rests on statistical theory formulated 70 years ago by Sir Ronald Fisher. Walter Shewhart, a Bell Laboratories physicist, designed the original version of SQC in the 1930s for the zero-defects mass production of complex telephone exchanges and telephone sets. During World War II, W. Edwards Deming and Joseph Juran, both former members of Shewhart’s circle, separately developed the versions used today. The Japanese owe their leadership in manufacturing quality largely to their embrace of Deming’s precepts in the 1950s and 1960s. Juran too had great impact in Japan. But U.S. industry ignored their contributions for 40 years and is only now converting to SQC, with companies such as Ford, General Motors, and Xerox among the new disciples. Western Europe also has largely ignored the concept. More important, even SQC’s most successful practitioners do not thoroughly understand what it really does. Generally, it is considered a production tool. Actually, its greatest impact is on the factory’s social organization. By now, everyone with an interest in manufacturing knows that SQC is a rigorous, scientific method of identifying the quality and productivity that can be expected from a given production process in its current form so that control of both attributes can be built into the process itself. In addition, SQC can instantly spot malfunctions and show where they occur—a worn tool, a dirty spray gun, an overheating furnace. And because it can do this with a small sample, malfunctions are reported almost immediately, allowing machine operators to correct problems in real time. Further, SQC quickly identifies the impact of any change on the performance of the entire process. (Indeed, in some applications developed by Deming’s Japanese disciples, computers can simulate the effects of a proposed change in advance.) Finally, SQC identifies where, and often how, the quality and productivity of the entire process can be continuously improved. This used to be called the “Shewhart Cycle” and then the “Deming Cycle”; now it is kaizen, the Japanese term for continuous improvement. But these engineering characteristics explain only a fraction of SQC’s results. Above all, they do not explain the productivity gap between Japanese and U.S. factories. Even after adjusting for their far greater reliance on outside suppliers, Toyota, Honda, and Nissan turn out two or three times more cars per worker than comparable U.S. or European plants do. Building quality into the process accounts for no more than one-third of this difference. Japan’s major productivity gains are the result of social changes brought about by SQC. The Japanese employ proportionately more machine operators in direct production work than Ford or GM. In fact, the introduction of SQC almost always increases the number of machine operators. But this increase is offset many times over by the sharp drop in the number of nonoperators: inspectors, above all, but also the people who do not do but fix, like repair crews and “fire fighters” of all kinds. In U.S. factories, especially mass-production plants, such nonoperating, blue-collar employees substantially outnumber operators. In some plants, the ratio is two to one. Few of these workers are needed under SQC. Moreover, first-line supervisors also are gradually eliminated, with only a handful of trainers taking their place. In other words, not only does SQC make it possible for machine operators to be in control of their work, it makes such control almost mandatory. No one else has the hands-on knowledge needed to act effectively on the information that SQC constantly feeds back. By aligning information with accountability, SQC resolves a heretofore irresolvable conflict. For more than a century, two basic approaches to manufacturing have prevailed, especially in the United States. One is the engineering approach pioneered by Frederick Winslow Taylor’s “scientific management.” The other is the “human relations” (or “human resources”) approach developed before World War I by Andrew Carnegie, Julius Rosenwald of Sears Roe buck, and Hugo Münsterberg, a Harvard psychologist. The two approaches have always been considered antitheses, indeed, mutually exclusive In SQC, they come together. Taylor and his disciples were just as determined as Deming to build quality and productivity into the manufacturing process. Taylor asserted that his “one right way” guaranteed zero-defects quality; he was as vehemently opposed to inspectors as Deming is today. So was Henry Ford, who claimed that his assembly line built quality and productivity into the process (though he was otherwise untouched by Taylor’s scientific management and probably did not even know about it). But without SQC’s rigorous methodology, neither scientific management nor the assembly line could actually deliver built-in process control. With all their successes, both scientific management and the assembly line had to fall back on massive inspection, to fix problems rather than eliminate them. The human-relations approach sees the knowledge and pride of line workers as the greatest resource for controlling and improving quality and productivity It too has had important successes. But without the kind of information SQC provides, you cannot readily distinguish productive activity from busyness. It is also hard to tell whether a proposed modification will truly improve the process or simply make things look better in one corner, only to make them worse overall. Quality circles, which were actually invented and widely used in U.S. industry during World War II, have been successful in Japan because they came in after SQC had been established. As a result, both the quality circle and management have objective information about the effects of workers’ suggestions. In contrast, most U.S. quality circles of the last 20 years have failed despite great enthusiasm, especially on the part of workers. The reason? They were established without SQC, that is, without rigorous and reliable feedback. A good many U.S. manufacturers have built quality and productivity into their manufacturing processes without SQC and yet with a minimum of inspection and fixing. Johnson & Johnson is one such example. Other companies have successfully put ma chine operators in control of the manufacturing process without instituting SQC. IBM long ago replaced all first-line supervisors with a handful of “managers” whose main task is to train, while Herman Miller achieves zero-defects quality and high productivity through continuous training and productivity sharing incentives. But these are exceptions. In the main, the United States has lacked the methodology to build quality and productivity into the manufacturing process Similarly, we have lacked the methodology to move responsibility for the process and control of it to the machine operator, to put into practice what the mathematician Norbert Wiener called the “human use of human beings.” SQC makes it possible to attain both traditional aspirations: high quality and productivity on the one hand, work worthy of human beings on the other. By fulfilling the aims of the traditional factory, it provides the capstone for the edifice of twentieth century manufacturing that Frederick Taylor and Henry Ford designed. Bean counters do not enjoy a good press these days. They are blamed for all the ills that afflict U.S. manufacturing. But the bean counters will have the last laugh. In the factory of 1999, manufacturing accounting will play as big a role as it ever did and probably even a bigger one. But the beans will be counted differently. The new manufacturing accounting, which might more accurately be called “manufacturing economics,” differs radically from traditional cost accounting in its basic concepts. Its aim is to integrate manufacturing with business strategy. Manufacturing cost accounting (cost accounting’s rarely used full name) is the third leg of the stool—the other legs being scientific management and the assembly line—on which modern manufacturing industry rests. Without cost accounting, these two could never have become fully effective. It too is American in origin. Developed in the 1920s by General Motors, General Electric, and Western Electric (AT&T’s manufacturing arm), the new cost accounting, not technology, gave GM and GE the competitive edge that made them worldwide industry leaders. Following World War II, cost accounting became a major U.S. export. But by that time, cost accounting’s limitations also were becoming apparent. Four are particularly important. First, cost accounting is based on the realities of the 1920s, when direct, blue-collar labor accounted for 80 % of all manufacturing costs other than raw materials. Consequently, cost accounting equates “cost” with direct labor costs. Everything else is “miscellaneous,” lumped together as overhead. These days, however, a plant in which direct labor costs run as high as 25 % is a rare exception. Even in automobiles, the most labor intensive of the major industries, direct labor costs in up-to-date plants (such as those the Japanese are building in the United States and some of the new Ford plants) are down to 18 % . And 8 % to 12 % is fast becoming the industrial norm. One large manufacturing company with a labor-intensive process, Beckman Instruments, now considers labor costs “miscellaneous.” But typically, cost accounting systems are still based on direct labor costs that are carefully, indeed minutely, accounted for. The remaining costs—and that can mean 80 % to 90 % —are allocated by ratios that everyone knows are purely arbitrary and totally misleading: in direct proportion to a product’s labor costs, for example, or to its dollar volume. Second, the benefits of a change in process or in method are primarily defined in terms of labor cost savings. If other savings are considered at all, it is usually on the basis of the same arbitrary allocation by which costs other than direct labor are accounted for. Even more serious is the third limitation, one built into the traditional cost accounting system. Like a sundial, which shows the hours when the sun shines but gives no information on a cloudy day or at night, traditional cost accounting measures only the costs of producing. It ignores the costs of nonproducing, whether they result from machine downtime or from quality defects that require scrapping or reworking a product or part. Standard cost accounting assumes that the manufacturing process turns out good products 80 % of the time. But we now know that even with the best SQC, nonproducing time consumes far more than 20 % of total production time. In some plants, it accounts for 50 % . And nonproducing time costs as much as producing time does—in wages, heat, lighting, interest, salaries, even raw materials. Yet the traditional system measures none of this. Finally, manufacturing cost accounting assumes the factory is an isolated entity. Cost savings in the factory are “real.” The rest is “speculation”—for example, the impact of a manufacturing process change on a product’s acceptance in the market or on service quality. GM’s plight since the 1970s illustrates the problem with this assumption. Marketing people were unhappy with top management’s decision to build all car models, from Chevrolet to Cadillac, from the same small number of bodies, frames, and engines. But the cost accounting model showed that such commonality would produce substantial labor cost savings. And so marketing’s argument that GM cars would lose customer appeal as they looked more and more alike was brushed aside as speculation. In effect, traditional cost accounting can hardly justify a product improvement, let alone a product or process innovation. Automation, for instance, shows up as a cost but almost never as a benefit. All this we have known for close to 40 years. And for 30 years, accounting scholars, government accountants, industry accountants, and accounting firms have worked hard to reform the system. They have made substantial improvements. But since the reform attempts tried to build on the traditional system, the original limitations remain. What triggered the change to the new manufacturing accounting was the frustration of factory-automation equipment makers. The potential users, the people in the plants, badly wanted the new equipment. But top management could not be persuaded to spend the money on numerically controlled machine tools or robots that could rapidly change tools, fixtures, and molds. The benefits of automated equipment, we now know, lie primarily in the reduction of nonproducing time by improving quality (that is, getting it right the first time) and by sharply curtailing machine downtime in changing over from one model or product to another. But these gains cost accounting does not document. Out of this frustration came Computer-Aided Manufacturing-International, or CAM-I, a cooperative effort by automation producers, multinational manufacturers, and accountants to develop a new cost accounting system. Started in 1986, CAM-I is just beginning to influence manufacturing practice. But already it has unleashed an intellectual revolution. The most exciting and innovative work in management today is found in accounting theory, with new concepts, new approaches, new methodology—even what might be called new economic philosophy—rapidly taking shape. And while there is enormous controversy over specifics, the lineaments of the new manufacturing accounting are becoming clearer every day. As soon as CAM-I began its work, it became apparent that the traditional accounting system could not be reformed. It had to be replaced. Labor costs are clearly the wrong unit of measurement in manufacturing. But—and this is a new insight—so are all the other elements of production. The new measurement unit has to be time. The costs for a given period of time must be assumed to be fixed; there are no “variable” costs. Even material costs are more fixed than variable, since defective output uses as much material as good output does. The only thing that is both variable and controllable is how much time a given process takes. And “benefit” is whatever reduces that time. In one fell swoop, this insight eliminates the first three of cost accounting’s four traditional limitations. But the new cost concepts go even further by redefining what costs and benefits really are. For example, in the traditional cost accounting system, finished-goods inventory costs nothing because it does not absorb any direct labor. It is treated as an “asset.” In the new manufacturing accounting, however, inventory of finished goods is a “sunk cost” (an economist’s, not an accountant’s, term). Stuff that sits in inventory does not earn anything. In fact, it ties down expensive money and absorbs time. As a result, its time costs are high. The new accounting measures these time costs against the benefits of finished-goods inventory (quicker customer service, for instance). Yet manufacturing accounting still faces the challenge of eliminating the fourth limitation of traditional cost accounting: its inability to bring into the measurement of factory performance the impact of manufacturing changes on the total business—the return in the marketplace of an investment in automation, for instance, or the risk in not making an investment that would speed up production changeovers. The in-plant costs and benefits of such decisions can now be worked out with considerable accuracy. But the business consequences are indeed speculative. One can only say, “Surely, this should help us get more sales,” or “If we don’t do this, we risk falling behind in customer service.” But how do you quantify such opinions? Cost accounting’s strength has always been that it confines itself to the measurable and thus gives objective answers. But if intangibles are brought into its equations, cost accounting will only raise more questions. How to proceed is thus hotly debated, and with good reason. Still, everyone agrees that these business impacts have to be integrated into the measurement of factory performance, that is, into manufacturing accounting. One way or another, the new accounting will force managers, both inside and outside the plant, to make manufacturing decisions as business decisions. Henry Ford’s epigram, “The customer can have any color as long as it’s black,” has entered American folklore. But few people realize what Ford meant: flexibility costs time and money, and the customer won’t pay for it. Even fewer people realize that in the mid-1920s, the “new” cost accounting made it possible for GM to beat Ford by giving customers both colors and annual model changes at no additional cost. By now, most manufacturers can do what GM learned to do roughly 70 years ago. Indeed, many go quite a bit further in combining standardization with flexibility. They can, for example, build a variety of end products from a fairly small number of standardized parts. Still, manufacturing people tend to think like Henry Ford: you can have either standardization at low cost or flexibility at high cost, but not both. The factory of 1999, however, will be based on the premise that you not only can have both but also must have both—and at low cost. But to achieve this, the factory will have to be structured quite differently. Today’s factory is a battleship. The plant of 1999 will be a “flotilla,” consisting of modules centered either around a stage in the production process or around a number of closely related operations. Though overall command and control will still exist, each module will have its own command and control. And each, like the ships in a flotilla, will be maneuverable, both in terms of its position in the entire process and its relationship to other modules. This organization will give each module the benefits of standardization and, at the same time, give the whole process greater flexibility. Thus it will allow rapid changes in design and product, rapid response to market demands, and low-cost production of “options” or “specials” in fairly small batches. No such plant exists today. No one can yet build it. But many manufacturers, large and small, are moving toward the flotilla structure: among them are some of Westinghouse’s U.S. plants, Asea Brown Boveri’s robotics plant in Sweden, and several large printing plants, especially in Japan. The biggest impetus for this development probably came from GM’s failure to get a return on its massive (at least $ 30 billion and perhaps $ 40 billion) investment in automation. GM, it seems, used the new machines to improve its existing process, that is, to make the assembly line more efficient. But the process instead became less flexible and less able to accomplish rapid change. Meanwhile, Japanese automakers and Ford were spending less and attaining more flexibility. In these plants, the line still exists, but it is discontinuous rather than tightly tied together. The new equipment is being used to speed changes, for example, automating changeovers of jigs, tools, and fixtures. So the line has acquired a good bit of the flexibility of traditional batch production without losing its standardization. Standardization and flexibility are thus no longer an either-or proposition. They are—as indeed they must be—melded together. This means a different balance between standardization and flexibility, however, for different parts of the manufacturing process. An “average” balance across the plant will do nothing very well. If imposed throughout the line, it will simply result in high rigidity and big costs for the entire process, which is apparently what happened at GM. What is required is a reorganization of the process into modules, each with its own optimal balance. Moreover, the relationships between these modules may have to change whenever the product, process, or distribution changes. Switching from selling heavy equipment to leasing it, for instance, may drastically change the ratio between finished-product output and spare-parts output. Or a fairly minor model change may alter the sequence in which major parts are assembled into the finished product. There is nothing very new in this, of course. But under the traditional line structure, such changes are ignored, or they take forever to accomplish. With competition intensifying and product life cycles shortening all the time, such changes cannot be ignored, and they have to be done fast. Hence the flotilla’s modular organization. But this organization requires more than a fairly drastic change in the factory’s physical structure. It requires, above all, different communication and information. In the traditional plant, each sector and department reports separately upstairs. And it reports what upstairs has asked for. In the factory of 1999, sectors and departments will have to think through what information they owe to whom and what information they need from whom. A good deal of this information will flow sideways and across department lines, not upstairs. The factory of 1999 will be an information network. Consequently, all the managers in a plant will have to know and understand the entire process, just as the destroyer commander has to know and understand the tactical plan of the entire flotilla. In the factory of 1999, managers will have to think and act as team members, mindful of the performance of the whole. Above all, they will have to ask: What do the people running the other modules need to know about the characteristics, the capacity, the plans, and the performance of my unit? And what, in turn, do we in my module need to know about theirs? The last of the new concepts transforming manufacturing is systems design, in which the whole of manufacturing is seen as an integrated process that converts materials into goods, that is, into economic satisfactions. Marks & Spencer, the British retail chain, designed the first such system in the 1930s. Marks & Spencer designs and tests the goods (whether textiles or foods) it has decided to sell. It designates one manufacturer to make each product under contract. It works with the manufacturer to produce the right merchandise with the right quality at the right price. Finally, it organizes just-in-time delivery of the finished products to its stores. The entire process is governed by a meticulous forecast as to when the goods will move off store shelves and into customers’ shopping bags. In the last ten years or so, such systems management has become common in retailing. Though systems organization is still rare in manufacturing, it was actually first attempted there. In the early 1920s, when the Model T was in its full glory, Henry Ford decided to control the entire process of making and moving all the supplies and parts needed by his new plant, the gigantic River Rouge. He built his own steel mill and glass plant. He founded plantations in Brazil to grow rubber for tires. He bought the railroad that brought supplies to River Rouge and carried away the finished cars. He even toyed with the idea of building his own service centers nationwide and staffing them with mechanics trained in Ford-owned schools. But Ford conceived of all this as a financial edifice held together by ownership. Instead of building a system, he built a conglomerate, an unwieldy monster that was expensive, unmanageable, and horrendously unprofitable. In contrast, the new manufacturing system is not “controlled” at all. Most of its parts are independent—independent suppliers at one end, customers at the other. Nor is it plant centered, as Ford’s organization was. The new system sees the plant as little more than a wide place in the manufacturing stream. Planning and scheduling start with shipment to the final customer, just as they do at Marks & Spencer. Delays, halts, and redundancies have to be designed into the system—a warehouse here, an extra supply of parts and tools there, a stock of old products that are no longer being made but are still occasionally demanded by the market. These are necessary imperfections in a continuous flow that is governed and directed by information. What has pushed American manufacturers into such systems design is the trouble they encountered when they copied Japan’s just-in-time methods for supplying plants with materials and parts. The trouble could have been predicted, for the Japanese scheme is founded in social and logistic conditions unique to that country and unknown in the United States. Yet the shift seemed to American manufacturers a matter of procedure, indeed, almost trivial. Company after company found, however, that just-in-time delivery of supplies and parts created turbulence throughout their plants. And while no one could figure out what the problem was, the one thing that became clear was that with just-in-time deliveries, the plant no longer functions as a step-by-step process that begins at the receiving dock and ends when finished goods move into the shipping room. Instead, the plant must be redesigned from the end backwards and managed as an integrated flow. Manufacturing experts, executives, and professors have urged such an approach for two or three decades now. And some industries, such as petroleum refining and large-scale construction, do practice it. But by and large, American and European manufacturing plants are neither systems designed nor systems managed. In fact, few companies have enough knowledge about what goes on in their plants to run them as systems. Just-in-time delivery, however, forces managers to ask systems questions: Where in the plant do we need redundancy? Where should we place the burden of adjustments? What costs should we incur in one place to minimize delay, risk, and vulnerability in another? A few companies are even beginning to extend the systems concept of manufacturing beyond the plant and into the marketplace. Caterpillar, for instance, organizes its manufacturing to supply any replacement part anywhere in the world within 48 hours. But companies like this are still exceptions; they must become the rule. As soon as we define manufacturing as the process that converts things into economic satisfactions, it becomes clear that producing does not stop when the product leaves the factory. Physical distribution and product service are still part of the production process and should be integrated with it, coordinated with it, managed together with it. It is already widely recognized that servicing the product must be a major consideration during its design and production. By 1999, systems manufacturing will have an increasing influence on how we design and remodel plants and on how we manage manufacturing businesses. Traditionally, manufacturing businesses have been organized “in series,” with functions such as engineering, manufacturing, and marketing as successive steps. These days, that system is often complemented by a parallel team organization (Procter & Gamble’s product management teams are a well-known example), which brings various functions together from the inception of a new product or process project. If manufacturing is a system, however, every decision in a manufacturing business becomes a manufacturing decision. Every decision should meet manufacturing’s requirements and needs and in turn should exploit the strengths and capabilities of a company’s particular manufacturing system. When Honda decided six or seven years ago to make a new, upscale car for the U.S. market, the most heated strategic debate was not about design, performance, or price. It was about whether to distribute the Acura through Honda’s well-established dealer network or to create a new market segment by building separate Acura dealerships at high cost and risk. This was a marketing issue, of course. But the decision was made by a team of design, engineering, manufacturing, and marketing people. And what tilted the balance toward the separate dealer network was a manufacturing consideration: the design for which independent distribution and service made most sense was the design that best utilized Honda’s manufacturing capabilities. Full realization of the systems concept in manufacturing is years away. It may not require a new Henry Ford. But it will certainly require very different management and very different managers. Every manager in tomorrow’s manufacturing business will have to know and understand the manufacturing system. We might well adopt the Japanese custom of starting all new management people in the plant and in manufacturing jobs for the first few years of their careers. Indeed, we might go even further and require managers throughout the company to rotate into factory assignments throughout their careers—just as army officers return regularly to troop duty. In the new manufacturing business, manufacturing is the integrator that ties everything together. It creates the economic value that pays for everything and everybody. Thus the greatest impact of the manufacturing systems concept will not be on the production process. As with SQC, its greatest impact will be on social and human concerns—on career ladders, for instance, or more important, on the transformation of functional managers into business managers, each with a specific role, but all members of the same production and the same cast. And surely, the manufacturing businesses of tomorrow will not be run by financial executives, marketers, or lawyers inexperienced in manufacturing, as so many U.S. companies are today. There are important differences among these four concepts. Consider, for instance, what each means by “the factory.” In SQC, the factory is a place where people work. In management accounting and the flotilla concept of flexible manufacturing, it is a place where work is being done—it makes no difference whether by people, by white mice, or by robots. In the systems concept, the factory is not a place at all; it is a stage in a process that adds economic value to materials. In theory, at least, the factory cannot and certainly should not be designed, let alone built, until the entire process of “making”—all the way to the final customer—is understood. Thus defining the factory is much more than a theoretical or semantic exercise. It has immediate practical consequences on plant design, location, and size; on what activities are to be brought together in one manufacturing complex; even on how much and in what to invest. Similarly, each of these concepts reflects a particular mind-set. To apply SQC, you don’t have to think, you have to do. Management accounting concentrates on technical analysis, while the flotilla concept focuses on organization design and work flow. In the systems concept, there is great temptation to keep on thinking and never get to the doing. Each concept has its own tools, its own language, and addresses different people. Nevertheless, what these four concepts have in common is far more important than their differences. Nowhere is this more apparent than in their assumption that the manufacturing process is a configuration, a whole that is greater than the sum of its parts. Traditional approaches all see the factory as a collection of individual machines and individual operations. The nineteenth century factory was an assemblage of machines. Taylor’s scientific management broke up each job into individual operations and then put those operations together into new and different jobs. “Modern” twentieth century concepts—the assembly line and cost accounting—define performance as the sum of lowest cost operations. But none of the new concepts is much concerned with performance of the parts. Indeed, the parts as such can only underperform. The process produces results. Management also will reflect this new perspective. SQC is the most nearly conventional in its implications for managers, since it does not so much change their job as shift much of it to the work force. But even managers with no business responsibility (and under SQC, plant people have none) will have to manage with an awareness of business considerations well beyond the plant. And every manufacturing manager will be responsible for integrating people, materials, machines, and time. Thus every manufacturing manager ten years hence will have to learn and practice a discipline that integrates engineering, management of people, and business economics into the manufacturing process. Quite a few manufacturing people are doing this already, of course—though usually unaware that they are doing something new and different. Yet such a discipline has not been systematized and is still not taught in engineering schools or business schools. These four concepts are synergistic in the best sense of this much-abused term. Together—but only together—they tackle the conflicts that have most troubled traditional, twentieth century mass-production plants: the conflicts between people and machines, time and money, standardization and flexibility, and functions and systems. The key is that every one of these concepts defines performance as productivity and conceives of manufacturing as the physical process that adds economic value to materials. Each tries to provide economic value in a different way. But they share the same theory of manufacturing.

The Emerging Theory of Manufacturing

Partner Center