https://hbr.org/1986/09/why-some-factories-are-more-productive-than-others
|
Why Some Factories Are More Productive Than Others
by Robert H. Hayes and Kim B. Clark
From the Magazine (September 1986) - HBR
Post
Post
Share
Save
Buy Copies
The battle for attention is over. The time for banging drums is long past. Everyone now understands that manufacturing provides an essential source of competitive leverage. No longer does anyone seriously think that domestic producers can outdo their competitors by clever marketing only—“selling the sizzle” while cheating on quality or letting deliveries slip. It is now time for concrete action on a practical level: action to change facilities, update processing technologies, adjust work-force practices, and perfect information and management systems.
But when managers turn to these tasks, they quickly run up against a stumbling block. Namely, they do not have adequate measures for judging factory-level performance or for comparing overall performance from one facility to the next. Of course, they can use the traditional cost-accounting figures, but these figures often do not tell them what they really need to know. Worse, even the best numbers do not sufficiently reflect the important contributions that managers can make by reducing confusion in the system and promoting organizational learning.
Consider the experience of a U.S. auto manufacturer that discovered itself with a big cost disadvantage. The company put together a group to study its principal competitor’s manufacturing operations. The study generated reams of data, but the senior executive in charge of the activity still felt uneasy. He feared that the group was getting mired in details and that things other than managerial practices—like the age of facilities and their location—might be the primary drivers of performance. How to tell?
Similarly, a vice president of manufacturing for a specialty chemical producer had misgivings about the emphasis his company’s system for evaluating plant managers placed on variances from standard costs. Differences in these standards made comparisons across plants difficult. What was more troubling, the system did not easily capture the trade-offs among factors of production or consider the role played by capital equipment or materials. What to do?
Another manufacturer—this time of paper products—found quite different patterns of learning in the same departments of five of its plants scattered across the United States. Although each department made much the same products using similar equipment and materials, they varied widely in performance over a period of years. Why such differences?
Our point is simple: before managers can pinpoint what’s needed to boost manufacturing performance, they must have a reliable way of ascertaining why some factories are more productive than others. They also need a dependable metric for identifying and measuring such differences and a framework for thinking about how to improve their performance—and keep it improving. This is no easy order.
These issues led us to embark on a continuing, multiyear study of 12 factories in 3 companies (see the appendix for details on research methodology). The study’s purpose is to clarify the variables that influence productivity growth at the micro level.
Appendix: Research Methods
There are three basic approaches for identifying the effects of management actions and policies on
...
The first company we looked at, which employs a highly connected and automated manufacturing process, we refer to as the Process Company. Another, which employs a batch approach based on a disconnected line-flow organization of work, we refer to as the Fab (fabrication-assembly) Company. The third, which uses several different batch processes to make components for sophisticated electronic systems, is characterized by very rapid changes in both product and process. We refer to it as the Hi-Tech Company. All five factories of the Process Company and three of the four factories of the Fab Company are in the United States (the fourth is just across the border in Canada). Of the three factories belonging to the Hi-Tech Company, one is in the United States, one in Europe, and one in Asia.
In none of these companies did the usual profit-and-loss statements—or the familiar monthly operating reports—provide adequate, up-to-date information about factory performance. Certainly, managers routinely evaluated such performance, but the metrics they used made their task like that of watching a distant activity through a thick, fogged window. Indeed, the measurement systems in place at many factories obscure and even alter the details of their performance.
A Fogged Window
Every plant we studied employed a traditional standard cost system: the controller collected and reported data each month on the actual costs incurred during the period for labor, materials, energy, and depreciation, as well as on the costs that would have been incurred had workers and equipment performed at predetermined “standard” levels. The variances from these standard costs became the basis for problem identification and performance evaluation. Other departments in the plants kept track of head counts, work-in-process inventory, engineering changes, the value of newly installed equipment, reject rates, and so forth.
In theory, this kind of measurement system should take a diverse range of activities and summarize them in a way that clarifies what is going on. It should act like a lens that brings a blurry picture into sharp focus. Yet, time and again, we found that these systems often masked critical developments in the factories and, worse, often distorted management’s perspective.
Each month, most of the managers we worked with received a blizzard of variance reports but no overall measure of efficiency. Yet this measure is not hard to calculate. In our study, we took the same data generated by plant managers and combined them into a measure of the total factor productivity (TFP)—the ratio of total output to total input (see the appendix for more details on TFP).
This approach helps dissipate some of the fog—especially because our TFP data are presented in constant dollars instead of the usual current dollars. Doing so cuts through the distortions produced by periods of high inflation. Consider the situation at Fab’s Plant 1, where from 1974 to 1982 output fluctuated between $45 million and $70 million—in nominal (current dollar) terms. In real terms, however, there was a steep and significant decline in unit output. Several executives initially expressed disbelief at the magnitude of this decline because they had come to think of the plant as a “$50 million plant.” Their traditional accounting measures had masked the fundamental changes taking place.
Another advantage of the TFP approach is that it integrates the contributions of all the factors of production into a single measure of total input. Traditional systems offer no such integration. Moreover, they often overlook important factors. One of the plant managers at the Process Company gauged performance in a key department by improvements in labor hours and wage costs. Our data showed that these “improvements” came largely from the substitution of capital for labor. Conscientious efforts to prune labor content by installing equipment—without developing the management skills and systems needed to realize its full potential—proved shortsighted. The plant’s TFP (which, remember, takes into account both labor and capital costs) improved very little.
This preoccupation with labor costs, particularly direct labor costs, is quite common—even though direct labor now accounts for less than 15% of total costs in most manufacturing companies. The managers we studied focused heavily on these costs; indeed, their systems for measuring direct labor were generally more detailed and extensive than those for measuring other inputs that were several times more costly. Using sophisticated bar-code scanners, Hi-Tech’s managers tracked line operators by the minute but had difficulty identifying the number of manufacturing engineers in the same department. Yet these engineers accounted for 20% to 25% of total cost—compared with 5% for line operators.
Just as surprising, the companies we studied paid little attention to the effect of materials consumption or productivity. Early on, we asked managers at one of the Fab plants for data on materials consumed in production during each of a series of months. Using these data to estimate materials productivity gave us highly erratic values.
Investigation showed that this plant, like many others, kept careful records of materials purchased but not of the direct or indirect materials actually consumed in a month. (The latter, which includes things like paper forms, showed up only in a catchall manufacturing overhead account.) Further, most of the factories recorded materials transactions only in dollar, rather than in physical, terms and did not readily adjust their standard costs figures when inflation or substitution altered materials prices.
What managers at Fab plants called “materials consumed” was simply an estimate derived by multiplying a product’s standard materials cost—which itself assumes a constant usage of materials—by its unit output and adding an adjustment based on the current variation from standard materials prices. Every year or half-year, managers would reconcile this estimated consumption with actual materials usage, based on a physical count. As a result, data on actual materials consumption in any one period were lost.
Finally, the TFP approach makes clear the difference between the data that managers see and what those data actually measure. In one plant, the controller argued that our numbers on engineering changes were way off base. “We don’t have anything like this level of changes,” he claimed. “My office signs off on all changes that go through this place, and I can tell you that the number you have here is wrong.” After a brief silence, the engineering manager spoke up. He said that the controller reviewed only very large (in dollar terms) engineering changes and that our data were quite accurate. He was right. The plant had been tracking all engineering changes, not just the major changes reported to the controller.
A Clear View
With the foglike distortions of poor measurement systems cleared away, we were able to identify the real levers for improving factory performance. Some, of course, were structural—that is, they involve things like plant location or plant size, which lie outside the control of a plant’s managers. But a handful of managerial policies and practices consistently turned up as significant. Across industries, companies, and plants, they regularly exerted a powerful influence on productivity. In short, these are the managerial actions that make a difference.
Invest Capital
Our data show unequivocally that capital investment in new equipment is essential to sustaining growth in TFP over a long time (that is, a decade or more). But they also show that capital investment all too often reduces TFP for up to a year. Simply investing money in new technology or systems guarantees nothing. What matters is how their introduction is managed, as well as the extent to which they support and reinforce continual improvement throughout a factory. Managed right, new investment supports cumulative, long-term productivity improvement and process understanding—what we refer to as “learning.”
The Process Company committed itself to providing new, internally designed equipment to meet the needs of a rapidly growing product. Over time, as the company’s engineers and operating managers gained experience, they made many small changes in product design, machinery, and operating practices. These incremental adjustments added up to major growth in TFP.
Seeking new business, the Fab company redesigned an established product and purchased the equipment needed to make it. This new equipment was similar to the plant’s existing machinery, but its introduction allowed for TFP-enhancing changes in work flows. Plant managers discovered how the new configuration could accommodate expanded production without a proportional increase in the work force. These benefits spilled over: even the older machinery was made to run more efficiently.
In both cases, the real boost in TFP came not just from the equipment itself but also from the opportunities it provided to search for and apply new knowledge to the overall production process. Again, managed right, investment unfreezes old assumptions, generates more efficient concepts and designs for a production system, and expands a factory’s skills and capabilities.
Exhibit I shows the importance of such learning for long-term TFP growth at one of Fab’s plants between 1973 and 1982. TFP rose by 96%. Part of this increase, of course, reflected changes in utilization rates and the introduction of new technology. Even so, roughly two-thirds (65%) of TFP growth was learning-based, and fully three-fourths of that learning effect (or 49% of TFP growth) was related to capital investment. Without capital investment, TFP would have increased, but at a much slower rate.
Exhibit I Capital Investment, Learning, and Productivity Growth in Fab Company’s Plant 2 1973–1982 These estimates are based on a regression analysis of TFP growth. We estimated learning-related changes by using both a time trend and cumulative output. The capital-related learning effect represents the difference between the total learning effect and the effect that remained one capital was introduced into the regression. The total capital effect is composed of a learning component and a component reflecting technical advance.
Such long-term benefits incur costs; in fact, the indirect costs associated with introducing new equipment can be staggering. In Fab’s Plant 1, for example, a $1 million investment in new equipment imposed $1.75 million of additional costs on the plant during its first year of operation! Had the plant cut these indirect costs by half, TFP would have grown an additional 5% during that year.
Everyone knows that putting in new equipment usually causes problems. Everyone expects a temporary drop in efficiency as equipment is installed and workers learn to use it. But managers often underestimate the costly ripple effects of new equipment on inventory, quality, equipment utilization, reject rates, downtime, and material waste. Indeed, these indirect costs often exceed the direct cost of the new equipment and can persist for more than a year after the equipment is installed.
Here, then, is the paradox of capital investment. It is essential to long-term productivity growth, yet in the short run, if poorly managed, it can play havoc with TFP. It is risky indeed for a company to try to “invest its way” out of a productivity problem. Putting in new equipment is just as likely to create confusion and make things worse for a number of months. Unless the investment is made with a commitment to continual learning—and unless performance measures are chosen carefully—the benefits that finally emerge will be small and slow in coming. Still, many companies today are trying to meet their competitive problems by throwing money at them—new equipment and new plants. Our findings suggest that there are other things they ought to do first, things that take less time to show results and are much less expensive.
Reduce Waste
We were not surprised to find a negative correlation between waste rates (or the percentage of rejects) and TFP, but we were amazed by its magnitude. In the Process plants, changes in the waste rate (measured by the ratio of waste material to total cost, expressed as a percentage) led to dramatic operating improvements. As Exhibit II shows, reducing the percentage of waste in Plant 4’s Department C by only one-tenth led to a 3% improvement in TFP, conservatively estimated.
Exhibit II Impact of Waste on TFP in Process Company Plants
The strength of this relationship is more surprising when we remember that a decision to boost the production throughput rate (which ought to raise TFP because of the large fixed components in labor and capital costs) also causes waste ratios to increase. In theory, therefore, TFP and waste percent should increase together. The fact that they do not indicates the truly powerful impact that waste reduction has on productivity.
Get WIP Out
The positive effect on TFP of cutting work-in-process (WIP) inventories for a given level of output was much greater than we could explain by reductions in working capital. Exhibit III documents the relationship between WIP reductions and TFP in the three companies. Although there are important plant-to-plant variations, all reductions in WIP are associated with increases in TFP. In some plants, the effect is quite powerful; in Department D of Hi-Tech’s Plant 1, reducing WIP by one-tenth produced a 9% rise in TFP.
Exhibit III Impact of Work-in-Process Reductions on TFP
These data support the growing body of empirical evidence about the benefits of reducing WIP. From studies of both Japanese and American companies, we know that cutting WIP leads to faster, more reliable delivery times, lowers reject rates (faster production cycle times reduce inventory obsolescence and make possible rapid feedback when a process starts to misfunction), and cuts overhead costs. We now know it also drives up TFP.
The trouble is, simply pulling work-in-process inventory out of a factory will not, by itself, lead to such improvements. More likely, it will lead to disaster. WIP is there for a reason, usually for many reasons; it is a symptom, not the disease itself. A long-term program for reducing WIP must attack the reasons for its being there in the first place: erratic process yields, unreliable equipment, long production changeover and set-up times, ever-changing production schedules, and suppliers who do not deliver on time. Without a cure for these deeper problems, a factory’s cushion of WIP is often all that stands between it and chaos.
Reducing Confusion
Defective products, mismanaged equipment, and excess work-in-process inventory are not only problems in themselves. They are also sources of confusion. Many things that managers do can confuse or disrupt a factory’s operation: erratically varying the rate of production, changing a production schedule at the last minute, overriding the schedule by expediting orders, changing the crews (or the workers on a specific crew) assigned to a given machine, haphazardly adding new products, altering the specifications of an existing product through an engineering change order (ECO), or monkeying with the process itself by adding to or altering the equipment used.
Managers may be tempted to ask, “Doesn’t what you call confusion—changing production schedules, expediting orders, shifting work crews, adding or overhauling equipment and changing product specifications—reflect what companies inevitably have to do to respond to changing customer demands and technological opportunities?”
Our answer to this question is an emphatic No! Responding to new demands and new opportunities requires change, but it does not require the confusion it usually creates. Much of our evidence on confusion comes from factories that belong to the same company and face the same external pressures. Some plant managers are better than others at keeping these pressures at bay. The good ones limit the number of changes introduced at any one time and carefully handle their implementation. Less able managers always seem caught by surprise, operate haphazardly, and leapfrog from one crisis to the next. Much of the confusion in their plants is internally generated.
While confusion is not the same thing as complexity, complexity in a factory’s operation usually produces confusion. In general, a factory’s mission becomes more complex—and its focus looser—as it becomes larger, as it adds different technologies and products, and as the number and variety of production orders it must accommodate grow. Although the evidence suggests that complexity harms performance, each company’s factories were too similar for us to analyze the effects of complexity on TFP. But we could see that what managers did to mitigate or fuel confusion within factories at a given level of complexity had a profound impact on TFP.
Of the sources of confusion we examined, none better illustrated this relationship with TFP than engineering change orders. ECOs require a change in the materials used to make a product, the manufacturing process employed, or the specifications of the product itself. We expected ECOs to lower productivity in the short run but lead to higher TFP over time. Exhibit IV, which presents data on ECO activity in three Fab plants, shows its effects to be sizable. In Plant 2, for example, increasing ECOs by just ten per month reduced TFP by almost 5%. Moreover, the debilitating effects of ECOs persisted for up to a year.
Exhibit IV Impact of Engineering Change Orders on TFP in Three Fab Company Plants
Our data suggest that the average level of ECOs implemented in a given month, as well as the variation in this level, is detrimental to TFP. Many companies would therefore be wise to reduce the number of ECOs to which their plants must respond. This notion suggests, in turn, that more pressure should be placed on engineering and marketing departments to focus attention on only the most important changes—as well as to design things right the first time.
Essential ECOs should be released in a controlled, steady fashion rather than in bunches. In the one plant that divided ECOs into categories reflecting their cost, low-cost ECOs were most harmful to TFP. More expensive ECOs actually had a positive effect. The reason: plant managers usually had warning of major changes and, recognizing that they were potentially disruptive, carefully prepared the ground by warning supervisors, training workers, and bringing in engineers. By contrast, minor ECOs were simply dumped on the factory out of the blue.
Value of Learning
If setting up adequate measures of performance is the first step toward getting full competitive leverage out of manufacturing, identifying factory-level goals like waste or WIP reduction is the second. But without making a commitment to ongoing learning, a factory will gain no more from these first two steps than a one-time boost in performance. To sustain the leverage of plant-level operations, managers must pay close attention to—and actively plan for—learning.
We are convinced that a factory’s learning rate—the rate at which its managers and operators learn to make it run better—is at least of equal importance as its current level of productivity. A factory whose TFP is lower than another’s, but whose rate of learning is higher, will eventually surpass the leader. Confusion, as we have seen, is especially harmful to TFP. Thus the two essential tasks of factory management are to create clarity and order (that is, to prevent confusion) and to facilitate learning.
But doesn’t learning always involve a good deal of experimentation and confusion? Isn’t there an inherent conflict between creating clarity and order and facilitating learning? Not at all.
Confusion, like noise or static in an audio system, makes it hard to pick up the underlying message or figure out the source of the problem. It impedes learning, which requires controlled experimentation, good data, and careful analysis. It chews up time, resources, and energy in efforts to deal with issues whose solution adds little to a factory’s performance. Worse, engineers, supervisors, operators, and managers easily become discouraged by the futility of piecemeal efforts. In such environments, TFP lags or falls.
Reducing confusion and enhancing learning do not conflict. They make for a powerful combination—and a powerful lever on competitiveness. A factory that manages change poorly, that does not have its processes under control, and that is distracted by the noise in its systems learns too slowly, if at all, or learns the wrong things.
In such a factory, new equipment will only create more confusion, not more productivity. Equally troubling, both managers and workers in such a factory will be slow to believe reports that a sister plant—or a competitor’s plant—can do things better than they can. If the evidence is overwhelming, they will simply argue, “It can’t work here. We’re different.” Indeed they are—and less productive too.
“Where the Money Is.”
Many companies have tried to solve their data-processing problems by bringing in computers. They soon learned that computerizing a poorly organized and error-ridden information system simply creates more problems: garbage in, garbage out. That lesson, learned so long ago, has been largely forgotten by today’s managers, who are trying to improve manufacturing performance by bringing in sophisticated new equipment without first reducing the complexity and confusion of their operations.
Spending big money on hardware fixes will not help if managers have not taken the time to simplify and clarify their factories’ operations, eliminate sources of error and confusion, and boost the rate of learning. Of course, advanced technology is important, often essential. But there are many things that managers must do first to prepare their organizations for these new technologies.
When plant managers are stuck with poor measures of how they are doing and when a rigid, by-the-book emphasis on standards, budgets, and exception reports discourages the kind of experimentation that leads to learning, the real levers on factory performance remain hidden. No amount of capital investment can buy heightened competitiveness. There is no way around the importance of building clarity into the system, eliminating unnecessary disruptions and distractions, ensuring careful process control, and nurturing in-depth technical competence. The reason for understanding why some factories perform better than others is the same reason that Willie Sutton robbed banks: “That’s where the money is.”
Recommended For You
Generative AI
Research: The Hidden Penalty of Using AI at Work
Leadership
Now Is the Time for Courage
Leadership Qualities
Why Leaders Should Bring Their Best Self—Not Their Whole Self—to Work
Podcast
How Leaders Undermine Their Own Authority
Readers Also Viewed These Items
HBR Guide to Better Recruiting and Hiring Toolkit
Book
Buy Now
The Activator Advantage: What Today's Rainmakers Do Differently
Book
Buy Now
Read more on Performance indicators or related topics Personal productivity and Operations and supply chain management
A version of this article appeared in the September 1986 issue of Harvard Business Review.
Robert H. Hayes is the Philip Caldwell Professor of Business Administration, emeritus, at the Harvard Business School.
Kim B. Clark served as dean of the faculty at Harvard Business School from 1995 to 2005. He also served as the fifteenth president of Brigham Young University–Idaho from 2005 to 2015 and Commissioner of Education for the Church of Jesus Christ of Latter-day Saints from 2015 to 2019. He is currently a Distinguished Professor of Management at Brigham Young University. He is the co-author of Leading Through: Activating the Soul, Heart, and Mind of Leadership (Harvard Business Review Press).
Post
Post
