THE SO-CALLED American System of Manufacturing, “the interchangeable system” which Americans had developed by the mid-nineteenth century, had sprung from a grand and simple new organizing idea which exploited American opportunities. It had been designed to produce large quantities of guns without numerous craft-skilled gunsmiths, and so made a virtue of New World limitations. A century later, the new science of statistics had provided a tool for shaping another characteristically American system of manufacturing in the mid-twentieth century. Few Americans were even aware of the revolution that had taken place, but the revolution was there, a symbol of the risks and opportunities of the quantitative approach to experience.
The earlier American system had been based on a new way of thinking about the “quality” of manufactured goods. For example, a Scottish engineer inspecting the Baldwin locomotive works in Philadelphia in the 1830’s noted that “the external parts, such as the connecting rods, cranks, framing, and wheels, were left in a much coarser state than in engines of British manufacture” while “those parts of the engine, such as the cylinder, piston, valves, journals, and slides, in which good fitting and fine workmanship are indispensable to the efficient action of the machine, were very highly finished.” Some of the visiting experts concluded that even the “best” American machinery was only “tolerably well finished, such as we would call second-rate machinery in our own country.” What they saw confirmed their conviction that skilled hand labor really was indispensable for finishing machine-made products. They predicted, in other words, that in the long run the American interchangeable system, which aimed to make the machine do all the work, could not succeed, for its products were bound to be of poor quality.
But Americans had come to a new definition of what they meant by “quality.” Function had come to displace perfection. When Americans judged a machine, a gun, a lock or a clock by its ability to do its job, all those fluted columns, ornamental arches, entablatures, curlicues and encrusted scrolls which adorned the best English-made machine were so much wasted effort. The more perceptive among overseas engineers did recognize what the Americans were doing. “In those American tools,” observed James Nasmyth, the famous Manchester machine-tool manufacturer, when he visited Samuel Colt’s pistol factory, “there is a common-sense way of going to the point at once, that I was quite struck with; there is great simplicity, almost a quaker-like rigidity of form given to the machinery; no ornamentation, no rubbing away of corners, or polishing; but the precise, accurate, and correct results.”
In that first, early-nineteenth-century American system of manufacturing, there was already discernible the concern for economy which would distinguish the late-twentieth-century form of the American system. That first American system aimed to make products just good enough for their purpose. The aesthetic-that-was-not-functional, the ornament that was merely traditional, had no place in the American scheme. For Americans, a high-quality machine was not one that was polished and ornamented, but one that worked.
From the point of view of the foreign observer it seemed that Americans had actually made a system for building products that were “imperfect.” To most English eyes, products that were “just good enough” for their job, that lacked the extra polish and ornament, fell below the highest standard. The late-twentieth-century American system went still further in outraging the Old World craftsman’s ideal. For while the nineteenth-century system aimed to manufacture products that were just good enough for their purpose, the twentieth century actually aimed to manufacture products that were no better than they needed to be. Perhaps, some Americans now suggested, the interchangeable system had itself become wasteful, requiring superfluous and costly precision. The most important twentieth-century American advance in the system of manufacturing would come from this explicit American acceptance of a still newer meaning of “imperfection.” And this was made possible by the rise of statistics.
THE NAME FOR this remarkable and little-celebrated American achievement was “Statistical Quality Control.” By the mid-twentieth century it had entered the engineering vernacular to describe a way of applying statistics to factory organization that translated quality into quantity. Although the basic theoretical concepts were borrowed from abroad, their productive new uses were substantially American.
The man most responsible for making statistics into a tool for factory organization was Walter A. Shewhart. In 1924, while he was working as a young engineer in the new inspection engineering department of the Western Electric Company (which made apparatus for the Bell Telephone Company), Shewhart puzzled over the factory inspection records. His assignment was to find ways to use inspection data to produce a uniform product more economically. Before Shewhart’s time, statistical sampling had been tried in factories abroad. But there the samples had been used to assess the quality of the finished product; in other words, to see how successful the production techniques had actually been. Shewhart offered the new idea that statistics could be used in production not only to discover how good the process had been but also to replan and control production while it was still going on. He transformed factory statistics from historical data into production tools, by making statistics an instrument of industrial prophecy.
Essential to Shewhart’s way of thinking was an elementary new concept that had been developed abroad in the late nineteenth and early twentieth centuries. By 1909 the word “tolerance” had entered the mechanic’s vocabulary with a new meaning—to describe “an allowable amount of variation in the dimensions of a machine or part.” The idea behind limit gauges and the new notion of tolerance in the world of machine production was, of course, very much the same as that behind the whole interchangeable system. For while the new American system of machine production had depended on a quest for precision, on the use of calipers, limit gauges and other measuring devices, on machine-made objects shaped more accurately than any machine-made objects before, it also depended on a new mechanics of compromise. There had to be new ways of identifying the product that was just good enough. Behind it all, as the historian Eugene Ferguson observes, was the acceptance of the idea of imperfectibility, the idea that if absoluteprecision was actually impossible, it was, anyway, superfluous. For the only prudent industrial objective was to achieve a precision that was sufficient. The old adage about striving for “perfection” had to be revised.
To aim vaguely at a “better” product, Shewhart explained, would not necessarily lead to a more satisfactory product at a lower price. The first requirement was to set specific, mathematically defined and limited goals for each part of the product. And now statistics and the mathematics of probability could be used to reach the prescribed goal at the least cost. Essential to Shewhart’s way of thinking was the idea of imperfection. He accepted the notion that “defects” were inevitable in the products of any system, and he insisted that before anything was produced, the maximum acceptable number of these defects had to be fixed in advance. Then the production could be organized so as to result in no more (but not necessarily less) than that number.
Shewhart’s brilliantly simple notion, then, was to make a system out of imperfectibility. It was based on the common-sense fact that while “excellence” must often be a matter merely of hunches and intuition, specific defects can be counted and measured. Just as the first American System a century before had managed to make the scarcity of skilled craftsmen the incentive to an interchangeable system, so this second American System aimed to make the very imperfections of the machine the basis for more economical production. Shewhart had the courage to give up the quixotic quest for craftsman’s “perfection,” which American manufacturers had been criticized for not producing. Instead he tried to make a virtue of American limitations. Now for the first time a nation could economize its resources by organizing factories to produce the allowable tolerances and “imperfections,” but no more. The new science of statistics had provided the techniques to make this possible.
Setting goals for production therefore first required the setting of “tolerance limits.” Obviously the tolerances for a barn-door hinge were wider than those for a watch spring. And then, of course, there was the question of cost. In setting tolerance limits, Shewhart observed, “it is not only what the engineer wants, but what he can get, or at least get economically, that must be taken into account.”
The pioneer treatise on quality control (apparently the first use in print of “quality control” in this sense) was by another American engineer, George S. Radford. In The Control of Quality in Manufacturing, Radford explained that until then, factory engineers had put “a disproportionate emphasis upon quantity of output, in the effort to effect economies.” He argued that “increased output and decreased costs are more certainly attained when manufacturing problems are approached with quality, instead of quantity, as the primary guide and objective.” But this paradox was only superficial, for Radford had used statistics to translate quality into quantity. His new system inspected “the flow of work in process.” Improved measuring instruments in the hands of impartial inspectors would identify products outside the tolerance limits.
The object was to attain minimum standards at minimum cost. “Granted that the ideal standard cannot be realized in practice because quality varies continually, practical manufacturing or working standards must be determined. These vary from the ideal standard by certain differences or allowed errors, and by adding them to the outline design or ideal standard, a complete design is obtained.” Repetition Manufacturing (in Radford’s phrase) depended on “the compromise in setting tolerances.” And this compromise was precisely what saved the manufacturer from costly uniformity-for-its-own-sake.
The thought of quality as something that is continually shifting and varying, when translated into form for use in the factory, gives rise, among other things, to the whole subject of tolerances and limits…. no design is sufficiently complete for intelligent manufacturing purposes unless the limits for each and every governing characteristic are known.
True manufacturing involves making a quantity of the same article, uniform within limits. In this respect it is the diametrical opposite of art work. The manufacturer seeks to make things alike, but the artist strives for the creation of things that are different and individualistic. The first system is far less costly; and therein lies the real value of manufacturing, because its product is thereby made more generally accessible to mankind. We make things alike because it is cheaper rather than for the sake of having them alike, although many secondary advantages accrue from this property of uniformity.
He gave the example of an inexpensive Ingersoll watch. Except for the crystal, the springs, and perhaps one or two minor parts, Radford explained, there was no special advantage, once the watch had been sold and was in use, in having any of the parts interchangeable. For it was not likely that any of the other parts would ever be replaced.
The prime purpose of an interchangeable system was “economical production.” And that meant not wasting resources to produce more precision (or a narrower limit of tolerance) than was needed for a satisfactory marketable watch. If Americans failed to take this new view, Radford warned, the interchangeable system with its temptation to emphasize a theoretical ideal standard for each part of a machine could become a fetish. Machine-made uniformity, “precision” for its own sake, could then be as unproductive as the hand-finishing standards of Old World craftsmen. That would get in the way of the social object: producing masses of usable objects inexpensively. Radford warned that there was such a thing as precision overkill. “When we generalize that it is best to make things uniform,” he concluded, “we must remember always that quality varies, and that what we really mean is likeness, uniformity, or standardization of quality within limits. This, in a word, is why quality requires control.”
OBVIOUSLY, THEN, there was a negative as well as a positive side to quality control. Since “quality” was a colloquial synonym for “excellence,” then to make the “control” of quality an objective of factory planning had some disturbing implications. For quality “control” actually aimed, as engineering historians have noted, to insure that any product was not one iota “better” than it needed to be. What must we think of a civilization that aims to make its products just barely as good as they need to be, and not one bit better? “She’s no better than she ought to be!” Was this the road to industrial progress, to the more democratic society, where the “quality” of objects was limited by the need to supply them to everybody?
Some engineers nostalgically lamented the rise of quality control as the decline of craftsmanship. “Much in modern design,” one of them complained, “serves its purpose, but it does not preserve the grade of civilization that we are used to.” Was there not a danger that products would be engineered to live down to the intentionally ambiguous claims of advertisers? Were these the perils of abandoning Old World aristocratic notions of quality to serve a New World democracy?
Quality control, shaped by Shewhart, Radford, and others, became a scheme for the more economical production of almost anything. Since quality control had to be statistical, they found ways to insure the “randomness” of samples; they designed new control charts and prescribed simplified ways of recording data. Although Shewhart had conceived his essential new ideas within a few days, he spent years elaborating and applying them. By the time his Economic Control of Quality of Manufactured Product appeared in 1931, the large design of his system was there, and fellow statisticians acknowledged that he had discovered momentous applications for their new science.
At the Bell Telephone Laboratories, which were responsible for improving the techniques for manufacturing telephone apparatus, Shewhart tried out his notions. Building on Shewhart’s work during World War II, the armed services further developed techniques of quality control for wartime purposes. “Sequential analysis” (or acceptance sampling) used an ingeniously simple new way of keeping cumulative statistics to economize the whole process of sampling. This made it possible to arrive at reliable conclusions with the smallest possible number of samples and was especially valuable where the sampling had to destroy the object being tested.
Although the consuming public remained unaware of these developments, among statisticians and engineers statistical quality control acquired the dignity of a new science. Shewhart himself taught the first college course in the subject at Stevens Institute of Technology in 1930, and formal training spread during the war, when it reached some ten thousand specialists. In 1946, seventeen local quality control societies joined to form the American Society for Quality Control. Publications on the subject multiplied. After Shewhart’s address on “The Future of Statistics in Mass Production” was published in 1939 in the Annals of Mathematical Statistics, numerous articles in professional journals applied Shewhart’s techniques to a vast range of activities: to verifying punch-card operations, to anticipating epidemics, to checking on overtime employment, and to designing laws to protect consumers.
Even after “SQC” had entered the everyday vocabulary of American industry, and governed the mass production of nearly everything that Americans bought, its esoteric mathematical vocabulary still obscured it from public grasp.
Yet there was no better symbol of the problems and paradoxes of a democratic America. Democracy required a new lexicon, a redefinition of what was excellent, of what was good, and—most significantly—of what was good enough. The American Standard of Living had depended on new ways of making large numbers of objects which, from the point of view of the consumer, were indistinguishable from one another. But the progress in the American system of production in the twentieth century would depend on scrutinizing and measuring and defining the allowable limits of differences. Radford, Shewhart, and their fellows aimed to save Americans from the fetish of perfectionism. Their democratic and equalizing purpose was to make as many things available to as many people as possible. In 1957 the standard textbook on quality control adopted Shewhart’s motto, “The object of control is to enable us to do what we want to do within economic limits.” And the author opened with Montaigne’s reassuringly democratic observation, “The most universal quality is diversity.”