W. Edwards Deming said of him:
Born in New Canton, Illinois to Anton and Esta Barney Shewhart, he attended the University of Illinois before being awarded his doctorate in physics from the University of California, Berkeley in 1917.
Bell Telephone’s engineers had been working to improve the reliability of their transmission systems. Because amplifiers and other equipment had to be buried underground, there was a business need to reduce the frequency of failures and repairs. When Dr. Shewhart joined the Western Electric Company Inspection Engineering Department at the Hawthorne Works in 1918, industrial quality was limited to inspecting finished products and removing defective items. That all changed on May 16, 1924. Dr. Shewhart's boss, George D Edwards, recalled: "Dr. Shewhart prepared a little memorandum only about a page in length. About a third of that page was given over to a simple diagram which we would all recognize today as a schematic control chart. That diagram, and the short text which preceded and followed it, set forth all of the essential principles and considerations which are involved in what we know today as process quality control. Shewhart's work pointed out the importance of reducing variation in a manufacturing process and the understanding that continual process-adjustment in reaction to non-conformance actually increased variation and degraded quality.
Shewhart framed the problem in terms of assignable-cause and chance-cause variation and introduced the control chart as a tool for distinguishing between the two. Shewhart stressed that bringing a production process into a state of statistical control, where there is only chance-cause variation, and keeping it in control, is necessary to predict future output and to manage a process economically. Dr. Shewhart created the basis for the control chart and the concept of a state of statistical control by carefully designed experiments. While Dr. Shewhart drew from pure mathematical statistical theories, he understood data from physical processes never produce a "normal distribution curve" (a Gaussian distribution, also commonly referred to as a "bell curve"). He discovered that observed variation in manufacturing data did not always behave the same way as data in nature (Brownian motion of particles). Dr. Shewhart concluded that while every process displays variation, some processes display controlled variation that is natural to the process, while others display uncontrolled variation that is not present in the process causal system at all times.
His work was summarized in his book Economic Control of Quality of Manufactured Product (1931).
Shewhart’s charts were adopted by the American Society for Testing and Materials (ASTM) in 1933 and advocated to improve production during World War II in American War Standards Z1.1-1941, Z1.2-1941 and Z1.3-1942.
From the late 1930s onwards, Shewhart's interests expanded out from industrial quality to wider concerns in science and statistical inference. The title of his second book Statistical Method from the Viewpoint of Quality Control (1939) asks the audacious question: What can statistical practice, and science in general, learn from the experience of industrial quality control?
Shewhart's approach to statistics was radically different from that of many of his contemporaries. He possessed a strong operationalist outlook, largely absorbed from the writings of pragmatist philosopher C. I. Lewis, and this influenced his statistical practice. In particular, he had read Lewis's Mind and the World Order many times. Though he lectured in England in 1932 under the sponsorship of Karl Pearson (another committed operationalist) his ideas attracted little enthusiasm within the English statistical tradition. The British Standards nominally based on his work, in fact, diverge on serious philosophical and methodological issues from his practice.
His more conventional work led him to formulate the statistical idea of tolerance intervals and to propose his data presentation rules, which are listed below:
Walter Shewhart visited India in 1947-48 under the sponsorship of P. C. Mahalanobis of the Indian Statistical Institute. Shewhart toured the country, held conferences and stimulated interest in statistical quality control among Indian industrialists.
He died at Troy Hills, New Jersey in 1967.
In 1938 his work came to the attention of physicists W. Edwards Deming and Raymond T. Birge. The two had been deeply intrigued by the issue of measurement error in science and had published a landmark paper in Reviews of Modern Physics in 1934. On reading of Shewhart's insights, they wrote to the journal to wholly recast their approach in the terms that Shewhart advocated.
The encounter began a long collaboration between Shewhart and Deming that involved work on productivity during World War II and Deming's championing of Shewhart's ideas in Japan from 1950 onwards. Deming developed some of Shewhart's methodological proposals around scientific inference and named his synthesis the Shewhart cycle.
As a man, he was gentle, genteel, never ruffled, never off his dignity. He knew disappointment and frustration, through failure of many writers in mathematical statistics to understand his point of view.
He was founding editor of the Wiley Series in Mathematical Statistics, a role that he maintained for twenty years, always championing freedom of speech and confident to publish views at variance with his own.
His honours included:
Progress in modifying our concept of control has been and will be comparatively slow. In the first place, it requires the application of certain modern physical concepts; and in the second place it requires the application of statistical methods which up to the present time have been for the most part left undisturbed in the journal in which they appeared.
1. All chance systems of causes are not alike in the sense that they enable us to predict the future in terms of the past.
2. Constant systems of chance causes do exist in nature.
3. Assignable causes of variation may be found and eliminated.
Based upon evidence such as already presented, it appears feasible to set up criteria by which to determine when assignable causes of variation in quality have been eliminated so that the product may then be considered to be controlled within limits. This state of control appears to be, in general, a kind of limit to which we may expect to go economically in finding and removing causes of variability without changing a major portion of the manufacturing process as, for example, would be involved in the substitution of new materials or designs.
The definition of random in terms of a physical operation is notoriously without effect on the mathematical operations of statistical theory because so far as these mathematical operations are concerned random is purely and simply an undefined term. The formal and abstract mathematical theory has an independent and sometimes lonely existence of its own. But when an undefined mathematical term such as random is given a definite operational meaning in physical terms, it takes on empirical and practical significance. Every mathematical theorem involving this mathematically undefined concept can then be given the following predictive form: If you do so and so, then such and such will happen.
Every sentence in order to have definite scientific meaning must be practically or at least theoretically verifiable as either true or false upon the basis of experimental measurements either practically or theoretically obtainable by carrying out a definite and previously specified operation in the future. The meaning of such a sentence is the method of its verification.
In other words, the fact that the criterion we happen to use has a fine ancestry of highbrow statistical theorems does not justify its use. Such justification must come from empirical evidence that it works.
Presentation of Data depends on the intended actions
Rule 1. Original data should be presented in a way that will preserve the evidence in the original data for all the predictions assumed to be useful.
Rule 2. Any summary of a distribution of numbers in terms of symmetric functions should not give an objective degree of belief in any one of the inferences or predictions to be made therefrom that would cause human action significantly different from what this action would be if the original distributions had been taken as evidence.