Walter Shewhart (1891-1967) was an influential US physicist and statistician who devised the statistical control chart and laid the early foundations of statistical process control and modern quality improvement practice.
Walter Andrew Shewhart was born in 1891 in a small farm just north of New Canton (pop 400) and four miles from the Mississippi River in Western Illinois. His great grandfather was a farmer who had brought his young family over from Germany in 1850. Walter's father, Antone, was the first member of the family to be born in the US and he was persuaded by his schoolteacher to change his surname to Shewhart from the original German spelling of Schuchardt. By the time Walter was born, several members of the Schuchardt /Shewhart family had farms in the area which is now known as Shewhart Hollow. Walter graduated from New Canton Grade School and then Barry High School. Throughout his early life, Walter had helped out on the farm, but in an exciting era when many new technologies were emerging (e.g. aeroplanes, automobiles, radio, telephones, phonographs and motion pictures) he decided his interests lay in the ordered precision of experimental physics rather than the tough uncertain world of farming.
In 1910 his parents paid for him to enrol at the University of Illinois in Urbana to study for a degree in Physics. He was awarded his BA degree in 1913 and in 1914 he was awarded his MA. Having secured an academic posting at the University of California (UC) together with a Whiting Fellowship he proposed to his hometown sweetheart Edna Elizabeth Hart and they married in 1914 before moving to Berkeley. Over the next three years he taught and undertook research work at UC and then back at the University of Illinois before being awarded his doctorate in 1917. With his strengthened C.V., Shewhart accepted the post of Head of Physics at the progressive La Crosse Normal School in Wisconsin. The USA had just (April 1917) declared war on Germany and Shewhart found the school full of war talk and patriotic fervour.
Joining Western Electric
After only six months at La Crosse, Walter Shewhart left to join the Western Electric Engineering Department at
463 West St, New York (
to conduct research mainly for war purposes). Walter and his wife rented a small
apartment just south of Prospect Park in Brooklyn, as he began his 38 year career with the US telecoms giant.
At that time Western Electric was busy fulfilling $22m orders for communications systems (around $25bn in 2015 values)
from the US Army, Navy and Army Air Force. Most of this equipment was being manufactured at
Western Electric's massive Hawthorne Works in the Chicago suburb of Cicero. Just two months later, the US government
took Western Electric and its parent AT&T group into public ownership. In October 1918 Shewhart wrote to Professor
Slate (one of his old UC supervisors) revealing that his first research project at Western Electric had been on a
device to be used by aviators which had attracted an initial government order of 40,000.
But only a few weeks later everything at Western Electric changed; World War 1 had ended and the government abruptly cancelled the vast majority of its contracts at a cost to Western Electric of around $4.5m (over $5bn in 2015 values)
The First Control Chart
By mid-1919 Western Electric had returned to private control and the Hawthorne production lines were rapidly modified to satisfy the massive demand for its latest consumer product: the now-classic 51AL candlestick telephone. Meanwhile Shewhart's short paper to the National Academy of Sciences in 1922 illustrated how his interests were now evolving from theoretical physics, to the study of process data. Shewhart's first paper in the prestigious Bell System Technical Journal did not appear till 1924. It consisted of 44 pages of closely reasoned, statistical theory and made scant reference to any Western Electric products.
Shewhart's managers may well have been more interested in papers on other more commercial Bell Labs projects. In that same year of 1924, for example, these would have included the early sound engineering work of Englishman George Groves. As Shewhart developed his arguments with reference to alpha particle movement and human skull dimensions, future Oscar-winner Groves was working on the in-house AT&T documentary film called ‘Hawthorne’ about the Western Electric works.
Later in 1924 Shewhart distilled some of his ‘relatively complex underlying theory’ into a much more accessible format when he sent his boss George Edwards a one page memo. About a third of that page was given over to a simple hand drawn diagram which we would now recognise as a control chart.
Shewhart's 1924 diagram and the short text which accompanied it, set forth all of the essential
principles of modern process quality control. The diagram monitored a production process over a period of time
using the percentage of defective products as an index. In addition to the upper and lower tolerance limits
for the product, Shewhart introduced two wider limit lines at three standard deviations above and below the
arithmetic mean and proposed that if a value exceeded these values there was some sort of problem with the process.
To emphasise this Shewhart wrote
this point indicates trouble against a point above the upper
limit in his diagram.
Shewhart's diagram was to change his life and is remembered to this day, but for many years the control chart
concept which it launched "did not take at all," at the Hawthorne works.
So, although Shewhart's colleagues, Harold Dodge and Harry Romig agreed that
Quality can't be inspected into
a product or service; it must be built in, the more conventional sampling inspection
disciplines that they jointly developed were seen as more practical quality contributions at that time.
Hawthorne Quality Problems
In 1925 Shewhart, in a paper to the American Statistical Association, refined his thinking on the use of statistics as an aid to quality . This was also the year that AT&T decided to consolidate the Western Electric Engineering Department with their other New York-based research facilities, into a new company named Bell Labs (full name: Bell Telephone Laboratories, Inc). Bell Labs then gradually moved work out of New York City to a purpose-built site at Murray Hill in New Jersey. Walter Shewhart was one of the first staff to be relocated and he and wife Edna moved from their Brooklyn apartment to a newly built upscale home in Mountain Lakes, New Jersey.
Quality problems persisted at the Hawthorne Works and in 1926, Shewhart, Dodge and Romig were part of six man Bell Labs team sent to Chicago to investigate. The team was led by Shewhart's boss George Edwards and was joined by eleven members from the Hawthorne works including a young graduate called Joe Juran.
After more than eight years at Western Electric and the Bell Labs, Shewhart was now evolving his thoughts on how
processes could be monitored using statistical analysis and his control charts. The progress of his thinking can
be observed in the ten papers he published in the Bell System Technical Journal and elsewhere between
1926 and 1929.
However the views of the 21 year-old Joe Juran (later to become to become a world-renown authority on quality)
probably summed up the prevailing assessment of Shewhart within AT&T.
That control chart is a delightful
device and a sensitive detector of change (but) there is no need for that. While being impressed by
Shewhart's enormous grasp of statistical theory, Juran added,
(Shewhart) exhibited flashes of brilliance,
but was mainly impractical and unintelligible. He had never been in a factory before and with respect to his
understanding of factory operations his ignorance was complete.
As AT&T management pondered how best to make use of Shewhart's statistical expertise, they allowed Shewhart considerable latitude in maintaining his extensive network of academic and other links in the US and abroad. One of these contacts was Charles R Kunsman who had followed Shewhart from the University of California to AT&T and was now at the Department of Agriculture. In 1927, Kunsman introduced one of his junior staff, Ed Deming, to Shewhart. This introduction was to prove pivotal for both individuals.
Shewhart's Quality Control Book
In 1930 Shewhart taught the world's first college course on statistical quality control at the Steven's Institute of Technology in nearby Hoboken. All the while he was busy refining the manuscript for his 500 page book on quality control. The book was eventually published in 1931 with the title ‘Economic Quality Control of Manufactured Product’. In it Shewhart starts with some simple examples of probability and variation like:
- the chance that one will die in the next year; or
- the variation when one hand-writes the letter ‘a’ several times.
He uses these examples to make the distinction between variations due to what he terms chance causes and what he calls assignable causes.
Shewhart defines chance causes as essentially random within a given process that is ‘under control’. Assignable causes lead to larger or else more systematic variations that can be attributed to some specific external factor which causes the process to be ‘uncontrolled’. He then goes on to describe:
- Concepts and advantages of statistical control
- Ways of expressing quality of product
- Basis for quality control specification
- Sampling fluctuations
- Allowable variability
- Use of control charts
- Quality control in practice
Shewhart's consolidation, into a single coherent publication, of the various elements which he had been refining and reporting on for many years, proved revolutionary. This book, which remains in print to this day, quickly established him as the international authority on statistical quality control.
Shewhart visits London
‘Economic Quality Control of Manufactured Product’ probably created even more interest in Europe that it did in the US. Leading British statistician, Egon Pearson, who was visiting the US in 1931 invited Shewhart to give a series of lectures on the topic in UK. As a result, in 1932 Shewhart gave three lectures at University College, London under the auspices of the British Standards Institution (BSI). Shewhart's UK visit is credited with persuading the BSI to develop BS600 (The Application of Statistical Methods to Industrial Standardisation and Quality Control) and the Royal Statistical Society to set up their Industrial and Agricultural Research Section.
During his three month trip he found time to visit Berlin to meet with
industrial researchers. Shewhart's
German visit is credited with persuading the German Standards Committee (Deutsche Normenausschluss) to standardise
their statistical methodology.
In 1933 Shewhart's control charts were also adopted by the American Society for Testing Materials (ASTM).
The publication of his landmark book coupled with his high-profile European trip, elevated Shewhart to statistical celebrity status. In 1934 an example of one of the many approaches Shewhart now had to contend with was from Leslie E. Simon, then a Captain at the neighbouring Picatinny Army Ordnance Arsenal. Shewhart was happy to provide advice on statistical methodology at the depot and the collaboration continued for the next ten years. Simon went on to become a general and the inaugural winner of the ASQ's Shewhart medal for technical leadership in quality control.
The Impact of Clarence Lewis
The early 1930s was also the time that Shewhart discovered ‘Mind and the World Order’ the latest work of Clarence Lewis who had been a fellow lecturer in his time at the University of California.
This densely argued 446 page book considers the fundamental nature of human knowledge and how it is acquired and developed. It examines statistical methodology and scientific and technological development in such areas as quantum and probability theory from a philosophical perspective.
While others struggled to absorb the relevance of Shewhart's writings, Shewhart was struggling to comprehend the relevance of Lewis' thinking to his own ideas. There is little doubt that Shewhart was profoundly influenced. He later confessed to Deming that he had read Lewis' book 14 times before he began to understand it. This absorption of Lewis' alternative perspectives and the increased public demands on Shewhart's time may have been factors in the reduction of his academic output, which amounted to only three short papers between 1932 and 1937.
Ed Deming and ‘the Shewhart Cycle’
In 1936 Deming followed Shewhart's lead by taking time out of his main job (lecturing at the Graduate School of the US Department of Agriculture) to lecture at University College London and to meet various European statisticians. When he returned to the Department of Agriculture in Washington DC he interspersed his own lectures with those of invited guest speakers. In 1938 these included Walter Shewhart. Despite Deming's pleading Shewhart refused to informalise his presentations, preferring his normal measured, meticulous delivery. An example of Shewhart's occasionally rather ponderous style can be seen in this extract from an internal Bell Labs paper that he wrote in 1935.
In such a case the quality judge must supply an inspection specification which will insure the following two
things: (1) that a satisfactory amount of data or evidence will be accumulated upon which to render judgment as
to the nature of the quality of the unsampled portion of the lot, and (2) that an operation will be indicated to
determine whether or not it should be rejected whenever the degree of belief in the satisfactoriness of the
unsampled portion of the lot upon the basis of evidence thus accumulated is insufficient to justify the acceptance
of the lot.
In the event, as Deming had feared, much of what Shewhart said in his four lectures to the USDA students went
over their heads. However Shewhart did agree to let Deming edit his lecture notes to
clarify their meaning.
A year later these edited notes reappeared in a book entitled
Statistical Method from the Viewpoint
of Quality Control by Shewhart
with the editorial assistance of Deming. The book had four sections:
- Statistical Control
- Establishing Limits of Variability
- Presentation of the Results of Measurement
- Specification of Accuracy and Precision
A key extract from this book shows Deming's tighter editorial presentation:
It may be helpful to think of the three steps in the mass production process as steps in the scientific method.
In this sense, specification, production and inspection correspond respectively to hypothesising, carrying out
an experiment and testing the hypothesis. The three steps constitute a dynamic scientific process of acquiring
This simply explained concept was promoted by Deming as the ‘Shewhart Cycle’ which he later augmented to Plan‑Do‑Study‑Act (PDSA).
Control Charts at AT&T
But at the outbreak of World War 2, although Shewhart and Deming's ideas had widespread acceptance by the
US government and armed forces US industry still preferred post-production inspection to designing in quality control.
So while Shewhart discussed the statistical significance of observed
runs above and below average
runs up and down and Deming was able to talk about "The Shewhart
Method of Quality Control", inside AT&T you could walk through this plant (Hawthorne), the
seed bed of the quality revolution, without seeing any control charts. Quality was left to
the inspection department.
After the war in 1946 Shewhart and Deming helped found the American Society for Quality Control (now the American Society for Quality), but continued to find that US industry was slow to accept their ideas.
There were the occasional breakthroughs, as when in 1947
Bonnie Small, one of the quality staff
from the Hawthorne works, paid a visit to Shewhart at Murray Hill. It was during the
nationwide strike against AT&T and Shewhart was on management picket duty checking identity badges
... because he could keep a book open and read some while he did it.).
But he recognised a kindred spirit and was happy to speak to her. It was a dialogue that was to continue for many years.
Quality Control in Japan
It was in post-war Japan that a pivotal turnaround in the attitude of business towards Shewhart and Deming's ideas
on quality control first occurred. After Japan's surrender in 1945 General Douglas MacArthur, as Supreme Commander
for the Allied Powers, effectively took control of the Japanese government. MacArthur quickly assembled teams that
would help the country rebuild itself using American
best practice. As part of this, Deming went to Japan to
assist in the preparation of Japan's first post-war census.
Homer Sarasohn, an American electronics engineer, was in the ‘Civil Communications’ team which was tasked with restoring Japan's communications industry. To facilitate this, Sarasohn started a series of management seminars and in 1950 he invited Shewhart to come to Japan and address Japanese executives on quality issues. Unfortunately Shewhart was just recovering from a bad bout of ‘flu and declined the invitation. At Shewhart's suggestion Ed Deming returned to Japan as his replacement and received a more attentive hearing than he had been getting in the USA.
Western Electric Rules
In 1956, the year of Shewhart's retirement from Bell Labs, Bonnie Small, chaired a Western Electric committee that published the first edition of an in-house Statistical Quality Control Handbook. By incorporating Shewhart's ideas on the interpretation of control charts these 'Western Electric Rules’ helped shop floor staff to detect "out-of-control" or non-random conditions on control charts.
In her preface to the handbook, Bonnie Small stated, "The book is written in non-technical language, and no attempt has been made to write for the professional statistician or the mathematician. The techniques described are essentially those which have been used in all types of industry since their development during the 1920's by Dr Shewhart. Perhaps the most distinctive features of the Western Electric program are the emphasis on:
- Engineering and Operating applications rather than Inspection, and
- The control chart and particularly the process capability study, as the foundation of the entire program."
The ‘non-technical language' included the replacement of Shewhart's ‘chance cause’ and ‘assignable cause’, with ‘natural pattern’ and ‘unnatural pattern’. Small also played a large part in the reintroduction of control charts in AT&T during the 1950s. By the 1960s it was estimated that the company maintained around 5,000 control charts
Shewhart's large network of contacts in the fields of quality and statistics had kept him busy and ensured a steady demand for his services from professional bodies. His memberships and fellowships included: International Statistical Institute, Royal Statistical Society, Econometric Society, Royal Economic Society, New York Academy of Science, American Mathematical Society, Mathematical Association of America, American Physical Society, American Society for Testing Materials, Psychometric Society, Acoustical Society of America, Philosophy of Science Association, and Association for Symbolic Logic.
Specific awards and appointments included:
- 1937 President of Institute of Mathematical Statistics
- 1941-48 Served on advisory committee of the Princeton mathematics department
- 1942-49 Council member American Association for the Advancement of Science
- 1943-65 Inaugural editor of the Mathematical Statistics Series of John Wiley and Sons.
- 1944 President of Institute of Mathematical Statistics
- 1944-46 Served on the National Research Council
- 1945 President of the American Statistical Association
- 1946 Founding member of the American Society for Quality Control
- 1947 Invited by Indian Statistical Institute for three months of lecturing and technical talks
- 1948 Inaugural Shewhart medal awarded to L.E. Simon by ASQC.
- 1954 Honorary professor at Rutgers University
- 1954 Awarded the Holley Medal by the American Society of Mechanical Engineers
- 1954 Second lecturing visit to India as guest of Indian Statistical Institute. (It was during this visit that Walter Shewhart met the Indian Prime Minister, Jawaharlal Nehru.)
- 1956 Walter Shewhart retires from Bell Labs
- 1958 Bell System Technical Journal marks the event by publishing a paper entitled
Nature and Origin of Standards of Qualitythat Shewhart had originally written in 1935.
- 1962 Last visit to India to receive an honorary D.Sc.
Retirement & Legacy
After his retirement in 1956 Walter Shewhart continued to communicate with his wide network of professional contacts. He and wife Edna remained gracious hosts in their New Jersey home to the many colleagues, relatives and friends who visited them. To younger relatives they were Uncle Wawa and Auntie Teddy. Following a serious auto accident and subsequent stroke, Walter Shewhart died on 11 March 1967 aged 75.
Since Shewhart's death the influence of technology on our lives has grown to a remarkable extent. Computer controlled processes and real-time data capture are now often the norm. Our shopping habits and lifestyles are routinely monitored, as is much of the electrical equipment that we use. This in turn has provided a prodigious quantity of statistical data for analysis by specialised computer systems and software.
Walter Shewhart's concept of process control, so revolutionary in his own time, is now seen as an axiomatic necessity,
as is the practice of ‘designing-in’ quality. This has been particularly true since the 1980s when USA
industry at last accepted the quality message after it had first proved so successful in Japan. Ed Deming and Joe
Juran were two major beneficiaries of this curious technical re-importation process. Throughout the remainder of his
long life Deming remained a passionate proponent of the Shewhart gospel. In 1986 he wrote,
Dr Shewhart is best
known for the least of his contributions — control charts. ... Another half-century may pass before the full
spectrum of his contributions has been revealed in liberal education, science, and industry.
Walter Shewhart was certainly ahead of his time. During his lifetime he inspired many key figures and since his passing his reputation and relevance has continued to grow.
Appendix - Shewhart's Work Locations at Western Electric
Previous biographies have suggested that for the first seven years of his Western Electric career, Shewhart was based at Western Electric's Hawthorne Works in the Chicago suburbs. However there are several reasons to believe this was not the case:
- After his marriage, Walter Shewhart was clearly reluctant to travel anywhere without his wife Edna. She travelled with him on his trips to Europe and to India. There is no evidence that they ever had a house in Chicago.
- In 1918, after six months at Western Electric, Shewhart gave his home address as, 409 East 21st St, Brooklyn.
- The 1920 US Federal Census shows Walter and Edna Shewhart living at 173 Lenox Rd, Brooklyn.
- In 1926 the 21 year old Joe Juran (who had only worked there for two years) was assigned to pilot ‘outside expert’ Walter Shewhart around the Hawthorne works.
- In Juran's biography he states that (in 1926) Shewhart
had never really been in a factory beforeand
with respect to his understanding of factory operations his ignorance was complete.
It therefore seems much more likely that from 1918 to 1925, Shewhart was primarily based in the Western Electric laboratories at 463 West St, New York, relying on remotely collected data and only occasional visits to the Hawthorne works for his assessments of production processes.