Which change in the numbers deserves a search for a cause

On 16 May 1924, Walter Shewhart set down a proposal for interpreting manufacturing data at Western Electric’s Hawthorne works. The MacTutor history, drawing on George Edwards’s recollection, describes a memorandum of about 1 page, with roughly 1/3 of its surface occupied by a diagram. That left very little room for an argument. The problem was how to tell whether variation belonged to the ordinary operation of a process or called for a search for an assignable cause.

A factory produces differences as well as products. For Shewhart, the difficult question was what those differences meant: which could be treated as chance variation, and which suggested that something specific required investigation. His memorandum gave that question a visual form, bringing the act of recording a change closer to the judgment of what it meant.

The memorandum had a destination. Written for work at Hawthorne, it addressed someone who needed to judge what a changing set of measurements warranted, and its usefulness depended on carrying a distinction across the small distance between an engineer’s reasoning and a colleague’s understanding. That makes its brevity a matter of communication as much as space: the question, the comparison and the reason to investigate had to remain intelligible when the paper passed to someone else.

What the numbers required

An assignable cause meant a particular explanation to pursue; chance variation described the fluctuations that could belong to the process as it was running. Confusing them would change the work an engineer undertook, because treating routine movement as a fresh fault creates an investigation, while treating a meaningful departure as routine leaves a question unasked.

The diagram supplied a way to hold those alternatives together. A written account can identify a suspect change and then move straight into explaining it, with the conclusion already suggested by the order of the sentences. Shewhart’s chart made the comparison itself part of the argument: the reader could consider a change against the variation of the process before deciding what sort of explanation to seek.

The diagram could indicate a reason to investigate, while the explanation of an assignable cause would still require work beyond the page, a separation that kept the memorandum’s promise both useful and limited. The cause still had to be found.

The drawing made the decision visible before the cause was known.

Space given to the argument

The proportion remembered by Edwards matters because the illustration occupied a substantial part of an already compact document. With roughly 1/3 of the sheet devoted to the diagram, visual explanation was doing work that a longer passage of prose would otherwise have had to carry, while the remaining space kept the image connected to an argument.

Taking up less space would have been an empty achievement if the page had also removed the uncertainty that made the factory’s numbers difficult to interpret. The drawing let the uncertainty stay in view. Its compactness depended on preserving the relationship between a changing process and the evidence available to the engineer, a relationship that could be lost in a neat verbal account of what had happened.

Edwards remembered the physical arrangement as part of the intellectual event. That is a telling emphasis in a recollection: the space taken by the diagram remained worth describing alongside the question it addressed. The object had made an abstract distinction available in the ordinary form of a colleague’s memorandum.

The argument acquired more pages

The later publication record supplies a different scale for the same work. In the October 1926 Bell System Technical Journal, volume 5, number 4, Shewhart’s Quality Control Charts appeared on pages 593 to 603. The journal article is behind a paywall. The title placed the chart itself at the centre of the published subject.

Then came Economic Control of Quality of Manufactured Product in 1931. The book’s title put the economic problem beside the technical one: control had to mean something in the manufacture of a product. Read alongside the dated memorandum and journal article, it gives the early page a place in a developing argument, with room for a more sustained treatment after the compact initial statement.

Seen in that sequence, the memorandum is an opening move in the written life of the method, a point at which the author had given the central distinction a form that could circulate while more developed accounts still lay ahead. The dates keep the short document in proportion. The compact proposal, the journal paper and the book represent different demands on the same writer, who needed room both to state the problem clearly and to pursue the argument beyond its initial outline.

The name attached to the chart

In 2024, the Institute for Healthcare Improvement marked the 100th anniversary in a publication titled 100th Anniversary of the Shewhart Chart. Technical commemoration has a particular way of preserving a person: the surname can remain in use as part of an instrument’s name, while the occasion of its making recedes. The anniversary returns attention to the act of making it.

The later story of Edwards Deming and Japan follows Shewhart’s pupil into a different setting. Following the student takes the narrative towards the reception of an idea; following this memorandum keeps it with the author trying to make a manufacturing distinction intelligible. The difference lies in the work under examination.

The memorandum answered a narrow question at one factory: which movements in the daily figures were worth a search for a cause, and which were the ordinary noise of the process. The chart made that line visible on a single sheet, which is why it could be handed to someone who had to decide before the shift ended. The wider career of the control chart came later and from other hands. What the 1924 page settled was the question an engineer had that week.