Quality Control Charts' 



By W. A. SHEWHART 



IRRESPECTIVE of the care taken in defining the production pro- 

 cedure, the manufacturer reahzes that he cannot make all units of 

 a given kind of product identical. This is equivalent to assuming the 

 existence of non-assignable causes of variation in quality ^ of product. 

 Of course, random fluctuations in such factors as humidity, tempera- 

 ture, wear and tear of machinery and the psychological and physiolog- 

 cal conditions of those individuals engaged in carrying out the manu- 

 facturing procedure may give rise to some of these apparently uncon- 

 trollable variations. Knowing this, the manufacturer contents himself 

 with trying to produce a product which is uniform and controlled — 

 one which does not vary from one period to another by more than an 

 amount which may be accounted for by a system of chance or non- 

 assignable causes producing variations independent of time. 



To make clear the significance of the terms "assignable causes" and 

 "non-assignable causes," we may make use of the following illustra- 

 tion. Suppose a person were to fire one hundred rounds at a target. 

 We know what probably would happen — the individual would not hit 

 the bull's-eye every time. Possibly some of the shots would fall within 

 the first ring, others within the second ring, and, in general, the shots 

 would be distributed somewhat uniformly about the center of the tar- 

 get. We have a more or less definite picture of some of the possible 

 reasons why the individual would not hit the bull's-eye every time, 

 but we probably cannot assign the reasons or causes for his missing 

 the bull's-eye in any particular instance — the causes of missing are 

 non-assignable. Suppose, however, that the individual tended to 

 shoot to the right of the bull's-eye. Naturally we would conclude that 

 there was some discoverable cause for this general tendency, i.e., we 

 would feel that the observed effect could be assigned to some particular 

 cause. 



The reason for trying to find assignable causes is obvious — it is only 

 through the control of such factors that we are able to improve the 

 product without changing the whole manufacturing process. But it 

 would be a waste of time to try to ferret out or assign some cause for a 



^ A brief description of a newly developed form of control chart for detecting lack 

 of control of manufactured product. 



2 Quality is some function of those characteristics X, Y, Z . . ., required to define 

 a thing. For our present purpose we shall consider that quality is a function of a 

 single characteristic X. 



S93 



