A METHOD OF RATING MANUFACTURED PRODUCT 357 



How low does a rate have to fall before lack of control is indicated? 

 Does a rate of —10 signify that something abnormal has happened? 

 The following discussion gives a method which can be used to detect 

 lack of control. 



First of all we must determine the value of rate to be expected, i.e. 

 establish a norm for expected quality. Past experience can usually be 

 used as a guide for this purpose. If the average quality during the 

 base period is considered satisfactory as an estimate of expected 

 quality under current conditions, then the expected rate is 0. If only 

 a portion of past data is judged suitable for this purpose, then the rate 

 figure corresponding to the selected data is the expected value. 



The method of establishing limits of expected variation for the rate 

 makes use of statistical methods which have been described elsewhere,* 

 but will be briefly reviewed. If the current rate deviates from the 

 expected rate by an amount which is greater than can be attributed to 

 chance, this will be taken as an indication of lack of control. 



Just how chance enters the discussion will perhaps be better under- 

 stood from the following. Each unit of product is the physical result 

 of fashioning and combining various materials by a large number of 

 manual and mechanical operations and processes. Every element in 

 the production process which contributes to the final detailed character 

 of a unit can be considered as a cause. Now the ideal state of affairs, 

 purely conceptual lo be sure but nevertheless one which is the goal in 

 all attempts to secure greater uniformity of quality, is one in which 

 each of the elemental causes or groups of causes (affecting a particular 

 trait of the product) functions continuously in the same manner to 

 produce a given elemental effect in the direction of defective quality. 

 Considering overall quality, one group of manufacturing causes is 

 responsible for one type of defect, another group for a second type, etc. 

 The aggregate of these many causes which cooperate to mould the 

 product may be considered as a system of causes. When the concept 

 of constancy-with-time is associated with all of the causes, the system 

 is spoken of as a "constant system of causes," ® i.e. one whose tendency 

 toward defective quality does not change with time. Product turned 

 out by such a system will be referred to as "uniform product." 



For product which is uniform in this sense the rates obtained week 



6 "Quality Control Charts," by W. A. Shewhart, Bell Sys. Tech. Jour., Vol. V, 

 pp. 593-603, October, 1926. 



^ "Application of Statistics as an Aid in Maintaining Quality of a Manufactured 

 Product," by W. A. Shewhart, Jour. Am. Stat. Ass'n, Vol. XX, pp. 546-548, De- 

 cember, 1925. It should be noted that the system of causes associated with the data 

 used in rating is all-inclusive, encompassing the causes which are responsible for in- 

 accuracies of measurement introduced by inspection as well as the manufacturing 

 causes which affect actual quality. 



