ECONOMIC QUALITY CONTROL OF PRODUCT 



377 



ence. About thirty typical items used in the telephone plant and pro- 

 duced in lots running into the millions per year were made the basis for 

 this study. As shown in this figure during 1923-24, these items showed 

 68 per cent control about a relatively low average of 1.4 per cent defec- 

 tive.^ However, as the assignable causes indicated by deviations in 

 the observed monthly fraction defective falling outside of control 

 limits were found and eliminated, the quality of product approached 

 the state of control as indicated by an increase of from 68 per cent to 

 84 per cent control by the latter part of 1926. At the same time the 

 quality improved; in 1923-24 the average per cent defective was 1.4 

 per cent whereas by 1926 this had been reduced to .8 per cent. Here 

 we get some typical evidence that, in general, as the assignable causes 

 are removed, the variations tend to fall more nearly within the limits as 

 indicated by an increase from 68 per cent to 84 per cent. Such evi- 

 dence is, of course, one sided. It shows that when points fall outside 

 the limits, experience indicates that we can find assignable causes, but 

 it does not indicate that when points fall within such limits, we cannot 

 find causes of variability. However, this kind of evidence is provided 

 by the following two typical illustrations. 



TABLE 2 



Electrical Resistance of Insulations in Megohms, 

 Should Such Variations be Left to Chance? 



B. In the production of a certain kind of equipment, considerable 

 cost was involved in securing the necessary electrical insulation by 

 means of materials previously used for that purpose. A research pro- 

 gram was started to secure a cheaper material. After a long series of 

 preliminary experiments, a tentative substitute was chosen and an 



1 Jones, R. L., "Quality of Telephone Materials," Bell Telephone Quarterly, June, 

 1927. 



