960 



THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1953 



surance that the units furnished from various segments of production 

 ^\^ll approximate the deviation pattern to be expected in random samples 

 from an appropriately controlled process. The fundamental objective is 

 to provide a parade of product-segments under the three-cell method 

 that ^vill continue to have distributions that meet the basic intent of the 

 distribution requirements as deliveries or replacements are made. At the 

 same time, it is desirable to make the three-cell method moderately more 

 restrictive than the control chart method and the batch method in order 

 to provide an incentive to attain a degree of control during production 

 that will permit sampling. The choice of three equal cells, packages of 5, 

 and the package distributions indicated in Fig. 6 were selected with 

 these things in mind. 



4.0 Operating Characteristics of Control Procedures 



In the preceding section a description was given of the control proced- 

 ures that are associated with specified distribution requirements. The 

 most important question to be answered is: ''What are the operating 

 characteristics of these procedures?" In other words, how well will these 

 procedures discriminate between quality that is good or bad, distribution 

 wise? How effectively will they assure realization of the objectives for 

 which they were set up? 



For the sake of simplicity the statistical models used to evaluate the 

 expected performance of the control methods are limited to Normal dis- 



ACCEPTABLE DISTRIBUTIONS IN PACKAGES OF FIVE 



Fig. 6 — Acceptable distributions of units in packages of 5, three-cell method. 



