A METHOD OF RATING MANUFACTURED PRODUCT 359 



from the warehouse and the resulting rates will be spread out more 

 widely around than when the entire product is inspected. These 

 results are exemplified by the two diagrams in Fig. 4. 



The probability curves of Fig. 4 represent the basis for setting con- 

 trol limits. The area under the curve between any two limits divided 

 by the total area represents the probability that a single rate will 

 fall between these limits. For a probability of .99 we can say that if 

 the product is controlled at a level corresponding to the expected 

 rate (zero in the illustration) then the chances are 99 in 100 that the 

 current rate will fall within the limits thus established and only 1 in 

 100 that it will fall outside the limits. 



For any product the spread between the limits is governed by two 

 factors, the number of pieces inspected and the value of the above 

 probability. It is necessary therefore to make an arbitrary choice of 

 probability, a choice which will depend on the use to be made of the 

 rate. 



The control lines are used primarily to distinguish between those 

 variations which may be attributed to chance causes and those which 

 are more probably the result of some significant change in manufactur- 

 ing conditions, either production or inspection, and therefore worthy of 

 investigation. The criterion of the suitability of the limits chosen is 

 the percentage of cases falling outside of the limits which on investiga- 

 tion are found to have resulted from some significant departures from 

 current standards of performance. 



In setting limits for rates the manufacturer has one point of view 

 and the purchaser another. The manufacturer wishes to detect lack 

 of control as early as possible and is willing to follow up false scents 

 occasionally in his endeavor to prevent the persistence of costly ir- 

 regularities. The purchaser is more interested in major swings or 

 trends in quality, is not so much concerned with the use of limits for 

 actual control and hence does not desire to instigate fruitless investiga- 

 tions frequently. For many telephone products, experience has 

 indicated that a probability value between .90 and .95 is economical 

 for shop control work while higher values such as .99 or above are better 

 suited for quality reports issued for purposes of general information. 



Inasmuch as the rate measures overall quality as determined by a 

 number of different characteristics, its control feature relates particu- 

 larly to final or partial assemblies of product. This control work 

 should, of course, be preceded by control activities based on the same 

 principles applied to the process inspection data for each of the essen- 

 tial characteristics of the parts which make up the whole. 



