Pyzdek, Thomas (1990, ASQC) Quality America Inc., Tucson, AZ
Statisticians and quality professionals have been taught, and have taught others, that classification of variation into two categories, variation from special causes and variation from common causes, is vital. The prevailing wisdom is that when variation is from special causes the source of the variation can be easily identified, and when variation is from common causes the source of the variation can not be identified easily or, more to the point, economically. In fact, some people even mistakenly believe that when variation is from a system of common causes it has no cause.The belief in common causes has become so established that it has become taboo to suggest that they don't actually exist. This paper discusses a bold new idea: there is really no such thing as a common cause of variation. The division of variation into these categories is utterly artificial. By maintaining this artificial distinction we often cripple problem solving and process improvement efforts. Continuous improvement is made much more difficult. The paper offers an alternative variation classification scheme that explicitly acknowledges the fact that all variation is caused. With the new system the two sources of variation would be variation from visible causes and variation from hidden causes. By recognizing that special causes always exist we open a new gateway for process improvement investigations.
The tool used to operationally define special cause variation is the Shewhart control chart. Shewhart control charts were designed for an era when Quality Control was enough. Quality Control implies maintaining the status-quo by identifying important departures from it. The operator is typically held responsible for identifying the source of the departure. Today we must go beyond Quality Control, continuous Quality Improvement is the only way a business can survive in today's environment. Since variability represents lower quality and added cost, we are compelled to refuse to accept any level of variation as acceptable. We must constantly search for new knowledge regarding the causes of variation in a process.
The paper suggests guidelines for assigning responsibility for identifying visible and hidden causes. It shows that, in general, different statistical and management tools are used for identifying the two different types of causes.