Dr. Yonatan Reshef
School of Business
University of Alberta
The following is an excerpt from:
VARIATION, MANAGEMENT AND W. EDWARDS
DEMING
Brian L. Joiner Marie A. Gaudard
The article was first published in Quality Progress, December 1990, pp.
29-37.
Variation is not a new concept. Statisticians and scientists have studied it for decades. What's new is that their awareness of variation and how it affects everyday activities is infiltrating the workplace. There are seven concepts about variation that everyone should know:
1. All variation is caused. There are specific reasons why your weight fluctuates every day, why sales go up, and why Maria performs better than Robert.
2. There are four main types of causes:
Common causes are the myriad of ever-present factors (e.g., process inputs or conditions) that contribute in varying degrees to relatively small, apparently random shifts in outcomes day after day, week after week, month after month. The collective effect of all common causes is often referred to as system variation because it defines the amount of variation inherent in the system.
Special causes are factors that sporadically induce variation over and above that inherent in the system. Frequently, special cause variation appears as an extreme point or some specific, identifiable pattern in data. Special causes are often referred to as assignable causes because the variation they produce can be tracked down and assigned to an identifiable source. (In contrast, it is usually difficult, if not impossible, to link common cause variation to any particular source.)
Tampering is additional variation caused by unnecessary adjustments made in an attempt to compensate for common cause variation.
Structural variation is regular, systematic changes in output. Typical examples include seasonal patterns and long-term trends.
3. Distinguishing between the four types of causes is critical because the appropriate managerial actions are quite different for each. Without this distinction, management will never be able to tell real improvement from mere adjustment of the process or tampering. In practice, the most important difference to grasp first is the difference between special cause variation and common cause variation.
4. The strategy for special causes is simple: get timely data. Investigate immediately when the data signal a special cause was present. Find out what was different or special about that point. Seek to prevent bad causes from recurring. Seek to keep good causes happening.
5. The strategy for improving a common cause system is more subtle. In a common cause situation, all the data are relevant, not just the most recent or offending figure. If you have data each month for the past two years, you will need to look at all 24 of these points.
In-depth knowledge of the process or system being improved is absolutely essential when only common causes are present. This knowledge can come from basic statistical tools, such as flowcharts, cause-and-effect diagrams, stratification analysis (used for measurement data such as process cycle time), and Pareto analysis (used for count data such as number of accidents). These and other tools can help identify fundamental changes to the system, but they should be tried on a small scale first to see whether results improve. Statistically designed experiments might also be helpful in identifying system innovations.
6. When all variation in a system is due to common causes, the result is a stable system said to be in statistical control. The practical value of having a stable system is that the process output is predictable within a range or band. For example, if a stable order entry system handles 30 to 60 orders a day, it will rarely slip to fewer than 30 or rise to more than 60.
If some variation is due to special causes, the system is said to be unstable since you cannot predict when the next special cause will strike and, therefore, cannot predict the range of variation. If the order entry system just described were unstable and subject to special cause variation, its capability might sporadically (and unpredictably) drop sharply below or rise sharply above the 30 to 60 range.
7. How much system variation is present can be determined by performing statistical calculations on process data. Thus control limits can be set. Control limits describe the range of variation that is to be expected in the process due to the aggregate effect of the common causes. Calculating these limits lets managers predict the future performance of a process with some confidence.