1. Random variation does not have a precise or recognizable origin. All the minor slack in a machine, small variations in raw material, variation due to vibration, and so forth amounts to background "noise". Because it comes from a variety of sources, random variation is stable and inevitable. If an operator adjusts a process in response to random variation the operator will be continually chasing a problem which vanishes and reappears. Consistency will actually suffer because of a new source of variation— the operator's corrections. It is not worthwhile to respond to this variation with an adjustment of the process. Since this variation is composed of numerous small effects, some one way, some another, they all add together to a pattern clustered near their average and less often at an extreme. This is the pattern of the normal distribution.

2. Assignable variation is attributable to a cause and is often not stable. Measurements changing due to a different operator, tool wear, or changes in materials cause nonrandom variation Assignable variation, in contrast to random variation, can and must be prevented to avoid loss.

Control charts help to separate random variation from assignable variation. Control charts help to identify the cause of nonrandom variation. Control Charts provide a way of detecting assignable causes, and provide a tool for identifying and eliminating variation in order to produce a more consistent product.

Consider a chart which tracks means of samples. If deviations from the average follow a normal distribution, then the process is most likely being affected by random events. The process is then said to be in statistical control. This is different than the usual way we think of "in control". Processes are usually under some sort of control, but they may not be in statistical control. In statistical control means the variation is random, and we can then expect the data we collect to follow a predictable distribution.

The next section shows an x-bar R control chart, a type of chart for tracking variable data. Other charts are appropriate for tracking attribute data.