Special Cause Variation: Detecting Patterns Over Time
In a previous blog post, we discussed the two types of variation: common cause and special cause. In order to take appropriate actions to reduce variation in our processes, we need to be able to differentiate between common cause and special cause. Reduced variation can lead to less rework, less waste, more predictability and control, and happier customers. As we discussed previously, common cause variation requires a holistic approach to process improvement (such as the DMAIC process – define, measure, analyze improve, control). However, special cause variation indicates a change to a process and must often be addressed immediately.
Plotting your data over time is one of the easiest ways to detect special cause variation. Before we look at plots of our data over time, keep these definitions in mind:
Average: center of the data, calculated by summing all data points and dividing by sample size
Median: another measure of center, the middle value in a data set
Control limits: visual reference lines that help us see significant deviations from average
The most common patterns that indicate special cause variation are shifts, trends, and outliers.
Example 1: Shift – 8 or more consecutive data points on one side of the median
Example 2: Trend – 6 or more consecutive points increasing or decreasing
Example 3: Outlier – 1 or more points that falls outside of the calculated control limits (note UCL = upper control limit and LCL = lower control limit)
Any one of the above examples signals that there is something going on that is different from the normal process and should be investigated. The patterns that you observed will help you focus your attention on a specific period of time so that you can ask impactful questions about potential changes to the process. For more information on identifying and investigating special cause variation, additional special cause signals, or creating time plots and control charts, contact us at email@example.com.