# A Six-Step Guide to Shewhart Control Charts Twelve customer-reported defects this month. Last month there were five, and the month before that, three. It looks like things are getting worse, and it might be time to worry. But without knowing the normal range of variation for your system, you can’t tell. The twelve might be a real signal that something has changed. It might just be an ordinary data point, well within the expected range for a system that varies by nature. Treating it as a crisis when it isn’t one costs time, energy, and credibility. Missing it when it is one costs more. Shewhart charts, also called process control charts, make that distinction visible. ![[Shewhart chart example.webp]] Source: [Wikipedia](https://en.wikipedia.org/wiki/Shewhart_individuals_control_chart) ## Building the Chart in Six Steps You need a time-series dataset to start, and fifteen to twenty data points are enough to work with. Cycle time, defect counts, deployment frequency, and incident rate are all candidates. Any metric that changes over time will do. 1. Plot the data as a time series, with time on the _x_-axis and values on the _y_-axis. 2. Calculate the average of all your data points and draw a horizontal line at that value. 3. Calculate the average moving range. Take the absolute difference between each consecutive pair of data points, add those differences up, and divide by the total number of data points minus one. 4. Multiply the average moving range by 2.667. This constant approximates three standard deviations, the natural spread of the data. 5. Add that number to your average and draw a horizontal line there. This is your upper control limit (UCL). 6. Subtract that number from your average and draw a horizontal line there. This is your lower control limit (LCL). Now you have a picture of your system’s natural variability. Points inside the limits are common cause variation, the normal fluctuation of a system going about its work. Points outside the limits are worth investigating. ## Common Cause vs. Special Cause Special cause variation means something specific pushed a result outside what your system normally produces. It shows up in three ways: - A data point that crosses either the upper or lower control limit. - Three or more consecutive data points near one of the limits, without crossing it. - Eight or more consecutive data points on the same side of the average. When you see any of these, something changed, and it’s worth finding out what. ## Don’t Recalibrate on a Schedule It’s tempting to recalculate the average and control limits on a regular schedule, quarterly, for example. Don’t. Once you do, the lines reflect whatever time period you happened to pick instead of the system, and the chart turns into noise. Recalibrate only when you have a real reason to believe the system has changed, such as a major process shift, a reorganization, or a new tool rollout. Let the data show you when that happened instead of imposing a calendar on it. Source: [Stop Reacting to Your Own Data: A Six-Step Guide to Shewhart Charts - IT Revolution](https://itrevolution.com/articles/stop-reacting-to-your-own-data-a-six-step-guide-to-shewhart-charts/)