Control Chart Explained: How to Make, Read and Report It
A control chart separates ordinary variation from signals that a process has changed. It plots measurements in time order against a centre line and control limits, so you can decide whether to investigate rather than react to noise.
What is a control chart?
Developed by Walter Shewhart, a control chart distinguishes common-cause variation (inherent to the process) from special-cause variation (a specific disturbance). Limits are typically set at the centre line ± 3 standard errors of the plotted statistic. They are not specification limits and are not tolerance limits set by customers.
Choosing the chart type
Continuous data in single observations: individuals and moving range (I-MR). Continuous data in subgroups: X-bar and R or S. Counts of defective units: p or np chart. Counts of defects per unit: c or u chart. Choosing the wrong type gives wrong limits.
Calculation: I-MR chart
MRi = |xi − xi−1| UCL/LCL = x̄ ± 2.66·MR̄ UCLMR = 3.267·MR̄
The constant 2.66 equals 3/d2 with d2 = 1.128 for moving ranges of span 2.
Original example (illustrative)
Ten consecutive bottle fill weights (g): 500.2, 499.8, 500.5, 499.6, 500.1, 500.4, 499.9, 500.3, 499.7, 500.0.
Mean x̄ = 5000.5 / 10 = 500.05. The nine moving ranges are 0.4, 0.7, 0.9, 0.5, 0.3, 0.5, 0.4, 0.6, 0.3 (sum 4.6), so MR̄ = 0.511. Limits: 500.05 ± 2.66 × 0.511 = 500.05 ± 1.36, giving LCL = 498.69 and UCL = 501.41. The MR upper limit is 3.267 × 0.511 = 1.67. All points lie inside the limits and the largest moving range (0.9) is below 1.67, so there is no signal of special-cause variation.
Ten points are used only for illustration. In practice, establish limits from roughly 20 to 25 or more in-control observations.
How to make it in StatFAS
Open the Control Chart Generator. Enter measurements in time order in one column (and subgroup information if needed).
Reading and reporting
Signals include a point beyond a limit, a long run on one side of the centre line, or a trend. Rule sets (such as the Western Electric rules) increase sensitivity but also raise false alarms. Report chart type, number of observations, centre line, limits and any signals: An I-MR chart of fill weight (n = 10) showed no points outside limits (LCL = 498.69 g, UCL = 501.41 g).
Common mistakes
Calculating limits from the overall standard deviation, using specification limits instead, recomputing limits after every point, ignoring autocorrelation, assuming a control chart tests normality, and reacting to every point inside the limits. A process can be in control and still not meet specifications; capability analysis is a separate step. See descriptive statistics to check the data first.
Conclusion
Use a control chart to decide when to intervene and when to leave a stable process alone.
1.
statfas-control-chart-interface.png, alt: "StatFAS Control Chart Generator interface"2.
statfas-control-chart-example-data.png, alt: "Example data entered in the StatFAS Control Chart Generator"3.
statfas-control-chart-settings.png, alt: "Control Chart Generator settings in StatFAS"4.
statfas-control-chart-result.png, alt: "Control Chart Generator result generated in StatFAS"References
- Shewhart, W. A. (1931). Economic control of quality of manufactured product. Van Nostrand.
- Montgomery, D. C. Introduction to Statistical Quality Control (Wiley). Add edition and year.