Control Charts: What They Are and What They Tell You
A control chart — also called a Shewhart chart, after Walter Shewhart who developed the concept at Bell Labs in the 1920s — is a time-ordered plot of process data with three reference lines added: a centre line (the process mean), an upper control limit (UCL), and a lower control limit (LCL). The control limits are set at three standard deviations above and below the mean, which in a stable process means that about 99.7 percent of data points will fall between them naturally. The control chart is not primarily a chart of what the process is doing — it is a chart for telling the difference between two kinds of variation: the kind that is normal and expected (common cause), and the kind that signals something unusual has happened (special cause). Knowing the difference is the foundation of effective process management.
Common Cause vs. Special Cause Variation
Common cause variation is the natural, inherent variability in any process. It comes from many small, random factors that are always present: slight differences in raw materials, minor temperature fluctuations, small differences in how operators perform a task. Common cause variation is predictable; it stays within the control limits over time. If only common cause variation is present, the process is said to be "in statistical control."
Special cause variation is unusual variation from a specific, identifiable source. A machine component that has started to wear. A new batch of material from a different supplier. An operator who was not trained on the correct procedure. A power fluctuation. Special cause variation typically appears as a data point outside the control limits, or as an unusual pattern within the limits (a long run above or below the mean, a trend in one direction, alternating high-low points).
The critical implication: if only common cause variation is present, do not adjust the process. Adjusting a process that is behaving normally — because the last data point looks a little high, for example — introduces additional variation. This is called tampering and makes the process worse, not better.
If special cause variation appears, stop and investigate. A data point outside the control limits is a signal that something has changed. Investigate the cause before doing anything else — not after the run ends, not at the next scheduled quality meeting.
How to Read a Control Chart in Practice
Plot the data in time order. A control chart loses most of its value if data is plotted out of order or averaged before plotting. The time sequence is what allows you to detect trends, runs, and shifts.
Calculate control limits from the data, not from specification limits. Control limits are statistical — they are derived from the actual variation in the process. Specification limits are what the customer or design requires. These are different things. A process can be in statistical control but still producing out-of-specification product; conversely, a process can meet specifications every time but show signs of instability. Do not draw control limits at the spec limits.
Look for the standard out-of-control signals. The most common are: a single point outside the control limits; seven or more consecutive points on the same side of the centre line (a run); six consecutive points all moving in the same direction (a trend). Each is a signal worth investigating.
Recalculate control limits when the process genuinely changes. If you implement a process improvement and the control chart shows a sustained shift, the previous control limits no longer describe the process. Recalculate from the new data. Control limits that no longer reflect the current process behaviour will generate false signals.
XNM supports public-sector and capital-project clients in applying statistical process control and Lean Six Sigma methods to operational and administrative processes. Connect with XNM's strategic advisory team to discuss how control chart monitoring could help your organisation distinguish real problems from normal variation.