Popularised by Deming, control charts recognise that there is variation in every process, the secret to quality is knowing what it is. There are, in fact, two quite distinct types of variation:
Normal or random variation – that which is invariably present in the process when it is working correctly; this is the something you cannot control within the process as it is currently established and trying to do so will almost inevitably worsen quality. Deming’s funnel experiment shows how trying to chase the centre reading doubled the spread of results.
Special cause variation – that which results from an externally applied change.
In its most basic form, you take a process that’s running under control and, selecting your most appropriate measure of “success” (say, the length of a bar being cut at a saw) you calculate the mean and standard deviation (sigma) of 20-30 recent (if not the latest) consecutive readings. On a graph with your chosen parameter on the y-axis, and time on the x-axis, mark the mean, the plus 2 sigma value and the -2 sigma value. Start the graph by plotting the readings used in your calculation and then carry on adding new values as they are made. If all is well, most points will be scattered between the control limits (the +/- 2 sigma marks) – statistically only about 5% should lay outwith these. Conversely, if you notice a pattern developing (a series of consecutive points all in the upper or lower half, a rising or reducing trend, or too many outwith the control limits, then it is likely that something untoward has happened – special cause variation – and it’s time to investigate and, probably, fix.
This is only a brief description – I recommend digging out one of the many QA/QC textbooks around (e.g. Juran’s Quality Control Handbook) to get more detail.
However, I’d like to suggest an alternative use for control charts. Rather than using it to detect unwanted special cause variation (and ignore normal variation, of course) it can be used to recognise intended change. For example, consider the situation where you’re trying to introduce an improvement (say a reduction in energy use); a control chart will show you what the normal pattern is and allow a more confident recognition of improvement.
For example, consider difficulties setting environmental improvement targets, where there are variables beyond direct control. A control chart will allow the normal variation to be seen; it may also show an underlying cycle (such as using less power for heating in summer than winter) but will also be able show if an improvement initiative is successful or not. However, remember that improvements need to be appropriate and ultimately based on desirable outcomes and not local, short-term outputs.