What is SPC?

Statistical Process Control (SPC) is an approach that helps to monitor and control the output of a process by assessing the stability of the process and the type of variation that is present. SPC charts can aid decisions about whether a process is ‘in-control’ and whether or not an adjustment is necessary.

The diagram on the right shows a typical SPC chart. As you can see, an SPC chart is very similar to a Time Series Plot, with the exception that it has control limits.

Each point on an SPC chart is tested against a number of statistical rules (the simplest of which checks if a point falls outside of the control limits). If a point fails any of the rules, it indicates that the process may be (statistically) out of control, and the result should be investigated further.

There are several different types of SPC chart available, depending on the type of data that is being analysed. Most of them have rather strange technical names or abbreviations, such as an I-MR chart, XBar-R chart, U chart and P chart.

A key point to remember when applying SPC is that control limits are very different from specification limits, and SPC charts should definitely not have specification limits on them! Specification limits define where the output of a process should be in order to produce an acceptable part or service. Control limits define only where the output of a process should be if it is statistically in control (stable). This confusion is the most common source of mistakes when applying SPC charts.

When should you use SPC?

There are two key uses of SPC charts, as follows:

1) For Historical Analysis: An SPC chart can be used to analyse the past performance of a process, in order to assess if it was statistically in control (stable) or not.

2) For Ongoing Control: An SPC chart can also be used to monitor the output of a process in real time, in order to detect (as soon as possible) if the process has become ‘out of control’ (i.e. if it has statistically changed in some way).

So, in summary, SPC charts:

• can be used to analyse if a process is statistically ‘in control’ (i.e. stable).
• incorporate control limits and statistical tests to analyse each process result.
• do not assess whether a process is capable or within specification.
• can be used to analyse historical performance or monitor an ongoing process in real time.