The Purpose and Limits of SPC
What statistical process control is designed to achieve, where it adds value, and where it is the wrong tool.
13 min
Activities: 0 of 2
A method for deciding when to act
Statistical process control (SPC) is often described as "using control charts". That description is too narrow, and it leads to programmes that produce many charts and little improvement. At its core, SPC is a decision rule. It tells the people running a process when a change in the data is large enough to justify action, and when it is not. Every other feature — limits, subgroups, run rules, capability indices — serves that decision.
Seen this way, SPC has two jobs. The first is process monitoring: detecting special causes quickly, so they can be found while the evidence is fresh. The second is process understanding: building a history of how the process behaves, which becomes the basis for improvement and for honest statements about capability.
What SPC cannot do
Advanced practitioners must be clear about the limits of the method, because over-claiming is how SPC loses credibility.
- It does not make a process capable. A stable process can predictably make defects. Reducing common-cause variation needs engineering change.
- It does not find causes. A chart shows when something changed. People, process knowledge and investigation find what changed.
- It does not protect against rare, isolated defects. A sampled chart will miss a single cracked part in ten thousand. That needs error-proofing or 100% detection.
- It does not work on bad data. If the measurement system cannot see the variation, the chart plots noise.
Where SPC adds most value
Worked example 1: the right tool
A grinding cell produces bearing races. Diameter drifts as the wheel wears and as coolant temperature changes during the day. An average and range chart at the machine shows drift well before parts leave tolerance, and the operator can dress the wheel or adjust the offset. SPC fits: the cause develops over time and the person at the process can act.
Worked example 2: the wrong tool
An airbag inflator weld has a rare, catastrophic defect: about one in fifty thousand has incomplete fusion. A chart of weld current every hour will not detect this. The right controls are a weld monitoring system on every part and a destructive test regime. SPC on weld current still helps, as a process monitor, but it must not be presented as the protection against the rare defect.
The customer requirement
Which statement best describes what a control chart tells you?