Common and Special Cause

Every process varies. The central insight of statistical process control is that variation has two sources, and confusing them makes processes worse.

Common cause variation

The natural, inherent variation of a stable process, arising from many small influences acting together — material variation within specification, minor temperature changes, machine play, measurement variation. It is predictable within limits, it is a property of the system, and reducing it requires changing the system.

Special cause variation

Variation from an identifiable, assignable source that is not part of the normal process — a tool breaking, a different material lot, a setting changed, a new operator, a failed sensor. It is unpredictable, it is a signal that something has changed, and it is removed by finding and eliminating that specific cause.

The two mistakes

• Tampering — reacting to common cause variation as though it were special. Adjusting a stable process after every measurement that is not exactly on target increases variation, often substantially. This is the single most common and most damaging misuse of process data.
• Ignoring a signal — treating a genuine special cause as normal variation, and allowing a real change to persist undetected.

A process producing scattered results is not necessarily out of control, and a process producing consistent results is not necessarily capable. The two questions are separate.