Surfstat.australia: an online text in introductory Statistics
STATISTICAL CONTROL CHARTS
Common and Special Causes of Variation
In the 1920's Walter Shewhart developed the idea of the control chart
to help decide when the output of a process was part of "a stable
system of chance causes", or whether there was an "assignable
cause". Shewhart viewed a stable system as one whose variation
arose as the result of many small perturbations (which we call
common cause variation); for a stable process the observations
could be described by a probability distribution - the system is said
to be "in a state of statistical control", or simply, in
control. An unusually large deviation suggested that the system had
been disturbed and hence there was an assignable cause for the
disturbance - the system is out of control, or unstable.
W. Edwards Deming
Deming substituted the term special cause for assignable cause.
Deming said that uncovering special causes was the
responsibility of the local work force (those who had day-to-day
contact with the process). Common causes were part of the system. The
system is the responsibility of management. If the common cause
variation is too large, it is the responsibility of management to
change the system. Deming, stated that 85% of the problems with
processes were system problems; later he increased this to over 94%,
based on his own experience.
Some authors regard this sharp delineation between special causes
and common causes, workforce responsibility and management
responsibility, as overly simplistic. For example, when a special cause
is signalled, and its cause found (rarely an easy task), the local
workforce may not have the authority to fix up the problem.
Nevertheless, the distinction between special and common causes of
variability is a useful one, and the recognition of responsibility
assignments to the workforce for sporadic problems and to management
for system problems is generally sound.
|Here are some examples of common and special causes of variation.
- Inappropriate procedures.
- Poor design.
- Poor maintenance of machines.
- Lack of clearly defined standard operating procedures.
- Poor working conditions,
e.g. lighting, noise, dirt, temperature, ventilation.
- Machines not suited to the job.
- Substandard raw materials.
- Measurement error.
- Vibration in industrial processes.
- Ambient temperature and humidity.
- Insufficient training.
- Normal wear and tear.
- Variability in settings.
- Computer response time.
- Operator absent.
- Poor adjustment of equipment.
- Operator falls asleep.
- Faulty controllers.
- Machine malfunction.
- Computer crashes.
- Poor batch of raw material.
- Power surges.