Page 123 - Six Sigma Demystified

104 Six SigMa DemystifieD

underlying root cause of variation is defined, the economic benefit of the
improvement is questionable.
The benefit of an improvement on a special cause depends on the underlying
process condition at the root of the special- cause variation. For example,
out- of- control conditions on a control chart can occur when multiple process
streams are shown on the same chart, as will be shown in an example later in
this section. In cases such as this, the process itself is not necessarily out of
control; it is our improper use of the control chart that provides the inaccurate
estimate of control. When the products are charted properly (on separate con-
trol charts or using short- run standardization techniques), the focus of the Six
Sigma project can be directed properly and its financial benefit calculated.
When out- of- control conditions are truly due to sporadic, unpredictable root
causes, the financial benefit of improvement can be known only when the root
cause of the behavior has been identified in process terms. While historical
evidence of the occurrence of similar patterns of behavior may be justification
to investigate the process, once an underlying cause is determined, an analysis
needs to link the cause to the past behavior because this past behavior may be
due to other (unidentified) root causes. Nonetheless, if there is financial burden
from the special causes, it would tend to justify a proper investigation, such as
a designed experiment as part of a Six Sigma project, into the causes.
Statistically, we need to have a sufficient number of data observations
before we can calculate reliable estimates of the common- cause variation and
(to a lesser degree) the average. The statistical “constants” used to define
control- chart limits (such as shown in Appendix 6) are actually variables and
approach constants only when the number of subgroups is “large.” For a sub-
group size of 5, for instance, the d value, used to calculate the control limits,
2

approaches a constant at about 25 subgroups (Duncan, 1986). When a lim-
ited number of subgroups are available, short- run standardization techniques
may be useful.
To distinguish between special causes and common causes, there must be
enough subgroups to define the common- cause operating level of the process.
This implies that all types of common causes must be included in the data. For
example, if the control chart is developed over a short time frame, such as an
eight- hour period, then the data do not include all elements of common- cause
variation that are likely to be characteristic of the process. If control limits are
defined under these limited conditions, then it is likely out- of- control groups
will appear owing to the natural variation in one or more of the process
factors.