In the 1920’s while working at
ATT Bell labs, Dr Walter Shewhart (mentor of W Edwards Deming) sought a way to
improve telephone transmission systems. Seeking to reduce variations and
failures, on May 16, 1924 Dr Shewhart wrote an internal memo introducing the
concept of the control chart [Wikipedia]. Today the control chart (AKA process
quality control chart or Shewhart chart) has become the standard statistical
tool for finding variation that lead to continual improvement. As with other
continual improvement tools, the control chart can be understood in a short
time, yet takes a lifetime to master as a truly powerful tool for statistical
and incremental improvements.
The control chart is a tool that
tells us whether a process or system is in a state of statistical control. The
state of statistical control is one in which a minimum or acceptable amount of
variation is acceptable while still meeting the desired output of the process. Shewhart
recognized that every process and system has a sort of natural “wobble” or
variation. Nothing follows a strict straight path without any variance or
exceptions. There are always
exceptions or variations in a process. The goal of process improvement then is
not to try to remove all variances or adhere to a perfect ideal, rather to
eliminate the variances that are outside of statistical control.
Reducing variances involves
plotting statistical data along a mean and identifying Upper and Lower Control
Limits that represent the acceptable level of variation within the process or
system. This plot of data done over time is known as a control chart. By
looking for data points that fall outside the acceptable or established limits
we can identify those situations where something unusual or out control took
place. Data points inside the established control limits are considered to be
part of the natural “ebb and flow” or “wobble” of a system. Time and effort
should be spent eliminating the causes (called “special causes” by Shewhart) of
the data points outside the established limits. Time and effort should not be
spent on the points inside the limits since these are the results of “normal
causes” or the natural variation of a system or process.
Perhaps one of the most important
points to remember about control charts and statistical control is that the
plot or results of the chart cannot be determined before actually conducting a
process and gathering data points. This means we cannot set an arbitrary target
and expect the system or process to conform to the target. We must first perform
the process, gather the data, plot it and see the natural mean and limits of
the current process. If we do not think the results are acceptable we must
change the process and gather new data and create a new chart to reflect the
new process. The old chart can no longer serve for an improved or changed
process. In other words control charts reflect
the process or system rather than directing or steering the process or system. This
concept was taken by Dr Deming as he went forth to talk about statistical
control in the form of the Plan-Do-Check-Act cycle. Checking (using control
charts) occurs after Doing of the process, not before.
Control charts are one of the
most important and powerful tools we have for Continual Service Improvement.
Spend some time learning about these tools and you will eap the benefit for
your organization.
Comments