Page 195 - Six Sigma Demystified

Chapter 8 co n t r o l S tag e 175

Unfortunately, prevention is at least sometimes more costly to implement
than detection, although once the cost of failure is factored in, then the true
cost of a detection-based control system often exceeds the cost of prevention-
based control. Recall the costs associated with “hidden factories,” and it is clear
that control systems for a Six Sigma organization should be prevention-
oriented in almost all cases.
As a result of designed experiments, we often discover input variables that
drive the process output. When we apply control schemes to these variables,
we can prevent errors from occurring. For example, if the number of incoming
orders drives the cycle time for order shipment, then a statistically significant
increase in the number of incoming orders is a clue to increase the number of
personnel in shipping. In this way, the cycle time for shipping (the process
output variable) is unaffected because of the previous effort.
The methods for process control include statistical process control (SPC),
engineering process control (EPC), and operational procedures. Statistical pro-
cess control refers to the statistical tools that detect process instability. SPC is
used to monitor output or input variables so that any lack of stability is detected.
When the process is statistically capable of meeting the requirements, charac-
terized by a process capability index Cpk of 1.5 or better (as defined in Part 3),
a control chart of the process output will provide a means for detecting whether
the process output at the Six Sigma level exceeds the requirements. When run-
test rules are applied, or if the process is highly capable at or above the Six
Sigma level, the control chart also will serve as a prevention tool because it is
likely to detect process shifts before out-of-control conditions are
experienced.
While SPC is often applied to process output, it is much better to apply SPC

to the key input variables. This was a focus for the experimentation used in the
analyze and improve stages of DMAIC. Establishing and monitoring statistical
control of the process drivers that determine the process output serves as a
prevention-oriented method of controlling the process.
Engineering process control refers to automated devices designed to respond
to process variation by adjusting one or more process input variables. A simple
example of this approach is the thermostat found in most homes. When the air
temperature inside the house reaches a set level, the air conditioner (or heater)
turns on to respond to the undesirable condition and control the temperature
within a set tolerance (usually determined by the manufacturer). Although a
household thermostat has rather simple programmable controls, industrial pro-
cesses often use more sophisticated algorithms with multiple input parameters