Page 36 - Six Sigma Demystified

Chapter 1 d e p loy m e n t s t r at e g y 17

perceptions and customer perceptions. To ensure thoroughness, conduct
a complete value- stream analysis. While site visits are a popular means of
collecting these data, they can be costly and may not be necessary. In-
deed, their usefulness certainly will be improved if surveys are conducted
beforehand. Critical- ncident surveys, described in the Six Sigma Hand-
i
book (Pyzdek and Keller, 2009), also can be a great source of customer
insight.

• Data mining sometimes is used for discovering opportunities but often is insuf-
ficient for conclusive decision making. Data mining involves the statistical
analysis of databases, either to understand the nature of a particular vari-
able (a directed analysis) or to search for patterns (an undirected analysis).
For example, customer data may be mined to look for buying patterns by
price and time of year. Because of the nature of this statistical analysis, it
is often wise to conduct designed experiments to verify the suspected
patterns before committing resources.

• Benchmarking, like data mining, can provide a wealth of ideas for defining
direction but often does not provide sufficient information for direct commit-
ment of resources. Benchmarking can be used to understand best prac-
tices and discover new methods. Often the information is readily
available from suppliers, books, magazines, and the Internet. Bench-
marking helps to define the potential for processes, especially those that
may represent a new direction, where no internal experience exists in
the organization. This information can be used to conduct pilot trials or
experiments that serve as valid data- collection strategies for decision
making.
• Process data is perhaps the most prolific and reliable source of data for deci-

sion making, given its relative ease of acquisition and low cost. Unfortunately,
process data often are analyzed incorrectly, which can lead to more pro-
cess degradation than improvement.

It’s not uncommon for management reports to use bar graphs or pie charts
to represent changes over time. Although bar graphs certainly are easy to inter-
pret, they may not really provide the necessary context for a decision.
In the bar graph shown in Figure 1.4, it would appear that the process error
rate has decreased in March. Apparently, the process change initiated in Febru-
ary was effective in preventing further increases in the failure rate, as observed
from January to February.