SIX SIGMA FUNDAMENTALS - Software Design for Six-Sigma: A Roadmap for Excellence - page 162

CTQs, or the outputs) 3 and their linkage to critical business levers as well as the
goal for improving the metrics. Business levers, for example, can consist of return on
invested capital, profit, customer satisfaction, and responsiveness.
The last step in this phase is to define the process boundaries and high-level inputs
and outputs using the SIPOC as a framework and to define the data collection plan.

7.8.2 Phase 2: Measure

The first step is to make sure that we have good measures of our Y's through validation
or measurement system analysis.
Next we verify that the metric is stable over time and then determine what our
baseline process capability is using the method discussed earlier. If the metric is
varying wildly over time, then we must first address the special causes creating
the instability before attempting to improve the process. Many times the result of
stabilizing the performance provides all of the improvement desired.
Lastly, in the Measure phase, we define all of the possible factors that affect the
performance and use qualitative methods of Pareto, cause-and-effect diagrams, cause-
and-effect matrices, failure modes and their effects, and detailed process mapping to
narrow down to the potential influential (significant) factors (denoted as the x's).

7.8.3 Phase 3: Analyze

In the Analyze phase, we first use graphical analysis to search out relationships
between the input factors (x's) and the outputs (Y's).
Next we follow this up with a suite of statistical analysis (Chapter 6) including
various forms of hypothesis testing, confidence intervals, or screening design of
experiments to determine the statistical and practical significance of the factors on
the project Y's. A factor may prove to be statistically significant; that is, with a certain
confidence, the effect is true and there is only a small chance it could have been by
mistake. The statistically significant factor is not always practical in that it may only
account for a small percentage of the effect on the Y's; in which case, controlling
this factor would not provide much improvement. The transfer function Y f(x) for

every Y measure usually represents the regression of several influential factors on the
project outputs. There may be more than one project metric (output), hence, the Y's.

=

7.8.4 Phase 4: Improve

In the Improve phase, we first identify potential solutions through team meetings and
brainstorming or through the use of TRIZ in product and service concepts, which are
covered in El-Haik and Roy (2005) and El-Haik and Mekki (2008). It is important at
this point to have completed a measurement system analysis on the key factors (x's)
and possibly to have performed some confirmation design of experiments.

3 See Chapter 5 for software metrics.