ROBUST DESIGN FOR SOFTWARE DEVELOPMENT - Software Design for Six-Sigma: A Roadmap for Excellence - page 467

Repair &
Categorization

Experience

Repairs

Programming
Skills

Maintenance

Backgrounds False Positives

Development
Team
Formation

Error (failure,
defect and
fault) Collection

Suppressed
Errors

VOC Collection

Validated Errors

Functional
Requirements
(FRs) Mapping

Issues &
Errors

Code Size

LEGEND

Verification
Preparation

Development
Activity

Code
Architecturing

Errors
List

Testing &
Inspection
Planning

I/O

FIGURE 18.1 Software developmental activities are sources of variation.

cost (Taguchi, 1986), (Taguchi & Wu, 1986), (Taguchi et al., 1989), (Taguchi et al.,
1999), and (Nair, 1992).
Quality is measured by quantifying statistical variability through measures such
as standard deviation or mean square error. The main performance criterion is to
achieve an on-target performance metric on average while simultaneously minimizing
variability around this target. Robustness means that a software performs its intended
functions under all operating conditions (different causes of variations) throughout
its intended life. The undesirable and uncontrollable factors that cause a software
code under consideration to deviate from target value are called “noise factors.”
Noise factors adversely affect quality, and ignoring them will result in software
not optimized for conditions of use and possibly in failure. Eliminating noise factors
may be expensive (e.g., programming languages, programming skill levels, operating
systems bugs, etc.). Many sources of variation can contribute negatively to software
quality level. All developmental activities in a typical process similar to the one
depicted in Figure 18.1 can be considered rich sources of variation that will affect
the software product. Instead, the DFSS team seeks to reduce the effect of the noise
factors on performance by choosing design parameters and their settings that are
insensitive to the noise.
In software DFSS, robust design is a disciplined methodology that seeks to find
the best expression of a software design. “Best” is defined carefully to mean that
the design is the lowest cost solution to the specification, which itself is based
on the identified customer needs. Dr. Taguchi has included design quality as one
more dimension of product cost. High-quality software minimizes these costs by
performing consistently at targets specified by the customer. Taguchi's philosophy of