SIX SIGMA FUNDAMENTALS - Software Design for Six-Sigma: A Roadmap for Excellence - page 157

and they ended up paying 4.5% higher price for three years. The majority of measures
in a service or process will focus on:
Speed
Cost
Quality
Efficiency as defined as the first-pass yield of a process step.
Effectiveness as defined as the rolled throughput yield of all process steps.
All of these can be made robust at a Six Sigma level by creating operational defini-
tions, defining the start and stop, and determining sound methodologies for assessing.
It should come as no surprise that “If you can't measure it, you can't improve it” is
a statement worth remembering and ensuring that adequate measurement sytems are
available throughout the project life cycle. Software is no exception.
Software measurement is a big subject, and in the next section, we barely touch
the surface. We have several objectives in this introduction. We need to provide some
guidelines that can be used to design and implement a process for measurement
that ties measurement to software DFSS project goals and objectives; defines mea-
surement consistently, clearly, and accurately; collects and analyzes data to measure
progress toward goals; and evolves and improves as the DFSS deployment process
matures.
Some examples of process assets related to measurement include organizational
databases and associated user documentation; cost models and associated user doc-
umentation; tools and methods for defining measures; and guidelines and criteria
for tailoring the software measurement process element. We discussed the software
CTQs or metrics and software measurement in Chapter 5.
7.7 PROCESS CAPABILITY AND SIX SIGMA PROCESS
PERFORMANCE
Process capability is when we measure a process's performance and compare it with
the customer's needs (specifications). Process performance may not be constant and
usually exhibits some form of variability. For example, we may have an Accounts
Payable (A/P) process that has measures accuracy and timeliness (same can be said
about CPU utilization, memory mangemnt metrics, etc.) For the first two months
of the quarter, the process has few errors and is timely, but at the quarter point, the
demand goes up and the A/P process exhibits more delays and errors.
If the process performance is measurable in real numbers (continous) rather than
pass or fail (discrete) categories, then the process variability can be modeled with a
normal distribution. The normal distribution usually is used because of its robustness
in modeling many real-world performance, random variables. The normal distribution
has two parameters quantifying the central tendency and variation. The center is the
average (mean) performance, and the degree of variation is expressed by the standard