Page 60 - Design for Six Sigma a Roadmap for Product Development

Six Sigma and Lean Fundamentals 37

steering performance. For each of these, you can further break down
to the next level of details.

The list of requirements can be long, but not all requirements are
equal in customers’ eyes. We need to analyze and prioritize those
requirements. This step can be done by Kano analysis or QFD, which
is covered in detail in Chap. 4. The list of high-priority customer
requirements is often called characteristics critical-to-quality (CTQ).

2.4.2 Stage 2: Measuring
process performance
Measure is a very important step. This step involves trying to collect
data to evaluate the current performance level of the process, and pro-
vide information for analysis and improvement stages.
This stage usually includes the following steps:
1. Select what needs to be measured. Usually, we measure the following:
■ Input measures
■ Output measure. CTQs, surrogates of CTQs, or defect counts.
■ Data stratification. This means that together with the collection
of output measures Y, we need to collect corresponding informa-
tion about the variables which may have cause-and-effect rela-
tionship with Y, that is, X. If we do not know what X is, we may
collect other information that may relate to X, such as stratifica-
tion, region, time, and unit factors, and by analyzing the variation
in performance level at different stratification factors, we might
be able to locate the critical X which may influence Y.
2. Develop a data collection plan. We will determine such issues as
sampling frequency, who will perform the measurement, the format
of data collection form, and measurement instruments. In this step,
we need to pay attention to the
■ Type of data (discrete or continuous). There are two types of data:
discrete and continuous. Discrete measures are those that enable
one to sort items into distinct, separate, nonoverlapping cate-
gories. Examples include car model types and types of credit
cards. Continuous measures are applied for quantities that can be
measured on an infinitely divisible continuum or scale, such as
price, cost, and speed. Discrete data are usually easy to collect,
easy to interpret, but statistically, are not efficient, and more data
need to be collected in data analysis.
■ Sampling method.
3. Calculate the process sigma level. For continuous data, we could use
the methods in process capability calculation described in the last