144 Quality Assurance
Given the answers to these questions, the team will develop an overall
plan and dene the requirements of the review process to make sure the
goals are being met. Typical questions may be
• What precisely did you notice?
• How did you feel?
• What (if anything) would you like to do differently?
Step 3. Define and Establish Process Controls
Once the rst two steps have been completed, then the owners of the process
should begin the determination of sampling, frequency, and calculating the
voice of the process (control limits). The notion of determining the appropri-
ate and applicable sampling and frequency is based on determining the cor-
rect control on critical and signicant process parameters as well as establish
dominant process factors that inuence process output.
The process of dening and establishing controls is very important and
depends on sampling and frequency, as well as the data involved in the
process. For sampling and frequency, the selection is based on the assump-
tion that both reect the current process. This means that the sample and
frequency are representative of the process pattern and a reection of the
population. As for the data, consideration must be given to what type of data
is used. There are two types:
1. Variable or measured: Examples are length, viscosity, weight, time,
temperature, pressure, dimension, and so on.
2. Attribute or countable: Examples are pass/fail, go/no go, present/
absent, yes/no, and so on.
The preferred choice of data is variable because there is a greater sensitiv-
ity of variability and there is greater exibility for analysis. With attribute,
both the sensitivity and exibility are missing. Depending on what data the
process generates or the process owner decides to choose, the selection of
control charts will be different. For example, with variable data, the X bar,
R chart, medium chart, individual moving range, standard deviation range
chart, CUSum, and other charts may be utilized. With attribute charts, the p,
u, np, c, and other charts may be used. The goal here is to remove all assign-
able causes, and develop a consistent and stable process so that capability
may be calculated. The calculated control limits will identify whether the
process is consistent and stable enough to pursue capability study. Only ran-
dom variation (common cause variation) is allowed for studying capability of
the process. Unless the process is normally distributed (all assignable causes
have been removed), capability cannot be calculated. If data is not normally
distributed, sometimes by transforming the data (via Wilk’s lambda [λ] or