Page 118 - Six Sigma Demystified

Chapter 5 m e a s u r e s tag e 99

tem. The senior sales staff initiating the project (one of whom served as the
sponsor and another as a team member) had identified several key errors or omis-
sions that would increase the campaign costs: e- mail address incorrect or missing,
license count incorrect or missing, and renewal date incorrect or missing. Refer-
ring to the detailed process map developed earlier in the measure stage, the team
confirmed that each of these was included in the process flow for the sales de-
partment’s activities. The team was able to identify where the data should reside
within the CRM system and use these historical data to analyze the errors. The
data analysis is shown in the next section.

Process baseline estimation

A process baseline provides an estimate of the current state of the process and
its ability to meet customer requirements. The baseline estimate typically is
made of the process metric used in operations and accepted within the organi-
zation as a reliable indicator of the process under study. The baseline will allow
the stakeholders to validate the costs associated with current process perfor-
mance (as calculated in the preceding section).
The context of the process baseline estimate must be clearly understood.
How much variation is there between the samples? Would additional samples
yield better, worse, or similar results? Do the samples provide a reliable esti-
mate of future samples?

enumerative Statistics

The classical statistics most of us have been exposed to are enumerative tech-
niques, used to compare samples randomly drawn from populations. Using
hypothesis- testing procedures, samples can be tested for the likelihood that
they came from a known population. Similarly, two samples can be compared
to gauge the likelihood that they came from the same population. The term
population simply refers to a group of data that meet a defined condition, such
as all customers purchasing a specific product. A key assumption is that the
samples are each representative of the population. A representative sample im-
plies that there is no bias in selection of the data: Each observation has an equal
chance of selection.
In a similar fashion, confidence intervals on point estimates may be con-
structed that will provide bounds (an upper and a lower bound) on the expected