Java Reference
In-Depth Information
a bad one is difficult. There is no mathematical formula that can be
used. A good clustering is one that “talks” to the marketers.
To get this profile information, we will use counts obtained from
the OMDB to compare profiles from psychographic attributes for
each obtained cluster, or we can use the profiles from buying patterns
with respect to each product line. These profiles will be presented to
and validated by the CMO. The marketing manager then uses this
information to select, for each product line, appropriate magazines or
TV channels.
Again, because the CMO expects this process to be done on a reg-
ular basis, especially because he would like to be able to compare
profiles easily for each cluster, HEW has decided to build a “tool” to
perform these operations. This tool will be referred as the Customer
Segmenter in this section.
The following are the step-by-step operations that should be sup-
ported by this tool:
HEW CMO commences the process of psychographic
The marketing manager uses the tool to start this process.
The tool selects the customers without psychographic
enhancement and extracts a file to be provided to the
service bureau with the customer identifier, name, and
HEW IT sends this extract to TSB for enhancement.
TSB uses proprietary algorithms to match the name and
addresses with their internal database and complement
each customer with demographic and psychographic
information. In this particular scenario, about 100 attribute
values are assigned to each customer. TSB sends back a file
with proper enhancement data to HEW.
HEW IT department loads the enhanced data as a specific
table in the OMDB.
HEW marketing manager then builds the clustering models.
In this scenario, HEW CMO estimates that this strategic
segmentation should generate between 5 and 10 clusters. So
the system will create different clustering models with
different maximum numbers of clusters. A table associating
each individual customer to cluster identifiers in the OMDB
will be generated. These clusters are kept in a specific table
Search WWH ::

Custom Search