Data Mining Software

Data mining refers to a specialized type of analytical software that looks for patterns and relationships
within the information contained in a comprehensive database, and it forms a class
unto itself that is not directly connected with the larger customer relations division of OSS
inasmuch as it does not involve flow-through procedures involving other software modules.
In the case of telecommunications networks, data mining could be used for a number of
purposes. The network operator may want to examine the demographic skew of certain service
offerings as indicated by the existing customer base and then adjust marketing strategy
accordingly. If, for instance, enterprises of a certain size or in a certain type of business tended
to be frequent users of conferencing services, the network operator may want to construct a
marketing campaign for acquainting similar types of subscribers with conferencing services on
the theory that they will be apt to purchase such services. Or, if a certain demographic grouping
tends to abuse flat-rate unlimited services, one may want to manage the network so as to
throttle bandwidth to such users and put provisions for doing so in place before the commencement
of service. Or, to cite yet another example, if a group defined by certain attributes
recorded in the database is subject to unusually heavy churn (telecom jargon for customer
turnover), one may decide to avoid active solicitation of individuals within that group or make
special efforts to determine the source of the churn.
Knowledge in the broadband access business is power, and data mining can provide network
operators with the same kind of in-depth information of user preferences amassed by
traditional market research companies. Even so, data mining has not been extensively used by
independent operators and has remained largely the province of large incumbents (though it
is by no means universal even there). Like all statistical techniques, data mining requires a reasonable
sample to produce accurate results, so in a small network it is of dubious utility.