Market segmentation 271
All of these can, of course, be identified from
the transactional data held in the database for
each participating customer.
The second approach refers to the digging
around in databases in a relatively unstruc-
tured way with the aim of discovering links
between customer behaviour and almost any
variable that might potentially be useful. This
second approach is discovery driven: ‘identifying
and extracting hidden, previously unknown
information... (to) scour the data for patterns
which do not come naturally to the analysts’ set
of views or ideas’ (Antoniou, 1997). Data-
driven marketers have tried a variety of
unusual or unexpected areas in which to mine.
For example, some have examined consumers’
individual biorhythms and star signs (Mitchell
and Haggett, 1997) as predictors of their pur-
chasing patterns.
At the heart of many data mining packages
is CHAID, as shown above. CHAID can use
any variable within the data set, so the old rules
have been overturned in favour of almost any
combination of whatever personal and further
overlaid profiling details the organization has
on its databases.
It is interesting to compare this with the
traditional use of market reports in which
entire countries’ consumption profiles might be
based on samples of just 1000, and profiled in
terms of gender, age and occupation. Inferences
would have to be made as to why consumption
patterns are as identified; profile characteristics
per sewould not necessarily be causal.
By contrast, the new data-driven segmen-
tation and targeting approaches might be based
on details of several million customers, each
with hundreds of transaction records overFigure 10.4 GIS data fusion
(b)