Travel Reference
In-Depth Information
achieve, Dolnicar and Leisch (2010) introduce the terms 'natural', 'reproducible' and 'constructive
clustering/segmentation':
• Natural segments are in line with the traditional conceptualization of market segmentation,
which is that groups of tourists exist and that the role of segmentation is to
identify
them.
• Reproducible segments result from a segmentation exercise which, if calculated multiple
times, leads to similar (not identical) segments. This indicates that there is some structure in
the data, but distinct natural segments do not exist.
• Finally, constructive clustering is the process which can be applied to data which neither
contains density clusters, nor any other data structure which would allow arriving at the
same result repeatedly. Instead, this approach implies that artifi cial segments are created in
line with management needs. At fi rst glance this appears like a sub-optimal outcome, but this
is not the case. Imagine a situation as illustrated in
Figure 15.2
:
tourists have been asked how
important the natural beauty of a tourist destination is and how important man-made
attractions are. As can be seen from the illustration, there are no distinct segments in this data.
Following the classical paradigm of market segmentation this would be the end of the
segmentation exercise (because no natural segments exist) and the logical consequence
would be to try mass marketing. But would that be the best option? Clearly not. A natural
heritage site which is of outstanding natural beauty but not permitted to develop any man-
made attractions would be best off developing offers for and marketing them to people in
the right bottom corner of
Figure 15.2
(shaded area); those who want natural beauty but do
not care about man-made attractions. The situation is the same at the other extreme: a theme
park would be inclined to focus their attention on people who are located in the top left
corner of
Figure 15.2
;
those to whom man-made attractions matter, but for whom natural
beauty of the tourist destination is not of importance.
In situations where only two pieces of information are used, as in
Figure 15.2
,
no segmentation
analysis is required, but when the number of pieces of information increases, the use of
segmentation algorithms is unavoidable. However, the basic principle does not change: if no
distinct segments exist, it still makes sense to divide the market into groups rather than mass
market and data analysts and managers have to work together to identify the groups of tourists
who best match what they as a provider are good at.
It should also be noted that the concept of market segmentation is independent of the size
of segments or the avenues by which they are communicated to. The current level of sophistica-
tion of online tools makes it possible to work with segments which contain only single indivi-
duals. Think of amazon.com. The moment a user stars shopping, amazon 'learns' about the
person's preference and, treating them as a segment of one, offers other products which may
also be of interest to this person. This offers great new opportunities for micro-marketing
segmentation. Such micro-segmentation, however, is not applicable to all problems. For
example, brand image campaigns cannot be individualized (how could you convince one
potential tourist that a destination is perfect for motorcycle groups and another that it is a
relaxing retreat for the retired?), so more conventional market segmentation approaches will
continue to be required.
Market segments can be identifi ed or created using a range of procedures and algorithms. In
terms of procedures, the two extreme options are referred to as a priori (Mazanec 2000) or
commonsense segmentation (Dolnicar 2003) and data-driven approaches (Dolnicar 2003), also
known as a posteriori (Mazanec 2000) or post-hoc segmentation (Myers and Tauber 1977); both
will be discussed in detail in the following sections.
