Marketing Strategy and Competitive Positioning pdf ebook

CHAPTER 8 SEGMENTATION AND POSITIONING RESEARCH 
In fact, an analyst does not have to choose between these two, because they can be used 
in combination, where Ward’s method is used to form the initial number of clusters (say, 
seven), and the K-mean approach used to refine that seven-cluster solution by moving 
observations around. If desired, after finding the best seven-cluster solution, Ward’s method 
can then be re-engaged to find a six-cluster solution that is again optimised using K-means. 
This may seem a computationally cumbersome approach, but fortunately packages are 
available to allow this process to be used. Arguably, Ward’s method in conjunction with 
K-means is the best approach for forming cluster-based segments; the analyst has removed 
the necessity to sort among numerous cluster alternatives and is able to choose between the 
clustering programs that are available.
While there is plenty of advice available on which techniques to use, the determination 
of the most appropriate number of segments to select following analysis is very much more 
Favoured name
Method
Aliases
Hierarchical methods
Single linkage
An observation is joined to 
another if it has the lowest level 
of similarity with at least one 
member of that cluster
Minimum method, linkage 
analysis, nearest neighbour 
cluster analysis, connectiveness 
method
Complete linkage
An observation is joined to a 
cluster if it has a certain level 
of similarity with all current 
members of that cluster
Maximum method, rank 
order typal analysis, furthest 
neighbour cluster analysis, 
diameter method
Average linkage
Four similar measures that 
differ in the way they measure 
the location of the centre of the 
cluster from which its cluster 
membership is measured
Simple average linkage analysis, 
weighted average, centroid 
method, median method
Minimum variance
Methods that seek to form 
clusters that have minimum 
within-cluster variance once a 
new observation has joined it
Minimum-variance method, 
Ward’s method, error sum of 
squares method, H GROUP
Interactive partitioning
K-means
Starts with observation