Methods used to Segment Markets
Factor Analysis
Factor analysis is used to reduce the data when it is collected on a wide variety of attitude and needs-
based items. if many of those items measure similar constructs, then subsequent analysis can be mis-
leading because some data are overweighed and other underweight. Irrelevant variables are also
dropped. We analyze the interrelationship among a large number of variables(attitudes questionnaire
responses) and then represent them in common, underlying dimension(factors). The
objective is to reduce the data from a large number of correlated variables to a
much smaller set of independent underlying factors.
Forming Segments by Cluster Analysis: Measures of association
In order to form segments of cluster, we do the following:
- Define a measure of similarity (or dissimilarity-distance) between all pairs of elements-individuals, families, decision making units.
- Develop a method for assigning elements to clusters or groups.
The most common method of grouping data elements, is the cluster analysis method. Its a unique
techniques for establishing groupings within a complex system of data, such as data matrix used
in segmentation analysis. In doing a cluster analysis, we select variables and create a measure of
association between all pairs of items.
Choice of Variables
Variables containing similar values from all respondents does not provide a good basis for
distinguishing between respondents. Counter to this is including variables that strongly differ
between respondents but are not relevant for the purpose at hand will provide misleading results.
To avoid this trap we include a number of variables so that adding or subtracting a variable will
not significantly affect the result.
Establishing Measures of Similarity
Most cluster analysis require you to define a measure of similarity for each pair of respondents.
Similarity measure falls into two categories depending on the data type available. For scaled data
you use distance measure. For nominal data such yes/no, male/female you use matching type measures.
when the data type is mixed, automatic interaction detection (AID) may be better.
Clustering Methods
- Hierarchical methods, in which you build up or break down the data row by row
- Partitioning methods, in which you break the data into a pre-specified number of groups and then reallocate or swap data to improve some measure of effectiveness
The dendogram below is an example of the hierarchical method using distance measures to form
clusters.
Cross-classification analysis
Cross-classification is sometimes called contingency table analysis. Data are
classified into two or more categories or dimensions. Cross-classification is
still widely used as a segmentation tool. However it can be unwieldy if two or
more classification variables are used. Cross-tabulation is also not recommended
if interaction exist among the variables.
Choice based segmentation
This is an increasing common approach to segmentation. Its highly used in direct
marketing. In choice-based segmentation the analysis is performed at the level of
the individual, relating that individual's likelihood of purchase or response to
a proposed marketing program to variables that the firm has in its database, such
as past purchase behavior for similar products, and attitudes or psychographics.
When the analysis is done at the individual level the customer database is sorted
in decreasing order of expected customer profitability and probability of purchase.
Please see segmentation checklist for more reading
Back to Home page