Cluster analysis is used to segment markets by grouping potential buyers into segments based on similar characteristics. It involves two stages: hierarchical clustering identifies the number of segments, which variables distinguish them, and segment profiles. K-means clustering assigns data points to centroids, moving centroids to the average of assigned points, until convergence. Cross tabulation analyzes the relationship between variables like demographics and geography to understand segment profiles. Cluster analysis uses distance calculations between variables to group them into clusters represented graphically. It partitions data into mutually exclusive clusters based on population density or distance between members.
2. Meaning
• Cluster Analysis is used for segmentating a market.
• A Market is made up of potential buyers.
• A Segment is a group of buyers who have the propensity or
inclination to purchase the same kind of product for a firm within
a product category.
• This desire to purchase a particular kind of product is linked to
segment characteristic s. These characteristic could be
demographics, pschographic, geographic and Behavioural.
3. STP - Target Marketing
1. Market Segmentation : Grouping potential Buyers
2. Market Targeting : Selection of groups to service, Evaluate attractiveness
of each segment & select target markets.
3. Product Positioning : Creating superior location for product in targetted
consumers’ mind
4. • Segmentation is done using Cluster Analysis.
• Cluster Analysis is conducted in two stages -
1. Hierarchial Clustering - It is use to identify the no. of segments or
clusters. It is used to understand;
i) Which person belongs to which segment?
ii) The varaibles which differentiates/distinguishes the segments.
iii) Segment profiles/descriptions based on their dominant distinguiing
characteristics.
Cluster Analysis - Segmentation -1
5. Cluster Analysis - Segmentation -2
2. K- Means Clustering - K-means clustering distinguishes itself from
Hierarchical since it creates K random centroids scattered throughout the data.
The algorithm looks a little bit like.
• Initialize K random centroids.
• You could pick K random data points and make those your starting points.
• Otherwise, you pick K random values for each variable.
• For every data point, look at which centroid is nearest to it.
• Using some sort of measurement like Euclidean distance.
• Assign the data point to the nearest centroid.
• For every centroid, move the centroid to the average of the points assigned to
that centroid.
• Repeat the last three steps until the centroid assignment no longer changes.
• The algorithm is said to have “converged” once there are no more changes.
6. 3. Cross Tabulation -
• It enables us to understand the Demographics & Geographic profiles of each
segment.
• Cross tabulation is a method to quantitatively analyze the relationship
between multiple variables. Also known as contingency tables or cross tabs,
cross tabulation groups variables to understand the correlation between
different variables.
• An example of categorical data is the region of sales for a product. Typically,
region can be divided into categories such as geographic area (North, South,
Northeast, West, etc) or state (Andhra Pradesh, Rajasthan, Bihar, etc). The
important thing to remember about categorical data is that a categorical data
point cannot belong to more than one category.
Cluster Analysis - Segmentation -3
7. Graphical
Represntation
• Varaibles sharing the
same attributes are
clustered togethere to
create pairs based on the
neareness of each
varaible to other.
• New clusters are then
represented to indicate
the relationship of the
new variables.
• Cluster analysis is done
on the basis of distance
between one to another
cluster as seen nearby in
the graph.
12. Conclusion
• The classification procedures used in cluster analysis are based on either density of
population or distance between members.
• These methods can serve to generate a basis for the classification of large numbers
of dissimilar variables.
• The resulting classes are mutually exclusive where the objects are partitioned
clearly into sets, so the clusters will not overlap in graphic projections. In density- or
mode-seeking techniques, clusters are identified and formed by locating regions in
a graphical representation that contains concentrations of data points.
• The clustering procedure is usually based on the Euclidean distance