Sei sulla pagina 1di 14

Market Segmentation &

Targeting
Cluster Analysis & Discriminant
Analysis

Segmentation Many Uses


Segmenting the market
benefit segmentation -- forming segments of consumers
that are relatively homogeneous in terms of benefits
sought
Selecting test markets
By grouping cities into homogeneous groups, it is possible
to select comparable cities to test various marketing
strategies
Identifying new product opportunities
competitive sets -- clustering brands competing more
fiercely with each otherEmerging needs (Opportunityfocused
segmentation)
Salesforce allocation/call planning
Emerging needs (Opportunity-focused

segmentation)
2

Segmentation, Targeting, &


Positioning
To identify and select groups of potential buyers
(organizations, buying centres, individuals)
Whose needs within-groups are similar and
between-groups are different
Who can be reached profitably
With a focused marketing program

Segmentation
Target segments may not be clearly
defined and reachable
In practice, segments may be hard
to define, fuzzy, and overlapping
Buyers can be classified into one or
more segments
Segmentation is not a static
classification but a process to
support business decisions
4

Find Value-Based Segments


Evaluate competencies vs. attractiveness

Competency in Segment

The Market
Vs.
Segments

High

Average

G
Low

I
Low

Average

High

Segment Attractiveness

Segments
Bases
Characteristics that tell us why segments differ
(needs, preferences, decision processes)
Descriptors
Characteristics that tell us how to find and reach
Business

Consumer

Industry
Size
Location
Organizational Structure

Age/Income
Education
Profession
Lifestyles
Media habits
6

What is Cluster Analysis ?


Objective of Cluster Analysis is
to separate objects (usually consumers) into
groups such that
each object is more alike other objects in its
groups than objects outside the group
Cluster Analysis assumes that
the underlying structure of the data involves an
unordered set of discrete classes;
these classes can be hierarchical in nature,
where some classes are divided into subclasses;
we do NOT use prior information to partition the
objects into groups;
we only assume that the data are partially
heterogeneous i.e. that clusters exist

Doing Cluster Analysis


a = distance from member
to cluster center
b = distance from I to III

Dimension
2

Perceptions or ratings data


from one respondent

III
b

II

Dimension 1

Market Structure Analysis using


Hierarchical Clustering
1 cluster

2 clusters

3 clusters

5 clusters

Maruti
Swift

Hyundai
Santro

Maruti
SX4

Honda Hyundai
City
Verna
9

Procedure - Cluster Analysis


Problem Formulation

Step 1

Select a Distance Measure

Step 2

Select a Clustering Procedure

Step 3

Decide on the Number of Clusters

Step 4

Interpret and Profile Clusters

Step 5
10

Distance or Similarity Measure

Euclidean Distance
City Block Distance
Correlation

11

Clustering Procedures
Hierarchical Clustering: A clustering procedure characterized by the
development of a hierarchy or treelike structure
Agglomerative Clustering -- each object starts out in a separate
cluster; clusters are formed by grouping objects into bigger and
bigger clusters
Divisive Clustering -- all objects start out in one group; clusters
are formed by dividing this cluster into smaller and smaller clusters
Non Hierarchical Clustering: Number of clusters are prespecified;
clusters built around cluster centres

Agglomerative Clustering Methods:


Linkage Methods -- Clusters objects based on computation of
the distance between them
Variance Methods -- Clusters are generated to minimize withincluster variance
Centroid Methods -- A method of hierarchical clustering in
which the distance between two clusters is the distance
between their centroids
12

Wards Minimum Variance


Agglomerative Clustering Procedure
First Stage:

A =

Second Stage:

Third Stage:

B =
AB
AC
AD
AE
BC

CDA = 38.0
AB =

4.5

Fourth Stage:

C =
BD
BE
CD
CE
DE

=
=
=
=
=

D =

=
=
=
=
=

4.5
24.5
32.0
84.5
8.0

CDB =

14.0

CDE = 20.66

AE =

84.5

BE = 50.0

ABCD =

41.0

ABE= 93.17

10

E = 15

12.5
50.0
0.5
18.0
12.5

CDE =

20.66

Fifth Stage:
ABCDE = 98.8

Blackberry Pearl - Preferences


RIM
Respondents /
Palm Treo Motorola
BlackBerry
Brands
700p
Q
Pearl

Sony
Sidekick3 Ericsson Segment
M600i

Nokia
9300

I
II
I
II
II
III
I
I
14

Blackberry Pearl - 9 Cluster Solution

929.86

Distance

335.86

116.86
61.75
56.04
41.86
40.95
38.02
1

4
8
Cluster ID

15

Blackberry Pearl Cluster Profiles


Segmentation variable / Cluster

RIM BlackBerry Pearl

Overall

Cluster 1

Cluster 2

Cluster 3

6.77

8.42

5.47

5.6

Palm Treo 700p

5.5

4.41

4.32

Motorola Q

5.5

7.79

3.06

4.68

Nokia 9300

6.06

6.21

7.19

4.36

Sidekick3

4.12

2.91

3.47

7.04

Sony Ericsson M600i

4.54

5.33

3.62

4.36

16

Segmenting the PDA Market


Cluster Analysis (Benefit Segmentation)
Identifying customers who differ in terms of
their usage of the various features of
ConneCtor -- data and voice interconnectivity
How many segments and how do they
differ?
Which segments should be targeted?
Pricing?
Product Line?
How do we reach these segments?
17

DISCRIMINANT ANALYSIS

18

Using Discriminant Analysis :


Typical Marketing Problems
Investigation of group differences
Whether groups differ from one another
Nature of these differences
Characteristics that differentiate between
Purchasers of our brand and those of
competing brands
Brand loyal and non-loyal consumers
Light and heavy users of the product
Good, mediocre, and poor sales
representative
19

Using Discriminant Analysis


Example: How do Brand-loyal and Switchers differ
in terms of their socio-economic profiles?
Simplistic Approach: Calculate the mean income,
age, education level, and so on for the brandloyals and switchers and compare and contrast
the 2 groups on these dimensions
Potential Problems
Variables may be correlated e.g. income and
education level
Which of these variables are more important?

20

10

Discriminant Analysis Vs.


Cluster Analysis
How does Discriminant Analysis differ from Cluster
Analysis?
In Discriminant Analysis, we form a priori groups
(e.g. loyals vs. switchers) and then ascertain
variables which explain these differences.
In Cluster Analysis, no a priori grouping but let
data tell the natural groupings

21

Discriminant Analysis
Analytical tool that considers the variables simultaneously
so as to take into account their inter-relationship and
partially overlapping information
Construct a linear combination of the variables i.e. a
weighted sum
So that the linear combination best discriminates among
the groups

Mathematical Model
D = b0 + b1X1 + b2X2 + + bMXM + e1
D = discriminant score
b = discriminant coefficients or weights
X = predictor or independent variables
The coefficients, b, are estimated so that the groups differ as much as
possible on the value of the discriminant function, D
Occurs when the ratio of between-group sum of squares to withingroup sum of squares for the discriminant scores is at a maximum
22

11

Two-Group Discriminant Analysis

Price
Sensitivity

XXOXOOO
XXXOXXOOOO
XXXXOOOXOOO
XXOXXOXOOOO
XXOXOOOOOOO

X-segment

Need for Data Storage

x = high propensity to
buy
o = low propensity to
buy

O-segment

ME Segmentation and
Targeting 2006 - 23

Procedure - Discriminant Analysis


Problem Formulation

Step 1

Estimate the Discriminant Function


Coefficients

Step 2

Determine the Significance of the


Discriminant Function

Step 3

Interpret the Discriminant Function

Step 4

Assess Validity of Discriminant


Analysis

Step 5

24

12

Interpreting Discriminant Analysis


What proportion of the total variance in the
descriptor data is explained by the
statistically significant discriminant axes?
Does the model have good predictability
(hit rate) in each cluster?
Can you identify good descriptors to find
differences between clusters? (Examine
correlations between discriminant axes and
each descriptor variable).

Discriminant Analysis: Basic


Concepts
Key Words
Canonical Correlation: Measures the extent of
association between the discriminant scores
and the groups. It is a measure of association
between the single discriminant function and
the set of dummy variables that define the
group membership
Centroid: Mean values for the discriminant
scores for a particular group

26

13

Discriminant Analysis: Basic


Concepts
Key Words
Confusion Matrix: Contains the number of
correctly classified and misclassified cases.
The correctly classified cases appear on the
diagonal, because the predicted and actual
groups are the same
Discriminant Loading: Represents the
simple correlation between the predictors
and the discriminant function. Higher
loadings mean that the descriptor variable
is important in explaining segment
membership

27

14

Potrebbero piacerti anche