WORK BOOK 8 – SEGMENTATION

A. Segmenting Customers- Cluster Analysis

Techtronics ltd is seeking new market opportunities. The company is focusing on voice recognition
market & has narrowed down to 3 segments: the fearful typists, the power users & the professional
specialists. The fearful typists don’t know much about computers – they just want a fast way to create e-
mail messages, letters & reports without errors. The power users know a lot about computers, use them
often & want a voice recognition program with lots of features. The professional specialist have jobs that
require a lot of writing . They don’t know much about computers but are willing to learn.

The marketing manager prepared a table summarizing the importance of each of three key needs in the
three segments.

Spread sheet 1

************** Features Easy to Easy to Learn

******* Importance of need ( 1= not important; 10 very important )

Techtronics sales staff conducted interviews with seven potential customers who were asked to rate
how important each of these three needs were in their work. The manager prepared a spreadsheet to
help him cluster each person in to one of the segments. The manager can then aggregate potential
customers in to the segment that is most similar

Spread Sheet 2
 Importance of need:

Potential Type of Feature Easy to Easy to

A Dell 8 1 2
B IBM 6 6 5
C Apple 4 9 8
D Apple 2 6 7
E IBM 5 6 5
F Dell 8 3 1
G Apple 4 6 8

a. The ratings for a potential customer appear on the first spread sheet. In to which segment would
you aggregate this person.

b. The response for 7 potential customers who were interviewed are listed in spread sheet 2.

Write down the similarity score for each segment. Repeat the process for each customer.

Based on your anlaysis indicate the segment in to which you would aggregate each customer.

Indicate the size ( number of customers ) of each segment.

c. In the interview each potential customer was also asked what type of computers he would

be using. The responses are shown in the table along with the ratings. Group the responses

based on the customer’s segment. If you were targeting the fearful typist segment, what

type of computer would you focus on when developing your software.

d. Based on your analysis, which customer would you say is least like any of the segments ?
B. Aham retail conducts market research amongst a sample of 100
consumer to develop appropriate segments. Loyalty, Ad exposure and
usage were considered as important variables. The data was scaled from
1-9, with 1 being very low and 9 being very high. Aham wants to find
best segments using cluster analysis

AD
Consumer LOYALTY USAGE EXPOSURE
1 9.05 9.31 8.15
2 9.34 8.88 6
3 9.61 8.80 8
4 9.08 9.03 6
5 9.60 8.72 8
6 8.02 8.33 7
7 8.87 8.37 10
8 8.34 7.87 10
9 8.25 7.65 6
10 8.28 8.74 6
11 8.31 8.74 9
12 9.78 9.31 8
13 9.96 6.71 9
14 9.68 8.59 9
15 9.84 7.96 8
16 9.13 7.79 8
17 7.09 4.92 9
18 7.75 6.75 8
19 7.25 5.92 10
20 7.27 7.27 9
21 6.31 6.94 7
22 6.58 6.80 6
23 6.26 4.83 7
24 6.43 6.04 5
25 8.83 2.91 6
26 8.14 4.38 5
27 8.52 1.78 7
28 8.02 2.15 5
29 8.68 1.40 7
30 8.55 1.17 4
31 8.21 2.14 4
32 8.50 1.86 5
33 8.91 3.48 6
34 8.22 1.75 6
35 8.36 1.23 6
36 8.03 5.42 5
37 8.17 5.48 4
38 8.23 2.12 7
39 2.19 2.74 4
40 5.97 1.97 4
41 5.96 2.27 1
42 5.78 1.71 5
43 5.35 0.96 1
44 5.74 1.79 3
45 4.59 2.89 3
46 4.69 2.80 4
47 4.02 1.95 4
48 4.22 0.69 2
49 4.07 3.73 4
50 4.13 3.16 5
51 4.73 7.66 3
52 4.19 6.08 5
53 4.47 7.76 2
54 4.72 6.44 2
55 4.58 3.65 1
56 3.50 2.93 4
57 3.94 1.98 3
58 3.65 3.21 2
59 3.98 4.06 4
60 5.64 2.77 3
61 2.89 6.24 3
62 2.55 6.46 3
63 2.25 5.70 2
64 2.47 5.89 5
65 2.23 5.93 4
66 2.41 5.58 3
67 2.66 9.16 5
68 2.45 8.56 7
69 5.06 2.69 3
70 5.25 3.86 3
71 5.78 3.72 5
72 5.46 3.24 7
73 5.78 4.36 3
74 5.49 5.06 4
75 5.51 4.42 6
76 5.13 3.03 7
 77 5.67 3.79 5
78 2.84 4.56 4
79 2.35 3.53 3
80 2.15 5.25 2
81 2.45 5.88 1
82 2.07 4.85 4
83 2.59 3.85 3
84 2.96 9.41 3
85 2.51 0.59 2
86 3.61 1.18 2
87 3.32 1.26 1
88 2.04 1.29 1
89 2.96 0.94 5
90 3.00 1.24 3
91 2.65 1.24 2
92 2.32 1.00 2
93 1.32 1.37 2
94 1.10 0.85 3
95 1.85 0.81 3
96 1.56 1.05 2
97 1.06 1.36 1
98 1.84 1.30 2
99 1.43 0.67 1
100 1.81 0.50 2

Demographic Details :

Age Gender Income Profession

Customer s : 4 6 9 10 21 To 24 36 39 to 100 : 26- 35 M /F > 5 lacs IT

Customers : 5 7 8 13 To 20 29 : 40-50 F (70%) > 3 lacs HW / Buz

Customers : 1 11 : 10-14 B /G NA Students

Customers : 2 25 26 28 30 To 35 37 : 18 -20 M (90%) < 1 lac BPO

Customers : 1 11 : > 60 M > 3 lac Retired

TASK :

1. Select one amongst the best clusters

2. Amongst the best cluster choose 1 segment

3. Understand the chosen segment structure

5. Suggest marketing interventions.

6. Develop marketing strategy for chosen segment.

C. R STUDIO

Activity 1 :
Data set : Age – Spend
Install packages : ggplo2

Activity 2:
• Top 5 Dim Summary Head Tail Str
• What is the pattern visible at this point?
• Whom to target? To improve advertising, the marketing team wants to
send more targeted emails to their customers.

Activity 3:
Create clusters: Plot the total spend and the age of the customers

Activity 4:
IMPORT DATA - iris (in-built data set)

Activity 5:
Packages : cluster, factoextra
Activity 6:
Boxplot for Iris data

Activity 7:
• Clustering with k-means

Activity 8:
Principal Component Analysis : PCA finds a new set of dimensions (or a set of
basis of views) such that all the dimensions are orthogonal (and hence linearly
independent) and ranked according to the variance of data along them.
Use PCA for display

Activity 9:
PLOT CLUSTER MAP

Activity 10:
Hierarchical Clustering :  Creating clusters that have a predetermined ordering from
top to bottom

Activity 11:
• PLOT DENDOGRAM

Activity 12:
SEGMENT LABELS
Activity 13:
Download dataset : mydata
Packages : a. cluster b. factoextra

Activity 14:
Data Preparation : Prior to clustering data, you may want to remove or estimate
missing data and rescale variables for comparability

Activity 15:
Determine number of clusters

Activity 16:
K-Means Cluster Analysis :
5 cluster solution
Get cluster means
Append cluster assignment

Activity 17:
Wide range of hierarchical clustering approaches.
Ward Hierarchical Clustering

Activity 18:
• cut tree into 5 clusters
• draw dendogram with red borders around the 5 clusters
Activity 19:
Plotting Cluster Solutions :
a. K-Means Clustering with 5 clusters
b. Cluster Plot against 1st 2 principal components

Activity 20:
Centroid Plot against 1st 2 discriminant functions

Activity 21:
Data set : Wholesale.customers.data

Activity 22 :
Assign & call with customer

Activity 23 :
Prepare the data for analysis. Remove the missing value & remove “Channel” and
“Region” columns because they are not useful for clustering

Activity 24 :
Standardize the variables

Activity 25 :
Determine the number of clusters

Activity 26 :
Fit the model and print out the cluster means

Activity 27 :
Plot the results

Activity 28 :
Outlier detection with K-Means

Activity 29:
Calculate the distance between each object and its cluster center, then pick those
with largest distances as outliers and print out outliers’ IDs

Activity 30:
Make it more meaningful

Activity 31:
Apply Hierarchical Clustering to the data

Activity 32:
Ward’s method

Activity 33
Customer Segmentation using RFM Analysis
Activity 34
• INSTALL PACKAGES
• library(data.table) library(dplyr)
library(ggplot2) library(tidyr)
library(knitr) library(rmarkdown)

Activity 35
• Load Dataset : E commerce data

Activity 36
• Assign & Call : df_data

Activity 37
DATA SENSE : Top 5
• glimpse(df_data)

Activity 38
Data Cleaning : Delete all negative Quantity and Price. Delete NA customer ID

Activity 39
Recode variables : To convert character variables to factors

Activity 40
Calculate RFM
Activity 41
• HISTOGRAM : Recency – How recently did the customer purchase?
• Frequency – How often do they purchase?
• Monetary Value – How much do they spend?
Activity 42
Data is skewed, use log scale to normalize

Activity 43
Clustering

Activity 44
CLUSTER DENDOGRAM

Activity 45
Cut