Upload
tuhin-chattopadhyay
View
53
Download
0
Embed Size (px)
Citation preview
CUSTOMER SEGMENTATION
By
Tuhin Chattopadhyay, Ph.D.
2
1. Business & Research Objectives
2. Executive Summary
3. Analytics Approach - Overview
4. Overall & Product Specific Segmentation
5. Decision Tree and Decision Rules
6. Appendix
Two Wheeler Loan Segment
Personal Loan Segment
Consumer Durable Loan Segment
Personal Loan Cross – Sell
Product and Overall Segments Mapping
Table of Contents
3
BUSINESS & RESEARCH OBJECTIVES
Segment the customers into unique segments to enable targeted marketing activities. Business
Objective
•Segment the customers into unique clusters. •Segmentation to be done for all customers of the client as well
as within each product category. •Provide distinct segments of customers along with their profile.
Research Objectives
Objective – Agent Profiling
4
Executive Summary • Segmentation done for 1.31 million customers • Demographic and Transactional variables considered based on business relevance and data availability • Variable transformation, outlier treatment and missing value imputation done based on requirement
• Profiles of macro and micro segments
• Map products purchased by each of the segments
• Decision Rules to Segment New Customers
Key Takeaways
Misclassification Error through Discriminant analysis
Model Validation
Agglomerative Hierarchical Clustering Methods & K-Means clustering used in tandem.
Statistical Modelling
Derive key variables and build rules to segment new customers
Decision Tree
Output
Overall Segmentation
5
ANALYTICAL APPROACH - OVERVIEW
Interpreting the Characteristics of the segment based on modelling output
Segment Profile
Statistical Modelling, Evaluation & Profiling
Discriminant analysis Misclassification Error
Validation Techniques
Agglomerative Hierarchical Clustering Method (Wards) & K-Means clustering in tandem.
Model Development
Age, Education, # Children, Work Experience, Gender, Marital Status, Occupation, Current Province, Income
Descriptive Analytics and Pattern Recognition
Variables Considered - Demographic
Exploratory Data Analysis
Data Understanding
Loan Amount, EMI, Interest Rate, Tenure, # of Contracts, DPD, SBV Bucket (G1, G2, G3, G4 & G5), Sales Channel, Interest Amount
Variables Considered - Transactional
Data Preparation
Data Set Creation
Created 5 data sets for modelling – • Overall Customer
base • Two Wheeler Loan • Consumer Durable
Loan • Personal Loan • Cross Sell & Up Sell
Variables Transformation
• Education in Years • Real Income
Data considered for all active customers from 1st January 2014 till 31st August 2015
Time Period
• In case of multiple loans the most recent contract considered
• Closed contracts considered in cases where customer has not taken an additional loan
• Separate analysis is done for charged off customers
Data Preparation
Customer Segmentation
7
Overall Customer Base Segmentation
Total Customers segmented: 1,314,582
Aspirers 434,802 (33.1%)
Desperate 275,274 (63.3%)
Mature 83,295 (19.16%)
Successful 76,233 (17.53%)
Pragmatic 358,771 (27.3%)
Wise 144,947 (40.4%)
Accumulator 213,824 (59.6%)
Affluent 521,009 (39.6%)
Homogeneous Segment
Note – The three macro and five micro segments have been identified after multiple iterations, to ensure that each segments are unique.
8
Product Mapping – Aspirers Segment
Aspirers 434,802 (33.1%)
Desperate 275,274 (63.3%)
Mature 83,295 (19.16%)
Successful 76,233 (17.53%)
Product Category No of Customers
Consumer Durable 272907 (99.14%)
Product Category No of Customers
Consumer Durable 59193 (71.06%)
Two Wheeler 13780 (16.54%)
PL New-to-bank 6547 (7.86%)
PL X-sell and Top-up 3775 (5.53%)
Product Category No of Customers
Two Wheeler 48978 (64.25%)
PL New-to-bank 11616 (15.24%)
Consumer Durable 7891 (10.35%)
PL X-sell and Top-up 7748 (10.16%)
9
Pragmatic 358,771 (27.3%)
Wise 144,947 (40.4%)
Accumulator 213,824 (59.6%)
Product Mapping – Pragmatic Segment
Product Category No of Customers
Consumer Durable 106470 (74.45%)
Two Wheeler 18215 9 (12.57%)
PL New-to-bank 16604 (11.46%)
PL X-sell and Top-up 3658 (2.52%)
Product Category No of Customers
PL New-to-bank 74029 (34.62%)
Two Wheeler 52319 (24.47%)
PL X-sell and Top-up 49248 (23.03%)
Consumer Durable 38338 (17.88%)
10
Product Mapping – Affluent Segment
Affluent 521,009 (39.6%)
Homogeneous Segment
Product Category No of Customers
PL New-to-bank 314659 (60.39%)
PL X-sell and Top-up 166828 (32.02%)
Two Wheeler 32873 (6.31%)
Consumer Durable 6649 (1.28%)
11
Overall Segmentation Dashboard
• The “Aspirers” segment is home to the youngest customers with the lowest income. Active in their finances and comfortable making tough financial decisions as shown with the high interest rate.
• “Pragmatic” segment comprises the oldest group of customers. Low interest & below average tenure show a thought through approach to financing
• The “Affluent” segment has the highest income consuming the highest amount of loan and with the longest tenure.
12
Overall Segmentation Dashboard
Occupation
Marital Status
• Highest number of students within “Aspirers” segment.
• Majority of the “Pragmatic” segment are self employed with a conservative approach to consume loans which is evident through loan amounts, interest rate and tenure
• “Affluent” group has the largest group of customers who hold a job (Blue Collar, White Collar) making them a secure segment. They also have the least number of students
31.18%
42.70%
18.58%
27.02% 25.75%
22.47%
18.71%
11.58% 9.17%
15.24%
12.02%
23.44%
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
Aspirer Pragmatic Affluent
SELF-EMPLOYED BLUE-COLLAR STUDENT WHITE-COLLAR
52.30%
76.16%
63.23%
39.38%
10.86%
29.21%
0%
10%
20%
30%
40%
50%
60%
70%
80%
Aspirer Pragmatic Affluent
Married Single
13
ASPIRERS
• The “Desperate” segment forms 63% of the “Aspirer” group. This group has the highest interest rates and lowest incomes amongst “Aspirers”
• Interest amount paid by the “Successful” segment is 3.6 and 4 times higher than the other micro segments
14
PRAGMATIC
• “Accumulator” segment is the oldest segment among all the micro segments
• Loan amount issued to “Accumulator” is 1.86 times that of the “Wise” segment” despite having an significantly higher interest rate.
• Given that the EMI to Income ration for “Accumulator” and “Wise” segment is 23%, and 18% respectively, they are good candidates for cross sell / up-sell.
15
PRAGMATIC
34%
49%
28% 24%
16%
9%
15%
10%
0%
10%
20%
30%
40%
50%
60%
Wise Accumulator
SELF-EMPLOYED BLUE-COLLAR STUDENT WHITE-COLLAR
• The “Accumulator” segment has the highest number of married customers at 84%. Well settled with family makes them an attractive segment for additional loans.
• 49% of “Accumulators” are self employed indicating the need for large loans.
• High Education levels among “Wise” segment shows their discretion in availing loans.
Occupation
65%
84%
23%
3%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
Wise Accumulator
Married Single
Marital Status
Rules for Segmenting New Customers
17
Decision Tree - Overview
What is a Decision Tree?
• Decision tree is a type of supervised learning algorithm (having a pre-defined target variable) that is mostly used in classification problems. It works for both categorical and continuous input and output variables.
• Decision trees generate the importance of variables for classification. These variables are used to define rules that will help classify customers.
• In this technique, we split the population or sample into two or more homogeneous sets (or sub-populations) based on most significant splitter / differentiator in input variables.
• The objective is to understand in which cluster a new customer will belong to.
• The 6 clusters viz. Desperate, Mature, Successful, Wise, Accumulator and Affluent are considered as the levels of the dependent variable.
• The demographic variables like age, income, education, number of children, work experience, occupation etc. as the independent variables.
Application of Decision Tree for New Customer Profiling
Order of Importance
Variable
First Income
Second Age
Third Work Experience
Fourth # Children
Fifth Occupation
Sixth Education (Yrs)
18
Indicative Rules for Segmenting New Customers
Aspirers
Desperate
Mature
Successful
IF INCOME>=2,000,000 INCOME<= 5,122,277 AND AGE >= 27 AND AGE <= 31
IF INCOME>=5,122,278 TO INCOME <=6,049,832 AND AGE>=24 TO AGE <=29
IF INCOME>= 6,049,833 TO INCOME <=7,000,000 AND AGE>=22 TO AGE<=28
Pragmatic
Wise
Accumulator
IF INCOME>=5,080,561 TO INCOME <= 6,448,612 AND AGE>=31 TO AGE<=40
IF INCOME>= 6,066,263 TO INCOME<= 7,353,570 AND AGE >= 41 TO AGE <= 65
Homogenous Segment
IF INCOME >= 6,511,105 AND AGE >= 29 TO AND AGE <= 34 Affluent
Note: Decision Tree throws number of rules for each of the segments. The indicative rules are
presented here. The exhaustive list are provided in the Technical Document.
Thank you ! 19
Appendix
Product Wise Customer Segmentation
22
Product Specific Segmentation
Two Wheeler (215260, 16.37%)
Young Turks (46320, 21.52%)
Diligent (84402, 39.21%)
Satisfied Entrepreneurs
(28825, 13.39%)
Risky Seniors (55713, 25.88%)
CDL (560920, 42.66%)
High Spenders (283230, 50.49%)
Affluent Young (159064, 28.36%)
Status Seekers (118626, 21.15%)
Personal Loan (466161, 35.46%)
High Earning Opportunists
(132517, 28.43%)
Promising (202121, 43.36%)
Middle Aged Conservatives
131523, 28.21%)
Top up & Cross Sell
(235634, 17.9%)
High Rollers (66050, 28.03%)
Up and Coming (111696, 47.40%)
Traditionalists (57888, 24.57%)
Note – The individual product level segments have been identified after multiple iterations, to ensure that each segments are unique.
Two Wheeler Loan Segment
24
TW Segment Profile - Overview
• “Young Turks” segment is a target for marketing activities as this is one of the youngest clusters with the second highest average income.
• “Diligent” have the highest EMI to income ratio leading to the lowest disposable income within the TW category.
• “Satisfied Entrepreneurs” have the highest disposable income within the Two Wheeler product category.
• The “Risky Seniors” and “Diligent” have similar Income and Loan appetite even though their average age is 42.78 and 26.67 respectively. Similarly “Satisfied Entrepreneurs” and “Young Turks” have similar transaction history given their average age is 42.9 and 27.11 respectively
2. Diligent
25
Occupation
TW Segment Profile
Province
• Over 50% of the older segments (Satisfied Entrepreneurs & Risky Seniors) are self employed compared to the younger segments who hold blue / white collar jobs
• “Young Turks” and “Satisfied Entrepreneurs” who have the highest income are primarily from Ho Chi Minh city compared to the “Diligent” and “Risky Seniors” who are from Binh Duong
• Over 80% of “Satisfied Entrepreneurs” and “Risky Seniors” are married with children.
Marital Status
36% 34%
55% 52%
25% 24% 24% 24%
16% 18%
9%
15% 13%
8%
0%
10%
20%
30%
40%
50%
60%
Young Turks Diligent SatisfiedEntrepreneurs
Risky Seniors
SELF-EMPLOYED BLUE-COLLAR STUDENT WHITE-COLLAR
11%
9% 10%
7%
5%
14%
4%
10%
7% 7% 7% 6%
0%
2%
4%
6%
8%
10%
12%
14%
16%
Young Turks Diligent SatisfiedEntrepreneurs
Risky Seniors
Ho Chi Minh City Binh Duong Dong Nai
52% 53%
85% 86%
41% 40%
5% 5%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Young Turks Diligent SatisfiedEntrepreneurs
Risky Seniors
Married Single
Personal Loan Segment
27
Segment Profile - Overview
• At 21% the “High Earning Opportunists” have the lowest EMI to Income ratio - High disposable income.
• “High Earning Opportunists” consume the largest loans amongst the PL group with a significantly larger tenure.
High Earning
Opportunist Promising Middle Aged
Conservatives
High Earning
Opportunist Promising Middle Aged
Conservatives
75%
56%
75%
15%
38%
13%
0%
10%
20%
30%
40%
50%
60%
70%
80%
High Earning Opportunist Promising Middle Aged Conservatives
Married Single
28
Segment Profile - Overview
Occupation
• 75% of the “High Earning Opportunists” segment hold a job where as only 20% are self employed
• 90% of the “Promising” Segment hold jobs where as only 5 % is self employed
• The % of students within all the segments is low indicating that most of the customers within the Personal Loan category are earning and not dependent on others
Marital Status
20%
5%
31%
25%
22% 21%
5% 4% 5%
23%
31%
20%
0%
5%
10%
15%
20%
25%
30%
35%
High Earning Opportunist Promising Middle Aged Conservatives
SELF-EMPLOYED BLUE-COLLAR STUDENT WHITE-COLLAR
Consumer Durable Loan Segment
30
Segment Profile - Overview
• “High Spenders” segment have the highest interest rate in the entire customer universe. This coupled with
• The “Affluent Young” segment enjoys a significantly lower interest rate (30.6 %) when compared to the other two segments, despite sharing a comparable income.
• Loans availed by “Affluent Young” are higher by over 50% compared to “High Spenders” and “Middle Aged Conservatives”
Affluent
Young
High
Spenders
Status
Seekers Affluent
Young
High
Spenders
Status
Seekers
31
Segment Profile
• 81% of the “Status Seekers” segment are married compared to the “High Spenders” and “Affluent Young” where the percentage is significantly lower.
• “Status Seekers” being the oldest group, also have the highest work experience.
Marital Status
48%
56%
81%
45%
33%
3%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
High Spenders Affluent Young Status Seekers
Married Single
Personal Loan – Top up & Cross Sell Segment
33
Segment Profile - Overview
• Loan amount of “High Rollers” twice that of “Up and Coming” and “Traditionalists”
• The “Traditionalists” are 13.7 years older than “Up and Coming” and 8.7years older than the “High Rollers”
High
Rollers Up and
Coming
Traditio
nalists
High
Rollers Up and
Coming
Traditio
nalists
19%
48%
35%
21% 22% 23%
13% 13%
19% 22%
10% 14%
0%
10%
20%
30%
40%
50%
60%
High Rollers Traditionalists Up and Coming
SELF-EMPLOYED BLUE-COLLAR STUDENT WHITE-COLLAR
34
Current Region
Occupation
• The younger groups, “High Rollers” and “Up and Coming” hold Blue / White collar jobs where are the “Traditionalists” are self employed.
• The top three regions for all the segments is Ho Chi Minh City, Binh Duong and Dong Nai
• 86% of the “Traditionalists” segment is married with an average of almost 2 children
Segment Profile - Overview
Marital Status 24%
17%
14%
16%
9% 8%
11%
7% 6%
0%
5%
10%
15%
20%
25%
High Rollers Traditionalists Up and Coming
Ho Chi Minh City Binh Duong Dong Nai
73%
86%
57%
19%
5%
34%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
High Rollers Traditionalists Up and Coming
Married Single
Mapping of Product Segments to Overall Segments
DESPERATE
MATURE
SUCCESSFUL
WISE
ACCUMULATOR
AFFLUENT
42
Charge Off Customer Cluster
97%
3%
Charged Off Status
Non-Charged Off Charged Off
26%
28%
19%
16%
10%
Current Region
Centre
Mekong
North
South
East
Charged off Customers by Current
region
There are 3% charged off customers. Out of that
54% are from South.
48%
33%
7%
13%
Product Group
TW
PL X-sell and Top-up
PL New-to-bank
CDL
Charged off Customers
by Product Group
There are 3% charged off customers. Out of
that 48% are CDL and 33% are PL (81%
together).