This presentation was given on October 12, 2013 at the Marketing EDGE Jacobs and Clevenger Casewriter's competition, where it received a Silver Award. The case outlines how to teach descriptive analytics, profiling and clustering for a fictional company.
Teaching Descriptive Analytics, Customer Profiling and Clustering
1. ECB.com: Customer
Profiling and
Segmentation
Dr. Debra Zahay-Blatz
Aurora University
Dr. Blodwen Tarter
Golden Gate University
Jacobs and Clevenger Casewriter’s
Competition Presentation, McCormick Place
10/12/2013
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
2. We Know Customer Information
Management Breeds Success
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
3. Problem: How can we teach
undergraduates/graduates the
fundamentals of data analysis?
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
4. ECB.com Business Model
• Sells coupons directly to consumers online
(formerly a print service)
• Coupons are used directly for types of
entertainment
• $25 coupon for a restaurant (the data used here)
is only $10.
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
5. Problem Setup:
• Data analytics team ponders segmentation
• Points of view: two junior analysts
• Existing customer segmentation: New, Engaged,
Lapsed, Inactive (RFM)
• Can we ‘beat the control’?
• Are there other insights?
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
6. Student Goals
•Understand how to read and interpret a data dictionary for a
customer database.
•Be able to run basic descriptive analysis on customer data and analyze
it.
•Be able to create a case summary report using a grouping variable,
such as cluster membership.
•Be able to use a TwoStep Cluster Analysis to create customer
segments in SPSS and interpret the results.
•Understand the basic principles of an outside segmentation scheme
created by a data vendor and what the segments might look like.
•(Optional): Be able to determine the effects of missing data in a
customer database.
•Understand the basics of what a Marketing Data Analyst does in
his/her job.
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
7. Case Approaches
• All or part of the case
• Analysis or interpretation
• Data vs. strategy
• SPSS vs. Excel vs. just interpretation
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
8. Understanding a Data Dictionary
is a Stand-Alone Lesson
DATA DICTIONARY
Variable Name
Unique ID
Transaction/Segment
Data
# days since last order
# of Orders
# of Certificates
Revenue
Tenure (in days)
Lifecycletenurerecovery
Average Order Value
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
Description
Unique Identifier for Customer
# of days since last order (recency) in the last 3
years
# of Orders in the last 3 years
# of Certificates in the last 3 years
Revenue in the last 3 years
# of days since first order with ECB
New = First order with ECB in the last 3 months
Engaged = Last order was within the last 6
months (and not new)
Lapsed = Last order was 6 - 12 months ago
Inactive = Last order was over 12 months ago
Revenue / # of Orders in the last 3 years
9. Q1: Typical Customer from
Descriptives
Descriptive Statistics
Days since Last order
N
Minimum Maximum
60000
1
1095
No. of Orders
60000
1
961
2.75
5.373
No. of Certificats
60000
1
1043
8.45
13.629
Revenue
60000
$.00
$5,193.90
$26.1003
$41.53822
Tenure in Days
60000
10
2612
589.10
446.789
Avg. Order Value
60000
$.00
$920.00
$11.7441
$13.93730
No. of orders
w/promotion
No. of orders w/o
promotion
60000
0
409
2.47
4.010
60000
0
44
.44
1.049
No. of orders w/ 80
percent off promotion
60000
0
156
1.53
2.530
No. of orders w/ 90
percent off promotion
Valid N (listwise)
60000
0
15
.16
.501
60000
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
Mean
Std. Deviation
386.57
295.136
10. Q2: Segment Behavior from
Case Summary
Case Summaries
lifecycletenurerecovery
Engaged
N
Dayssincelast Numberoford
order
ers
13521
13521
avgordervalu
e
Revenue
13521
13521
Mean
N
26177
26177
26177
26177
673.87
1.81
$19.7037
$12.4295
N
15610
15610
15610
15610
263.46
2.62
$25.6383
$11.7408
N
4692
4692
4692
4692
41.93
1.38
$17.5838
$13.4102
N
60000
60000
60000
60000
Mean
Total
$9.8431
Mean
New
$41.9729
Mean
Lapsed
5.19
Mean
Inactive
92.07
386.57
2.75
$26.1003
$11.7441
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
11. Q2: Try to “beat the control”
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
12. Q2: Use the exercise to explain
clustering procedures
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
13. Q3: Data Vendors
• Nielsen PRIZM (formerly Claritas)
• Experian
• Axciom
• ESRI Tapestry
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
14. Bonus: 80/20 Rule
Case Summaries: Current Segmentation Scheme, about 42% of
revenue in last three years comes from Engaged and New
Customers (30%, (13521+4692)/60000) under the old
Segmentation Scheme
Revenue
lifecycletenurereco
very
Engaged
% of Total
Sum
N
Mean
Sum
13521 $41.9729 $567,515.84
36.2%
Inactive
26177 $19.7037 $515,782.78
32.9%
Lapsed
15610 $25.6383 $400,214.35
25.6%
4692 $17.5838 $82,502.97
5.3%
New
Total
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
60000 $26.1003
$1.57E6
100.0%
15. Contact information
• Dr. Debra Zahay-Blatz
• dzahayblatz@aurora.edu
• Cell 630-300-8838
• Work 630-844-3825
• Dr. Blodwen Tarter
• btarter@ggu.edu
• Work 415-442-6587
Copyright by Debra Zahay-Blatz and Blodwen Tarter 2013
Hinweis der Redaktion
Students know information mangement breeds success but not the techniques to do so