MIS784 Marketing Analytics: Customer Churn & RFM Analysis-TESCO

This report provides an analysis of customer data from TESCO, focusing on predicting customer churn and performing Recency Frequency Monetary (RFM) analysis. The analysis uses decision trees in IBM SPSS and Microsoft Excel to evaluate model performance. The report includes a literature review on customer satisfaction, outlines the methodology used (descriptive and quantitative research designs), and presents results related to customer segmentation and churn prediction. The goal is to help TESCO improve customer experience and loyalty programs by understanding customer behavior and prioritizing customer service efforts. Desklib offers a range of resources, including similar reports and solved assignments, to aid students in their studies.

Deakin Business School
Department of Information Systems and Business Analytics
Assignment 2
Name of Student:
Student ID:

Contents
Introduction.................................................................................................................................................3
Literature Review........................................................................................................................................3
Methodology...............................................................................................................................................5
Results.........................................................................................................................................................7
Conclusion.................................................................................................................................................13
Recommendations.....................................................................................................................................13
References.................................................................................................................................................15
Appendix...................................................................................................................................................16

Introduction
This section gives an overview of the report and the problem that this report is solving. The
primary purpose of this report is to provide an analysis of some of the important aspects of
customers from TESCO public limited company. The analysis is meant to assist Telco in making
important decisions which are data- driven as well as being conscious of the nature of customer
base that they have.
TESCO is public limited company leading international supermarket based in the United
Kingdom (UK). TESCO Public limited company has a wide base of customers. TESCO public
limited company would like to improve their customer experience based on the experiences that
they have with the customers in the recent past. As a result, TESCO collected some important
information from their customers. The information that was collected for each client included
customers’ loyalty, customers’ churn, customers’ purchasing amount and the last time of
purchase just to mention a few.
The problem that is being solved in this report is to help TESCO public limited company in
predicting their customer’s churn as well as the Recency Frequency Monetary (RFM). Therefore,
the focus of TESCO is to be able to predict their customers’ church based on the parameters that
they have. Similarly, TESCO is interested in segmenting customers into different categories
based on their churn or loyalty. Segmentation of the customers will enable the management to
properly prioritize the customers and improve on their customer service as well as experience.
Finally, TESCO public limited company is interested in determining the Recency Frequency
Monetary (RFM) of the customers. Recency Frequency Monetary (RFM) method is a technique
that is commonly used in marketing and sales field to quantitatively analyze and determine the
best customers for given product or business (Vermeeren, et al., 2011). Similarly, Recency

Frequency Monetary (RFM) will also help TELCO to evaluate and improve on their customer
loyalty scheme that they launched in 1995.
Literature Review
There are several studies that have been done on marketing, customer service, and customer
satisfaction. Several scholars have published articles and journals on customer satisfaction; ways
of improving customer satisfaction, benefits of improving customer satisfaction and why it is
important to improve customer satisfaction (both to the business and to the customer).
Customer satisfaction is the way of quantifying the extent or the degree to which a client or
customer is happy with a service, a product or even with experience during the service
(Chesnokova, et al., 2013). A study by (Bourne & Paul, 2016) demonstrates why customer
satisfaction is a very important concept for business. The study outlines that a satisfied customer
is a loyal customer. A loyal customer is a treasure that a business should keep and hide from the
rest of the business. The study demonstrates that a loyal customer is very crucial in ensuring that
a business is on par with its competitors. However, this article did outline how to identify some
of the loyal customers that a business should treasure.
A study by (Jung & Yoon, 2013) outlines that customer satisfaction can be used as a tool to stand
out of the rest of the competitors. Therefore, a business with stiff competition, it is important that
they improve on their customer service and experience so that they can attract more loyal
customers (Cheng, et al., 2009). However, this study did not outline the metrics or the techniques
that business can use to identify loyal customers as well as keep the existing ones.
Exceptional customer service or experience can help a company to build a brand (Netemeyer, et
al., 2010). When a company builds a brand of itself or a product, it will be very easy to reach a

wide range of market as well as stand out of the rest of the competitors (Deng, et al., 2013).
Similarly, building a brand will enable the business to be sustainable enough thereby fulfilling
the concept of going concern. Therefore, it should be a primary objective for a business to offer
exceptional customer experience to its customers as well as its prospective customers (Sun, et al.,
2013).
A study on the analysis of market- based approach to customer satisfaction reveal that every
customer should be treated as if they are “very important persons” ( VIP) (Hur, et al., 2015). The
researchers argue that treating customers in a manner to suggest that they are highly valued and
very important will improve their loyalty. The loyalty of a customer is one way of getting
customers outstanding from the pool of customers. The study reveals that some of the best
practices that can make a customer feel very important are when customer service officers make
endless efforts to assist the customer in every way they can as well as when the customer service
officers (and the entire staff of the business) keep their promises and remain honest. However,
the study did not point out some of the important aspects that a business should consider doing to
their customers so that the customers can feel they are treated with utmost respect.
The other way of improving customer satisfaction is by constantly measuring the level of
customer satisfaction (Deng, et al., 2013). In this dynamic market, a business should make it a
habit to constantly measure the level of their customers’ satisfaction. They should get timely
feedbacks o that they can strategize and make necessary adjustments to improve on their
customer experience. Similarly, constantly evaluating the level of customer satisfaction will
enable the business to put the right technology in marketing and offering their services. Putting
the right technology in place will ensure that a business is sustainable and is at par or above par
with the rest of the competitors. However, this study did not outline how customers can be put

into different segments or categories for a better experience. Moreover, the study did not outline
some of the best ways and parameters for measuring the level of customer satisfaction.
The other way of ensuring exceptional customer service and experience is maintained is by
knowing the right way to survey the customers (Setiawan & Budi, 2014). There are numerous
ways and platforms of surveying the customers to get their feedback on their experience.
However, not all the survey methods are suitable or can work for every business (Setiawan &
Budi, 2014). The nature of the survey should be determined based on the nature of the products
and services. However, this study did not reveal how to suitably determine a survey method.
A study by (Bueshken, 2009) reveals that it is very crucial to keep an eye on what the customers
are saying about the business on social media. Similarly, it is important to keep an eye on what
the customers are saying about the industry on social media. Therefore, it is very crucial for a
business to build and maintain a serious and working online social media platform for better
customer experience (Bueshken, 2009). The study further suggests that customer analytics and
intelligence can be applied to such information to determine whether there should be an
improvement. Similarly such feedback will help the business to know their position compared to
the competitors in the industry (Bueshken, 2009).
Methodology
This section describes the analytical tools as well as the analytical techniques that have been
conducted to meet the objectives. Similarly, this section also outlines the research design, the
model evaluation metrics, the working data, and the model building process. The analytical tools
that have been used are Microsoft Excel and IBM SPSS (Statistical Packages for Social

Sciences). IBM SPSS has been used to construct a model for predicting customer churn using
binary classification trees. Microsoft Excel has been used to evaluate the performance of the
model. Microsoft Excel has also been used to evaluate the performance of the model against the
Recency Frequency Monetary (RFM) method.
As mentioned above, the analytical techniques that have been used are the decision trees and the
RFM method. Decision trees have been chosen because building a decision tree is an easier and
simpler way of making predicting and comparisons using SPSS (Reyes, et al., 2010). Similar,
decision trees are crucial in identifying the groups in a data set, demonstrating the relationships
among the groups and among the variables and predicting future occurrence of the events.
Decision trees also outline the probability or the possibility of the occurrence of such events
(Boos, et al., 2013).
Recency Frequency Monetary (RFM) method is a technique that is commonly used in the
marketing and sales field to quantitatively analyze and determine the best customers for given
product or business. The determination of the best customers for products and services is done
based on the examination of the most recent purchases that the customers have made, the
frequency of purpose of the customers and the amount of money that the customers spend for
goods and services. Customers who are considered to be the best for business are those who
often purchase from the business and with significantly large amounts of money (Yin, 2012).
RFM analysis uses sales data to create a segment of a pool of customers who would be best
suited for the business (Schafer, 2015).
The research designs that have been applied in this study are the descriptive research design and
the quantitative research design (Boos, et al., 2013). A descriptive research design is a technique
that is applied to provide an explanation or description of an event or a phenomena. In our

scenario, a descriptive research design has been used to outline the nature of the customers that
TESCO public limited company have. The customers have been divided into either churners or
non- churners (Boos, et al., 2013). A prediction model has been produced to forecast the number
and nature of the customers that TESCO public limited company expects to have or receive for
their goods and services (Boos, et al., 2013).
A quantitative research design is a technique that is used to describe a phenomenon in terms of
the number or frequency of occurrence (Boos, et al., 2013). A quantitative research design is
applicable when the parameters or the variables at hand are quantitative in nature (Boos, et al.,
2013). A quantitative research design has been applied in this research to determine the number
of churners and non- churners. Similarly, a quantitative research design has been applied in this
research to determine the profit as well as determine the frequency of purchase for different
categories of goods.
The data set consisted of the information that was collected from the previous customers that
visited the TESCO public limited company (plc). The variables of the data set are shown below.
Variable Name Description
ID The unique ID of customers
Purchase Number of purchases during the observation period1
T.last The time gap between customer’s first purchase and last purchase during
the observation period
T.active The time gap between customer’s first purchase and last day of the
observation period
Loyalty A binary variable to show membership level: (0) Silver (1) Gold
Service Failure Number of service failures during the observation period
Total Profit Total profit generated by the customer during the observation period
AP.spent Total spending on the Apparel category during the observation period
BH.spent Total spending on Bakery category during the observation period
DL.spent Total spending on Deli category during the observation period

DY.spent Total spending on Dairy category during the observation period
FV.spent Total spending on Fresh Produce category during the observation period
GM.spent Total spending on General Merchandise category during the observation
period
GR.spent Total spending on the Grocery category during the observation period LQ
LQ.spent Total spending on Liquor category during the observation period
MT.spent Total spending on Meat category during the observation period
Socio.Economic Socio-Economic status of the customer on a scale from 1(lowest) to 10
(highest)
churn A binary variable to show the churn status of the customer in the prediction
period2 (0) non-churner (1) churner
Results
The results of the model for that can be used to predict the churn are shown below. The results
have been obtained by use of SPSS. The dependent variable was the churn while the rest of the
variables except the customer id are the independent variables. The table below shows the model
summary. The summary outlines that various segments of grouping the customers (Hasnelly &
Eddy, 2012).
Model Summary
Specifications
Growing Method CRT
Dependent Variable Churn
Independent Variables
T.last, T.active, Loyalty, Service Failure,
Total Profit, AP.spent, BH.spent,
DL.spent, DY.spent, FV.spent,
GM.spent, GR.spent, LQ.spent,
MT.spent, Socio.Economic, Purchase
Validation None
Maximum Tree Depth 5
Minimum Cases in Parent Node 100
Minimum Cases in Child Node 50

Results
Independent Variables Included
Purchase, Total Profit, GR.spent,
DY.spent, T.last, BH.spent, FV.spent,
MT.spent, DL.spent, T.active, GM.spent,
Service Failure, LQ.spent, Loyalty,
AP.spent, Socio.Economic
Number of Nodes 57
Number of Terminal Nodes 29
Depth 5
The table below shows the prior probabilities. A prior probability is a probability value obtained
from the data set that we had. From the prior probabilities, it is clear that the probability of
getting a churner customer is 0.428 while the probability of getting a non- churner customer is
0.572.
Prior Probabilities
churn Prior Probability
0 .572
1 .428
Priors are obtained from the training sample
Misclassification Costs
Observed Predicted
0 1
0 .000 1.000
1 1.000 .000
Dependent Variable: churn
Risk
Estimate Std. Error
.213 .003
Growing Method: CRT
Dependent Variable: churn

The table below outlines the importance of all the variables as well as the normalized
importance.
Independent Variable Importance
Independent Variable Importance Normalized Importance
T.last .231 100.0%
Purchase .129 55.8%
Total Profit .128 55.5%
GR.spent .123 53.4%
DY.spent .106 46.0%
BH.spent .066 28.6%
FV.spent .054 23.3%
MT.spent .048 20.8%
T.active .045 19.6%
GM.spent .034 14.8%
DL.spent .033 14.5%
Service Failure .030 12.9%
Loyalty .017 7.3%
AP.spent .007 2.9%
LQ.spent .007 2.8%
Socio.Economic .001 0.3%
Growing Method: CRT
Dependent Variable: churn
The table below shows the classification of the predicted churns. There are a total of 9238 cases
of non- churners while the cases of churners are 2060.
Classification
Observed Predicted
0 1 Percent Correct
0 9239 2203 80.7%
1 2060 6498 75.9%
Overall Percentage 56.5% 43.5% 78.7%
Growing Method: CRT
Dependent Variable: churn

The bar graph of a normalized and importance. From this graph, we can say whether the
variables are normally distributed. The data does not portray the characters of a normal
distribution.
Using the analysis of a confusion matrix, it is clearly demonstrated here that the accuracy is
78.6%. This result demonstrates the model that we have developed above has predicted 78.6% of
the real churners. Apart from the calculations below, this assertion can also be seen in the
classification table. According to the records in the classifications table, the model has predicted
78.7% of the churners.
Actual Prediction
0 1
0 True Negative (TN) False Positive (FP)
1 False Negative (FN) True Positive (TP)
Accuracy= TN +TP
( N + P) =(4564+3295)
(5749+ 4251) =78.6 %
The rate of Misclassification= FN +FP
(N +P) = 956+1185
(5749+4251) =21.41%
sensitivity=TP
P = 3295
4251 =77.51 %

The cumulative lift chart
The table below has been used to produce a graph of the cumulative lift.
Decile Decision Tree RFM Random
0 0 0 0
1 18.34862 18.70148 10
2 36.83839 36.81487 20
3 55.11644 55.91625 30
4 71.58316 71.44201 40
5 80.52223 82.99224 50
6 84.61538 91.53140 60
7 88.26159 96.21266 70
8 92.23712 97.90637 80
9 96.23618 99.76476 90
10 100 100 100
The full output of the model can be found in the appendix section of this report. The appendix
provides the actual predicted values as well as the surrogate values.

Comparing RFM and CRT
The table below outlines the output of the comparison of the RFM and the CRT.
Decile Decision Tree RFM Random
0 0 0 0
1 18.34862 18.70148 10
2 36.83839 36.81487 20
3 55.11644 55.91625 30
4 71.58316 71.44201 40
5 80.52223 82.99224 50
6 84.61538 91.53140 60
7 88.26159 96.21266 70
8 92.23712 97.90637 80
9 96.23618 99.76476 90
10 100 100 100
0 1 2 3 4 5 6 7 8 9 10
0
20
40
60
80
100
120
Comparing FRM and CRTDecile Decision Tree
RFM
Decision Tree RFM Random

Conclusion
The analysis results demonstrate that TESCO can actually pull out their customers from a pool of
customers. Pulling out customers from a pull of other customers ensures that better customers
experience and service is maintained. From the results, it is clear that the model can pull out 78.6
% of the churners. Therefore, if TESCO will choose to use this model for predicting the churns
in the pull of customers, it is possible that they will a significant number of customers (Stock &
Bednarek, 2014). The number of churns that they will get will be sufficient to ensure significant
profits (Shestakov, et al., 2010).
The comparion of the RFM and CRT reveals that FRM will yield higher income compared to the
random (test) group of customer. Therefore, it would be prudent of the management of TESCO
to use RFM (Setiawan & Budi, 2014). This will ensure that they manage a large pull of
customers so that they keep at par with the competitors. Similarly, it will enable them to easily
pull out their customers from the large pool of customers for better customer experience
(Otterbring, et al., 2018).
Recommendations
It is recommended that TESCO does further research on the distribution of their customers. The
research should focus on comparing the customer bases and income. Moreover, the research
should focus on evaluating the level of customer satisfaction and experience in those areas
(Nahm, 2013). Conducting such studies will enable TESCO to determine the best ways of
distributing their supplies (Lee, et al., 2015). Similarly, the study will help TESCO to make
necessary improvements for better and sustainable customer experience (Kapustina &
Babenkova, 2010).

It is recommended that TESCO should conduct proper market research aimed at determining the
level of customer satisfaction of their competitors (Jacobson, et al., 2009). The research will help
TESCO to know their position in terms of customer experience that they offer (Helia, et al.,
2018). Therefore, by conducting the research, TESCO will be at a point of making necessary
adjustments in terms of their customer service and experience (Hur, et al., 2015). Moreover, the
research will be crucial in ensuring the sustainability of TECO public limited company (plc)
(Bueshken, 2009).

References
Boos, Dennis & Stefanski, L. A., 2013. Roles of Modeling in Statistical Inference. Essential
Statistical Inference, Volume 120.
Bourne & Paul, A., 2016. Customer Satisfaction of Policing the Jamaican Society: Using
SERVQUAL to Evaluate Customer Satisfaction. Journal of Healthcare Communications, 1(03).
Bueshken, J., 2009. Does Improving Customer Satisfaction Really Increase the Market Value of
Equity? - Revisiting the ACSI Customer Satisfaction Data. SSRN Electronic Journal.
Cheng, et al., 2009. The Customer Satisfaction Matrix: A Method to Analyze, Evaluate and
Improve Customer Satisfaction. international Conference on Management and Service Science.
Chesnokova, A. V., Barvenko, O. G. & Radina, O. I., 2013. Identification Mechanisms of
Contact Points with a client as Tools of Improvement of Customer satisfaction. 25(02).
Deng, W. J., Yeh, M. L. & Sung, M. L., 2013. A customer satisfaction index model for
international tourist hotels: Integrating consumption emotions into the American Customer
Satisfaction Index. International Journal of Hospitality Management, 35(12).
Hasnelly & Eddy, Y., 2012. Analysis of Market-Based Approach on the Customer Value and
Customer Satisfaction and Its Implication on Customer Loyalty of Organic Products in
Indonesia. Volume 40, p. 1.
Helia, et al., 2018. Analysis of customer satisfaction in the hospital by using Importance-
Performance Analysis (IPA) and Customer Satisfaction Index (CSI). MATEC Web of
Conferences, Volume 154.
Hur, et al., 2015. Customer response to employee emotional labor: the structural relationship
between emotional labor, job satisfaction, and customer satisfaction. Journal of Services
Marketing, 29(1).
Jacobson, Robert, Mizik & Natlie, 2009. The Financial Markets and Customer Satisfaction:
Reexamining Possible Financial Market Mispricing of Customer Satisfaction. Journal of
Marketing Science, 09(05).
Jung, H. S. & Yoon, H. H., 2013. Do employees’ satisfied customers respond with a satisfactory
relationship? The effects of employees’ satisfaction on customers’ satisfaction and loyalty in a
family restaurant. International Journal of Hospitality Management, 34(09).
Kapustina, L. M. & Babenkova, A. V., 2010. Assessment of Business Customers Satisfaction
with the Products and Service of Pnevmostroimashina, JSC, on the B2B Market. Issue 4.
Lee, et al., 2015. Does advertising exposure prior to customer satisfaction survey enhance
customer satisfaction ratings?. Journal of Marketing Letters, 26(4).
Nahm, Y.-. E., 2013. A novel approach to prioritize customer requirements in QFD based on
customer satisfaction function for customer-oriented product design. Journal of Mechanical
Science and Technology, 27(12).

Netemeyer, R. G., Maxham, J. G. & Lichtesnstein, D. R., 2010. Store manager performance and
satisfaction: Effects on store employee performance and satisfaction, store customer satisfaction
and store customer spending growth. Journal of Applied Psychology, 95(03).
Otterbring, Tabias, Lu & Chaoren, 2018. Clothes, condoms, and customer satisfaction: The
effect of employee mere presence on customer satisfaction depends on the shopping situation.
Journal of Psychology & Marketing, Issue 04.
Reyes, J.-. R., Guillermo, L. & Alfredo, C.-. S., 2010. Teaching undergraduate students to model
use cases using tree diagram concepts. 18(01).
Schafer, C. M., 2015. A Framework for Statistical Inference in Astrophysics. Annual Review of
Statistics and Its Application, 02(01).
Setiawan & Budi, 2014. Customer Satisfaction Index Model on Three Level Of Socioeconomic
Status In Bogor Case Study: Customer Satisfaction on Branded Cooking Oil Product. ASEAN
Marketing Journal, 06(1).
Shestakov, A. L., Sidorov, A. I., Shefer, L. A. & Gickina, E. V., 2010. A quality management
system for an institution of higher education: analysis of customer satisfaction. Volume 08.
Stock, R. M. & Bednarek, M., 2014. As they sow, so shall they reap: customers’ influence on
customer satisfaction at the customer interface. Journal of the Academy of Marketing Science,
Volume 42.
Sun, Kyung-A, Kim & Dae- Young, 2013. Does customer satisfaction increase firm
performance? An application of the American Customer Satisfaction Index (ACSI).
International Journal of Hospitality Management, 35(12).
Vermeeren, B., Kuipers, B. & Steijn, B., 2011. Two Faces of the Satisfaction Mirror: A Study of
Work Environment, Job Satisfaction, and Customer Satisfaction in Dutch Municipalities. Review
of Public Personnel Administration, 31(02).
Yin, Y., 2012. Using Tree Diagrams as an Assessment Tool in Statistics Education. Journal of
Educational Assessment, 17(1).

Appendix
Appendix 1: Tree Tables
Tree Table
Node 0 1 Total Predicted
Category
Parent Node Prim
N Percent N Percent N Percent Variable
0 11442 57.2% 8558 42.8% 20000 100.0% 0
1 3461 34.1% 6689 65.9% 10150 50.8% 1 0 Purchase
2 7981 81.0% 1869 19.0% 9850 49.2% 0 0 Purchase
3 1662 23.4% 5429 76.6% 7091 35.5% 1 1 T.last
4 1799 58.8% 1260 41.2% 3059 15.3% 0 1 T.last
5 712 41.8% 990 58.2% 1702 8.5% 1 2 T.last
6 7269 89.2% 879 10.8% 8148 40.7% 0 2 T.last
7 868 18.7% 3784 81.3% 4652 23.3% 1 3 T.active
8 794 32.6% 1645 67.4% 2439 12.2% 1 3 T.active
9 1023 52.1% 941 47.9% 1964 9.8% 0 4 T.last
10 776 70.9% 319 29.1% 1095 5.5% 0 4 T.last
11 132 23.8% 422 76.2% 554 2.8% 1 5 T.last
12 580 50.5% 568 49.5% 1148 5.7% 0 5 T.last
13 1394 76.1% 438 23.9% 1832 9.2% 0 6 T.last
14 5875 93.0% 441 7.0% 6316 31.6% 0 6 T.last
15 610 15.4% 3357 84.6% 3967 19.8% 1 7 T.last
16 258 37.7% 427 62.3% 685 3.4% 1 7 T.last
17 377 28.5% 945 71.5% 1322 6.6% 1 8 Loyalty
18 417 37.3% 700 62.7% 1117 5.6% 1 8 Loyalty
19 526 59.7% 355 40.3% 881 4.4% 0 9 T.active
20 497 45.9% 586 54.1% 1083 5.4% 1 9 T.active
21 430 63.7% 245 36.3% 675 3.4% 0 10 GR.spent
22 346 82.4% 74 17.6% 420 2.1% 0 10 GR.spent
23 38 13.2% 249 86.8% 287 1.4% 1 11 Loyalty
24 94 35.2% 173 64.8% 267 1.3% 1 11 Loyalty
25 181 70.4% 76 29.6% 257 1.3% 0 12 T.active
26 399 44.8% 492 55.2% 891 4.5% 1 12 T.active
27 944 86.1% 152 13.9% 1096 5.5% 0 13 T.active
28 450 61.1% 286 38.9% 736 3.7% 0 13 T.active
29 1950 87.5% 279 12.5% 2229 11.1% 0 14 T.last
30 3925 96.0% 162 4.0% 4087 20.4% 0 14 T.last
31 269 11.2% 2141 88.8% 2410 12.0% 1 15 Loyalty
32 341 21.9% 1216 78.1% 1557 7.8% 1 15 Loyalty
33 96 27.7% 251 72.3% 347 1.7% 1 16 Loyalty

34 162 47.9% 176 52.1% 338 1.7% 1 16 Loyalty
35 187 34.6% 353 65.4% 540 2.7% 1 17 T.last
36 190 24.3% 592 75.7% 782 3.9% 1 17 T.last
37 227 51.9% 210 48.1% 437 2.2% 0 19 T.last
38 299 67.3% 145 32.7% 444 2.2% 0 19 T.last
39 204 38.3% 328 61.7% 532 2.7% 1 20 Loyalty
40 293 53.2% 258 46.8% 551 2.8% 0 20 Loyalty
41 333 60.2% 220 39.8% 553 2.8% 0 21 T.last
42 97 79.5% 25 20.5% 122 0.6% 0 21 T.last
43 121 74.7% 41 25.3% 162 0.8% 0 22 Total Prof
44 225 87.2% 33 12.8% 258 1.3% 0 22 Total Prof
45 64 58.2% 46 41.8% 110 0.5% 0 25 T.last
46 117 79.6% 30 20.4% 147 0.7% 0 25 T.last
47 205 39.1% 319 60.9% 524 2.6% 1 26 T.last
48 194 52.9% 173 47.1% 367 1.8% 0 26 T.last
49 288 78.5% 79 21.5% 367 1.8% 0 27 Purchase
50 656 90.0% 73 10.0% 729 3.6% 0 27 Purchase
51 312 56.4% 241 43.6% 553 2.8% 0 28 T.last
52 138 75.4% 45 24.6% 183 0.9% 0 28 T.last
53 1298 93.2% 94 6.8% 1392 7.0% 0 29 T.active
54 652 77.9% 185 22.1% 837 4.2% 0 29 T.active
55 901 90.7% 92 9.3% 993 5.0% 0 30 GR.spent
56 3024 97.7% 70 2.3% 3094 15.5% 0 30 GR.spent
Growing Method: CRT
Dependent Variable: churn
Appendix 2: Surrogates
Surrogates
Parent Node Independent Variable Improvement Association
0 Primary Purchase .110
Surrogate Total Profit .110 .799
GR.spent .107 .747
DY.spent .095 .738
T.last .154 .666
BH.spent .062 .583
FV.spent .051 .543
MT.spent .045 .497

DL.spent .032 .425
T.active .035 .399
GM.spent .033 .359
Service Failure .028 .330
LQ.spent .006 .120
Loyalty .013 .087
AP.spent .006 .081
Socio.Economic .000 .017
1
Primary T.last .027
Surrogate
Purchase .014 .298
Total Profit .012 .198
GR.spent .009 .166
DY.spent .008 .144
FV.spent .003 .048
BH.spent .003 .042
MT.spent .002 .035
Service Failure .001 .023
DL.spent .001 .011
LQ.spent .001 .009
GM.spent .000 .009
AP.spent .000 .002
3
Primary T.active .003
Surrogate
Total Profit .001 .059
T.last .002 .057
Purchase .001 .051
GR.spent .001 .043
DY.spent .001 .035
MT.spent .000 .018
FV.spent .000 .012
BH.spent 5.999E-005 .010
AP.spent 4.802E-005 .009
DL.spent 5.522E-006 .004
Service Failure 3.299E-005 .002
GM.spent .000 .001
Socio.Economic 1.649E-005 .000
7 Primary T.last .003
Surrogate Purchase .001 .191
Total Profit .001 .098
DY.spent .001 .080

GR.spent .001 .058
MT.spent .000 .019
Service Failure .000 .018
FV.spent 4.506E-005 .016
BH.spent .000 .009
LQ.spent 6.465E-006 .001
15
Primary Loyalty .001
Surrogate
Service Failure 3.857E-005 .004
DL.spent 4.017E-007 .004
FV.spent 8.679E-007 .002
DY.spent 3.277E-005 .001
Total Profit 4.731E-006 .001
LQ.spent 7.163E-005 .001
GR.spent 4.686E-006 .001
AP.spent 2.365E-006 .001
16
Primary Loyalty .001
Surrogate
DY.spent 6.705E-005 .124
DL.spent 5.671E-005 .115
Socio.Economic .000 .104
BH.spent 6.584E-005 .095
FV.spent 4.080E-006 .089
Total Profit .000 .080
GM.spent 1.732E-005 .077
T.active .000 .044
T.last 8.245E-005 .038
LQ.spent 2.018E-005 .030
MT.spent 8.903E-006 .024
Service Failure 2.170E-008 .024
GR.spent 3.302E-005 .018
Purchase 1.020E-006 .009
AP.spent 5.652E-007 .003
8 Primary Loyalty .000
Surrogate Socio.Economic 1.978E-009 .053
Total Profit 9.287E-007 .018
GR.spent 3.282E-006 .017
AP.spent 3.990E-005 .015
DY.spent 2.150E-006 .013
Service Failure 6.910E-006 .013
GM.spent 2.393E-006 .012
DL.spent 1.430E-005 .008

FV.spent 2.283E-005 .007
BH.spent 1.963E-006 .005
MT.spent 1.277E-005 .004
T.last 2.782E-006 .004
17
Primary T.last .000
Surrogate
Purchase .000 .943
Total Profit 5.986E-005 .561
GR.spent 3.842E-009 .493
DY.spent 1.481E-005 .411
BH.spent 1.967E-005 .146
MT.spent 2.172E-006 .113
FV.spent 1.257E-007 .111
DL.spent 1.486E-005 .020
Socio.Economic .000 .007
4
Primary T.last .002
Surrogate
GR.spent .001 .017
Total Profit .001 .015
DY.spent .000 .009
DL.spent 3.179E-005 .006
BH.spent .000 .005
MT.spent .000 .005
Service Failure 1.854E-006 .003
GM.spent 1.195E-005 .002
LQ.spent 7.787E-006 .002
T.active 4.941E-006 .002
FV.spent 5.572E-006 .001
Socio.Economic 1.854E-006 .001
9
Primary T.active .001
Surrogate
Socio.Economic 3.271E-005 .008
BH.spent 6.483E-005 .007
GM.spent 8.277E-006 .007
LQ.spent 4.408E-005 .006
DL.spent 2.714E-005 .001
Total Profit 2.297E-005 .001
FV.spent 2.297E-005 .001
GR.spent 2.297E-005 .001
DY.spent 1.737E-007 .001
19 Primary T.last .001
Surrogate Purchase .000 .117
BH.spent 2.244E-005 .096

Total Profit .000 .076
GR.spent 5.843E-005 .076
Socio.Economic .000 .064
T.active 1.616E-005 .053
DY.spent 9.067E-006 .050
FV.spent 3.596E-006 .046
DL.spent 3.382E-010 .039
MT.spent 2.787E-009 .021
GM.spent 9.779E-006 .016
LQ.spent 2.127E-006 .011
Service Failure 2.928E-006 .009
20
Primary Loyalty .001
Surrogate
DY.spent .000 .113
Socio.Economic 1.823E-005 .098
DL.spent 5.523E-007 .094
T.active 5.761E-005 .064
BH.spent 2.715E-006 .062
Total Profit .000 .055
T.last .000 .038
GR.spent 7.911E-005 .038
Service Failure 3.508E-009 .034
MT.spent 3.247E-005 .030
FV.spent 2.552E-006 .030
AP.spent 2.833E-005 .024
LQ.spent 3.285E-005 .017
GM.spent 2.404E-006 .013
Purchase 1.408E-006 .013
10 Primary GR.spent .001
Surrogate Total Profit .001 .707
DY.spent .000 .329
Purchase .000 .110
MT.spent 5.870E-005 .102
BH.spent .000 .086
LQ.spent 3.503E-005 .062
DL.spent 5.944E-005 .055
GM.spent 1.201E-005 .050
FV.spent 8.237E-005 .048
AP.spent 6.066E-006 .021
T.last 6.048E-005 .012
Socio.Economic 1.083E-006 .007

Service Failure 2.947E-005 .002
21
Primary T.last .000
Surrogate Total Profit 3.970E-005 .008
LQ.spent 1.319E-005 .008
22
Primary Total Profit .000
Surrogate
GR.spent 8.878E-005 .531
DY.spent 1.252E-006 .265
BH.spent 2.413E-005 .080
Purchase 8.006E-005 .074
Socio.Economic 7.354E-005 .031
T.last 2.857E-006 .031
Service Failure 9.380E-006 .006
AP.spent 2.869E-007 .006
2
Primary T.last .032
Surrogate
T.active 2.076E-005 .028
GR.spent .002 .022
Total Profit .002 .013
5
Primary T.last .003
Surrogate
Total Profit 3.424E-005 .011
GR.spent 1.789E-005 .005
GM.spent 3.424E-005 .004
LQ.spent 3.385E-005 .002
BH.spent 5.259E-005 .002
AP.spent 1.751E-005 .002
11
Primary Loyalty .001
Surrogate
Socio.Economic 2.301E-005 .120
GR.spent .000 .082
Total Profit 1.326E-005 .067
GM.spent 2.045E-006 .064
DY.spent 1.108E-007 .045
MT.spent 8.664E-005 .041
T.active 1.913E-005 .041
BH.spent 4.301E-007 .034
Purchase 3.012E-006 .019
Service Failure 7.389E-007 .019
DL.spent 6.064E-008 .019
FV.spent 2.362E-005 .011
AP.spent 1.182E-005 .011
T.last 1.622E-006 .011
LQ.spent 3.103E-006 .004

12 Primary T.active .001
Surrogate BH.spent 7.839E-006 .004
25
Primary T.last .000
Surrogate
DY.spent 2.175E-006 .064
Total Profit 3.653E-005 .045
Socio.Economic 7.594E-006 .045
Purchase 9.793E-005 .036
GM.spent 2.613E-005 .036
GR.spent .000 .027
T.active 6.552E-007 .027
FV.spent 1.680E-005 .018
BH.spent 1.763E-005 .018
Service Failure 4.979E-005 .009
DL.spent 3.074E-005 .009
MT.spent 4.294E-007 .009
26
Primary T.last .000
Surrogate
Purchase .000 .049
GR.spent 7.259E-008 .019
DY.spent 1.983E-005 .019
DL.spent .000 .011
Total Profit 3.494E-005 .008
AP.spent 5.752E-006 .008
FV.spent 5.888E-005 .005
MT.spent 1.572E-005 .005
Service Failure 2.199E-006 .005
T.active 2.008E-005 .003
LQ.spent 1.055E-005 .003
BH.spent 1.031E-008 .003
6
Primary T.last .004
Surrogate
T.active 2.018E-007 .397
GR.spent 9.972E-005 .002
Total Profit 1.164E-006 .001
13 Primary T.active .003
Surrogate T.last 5.409E-005 .030
AP.spent .000 .011
Service Failure 1.484E-005 .005
Total Profit 5.719E-006 .001
GR.spent 5.719E-006 .001
DY.spent 2.669E-006 .001
LQ.spent 5.793E-005 .001

BH.spent 2.669E-006 .001
27
Primary Purchase .000
Surrogate
Total Profit 8.678E-005 .286
GR.spent 8.531E-005 .223
DY.spent 9.521E-005 .204
T.active .000 .041
BH.spent 1.550E-005 .038
T.last 3.659E-005 .014
FV.spent 2.383E-005 .003
LQ.spent 1.140E-005 .003
28
Primary T.last .000
Surrogate
Purchase 4.992E-006 .011
Total Profit 1.512E-005 .005
AP.spent 1.512E-005 .005
MT.spent 1.512E-005 .005
14
Primary T.last .001
Surrogate
T.active 2.027E-005 .412
Purchase .001 .212
Total Profit .001 .117
GR.spent .001 .087
DY.spent .000 .069
BH.spent 4.876E-007 .000
29
Primary T.active .001
Surrogate
T.last 9.628E-008 .012
DL.spent 8.189E-006 .007
Total Profit 2.806E-006 .004
Purchase 2.813E-005 .002
DY.spent 4.561E-011 .002
GR.spent 2.813E-005 .002
GM.spent 7.851E-006 .001
AP.spent 1.567E-006 .001
FV.spent 8.808E-005 .001
30
Primary GR.spent .000
Surrogate
Total Profit .000 .632
DY.spent .000 .277
Purchase .000 .245
T.active 1.102E-006 .001
Growing Method: CRT
Dependent Variable: churn