See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/262949102

A new model for assessment fast food customer behavior case study: An
Iranian fast-food restaurant

Article  in  British Food Journal · April 2013

DOI: 10.1108/00070701311317874

CITATIONS READS

22 3,516

3 authors:

Niyoosha Jafari Somayeh Alizadeh

Khaje Nasir Toosi University of Technology Khaje Nasir Toosi University of Technology
5 PUBLICATIONS   58 CITATIONS    51 PUBLICATIONS   987 CITATIONS   

SEE PROFILE SEE PROFILE

Mahyar Sharif Vaghefi

9 PUBLICATIONS   55 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Knowledge Extraction of Diabetics' Data by Decision Tree Method View project

All content following this page was uploaded by Niyoosha Jafari on 08 June 2019.

The user has requested enhancement of the downloaded file.

 The current issue and full text archive of this journal is available at
www.emeraldinsight.com/0007-070X.htm

Fast food
A new model for assessment fast customer
food customer behavior case behavior
study
601
An Iranian fast-food restaurant
Niyoosha Jafari Momtaz, Somayeh Alizadeh and
Mahyar Sharif Vaghefi
Department of Industrial Engineering, K.N. Toosi University of Technology,
Tehran, Iran
Abstract
Purpose – Nowadays, because of more availability of products, there is an increasing need for
companies to establish a strong relationship with their customers. As the fast food industry is not an
exception and has a competitive environment, analyzing customers’ behavior helps bridge this gap.
Data mining techniques help to segment customers as well as to drive improved customer relationship
management. This paper seeks to address these issues.
Design/methodology/approach – This study proposes a new model based on RFM model for
defining customers’ value as well as using K-means algorithm to segment restaurants’ customers. In
addition, the authors combine a new category in the account portfolio analysis in order to analyze the
behavior of each cluster.
Findings – A real dataset of an Iranian fast food restaurant chain is employed to show the procedure
of the authors’ model. The customers are segmented into four clusters. The clusters are analyzed and
named based on categories in the account portfolio analysis. The result of this analysis shows that
there is no significant difference between the behavior of the most valuable customer and customers
who have left the restaurant. Therefore, restaurant managers should seek other reasons for detecting
churn behavior.
Originality/value – This paper helps managers in the fast food industry to readily analyze their
customer behavior in order to understand their needs and establish strong relationships.
Keywords Customer relationship management, K-means algorithm, RFM model, Fast food restaurant,
Customer satisfaction, Fast foods, Iran, Consumer behaviour
Paper type Research paper

1. Introduction
In recent years, customers in different industries have become the most important asset
in each business unit. So an effective key for organizations to get excelled is developing
long-term relationships with customers. The consumer behavior trends have changed
recently. The three important changes include an increased need for individual
solutions, more attention to health and product availability (Bijnen et al., 2002). These
new trends made supplier to change the supply chain. In fact, these changes, coupled
with an increasing level of competition, reveal that new customizations for each group
of customers are needed. One of the industries that have recently changed a lot is fast
food industry. While customers become intensively health-conscious and different fast British Food Journal
Vol. 115 No. 4, 2013
pp. 601-613
q Emerald Group Publishing Limited
The authors would like to thank the managers of Shila restaurant for their guidance during the 0007-070X
project and making available the data. DOI 10.1108/00070701311317874
BFJ food restaurants are more available, food industry operators have to change their
115,4 strategies. Increasing competition in an already struggling market causes fast food
restaurants to plan for customized solutions in order to achieve customer satisfaction
and prevent customers from eating at other rival restaurants. In the food industry, the
concern is on identifying customers’ characteristics and needs in different segments
and offering special options related to each segment (Johnsa and Pineb, 2002).
602 Many organizations, including fast food restaurants, collect and store a wealth of
data about their current customers. However, most of them are unable to transform
raw data into valuable and useful knowledge and discover valuable information
hidden in them (Berson et al., 2000). Data mining is the process of extracting and
detecting hidden patterns or knowledge from the enormous amount of data using
statistical, mathematical, artificial intelligence and machine-learning techniques
(Turban et al., 2007). The common theme in all definitions of data mining is extracting
and identifying hidden knowledge from large databases (Turban et al., 2007; Ahmed,
2004; Berry and Linoff, 2004). This popular means by analyzing customer data could
help organizations to discover useful information and subsequently employ this
knowledge in decision-making and generating new opportunities based on customer
characteristics.
The aim of this research is to provide a model to analyze customer behavior in food
industry by using data mining tools. The structure of this paper is organized as
follows: In the next three sections, a literature review including Customer Relationship
Management, RFM analysis and K-means algorithm is presented. Section 2 explains
the model of this research and in the third section in a case study the behavior of
customers in a branch of a fast food chain is analyzed. At the end, the conclusion and
our future work are discussed.

1.1 The CRM concept

Customer Relationship Management is the process of designing an organization around
its customers, so that customer’s needs become the main focus of each organization in
decision-making process. Interest in Customer Relationship Management began to grow
in 1990s. CRM introduced a view that a group of processes construct a strong long-term,
lucrative relationship with particular customers (Ling and Yen, 2001; Xu et al., 2002). An
improved relationship with a customer can eventually lead to greater customer loyalty,
maintenance and profitability. In fact, the old product-oriented view is being replaced by
a customer-oriented view.
Although CRM has been known as one of the most significant approaches used in
business, there does not seem to exist any comprehensive definition for it (Ling and Yen,
2001; Ngai, 2005). For example, CRM is defined as an ‘‘enterprise approach to
understanding and influencing customer behavior through meaningful communications
in order to improve customer acquisition, customer retention, customer loyalty, and
customer profitability” (Swift, 2001). Furthermore, in another definition, CRM is ‘‘the
strategic use of information, processes, technology, and people to manage the customer’s
relationship with your company (Marketing, Sales, Services, and Support) across the
whole customer life cycle” (Kincaid, 2003). Parvatiyar and Sheth (2001) define CRM as ‘‘a
comprehensive strategy and process of acquiring, retaining, and partnering with
selective customers to create superior value for the company and the customer. It
involves the integration of marketing, sales, customer service, and the supply chain
functions of the organization to achieve greater efficiencies and effectiveness in Fast food
delivering customer value”. These definitions seem to show that CRM should be paid customer
more attention as a whole process of managing the customers.
The CRM can be divided into operational and analytical parts (Berson et al., 2000; behavior
He et al., 2004; Teo et al., 2006). Operational CRM is related to automation of the
business process, while analytical CRM is related to the analysis of customer
characteristics in order to help organizations perform their customer management 603
strategies more efficiently. In addition, analytical CRM helps organizations to identify
their most valuable customers and allocate resources between them more practically.
CLV is a concept in CRM that can be used to facilitate this identification. To an
organization, CLV is the present values of all future profit and loss, which is gained
from a customer during their entire purchasing lifetime (Berger and Nasr, 1998). To
sum up, CLV has been playing a vital role in CRM modeling.

1.2 RFM
RFM analysis proposed by Hughes in 1994 is one way that can be useful in CLV
concept (Hughes, 1994). This analytical model has been used for several decades in
direct marketing (Baier, 2002). This method categorizes customer behavior and
identifies important customers according to three variables as follows:
(1) Recency: This is the last date of customer’s purchase.
(2) Frequency: This is the number of customer’s purchases or transactions.
(3) Monetary: This is the value a customer has paid for its purchases so far.

Using RFM analysis, it is possible to assign a RFM score to each customer, based on
their behavior and subsequently rank them.

1.3 K-means algorithm

Clustering is the process of segmenting an incongruous population of physical or abstract
objects into groups of similar objects (Ahmed, 2004; Berry and Linoff, 2004). A cluster is a
collection of data objects that have more similarity to one another within the same cluster
and less dissimilarity to the objects in other clusters (Han and Kamber, 2006).
K-means is one of the simple and famous algorithms for clustering used in this
study to build clusters by attributes (i.e. RFM scores). This algorithm introduced by
MacQueen is based on the mean value of the objects in the cluster and assigns each
item to the cluster with the nearest centroid (mean) (MacQueen, 1967). The user should
pass the number of clusters (K) as an input to K-means algorithm.

2. Fast food customer behavior assessment model

In this section, the research model of this study for analyzing customer behavior in fast
food restaurants is presented. This model is based on using data mining techniques to
analyze customer’s value and customer’s behavior.
As mentioned before, Customer Relationship Management (CRM) is the process of
designing an organization around its customers, so the customers’ needs are
considered as the main focus of each organization in decision-making process. In the
food industry, customers are considered as the most important assets. As a result,
retaining them because of special circumstances in this industry is much harder than
retaining them in other industries. In this study, a special clustering of customers in the
BFJ form of a model is presented to analyze customer behavior by using data mining
115,4 techniques. The main aim of this model is to help restaurants owners better understand
their customers’ needs and address them. So although this model is applied in an
Iranian fast food restaurant chain, it could be helpful for all types of restaurants.
To direct the process of data mining, a common standard is developed by experts
from the data mining industry. This model is called CRoss Industry Standard Practice
604 (CRISP) (Han and Kamber, 2006) and describes the life cycle of every data-mining project
through six phases: business understanding, data understanding, data preparation,
modeling, evaluation and deployment. The main phases of our methodology are
developed based on CRISP methodology to help restaurants owners benefit from data
mining techniques and better understand their customer’s needs. The main phases of
this model include: data preprocess, modeling, and analyzing the model output. Since the
purpose of this research is providing some clusters of customers in fast food restaurants,
first, customers are clustered based on RFM concept by K-means algorithm; Second,
each cluster is labeled based on special categories in the concept of customers’ account
portfolio; finally, for each cluster, customers’ behaviors are analyzed. Figure 1 shows the
main phases as well as detailed steps of our model.

2.1 Data gathering and data preprocessing

At first business has to be understood and all related databases recognized. Preprocessing
is the first phase of data mining project. The data in the databases usually have not been
preprocessed and are incomplete and noisy. It is estimated that preprocessing takes about
60 percent of the whole data-mining project (Han and Kamber, 2006). This part includes
three steps, i.e. data cleaning, data conversion and data integration.
In most restaurants, there are two categories of customers. The first category includes
customers who directly buy from the restaurant. As this group does not subscribe, its

Figure 1.
Fast food customer
behavior assessment
model
effective information is not saved and just some simple data about purchase transaction Fast food
like time and date of purchase or food type are saved. This data are isolated and could not customer
be traced. The second group is customers who have already subscribed. These customers
order their food through their subscription ID. For this group, besides data about behavior
purchases, some personal information is saved, so that restaurants can analyze these
customers’ behavior. For example, they can find the most popular food for each group of
customers. Hence, customer behavior analysis is possible just for the second group. 605
One important factor that could change the consumer behavior analysis in the fast
food restaurant is the date of subscription. Because older customers could have a
higher number of purchases and as a result more monitory, their RFM score is higher
than new customers. Therefore, customers should be divided into two groups
according to their membership date to analyze separately. Based on this division, each
customer is labeled with one of the two labels, “old” or “new”. This labeling would be
essential for RFM analysis, which will be explained in the next section.

2.2 RFM parameters value (Score) to customers

In this step, several useful fields should be chosen from those data that have been
passed through preprocessing step, for ranking customers. As mentioned before, RFM
analysis is a method that is typically used to rank customers due to analyze their
behavior. This method uses three factors, which are: recency of the last purchase,
frequency of the purchases and monetary value of purchases.
In each industry with its exclusive characteristics and according to different data
that is saved, various approaches are used to rate and weigh three well-known RFM
factors. Typically, these factors are divided into the appropriate number of ranges.
Then, each range receives a RFM score, according to its value in the industry. RFM
factors could have different weights according to the special type of industry and the
importance of each factor in that industry. However, there are some scientists such as
Hughes (1994), who believe there should not be a weighing system for these variables
and all of them should be at the same level.
Using traditional RFM values for analyzing customers’ behavior is not enough in
this industry, since one important factor changing the customer values is customers’
membership date. For this reason, besides the three features of frequency, monetary
and recency, customers could also be analyzed based on the old or new labels they have
received from the preprocess step.

2.3 Clustering customers value with K-means

Responding to the different necessities of customers, there is a need to cluster
customers and identifying the behavior of each segment to provide a customized
strategy. Accordingly, the K-means algorithm based on the three values identified in
the RFM analysis would be used in clustering. In accordance with this clustering,
customers with similar behaviors would be placed into the same cluster.
Output and quality of the clusters in K-means algorithm largely depend on selecting
the proper number of clusters. For selecting the optimal number of clusters, there are
many indicators such as Davies– Bouldin Index (Davies and Bouldin, 1979), the
Silhouette Width (Rousseeuw, 1987) or Dunn Index (Dunn, 1974).
The aim of all these indexes is to have meaningful clusters where the data objects
within the same cluster are similar to one another and dissimilar to the objects in other
BFJ clusters (Bijnen et al., 2002). In this research, we evaluate the optimum number of K
115,4 based on Dunn Index in order to use in K-means algorithm as the number of clusters.

2.4 Labeling customers clusters with typical labels in customer account portfolio
In order to manage customers, the customer behavior in each cluster should be
analyzed. In this step the cluster map with different categories extracted customers’
606 account portfolio. According to individual expert opinion, no matter how many clusters
we have, these categories will match with the customers’ behavior in the food industry.
So, in our model we produce some cluster based on customer behavior, and map this
cluster with customers’ account portfolio categories (Gök, 2009).

2.5 Analyzing each segment using customer purchase behavior

The aim of clustering is to segment customers into some groups to study their behavior
in order to plan appropriate strategies for each cluster. According to this, each cluster
should be analyzed based on various factors. One of the most important things in the
food industry, which has to be recognized carefully, is peak times. Busy days and hours
are very important and analyzing this related information helps restaurant managers to
overcome these difficulties. For example, a proper decision would be to increase the
number of staff in peak times to deliver services on time. Another factor that plays an
important role in achieving competitive advantages among rivals is to know which type
of foods has the most popularity as well as which foods are eaten together.

3. Case study
Fast food restaurants have an old history in Iran. Popular fast food items in Iran
include “joojeh kabab” (grilled or broiled chicken), “naan o kabab” (literally bread with
kabab), “kabab sandwiches”, and a number of different derivatives of traditional
slow-cooked meals. However, the impact of American style food, especially among a
younger generation, is undeniable. Today, fast food restaurants in Iran serve different
foods, including pizza, hamburgers, chicken burgers, etc. First fast food restaurant
chain was founded in 1995. Then, fast food chains grow and now there are almost 20
fast food chains in Iran.
In this section, the developed model is implemented in a fast food restaurant chain in
Iran named Shila (founded in 2002). Shila aims at providing healthy foods as well as
focusing on customer relationship management. Shila restaurant has more than 16
branches in Iran. We have worked on the Tajrish branch, which is the most successful
branch with the highest annual sell, and we are planning to work on the other branches
in our future researches. This branch was chosen because it is located in a competitive
region; it also covers both home and office customers. This case study shows how our
model could be applied in analyzing customers’ behavior. In this paper SPSS Clementine
which is a mature data mining toolkit has been used to analyze the restaurant’s data and
follow the model’s steps. The computing process can be expressed in detail as follows.

3.1 Step 1: data gathering and data preprocess

In the Tajrish branch, 5,400 customers have subscribed since 2005. To begin with,
two tables from existing tables in the database are selected. The first table
represents each customer feature, and the second table includes information of all
customers’ purchase transactions. Then, as the raw data includes noisy and
incorrect information, some records containing inaccurate values and missing data Fast food
are deleted. Moreover, redundant attributes are eliminated. In addition, the date of customer
purchase, which is shown in three columns (for day, month and year), is changed
into suitable formats in one column. In the first table, date of last purchase, date of behavior
receiving subscription, total amount of money paid by customers in all purchases,
and number of purchases are selected. Also, in the second table, some features like
time and day of buying and goods purchased in each purchase are identified as the 607
main features. These critical operations caused to delete at least 650 customers.
In the second stage of our model, customers are divided into two clusters according
to the date of receiving subscription by K-means algorithm. As a result, new customers
are those who have received their subscription in the past 13 months, and the second
cluster encompasses old customers who have received their subscription before these
13 months. Thus, all customers at this stage are labeled with old and new customers.
This segmentation helps us focus on customer behavior precisely.

3.2 Step 2: assigning value to customers using RFM concept

In this step, the following three fields are selected to rank customers based on RFM
analysis:
(1) last purchasing date; which refers to the recency;
(2) total amount of money paid by customers in all purchases which refers to the
monetary; and
(3) number of purchases; which refers to the frequency.

However, frequency is dependent on the number of purchases and customers with

longer duration of subscription have higher frequency. So, evaluating old and new
customers is done separately to allow both time durations of customer’s subscription
and the number of purchase effect on the analysis separately.
These three fields, with respect to customer’s behavior in this industry, have been
classified and shown in Table I.
Accordingly, each customer receives a score between 3 and 15, where the higher
score indicates a higher value for the restaurant.

3.3 Step 3: clustering customers’ value with K-means

In this step, customers should be segmented and analyzed according to the behavior of
each cluster to provide a customized strategy. Therefore, the K-means algorithm based
on the values identified in the RFM analysis is used for clustering. Output and quality

Scores Frequency Recency Monetary

5 20#F , 315 2010-02-22#R , 2010-04-24 119150#M , 2133723

4 10#F , 20 2009-08-17#R , 2010-02-22 58400#M , 119150
3 6#F , 10 2008-11-02#R , 2009-08-17 32850#M , 58400
2 4#F , 6 2007-12-14#R , 2008-11-02 18650#M , 32850
1 2#F , 4 2006-10-12#R , 2007-12-14 3500#M , 18650
Note: Accordingly, each customer receives a score between 3 and 15 where the higher score indicates Table I.
a higher value for the restaurant RFM scores
BFJ of the clusters in K-means algorithm largely depend on the proper selection of K.
115,4 Appropriate interval for K is identified by experts and managers of the restaurant
between 2 and 7. Then, for choosing optimal K, in this interval Dunn indicator is used.
Table II shows the Dunn index value for K ¼ 2 to K ¼ 7.
Lower Dunn index indicates good clustering with high similarity within a cluster
and low similarity between clusters. The least occurs in K ¼ 2: Note that by selecting
608 K ¼ 2; customers are divided into two good and bad categories, for choosing optimal
K two clusters are ignored. As shown in Figure 2, the second least Dunn index is
0.516399. Therefore, K ¼ 4 is selected as the optimal number of clusters for K-means
algorithm. As a result, restaurant’s customers are divided into four categories.
Table III shows the specification of these clusters after applying K-means algorithm
to the data, and Figure 3 shows these clusters.

3.4 Step 4: labeling customers’ clusters with typical labels in customer account portfolio
As mentioned before, for better understanding customer’s behavior, some models are
employed to categorize them in some conceptual groups. Here is one of the best and
recent conceptual groups used, and customer clusters map to these four categories
(Gök, 2009). These categories are as follows:
(1) Today’s loyal: these customers are the most valuable and loyal customers.
(2) Competitors’ customers: the customers who have a weak relationship with the
supplier. They could be old valuable customers who have now become rival’s
customers.

K 2 3 4 5 6 7
Table II.
Dunn index Dunn index 0.480185 0.597044 0.516399 0.682628 0.606022 0.73056

Figure 2.
Dunn index

Cluster Frequency Monetary Recency Members number

1 4.453 4.568 4.313 990

2 1.66 1.527 1.64 969
Table III. 3 1.693 2.164 4.048 791
Cluster parameters 4 3.741 3.656 1.984 810
 Fast food
customer
behavior

609

Figure 3.
Fast food customer
clusters

(3) Tomorrow’s loyal: the length of their relationship with the supplier is low, and
they have the potential to show loyal behavior in near future and become
valuable customers.
(4) Open to competitors’ offerings: these customers do not have strong relationship
with the supplier, and they are ready to accept a better suggestion and become a
rival’s customers.

After mapping customer clusters with these four categories, these four clusters are
labeled based on the customer account portfolio categories:
(1) These customers have high frequency of purchases and good volume of money.
Furthermore, they have bought recently and in fact, are the most valuable
customers for the restaurant. Customers of this cluster are called “Today’s
loyal”.
(2) These customers are those who do not have large number of purchases and
because of low recency, it seems that they were not satisfied by restaurant’s
services and have left the restaurant. It can be said that these customers have
preferred rival’s products to our products. Therefore they are called
“Competitors’ customers”.
(3) These customers are those who have not bought in large numbers but during
their purchases, have paid high amount of money. These customers have
bought recently and it seems that they are newly familiarized with the
restaurant. They have an ability to transform to the first category and keeping
them is so important for the restaurant. We can call them “Tomorrow’s loyal”.
(4) These customers have paid a high amount of money, but have not purchased
recently. In fact, this cluster contains old valuable customers who have left the
restaurant, or they are ready to leave us. So, we can call them “Open to
competitors’ offerings”.
BFJ 3.5 Step 5: analyze each segment using customer purchase behavior
115,4 In this step, we analyze the customers’ purchase behavior of clusters in different days
and hours of the week. The results of analysis are shown in Figures 4 and 5.
The striking point is that the purchase’s behavior of all clusters in different hours is
rather identical. The only difference is amount of purchase. Most purchases have been
done by cluster 1, while cluster 3 has done the minimum number of purchases.
610 The results of analysis of different days reveal that although the maximum
purchase of cluster 1, cluster 2 and cluster 3 is on Thursday, there is no important
distinction in purchase behavior of cluster 4 in different days.
In the food industry, the churn behavior of the customers and the reason for this
behavior are very important. On the other hand, keeping valuable customers is
important too, because they can invite other people to try the restaurant. So analyzing the
behavior of cluster 1 (Today’s loyal) and cluster 2 (Competitors’ customers) is very
important. Although, the minimum purchase of cluster 1 and cluster 2 is on Friday, the
maximum purchase in the night is also on Friday. It shows that restaurant has lost a
large number of customers in the Friday noon. The minimum amount of nightly
purchases is on Saturday. In addition, the second least amount of purchases for these two
clusters is on Saturday. Besides, these diagrams show the behavior of Today’s loyal

Figure 4.
Purchase behavior in
different hours
 Fast food
customer
behavior

611

Figure 5.
Purchase behavior in
different days

customers who are in cluster 1 are more or less similar in other days than Friday,
however the Competitors’ customer’s behaviors (cluster 2) are different during the week.
On the other hand, clusters have been analyzed based on the type of foods ordered.
The result shows the foods that have been eaten by all clusters are almost the same.
The most popular drinks and foods eaten by customers of four clusters are shown next:
.
coca;
. fries;
.
mix hot dog;
.
royal hot dog;
.
hot dog with Gouda cheese; and
.
cabbage salad.

One reason of this behavior is that this restaurant have just sandwich and there is no
food diversity. According to the result of our analysis, managers of Shila restaurant are
planning to add five new items including pizza to their menu.
It can thus be concluded that the food is not the reason of customers’ churn. One
important factor that can properly affect in this behavior could be the customers’ distance
from the restaurant, and we are planning to check this factor in our future works.
BFJ 4. Conclusion
115,4 This study has proposed a new model to assess fast food restaurant customers’
behavior. This model uses RFM attributes and K-means algorithm to segment
customers. In the analyzing phase of the model, for defining the behavior of customers
in each cluster we benefit from a new category in account portfolio analysis. A
practical dataset from a branch of a fast food chain in Iran is employed in the empirical
612 case study to illustrate the proposed model. The purchase behavior of customers in
each cluster is analyzed based on the days and hours of the week, as well as their foods.
The results show that there is no significant difference between the behavior of the
most valuable customers and customers who have left the restaurant. Thereupon, the
restaurant should seek other reasons for detecting churn behavior. Two probable
reasons could be the lack of food diversity as well as distance of customers from the
restaurant. For future research, we are planning to study these factors moreover,
investigating other branches and compared with each other.

References
Ahmed, S.R. (2004), “Applications of data mining in retail business”, in Information Technology:
Coding and Computing Proceedings of the International Conference, Vol. 2, pp. 455-459.
Baier, M., Ruf, K.M. and Chakraborty, G. (2002), “Contemporary data base marketing: concepts
and applications”, Direct Marketing: An International Journal, Vol. 2 No. 3, p. 186187.
Berger, P.D. and Nasr, N.I. (1998), ““Customer lifetime value: marketing models and
applications”, Journal of Interactive Marketing, Vol. 12 No. 1, pp. 17-30.
Berry, M.J.A. and Linoff, G.S. (2004), Data Mining Techniques for Marketing, Sales, and
Customer Relationship Management, 2nd ed., Wiley Computer Publishing.
Berson, A., Smith, S. and Thearling, K. (2000), Building Data Mining Applications for CRM,
McGraw-Hill, Manhatten, NY.
Bijnen, F.G.C., van Aalst, H., Baillif, P.-Y., Blonk, J.C.G., Kersten, D., Kleinherenbrink, F., Lenke,
R. and vander Stappen, M.L.M. (2002), “In-line structure measurement of food products”,
Powder Technology, Vol. 124 No. 3, pp. 188-94.
Davies, D.L. and Bouldin, D.W. (1979), “A cluster separation measure”, IEEE Transactions on
Pattern Analysis and Machine, Vol. 24 No. 2, pp. 224-7.
Dunn, J.C. (1974), “Well separated clusters and fuzzy partitions”, Cybernetics and Systems: An
International Journal, Vol. 4 No. 1, pp. 95-104.
Gök, O. (2009), “Linking account portfolio management to customer information: using customer
satisfaction metrics for portfolio analysis”, Industrial Marketing Management, Vol. 38
No. 4, pp. 433-9.
Han, J. and Kamber, M. (2006), Data Mining Concepts and Techniques, 2nd ed., Morgan
Kaufmann Publishers, San Francisco, CA.
He, Z., Xu, X., Huang, J.Z. and Deng, S. (2004), “Mining class outliers: concepts, algorithms and
applications in CRM”, Expert Systems with Applications, Vol. 27 No. 4, pp. 681-97.
Hughes, A.M. (1994), Strategic Database Marketing, Probus Publishing Company, Chicago, IL.
Johnsa, N. and Pineb, R. (2002), “Consumer behaviour in the food service industry: a review”,
International journal of Hospitality Management, Vol. 21 No. 2, pp. 119-34.
Kincaid, J.W. (2003), Customer Relationship Management: Getting It Right, Prentice Hall, Upper
Saddle River, NJ.
 Ling, R. and Yen, D.C. (2001), “Customer relationship management: an analysis framework and Fast food
implementation strategies”, Journal of Computer Information Systems, Vol. 41 No. 3,
pp. 82-97. customer
MacQueen, J.B. (1967), “Some methods for classification and analysis of multivariate behavior
observations”, Proceedings of 5-th Berkeley Symposium on Mathematical Statistics and
Probability, University of California Press, Berkeley, CA, pp. 281-297.
Ngai, E.W.T. (2005), “Customer relationship management research (1992–2002): an academic 613
literature review and classification”, Marketing Intelligence & Planning, Vol. 23,
pp. 582-605.
Parvatiyar, A. and Sheth, J.N. (2001), “Customer relationship management: emerging practice,
process, and discipline”, Journal of Economic & Social Research, Vol. 3 No. 2, pp. 1-34.
Rousseeuw, P.J. (1987), “Silhouettes: a graphical aid to the interpretation and validation of cluster
analysis”, Journal of Computational and Applied Mathematics, Vol. 20 No. 1, pp. 53-65.
Swift, R.S. (2001), Accelarating Customer Relationships: Using CRM and Relationship
Technologies, Prentice Hall, Upper Saddle River, NJ.
Teo, T.S.H., Devadoss, P. and Pan, S.L. (2006), “Towards a holistic perspective of customer
relationship management implementation: a case study of the housing and development
board, Singapore”, Decision Support Systems, Vol. 42 No. 3, pp. 1613-27.
Turban, E., Aronson, J.E., Liang, T.P. and Sharda, R. (2007), Decision Support and Business
Intelligence Systems, 8th ed., Pearson Education, Upper Saddle River, NJ.
Xu, Y., Yen, D.C., Lin, B. and Chou, D.C. (2002), “Adopting customer relationship management
technology”, Industrial Management & Data Systems, Vol. 102 No. 8, pp. 442-52.

About the authors

Niyoosha Jafari Momtaz earned a Bachelor’s degree in computer engineering from Bu-Ali Sina
University, Hamedan, Iran. She is now working on her thesis in order to earn a Master’s degree in
information technology from K.N. Toosi University of Technology, Tehran, Iran. Her research
has focused on social network analysis in marketing activities. She is also interested in the fields
of consumer behavior in electronic commerce, data mining and metaheuristic algorithms.
Somayeh Alizadeh is an Assistant Professor in the Department of Industrial Engineering at
K.N. Toosi University of Technology, Tehran, Iran. She received her BSc from Sharif University
of Technology in Tehran, and MSc and PhD from IUST University in Tehran. Her main research
interests are in information technology, data mining and social network. Somayeh Alizadeh is
the corresponding author and can be contacted at: [email protected]
Mahyar Sharif Vaghefi earned a Bachelor’s degree in computer engineering from Azad
University, Tehran, Iran. He is now working on his thesis in order to earn a Master’s degree in
information technology from K.N. Toosi University of Technology, Tehran, Iran. His research
has focused on agents’ satisfaction in electronic commerce. He is also interested in the fields of
social networks, marketing, data mining, and electronic auctions.

To purchase reprints of this article please e-mail: [email protected]

Or visit our web site for further details: www.emeraldinsight.com/reprints

View publication stats