Review on Customer Segmentation Technique on Ecommerce

Article (PDF Available)inJournal of Computational and Theoretical Nanoscience 22(10):3018-3022 · October 2016with 3,775 Reads 
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
DOI: 10.1166/asl.2016.7985
Cite this publication
Abstract
Ecommerce transactions are no longer a new thing. Many people shop with ecommerce and many companies use ecommerce to promote and to sell their products. Because of that, overloading information appears on the customers’ side. Overloading information occurs when customers get too much information about a product then feel confused. Personalization will become a solution to overloading problem. In marketing, personalization technique can be used to get potential customers in a case to boost sales. The potential customer is obtained from customer segmentation or market segmentation. This paper will review customer segmentation using data, methods and process from a customer segmentation research. The data for customer segmentation were divided into internal data and external data. Customer profile and purchase history were treated as the internal data while server log, cookies, and survey data were as the external data. These data can be processed using one of several methods: Business Rule, Magento, Customer Profiling, Quantile Membership, RFM Cell Classification Grouping, Supervised Clustering, Customer Likeness Clustering, Purchase Affinity Clustering and Unsupervised Clustering. In this paper, those methods were classified into Simple technique, RFM technique, Target technique, and Unsupervised technique and the process was generalized in determining the business objective, collecting data, data preparation, variable analysis, data processing, and performance evaluation. Customer behavior in accessing ecommerce when viewing a product on ecommerce was recorded in server log with time. Duration when seeing the product can be used as customer interest in the product so that it can be used as a variable in customer segmentation.
Figures - uploaded by Juni Nurma Sari
Author content
All content in this area was uploaded by Juni Nurma Sari
Content may be subject to copyright.
Advertisement
RESEARCH ARTICLE
Adv. Sci. Lett. 4, 400–407, 2011
1Adv. Sci. Lett. Vol. 4, No. 2, 2011 1936-6612/2011/4/400/008 doi:10.1166/asl.2011.1261
Copyright © 2011 American Scientific Publishers Advanced Science Letters
All rights reserved Vol. 4, 400–407, 2011
Printed in the United States of America
Juni Nurma Sari
1,2
, Lukito Edi Nugroho
1
,Ridi Ferdiana
1
,P. Insap Santosa
1
1
Department of Electrical Engineering and Information Technology, University of Gadjah Mada, Jogjakarta, Indonesia
2
Department of Informatics Technology, Polytechnic Caltex Riau, Pekanbaru, Indonesia
Ecommerce transactions are no longer a new thing. Many people shop with ecommerce and many companies use ecommerce
to promote and to sell their products. Because of that, overloading information appears on the customers’ side. Overloading
information occurs when customers get too much information about a product then feel confused. Personalization will
become a solution to overloading problem. In marketing, personalization technique can be used to get potential customers in
a case to boost sales. The potential customer is obtained from customer segmentation or market segmentation. This paper will
review customer segmentation using data, methods and process from a customer segmentation research.The data for
customer segmentation were divided into internal data and external data. Customer profile and purchase history were treated
as the internal data while server log, cookies, and survey data were as the external data. These data can be processed using
one of several methods: Business Rule, Magento, Customer Profiling, Quantile Membership, RFM Cell Classification
Grouping, Supervised Clustering, Customer Likeness Clustering, Purchase Affinity Clustering and Unsupervised Clustering.
In this paper, those methods were classified into Simple technique, RFM technique, Target technique, and Unsupervised
technique and the process was generalized in determining the business objective, collecting data, data preparation, variable
analysis, data processing, and performance evaluation. Customer behavior in accessing ecommerce when viewing a product
on ecommerce was recorded in server log with time. Duration when seeing the product can be used as customer interest in
the product so that it can be used as a variable in customer segmentation.
Keywords: Ecommerce, Customer Segmentation, Personalization
1. INTRODUCTION
Ecommerce development began when the internet is
growing and growing until today, especially in B2C
ecommerce (Business to Customer). When shopping use
ecommerce, a user finds it easy and faster. The ease of
using ecommerce encourages customers to buy using
ecommerce. With these conditions the problem that
comes up is the overloading information because of many
products offered by ecommerce
1
. Overloaded information
can be overcome by an implementation of personalization
in ecommerce services such as providing product
recommendation, links recommendation, ads or text and
graphics that correspond to the users’ characteristics and
needs
2
. In addition to solving the problem of overloaded
information, personalized services in ecommerce can
maintain customer loyalty of existing customer
3
, getting
new customers by providing service to customers in
accordance with their needs and characteristics. It will
generate more profits for the company. Before the
personalization is implemented, customer segmentation
*
Email Address: juni.s3te14@mail.ugm.ac.id
should be conducted because the result from customer
segmentation process will be used as inputs to personalize
ecommerce services, resulting in dynamic personalization
ecommerce services based on current customer conditions.
Customer segmentation is currently performed by
processing customer database, i.e. demographic data or
purchase history. Several researchers discuss the customer
segmentation method on their papers, such as Magento
4
,
who used several variables to perform customer
segmentation, namely transaction variable, product
variable, geographic variable, hobbies variable and page
viewed variable; Baer
5
and Colica
6
discuss customer
segmentation methods of Business Rule, Quantile
membership, Supervised Clustering, Unsupervised
Clustering, Customer Profiling, RFM Cell Classification
Grouping, Customer Likeness Clustering and Purchase
Affinity Clustering. Some of these methods have
similarity. Other researchers discuss the implementation
of customer segmentation. This paper will classify
customer segmentation methods based on data processing.
Review on Customer Segmentation
Technique on Ecommerce
Adv. Sci. Lett. 4, 400–407, 2011
RESEARCH ARTICLE
2
2.CUSTOMER SEGMENTATION
In marketing, one way to increase profits is to
communicate with customers to determine customer
wishes
5
. Communication is built according to the
characteristics of the customer. Communication is very
difficult to create using personal approaches. So it is
necessary to divide customers into groups that have the
same characteristics, and this is called customer
segmentation. Schneider
7
also called market segmentation
that divides potential customers into a group. Magento
4
,
an ecommerce platform, in its ebook mentions that
customer segmentation is an activity to divide customers
into groups that have the same characteristics. Customer
segmentation has several benefits: it enables us to match
between the customer and an offer of similar products; it
changes the way we communicate with the customer
based on customer data; it identifies the most profitable
customers; and it enables us to update the products and
services to meet customer needs. Baer
5
states that
customer segmentation is the activity to categorize or to
classify an item or subject to a group that has been
identified to have in common. In his research, Baer
discusses Customer Segmentation Intelligence to improve
marketing in offering products or services that meet the
needs of each customer group. Segmentation according to
Collica
6
is the process to categorize or classify an item
into a group that has a similarity in characteristic and in
CRM (Customer Relationship Management)
segmentation is used to classify customer based on some
similarities by segmenting the records of customer
database. This chapter will discuss the customer data for
customer segmentation, customer segmentation methods
and customer segmentation process and then the methods
will be classified based on data processing.
A. Data for Customer Segmentation
Customer segmentation requires customer data from
various sources. Magento
4
categorizes the data into
internal data and external Data. Customer registration,
customer profile, and purchase history are the internal
data obtained from the database of an ecommerce. While
external data are census data, media browsing, surveys
and market search, cookies, web and social media
analysis. Information about customer lifestyle, attitude,
activity and shopping preferences are obtainable through
surveys and market search and social media. Browsing
history can be seen from server log or cookies. Baer
5
in
his research, Customer Segmentation Intelligence, uses
internal data by looking the demographic data from
customer profile and purchase history. Likewise, Colica
6
,
uses the customer database and purchase history on
customer segmentation methods.
B. Methods of Customer Segmentation
Customer segmentation can be performed using
various approaches. Theoretically, Schneider
7
divides
customer segmentation methods into geographic,
demographic, psychographic, behavioral/occasion, usage-
based market segmentation. Geographic segmentation is
based on location. Demographic segmentation is based on
age, gender, family size, income, education, religion or
ethnic. Psychographic segmentation is based on social
class, personality or their approach to living. Behavioral
segmentation is based on customer behavior but when
customer behavior occurs in specific time or occasion,
Schneider called it Occasion segmentation. Usage-based
Market segmentation is based on behavior pattern of each
visitor, which includes a set of categories of customer
namely browser, buyer and shopper. Browsers are visitors
that just browse a site; buyers are visitors that make a
purchase; and shoppers are customers that want to buy,
but want to read product reviews and the list of features
before buying.
Almost the same with Schneider, Magento divides
customer segmentation methods into Profit Potential, Past
Purchase, Demographic, Psychographic and Behavior. In
Magento
4
there are several variables used:
1)
Profit Potential: using variable transaction frequency,
date of last purchase, average order value, customer
lifetime value.
2)
Past Purchases: using the variable of product
type/attribute, product price, payment/shipping method
used, product benefit sought (price, quality, prestige),
product satisfaction.
3)
Demographic: using the variable of geographic
location (city state, country, region), age, gender,
household size, income, occupation, education,
ethnicity, browsing device (laptop, PC, tablet,
smartphone) and type (vendor and model), traffic
source (organic search, banner link, referral site).
4)
Psychographic: using the variable of hobbies and
interest, leisure and recreational activity, affiliations
(religious, professional, cultural, political,
institutional), personal traits (social vs. private;
modern vs. traditional; spontaneous vs. cautious).
5)
Behavior: using the variable of pages viewed,
responses to offers and promotions, participation in
reward programs, channel management.
Magento also performed an analysis of purchase
history to get the best customer, unprofitable customers,
potential customer profit. Best customer is when the
customer is a frequent shopper and a repeated customer,
with high average order value, low return, providing
review and response customer. Unprofitable customer
when the customer has high rate product return, low
average order value, high rate customer service calls,
wants the lowest price. Potential customer profit is
determined by counting customer lifetime values.
Baer
5
segments customer using business rules
method, quantile membership method, supervised
clustering with decision tree method and unsupervised
clustering method using k-means algorithm. Demography
RESEARCH ARTICLE
Adv. Sci. Lett. 4, 400–407, 2011
3
data and purchase pattern are used to segment costumers.
Here are Baer customer segmentation methods:
1) Bussiness Rule: in this method, customers are grouped
into specific groups based on a predetermined class,
such as:
a) Grouping based on demographic data, such as age,
gender, income and education, etc. This method
has similarity with Magento and Schneider.
b) Grouping based on customer interaction with the
company based on data purchase pattern such as
the type of product or service provided or RFM
data, where R is Recency (when customer last
shopped), F is Frequency (how often the customer
shops) and M is Monetary (how much the customer
spends)
According to Baer, the lack of business rule does not
reflect the actual customer behavior and a segment
similar to another segment.
2) Quantile Membership, this method uses data Recency,
Frequency, and Monetary. Here is the quantile
membership methods:
a) Recency divided into five groups of intervals, for
example, starting from 0 days up to 730 days then
classify it with label A until E, where A is very
valuable customer and E is low-value customer.
Also with Frequency and Monetary. When 3 RFM
is combined, there is label AAA until EEE.
b) Map two components of RFM to a table.
c) Divided into two groups A, B with the
classification most valuable customer and two
groups D, E to the classification of least valuable
customer. C is average value customer.
d) The result can be inferred for example good
frequency (A or B), good monetary (A or B) but
poor recency (D or E), and then the advice that
given is upgrade the promotion strategy to make
the old customer come back
3) Supervised Clustering with decision tree: this method
uses a specific target, or dependent variable and target
would predict differences in independent variables
(input). Data utilized in this method is previous
purchase pattern and customer demographic.The
algorithm that used is decision tree with the target on
their nodes. According to Baer, although this method
connects the target with the other customer attributes,
it shows only one aspect of customer behavior.
4) Unsupervised Clustering: this method uses any
number of customer attributes then measure the
similarity among customer, each customer attribute
use Euclidean distance
8
(1) then cluster the customer
use k-means clustering
9
(2). If the distance is the
shortest distance between customer data and cluster,
then customer is included in that cluster.
Euclidean distance=(
− 
)
+ . .. +(
− 
)
(1)
C(i)= arg min ∑ ∑ || − 
||
∈

(2)
Colica has several methods are almost the same.
Colica has segmentation methods as follows: Customer
Profiling, Customer Likeness Clustering, RFM Cell
Classification grouping and Purchase Affinity Clustering.
In Customer Profiling method, the required information
about customer is the fourW's (who, what, where and
when) from customer database.It can be done by using a
query on the customer database or using the clustering
algorithm when the data is huge. Customer likeness
clustering method is used in franchise stores to know
whether the profits and turnover of each product in each
store are similar, then to review other variables such as
demographics. Colica also uses a decision tree for simple
clustering the same with Baer. Method RFM (Recency
Frequency Monetary) Cell Classification Grouping uses
three dimensions to classify each customer in one cell
after labeling each level of RFM. Colica names it the
Segmentation Using Cell-Based Approach. This method
is similar to Baer’s quantile membership. Another method
used by Colica is Purchase Affinity Clustering. This
method uses scoring on interesting in certain products
then clusters customer database based on that score to get
a similar group.
Table 1. Methods of Customer Segmentation
Paper
Method
Data
Advantage
Disadvantage
Magento
(2014)
Magento Demographic,
Purchase
History, Data
Product, Data
Media, Data
Marketing,
Server Log
Have clear
variable
customer
segmentation
There is no
data processing
for each
variable
Baer
( 2012)
Bussiness
Rule
Demographic,
Purchase
history
Easy to apply,
Use database
query
Not focus on
customer
behavior
Quantile
membership
Purchase
history
Can process
small data,
can be used
with other
data
Good result
obtained when
determining a
good
classification
Supervised
Clustering
with decision
tree
Demographic,
Purchase
history
Classify
customers
according to
target
Use one
variable to
cluster
Unsupervised
Clustering
Purchase
history
Use any
number of
customer
attributes
Speed of
computation
depends on k
values
Colica
(2011)
Customer
Profiling
Demographic,
Purchase
history
use database
query if data
is small
Not focus on
behavior
Customer
Likeness
Clustering
Demographic,
Purchase
History, Data
product
classify
customers
according to
the target
Problem arises
when there
are different
unit in record
RFM Cell
Classification
Grouping
Purchase
history
Efficient three
-dimensional
mapping
Good result
obtained when
determining a
good
classification
Purchase
Affinity
Clustering
Purchase
history, Data
product
know the
products most
in demand
Spesific to
product
segmentation
Adv. Sci. Lett. 4, 400–407, 2011
RESEARCH ARTICLE
4
There are some researches that implement customer
segmentation methods according to the table above such
as Lieberman
10
, who uses combination Business Rule,
Customer Profiling, Magento to find how much customer
spend money monthly on clothing and how many
customers visit monthly; Dodwell
11
, who uses RFM
Analysis to segment email marketing for potential
customer; Birant
12
, who uses combination RFM Analysis
and Data Mining (Classification Rules and Association
Rules) to provide better product recommendation; Han
13
,
who uses Decision tree model to identify high-value
customer; Ma
14
, who uses Association Rules and Decision
Tree to improve customer loyalty, attract new customer
and expand the market effectively; Baer
15
, who uses
Market Base Analysis, K-means Clustering, and
Doughnut Clustering to segments customer based on
product, and Ezenkwu
9
and Venkatesan
8
, who use K-
means Clustering to segment customer.
Based on table and researches above, customer
segmentation methods can be classified into: Simple
technique, because this method uses database query and
statistical data; RFM technique, because this method uses
RFM analysis; Target technique, because this method
must have target to segments customer, for instance,
customer segmentation focus on product, focus on
purchase; and Unsupervised technique, because this
method uses dynamic data. Figure 1 describes Customer
Segmentation Classification.
Figure 1. Customer Segmentation Classification
C. Process of Customer Segmentation
Customer Segmentation is associated with the
business objective. The first step of segmentation is
deciding business objective. Chen
16
discusses customer
segmentation process begins by determining the business
objective such as the identification of high profitable
customer groups, improve product for that customer. The
next step is collecting the necessary data such as
demographic data, transaction data, and promotional data,
then determining the method of customer segmentation
and standardization measurement. After that, the next step
is exploration data by analyzing the statistics and look for
relationships between variables. Results of analysis can
be used to measure the similarity among the customers
using Euclidean distance to measure two points in a
multidimensional space where the point is customer data.
The cluster is validated by calculating the ratio of the
between-cluster variantto within cluster variants(RSQ/1-
RSQ).
Process Customer Segmentation on Lieberman
10
research begins with determining the business rule,
collecting data spread the questioner, then data processing
with logistic regression and waterfall and analyze statistic
data. Birant
12
has a more complex process than
Lieberman because he combines RFM Analysis and Data
Mining to find product recommendation. Birant starts the
process of defining the business objective, collecting data,
and then data processing with the first method of RFM
analysis that uses quantile membership to find customer
level of Recency, Frequency, and Monetary. The second
method is Clustering with RFM Cell Classification
Grouping to find customer segmentation. After
segmentation, there is prediction of customer behavior, it
uses Association Rule method. Finally, the product
recommendation uses Classification method. Process
Customer Segmentation on Ma
17
research starts with
defining the business objective, choosing variables that
relate to purchasing then form data set, finding frequent
item set use generalized association rule, cleaning non-
interest rule, building tree process, prunning decision tree,
extract rules from pruned decision tree in if-then format.
Ezenkwu
9
process also starts with determining the
business rule, choosing data variable, namely the amount
of goods purchased by customer monthly and the average
number of customers visiting monthly; the data
processing with k-mean clustering which is normalization
alongside centroids, initialization step, assignment step
and updating step after that performance evaluation.
Process of customer segmentation can be simplified into
defining business objective, collecting data, data
preparation, analyzing variable, data processing, and
performance evaluation as describe in figure 2.
3. FUTURE WORK
One of the data used for customer segmentation is
customer behavior in accessing ecommerce. Customer
behavior data are obtainable from server log. Variables
contained in server log are IP address of customer, date,
time, HTTP request. Here an example of server log data:
05:09:49 GET /detail-item.php?item=ilford-delta-100 HTTP/1.0
05:09:53 GET /detail-item.php?item=ilford-pan-f-50 HTTP/1.0
Customer
Segmentation
Technique
Simple Technique
Business Rule
Customer
Profiling
Magento
RFM Technique
Quantile
Membership
RFM Cell
Classification
Grouping
Target Technique
Supervised
Clustering
Customer
Likeness
Clustering
Purchase Affinity
Clustering
Unsupervised
Technique
RESEARCH ARTICLE
Adv. Sci. Lett. 4, 400–407, 2011
5
Time shows when a customer accesses page, the
difference of time between the customer’s visit to the first
page and the second page is the duration of customer’s
visit to the first page. The first data is page detail-
item.php with the first product of ilford-delta-100 and the
second data is page detail-item.php with the product of
ilford-pan-f-50. Knowing data duration, we can determine
the user's attention to the product. If the user's attention to
the product is in long duration, then the customer has an
affinity for product. It can also be used for customer
segmentation based on the interest in the product. Such
information can be utilized for the promotion of a product.
The disadvantage of this method is when customer
position isn't in front of computer but server still record
the activity, so the solution is using an eye-tracker to
record the customer’s attention.
Figure 2. Process of Customer Segmentation
4. CONCLUSION
Customer segmentation is a way to improve
communication with the customer, to know the wishes of
the customer, customer activity so that appropriate
communication can be built. Customer Segmentation
needed to get potential customers used to increase profits.
Potential customer data can be used to provide service the
characteristics of customer including ecommerce services
as a media buying and selling online.
This paper discusses several components to do
customer segmentation, which is:
Customer segmentation is an activity to divide
customers or item into groups that have the same
characteristics.
Data that needed for customer segmentation are
internal data and external data. The internal data
include demographic data and data purchase history,
while the external data include cookies and server
logs. Internal data can be obtained from a database
when customer do registration or transactions and
external data can be obtained from web server or
other source.
Methods of Customer Segmentation can be classified
into Simple technique, RFM technique, Target
technique, and Unsupervised technique. On Target
technique, researcher focus on one variable, it can be
product or purchase. Unsupervised technique was
used when clustering process reseacher have many
variable
Process of Customer Segmentation can be
simplified into defining business objective,
collecting data, data preparation, analyzing variable,
data processing, and performance evaluation.
REFERENCES
[1] Al-Qaed F, Sutcliffe A. Adaptive Decision Support System
(ADSS) for B2C E-Commerce. 2006 ICEC Eighth Int Conf
Electron Commer Proc NEW E-COMMERCE Innov Conqu
Curr BARRIERS, Obs LIMITATIONS TO Conduct Success
Bus INTERNET. 2006:492-503.
[2] Mobasher B, Cooley R, Srivastava J. Automatic
Personalization Based on Web Usage Mining. Commun
ACM. 2000;43(8).
[3] Cherna Y, Tzenga G. Measuring Consumer Loyalty of B2C
e-Retailing Service by Fuzzy Integral: a FANP-Based
Synthetic Model. In: International Conference on Fuzzy
Theory and Its Applications iFUZZY.; 2012:48-56.
[4] Magento. An Introduction to Customer Segmentation. 2014.
info2.magento.com/.../
An_Introduction_to_Customer_Segmentation...
[5] Baer D. CSI : Customer Segmentation Intelligence for
Increasing Profits. SAS Glob Forum. 2012:1-13.
http://support.sas.com/resources/papers/proceedings12/103-
2012.pdf.
[6] Colica R. Customer Segmentation And Clustering Using
SAS Enterprise Minner Part I The Basics. 2011:1-14.
[7] Schneider G. Electronic Commerce, 9th Edition.; 2013:643.
doi:10.1002/1521-3773(20010316)40:6<9823::AID-
ANIE9823>3.3.CO;2-C.
[8] Venkatesan R. Cluster Analysis For Segmentation. 2007.
[9] Ezenkwu CP, Ozuomba S. Application of K-Means
Algorithm for Efficient Customer Segmentation : A Strategy
for Targeted Customer Services. 2015;4(10):40-44.
[10] Lieberman M. Target “ golden egg ” consumer to achieve
maximum ROI. 2009;(May):50-51.
[11] Dodwell A. Effective Marketing email startegy _
Segmentation RFM. 2015.
http://www.sailthru.com/marketing-blog/written-effective-
email-marketing-strategies-segmentation-rfm/.
[12] Birant D. Data Mining Using RFM Analysis. Knowledge-
Oriented Appl Data Min. 2011;(iii):91-108.
doi:10.5772/13683.
[13] Hua S, Xiu S, Leung SCH. Expert Systems with
Applications Segmentation of telecom customers based on
customer value by decision tree model. Expert Syst Appl.
2012;39(4):3964-3973. doi:10.1016/j.eswa.2011.09.034.
[14] Ma H. A Study on Customer Segmentation for E-Commerce
Using the Generalized Association Rules and Decision Tree.
2015;(December):813-818.
[15] Baer D, Ph D. Product Affinity Segmentation Using The
Doughnut Clustering Approach. Cust Intell SAS Glob Forum
2013. 2013.
[16] Chen J. Retail Customer Segmentation. 2014;(April).
[17] Chan C, Swatman PMC. Management and business issues
for B2B ecommerce implementation. Proc 35th Annu
Hawaii Int Conf Syst Sci. 2002;00(c):1-11.
doi:10.1109/HICSS.2002.994303.
Defining Business
Objective Collecting Data
Data Preparation
Analyzing variable and
looking for the
relationship amongst
the variables
Data Processing with
Selected Method
Performance
evaluation
  • ... However, such an arbitrary choice of features is limited and does not represent the relations among distinct features. The unsupervised clustering technique has been identified in [21] as the most flexible method for automatic detection of underlying behavioral patterns. In [23] the authors scale up the user neighborhood formation process through the use of bisecting k-means clustering for an e-commerce application. ...
    Preprint
    Full-text available
    The traditional offline approaches are no longer sufficient for building modern recommender systems in domains such as online news services, mainly due to the high dynamics of environment changes and necessity to operate on a large scale with high data sparsity. The ability to balance exploration with exploitation makes the multi-armed bandits an efficient alternative to the conventional methods, and a robust user segmentation plays a crucial role in providing the context for such online recommendation algorithms. In this work, we present an unsupervised and trend-responsive method for segmenting users according to their semantic interests, which has been integrated with a real-world system for large-scale news recommendations. The results of an online A/B test show significant improvements compared to a global-optimization algorithm on several services with different characteristics. Based on the experimental results as well as the exploration of segments descriptions and trend dynamics, we propose extensions to this approach that address particular real-world challenges for different use-cases. Moreover, we describe a method of generating traceable publishing insights facilitating the creation of content that serves the diversity of all users needs.
  • ... However, such an arbitrary choice of features is limited and does not represent the relations among distinct features. The unsupervised clustering technique has been identified in [21] as the most flexible method for automatic detection of underlying behavioral patterns. In [23] the authors scale up the user neighborhood formation process through the use of bisecting k-means clustering for an e-commerce application. ...
    Conference Paper
    Full-text available
    The traditional offline approaches are no longer sufficient for building modern recommender systems in domains such as online news services, mainly due to the high dynamics of environment changes and necessity to operate on a large scale with high data sparsity. The ability to balance exploration with exploitation makes the multi-armed bandits an efficient alternative to the conventional methods, and a robust user segmentation plays a crucial role in providing the context for such online recommendation algorithms. In this work, we present an unsupervised and trend-responsive method for segmenting users according to their semantic interests, which has been integrated with a real-world system for large-scale news recommendations. The results of an online A/B test show significant improvements compared to a global-optimization algorithm on several services with different characteristics. Based on the experimental results as well as the exploration of segments descriptions and trend dynamics, we propose extensions to this approach that address particular real-world challenges for different use-cases. Moreover, we describe a method of generating traceable publishing insights facilitating the creation of content that serves the diversity of all users needs.
  • Chapter
    A social network refers to a network of social interactions using various social media sites such as Facebook, WhatsApp, Twitter, Snapchat, Instagram, Pinterest and LinkedIn. This chapter will first explain various concepts of social commerce involving the use of social networks and then elaborate on the key characteristics of social commerce and will highlight how it is different from e-commerce. A discussion is included on the concept of cross-border social commerce and its specific characteristics. Further, this chapter examines the trends in social commerce adoption by companies to promote international sales and the related benefits and challenges. It also presents a discussion on the use of social networks to understand and influence cross-cultural consumer behaviour in social commerce. Additionally, this chapter explores how social commerce can specifically benefit SMEs and born global firms to reach a larger audience in foreign countries and includes a discussion on cross-border social commerce strategy for business. Keywords: Social network, Social media, Social support, E-commerce, Social commerce, Born global.
  • Chapter
    Customer satisfaction represents a crucial goal for every seller. In e-commerce, it is possible to increase this factor by a better understanding of customers purchasing behavior based on collected historical data. In a period of a continually growing amount of data, it is not an easy task to effectively pre-process and analyses. Our motivation was to understand the buying behavior of the on-line e-shop customer through appropriate analytical methods. The result is a knowledge set that retailers could use to deliver products to specific customers, to meet their expectations, and to increase his revenues and reputation. For recommendations generation, we used a collaborative filtering method and matrix factorization associated with Singular Value Decomposition (SVD) algorithm. For segmentation, we selected the K-Means algorithm and the RFM method. All methods produced interesting and potentially useful results that will be evaluated and deployed into practice.
  • An Introduction to Customer Segmentation
    • Magento
    Magento. An Introduction to Customer Segmentation. 2014. info2.magento.com/.../ An_Introduction_to_Customer_Segmentation...
  • Customer Segmentation Intelligence for Increasing Profits
    • D Baer
    • Csi
    Baer D. CSI : Customer Segmentation Intelligence for Increasing Profits. SAS Glob Forum. 2012:1-13.
  • Article
    Full-text available
    With the rapid development of e-commerce, e-commerce is becoming more and more competitive. How to improve customer loyalty, attract more new customers, and expand the market effectively, it is very important for the e-commerce enterprise. In this paper, a comprehensive model is proposed, which is based on generalized association rules and decision tree technology. The model is used for customer segmentation of e-commerce website. It can help e-commerce companies understand customers, support decision-making, so as to provide customers with more targeted services.
  • Article
    Full-text available
    The emergence of many business competitors has engendered severe rivalries among competing businesses in gaining new customers and retaining old ones. Due to the preceding, the need for exceptional customer services becomes pertinent, notwithstanding the size of the business. Furthermore, the ability of any business to understand each of its customers' needs will earn it greater leverage in providing targeted customer services and developing customised marketing programs for the customers. This understanding can be possible through systematic customer segmentation. Each segment comprises customers who share similar market characteristics. The ideas of Big data and machine learning have fuelled a terrific adoption of an automated approach to customer segmentation in preference to traditional market analyses that are often inefficient especially when the number of customers is too large. In this paper, the k-Means clustering algorithm is applied for this purpose. A MATLAB program of the k-Means algorithm was developed (available in the appendix) and the program is trained using a z-score normalised two-feature dataset of 100 training patterns acquired from a retail business. The features are the average amount of goods purchased by customer per month and the average number of customer visits per month. From the dataset, four customer clusters or segments were identified with 95% accuracy, and they were labeled: High
  • Conference Paper
    Customer store loyalty is one of the most important motivations that facilitate the intentions of re-patronizing a retailer's website and create marketing values for firms. Research on survey-based for measuring consumers' loyalty to persist in online shopping environment has been developed. But while the measured values of human perceptions towards various attributes are often uncertain data, most conventional measurement methods could not precisely clarify the incomplete information by numerical format. Thus, this paper focused on employing the proposed fuzzy multiple attribute decision making (FMADM) model to cope with this issues within context of B2C retailer loyalty construct. First, the analytic network process (ANP) will be used for determining the relative importance weights for each attributes and identifying key factors that promote B2C loyalty. Second, the logic of triangular fuzzy numbers will be leveraged to transform the satisfaction level of consumers into explicit numbers. Further, considering the complex characteristic between attributes that interdependently interact with one another, we applied the addictive simple weighted aggregation and the non-addictive type fuzzy integral to obtain the fuzzy synthetic utility measures representing for loyalty level and assess the alternatives. In this paper, an empirical case assessed the consumer loyalty towards three Taiwanese B2C ecommerce websites will demonstrate the feasibility of this approach. In addition, we validated the satisfactory degree of a new elements of Personalized Product Recommendations (PPRs) attribute in B2C e-retailing in subject of online marketing by FMADM. In the evaluation process, five main antecedents of B2C retailing services loyalty are obtained and the fuzzy synthetic utility scores are measured for loyalty ranking of the case companies. The results also revealed the different ranking result derived from simple aggregation method and from non-addictive perspective of multi- le attributes utility. The multi-attributes consumers' loyalty key measures derived from this study could further support decision making in B2C e-retailing and provide strategic directions to service innovation in markets.
  • Article
    Full-text available
    Until recently little research has been undertaken into the process of eCommerce implementation, especially in relation to the implementation of business-to-business (B2B) relationships. Given the complexity of this process we have, in this paper, endeavoured to contribute to what we perceive as a gap in the body of theory surrounding the implementation process in the extant business-to-business relationship literature. We describe the findings of a series of multiple case studies comprising ten major Australian eCommerce initiators. In addition to confirming our earlier finding of the importance of non-technical factors for the success of the implementation process we also present, through our case studies, the various management and business issues associated with the success or otherwise of B2B eCommerce implementation.
  • Conference Paper
    Full-text available
    This paper covers the research problem of supporting users' decision making in E-Commerce systems with complex choices, and design of an Adaptive Decision Support System (ADSS) which matches the appropriate tool support and decision strategy advice to the user's preferences and motivations. A preliminary requirements investigation will be described which used an adapted Wizard of Oz approach to test users' reaction to mock-ups of the ADSS with and without system advice. The requirements study tested users' reaction to the proposed tools and whether they were influenced by the system advice. This discussion will be followed by a description of the design and development of the ADSS and its architecture. The paper concludes with plans for future research.
  • Article
    Full-text available
    this paper we describe an approach to usage-based Web personalization taking into account the full spectrum of Web mining techniques and activities. Our approach is described by the architecture shown in Figure 1, which heavily uses data mining techniques, thus making the personalization process both automatic and dynamic, and hence up-to-date. Specifically, we have developed techniques for preprocessing of Web usage logs and grouping URL references into sets called user transactions [CMS99]. A user transaction is a unit of semantic activity, and performing data mining on them is more meaningful. We describe and compare three different Web usage mining techniques, based on transaction clustering, usage clustering, and association rule discovery, to extract usage knowledge for the purpose of Web personalization. We also propose techniques for combining this knowledge with the current status of an ongoing Web activity to perform real-time personalization. Finally, we provide an experimental evaluation of the proposed techniques using real Web usage data.