Advances in Systems Science and Applications (2013) Vol.13 No.2 167-181

E-Market Segmentation for Internet-Mediated
Fashion Brands: A Conceptual Framework with

Mean-Variance Consideration
Tsan-Ming Choi, Pui-Sze Chow and Jin-Hui Zheng

Institute of Textiles and Clothing, The Hong Kong Polytechnic University,
Hung Hom, Kowloon, Hong Kong.

Abstract

E-market segmentation takes a crucial role in marketing for fashion brands. Its
applications for collaborative functions such as estimating the advertisement bud-
get, retaining customers, carrying out direct tailored marketing, and implement-
ing dynamic pricing are essentially important and critical for their e-business.
Unlike the bricks-and-mortar traditional store retailers, fashion brands which
operate online can keep track of the details of the online customers easily and
precisely. How to make use of these customers details in segmenting the e-market
for each particular function becomes an important issue. As a result, we propose
and discuss in this paper a conceptual model for carrying out e-market segmen-
tation and we focus on the areas of dynamic pricing and advertisement budget
estimation. Through extensive discussions with mean-variance consideration, we
believe that the model can be incorporated into other existing market segmenta-
tion analyses. Managerial implications are discussed.
Keywords Fashion branding, e-market segmentation, internet marketing, mean-
variance

1 Introduction

For fashion brands which operate solely in the traditional bricks-and-mortar re-
tailing mode, keeping track of the buying behaviours and preferences of each
specific customer is a difficult, if not impossible, task. Even if the fashion brands
can collect some customers’ data by observation, the results are relatively sketch
[1]. Unlike the bricks-and-mortar stores, fashion brands which have its online
channel (we call them EFBs (e-fashion-brands)) can precisely keep track of the
online customers’ buying behaviours easily. Details such as the surfing habit,
buying preference, price sensitivity, loyalty, etc can all be recorded and estimat-
ed. Obviously, information of these details can help the EFBs in making good
business decisions. One of the most intuitive uses of these observed customers
buying data is for marketing purpose (see [2] for the discussion of the issue of
marketing on the Internet).

Market segmentation∗ , as defined in Chaffey [4], is the “identification of dif-
ferent groups within a target market in order to develop different offerings for

∗See [3] for a recent review and discussion on a framework for market segmentation and its
applications.



Tsan Ming Choi : E-Market Segmentation for Internet-Mediated Fashion Brands... 168

the groups”. It involves identifying segments in the market, grouping customers
into segments, and targeting, positioning and developing a differential advantage
over competitors [5]. It also represents a rational and precise adjustment of the
products and services for customers [6]. A good market segmentation scheme
implies a good analysis of the target market and an identification of the true
requirements of each specific group of customers [4, 7-9]. In modern business
world, market segmentation has been realized as an essential part in nearly all
marketing projects [4,10-11].

According to Dibb and Stern [8], the literature of market segmentation mainly
goes to two distinct streams. The first stream treats market segmentation as a
technique in determining and studying different segments in the market. The sec-
ond stream views market segmentation as an approach for an efficient allocation
of resource to different segments in the market. This paper basically follows the
viewpoint of the second stream. In both the academic literature [10-12] and the
industrial practice [13], in general, we need two types of information for market
segmentation. The first type is called the “classification variables” which include
four types of variables: Demographic variables (e.g. age, gender, etc), geographic
variables (e.g. city, country, etc), psychographic variables (e.g. risk attitude,
lifestyle, etc) and behavioral variables (e.g. brand loyalty, usage level, etc). The
second type is the descriptor variables which describe each particular segment
and are used to distinguish one segment from another. Many classification vari-
ables would function as the descriptor variables, too.

Market segmentation can be a complicated process in business. For example,
suppose that after adopting the conventional segmentation steps with the classifi-
cation according to the geographical and demographic variables, we have obtained
a number of different segments. An important question to ask is: “Should we
continue to break down the segments into smaller segments?” No matter the
answer is “yes” or “no”, we need to have a good reason for it. In other words, a
good market segmentation scheme requires a good stopping condition. It is true
that there are a few rules of thumbs for decision makers to decide the number
of segments and when to stop the segmentation process. Some factors [13] that
decision makers will bear in mind include:

(1). The size of the segments must be large enough.
(2). The segments must be reachable by the company’s marketing strategies

(e.g. promotion, pricing, etc).
(3). The segments must be relevant to the company’s products and different

segments should be clearly and sufficiently different.
These guidelines are important and practical but they do not provide a precise
decision model for decision makers for making a wise and optimal decision for a
specific collaborative function with market segmentation.



169 Advances in Systems Science and Applications (2013) Vol.13 No.2

With the advance of science and technology, market segmentation can now be
done by a computerized decision support system. Methods which rely on data
mining and artificial intelligence [14-15], advanced database system [16], opera-
tions research and management science optimization techniques [6,10,17-18] , and
others [11-12,14,19-20] have been widely tested and applied. However, none of
these methods dominate the literature and each of them has its pros and cons.

As many recent research projects revealed and discussed, market segmentation
is very important for Internet-enabled e-commerce [4,16,21-22]. This also gives
rise to the term “e-market segmentation” which represents the segmentation of
the electronic market. In fact, good e-market segmentation is generally believed
to be not only important for maintaining the company’s competitive edge, but
it is also essential and crucial, for EFBs in the knowledge-based economy. When
we look deeper into the potential collaborative applications of the e-market seg-
mentation scheme for EFBs, we can identify several major functionality areas for
it. Some of them include the following:

1. Estimating the advertisement budget for acquiring different groups of cus-
tomers: Different customers have different values for the EFBs. Some of them
only surfed around and did not buy a single item at all; some made one purchase
and then did not come back; some made repeated purchases and showed a strong
sense of trust on the EFB (see Moe and Fader [23] and Betts [24] for the identifi-
cations of the four types of online shopping visits). Since it can be very expensive
to acquire new customers on the Internet [25], the advertising strategies should
be more focused. The expenses spent on advertisement for different groups (i.e.
segments) of customers should be made different, too. This can be achieved by
a good e-market segmentation scheme. On the other hand, the information from
the estimation of advertisement budget can be useful for e-market segmentation.

2. Direct marketing by providing tailor-fit services and products to the interest-
ed customers: Different customers have different preferences and requirements.
If an EFB can provide the right product to the right customer at the right price
at the right time in the right place, then a transaction results. Even though it
is impossible for the EFB to exactly predict all of these aspects, good market
segmentation can help to narrow down the error in terms of the direct marketing
offers of product type, selling price, etc. It can also help in turning browsers into
buyers [23-25] and improve customer relationship management [26].

3. Dynamic pricing: Different consumers have different sensitivities towards
prices and other attributes for the products [1,27]. Some consumers are highly
price-sensitive while some care more about service and reliability. Surveys also
show that many customers purchasing online do not shop around and they just
buy from the EFB which they first visit. According to a survey as reported in
Baker et al. [27], the percentage of consumers who buy online from the first



Tsan Ming Choi : E-Market Segmentation for Internet-Mediated Fashion Brands... 170

website they visit ranges from 76% to 89% for various products including CDs,
books and electronics. In light of this, EFBs should try to attract the customers
at the very beginning by bringing the products in front of the consumers when
they are in need (it is mentioned in the last paragraph). At the same time, how
to set the selling prices dynamically for different groups of customers so that the
profit for the EFB is maximized becomes vitally important. Failing to do so can
lead to the collapse of the business (see the example of Sun Country in Dutta et
al.[28]). In order to exercise a profit-making dynamic pricing scheme for different
groups of customers, we need to segment the market properly.

4. Retaining customers: For all kinds of retail businesses, there are always
some loyal customers. For every loyal customer, there is a sense of trust between
the retailer and herself. However, the importance of customer’s loyalty varies
among different retailers. For instance, a retail store whose customers are mainly
the tourists from overseas countries care relatively less about the loyalty of the
customers since it is unlikely for the overseas tourists to come back again in the
near future while a cosmetics retailer always wants to maintain a loyal customer
base (since the loyal customers are more likely to try some other products and
have repeated purchase’). Since most of the customers buying online have to
contribute some private data to the EFB (e.g. the credit card number) but they
cannot touch the product and cannot visit the store to talk with the sales staff
face-to-face, they simply won’t make a purchase without trusting the EFB. As
a consequence, compared with the bricks-and-mortar stores, the trust and loy-
alty of customers is especially important for EFBs. Moreover, having the loyal
customers can help the EFBs in at least two ways. First, a loyal customer is
more likely to repeat her purchase. Second, a loyal customer tends to refer new
customers to the EFB she is loyal to. It is why Reichheld and Schefter [1] have
proposed that “price does not rule the web; trust does”. In order to gain the trust
and keep the loyalty of the customers, we have to focus on their needs. Without
proper market segmentation, building and keeping the loyalty of the customers
becomes more difficult.

From the above description, we can see that a good e-market segmentation
scheme is undoubtedly crucial for the success of EFBs and it can be applied to
different collaborative functionality areas. However, what is a good e-market
segmentation scheme for each particular collaborative function? Obviously, a
good e-market segmentation scheme depends on its specific targeted functionali-
ty area. For example, a good e-market segmentation scheme for dynamic pricing
may not be good for the estimation of advertisement budget. As a result, we
propose in this paper a decision model for carrying out e-market segmentation
for different specific functionality areas. We focus our attention on two impor-
tant functionality areas: Advertisement budget estimation, and dynamic pricing.



171 Advances in Systems Science and Applications (2013) Vol.13 No.2

These are important functions for EFBs and they can be collaborated with the
conventional e-market segmentation scheme. Through the incorporation of the
performance measure (for the collaborative function) into the e-market segmen-
tation decision model, we can provide more precise e-market segmentation results
for the marketing managers to make an optimal decision (with each particular
function). The idea behind this proposed decision framework is inspired by the
classical Markowitz’ mean-variance theory in financial portfolio management [29]
with which we quantify and control the uncertainty associated with a decision by
the variance of that measure (this point will be discussed in the next section).
With the illustrative examples, we demonstrate the applicability of the proposed
model.

The organization of the rest of this paper is as follows. We first present the
mean-variance decision framework which combines the conventional e-market seg-
mentation and the collaborative function in Section 2. The detailed e-market seg-
mentation schemes for estimation of advertisement budget and dynamic pricing
are proposed in Sections 3 and 4. The e-market segmentation schemes for other
functionality areas are discussed in Section 5. We conclude with the discussion
of managerial insights in Section 6.

2 Mean-Variance Decision Models

Before we present each particular e-market segmentation model for the specific
function area, we propose in this section the basic general decision Model. In
performing e-market segmentation, as we mentioned earlier, it is usual that the
marketing managers would make use of the demographic variables, geographic
variables, psychographic variables and behavioral variables. Despite the intu-
itive physical meanings behind these variables, some of the classifications with
these variables may not be very helpful for all specific collaborative functions.
For example, when the objective of a particular market segmentation project is
to decide the advertisement budget (and hence the advertisement strategy), the
profit that can be generated by each customer becomes a key measure. As a
result, we should incorporate a measure of the profit generated by the customers
in the market segmentation scheme. However, different customers can carry d-
ifferent profit-values to the company, a precision control rule is hence essential
for building a good market segmentation decision. In the advertisement budget
estimation example we mentioned above, a precision control rule can be imposed
on the degree of uncertainty of the profit. Thus, a mean-variance consideration
with which the average profit is used as the performance measure variable and
the variance of profit is applied as a precision control variable. In this paper, we
call the market segmentation scheme which includes the average objective per-
formance measure and the variance of this performance measure for a particular



Tsan Ming Choi : E-Market Segmentation for Internet-Mediated Fashion Brands... 172

market segmentation project the “mean-variance performance measure market
segmentation scheme (MVPM)”. With MVPM, we can carry out market seg-
mentation following a mean-variance consideration with which the segmentation
is performed with a measure by the “mean” and its precision is controlled by
the “variance”. Since the variance is an absolute measure, we make use of the
relative measure of the coefficient of variation, defined as “the standard deviation
divided by the mean” as the precision measure.

The general idea behind MVPM is that: The e-market segmentation scheme
follows the conventional type of segmentation policy and yields a number of e-
market segments. This segmentation process is treated as an initial and basic
e-market segmentation. After that, depending on different collaborative func-
tions, the manager of each function would impose an additional measure on each
of the segments. Systematic and precise evaluation is carried out and further
segmentation or re-segmentation may be required depending on the evaluation
results. By doing so, tailor-fit e-market segmentation results are provided to each
collaborative function and optimal decision can hence be made.

In the following sections, we outline the use of the concept of MVPM for several
important collaborative functionality areas with e-market segmentation.

3 Segmentation for Estimation of Advertisement Budget

The expenses companies spent on online advertisement is expected to increase in
the coming years. As reported in Bhatnagar and Papatla [21], Forrester Research
has estimated that the spending on online advertisement in the United States will
reach US$22 billion by 2004, which is more than 8% of the total spending for ad-
vertising in the United States. In fact, acquiring customers on the Internet can
be very expensive. As estimated and shown in Hoffman and Novak [25], many
EFBs have spent more than US$100 to acquire a new customer and some have
even spent US$500! However, the “values” of most new customers, as measured
by the expected lifetime spending on the EFB, are less than these advertising ex-
penses. In fact, some consumers shopping around the Internet only surf and buy
nothing; some of them may only make one purchase and never come back; some
may have repeated purchases and are loyal customers. As we mentioned above,
since it is expensive to acquire new customers on the Internet, the advertising
strategies should be more focused.

Recalling from the well-agreed Pareto rule (or called the 80-20 rule), the ma-
jority of profit is actually generated by a relatively small amount of customers.
It is thus a wise decision to focus the company’s resource on promoting to the
customers which can bring higher profits. Furthermore, the expenses spent on
advertisement for different groups (i.e. segments) of customers should be made
different. This can be achieved by a good e-market segmentation scheme. In the



173 Advances in Systems Science and Applications (2013) Vol.13 No.2

literature, there are a number of methods proposed to help. They include the
use of the customer’s searching behavior [21] and the use of personalized adver-
tisement [9,14]. In this section, we apply the MVPM to build a decision model
for the estimation of the advertisement budget for different market segments.

Under our proposed mean-variance model, MVPM, the EFB should incorpo-
rate the performance measure and the precision control measure for “the prof-
it that can be generated by the customer” into the decision model for market
segmentation. Thus, when EFBs try to evaluate the “value” of the customers
purchasing online, they should investigate the e-market segments with respect to
the profit generated by the customers in each of these segments with a precision
control. To be specific, we have the following market segmentation decision mod-
el:

“The EFB first segments the customers according to the conventional segmen-
tation scheme by, for example, the geographic and demographic variables. Then,
the EFB checks (and/or further segments) each group with respect to the aver-
age generated profit and the variance of profit from the members of that group.
The objective is to ensure that the coefficient of variation of profit, defined as
the standard deviation of profit divided by the average profit, for each segment is
under the EFB’s precision control and the segments size is large enough. After
that, the final outcome from each market segment will have an average value for
the customers in that segment and this average value can be a good representative
measure since its variation is under the EFB’s control.”

To illustrate the above statement, let us consider a simple example. Suppose
that an EFB has classified his customers according to the conventional segmen-
tation scheme with respect to the customers’ variables of location, gender, age
and education, and under a constraint on the size of each segment. After this
segmentation scheme, he has obtained different market segments. When the EFB
looks deeply into each obtained market segment, he can identify the profit that
has been generated by each customer inside each market segment. He can then
obtain the average profit and the variance of profit generated by all the members
in each market segment. For a particular segment, if the coefficient of variation
of profit is larger than a certain threshold (decided by the EFB), the level of
uncertainty of the profit generated by the members inside this market segment is
too large. Further market segmentation should be carried out by adding another
attribute (e.g. the purchear frequency of the customers). If the coefficient of vari-
ation of profit is less than a certain threshold (decided by the EFB), the market
segmentation that has been done is good enough and the EFB can stick with
it. However, if the coefficient of variation of profit is too large but the market
segments size is also relatively small, further market segmentation should not be
carried out. In this case, the EFB needs to reconsider carrying out the market



Tsan Ming Choi : E-Market Segmentation for Internet-Mediated Fashion Brands... 174

segmentation in another way with consideration of other variables at the very
beginning. We summarize this approach in the following proposition.

Proposition 1. The e-market segmentation scheme for estimating the adver-
tisement budget can be stopped when the coefficient of variation of profit generated
by the members of that segment is less than a threshold α. If the coefficient of
variation of profit is larger than α, then further segmentation or re-segmentation
is needed.

To give a better picture of the proposed decision model in Proposition 1, let
us have an illustrative numerical example below.

Example 1. An EFB has performed a market segmentation scheme by using
the conventional variables of city, gender and age and identified 18 e-market seg-
ments. Suppose that this EFB looks into two distinct segments: Segment 1 and
Segment 2, where Segment 1 refers to the group of customers who live in City 1,
male, and aged between 22-25, and Segment 2 refers to the group of customers
who live in City 2, female, and aged between 18-21. The desirable minimum size
of each segment is 500. For Segment 1, there are 1600 customers and for Segment
2, there are 850 customers. The profit generated by each one of the customers
can be found from the EFB’s database. When the EFB calculates the average
profit per head (AP), variance of profit (VP), standard deviation of profit (SDP),
and coefficient of variation of profit (CVP) generated by the members of each
segment, he has the following results,

Table 1.1 Example 1

Segment 1 Segment 2

Average Profit (in $) 50 180

Variance of Profit (in $2) 352 802

Standard Deviation of Profit (in $) 35 80

Coefficient of Variation of Profit 0.70 0.44

Obviously, even though the VP of the members in Segment 1 is smaller than the
members in Segment 2, the CVP is much larger. In fact, the profit uncertainty for
Segment 1 is too large to be ignored. Thus, if the precision threshold of the EFB
(α) is 0.5, then the e-market segmentation for Segment 2 is good enough because
its CVP is less than 0.5. However, the e-market segmentation for Segment 1 is
not good enough because Segment 1’s CVP is larger than α. Since the size of
this segment is 1600 and the minimum segment size is 500, the EFB can consider
carrying out further segmentation on Segment 1 by using another classification
variable. After the e-market segmentation scheme with the precision control over
the profit generated by the members of each segment, the EFB can make use
of the estimated average profit for each segment as an indicator to decide the



175 Advances in Systems Science and Applications (2013) Vol.13 No.2

amount of budget for advertisement on each segment. In this example, the value
for each customer in Segment 2 is $180 and a reasonable budget (say $100 per
head) for advertising towards customers in this market segment can hence be
estimated.

As summarized in Proposition 1 and illustrated by Example 1, we can see
that a decision model which provided a tailor-made stopping condition for the
e-market segmentation process has been proposed. As we mentioned earlier, ad-
vertisement for e-commerce can be very expensive. Nowadays, on-site banner,
pup-up screen, emailing, affiliate program, TV and radio commercials, etc are
all popular means of advertisement. However, obviously, different means of ad-
vertisement carry different costs. As a result, it is a wise decision to decide the
advertisement budget for each specific group of customers before considering the
specific means of advertisement. By having the market segmentation scheme
as described in Proposition 1 where the members of each segment are grouped
together with the consideration of the average profit generated under precision
control, we can identify precisely the “value” of each member of that particular
market segment. As a consequence, the EFB can allocate the optimal advertise-
ment resource to focus on the most profitable customers, and decide the most
appropriate advertisement scheme for them.

4 Segmentation for Dynamic Pricing

The online consumers have different sensitivities towards price and other non-
price factors. Some consumers are highly price-sensitive and they like to use the
shop bots for finding the EFBs which offer the lowest prices while some care more
about service, reliability and trust. As a result, how to set the right selling prices
dynamically for different groups of customers so that the EFB’s profit is maxi-
mized becomes very important and, in fact, crucial (the failure stories due to the
lack of good pricing policy can be found in [28]). One of the effective ways for
dynamic pricing is to carry out dynamic price testing. The idea of the dynamic
price testing [27,30] is that: From changing the listed price showing on an EFB’s
website and keeping track of the customers’ purchasing rates at that price, the
EFB can know the expected profitability of each listed price. For example, when
the EFB sets the product’s listed price as $10, it is observed that 2 out of 10
visitors will buy the product; when the EFB changes and tests the price at $9.5,
it is found that 3 out of 10 visitors will buy the product. The purchasing rates are
20% and 30% for $10 and $9.5, respectively. The EFB can then decide whether
$9.5 is better than $10 or not based on the corresponding observed purchasing
rates and profit margins. However, in order to exercise an effective dynamic price
testing scheme (and hence a good dynamic pricing), the EFB needs to segment
the e-market properly. From the sales record of the customers for that product



Tsan Ming Choi : E-Market Segmentation for Internet-Mediated Fashion Brands... 176

(or a closely related product if the data for that product is not sufficient or avail-
able) in the past, the EFB can keep track of the exact purchasing price of each
customer. As a result, the EFB can first segment the e-market for that product
according to the conventional approach by the customers’ age, gender, etc. After
that, the EFB can check the average purchasing price and the variation of the
purchasing price per each segment following the concept from MVPM. Similar to
the proposed method for the segmentation for advertisement budget estimation,
we have the following model:

“From the database of the customers, the EFB first segments the customers
according to the conventional segmentation scheme by, for example, the geograph-
ic and demographic variables. Then, the EFB further segments each group with
respect to the average purchasing price and the variance of purchasing price from
the customers in that group. The objective is to ensure that the coefficient of
variation of the purchasing price, defined as the standard deviation of purchasing
price divided by the average purchasing price, for each segment is under the EF-
B’s precision control and the segments size is large enough.”

With the above method, the EFB can effectively identify the group of cus-
tomers with a specific average purchasing price. After that, a tailor-made dy-
namic price testing scheme can be arranged for that particular market segment.
It is thus more effective than performing the dynamic price test blindly. We
summarize this proposed method in Proposition 2 below.

Proposition 2. The e-market segmentation scheme for dynamic price testing
(and hence dynamic pricing) can be stopped when the coefficient of variation of
the purchasing price of the members in that segment is less than a threshold θ.
If the coefficient of variation of profit is larger than θ, then further segmentation
or re-segmentation is needed.

Example 2 below gives an illustrative numerical example for the proposed e-
market segmentation model presented in Proposition 2.

Example 2. An EFB has performed conventional market segmentation (by
using the variables such as age, city, gender, etc) for the customers of a specific
product and has identified 20 e-market segments. Suppose that this EFB looks
into two distinct segments: Segment A and Segment B. The desirable minimum
size of each segment is 500. For Segment A, there are 1300 customers and for
Segment B, there are 1500 customers. When the EFB calculates the average
purchasing price, variance of purchasing price, standard deviation of purchasing
price, and coefficient of variation of purchasing price generated by the members
of each segment, he has an khown in Table 1,

From Table 1, the coefficient of variation of the purchasing price for customers
in Segment B is much larger than the customers in Segment A. Suppose that the
EFB has set the precision threshold θ(for the coefficient of variation of the pur-



177 Advances in Systems Science and Applications (2013) Vol.13 No.2

Table 1.2 Example 2

Segment A Segment B

Average Purchasing Price (in $) 100 90

Variance of Purchasing Price (in $2) 252 502

Standard Deviation of Purchasing Price (in $) 25 50

Coefficient of Variation of Purchasing Price 0.25 0.56

chasing price) to be 0.5. By Proposition 2, the e-market segmentation result for
Segment A is good enough because the coefficient of variation of the purchasing
price is less than θ. Thus, for consumers in Segment A, the EFB can carry out
the dynamic price test with a reference price of $100 and the dynamic testing
prices can be set within a reasonable range. On the other hand, the e-market
segmentation for Segment B is not good enough because Segment B’s coefficient
of variation of the purchasing price is larger than θ. Since the size of this seg-
ment is 1500 and the minimum segment size is 500, the EFB can consider further
segmenting Segment B by using another classification variable.

5 Segmentation for Retaining Customers, Direct Marketing and Others

In Sections 3 and 4, we have discussed the e-market segmentation schemes, follow-
ing the concept of MVPM, for two important collaborative functions for EFBs.
We will discuss more functionality areas where MVPM can be applied for e-
market segmentation in this section.

As we mentioned earlier, trust and customers’ loyalty are two important issues
for EFBs doing business on the Internet. Without proper e-market segmentation,
building and keeping the loyalty of the online customers becomes more difficult.
As a result, when the EFB performs the e-market segmentation, it is important
for him to bear in mind that he has to be able to identify precisely the need
of the customers inside that market segment and be focused [1]. This objective
follows exactly the MVPM where the need of the customers inside each e-market
segment refers to the average measure of that need and the precision for the
understanding of this need in the corresponding e-market segment is reflected
by the coefficient of variation of that need measure. This is what the MVPM
captures. As an example, suppose the EFB would like to retain the customers
by providing a membership system. In order to attract the customers, the EFB
would like to provide bonus points for the customers who join the membership
and make some purchases. Obviously, in order to attract more customers to be
members, the amount of the bonus points and the amount of required purchases
should be offered differently to customers in different segments. The concept of
MVPM can hence be applied for providing a measure for this purpose.



Tsan Ming Choi : E-Market Segmentation for Internet-Mediated Fashion Brands... 178

Similarly, for an EFB who wants to provide the right product and service to
the right customer for the right price at the right time in the right place with
direct marketing strategy [31-32] , the MVPM also works where it provides a
mechanism for e-market segmentation which helps to narrow down the error in
terms of the product type, selling price, etc. For instance, the EFB can make
use of the data on the products bought by the customers in the past to help in
segmenting the customers and measuring the customers likelihood of buying the
product that will be direct marketed.

6 Conclusion and Managerial Implications

We have proposed in this paper a conceptual decision model, called the “mean-
variance performance measure market segmentation scheme (MVPM)”, for e-
market segmentation for different specific collaborative functions. As we all know,
the traditional market segmentation relies on the classification variables like the
demographic, geographic, psychographic and behavioral variables. However, in
spite of the intuition behind the classification by using these variables, researchers
have questioned the reliability of the available market segmentation techniques
(e.g. [8]). Furthermore, the classifications with these conventional variables may
not be very helpful for a specific collaborative function. As a result, we propose
to incorporate the collaborative function’s performance measure and its precision
into the e-market segmentation scheme for effective decision making by the man-
agers of the corresponding functions.

A mean-variance consideration with which the average or expected perfor-
mance measure (e.g. the average profit) is used as the performance measure
variable and the degree of variation of the performance measure (e.g. the co-
efficient of variation of profit) is applied as a precision control variable. This
market segmentation scheme is thus called the mean-variance performance mea-
sure market segmentation scheme (MVPM). With MVPM, EFBs can carry out
e-market segmentation for each particular collaborative functionality area fol-
lowing a mean-variance consideration. We have discussed the use of MVPM for
EFBs with the estimation of advertisement budget, dynamic pricing, retaining
customers and direct marketing.

Since brand managers are involved with the resource allocation and decision
making for the EFBs, effective and precise market segmentation schemes can pro-
vide them with good assistance in making a sound and optimal decision. Under
MVPM, the e-market segmentation decision is under a control on the precision of
the specific performance measure. As a consequence, better resource allocation
decisions for different collaborative operations can be made by the fashion brand
managers optimally and precisely. We can thus view MVPM as an uncertainty
control model which targets at reducing and constraining the degree of uncertain-



179 Advances in Systems Science and Applications (2013) Vol.13 No.2

ty associated with the performance measure. We believe that MVPM is especially
important for e-commerce (and mobile commerce) because the e-market is highly
volatile with a large amount of uncertainty sources and customers details can
easily be recorded and analyzed. Notice that, even though we focus on the use
of MVPM for Internet enabled e-market segmentation schemes, the conceptual
framework of MVPM can actually be applied to the general market segmentation
systems. The model of MVPM can also be implemented into an intelligent com-
puterized decision support system which automatically helps managers in making
wise and scientifically sound decisions.

Acknowledgements
The authors are indebted to the anonymous referees for their critical comments
(on the earlier version of the paper) which improve this paper substantially. This
paper is supported in part by The Hong Kong Polytechnic University with re-
search funding account of G-YJ23.

References

[1] Reichheld F and Schefter P. (2000), E-Loyalty: Your secret weapon on the
web, Harvard Business Review, Vol.78, No.4, pp.105-113.

[2] Kiang M.Y, Raghu T.S and Shang K.H.M. (2000), “Marketing on the Inter-
net - who can benefit from an online marketing approach?”, Decision Support
Systems, Vol.27, pp.383-393.

[3] Liu Y, Kiang M and Brusco M. (2012), “A unified framework for market seg-
mentation and its applications”, Expert Systems with Applications, Vol.39,
pp.10292-10302.

[4] Chaffey D. (2002), E-Business and E-Commerce Management, 1st Edition,
Prentice Hall.

[5] Dibb S, Simin L, Pride W and Ferrell O. (2000), Marketing Concepts and
Strategies, 4st Edition, Houghton Mifflin, Boston, MA.

[6] Smith W.R. (1956), “Product differentiation and market segmentation as
alternative product strategies”, Journal of Marketing, Vol.20, No.7, pp.3-8.

[7] Gary R.K. (2001), “New dimension of Internet buyer behavior: Strategic
marketing implications”, The Proceedings of The First International Con-
ference on Electronic Business, Hong Kong.

[8] Dibb S and Stern P. (1995), “Questioning the reliability of market segmen-
tation techniques”, Omega, Vol.23, No.6, pp.625-636.



Tsan Ming Choi : E-Market Segmentation for Internet-Mediated Fashion Brands... 180

[9] Raghu T.S, Kannan P.K, Rao H.R and Whinston A.B. (2001), “Dynamic
profiling of consumers for customized offerings over the Internet: a model
and analysis”, Decision Support Systems, Vol.32, pp.117-134.

[10] Haley R.I. (1968), Benefit segmentation: “A decision-oriented research tool”,
Journal of Marketing, Vol.32, pp.30-35.

[11] Johnson R.M. (1971), “Market segmentation: A strategic management tool”,
Journal of Marketing Research, Vol.8, pp.13-18.

[12] Darden W.R and Perreault W.D. (1977), “Classification for market segmen-
tation: an improved linear model for solving problems of arbitrary origin”,
Management Science, Vol.24, No.3, pp.259-271.

[13] DSS Research, Understanding market segmentation, Downloadable from
http://www.dssresearch.com/marketsegment/Library/Segment/understan-
ding.asp

[14] Kim J.W, Lee B.H, Shaw M.J, Chang H.L and Nelson M, (2001), “Applica-
tion of decision-tree induction techniques to personalized advertisements on
Internet storefronts”, International Journal of Electronic Commerce, Vol.5,
No.3, pp.45-62.

[15] Shaw M.J, Subramaniam C, Tan G.W and Welge M.E. (2001), “Knowledge
management and data mining for marketing”, Decision Support Systems,
Vol.31, pp.127-137.

[16] Montgomery L. (2001), “Applying quantitative marketing techniques”, In-
terfaces, Vol.31, No.2, pp.90-108.

[17] Brusco M.J, Cradit J.D and Stahl S. (2002), “A simulated annealing heuristic
for a bicriterion partitioning problem in market segmentation”, Journal of
Marketing Research, Vol.39, pp.99-109.

[18] Novak T.P, Hoffman D.L and Yung Y.F. (2000), “Measuring the customer
experience in online environments: A structural modeling approach”, Mar-
keting Science, Vol.19, No.1, pp.22-42.

[19] Moorthy K.S. (1984), “Market segmentation, self-selection, and product line
design”, Marketing Science, Vol.3, No.4, pp.288-307.

[20] Winter F.W. (1989), “Market segmentation using modeling and Lotus 1-2-
3”, Interfaces, Vol.19, No.6, pp.83-94.



181 Advances in Systems Science and Applications (2013) Vol.13 No.2

[21] Bhatnagar A, and Papatla P. (2001), “Identifying locations for targeted ad-
vertising on the Internet”, International Journal of Electronic Commerce,
Vol.5, No.3, pp.23-44.

[22] Turban E, Lee J, King D and Chung H.M. (2012), Electronic Commerce -
A Managerial Perspective, 7th Edition, Pearson.

[23] Moe W. W and Fader P.S. (2001), “Dynamic conversion behaviour at e-
commerce sites”, Management Science, Vol.50, pp.326-335.

[24] Betts M. (2001), “Turning browsers into buyers”, MIT Sloan Management
Review, Winter, pp.8-9.

[25] Hoffman L and Novak T.P. (2000), “How to acquire customers on the web”,
Harvard Business Review, Vol.78, No.3, pp.179-188.

[26] Chatranon A, Chen J.C.H, Chong P.P and Chen Y.S. (2001), “Customer
relationship management (CRM) and e-commerce”, The Proceedings of The
First International Conference on Electronic Business, Hong Kong.

[27] Baker W, Marn M, and Zawada C. (2001. February), “Price smarter on the
net”, Harvard Business Review, Vol.79, No.2, pp.122-127.

[28] Dutta S, Bergen M, Levy D, Ritson M and Zbarachi M. (2002), “Pricing as
a strategic capability”, MIT Sloan Management Review, Spring, pp.61-66.

[29] Markowitz H.M. (1959), Portfolio Selection: Efficient Diversification of In-
vestment, New York: John Wiley & Sons.

[30] Choi T.M, Chow P.S and Xiao T. (2012), “Electronic price-testing scheme
for fashion retailing with information updating”, International Journal of
Production Economics, Vol.140, pp.396-406.

[31] Kenny D and Marshall J.F. (2000), Contextual marketing: the real business
of the internet, Harvard Business Review, Vol.78, No.6 pp.119-125.

[32] Rust R.T and Lemon K.N. (2001), “E-service and the consumer”, Interna-
tional Journal of Electronic Commerce, Vol.5, No.3, pp.85-101.

Corresponding author
Tsan-Ming Choi can be contacted at jason.choi@polyu.edu.hk.