Faculty of Economics and Business MASTER’S THESIS
Customer Segmentation on the Basis of Mobile shopping and the Use
of Shopping Applications
Xiaotong Wang (10864636)
Msc. in Business Administration – Marketing track Supervisor: Dr. U. (Umut) Konus
2 Statement of Originality
This document is written by Student Xiaotong Wang who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 Abstract
With a number of innovations in mobile commerce services, shopping via the mobile channel brings a great opportunity and would be the major business channel in the future. The mobile shopping applications have been developed and already in use, which facilitates mobile shopping process. In this paper, we aim to expand current understanding of consumer usage behavior in mobile shopping context by executing segmentation study. Our research specifically focuses on mobile shopping applications which are regarded as the trending shopping channel in recent years. Besides, the study focuses on the Chinese market since it is the typical emerging market that attracts many scholars and practitioners’ attention. Specifically, survey data has been collected from 275 Chinese customers who have online shopping experience; the segmentation study based on their usage purpose and frequency are conducted separately by a latent class cluster analysis (LCCA). The authors aim to (1) segment consumers based on the usage of mobile shopping applications in three product categories (2) examine whether or not the demographic, psychological variables and expected benefits are significantly associated with segment membership3)All identified segments are labelled and compared. Main findings indicated the appearance of flight tickets-averse usage segment, no-oriented usage purpose segment and non-usage segment in terms of customer usage purpose; and the appearance of light usage segment, moderate usage segment, heavy usage segment, heavy usage only for clothes segment and light usage only for flight tickets segment were identified on the basis of usage frequency. The active covariates that could predict the segment membership are also identified, including gender, education level,
4 innovativeness, privacy and ease of use. The implications and limitations are also discussed for further research.
Keywords: Mobile shopping; Mobile shopping applications; Customer segmentation;
Usage behavior (usage purpose, usage frequency); Covariates; Latent Class Cluster Analysis (LCCA)
5 Table of Content Statement of Originality ...2 Abstract ...3 1. Introduction ...9 2. Literature review ... 14
2.1 Mobile Shopping and Mobile Shopping Applications ... 14
2.1.1 Mobile shopping and applications development ... 14
2.1.2 The usage behavior of mobile shopping applications ... 16
2.1.3 Product categories and mobile shopping applications usage ... 18
2.1.4 The drivers of mobile shopping adoption ... 19
2.2 Customer Segmentation ... 22
2.2.1 Customer segmentation in marketing research ... 22
2.2.2 Customer segmentation approaches ... 23
2.2.3 Customer segmentation in multichannel research ... 26
2.2.4 Demographic and psychographic characteristics in customer segmentation .. 29
3. Conceptual Framework ... 30
3.1 The usage behavior as segmentation indicator ... 31
3.2 Covariates of mobile shopping applications usage ... 32
3.2.1 Demographical covariates ... 33
3.2.2 Psychological covariates... 34
6
3.3 Conceptual framework and expectations ... 37
4. Methodology ... 39
4.1 Research design ... 39
4.2 Pilot study ... 40
4.3 Main study... 40
4.4 Latent-class cluster model... 41
5. Analysis and results ... 43
5.1 Descriptive analysis ... 43
5.2 Reliability analysis ... 44
5.3 Segmentation analysis ... 44
5.3.1 Optimal model selection... 45
5.3.2 Interpreting usage purpose segment model ... 47
5.3.3 Interpreting usage frequency segment model ... 52
5.3.4 Comparing segments in two models ... 57
5.3.5 Combination of main segments in two models ... 58
6. Discussion... 61
6.1 Main findings ...61
6.1.1 Identified customer segments ... 62
6.1.2 Active covariates ... 64
6.2 Theoretical implications... 66
6.3 Managerial implications ... 67
7
7. Reference ... 71
8. Appendix ... 80
Appendix 1 China Mobile Shopping Market in 2015 ... 80
Appendix 2 The questionnaire ... 80
Appendix 3 Descriptive Analysis ... 83
9. Acknowledgement... 84
List of Tables Table 1. Drivers of mobile shopping adoption ... 21
Table 2. Customer segmentation approaches ... 25
Table 3. Common segmentation bases ... 25
Table 4. Customer segmentation in multichannel research ... 29
Table 5. Indicators and covariates in the segmentation study ... 43
Table 6. Reliability Analysis ... 44
Table 7. Optimal model selection (usage purpose model) ... 46
Table 8. Optimal model selection (usage frequency model) ... 47
Table 9. Composition analysis (usage purpose model) ... 48
Table 10. Profiling usage purpose model ... 51
Table 11. Usage purpose model summary ... 52
Table 12. Composition analysis (usage frequency model) ... 53
Table 13. Profiling usage frequency model ... 56
8 Table 15. Two models comparison... 58 Table 16. Combination of main segments in two models ... 60
List of Figure
9 1. Introduction
In the past, web-based online shopping (via a PC browser) had overtaken store shopping in popularity. This trend is changed by the emergence of smart phone that provides customers with a new shopping channel. And the mobile shopping is believed to be the alternative of online shopping. Mobile shopping is a typical example of mobile commerce that consists of any indirect or direct transactions with a monetary value. Wireless telecommunication network makes mobile commerce work (Kleijnen, Ruyter and Wetzels, 2007).Compared with other shopping channels, the mobile shopping channel offers consumers many benefits. For example, they could purchase anything they prefer without leaving their place. Mobile shopping does not just mean visiting the shopping website via mobile phone. The sudden growth of latest mobile applications technology has totally changed the smartphone users to be the potential smartphone shopper. Many mobile applications are accepted by an increasing number of consumers, such as mobile advertising, entertainment services. Among all these applications, mobile shopping applications are considered to have the largest potential. Much evidence from industry sources in developed countries shows that using mobile phone applications, especially in shopping process, is increasing significantly. From looking at mobile applications usage research from Flurry Analytics (Bosomworth ,2015) , the largest two growth categories were shopping applications and lifestyle applications, which showed a 174 percent increase from 2013, more than double the average of the rest of the categories. The massive year-over-year growth shows that consumers are willing to accept the mobile
10 applications for shopping and other daily tasks. However, if the focus only targets at the developed countries, the results may not be sufficient. Current marketing researchers shift their interest to the developing countries, such as China. The latest data from I-Research consulting group indicated that mobile shopping occupied 47.8% in online shopping market in the first quarter 2015 and predicted that this proportion will exceed over 50 percent in China 2015(I research consulting group, 2015). It is obvious that mobile shopping experienced remarkable development in China. In light of this, understanding the usage behavior of mobile shopper is of importance. As Wu and Huang (2009) suggests “understanding consumer behavior is critical for successful management and development of mobile shopping”.
In mobile shopping context, the decision making process is very similar with the traditional consumer decision model. In the traditional model, consumer’s decision process includes need awareness, information searching, alternative comparison, purchase decision and finally post-purchasing behavior. One major difference in mobile shopping context is the shopping environment. In the specific mobile technology-mediated shopping environment, the limited understanding of consumer behavior will hinder marketers’ ability to develop appropriate shopping services in every stage of mobile shopping process. Therefore, Predicting and understanding the consumer usage behavior in mobile shopping context is important since it will help marketers get approval of customers easily. In other words, if the marketers deepen their insight of consumers’ usage behavior in shopping process and their expected benefits derived from mobile shopping applications, they will both attract new
11 customers and serve the current customers better.
The segmentation study could contribute to a better understanding and characterization of usage behavior. Segmentation is defined as the process of developing meaningful customer groups; in each group, people have similarity in characteristics and behaviors (Wayne Thompson, 2008). This process starts with observing customers actions and ends with the learning of demographic and psychographic characteristics. Customer segmentation could help companies deepen their insight into consumers’ usage behavior of channel usage; in this way, the firms could add channel value to customers. The disclosure of information via customer segmentation enables marketers to not only understand the consumer behavior but also refine marketing strategies to meet customers’ expectations (Wu and Chou, 2011). Customer segmentation is also closely related to the marketing strategy. Mobile strategies associate with channel optimization and resource allocation, which determines company’s success. As a result, customer segmentation is important and necessary for the mobile shopping research.
According to the previous literatures, many studies associated with customer segmentation include the demographical and psychological profile of consumers. Profiling customer segmentation is also important for marketers because they could use consumers’ demographical and psychological characteristics to predict their behavior and needs, thus tailoring marketing mix to the different customer groups. Even though the customer segmentation is important for the understanding of mobile shopping behavior, relevant research in mobile shopping context is relatively limited.
12 Regarding to the customer segmentation, previous literatures engage in segmenting consumers only for pc channel and general multichannel channel (Konus, Verhoef and Neslin, 2008; Bhatnagar and Ghose, 2004; Andrew and Vaitha, 2004). For the specific mobile shopping applications channel, there is nothing research about marketing segmentation studies. Besides, most relevant studies about customer segmentation investigate those customers outside china. But it is meaningful and urgent to study the mobile shopping market in Chinese context due to its obvious intriguing development and huge market opportunity. In light of this, there is a research gap between usage behavior in using mobile shopping applications and relevant customer segmentation, especially in Chinese market. Our paper aims to fill this research gap by segmenting customer based on mobile shopping applications usage. Since the Chinese market is our research context, the samples will be only Chinese customers. To be more specific, our research question is
1．Whether or not we have distinct identified segments based on the usage of mobile shopping applications?
2．Whether or not there are differences in demographic and psychological characteristics based on the usage of mobile shopping applications among segment membership?
3．Whether or not there are differences in expected benefits based on the usage of mobile shopping applications among segment membership?
This study contributes to the current state of mobile shopping market, particularly among Chinese market. From the academic perspective, this paper focuses on the
13 mobile shopping applications that are considered as the trending shopping channel. More importantly, we conduct the customer segmentation study based on this new shopping channel, which makes the theoretical contribution in multichannel shopping research. This is because relevant literatures about the consumer behavior of mobile applications shopping channel are few and there is no literature about the customer segmentation study in this area. Hence, this paper fills this research gap and thus makes the academic contribution to this trending topic. In practice, practitioners are constantly attempting to identify the most profitable markets and valuable customers. To do this, companies need to employ some form of market segmentation which is the second foundation block of relationship marketing (Zeithaml and Bitner, 2006). Different segments need different strategies for each segment; company also should use resources in different ways for different segments. Therefore, this paper also minimizes “discrepancies between academic developments and real world practice” (Wind, 1978) in market segmentation. Adopting a segmentation approach will help marketers better understand consumers and their target groups, thus better informing their strategic choices. It also has solid benefits for the effective use and allocation of resources, such as time, budget and human resources.
The paper structure is as follows. In the next section, we discuss the relevant previous literature; In Section 3, we present our conceptual framework and expectations. Section 4 describes the methodology and conduct the segmentation study. Finally, we conclude in Section 5 with managerial implications and directions for future research.
14 2. Literature review
2.1 mobile shopping and mobile shopping applications 2.1.1 Mobile shopping and applications development
The emergence of mobile commerce allows consumers to purchase via electronic store. It takes places on the mobile phone and provides consumers with shopping experience similar to the online shopping through laptop computer (Aldás, Ruiz-Mafe and Sanz-Blas, 2009). With the trend of mobile commerce, the relevant mobile applications also have been developed and aim to facilitate the mobile commerce service. These applications help marketers achieve different types of business functions via smartphone, such as banking, advertising and shopping (Wu and Wang, 2006). Among these innovative mobile commerce services, shopping via mobile channel brings a great potential opportunity (P. May, 2001).
According to the definition by Ko, kim and Lee (2009), the mobile shopping refers to the usage of wireless internet services while shopping via smart phones. The shopping products include consumer goods and services, such as retail items and booking tickets online (Ozok and Wei, 2010). The mobile shopping has several unique characteristics, including ubiquity, net-access convenience, localization, personalization, and various types of mobile applications (Kim et al., 2007; Lee et al., 2005; Lee and Park, 2006; Liao and Xu, 2005). These features create different consumers’ needs compared with other shopping channels (e.g. PC online shopping, in-store, catalog). The study by Barnes (2002) indicates that the new mobile technologies equipped with mobile applications and micro-browser provide internet ‘in your pocket’.
15 The mobile applications, also called native applications, are trending in recent years. It offers a user-friendly experience to search information and purchase via smartphones or tablets (Grøtnes, 2009). Retailers normally have two traditional mobile channel choices, including the mobile web site and website applications. The mobile website refers to the online website viewed on the smartphone browser (Wong, 2012); the way to use web applications is that consumers are access to the retailer’s website via mobile browser which has been formatted for mobile usage(Goldman, 2010).All these two traditional forms need to use mobile browser to be access to retailers’ website. By contrast, native applications offer marketers a new option. Users could download the native applications from mobile application store. The store is kind of database which includes different types of available applications. Consumers download their needed applications from database and install it onto their mobile device (Wong, 2012).Some of retailers also benefits from the emergence of native applications. Sales via mobile shopping applications increased dramatically. For example, IMRG (IMRG, 2011b) shows that there were 4.2 million consumers who visited retailers’ website, nearly a third of mobile shoppers use native applications to purchase (IMRG, 2011b). The growth of mobile shopping applications usage is becoming apparent. In light of this, our research focuses on native mobile shopping applications.
In the mobile digital age, marketers aim to strongly influence consumers’ shopping habits, which require online retailers to be up-to-date with mobiles shopping applications. There are some common types of mobile shopping applications based on
16 the usage purpose. The first type is relevant to discount and deal givers. Consumers are able to not only acquire information about competitors’ promotion but also look into the popular deals. Social shopping applications, as the second type, are adopted widely by shopping lovers. For instance, Pinterest provides fashion people with a platform where they even could boost their own business. Consumers mostly vote their favorite items and compare prices on this kind of social platform. The third one is the price comparison helpers. As the name indicates, this type of applications is specifically made for comparing products and price across markets. Humongous marketplaces are fourth type of mobile shopping applications. An increasing number of huge marketplaces have entered mobile shopping market. Amazon is a good example for allowing mobile shoppers to search information and purchase via smartphone. The last type refers to the payment applications. This kind of digital wallets facilitate the payment process during the purchase stage (Park, 2011). These different types are closely associated with consumers’ usage behavior and purpose in mobile shopping process.
2.1.2 The usage behavior of mobile shopping applications
Mobile shopping applications help customers to find products at competitive prices without limitation of place. It offers the whole flow of the shopping experience that is similar to the traditional shopping process, including information search, price comparison, alternative evaluation, purchasing and after service (Ranganathan and Ganapathy,2002).Compared with the five stages model of consumer behavior in traditional shopping environment, Haübl and Trifts (2000)suggest that potential online
17 consumers use two-stage process of screening products to identify a promising subset and then making a purchase decision. As for the mobile channel, the industry sources also indicates that consumers use mobile shopping channel not only to just purchase the products but also to search information about price and reviews from other customers (Charlton, 2011).
The information searching is a significant usage purpose in mobile shopping behavior (Seock and Norton, 2007). Before they purchase, enough information is necessary since it could allow customers to compare prices and quality among similar products; during this stage, the well-designed and clear layout mobile applications is more attractive for consumers to acquire useful knowledge so that they are more likely to purchase products and services (Song, Koo and Kim., 2008).Purchasing is another important purpose for the usage of mobile shopping applications. For example, customer sometimes searches information about products via offline channel, such as retail store; but they choose purchase products via mobile shopping applications since mobile channel may provide them with some discounts or they just believe the applications are convenient for buying products. In term of these two main usage purposes, the previous study of multichannel shopping usage behavior shows that a great number of consumers use the same channel for information and purchasing purposes (Schröder and Zaharia, 2008). However, the research from Holmes, Byrne ,and Rowley (2013) suggests that consumers have high level of using mobile phone in information search and other pre-purchase activities; instead, they have low level of using it for purchase purpose. As a result, consumer may have different usage
18 purposes when they use mobile shopping applications. In our study, we employ two main purposes- information search and purchase-to investigate usage behavior of mobile shopping applications.
2.1.3 Product categories and mobile shopping applications usage
Previous research about consumer behaviors shows that the type of product and consumers’ decisions influence how consumers perceive the value of these products, in both purchase-intention stage and post-purchase stage (Levin, Irwin, and Weller, 2005).The difference of product types may influence the usage of mobile shopping applications. Earlier studies have focused on why online shopping differs across products. Several studies have shown that “high touch” products, which means consumers feel they need to touch or try, are popular in offline shopping channel at least at purchase stage (Dholakia, 2003)(Kim and Jin, 2006). Levin et al. (2005) shows that “low touch” products like flight tickets and electronic products that generally favor online services due to the convenience derived from online shopping. Compared with online shopping, the research on the usage of mobile shopping applications also needs to take the difference in product categories into consideration. The findings from Brown, Pope, and Voges (2003) also supported that the product type is more likely to be predictors of online purchase intentions since the shopping orientation is different among product categories. And the statistics (Statistics Denmark, 2007) showed that clothing is one of the most common product categories purchased online.
19 2.1.4 The drivers of mobile shopping adoption
Consumers have different shopping motivation in using mobile shopping applications, which may lead to dramatically different shopping patterns. Previous studies show that identifying motivations could assist marketers to proactively design mobile shopping services and thus drive customers to use mobile channel (Yang and Kim, 2012).
Many previous researches engaged in the investigation of drivers of mobile shopping acceptance. The adoption measures include a variety of perspective, such as behavioral intention to use, purchase intention and consumption experience. As the table1 shown, researches on the acceptance of mobile technology in shopping context normally based on the technology acceptance model (TAM)(Davis,1989).It was developed to explore the drivers of affecting the acceptance of mobile commerce. The traditional TAM model is the leading model which has been used widely to explain users’ acceptance towards a new technology. The model includes the perceived usefulness, the perceived ease of use, and the attitude towards using a technology. The advantage of TAM is that this model focuses on the technological aspect in various contexts; by contrast, the weakness is insufficient consideration of the effect of individual factors on the technology adoption.
Based on the TAM model, other theories are then used and extended with drivers which influence consumers’ acceptance of mobile service. As indicated in the Table1, the theoretical bases consist of Revised TAM model (Wu and Wang, 2005; Lu and Su, 2009).Wu and Wang (2005) employed the Revised TAM model to explore the factors
20 that determine the mobile commerce acceptance, including the perceived risk, cost, compatibility and perceived usefulness. Lu and Su (2009) also use this model to investigate those factors that affect the purchase intention on mobile shopping websites. The result shows that anxiety has negative relationship with purchase intention whereas enjoyment, usefulness, and compatibility are positively correlated with behavioral intentions.
The information system success theory is developed by De Lone and McLean (2003). System quality, information quality, and service quality are considered as three main success attributes of information system, which significantly influence users’ usage and satisfaction. Recent research extend the information system success theory to identify perceive ease of use, perceived usefulness and trust have significant effect on satisfaction (Tao Zhou, 2011).this theory is also suitable to the mobile context. Those main factors from the information system success theory reflect users’ perception of mobile websites (Chung and Kwon, 2009)
The emotion-related theory is developed by Li and Chen (2012).As the table1 indicates, it is suggested that mobile commerce should focus on mobile commerce experience which are determined by both utilitarian and hedonic factors. Utilitarian shopping motivation stresses that the shopping behavior depends on the functional features and financial desires (Kim, 2006), including efficiency and achievement; by contrast, hedonic shopping motivation refers to the emotion feeling and psychological sensations, including adventure, stimulation, shopping for excitement, and experiencing a different environment (Arnold and Reynolds, 2003; Kim, 2006).the
21 finding shows that utilitarian factors has negative effect on consumer experience while hedonic factors are positively related to the consumer experience. Other emotion-related researches from Ko et al.(2009) and Li, Yeh (2010) also examines the influence of other hedonic factors on the experience ,such as aesthetics, enjoyment and escapism.
Lastly, the diffusion of innovation theory in the table 1 explores factors which affect the adoption of new technology. According to Rogers (1995), these factors are complexity, compatibility, tialability, observability, and relative advantage. Previous study have consistent conclusion about these factors, especially for the ease of use, compatibility and relative advantage; these three attributes act as most frequently important factors for the adoption of internet and mobile technology (Koening-Lewis, Palmer and Moll.; Liu and Li 2010; Papies and Clement 2008).
Table1: Drivers of mobile shopping adoption
Theoretical base Drivers of mobile shopping adoption
Representative literatures TAM model Perceived usefulness;
perceived ease of use;
subjective norm; self-efficacy; enjoyment; network externality Davis,1989 Li and Yeh, 2010 Revised model(TAM)
perceived risk, perceived cost, compatibility, and perceived usefulness
anxiety, enjoyment, usefulness, and compatibility Wu and Wang , 2005 Lu and Su , 2009 The information system success model
System quality, information quality and service quality ease of use;
usefulness
Delone and McLean, 2003 Wang and Liao, 2007 Kim et al., 2009; Lee and Chung, 2009; Chung and Kwon, 2009 Emotion-related utilitarian factor Arnold and Reynolds, 2003
22
theory hedonic factors Kim, H. S. , 2006
Ko et al., 2009; Li and Yeh, 2010 Li et al. 2012 Innovation Theory complexity, compatibility,
tialability , observability, and relative advantage
Rogers, 1995;
Papies & Clement 2008 Koenig-Lewis et al. 2010; Liu & Li , 2010
2.2 Customer segmentation
2.2.1 Customer segmentation in marketing research
With the change of channel and consumer buying behavior, it is important to understand the nature of their shopping behavior. The conception of segmentation becomes more relevant for understanding consumers, the distribution of resources and also the marketing approaches development(Palmera and Millierb, 2004).Customer segmentation, is also known as market segmentation, is the process in which some homogenous sub-groups are identified in the heterogeneous aggregate market. The segmentation basis refers to those variables that are used to distribute potential customers to homogeneous groups (Wedel and Kamakura, 2000). This definition addresses that the essence of segmentation is to differentiate consumers by using certain variables. From both marketing research and practice perspective, the segmentation refers to the act of defining sub-groups of individuals which are meaningful for company profits (Hajer, Kamel and Noureddine, 2006).The segmentation approach is important and often employed in direct marketing; in this way, marketers could target those profitable segments, allowing the firm to maintain valuable customers. Specifically, marketers could make predictions about the
23 behavior of those groups when the segmentation is identified. They use the segmentation results to align marketing strategies so that formulate better-targeted policies. Segmentation is important for marketing strategy, which is widely recognized by marketers (Wedel and Kamakura, 2000). Appropriate segmentation analysis is crucial for developing sustainable marketing strategies (Thorsten Teichert, 2008). Channel-based segments have important implications for marketers since the previous study shows that customers have different characteristics based on their channel usage (Kushwaha and Shankar, 2008a). Marketers could design marketing programs by channel if channel segments are identified. Therefore, the channel-based segments are advisable. One option segment basis is the customer usage of different channels (Neslin and Shankar, 2009).
2.2.2 Customer segmentation approaches
The segmentation encompasses different types of approaches; the choice of approaches depends on the resources and skills in each project approaches (Wedel and Kamakura, 2000; Darnton and Sharp, 2006).Historically, these approaches generally can be divided into two types. The first one is pre-determined (a priori) segmentation based on known characteristics (Chin‐Feng Lin, 2002). The groups are selected from people in advance and formed ‘segments’. The characteristics and variables are determined by marketers or researches. They normally choose based on the past research. As a result, this approach is influenced by the researchers’ conceptual limitations. As table 2 shown, decision tree learning, as the statistical technique, is commonly used in the priori segmentation.
24 The second type is the Market-defined (post hoc) segmentation. In this approach, the segments are identified by multivariate analysis. The main difference from the pre-determined segmentation is that market-defined segments are identified by collecting data rather than researchers; and both segment numbers and cluster size remain unknown until the process is completed(Anable, 2005).Respondents are clustered based on their similarity on multivariate profiles on any number of combinations of variables. When using the multivariate statistical analysis, it may include one characteristics or a combination of two characteristics, such as demographic and behavioral characteristics. The main types of these variables are listed in the table 3, including geographic, demographic, and psychographic variables. In addition to these personal characteristics, perceived benefit and behavioral variables also act as the basis of market segmentation. For example, the segment basis could be the usage purpose or frequency that belongs to behavioral basis. Post-hoc segmentation methods are normally based on latent variables and employ latent class model into segmentation. As table 2 indicates, latent class cluster analysis is one statistical technique for this approach. It could be regarded as a special form of cluster analysis. There are a number of algorithms available for clustering which are broadly classified into two groups: hierarchical and nonhierarchical clustering technique (Blattberg, R. C., Kim, P., and Neslin, S. A., 2008). Latent class cluster analysis is also known as probabilistic clustering that belongs to nonhierarchical clustering technique. Unlike most of the ad-hoc clustering methods which merely present convenient heuristics for deriving segments in a sample, latent class models allow for
25 segment based estimation within the framework of standard statistical theory (Wedel and Kamakura, 2000). The fundamental assumption underlying latent class models is that of local independence which states that objects of the same segment share a common joint probability distribution among the observed variables, enabling thus the simultaneous identification of classes and description of the within-class structure of the observations (Vermunt and Magidson, 2003). In our study, we use latent class analysis as statistical technique. This approach has the advantage since the segments are more likely to link to actual marketplace behaviors and preferences (Allenby et al.,2002) Latent GOLD is one popular commercial software in marketing segmentation research that was developed by Statistical Innovations to implement the latent class clustering (Vermunt and Magidson, 2002). We employ this software to conduct segmentation study.
Table 2: customer segmentation approaches
Two approaches Segmentation method Statistical technique priori
approaches
the segment based on known
characteristics which are determined by marketers and researchers
Decision tree learning
Post-hoc approaches
The segment is identified by using multivariate analysis
The similarity on multivariate profile on any number of combinations of variables
Latent class analysis (non-hierarchy cluster analysis )
Table3: Common segmentation bases
Segmentation base Description of each main consumer
segmentation base
Geographic segmentation Segmenting by country, region, city or other geographic basis.
Demographic segmentation Segmenting based on identifiable population characteristics, such as age, occupation, marital status and so on.
26 Psychographic segmentation This segmentation approach involves an understanding of a consumer’s lifestyle, interests, and opinions.
Benefit segmentation This approach segments consumers on the basis of specific benefits they are seeking from the product, such as convenience, or status, or value, and so on.
Behavioral segmentation Segmenting the market based on their relationship with the product or the firm. Examples include: heavy or light users, brand loyal or brand switchers, and so on.
2.2.3 Customer segmentation in multichannel research
There are numbers of previous literature about mobile shopping and customer segmentation in different channels. Table 4 summarized those researches and showed whether or not it includes demographical, psychological characteristics and expected benefits in their research.
The research from Bhatnagar and Ghose (2004) engaged in segmenting e- shopper based on their sensitivity to the benefits and risks of internet shopping. And the segments are profiled by demographic variables. This paper firstly employs the latent segmentation in the research. Andrew and Vanitha (2004) developed segments based four motivations for online shopping, including online convenience, physical store orientation, and information use in planning and shopping, and variety seeking in the online shopping context. Wetzels, Ruyter and Feinberg (2004) segmented customers based on their preference for retail or online shopping. Those segments are profiled and labeled by demographical characteristics. The paper from Konus et al. (2008) includes multi-shopping channel, such as store, internet and catalog, to segment shoppers based on their attitudes toward different channels in searching and
27 purchasing stages. Both psychological and demographical covariates are examined in this study. The result shows that only psychological variables could predict multichannel segment membership. Bigné-Alcañiz et al. (2008) firstly identified the influence of personality characteristics on mobile shopping acceptance. The paper examined the effect of innovativeness, compatibility and affinity on the behavioral intention to adopt mobile shopping. The finding shows that personality variables have positive effect on the adoption of mobile shopping. Moreover, drivers or motivations of intention to use also play an important role in customer segmentation of mobile shopping. Schröder and Zaharia(2008) investigates the buying behavior of 525 Germany customers of multichannel retailers based on their shopping motivations and also their socio-demographic characteristics, including sex, age, net income. The result shows that most customers only choose single channel during the shopping process; the choice of channels depend on their shopping motivations in each situation. And it is also suggested that the customer segments do not differ in terms of socio-demographic characteristics. The paper from Lu and Su (2009) aims to analyze customers’ perception of mobile shopping website and the expected benefits and barrier are investigated by researchers. The finding indicates that anxiety is the negative indicator of behavioral intention to use mobile shopping website; by contrast, the enjoyment, usefulness and compatibility positively influence the behavioral intentions. The study by Hernández, Jiménez and Martín (2011) aims to analyze the influence of consumers’ demographic characteristics on the online shopping behavior. Those characteristics include age, gender, and income. The result shows that these
28 factors do not influence the use and perception of e-commerce. In the research by Yang and Kim (2012), the comparison between mobile shoppers and non-mobile shoppers based on the motivation is conducted. The result indicates that efficiency, idea, adventure and gratification act as the main factor of mobile shopping usage. San-Martin et.al (2013) furthermore uses latent class cluster to identify three groups of mobile shopping users. Those segments are profiled based on their perceived drivers and impediments; and this is the first study that employs drivers and impediments to segment in the mobile shopping area. The personal characteristics are also examined in this research.
The previous literature provides deep insight into mobile shopping and customer segmentation. But few literatures focus on the segmentation of mobile shopper, especially for mobile shopping applications users. This is probably because the mobile native applications are relatively new shopping channel. Even though there are some studies about mobile shopping, in fact, their topic is more relevant to mobile shopping adoption rather than mobile shopping segmentation. Besides, those researches do not divide mobile shopping channel into specific types, such as the native shopping applications. As a result, our research focuses on mobile shopping applications and segment consumer based on the usage behavior. In our paper, the covariates consist of demographical and psychological characteristics. Besides, the perceived expected benefit is also included since it may influence the usage behavior of mobile shopping applications.
29 Table4: Customer segmentation in multichannel research
2.2.4 Demographic and psychographic characteristics in customer segmentation Many Companies identify segments based on demographic and psychological characteristic, which could help marketers understand their targeted group better and then refine the relevant marketing tactics and strategies. The demographic profiling is an important segment process in which marketers split the market by using personal demographic information, such as gender, age and education level. Such descriptive attributes are also used in the marketing surveys (Okazaki, 2006). Swinyard and Smith (2003) found that online shoppers were normally younger, wealthier, better educated, more computer literate and more likely to spend time on the computer, more likely to find online shopping to be easy and entertaining, and less fearful about
30 financial loss resulting from online transactions. Lohse and Bellman (2003) also found Internet shoppers to be younger, more educated and wealthier and to have a more “wired lifestyle,” but also to be more time-constrained than non-Internet shoppers. A study found that the influence of gender on the usage of online shopping.it shows that the number of women who plan to shop online in the Christmas season of 1999 is substantially larger than the number of men (Rodgers and Harris, 2003). Another NPD research (2000) found that the female segment will soon outnumber male shoppers in every product category that the survey examined. We also found some evidence from mobile context. A survey about mobile banking adoption in South Africa indicated that the majority (67%) of the respondents were ‘young, educated groups, either employed or studying or both’. Like demographic characteristics, some personality traits also acted as the predictors of shopping channel usage. The research from Konus et.al (2008) and Heitz-spahn (2013) suggested that the usage benefits are associated with psychological traits. Even though the strong evidences were mostly found in online shopping context, we assume mobile shoppers may have similar characteristics. The specific explanation of these factors will be presented in the conceptual framework part.
3. Conceptual framework
We start to identify the indicators and covariates of segmentation. Through reviewing previous articles, it could help our study get an idea of the current view about the indicators and profiling variables in customer segmentation; then, we describe variables used in our study and discuss why these variables are selected as the
31 indicator and covariates; finally, we present the conceptual framework and give our expectations.
3.1 The usage behavior as segmentation indicator
Behavioral segmentation is an important base for segmenting consumer market. In our research, the usage behavior of mobile shopping applications is the only indicator for customer segmentation. Using usage behavior as indicator of segmentation could bring some advantage; firstly, past behavior is always regarded as the good predictor of future behavior; besides, the reliability could be improved compared with segmentation based on other indicators, such as attitude and purchase intentions (Vermunt and Magidson, 2004). Since the usage behavior is abstract and broad, we employ usage purposes (information search and purchase) and usage frequency (the intensity of searching or purchasing) to measure usage behavior. These two variables could be regarded as two alternative indicators, which are also commonly applied in the behavioral segmentation study. In addition, we select three common product categories (clothes, flight tickets and electronics) to measure the usage behavior which makes usage behavior more specific.
In the whole shopping process, information search and purchase are two main purposes when people use mobile shopping applications. The information search is a significant aspect in mobile shopping behavior (Seock and Norton, 2007). The useful characteristic of internet exists in the pre-purchase stage because it is easy for customers to use internet for information searching (Urbany, Dickson and Sawyer, 2000). When customers have enough information, they will make purchase decisions.
32 Purchasing is another important purpose for usage of mobile shopping applications. In addition to the usage purpose, usage frequency is another commonly basis applied into behavioral segmentation. Individuals are normally clustered as heavy, medium or light users based on the usage rate. To be more specific, we employ the questions, like ‘how many times of search or purchase for specific product in recent three months’, to measure the usage rate. The reason why we choose recent three months as time period is that usage behavior in recent three months could reflect the current shopping behavior and is also easily recalled by consumers.
In this study, three product categories are selected as one dimension of the segmentation indicator. The reason why this paper includes different categories in the segmentation is that product categories may influence usage behavior in some extent. For example, if someone is the big fan of electronic but not into the clothing, he or she will search and purchase only for electronic products. The product category, in this case, has effect on the customers’ usage behavior. This paper uses clothing, flight tickets and electronic as three product categories since the previous literature shows the difference among these three categories in online shopping context (Bhatnagar, and Ghose, 2004a).
3.2 Covariates of mobile shopping applications usage
Covariates play an important role in the segmentation study; those variables may influence the classification probabilities and thus have indirect effect on the segment membership. Their function is normally profiling identified segment (Vermunt and Magidson, 2002).In our study, we have three types of covariates: demographic
33 variables, psychological variables and expected benefits. To be more specific, this paper includes age, gender and education level as demographic variables to investigate. Innovativeness, impulsiveness and time pressure act as psychological covariates. Besides, as the previous literature review, we also use some drivers of mobile shopping adoption as expected benefits that serve as covariates. People believe that these kinds of drivers could bring the expected benefits when they use mobile shopping channels so that these factors may influence the usage behavior. For example, the previous study from Ko and Lee (2009) indicates that the usage of mobile shopping service is associated with both utilitarian and hedonic benefits in the Korean mobile shopping context. As a result, we choose privacy (security), ease of use and shopping enjoyment as covariates in this study.
3.2.1 Demographical covariates Gender
Many previous researches aim to investigate the relationship between gender and the usage behavior of online shopping. Allen (2001) found that men are more likely than women to make purchase online. However, the literature from Levin (2005) indicates that females are more likely than males to prefer online shopping for clothing. Within the university sample, males were more likely to prefer online shopping for electronic and computer products. Besides, other study indicates the gender difference is associated with the perception of online shopping and thus affects purchase decision (Chiu et al., 2005; Rodgers and Harris, 2003; Slyke et al., 2002). In light of this, I
34
the usage of mobile shopping applications.
Age and education level
As for the influence of age, the study from Levin (2003) suggests that younger students are more likely to prefer online search for electronic products. In the mobile shopping context, the research from Yang and Kim (2012) indicates that the people between 21-30 years old are main mobile shoppers. Besides, the education level is also examined in this study. They found that most of mobile shoppers have college degree (Yang and Kim, 2012). The relationship between education level and online shopping behavior is widely proved by many researches. As a result, I expect that
both age and the education level are significantly associated with the segment membership based on the usage of mobile shopping applications.
3.2.2 Psychographic covariates Innovativeness
Many previous literatures focus on the relationship between innovativeness and usage pattern of online shopping. The study shows that the innovative people are heavy users of interactive electronic shopping media, which reflects the usage rate is related to the personality traits (Eastlick and Lotz ,1999). The study by Goldsmith (2000) also supported that the innovativeness is the predictor of the usage frequency of online shopping and future shopping intention. Citrin et al. (2000) believe that innovativeness with internet usage have directly effect on adoption behavior of online shopping. The association between innovativeness and mobile shopping behavior is also examined by previous research. The study indicates that innovativeness has
35 positive effect on mobile shopping intention (Aldás et.al, 2009). As a result, I expect
that innovativeness has significantly influence on the segment membership based on the usage of mobile shopping applications.
Time pressure
Time pressure means consumers incline to the view that time is scarce and valuable and plan its use carefully (Kleijnen, 2007). Verhoef and Langerak (2001) demonstrate a positive relationship between time pressure and the relative advantage of an online channel. Pechtl (2003) suggested that the time-stressed consumer could be regarded as the potential online shopper since they could receive benefit from convenient internet shopping. The finding from Xu-priour, Cliquet and Fu (2012) supported that high time-pressured consumers prefer online channel for shopping due to the convenience. As for the mobile shopping channel, it could provide time-pressured consumers with an opportunity to go shopping without time and place limitations (Yang and Kim, 2012).Therefore, I expect that time pressure is significantly associated with the
segment membership based on the usage of mobile shopping applications.
Impulsiveness
Impulsiveness is the personality trait which reflects an urge to act spontaneously without thinking or planning ahead for the consequences of your actions (Harris, 2014). In the mobile shopping context, if a person with impulsiveness trait, he or she is more likely use mobile shopping applications to purchase product without any hesitation only because this product is attractive for them. Trait impulsiveness has impact on psychological processes and behaviors, including decision-making
36 (Bechara, 2007). Consequently, I expect that impulsiveness has significant influence
on the segment membership based on the usage of mobile shopping applications.
3.2.3 Expected benefits as covariates Privacy (security)
Privacy means the security of using mobile shopping channel. The improved security is identified as a critical successful factor in mobile commerce (Vrechopoulos, 2003).The previous study found that the privacy or security is a strong predictor of online shopping satisfaction (Liu et.al , 2008). The study from Venkatesh, Hofacker and Naik (2010) suggested that mobile users regard privacy and security of information as the most important factors when they use mobile commerce service; thus, I expect that privacy(security) is significantly associated with the segment
membership based on the usage of mobile shopping applications.
Ease of use
The ease of use is derived from the traditional technology acceptance model (TAM model). It refers to the required effort to use the system, and then to the people’s perception that using a particular technology will be easier (Davis, 1989).However, the previous literatures do not have consistent results on the impact of the perceived ease of use on usage behavior of online shopping (Kulviwat et al., 2007; Nysveen, Pedersen, and Thorbjørnsen, 2005).However, in the mobile shopping context, the study from Kleijnen, Ruyter, and Wetzels , 2007) found that cognitive effort of usage has negative effect on the perceived value of the usage of mobile shopping channel, which means that ease of use could increase the perceived value of mobile shopping
37 channel. As a result, I could expect that ease of use has significant effect on the
segment membership based on the usage of mobile shopping applications.
Shopping enjoyment
The shopping enjoyment reflects affective and instinct benefits of using the technology (Kim, Chan and Gupta, 2007). Several previous literatures supported that the enjoyment is the important predictor for using internet and mobile technology(Sánchez-Franco and Roldán, 2005; Nysveen et al., 2005). Lu (2009) then used a revised TAM model to explore the factors which affect purchase intention on mobile shopping websites. The result shows that the shopping enjoyment has a positive impact on customers’ behavioral intentions. Other relevant researches also have consistent results (Yang and Kim, 2012; Li et al., 2012). Therefore, I expect that
shopping enjoyment significantly influence the segment membership based on the usage of mobile shopping applications.
3.3 Conceptual framework and expectations
The figure 1 is the conceptual framework of our study. The objective of research is to identify distinct segments based on the usage of mobile shopping applications. We assume people use mobile shopping applications to search and-or purchase for three product categories. Therefore, the segmentation indicator is usage behavior that is measured by two alternative indicators: usage purpose and usage frequency. We include the expected benefits, demographic and psychological characteristics as covariates, which aims to profile identified segments. The latent class cluster analysis will be employed in our segmentation study; so, the number of segments cannot be
38 determined before analyzing data. Before analyzing data, we could expect that
1. We could identify distinct segments based on mobile shopping applications usage in three different types of product categories.
2. There are identified distinct segments based on the usage purpose of mobile shopping applications
3. There are identified distinct segments based on the usage frequency of mobile shopping applications
Besides, other expectations about covariates are also included in our study, which are already specifically described in the previous part (3.2).
39 4. Methodology
4.1 Research design
We use the non-probability sampling technique to select the respondents. The data will be collected through an online survey, which will be created on www.qualtrics.com. The survey is distributed within the personal network of the researchers through the social media website and by e-mailing acquaintances. A questionnaire with closed-end questions is used for this study.
We use the filter question at the beginning of the questionnaire. It aims to select respondents who have online shopping experience and fulfil the condition to participate the survey. The survey has three different parts. In the first part, the respondents are asked whether or not they use mobile shopping applications for searching information and for purchasing separately. Each question is repeated for three types of product categories; in this part, all the questions are yes or no questions. If the answer is yes, the further question is about the usage frequency. The second part aims to investigate the psychological factors about the respondents. There are three types of traits included in our research. For each trait, several questions represent this type of trait. In the third part, several questions which stand for three kind of expected benefits are asked. For the second and third part, we use five point-Likert scales to measure the agreement level for each question. Items will be rated on a scale ranging from 1 (strongly disagree) to 5 (strongly agree). Finally, respondents are asked about their basic demographic information, including age, gender and education level. Since the questionnaires are developed by adopting the measurements from research written
40 in English, this research go through back translation processes following the recommendation by Brislin (1970).
4.2 Pilot study
Before administering the questionnaire in the main study, a pilot study was firstly conducted. Ten Chinese students from the University of Amsterdam participated in this pilot study. The reason why we chose these ten students is that all of them are Chinese users of mobile shopping applications and also could provide some meaningful suggestions which are helpful to the further improvement of questionnaires. They were required to complete the online survey separately in order to make sure they do not watch each other. In the pilot study, respondents need to answer all the survey. After finishing questionnaire, they were asked to provide their suggestions about the survey. Pilot study could not only make sure the appropriateness of questions to the target population but also identify the problem which may lead to biased answers.
4.3 Main study
Data collection was finished in seven days. The 350 questionnaires were distributed online; but there were 48 participates who did not fill any questions. These samples were invalid so that excluded from research. We did not find any missing value in other 302 questionnaires which could be considered as valid samples. After cleansing data, the filter question was used to choose samples included in our research. We assume that those people who never use internet for shopping are no likely to use mobile application for shopping. So, this kind of people is also excluded in our
41 research.in our sample collection, 27 respondents are not online shopping users and thus deleted from data set. Therefore, there are 275 valid samples in our research. Lastly, In order to facilitate the further analysis, the demographic question – age- is recoded into two groups: less than or equal to 30 years old and more than 30 years old.
4.4 Latent-class cluster model
This study we employ latent class models (also called finite mixture models) in our research. The models include one or more discrete unobserved variables that refer to latent variables. In the marketing research, this kind of models normally interprets the categories of these latent variables (classes) as clusters or segments (Dillon et al., 1994; Wedel and Kamakura, 1998).Latent class analysis could be considered as a new powerful tool to identify important marketing segments.
The strength of Latent class cluster analysis is obvious. Firstly, the previous study suggested that the use of latent class model could yield powerful improvements to cluster compared with traditional approaches (Vermunt and Magidson, 2000a, 2000b).Besides, this approach could avoid the biases related to the data not conforming to model assumptions. More importantly, this approach could employ mix scale types in the same analysis(Vermunt and Magidson, 2000a, 2000b).In addition, the description of segment in this approach is also improved; the relationship between the latent class and covariates(external) variable could be examine simultaneously when the cluster is identified. But in the traditional cluster analysis, we can only conduct discriminant analysis after identifying segments.
42 In the latent-class cluster analysis, the latent variable (customer segments) is always regarded as categorical variable which have K possible values and thus represent K segments. Latent classes are unobservable segments. Cases within the same latent class are homogeneous on certain criteria, while cases in different latent classes are dissimilar from each other in certain important ways. And the segmentation structure will be influenced by active covariates. To be more specific, in our study, the model this could be written as follows:
f (URic|𝑍𝑖) =∑𝐾 [∏𝑐=13 g (𝑈𝑅𝑖𝑐|𝑍𝑖, 𝑆𝑖) ]
𝑋=1 p (𝑆𝑖 = 𝑥|𝑍𝑖)
URic Respondents (i) Usage purpose (U) and frequency rate (R) of mobile shopping applications usage across different product categories (c) in the decision process, in which c= clothing, electronics or flight tickets
Si indicator of respondent (i) Usage purpose (U) and Frequency rate (R) segment
Zi Descriptive variables: expected benefits, psychological and demographical traits that represent the covariates for respondents (i)
f (URic|𝑍𝑖) Probability distribution for respondents (i) Usage purpose(U) and Frequency rate(R) of mobile shopping applications usage across product categories given the respondents’ set of descriptive variables.
g (𝑈𝑅𝑖𝑐|𝑍𝑖, 𝑆𝑖) Probability distribution for respondents’ usage purpose and frequency rate of usage behavior towards mobile shopping applications across product categories given the respondents set of indicator and descriptive variables and given that the respondents in segment Si
P (𝑆𝑖 = 𝑥|𝑍𝑖) Probability that respondent (i) is in segment x given the respondents
descriptive variables.
According to the model, two alternative indicators are used in the segmentation process: usage purpose (information search and purchase) and usage frequency rate (how many times using mobile shopping applications for searching or purchasing in
43 recent three months). In addition, three types of covariates are also included in order to profile the class membership. Combing mixed indicators with covariates, the latent class cluster model in our study is formulated (Vermunt and Magidson, 2002). In order to clarity the indicators and covariates used in our study, all variables included in the study are summarized in the table 5.
Table 5: Indicators and covariates in the segmentation study
Indicators Usage purpose Information search Product categories: Clothes Flight tickets electronics Purchase Usage frequency
Usage rate in searching information
Usage rate in purchasing Covariates Demographic
covariates
Age ; Gender ; Education level Psychological
covariates
Innovativeness; Time pressure; Impulsiveness Expected
Benefits
Privacy(security); Ease of use; Shopping enjoyment
5． Analysis and results
In the analysis part, we conduct the descriptive analysis and reliability analysis firstly by using SPSS. Then, the main segmentation result is analyzed by Latent Gold software in order to test our expectations.
5.1 Descriptive analysis
Descriptive analysis is important since it helps us to present the meaningful data. It is especially critical for our research because some demographic variables are included in our analysis part, such as age and gender. It means that we have to try to balance the distribution of the age and gender. Descriptive analysis makes data visual and facilitates us to check the distribution of samples in terms of their age and gender. The
44 result shows that the gender distribution is balanced since the male and female accounted for 48 percent and 52 percent respectively. As for the age distribution, the respondents who are less than or equal to 30 years old occupy 64 percent of total. It also could be acceptable. The third demographic variable is education level. The bachelor degree had the highest percentage (65%) compared with other four education levels. Besides, we also calculate the non-mobile shopping users who never use the mobile shopping applications for both searching and purchasing process in different three categories; in other words, their all answers about usage purpose questions are ‘no’. The non-users group accounts for only 14 percent of total respondents.
5.2 Reliability analysis
Reliability is used to measure the internal consistency, indicating how well items are correlated with one another. The Cronbach’s Alphas ranges from 0.00 to 1.00.The higher the value, the more reliable the test score. Nunnaly (1978) indicated that 0.7 is an acceptable reliability coefficient. The table6 shows that all variables’ Alphas are more than 0.7, which means items are measuring the same construct.
Table 6 Reliability Analysis
Items Cronbach's Alpha N of Items
Privacy 0.743 3 Ease of Use 0.876 2 Shopping Enjoyment 0.773 3 Innovativeness 0.716 3 Time pressure 0.751 2 Impulsiveness 0.781 3 5.3 Segmentation analysis
45 segment analysis: usage purpose and usage frequency; thus, we need to conduct two segmentation analyses separately. In the first latent class cluster analysis (LCCA) model, two searching and purchasing purpose for different three categories consist of six variables which are used as segmentation indicator variables. Other psychographic, demographic variables and expected benefits play the role of covariates. Each segment will be profiled in terms of covariates. The second LCCA model has a different indicator whereas covariates remain same since their function is also used to profile each segment. The indicator of this model is the usage frequency for two purposes in different three categories. Distinct six indicator variables will be analyzed to form different segments and then profiled by same covariates. After identifying segments, we will examine following questions in the result part. Firstly, we will check whether or not there are different segments based on two models. Secondly, the main segments based on each model are identified. Their profiling information will give the insight into segment characteristics. In this way, we could see the difference among clusters for each model. Next, we compare these two models to find the similarity among those segments. Finally, we explore those similar main segments from two models and combine them.
5.3.1 Optimal model selection
One fundamental principle of latent cluster class model is that the number of segments is a priori unknown. This means that the analysis begins with optimal model chosen. The procedure of model selection is followed by Wedel and Kamakura (1998) recommendation. We started to estimate process from a single segment model to eight
