Segmenting customers on the basis of their mobile shopping app orientation, behavior and other drivers of mobile shopping

(1)

Master Thesis

‘Segmenting Customers on the basis of their mobile shopping app orientation,

behavior and other drivers of mobile shopping’

University of Amsterdam

Faculty of Economics and Business

MSc. in Business Administration – Digital Business Track

Supervisor: Dr. Umut Konuş

Student: Şebnem Alpaylı

Student Number: 11642629

(2)

Master Thesis Şebnem Alpaylı (2018)

University of Amsterdam 1

Statement of Originality

This document is written by Şebnem Alpaylı who declares to take full responsibility for the contents of the document.

I declare that the text and the work presented in this document is 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

Mobile shopping has been highly adopted during the last decade. Today it is a major shopping channel, so that most of the global e-commerce transactions were made via the mobile devices in 2017. Meanwhile mobile apps have emerged as the biggest mobile channel with the highest volume of mobile transactions. In such a mobile landscape, a comprehensive approach to mobile app shopping behavior and its drivers has become more valuable than before.

Current academic literature focuses on mobile shopping behavior, the effects of psychographic and demographic variables, and attitudes towards mobile shopping. However, no study solely focuses on mobile app shoppers and offers distinct mobile app shopper segments based on the shopping behavior and attitudes. Also, none of the previous studies provide background

information about mobile app shoppers, which can generate important insights for marketing professionals. Therefore, this study aimed to (1) segment customers based on their mobile shopping app behavior (during search and purchases phases) and their attitudes, (2) profile these segments based on their demographic, psychographic, and device related attributes, and (3) explore whether mobile shopping app behavior differs across three different product categories (fashion, consumer electronics, and flight tickets). To achieve these goals, this study segmented customers using Latent Class Cluster Analysis (LCCA), and the survey data from Turkish and Dutch respondents.

As a result, the study spotted six different segments which are as follows :(1)‘Mobile Rookies’, (2)‘Non-Mobile Shoppers’, (3)‘Multiscreen Enthusiasts’, (4)‘Mobile App Explorers’, (5)‘Mobile App Adopters’ and (6)‘Mobile App Window Shoppers’. These findings enhance the previous literature, and they provide business managers with a base for mobile customer segmentation practices.

(4)

Having this base, managers can develop customized marketing strategies, and select the most appropriate marketing communications techniques for each of these segments. Further, managers can execute their marketing strategies through the most relevant targeting and retargeting tactics, value propositions and content. Therefore, they might higher their customer loyalty levels, and profits (Wu, 2011).

Keywords: mobile shopping, mobile shopping apps, mobile behavioral segmentation, drivers of mobile app shopping, mobile shopping journey.

(5)

Master Thesis Şebnem Alpaylı (2018) University of Amsterdam 4

Master Thesis Şebnem Alpaylı (2018) University of Amsterdam 5 4.3.1.1.1 Convenience 27 4.3.1.1.2 Price Saving 27 4.3.1.1.3 Entertainment 28 4.3.2 Passive Covariates 28 4.3.2.1 Psychographic Covariates 28 4.3.2.1.1 Innovativeness 28 4.3.2.1.2 Price Consciousness 29 4.3.2.1.3 Time Pressure 29 4.3.2.1.4 Motivation to Conform 29 4.3.2.2 Demographic Covariates 30 4.3.2.2.1 Age 30 4.3.2.2.2 Gender 30 4.3.2.2.3 Education 30 4.3.2.2.4 Relationship Status 31

4.3.2.3 Device Related Covariates 31

5. Research Plan 31 5.1 Sample 32 5.2 Research Design 33 5.2.1 The Survey 34 5.3 Procedure 34 5.3.1 Survey Pilot 35 5.3.2 Main Study 35

(7)

5.4 Latent Class Cluster Analysis (LCCA) 36

5.4.1 Segmentation Indicators 37

5.4.2 Covariates 38

5.4.2.1 Active Covariates 38

5.4.2.2 Passive Covariates 39

6. Analysis and Results 41

6.1 Preliminary Analysis 41

6.1.1 Reliability 42

6.2 Results of Segmentation 43

6.2.1 Mobile shopping app segments based on actual use (behavior) 43

6.2.1.1 Cluster Analysis for Segmentation Indicators 44

6.2.1.2 Cluster Analysis for Active Covariates 49

7. Discussion 51

7.1 Conclusions 52

7.2 Theoretical Implications 54

7.3 Managerial Implications 54

7.4 Limitations and Further Research 57

8. References 58

List of Tables

Table 1. Review of Mobile Shopper Empirical Literature 19

(8)

Table 3. Sample Characteristics 33

Table 4. Model Selection – Log Likelihood Statistics 43

Table 5. Description of the Segmentation Indicators 48

Table 6. Description of the Segments 50

List of Figures

Figure 1. Mobile as Share of Total E-Commerce 8

Figure 2. Share of Online Activities Starting with a Smartphone 13

Figure 3. Share of Mobile Apps within the Total E-Commerce in the United States 14

Figure 4. Conversion Rate Comparison Between Mobile App, Web and Desktop in the United States 14

Figure 5. Conceptual Model 24

(9)

1. INTRODUCTION

Mobile shopping, as it is called a phenomenon of our time (Pousttchi, 2015), has been growing continuously in terms of volume and the number of users in the last years. In 2017, smartphones had more share of the global internet traffic compared to desktop and laptop devices. In the same period most of the global

e-commerce transactions were made via mobile devices (We are Social, 2018; Statista, 2018). Further, looking at the industry estimates, the growth of mobile shopping does not seem to come to an end in the following years (Statista, 2018).

As mobile shopping has become the biggest branch of e-commerce in 2017, the use of mobile shopping apps has been on rise as well (App Annie 2017 Retrospective, 2018). In 2017, the share of mobile shopping app transactions has reached to 44%, surpassing the shares of mobile web and desktop which account for 23% and 33% of the e-commerce sales respectively. During the same period, mobile app shopping has grown in the Europe, Africa and Middle East regions as well (Criteo, 2018). In such a mobile landscape, the value of mobile app shopping has become more important for online and traditional retailers.

Figure 1: Mobile as share of total e-commerce

(10)

Regarding the growth of mobile shopping apps, the industry statistics indicate the better user experience design, faster checkout processes due to the stored settings, and higher speed is the main reasons that users prefer shopping via mobile apps (Criteo, 2018).

On the other hand, mobile shopping apps have been desirable channels also for the companies, since the apps enable advanced marketing techniques such as location-based targeting,

promotions, and push notifications. Also, mobile shopping apps might contribute to companies’ brand engagement levels, since they are located on the main screens of customer smartphones, as a constant reminder of the brand.

Despite all the mentioned merits of mobile app shopping, currently there is no research investigating different types of mobile app shoppers and their behaviors. Previous research indicates that different customer segments based on the channel use require different marketing approaches (Neslin, 2008). Therefore, companies might need different marketing strategies to reach these different segments (Neslin, 2008). At this stage, a comprehensive study focusing on mobile app shopper segmentation based on the behavior and its drivers would highly benefit the industry. Such a study can enable companies to develop relevant marketing strategies and marketing mix propositions for each of the different customer groups and leverage their full business potential.

Current literature about mobile app shopping behavior focuses on the effects of psychographic and demographic variables, and attitudes towards mobile shopping. However, there is no study which solely focuses on mobile app shoppers and offers a segmentation approach based on the actual behavior and its drivers. Therefore, the mobile app focus will be the main differentiator of

(11)

this study. Also, unlike the previous research, this study will profile mobile app shopper segments based on their demographic, psychographic and device-related attributes, which might indicate important insights for marketing professionals.

In a growing mobile shopping landscape, the findings of this study are expected to have important business implications.

Firstly, segmenting mobile app shoppers into different groups enables managers to serve customers regarding their specific needs. Therefore, companies can avoid applying the same marketing strategy to all of their customers and can create relative strategies to reach each of the segments. Doing these, companies can gain competitive advantage against their competitors. When it comes to executing these marketing strategies, such a segmentation approach can help managers target and retarget customers at right times and places. Also, managers can serve customers with well-targeted value propositions and allocate their digital marketing budgets more efficiently. Therefore, companies might higher their customer loyalty levels, and profits (Wu, 2011). In addition, understanding motivations behind the shopping behavior of each customer group would allow companies to create best possible mobile app experience in terms of the shopping journey design and content.

In order to reach these mentioned research objectives, this study was organized in in five sections. First, the current literature regarding mobile shopping and its drivers, and mobile shopping

segmentation will be reviewed. Secondly, the research model and its components will be explained within the conceptual framework. Thirdly, the research method and sample will be

(12)

reviewed and then the results of this study will be analysed. In the last section, the main conclusions of the study, including the managerial and academic implications will be explained. Also, the main research limitations and future research areas will be explained.

2. LITERATURE REVIEW

2.1 Online and Mobile Shopping: Mobile Shopping and Its Emerge

Regarding the roots of mobile shopping, Strom et al. (2014) suggests that shopping through mobile channels is an extension of previously adapted online shopping (non-mobile) behaviors. This behavior might be explained through the younger history of the mobile devices, since most of the users have been introduced to e-commerce via desktop or laptop computers. Therefore, the users might have adopted their mobile shopping behaviors from the initial non-mobile shopping experiences (Strom, 2014). Further, Kim et al. (2017)’s study shows that mobile purchase decisions are positively affected by the user’s digital experiences (via online and mobile channels).

Being in in line with these studies, mobile shopping has been adopted as a major shopping

channel, and the volume of mobile shopping has increased significantly during the last decade (We are Social, 2017).

Accordingly, today the mobile’s share has reached to 60% of the global e-commerce (Figure 1) (Statista, 2018). Therefore, it can be said that mobile shopping has become a digital phenomenon of our time (Pousttchi, 2015).

Further, the growth of mobile shopping does not seem like slowing down in the following years. As one of the leading markets, the United States is expected to have mobile shopping growth of 18%

(13)

(by a compound annual rate) for the next five years (Forrester, 2018). Also, the forecasts indicate that the smartphones will be used for more than one-third of the U.S. retail sales (more than 1 trillion USD) along the customer journeys (research, price comparison, and purchases) in 2018 (Forrester, 2018).

As the mobile shopping expands by volume, mobile devices get more involved into different phases of an online shopping journeys (Forrester, 2018). Therefore, understanding the different phases of mobile shopping has become more important than before.

Based on the literature, mobile shopping (m-shopping) has been defined in several ways. Although the term ‘shopping’ might give more emphasis to the purchase stage of mobile shopping, it is a mobile business process including pre-purchase, purchase and post-purchase stages (Tiwari, 2007). Mobile shopping has been classified as a type of online shopping which involves monetary transactions and non-monetary transactions, such as post-purchase and search stages (Tiwari, 2007). Further, the customers can search, compare and purchase products; create shopping lists; and manage post-purchase actions along the mobile shopping process (Shankar, 2010). In this regard, mobile devices have been used increasingly for all of the mentioned phases as mobile shopping expands over the last decade (Statista,2018). Especially, mobile devices are used mostly during search phase of mobile customer journeys (Holmes, 2013).

(14)

On the other hand, Google’s study indicates that 65% of the online

journeys (mobile and non-mobile) start with a smartphone (Google, 2012). The findings of Holmes et al. and Google (2012) can point that mobile devices are mostly used for searches rather than purchases. However, the fact that the mobile is first step of most of the

customer journeys might indicate the further importance of that channel.

Accordingly, today the mobile stands for an expanding sales channel, as well as a common first step online multichannel customer journeys (Google, 2012) (Statista, 2018).

2.2 Mobile Shopping Apps

After discussing the rise of mobile shopping, now mobile apps will be reviewed as a component of mobile shopping and customer shopping journeys. Mobile internet and shopping have initially emerged via mobile web browsers, due to the younger history of mobile apps. However, mobile apps have been highly adopted by users during the last decade (Criteo, 2018). Even further, the users have spent most of their online time on mobile devices in 2017 (ComScore, 2017).

In such a growing mobile landscape, mobile shopping apps were no exception. In 2017, mobile shopping apps accounted for 44% of the total e-commerce sales in the United States, having more share than mobile web and desktops which accounted for 23% and 33% of the e-commerce sales

Figure 2: Share of Online Activities Starting with a Smartphone

(15)

respectively (Business Insider, 2018). Further, in 2017 mobile shopping apps have followed a growth pattern in Europe, the Middle

East and Africa as well (Criteo, 2018). On the other hand, in 2017 mobile apps dominated the time spent on mobile devices by 87%, while mobile web accounted for 13% of the time in the U.S (ComScore, 2018).

Therefore, today the apps stand for the strongest channel for mobile shopping.

Following this growth (Criteo, 2018), the academia and managers have highlighted various mobile app shopping benefits which make it a desirable channel for mobile shoppers and businesses. From the user perspective, convenience, speed and stored user settings have been suggested as the top three drivers making customers prefer mobile apps for shopping (JMango, 2017).

Accordingly, mobile apps are reported to offer a better user

experience and higher speed (load and response time) compared to mobile websites. Therefore, they can serve customers a smoother check out process, and as a result shopping cart abandonment is dramatically lower for shopping apps than it is for mobile web and desktop (JMango, 2017). Also, the shopping apps have reported to have three times higher conversion rates compared to mobile web

Figure 4: Conversion rate comparison

between mobile app, web and desktop

Figure 3: Share of Mobile Apps within the Total E-Commerce in

the United States

Source: Business Insider & ComScore, 2018.

(16)

(Criteo, 2018). On the other hand, stored user settings and payment account shorten the checkout process and can lead to a more convenient online experience.

These mentioned merits of mobile shopping apps also make them a desired sales channel for the companies. Accordingly, the companies might enjoy higher engagement (session time)

(ComScore,2018), and conversion rates (Criteo, 2018) if they can pull more customers into their shopping apps. Therefore, today many companies try different ways to acquire new mobile customers. In this regard, the research shows that digital experience via non-shopping apps has a positive relation with shopping app ownership and purchase decisions (Kim, 2017). Hence, the companies can acquire new users via non-shopping apps as a first step and then encourage the new users to adopt their shopping apps.

In addition, mobile apps enable business managers to track the user behavior and offer them more relevant values. Considering the increased time spent within mobile apps (ComScore, 2018), today the companies have potential to generate more in-app behavior data. Doing so, managers can gain further insights about mobile customer behavior, attitudes, and attributes. Therefore, they can segment customers in a more sophisticated way, and enjoy higher sales and profits (Wu,2011).

2.3 Behavioral Customer Segmentation on Online-Mobile Shopping

The academia has been well-aware of the need for mobile shopper segmentation during the last decade, as mobile shopping became a digital phenomenon of our time (Pousttchi, 2015).

(17)

Currently, most of the mobile shopping segmentation literature has focused on the drivers of adoption, and the effects of demographic and psychographic variables on mobile shopping

behavior. These studies focus on mobile shopping with different perspectives while some of them has a specific focus on mobile app shopping.

Regarding mobile app shopping behavior, Kim et al. (2017)’s study has pointed out a positive relation between non-shopping app possession with shopping app ownership and use. Further, the study suggests that digital experience (mobile and other online experiences) and

non-shopping app browsing activities have a positive relationship with mobile app purchase decisions. Therefore, these findings indicate the importance of non-shopping app adoption as preliminary step of adopting mobile app shopping (Kim, 2017).

Further, Natarajan et al. (2017)’s study suggests that personal innovativeness has an important role in mobile shopping app adoption. Also, the users with higher levels of innovativeness have been observed to be less price sensitive (Natarajan,2017). Therefore, this article suggests price sensitivity can be important segmentation criteria in order to develop different pricing strategies for different customer groups. Applying different pricing strategies, the companies can achieve higher sales as well as higher profits (Natarajan, 2017). On the other hand, Natarajan et al. (2017) has found out that perceived risk has an impact on mobile shopping app behavior, and lower perceived risk towards mobile shopping has a positive effect on the shopping satisfaction.

Therefore, perceived risk level can be an important factor to segment the customers and to design mobile shopping app interfaces (Natarajan, 2017).

(18)

Regarding the impacts of mobile shopping adoption on the shopping behavior, Wang et al. (2015) points the order frequency and size increase as customers adopt mobile shopping. The study has noted that this effect was more visible for the low-spending customers.

Further this study indicates that, the customers engage in mobile shopping mostly for purchasing habitual products with short life span (low involvement products) and from the companies with whom they have a purchase history. Therefore, this behavior indicates the customers try to

minimize potential post-purchase regret due to the mobile’s screen size constraint while searching the most relevant products (Wang, 2015). Considering the findings of this study, the shopping volumes and product selection patterns might indicate an effective approach to segment the mobile shoppers.

Adding more on the product involvement level, Holmes et al. (2013) suggests that customers use the mobile channel mostly for product search and reviewing the alternatives before they purchase high involvement products (products with higher capital value, which are generally purchased after careful consideration).

On the other hand, Konus et al. (2008) segmented customers based on their search and purchase behavior, and attitudes towards different shopping channels in a multichannel shopping

environment. Further, this study has examined the relationship among psychological, economic, and sociodemographic covariates and segment membership across different product categories, namely fashion, consumer electronics, and flight tickets.

In the end, this study has identified three customer segments, which are multichannel enthusiasts, uninvolved shoppers and store focused consumers (Konus, 2008). Although Konus et al. has not

(19)

focused on the mobile as a channel, his findings point the presence of multichannel customer segments based on the behavior and attitudes. Considering mobile app shopping has expanded significantly (Criteo, 2018) since Konus et al.’s study was released in 2008, meanwhile a research on mobile app shopper segmentation with a focus on the customer attitudes and different product categories has emerged as a research gap.

In the end of this review, it was noticed that most of the research on mobile shopping behavior focus around the drivers of adoption, demographic and psychographic variables. Although there is past literature about online shopping behavior-based segmentation, either these studies do not have the mobile channel within their scope (Konus, 2008), or they do not specifically focus on mobile shopping apps. Therefore, this study aims to expand the knowledge in this field through combining new covariates which were not combined in the previous studies with a mobile shopping app focus. In addition, this paper includes different phases of a shopping journey, such as search and purchase, which have been rarely used before.

Table 1 provides an overview of the previous research done about mobile shoppers and the variables used in each study.

(20)

2.4 An Integrated Segmentation on Mobile App Shopping Behavior and Its Drivers

As mentioned, this study aims to expand the current academic literature on mobile shopping behavior with a focus on mobile shopping apps. In this regard, the presence of customer segments based on mobile app shopping behavior and attitudes will be investigated. On the other hand, the study will employ demographic, psychographic and device related variables as passive covariates. Meaning that, these variables will not play role in segmentation, however, they will be used to provide background information about the clusters. Furthermore, this study will be focusing on three different product categories which are fashion, consumer electronics and flight tickets, as they have the highest three e-commerce sales volume globally.

Current academic research has various behavior-based segmentation studies within the online and/or mobile shopping landscape. However, there is no previous research incorporating these

(21)

variables and product categories with a mobile shopping app focus. Therefore, the findings of previous online (non-app) segmentation studies might differ from the dynamics of mobile shopping app behavior. Mobile app shopping might have different customer behavior dynamics caused by the user experience they provide to users (i.e. speed, and saved settings), the

conversion rates and the frequency of use (Criteo, 2018).

Therefore, this study aims to fill this academic gap via combining relevant variables with a mobile app shopping focus.

2.5 Drivers of (Factors Behind) Mobile Shopping

The research framework of this study employs different demographic, psychographic and

attitudinal variables in a mobile app shopping context. In this section there will be an overview of the past literature focusing on such variables with a mobile shopping perspective.

2.5.1 Personality Traits and Demographic Variables

Kang et. al (2013) has studied the relation among the mobile shoppers’ personality traits and their willingness to purchase on mobile devices. This study indicated extroversion and openness to experiences are positively related to engaging in mobile shopping (Kang et. al, 2013). Further, Gupta et al. (2017)’s study has indicated the positive relation between the personal trait, openness to change, and mobile shopping adoption.

On the other hand, affinity to mobile phones, compatibility and innovativeness have found to have a positive effect on the mobile shopping intention (Aldas-Manzano, 2009). The study also suggests that perceived usefulness does not have a direct effect on attitudes towards mobile shopping (Aldas-Manzano, 2009). Furthermore, Feng et al. (2017)’s study suggests that time pressure

(22)

positively affects shopping intention, while search and transaction convenience mediates the effect of time pressure on mobile shopping.

Regarding the demographic variables, Okazaki and Mendez et al. (2013) have investigated the moderating effect of gender in mobile commerce and the importance of convenience. In addition, Chong et al. (2013)’s study has showed that age and educational background are related to mobile shopping adoption.

2.5.2 Expected Benefits and Costs

Agrebi et al. (2015) has identified two profiles of individuals which are the purchasers and non-purchasers and analyzed the factors affecting mobile purchase intention. The study has showed that, mobile shoppers can be encouraged to purchase more through enhanced perception of usefulness and enjoyment. For the non-purchasers, perceived usefulness had an effect on the intention of purchase.

Furthermore, Kim et. al (2012) has suggested that service quality has a positive effect on utilitarian and hedonic mobile shopping values. Also, Wu and Wang et al. (2005) has pointed out that the level of perceived risk impacts mobile shopping adoption. Therefore, trust is spotted as an important factor to enhance the adoption (Wu and Wang et al., 2005).

More on the drivers of mobile shopping adoption, previous literature points out that emotions and enjoyment as significant drivers of this adoption (Li et al., 2012; Yang and Kim, 2012). Also, Kleijnen et. al (2007) have concluded the time convenience, user control benefits, and as well as the perceived risk and cognitive effort costs have significant impact on mobile channel value

(23)

perceptions. Further, San-Martin et al. (2013) has indicated the effects of motivational traits on mobile shopping adoption.

In addition, Yang et. al (2012) suggests that efficiency, adventure and gratification are significant drivers of mobile shopping and therefore, they can be employed within the design of mobile shopping channels.

In summary, previous research on the drivers of mobile shopping and its adoption have different perspectives and focuses.

After reviewing the most relevant past literature regarding mobile shopping behavior and mobile shopper segmentation, a research framework has been built regarding the current literature and a research gap within this field.

In the next two chapters the research gap, research questions and accordingly, the research framework will be explained in detail.

3. RESEARCH GAP AND RESEARCH QUESTION

Current studies about mobile shopper segmentation have not combined mobile app shopping behavior and the attitudes towards mobile shopping, combined with psychographic, demographic and product-related covariates across different product categories.

Therefore, this study aims to answer the following questions:

Research Question:

Can customers be segmented based on their mobile shopping app behavior, and attitudes towards mobile shopping?

(24)

Research Sub questions:

1) Are there customer segments which can be identified based on their mobile shopping app behavior and attitudes (expected benefits) towards mobile shopping?

2) If so, can these segments be profiled by their demographic, psychographic and device related attributes?

4. CONCEPTUAL FRAMEWORK

The conceptual framework, as visualized below, summarizes the scope of this study. This framework includes actual mobile shopping app behavior during search and purchase, as the segmentation indicator. Also, this framework employs the expected benefits from mobile shopping, as the active covariates, while it uses demographic, psychographic and device-related variables as the passive covariates. The passive covariates will not be deterministic while

segmenting customers based on their similar characteristics, however, they will be used to profile demographic, psychographic and device-related attributes of the (possibly) observed clusters. Due to the nature of this segmentation study, no hypothesis is included within this framework.

(25)

4.1 Framework

This research aims to uncover the mobile app shopper segments as a product the segmentation indicators and active covariates. Accordingly, this research framework approaches mobile app shopper segments as an output of the actual shopping behavior and attitudes towards mobile shopping (expected benefits). Within this framework, we incorporated convenience, price saving and entertainment as the attitudes towards mobile shopping, which indicate the active covariates. Also, a number of demographic, psychographic and device-related covariates are employed as the passive covariates. All of the variables within the scope and their measurement methods can be seen in Table 2.

(26)

In the next section, all of the indicators and covariates which are employed within the conceptual framework will be explained via examples from the past research.

(27)

4.2 Segmentation Indicators

The mobile shopper segmentation will be based on the actual mobile app shopping behavior during different phases of shopping (search and purchase), and for the three different product categories. These indicators will be used to examine the behavioral variations across different products categories and shopping phases.

4.2.1 Actual Behavior

The actual mobile app shopping behavior will be used as the segmentation indicator of this study. Previously, the actual behavior has been used as a segmentation indicator for multichannel customers (Thomas, 2005) (Konus, 2008). Although, this variable has not been used in a mobile shopping app context in isolation, the presence of mobile shopping in multichannel shopping journeys indicates relevance of this variable for our study.

4.3 Covariates

4.3.1 Active Covariates

The customer attitudes toward mobile shopping have been employed as the active covariates in this research. These variables (convenience, price saving and entertainment) will be used to predict the clusters during Latent Class Cluster Analysis.

(28)

4.3.1.1 Attitudes (Expected Benefits)

In this study, customer attitudes refer to the motivation to engage in shopping via mobile apps. In other words, attitudes indicate the expected benefits customers aim to get from engaging in mobile shopping process (Aliawadi et. al, 2001; Konus et. al, 2008).

Early research indicates that expected benefits alter customers’ behavior and channel choices (Aliawadi, 2001). As a further step, this study investigates effects of the attitudes in mobile

shopping app customer segmentation. Therefore, the attitudes will be taken as a part of the active covariates.

4.3.1.1.1 Convenience

Mobile shopping enables customers to have mobility and accessibility to online services with less physical limitations. Therefore, during a customer journey, mobile shopping is expected to enhance convenience, which refers to availability, accessibility and agility within the shopping process (Okazaki and Mendez, 2013). Likewise, convenience has been identified as an important indicator of online shopping attitudes (Keijnen et. al, 2007). Therefore, this study expects

convenience to be effective and relevant to the mobile shopper segments’ formation.

4.3.1.1.2 Price Saving

Previously academic research has employed the motivation for price saving as a driver to search for the best possible offer during a multichannel shopping journey (Konus et. al, 2008). In this study, price consciousness will be examined in terms of its effectiveness on attitudes and behaviors of mobile shoppers.

(29)

4.3.1.1.3 Entertainment

Entertainment stems for the enjoyment customers experience during their mobile shopping journeys. Previously, research has indicated that entertainment has a significant effect on the customers’ attitudes towards online shopping (Childers, 2001). In addition, research suggests that customers engage in mobile shopping more in case they find it entertaining (Chong, 2013).

Therefore, this study expects entertainment to be a significant variable in terms affecting mobile shopper behavior segments.

4.3.2 Passive Covariates

The conceptual framework employs three different types of passive covariates, which are

demographic, psychographic and product-related ones. These variables will not be indicative while identifying the clusters via Latent Class Cluster Analysis, however, they will be used to profile the demographic, psychographic and device-related attributes of the (possibly) observed clusters. In the following section, the covariates and their possible effects will be discussed.

4.3.2.1 Psychographic Covariates 4.3.2.1.1 Innovativeness

Innovativeness refers to one’s willingness to try new experiences, and within this study’s context, different products and services (Midgley and Dowling 1978). Mobile shopping offers

unconventional experiences to customers such as location-based services, which might not be possible during offline/desktop/laptop shopping journeys. Therefore, I expect innovativeness to have a visible effect on the segmentation process.

(30)

4.3.2.1.2 Price Consciousness

Price consciousness refers to the extent of one’s willingness to pay lower prices for products and services (Lichtenstein, Netemyer, and Burton 1990). Considering the multichannel nature of online customer journeys, this study assumes that customers can easily engage with price comparison during mobile app shopping. Therefore, this variable is expected to be effective as a covariate in order to profile the cluster attributes.

4.3.2.1.3 Time Pressure

Time pressure refers to whether customers perceives time as scarce and limited resource and try to make most out of it (Kleijnen, 2007). This study expects to observe time pressure as an effective psychographic covariate to profile segments, because mobile app shopping enables mobility and location-based services, which could also enhance convenience for customers.

4.3.2.1.4 Motivation to Conform

Self-expression has been identified as a hedonic benefit that customers experience during their shopping journeys (Aliawadi et al., 2001). Further, the academics identify motivation to conform as a way of self-expression, which indicates the need for social endorsement during the shopping decision (Ailawadi et al. 2001; Chandon et al. 2000). Considering today mobile app shopping has become a mainstream transaction (We Are Social, 2017), I expect the motivation to conform to have a positive effect on mobile shopping app adoption.

(31)

Previous research suggests that demographics do not have a direct effect on customer behaviors, however they might have effects on psychographic characteristics (Aliawadi et. al, 2001).

Therefore, demographics might have effect on shaping mobile app shopper segments and this study employs demographics characteristics as passive covariates.

4.3.2.2.1 Age

Further research does not offer a comprehensive understanding on the age’s effect on mobile shopping behavior. While one research suggests a negative relationship between age and mobile shopping engagement (Chong et. al, 2013), another study opposes the relationship between age and online shopping (Teo et. al, 2001). For instance, Teo et. al (2001) suggests that there is no relationship between age and online purchases.

Considering the lack of well-defined relationship between age and mobile shopping, this study aims to observe this variable among the profiles of different segments.

4.3.2.2.2 Gender

Previous research has not yet come up with comprehensive findings regarding the effects of gender on attitudes towards mobile shopping. Likewise, Chong et. al (2013) has stated that men and women do not diverge in how they engage in mobile shopping.

This study will investigate the gender allocation of the observed mobile app shopper segments.

4.3.2.2.3 Education

Previous research has suggested relationship between one’s educational background and multichannel shopping behavior, due to education’s potential effects on analytical capabilities which may enable to unlock potential benefits of multichannel shopping (Konus et. al, 2008). Also, there are other studies suggesting this relationship for the mobile shopping context as well (Chong

(32)

et. al, 2013; Yang et. al, 2012). Therefore, this study employs educational background as a passive covariate and expects to see the educational profile distribution among the segments.

4.3.2.2.4 Relationship Status

Previous research has examined the effects of having children on the attitudes toward mobile shopping (Yang, 2005). This study aims to examine the possible effects of one’s relationship status on mobile app shopping.The relationship status indicates in this context that whether a romantic partner is present in customer’s life, and if so the length and status of this relationship.

4.3.2.3 Device Related Covariates

Device related covariates indicates the variables regarding the operating system a device uses, device memory storage, and tablet ownership.

The industry reports suggest that customer behave and purchase differently on Android and iOS operating systems (Criteo, 2015).On the other hand, this study aims to investigate the relation between smartphone storage capacity and segment membership based on the app usage. Therefore, these variables will be employed as passive covariates.

5. RESEARCH PLAN

This section represents the research plan which is dedicated to investigating the following research questions in a structured way: (1) Are there customer segments which can be identified based on their mobile shopping app behavior and attitudes (expected benefits) towards mobile shopping, and (2) if so, can these segments be profiled by their demographic, psychographic and device related attributes?

(33)

This section explains: the sample, research design, procedure, and method that have been used while investigating the presence of consumer segments based on their mobile app shopping behavior.

5.1 The Sample

The target for this study is Dutch and Turkish consumers who use smartphones during their shopping journeys and are older than 18 years old. The population from Netherlands is around 13 million and 75% of the total population (CBS, 2016) and from Turkey is 56,9 million and is 71% of the population (TUIK, 2017). This sample was targeted through professional networks (LinkedIn), and social media (Facebook, Twitter and Whatsapp groups).

As a next step, the snowball distribution method was selected to reach potential respondents. Accordingly, the survey was distributed via social networks LinkedIn, Twitter, Facebook and Whatsapp in the Netherlands and Turkey. Due to the nature of snowball distribution method, the response rate could not be estimated. Therefore, the survey was distributed through multiple channels in order to achieve maximum possible number of respondents in order to gather

sufficient amount of data for this study. Another reason for using various social media channels for the distribution was that I aimed to reach a representative sample which includes respondents from different age groups, educational backgrounds and genders. So that, the research findings can have higher potential to be generalized to a bigger sample.

Following the distribution, the responses have been collected for two weeks, and a total of 356 responses were collected by the 13th of May 2018. From these 356 responses, 91 of them were removed since they were not completed. Also, 14 responses were deleted due to invalid entries

(34)

within the free text responses.

In the end, 251 complete responses were eligible for the analysis stage. Eventually, the further steps (preliminary reliability analysis and latent class segmentation) were performed using the data captured from these 251 responses.

5.2 Research Design

This research has been designed and conducted as an inductive quantitative research, in which no hypothesis was determined from the beginning. Following that, the research data was collected cross-sectional via a self-administered online questionnaire. After the data was collected, the research was planned to continue with a preliminary analysis in SPSS, followed by a Latent Class

(35)

Cluster Analysis (LCCA) in order to segment respondents into different clusters based on the segmentation indicators, and active covariates. At this stage, it would be possible investigate the presence of customer segments based on mobile shopping app behaviors. Furthermore, LCCA would enable to identify the psychographic, demographic and mobile device related

characteristics of these potential segments.

5.2.1 The Survey

The survey was designed only in English. Therefore, only English-speaking respondents were within the scope for Turkey. The survey was divided into three sections, based on the content of the questions (Appendix 1). The first section includes questions about shopping behavior during product search and purchase phases for three different product categories: fashion, consumer electronics, and flight tickets. Prior research indicates the customer behavior may vary for different product categories (Neslin et al, 2008). Therefore, the first section has three sets of the same questions (Appendix 1), in order to observe consumer behavior for each of the categories separately. The second section includes questions about the consumer attitudes (expected benefits) towards mobile shopping. Finally, the third section includes questions focusing on psychographic, demographic, and mobile device related attributes of the participants.

5.3 Procedure

The survey was operated in English, using Qualtrics software via the student account provided by the University of Amsterdam. The procedure includes two stages, the pilot release to test the user experience and understandability of the survey, and the actual survey launch to collect responses

(36)

for the research.

5.3.1 Survey Pilot

Three pilots were performed between the 21st and 25th of April with participation of six pilot respondents. The sequence and understandability of questions, length of introductory letter, and overall user experience was tested with six participants with the following profiles: one student, five working professionals (having one native English speaker profile). Following the pilot, the feedbacks were evaluated and implemented before launching of the survey.

5.3.2 Main Study

The survey was distributed on the 1st of May via Facebook, Twitter and Whatsapp. Initially, the survey was only sent to my personal contacts. Also, the contacts were encouraged to share the survey within their personal networks. Therefore, an extended reach was aimed using snowball distribution method. In addition, the survey was shared in various public groups on Facebook in order to boost visibility and receive participants with various profiles. In the end of that two-weeks response collection period, 356 respondents have started the survey. However, 265 of them have completed it, and 251 of the responds have been qualified for the study.

(37)

This study has employed Latent Class Cluster Analysis (LCCA) methodology due to its merits in terms of investigating different mobile app shopper clusters. Also, previous academic research points the benefits of this methodology.

Firstly, the method inherits model selection criteria as well as probability-based classification. Secondly, LCCA focuses on the latent constructs, rather than visible relations among the variables. Also, this method enables analyzing different variable types at the same time. Moreover, LCCA is the only method that provides an equation for probabilities of membership (scoring equation) (XLSTAT, 2016). Regarding its clustering method, LCCA has a latent variable which has K categories, while each category indicates a cluster. These clusters share common characteristics such as behaviors, motivations or demographics (Vermunt, 2004).

Following this methodology, the respondents were clustered (into K latent classes), based on the indicators (Si) which are used to define the latent classes. Within the analysis, the actual use (U) during the search and purchase phases (a) was used as the segmentation indicator. In addition to the indicator, each of the respondents (i) was assigned to the clusters also regarding the

antecedent variables, which are used to predict the latent classes.

On the other hand, the descriptive measures (passive covariates) have not used to define latent classes.

(38)

The relationships between these variables are formulated as following:

𝑓(𝑈

_𝑖𝑎

|𝑍

_𝑖

)=∑[

′

∏

′

𝑓(𝑈

_𝑖𝑎

|𝑍

_𝑖

, 𝑆

_𝑖

)]𝑝(𝑆

_𝑖

=𝑥|𝑍

_𝑖

)

′

5.4.1 Segmentation Indicators

This research employs actual use (U) as the main segmentation indicator. The actual use/behavior was measured during search and purchase phases for three different product categories (fashion, consumer electronics, and flight tickets). The use of different shopping channels was asked via the following questions: ‘Which channels do you use for product search (collect information, view

different brands/models and compare prices) before you purchase a fashion product/apparel?’, ‘Which channels do you use for purchasing fashion/clothing products?’, and ‘Which channel do you use most for purchasing fashion products?’ (these questions were repeated for all the different

product categories: consumer electronics products and flight tickets).

As a next step, the channel use was measured via giving respondents five possible options where

X=1 K

c=1

(39)

the following selections were possible: ‘Offline (Stores)’, ‘Online, non-mobile (laptop, desktop)’, ‘Smartphone / Mobile Web Browser (non-app)’, ‘Smartphone / App’, and ‘Tablet’.

In addition, the following question was asked to respondents in order to measure the share of tablet and smartphone use during mobile shopping: ‘Considering all the purchases on your mobile

devices (smartphone and tablet), approximately what percentage of these purchases do you make through a smartphone compared to a tablet?’ This item was measured via a metric scale from 0 to

100. Further details about the survey questions and scales can be found in Appendix 1.

5.4.2 Covariates

5.4.2.1 Active Covariates

The active covariates are involved in cluster building and classification along the segmentation process, they are as follows: (1) the attitudes towards mobile shopping (convenience, price saving, and entertainment), and (2) share of online promotions within all the purchases (online and offline) made during discount periods.

Firstly, the survey measures the attitudes (expected benefits) ‘convenience’, ‘price saving’ and ‘entertainment’ using validated scale items. The items were taken from previous research and measured via 5-point Likert scale. However, some of the validated scales were changed in order to adapt them to mobile shopping app use context. The reason was that previous literature have not covered these attitude measures within a mobile shopping app usage focus.

The attitude-related questions were asked using the following format, ‘Please read the following

statements and indicate to what extent do you agree’. Accordingly, the respondents indicated

(40)

refers to strongly disagree and refers to 5 strongly agree).

Secondly, the share of online promotions within all purchases (online and offline) made during discount periods was used as an active covariate. Accordingly, this variable has aimed to measure expected price saving from the online shopping channels, and it was measured using the following question, ‘Considering all your purchases (online and offline) during sales (discount) seasons, for

what percent of your purchases do you use online promotions?’. As a next step, the variable was

measured via a metric scale from 0 to 100. Further details about these questions and the scaled measurement items can be found at Appendix 1.

5.4.2.2 Passive Covariates

Unlike active covariates, these covariates do not define the clusters and do not have an active role in the cluster formation. These covariates were employed to profile the clusters and reveal their psychographic, demographic, and device related attributes.

Firstly, the psychographic variables (price consciousness, time pressure, innovativeness, and motivation to conform) were measured via previously measured scale items. However, some of the validated scales were changed in order to adapt them to mobile shopping app use context. Then, these values were measured using a 5-point Likert scale (where 1 refers to strongly disagree and refers to 5 strongly agree).

Secondly, the study included various demographic variables. Namely, age, education, marital status, gender, and country of residence variables were asked to the respondents through various question types. The age variable was asked via free text entry method and no age interval was used. In order to measure the gender variable, three answer options were given ‘Male’, ‘Female’,

(41)

and ‘Prefer not to disclose’. In order to measure education five possible answer options were given ‘Primary school’, ‘Secondary school / High school’, ‘Bachelor's degree’, ‘Master’s Degree’ and ‘Other’. When respondents select the other option, they could type down their education type. Regarding the marital status variable, six different answers were given to the respondents ‘Single’, ‘In a relationship’, ‘Married’, ‘Divorced or separated’, ‘Widowed’, and ‘prefer not to disclose’. Lastly, the country of residence was measured via three optional answers ‘Netherlands’, ‘Turkey’ and ‘Other’. When respondents select the other option, they can type down their

education type.

Thirdly, device related variables were measured via questions regarding tablet ownership, smartphone memory, and the operating system. Considering potential behavior variations based on the mobile device type or operating systems, these variables aimed to profile device

characteristics amongst the clusters. Firstly, the operating system type was asked to respondents through four answer options: ‘iOS’, ‘Android’, ‘Both’, and ‘Other’ where free entry was possible. Secondly, the respondents received a question about the memory capacity of their smartphone. The question was followed by six answer options, respectively, ‘Less than or equal to 16 GB’, ‘16– 32 GB’, ‘33–64 GB’, ‘65–132 GB’, ‘133–256 GB’ and ‘More than 256 GB’. This question aimed to profile any potential effect that smartphone memory might have on mobile app usage. Lastly, the respondents were asked whether they had a tablet device via the answer options ‘Yes’ and ‘No’.

(42)

6. ANALYSIS AND RESULTS

This section presents the preliminary analysis performed in SPSS, and the Latent Class Cluster Analysis (LCCA) performed in Latent Gold. The preliminary analysis was operated in order to confirm the validity of responses and prepare the data for the Latent Class Cluster Analysis(LCCA) in Latent Gold.

6.1 Preliminary Analysis

As mentioned previously, the data analysis was performed using 251 complete survey responses. Initially, 356 respondents have started the survey. However, 91 of them were removed from the sample due to incomplete responses. As a next step, the frequency of the responses was checked in order to detect possible errors or missing data. In the end of this analysis, 14 responses with missing and abnormal values were detected. Then, the responses with missing values were removed via Listwise deletion method. Accordingly, only the cases without missing values were included in the analysis. At this stage, the listwise deletion method has lowered size of the sample and that might have caused bias along the analysis. However, number of the deleted responses has been relatively small compared to the whole sample. Therefore, the bias effect caused by the deletion was expected to be minor (A. Field, 2013).

In addition, the presence of any counter indicative items was checked, and none was found. Therefore, re-coding for counter indicative items was not necessary. Also, in order to have a simplified LCCA process, the mean values were calculated for some of the psychographic, and attitude related variables.

(43)

6.1.1 Reliability

The scale items for the attitudes (active covariate) and psychographic variables (passive

covariates) were tested in terms of their reliability to measure the intended values (the attitudes and psychographic attributes). Accordingly, the reliability analysis was performed for the items in SPSS: their Cronbach’s alphas were calculated, and the reliabilities of each item were evaluated. In the end of that analysis, Cronbach’s Alphas for the items of variables ‘entertainment’,

‘innovativeness’, ‘price consciousness’, and ‘time pressure’ were equal to or higher from 0.70. On the other hand, attitude-related values ‘convenience’, ‘price saving’, and psychographics-related value ‘motivation to conform’ had Cronbach’s alphas lower than 0.70. Therefore, ‘price saving’ variable was measured only with the following item for the rest of the analysis: ‘I get more price

deals while I shop through online and mobile channels’.

Regarding the convenience, the items ‘Shopping through mobile apps is faster than mobile

browser web (non-app)’ and ‘Shopping through mobile devices saves time’ were removed from the

measurement scale. Accordingly, this variable was measured via the following items for the rest of the analysis: ‘Shopping should be easy’, and ‘Shopping should not take too much time’. The

Cronbach’s alpha for these items has been equal to 0.70.

Furthermore, for ‘motivation to conform’ variable, the items ‘It is important for me to fit in.’ and ‘I

like to solve problems without much thinking.’ were excluded from the analysis due to low value of

their Cronbach’s alphas. Accordingly, this variable was only measured via the following item: ‘It

bothers me when other people criticize my behaviors’.

After checking reliability of the measures and results, the segmentation was executed in Latent Gold in order to investigate the presence of different mobile shopper clusters.

(44)

6.2 Results of Segmentation

6.2.1 Mobile shopping app segments based on actual use (behavior)

Due to nature of the segmentation study, the number of segments and their characteristics cannot be estimated from the beginning of the analysis. Therefore, we need to find the appropriate number of clusters to be able to observe representative and distinctive segments. Accordingly, two criteria were selected in order to compare the appropriateness of the clusters, namely, Akaike Information Criteria (AIC), and Akaike Information Criteria 3 (AIC3).

Then the models were compared based on these criteria. As a result, Model 6 in Table 4 was selected as the most appropriate model, having a combination of a lower AIC(LL) value (10,227), AIC3(LL) value (10,471), and classification error. On the other hand, the clusters within this model can be distinguished significantly. Therefore, this model was selected for further analysis.

(45)

6.2.1.1 Cluster Analysis for Segmentation Indicators

After selecting the appropriate model, the clusters were identified based on the segmentation indicators below. Further details and the relevant data can be seen in Table 5 and Table 6.

Cluster 1. This cluster represents 41% of the total sample. Non-mobile channels (laptop/desktop)

are the most used channels for search and purchases within all of the categories. Also, the users highly prefer offline for purchases within apparel and consumer electronics categories, while they do not use offline channels significantly for search purposes.

Regarding the mobile channels, the users prefer mobile browsers highly for product search within the three categories, however, mobile browsers are also preferred purchase channels for flight tickets. On the other hand, the cluster tend to use mobile apps moderately for search and purchases for fashion and flight tickets categories. This cluster uses tablets very rarely.

Cluster 2. This cluster accumulates 23% of the sample. Offline and non-mobile/online channels

(desktops and laptops) are the most popular channels within that cluster. This cluster has a higher preference towards non-mobile online channels (laptops and desktops) compared to the mobile channels.

The users make most of their product searches via desktop or laptop devices for all of the categories, and most of the flight tickets are purchased via these channels. On the other hand, offline is the most preferred purchase channel for fashion and consumer electronics products. Regarding the mobile channels, this cluster uses mobile web browsers and apps moderately for search and purchases. Although, most of the users own tablet devices, the use of tablets is almost invisible.

(46)

Cluster 3. This cluster illustrates 10% of the sample. This cluster is identified with the high levels of

tablet usage during search and purchase phases. Accordingly, these users make 70% of their mobile (smartphone and tablet) purchases via tablets. Especially, tablets are highly used during search phase. Respectively 95%, 84% and 75% of the cluster uses tablets for search in fashion, electronics and flight tickets categories.

Regarding the purchases, at least 40% of the cluster uses tablet devices for all of the categories. On the other hand, the cluster moderately uses mobile apps during their search and purchase phases in fashion and flight ticket categories.

Although this cluster has a high preference towards tablets, still most of the purchases are made via the other channels. The most used purchase channel is non-mobile (laptop and desktop) for flight tickets and consumer electronics category, while it is the offline channel for fashion category.

Cluster 4. This cluster represents 9% of the sample, and it is identified with high level of mobile

app usage within search and purchase phases across the three categories. Cluster 4 is the only group using mobile apps as the most preferred purchase channel for one of the three categories, namely, flight tickets. In terms of search, 83% of the users use mobile apps for searching for flight tickets, while 66% and 43% of the group use apps for the fashion and electronics category

searches, respectively.

Regarding non-app mobile use, mobile browsers are highly used for search and purchases in fashion and consumer electronics categories. However, the users make most of their purchases from the offline channels within these categories. Also, this group uses offline as an important search channel for fashion and consumer electronics category.

(47)

Cluster 5. This cluster represents 8% of the sample. It is identified with high mobile app and tablet

use during search and purchase phases. Further, this is the only cluster making most of their purchases via mobile apps for all of the categories. In addition, this group uses tablet devices moderately for searches and purchases.

Regarding the non-app mobile use, mobile web browsers are highly used for searches, and they are preferred moderately for purchases. On the other hand, this cluster differs with its lower preference towards offline channels. Unlike the other clusters, offline is one of the least preferred channels for search and purchases.

Cluster 6. This cluster represents 8% of the sample. The group uses mobile apps significantly along

different phases of shopping (search and purchase). However, the users prefer mobile browsers more than apps for search and purchases within all the categories. Considering the high use of mobile devices for search and purchases, this cluster might hold a potential for a higher level of mobile app adoption. On the other hand, currently the users prefer to purchase goods mostly from offline and non-mobile channels (desktops, laptops). In other words, the users engage with mobile touch points during shopping journeys and prefer to make to purchase via non-mobile channels.

(48)

(49)

6.2.1.2 Cluster Analysis for Active Covariates

At this section, the clusters are further analyzed based on the active covariates. Accordingly, each cluster’s attitude towards shopping and their level of online shopping promotions use will be discussed.

Firstly, all of the clusters were observed to value convenience and price saving, while Cluster 4 values convenience moderately less, and price saving more compared to the others. Since all the clusters have these attitudes to an extent, they are not very significant at identifying

characteristics for the segments. On the other hand, Cluster 4, which is identified by the highest level of mobile shopping app use during search and purchase phases, has the highest shopping entertainment level. Hence, this relationship might indicate a positive relationship with the app use and entertainment expectancy from mobile app shopping. Further explanations about the segmentation indicators and cluster characteristics can be seen in Table 5 and Appendix 3.

(50)

(51)

7. DISCUSSION

This research aims to enhance the current academic research regarding the mobile app shopping behavior in a growing mobile shopping landscape. Therefore, it focuses on investigating the

presence of different mobile app shopper segments, based on the actual shopping behavior across three different product categories. Meanwhile, this study includes two phases of a shopping journey: product search and purchase. In order to do so, a latent class cluster analysis (LCCA) was performed. After this analysis, six different clusters (segments) have been identified based on the shopping behavior during the search and purchase phases, and the attitudinal differences towards mobile shopping.

In this last chapter, the results of the study will be discussed, also the theoretical and managerial implications of the results will be presented. Lastly, the limitations of the research, and the future research areas will be explained.

Segmenting customers on the basis of their mobile shopping app orientation, behavior and other drivers of mobile shopping

Master Thesis