Mobile Shopping Through Applications: Understanding Application Possession and Mobile Purchase

Available online at www.sciencedirect.com

ScienceDirect Journal of Interactive Marketing 39 (2017) 55 – 68 www.elsevier.com/locate/intmar

Mobile Shopping Through Applications: Understanding Application Possession and Mobile Purchase Mingyung Kim a & Jeeyeon Kim b & Jeonghye Choi b,⁎& Minakshi Trivedi c a

Department of Statistics, University of California, Berkeley, Berkeley, CA 94720, USA b School of Business, Yonsei University, Seoul 03722, Republic of Korea c Neeley School of Business, Texas Christian University, Fort Worth, TX 76109, USA

Abstract Smartphones have penetrated rapidly and mobile shopping provides promising market opportunities for retailers. However, little is known about mobile shopping patterns and inferring these patterns from online shopping may provide misleading insights. We combine mobile log data and a mobile panel survey, and examine two stages in mobile shopping: the possession of shopping applications (hereafter, apps) and the purchase via shopping apps. Our exploratory investigation of mobile data and its empirical analyses provide three substantive findings. First, online experience and mobile experience both positively relate to the possession of shopping apps. Second, browsing behavior for non-shopping apps helps understand the possession of shopping apps as it reflects user preferences for acquiring more apps. Third, purchasing decisions are explained by digital experience (i.e., online experience and mobile experience) and browsing information from shopping apps, with other factors being of little predictive value. The implications for mobile retailing research and practice are discussed. © 2017 Keywords: Smartphone; Mobile shopping; Mobile applications; Mobile browsing; Mobile experience; Online experience

Introduction Smartphones have penetrated rapidly since their advent, and presently, more than 50% of mobile owners use smartphones in many countries (comScore 2015; eMarketer 2014a, b, c).1 With the prevalence of smartphones, the mobile channel has become the third marketplace, following the offline and online channels; however, little is known about this mobile channel (Bang et al. 2013; Kleijnen, De Ruyter, and Wetzels 2007). Furthermore, despite the remarkable growth in users and the

⁎ Corresponding author. E-mail addresses: [email protected] (M. Kim), [email protected] (J. Kim), [email protected] (J. Choi), [email protected] (M. Trivedi). 1 Country-specific smartphone penetration was released by different organizations. comScore (2015) announced that smartphones were in use by more than 50% of US consumers in 2014. eMarketer (2014a, b, c) reported the same penetration rate in 2014 in European countries (e.g., UK, Denmark, and Sweden) and Asian countries (e.g., Korea, China, and Japan). http://dx.doi.org/10.1016/j.intmar.2017.02.001 1094-9968/© 2017

consequent market potential, revenue from mobile shopping still accounts for a small percentage of the overall retailing sector (eMarketer 2014a, b, c).2 Therefore, there is a growing need to understand mobile shopping and its drivers. Mobile shopping requires smartphones and this shopping behavior cannot be directly inferred from computer-based online shopping behavior. For instance, smartphones provide ubiquitous shopping opportunities; however, inconvenient interfaces increase search costs and inhibit mobile purchasing (Bang et al. 2013; Chong 2013; Ghose, Goldfarb, and Han 2013; Goh, Chu, and Wu 2015). Thus, this study aims to contribute to the understanding of mobile shopping. We obtain mobile log data, the mobile version of clickstream data, across mobile retailers, and compare and contrast two stages in mobile shopping: the possession of shopping applications (hereafter, apps) and purchasing through shopping apps. 2

eMarketer (2014a, b, c) said the m-commerce market is expected to reach $98 billion in the US by 2016, and this would account for about 1% of the total retail sector.

56

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

We offer three substantive findings. First, online experience (i.e., experience accumulated through online shopping) and mobile experience (i.e., experience through smartphone usage) both positively relate to the possession of shopping apps. Second, browsing behavior for non-shopping apps helps understand the possession of shopping apps. However, in contrast to online sales channels where visits to any site impact online shopping, preloaded apps fail to explain mobile shopping. Lastly, mobile purchases through shopping apps are explained unsurprisingly by the browsing behaviors for these shopping apps. In fact, mobile purchases are determined solely by digital experience (i.e., online experience and mobile experience) and the browsing patterns of the shopping apps along with all other factors, are of little-to-no predictive value. The remainder of this study proceeds as follows. The next section reviews the related literature and describes our hypotheses. The subsequent section describes the data and measures. We then introduce the exploratory findings, which prompt the empirical analyses. After we discuss empirical models, we report empirical findings. The concluding section discusses implications for academics and practitioners and avenues for future research. Related literature We first review prior research in the areas of online and mobile shopping. Using evidence from these areas, we identify research directions and set up our hypotheses for mobile shopping behavior. The Internet and Online Shopping Early research on the Internet focused on online experience and online activities. Emmanouilides and Hammond (2000), for example, conducted a survey measuring online experience by how long an individual has been using the Internet, and found that online experience is a predictor of online activities such as frequent browsing of online sites. Kraut et al. (1999) also revealed that Internet users showed preferences for activities they had experienced for a longer time because the benefits of such activities could be easily understood. As the Internet has emerged as a shopping channel, research has moved toward identifying key factors affecting online shopping. Seminal research has used survey data to identify these. Forsythe and Shi (2003) proved that heavy online shoppers are likely to be more experienced ones (i.e., they have used the Internet for four or more years) than light and window shoppers because perceived risks of online purchasing decrease as years of online experience increase. Brengman et al. (2005) found that heavy online browsers of non-shopping sites (e.g., sites for information and entertainment) are likely to be frequent online shoppers. Venkatesh and Agarwal (2006) asked panelists to visit multiple sites and then, after six months, to recall their browsing pattern (i.e., frequency, duration, and intensity) and purchase experience (i.e., frequency and average amount) at each website via telephone interviews. They found that online browsing had a positive impact on online purchases. Furthermore, Konus,

Verhoef, and Neslin (2008) confirmed that product characteristics influence consumer multichannel shopping behavior using the Internet, catalogs, and stores. For example, consumers favor multichannel shopping when purchasing electronics but not when purchasing clothes. Internet clickstream data, the electronic record of online activity, enriches research on online shopping, particularly by providing browsing data across sites and competitive retailers. Using such data, Moe and Fader (2004) discovered the positive effect of browsing on online shopping. Specifically, the more visits a consumer makes, the more likely he/she is to purchase products. Sismeiro and Bucklin (2004) investigated the purchase process using three steps: interest (i.e., the completion of product configuration), desire (i.e., the input of personal information), and finally purchase (i.e., the order confirmation after providing credit card information). They found that online exposure variables, such as number of links, are significant in the first two stages but insignificant at the purchase stage. Huang, Lurie, and Mitra (2009) found that consumers engage in extensive online searching on shopping sites prior to purchasing: For a single online purchase, consumers visit 3.4 sites, create 124 sessions, and spend 78 minutes (on average).3 Smartphones and Mobile Shopping The advent of smartphones allows scholars to expand the scope of mobile research beyond basic functions, such as calling and texting, and to compare and contrast online and mobile behaviors. Smartphone apps in many cases are the mobile versions of online sites and companies usually design and launch apps similar to their online sites when expanding their business to the mobile platform (Bang et al. 2013). Still, stark differences remain between smartphones and computers. Goh, Chu, and Wu (2015) found that information search behaviors using mobile phones are different from desktop computer search behaviors. Mobile users intermittently read content because mobile content is shown on smaller screens. Ghose, Goldfarb, and Han (2013) found a similar result that the smaller screens of mobile devices increase search costs, which in turn makes the relative attractiveness of the first search result over the second greater on mobile devices than on computers. Chong (2013) showed that mobile users who value ubiquitous access prefer mobile phones to computers for watching videos and listening to music. These findings underline the distinctions between the online and mobile channels. Therefore, inferring mobile behavior from research on online behavior could result in misleading information. The prevalence of smartphones offers a third channel of shopping, following offline and online channels. Kleijnen, De Ruyter, and Wetzels (2007) developed a conceptual model that incorporates the benefits (i.e., time convenience and user control) and costs (i.e., risks and cognitive efforts) of mobile shopping. In their model, time-related gains in efficiency 3 It is possible that a consumer visits a site and then leaves sessions unattended. To avoid the inflating effect of such sessions, the authors discarded the sessions when the idle time exceeded 5 minutes.

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

increase the perceived value of mobile shopping, resulting in higher purchase intention. Bang et al. (2013) analyzed browsing and purchase data for the online retailer that launched a mobile app. They found that the online shopping experience was positively related to the downloading of its shopping app because a mobile device allows for ubiquitous access to the shopping environment similar to that of its online site. However, consumers browsing multiple categories were less likely to adopt shopping apps, as the mobile interfaces were inferior to the interfaces on computer screens.

Our Study Research on mobile shopping is still scarce and more research is called for to expand understanding of this emerging mobile channel. As in online research (Sismeiro and Bucklin 2004; Venkatesh and Agarwal 2006), by studying multiple stages in mobile shopping across multiple retailers, we can advance theories specific to the mobile interface. To this end, the mobile version of the Internet clickstream data will provide detailed mobile specific information. Thus, we segment mobile shopping into two stages: the possession of shopping apps and the purchasing via these apps, and investigate mobile specific shopping using relevant log data from multiple retailers.

Hypotheses Smartphones have become widespread devices but research on mobile shopping is in its infancy. Seminal research on mobile devices warns against blindly inferring mobile shopping behavior from online shopping due to some intrinsic differences between the two channels (Chong 2013; Ghose, Goldfarb, and Han 2013). Therefore, we examine this unexplored area of mobile shopping by testing the following hypotheses with the aim to offer new insights into mobile shopping. A conceptual representation of the hypotheses involved is given in Fig. 1.

57

Digital Experience and App Possession (H1) Experienced online shoppers are familiar with the online shopping environment (e.g., website interfaces and product displays); this higher familiarity lowers barriers to shopping apps as the app environment is similar to that of websites (e.g., Bang et al. 2013). Thus, we infer that experienced online shoppers possess a greater number of shopping apps. It is unsurprising that online users exposed to more online activities perceive decreasing risks in conducting these activities as their years of online experience increase (Emmanouilides and Hammond 2000; Forsythe and Shi 2003). Similarly, as experience in using smartphones accumulates, users are likely to be exposed to a greater number of apps. Mobile experience is therefore expected to lower the perceived risks of downloading apps, which thereby leads to the possession of a greater number of shopping apps. Thus, we hypothesize the following: H1. Digital experience increases the possession of shopping apps. Specifically, (a) more experienced online shoppers possess a greater number of shopping apps, and (b) more experienced smartphone users possess a greater number of shopping apps. Non-shopping Apps and App Possession (H2) Prior experience lowers the costs associated with conducting similar activities going forward. Previous visits to online sites lower the perceived costs of visiting additional sites (Emmanouilides and Hammond 2000), and prior experience purchasing online increases the likelihood of making subsequent purchases (Venkatesh and Agarwal 2006). We follow this thinking and infer that as mobile users download and use more apps, they experience the decreasing costs (e.g., time and effort) of acquiring additional apps. That is, the greater the number of non-shopping apps users download and install, the greater number of shopping apps they will have. On the other hand, personal resources for using apps are limited (Chong 2013; Kleijnen, De Ruyter, and Wetzels 2007). If users browse

Fig. 1. Conceptual framework.

58

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

non-shopping apps frequently and heavily, they have less time for shopping apps and, as a result, users may possess a smaller number of shopping apps. Thus, we advance the following hypothesis: H2. Browsing of non-shopping apps helps understand the possession of shopping apps. Specifically, the possession of shopping apps is greater (a) when users browse a greater number of non-shopping apps and (b) when users browse non-shopping apps less frequently and less heavily. Digital Experience and Mobile Purchase (H3) Mobile shoppers go through several decision steps (e.g., choose products, share personal information, and provide credit card information) similar to those for online purchasing (e.g., Bang et al. 2013; Bellman et al. 2011; Sismeiro and Bucklin 2004) and can use purchase-related information they have accumulated online. Experience in online purchasing can thus lower learning costs related to mobile purchasing, which can increase the number of shopping apps being used for mobile purchases. Prior experience relieves concerns over related activities because experienced users have learned how to handle those concerns. For instance, online experience lowers security and privacy concerns over online purchasing (Forsythe and Shi 2003). Similarly, mobile experience may relieve mobile users of similar concerns regarding mobile purchasing across shopping apps. Mobile experience is thus expected to help users in purchasing via a greater number of shopping apps. Thus: H3. Digital experience increases mobile purchasing using shopping apps. Specifically, (a) more experienced online shoppers make mobile purchases using a greater number of shopping apps, and (b) more experienced smartphone users make mobile purchases using a greater number of shopping apps. Shopping Apps and Mobile Purchase (H4) Greater browsing of online shopping sites leads to greater online purchasing (Moe and Fader 2004; Sismeiro and Bucklin 2004). This is because online shoppers who have higher purchase intentions are likely to spend more time and effort on such online shopping sites. Whether this behavior translates to the mobile context however, is not clear. Indeed, the smaller screens of mobile phones have been shown to reduce readability and hamper information search (Bang et al. 2013; Ghose, Goldfarb, and Han 2013; Goh, Chu, and Wu 2015). For instance, mobile users are reluctant to search for and purchase products from multiple categories (Bang et al. 2013). If in spite of these mobile shortcomings, users browse mobile shopping apps more extensively, they are also likely to engage in mobile shopping and in turn make mobile purchases using a greater number of shopping apps. Thus, we hypothesize as follows: H4. Browsing of shopping apps helps understand mobile purchasing via shopping apps. Specifically, (a) if users browse a greater number of shopping apps, they purchase via a greater

number of shopping apps, and (b) if users browse shopping apps more frequently and for a longer time, they make purchases using a greater number of shopping apps. Data and Measures We describe the mobile panel data obtained from TNS, a global market research company, and then introduce the variables to be used in our empirical analyses. Data Sources TNS developed a mobile app to collect mobile data at the individual phone level and then launched a mobile intelligence service using the mobile data in one country after another. TNS Korea, the Korean subsidiary of TNS, collected mobile data starting in July 2012 and launched the service in September 2012. We obtained the following two datasets from TNS Korea: (1) mobile log data and (2) panel survey data. Mobile Log Data TNS Korea recruited a panel of 978 Android smartphone users.4 Once the tracking app is installed on smartphones, it collects phone-related information such as type of cellular network supported, smartphone brand, and mobile service operator. The main function of the app is to track every activity that the panel members conduct on their smartphones. The log data include the name of the app activated, duration of time in use, and whether the app was terminated because another app was initiated or the smartphone went idle. This mobile log data is the same as the Internet clickstream data that track every interaction with Internet browsers. Following Bucklin and Sismeiro (2003) who used the one-month Internet clickstream data, we analyze one-month mobile log data collected in July 2012. We define shopping apps as those that allow users to make “shopping” transactions through their smartphones.5 The shopping apps of interest include eight apps offered by traditional retailers and seven apps by pure click retailers, and cover product categories such as grocery, clothing, beauty, electronics, and tickets.6 We focus on these top 15 shopping 4 Market research companies had difficulty collecting the log data of iOS smartphones due to technical problems during the data period. However, about 75% of the smartphone users in Korea had Android smartphones as of September 2012 and TNS Korea confirmed that its mobile panel is not subject to any bias in representing Korean smartphone users during the same data period. 5 We define shopping apps from the industry and consumer perspectives. Consumers can make a purchase using various apps such as portal apps and game apps; however, the industry practice does not classify these into the shopping category. Moreover, consumers will perceive the categories of different apps based on the main purpose of using them, other than supplementary functions. Thus, our definition of shopping apps follows the category definition in the Android Market and includes the major players in the retailing industry. 6 Traditional retailers sell through their website as well as their traditional bricks-and-mortar stores whereas pure click retailers do not operate offline stores.

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

apps because they account for 95% of the total number of sessions and 97% of the total time spent in the shopping category in July 2012.7 We segment the remaining apps into non-shopping and preloaded apps. Non-shopping apps refer to those that users download and install by themselves but are unrelated to mobile shopping such as game apps (e.g., Temple Run) and social networking apps (e.g., Facebook). Preloaded apps are those preinstalled by smartphone manufacturers and mobile service operators when the smartphones are purchased, and include, for example, apps for calling, texting, and emailing. Panel Survey Data Mobile purchase data cannot be collected through the tracking app due to privacy policies, and thus TNS Korea conducted a survey in late August 2012.8 Specifically, it asked the mobile panel members to recall the number of shopping apps they used in one month, out of 15 possible shopping apps, to make mobile purchases (see Venkatesh and Agarwal 2006, for the same approach).9 A one month time period is long enough to measure mobile shopping as users' browsing patterns do not change much from month to month (e.g., Johnson et al. 2004). Less than that captures too much noise and more than that substantially increases recall errors (e.g., De Mel, McKenzie, and Woodruff 2009). A total of 785 panelists completed the survey, leading to a high response rate of 80%. The survey also contained questions about digital experience (i.e., online experience and mobile experience), smartphone hardware, and socio-demographics. Smartphone hardware includes type of cellular network supported, smartphone brands, and mobile service operators, which align perfectly with the phone-related information collected through the tracking app. This indicates that mobile users pay enough attention to high involvement products like smartphones so that their memories should be highly reliable. Variable Description Mobile Shopping We study two stages in mobile shopping, the possession of shopping apps and the purchase via shopping apps. Mobile users (or smartphone owners) become “potential” mobile shoppers if they have shopping apps installed on their smartphones, and thus we define possession by the number of shopping apps installed as of the beginning of August using the mobile log data. Mobile purchase experience turns “potential” mobile buyers into “actual” 7 Those who have obscure shopping apps might be more inclined to adopt the top 15 focal apps and thus we include the possession of non-focal shopping apps in the possession equation as a robustness check. The empirical results remain consistent after we control for the number of non-top 15 shopping apps as of the last day of July. 8 The survey was conducted on August 26, 2012, which helps establish temporal precedence of the explanatory variables measured using mobile log data in July. As a robustness check, we establish clearer temporal precedence by focusing on log data between July 1 and July 26 and the empirical results remain qualitatively identical. 9 Asking consumers to recall purchase experience is unlikely to distort empirical findings. See Venkatesh and Agarwal (2006) for detailed explanation.

59

buyers, thus purchase is defined by the number of shopping apps used for purchasing using the panel survey data. Digital Experience Digital experience helps explain mobile shopping behaviors (e.g., Emmanouilides and Hammond 2000; Forsythe and Shi 2003) and is operationalized using two variables. Online experience is a categorical variable having three levels of online purchase intensity. Specifically, those who buy online once or more per day are defined as having high online experience. Similarly, those who buy online once or more a week are defined as having medium online experience, and those remaining are defined as having low online experience. Mobile experience is also classified into three levels. Those who have used smartphones for more than two years are defined as having high mobile experience. Similarly, those with one-to-two years of experience are defined as having medium mobile experience, and those remaining as having low mobile experience. Mobile Browsing Following prior research on online browsing (Emmanouilides and Hammond 2000; Kraut et al. 1999), we measure three dimensions of mobile browsing: breadth, frequency, and length, using the mobile log data.10 Table 1 shows hypothetical user i who installed three shopping apps, and the ways user i uses each app over three days. Breadth measures the number of mobile apps used at least once, thus, the breadth is equal to two. Frequency and length measure on average how often the apps are activated and how long they are in use, respectively. The frequency is the average number of sessions used per app and computed as (1 + 3) / 2 = 2 (sessions per app). The length is the average time spent per session and measured by (2 + 4) / (1 + 3) = 1.5 (minutes per session). As there are two key types of mobile apps, we define the mobile browsing variables of breadth, frequency, and length, separately for shopping apps (i.e., SBi, SFi, and SLi) and non-shopping apps (i.e., NBi, NFi, and NLi).11 We also measure the mobile browsing variables for preloaded apps (i.e., PBi, PFi, and PLi). Notably, the decisions on the preloaded apps may be independent of smartphone users' preferences for apps and thus, the corresponding browsing variables serve as control variables in the following empirical analyses. Control Variables We control for smartphone hardware that is likely to influence mobile shopping. LTE supported: Yes is a binary variable taking 1 if a smartphone supports both LTE and 3G cellular networks and 0 if it supports only 3G. As the LTE network is more advanced than the 3G network, consumers using LTE phones tend to be tech-savvier than those using 3G-based smartphones, thus are more likely to possess a greater 10 Following Huang, Lurie, and Mitra (2009), we define that a session of an app ends when (1) another app is initialized or (2) a device goes idle for more than 5 minutes. In addition to 5 minutes, we have tested 2, 3, 7, and 10 minutes and all the results remain consistent. 11 The depth variables in the empirical analyses are measured in seconds per session.

60

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

Table 1 Example of breadth, frequency, and length. Apps in use App 1 Number of sessions Day 1 Day 2 Day 3 Total

0 0 1 1

Time of sessions 0 0 2 2

App not in use App 2

minutes minutes minutes minutes

Number of sessions 0 2 1 3

App 3

Time of sessions 0 2 2 4

minutes minutes minutes minutes

Number of Time of sessions sessions – – – –

– – – –

Notes: • Breadth = 2 (i.e., the number of apps in use). • Frequency = 2 (i.e., total number of sessions / number of apps in use). • Length = 1.5 (i.e., total time of sessions / total number of sessions).

number of shopping apps. The smartphone brand is a categorical variable that takes three values, one for each of the top two brands and one for the remaining brands. In empirical models, the brand with the largest market share becomes a reference group. Mobile service operator represents three mobile carriers, with the major operator as a reference.12 We also include socio-demographic variables that prior research finds as relevant to both online and mobile shopping (Choi, Bell, and Lodish 2012; Ghose and Han 2011). These variables include both age and age-squared variables to handle nonlinear effects of age for mobile shopping behaviors (e.g., Koyuncu and Lien 2003). Except for the age variable, the remaining socio-demographic variables are categorical. Gender is equal to 1 if an individual is female and 0 otherwise. Occupation represents three categories: white collar, blue collar, and no occupation, with white collar as a reference.13 We expect blue-collar workers to use a smaller number of shopping apps than white-collar workers. Blue-collar workers may adopt mobile shopping more slowly because they may spend less time on the Internet at work and be less familiar with digital activities. Income also includes three categories based on monthly earning: more than 5,000 dollars, 3,000 to 5,000 dollars, and less than 3,000 dollars, with the highest income group as a reference group. Finally, Region is 1 if a user lives in the capital city and 0 otherwise. Final Dataset As mentioned above, 785 mobile users completed the panel survey, but nine were dropped from the final dataset due to sporadic missing answers. The final dataset has 776 smartphone users, including 380 users who installed at least one shopping 12 The top two smartphone brands in Korea have market shares of 65% and 20%, respectively. The other brands together have the remaining market share of 15%. Also, there are three mobile operators in Korea. The largest operator has a market share of 50%, the next has 32%, and the last one, the remaining 18%. 13 We categorize students and housewives as those with no occupation since students and housewives are generally classified as economically inactive and dependent. Note that we do not include education variables due to the multicollinearity problem. Moreover, Koreans are highly educated with 80% of the population aged between 20 and 34 years having university degrees.

app, and 164 users who purchased via such shopping apps. Table 2 shows the variables and their descriptive statistics. The correlation matrices show that the independent variables are not highly correlated with each other. We also confirm that the multicollinearity would not occur with all of the VIF scores less than 3, which is below the typical cut-off of 10 (Hair et al. 2006). Exploratory findings Marketers often obtain practical insights into marketing strategies by dividing consumers into groups and comparing how these groups differ through decision stages (Konus, Verhoef, and Neslin 2008; Sismeiro and Bucklin 2004). We also look at differences across user groups through two stages of mobile shopping: possession and purchase. To explore interesting patterns and compare them with the hypotheses, Table 3 shows the descriptive results of the key variables: digital experience and mobile browsing. First, we segment total mobile users into two groups based on whether or not they possess shopping apps. The group without shopping apps represents about half of the mobile users, and the group with shopping apps the other half. According to the results, users who have shopping apps are more experienced with online shopping than those without shopping apps (38% vs. 21% for high online experience; 43% vs. 40% for medium online experience) (i.e., consistent with H1(a)). They are also more experienced smartphone users (21% vs. 12% for high mobile experience; 40% vs. 38% for medium mobile experience) (i.e., consistent with H1(b)). Moreover, while they tend to download a greater number of non-shopping apps (41 vs. 31), they browse those apps less frequently (30 vs. 38) (i.e., consistent with H2). This could be because acquiring additional apps becomes easier with accumulated experience in downloading apps, but when it comes to actual app usage, more available apps lead to less time per app given an individual's allotted time for mobile browsing. This suggests that mobile retailers can increase the pool of potential mobile shoppers by targeting users who are more experienced in online shopping and smartphone usage, and have a greater number of non-shopping apps but spend less time on them. Next, we examine how mobile buyers are different from mobile “window-shoppers.” We focus on the group with shopping apps and further define subgroups based on whether or not they make purchases via mobile shopping apps. Of note, roughly 20% of the mobile users made mobile purchases in the month. These mobile buyers are more experienced online shoppers than the non-buyers (45% vs. 32% for high online experience; 45% vs. 42% for medium online experience) (i.e., consistent with H3(a)), and have used their smartphones longer (25% vs. 19% for high mobile experience; 43% vs. 38% for medium mobile experience) (i.e., consistent with H3(b)). In addition, these buyers use a greater number of mobile shopping apps (2 vs. 1), visit them more frequently (12 vs. 4), and browse them longer (162 vs. 82 minutes) (i.e., consistent with H4). This implies that mobile retailers can increase purchase conversion rates by targeting heavy browsers of shopping apps who are experienced in online shopping and smartphone usage.

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

61

Table 2 Descriptive statistics and correlations among all covariates. [A] Means and standard deviations Variables

Possession

Purchase

Mean Dependent variables Digital experience Online Experience: Higha Online Experience: Med Mobile Experience: Highb Mobile Experience: Med Mobile browsing Shopping apps SB: N of apps SF: N of sessions per app SL: Time per session Non-shopping apps NB: N of apps NF: N of sessions per app NL: Time per session Control variables Smartphone hardware LTE supported: Yes Smartphone brand: B Smartphone brand: Others Mobile service operator: B Mobile service operator: C Socio-demographics Age Gender: Female Occupation: Blue collar Occupation: None Income: Below $3,000 Income: $3,000 to $5,000 Region: Capital Browsing of preloaded apps PB: N of apps PF: N of sessions per app PL: Time per session

SD

Mean

SD

1.134

1.699

.705

1.110

.293 .414 .165 .390

.455 .493 .371 .488

.376 .429 .213 .403

.485 .496 .410 .491

– – –

– – –

1.442 7.634 116.353

1.525 13.942 162.988

35.829 34.240 154.526

22.640 34.197 76.836

41.018 30.235 157.013

24.065 24.874 74.123

.286 .174 .197 .321 .184

.452 .379 .398 .467 .388

.347 .147 .129 .289 .145

.477 .355 .336 .454 .352

31.738 .452 .173 .379 .299 .412 .361

10.088 .498 .378 .485 .458 .493 .481

31.621 .489 .158 .361 .303 .421 .361

9.503 .501 .365 .481 .460 .494 .481

20.823 43.622 101.587

6.128 27.559 62.583

21.289 43.935 98.093

6.057 26.660 61.530

Notes: Specifically, it refers to online shopping experience, classified into three levels. Those who shop online once or more per day are defined as having high online experience. Similarly, those who shop once or more per week are defined as medium, and those remaining as defined as low. b It is classified into three levels. Those who use smartphones more than two years are defined as having high mobile experience. Similarly, those with more than one year are defined as medium, and those remaining are defined as low. a

[B] Correlation Matrix for the Possession Stage Variable

1

2

3

4

5

6

7

8

9

10

11

12

13

1. DV: Possession 2. Online Experience: High 3. Online Experience: Med 4. Mobile Experience: High 5. Mobile Experience: Med 6. SB: N of apps 7. SF: N of sessions per app 8. SL: Time per session 9. NB: N of apps 10. NF: N of sessions per app 11. NL: Time per session 12. LTE Supported: Yes 13. Smartphone Brand: B 14. Smartphone Brand: Others 15. Mobile Service Operator: B 16. Mobile Service Operator: C

1.00 .16⁎ .03 .16⁎ .01 — — — .23⁎ -.11⁎ .02 .08⁎ -.06 -.15⁎ -.03 -.07⁎

1.00 -.54⁎ .03 .05 — — — .03 .02 -.06+ .00 .02 .01 .01 .05

1.00 .01 -.00 — — — -.03 -.01 -.02 .02 .01 -.03 -.07⁎ -.03

1.00 -.36⁎ — — — .03 -.07+ .06+ .23⁎ .01 -.05 -.10⁎ .06

1.00 — — — -.08⁎ .01 .04 -.12⁎ .06 .02 -.10⁎ -.01

— — — — — — — — — — —

— — — — — — — — — —

— — — — — — — — —

1.00 -.12⁎ -.13⁎ .04 -.02 -.06+ -.06+ -.02

1.00 -.17⁎ -.06 .03 .11⁎ .04 .01

1.00 .04 -.03 -.10⁎ -.06 -.05

1.00 .15⁎ -.16⁎ -.20⁎ .29⁎

1.00 -.23⁎ -.13⁎ .38⁎

(continued on next page)

62

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

Table 2 (continued) [B] Correlation Matrix for the Possession Stage Variable

1

2

3

4

5

6

7

8

9

10

11

12

13

17. Age 18. Gender: Female 19. Occupation: Blue collar 20. Occupation: None 21. Income: Below $3,000 22. Income: $3,000 to $5,000 23. Region: Capital 24. PB: N of apps 25. PF: N of sessions per app 26. PL: Time per session

.00 .03 -.06 -.03 -.03 .05 -.02 .11⁎ .04 -.04

-.03 .14⁎ -.01 -.05 -.03 -.01 .01 .04 -.00 -.06+

.02 -.07+ .07+ -.06 -.06 .04 .00 -.00 .06 .02

.07+ -.07+ -.02 -.12⁎ .01 -.01 .01 .07+ -.04 .05

-.07⁎ .01 .04 -.03 -.02 .00 -.04 -.14⁎ -.05 -.03

— — — — — — — — — —

— — — — — — — — — —

— — — — — — — — — —

-.23⁎ .01 -.09⁎ .11⁎ .05 .00 -.06 .41⁎ .07+ -.16⁎

-.22⁎ .15⁎ -.05 .16⁎ .08⁎ -.04 -.02 -.05 .27⁎ -.15⁎

.07+ -.08⁎ .03 -.08⁎ -.02 .05 .03 -.14⁎ -.18⁎ -.29⁎

.04 .00 .04 -.02 .04 -.02 .02 .16⁎ -.06+ .04

.05 -.04 .05 -.02 .00 .04 .02 .15⁎ .03 .08⁎

14. Smartphone Brand: Others 15. Mobile Service Operator: B 16. Mobile Service Operator: C 17. Age 18. Gender: Female 19. Occupation: Blue collar 20. Occupation: None 21. Income: Below $3,000 22. Income: $3,000 to $5,000 23. Region: Capital 24. PB: N of apps 25. PF: N of sessions per app 26. PL: Time per session

14

15

16

17

18

19

20

21

22

23

24

25

26

1.00 .23⁎ -.14⁎ -.10⁎ .04 -.02 .15⁎ .01 .05 -.04 .02 -.19⁎ .05

1.00 -.33⁎ -.06+ -.04 -.04 .13⁎ .01 .04 -.07⁎ .02 -.05 -.08⁎

1.00 .01 .04 -.02 -.01 .05 -.03 -.02 .14⁎ .03 .00

1.00 -.35⁎ .19⁎ -.47⁎ -.21⁎ .05 .09⁎ -.01 .13⁎ .05

1.00 -.14⁎ .34⁎ .10⁎ -.02 -.03 -.01 .00 -.08⁎

1.00 -.36⁎ .01 -.02 .00 -.01 -.04 .-00

1.00 .18⁎ -.01 -.04 -.01 -.14⁎ .00

1.00 -.55⁎ -.10⁎ -.01 -.06+ -.02

1.00 -.02 -.02 .02 .04

1.00 -.03 -.01 .08⁎

1.00 .11⁎ -.06+

1.00 -.12+

1.00

[C] Correlation Matrix for the Purchase Stage Variable

1

2

3

4

5

6

7

8

9

10

11

12

13

1. DV: Purchase 2. Online Experience: High 3. Online Experience: Med 4. Mobile Experience: High 5. Mobile Experience: Med 6. SB: N of apps 7. SF: N of sessions per app 8. SL: Time per session 9. NB: N of apps 10. NF: N of sessions per app 11. NL: Time per session 12. LTE Supported: Yes 13. Smartphone Brand: B 14. Smartphone Brand: Others 15. Mobile Service Operator: B 16. Mobile Service Operator: C 17. Age 18. Gender: Female 19. Occupation: Blue collar 20. Occupation: None 21. Income: Below $3,000 22. Income: $3,000 to $5,000 23. Region: Capital 24. PB: N of apps 25. PF: N of sessions per app 26. PL: Time per session

1.00 .14⁎ -.01 .11⁎ -.01 .58⁎ .23⁎ .17⁎ .09+ -.05 -.03 .03 .00 -.07 -.01 .12⁎ .01 .01 .03 -.07 -.05 -.02 -.09+ .03 .04 -.07

1.00 -.67⁎ -.01 .00 .13⁎ .09+ .15⁎ -.01 .10⁎ -.06 -.06 .04 .01 .02 .11⁎ -.05 .10+ .01 -.06 -.02 -.08 -.01 .04 -.01 -.05

1.00 .03 .06 -.04 -.04 -.10+ -.08 -.07 .01 .04 -.00 -.06 -.07 -.07 .05 -.04 .06 -.03 -.07 .11⁎ .00 -.01 -.00 .05

1.00 -.43⁎ .05 -.09+ -.06 -.04 -.05 .11⁎ .24⁎ .04 -.07 -.06 .04 .07 -.09+ -.08 -.11⁎ -.01 -.00 -.02 -.00 -.08 .06

1.00 -.02 .07 -.05 -.06 .08 -.01 -.16⁎ .02 -.03 -.11⁎ .04 -.05 .01 .07 -.07 .01 .03 -.08 -.13⁎ .01 -.09+

1.00 .25⁎ .24⁎ .17⁎ -.03 -.07 .01 .06 -.07 .08 .11⁎ .05 .02 -.01 .04 -.05 .04 -.06 .19⁎ .10+ -.09+

1.00 .15⁎ .07 .11⁎ -.09+ -.08 .08 -.02 .06 .04 .01 .02 .09+ .02 .00 .01 -.09+ .12⁎ .14 -.05

1.00 .08 -.06 .02 -.04 .01 -.05 .01 .04 -.03 .01 .03 -.01 -.04 .09+ -.05 .03 .04 -.03

1.00 -.08 -.14⁎ .00 .01 -.06 .07 .04 -.18⁎ -.00 -.09+ .15⁎ .02 .02 -.06 .44⁎ .14⁎ -.16⁎

1.00 -.24⁎ -.08 .03 .12⁎ -.02 .03 -.33⁎ .33⁎ -.06 .18⁎ .14⁎ -.10⁎ -.03 -.03 .30⁎ -.23⁎

1.00 .05 -.05 -.03 -.05 -.03 .13⁎ -.13⁎ .07 -.12⁎ -.09+ .14⁎ .07 -.14⁎ -.19⁎ .36⁎

1.00 .07 -.18⁎ -.25⁎ .30⁎ -.02 .00 .00 .00 .04 -.03 -.02 .13⁎ -.08 .04

1.00 -.16⁎ -.10⁎ .34⁎ .07 -.08 .04 -.00 -.10+ .05 -.00 .18⁎ -.06 .09+

14

15

16

17

18

19

20

21

22

24

25

26

1.00 .26⁎

1.00

14. Smartphone Brand: Others 15. Mobile Service Operator: B

23

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

63

Table 2 (continued) 14 16. Mobile Service Operator: C 17. Age 18. Gender: Female 19. Occupation: Blue collar 20. Occupation: None 21. Income: Below $3,000 22. Income: $3,000 to $5,000 23. Region: Capital 24. PB: N of apps 25. PF: N of sessions per app 26. PL: Time per session

+

-.09 -.12⁎ .05 -.06 .15⁎ .05 .04 -.04 .02 -.15⁎ .03

15

16

17

18

19

20

21

22

23

24

25

26

-.26⁎ -.04 -.02 -.09+ .17⁎ -.00 .06 -.08 .04 -.03 -.09+

1.00 -.02 .02 -.01 .00 -.01 .01 -.04 .14⁎ -.05 .03

1.00 -.42⁎ .15⁎ -.40⁎ -.20⁎ .04 .07 .03 .10⁎ .08

1.00 -.09+ .34⁎ .09+ -.04 -.01 .02 .02 -.11⁎

1.00 -.33⁎ .06 -.03 .04 -.02 .00 .00

1.00 .14⁎ .03 -.03 .06 -.10⁎ .01

1.00 -.56⁎ -.15⁎ -.02 -.03 -.03

1.00 .04 .00 .08 .06

1.00 -.06 -.02 .13⁎

1.00 .09+ -.11⁎

1.00 -.15⁎

1.00

Note: * and + indicate significance at p b .05 and p b .10, respectively.

Notably, non-shopping app browsing behavior does not differ much across the two groups of mobile buyers and non-buyers. In sum, the exploratory findings indicate that different factors affect different stages of mobile shopping (see Sismeiro and Bucklin 2004, for the same online shopping findings). Specifically, digital experience and the propensity to use non-shopping apps play important roles in the possession of shopping apps, and mobile purchasing is primarily influenced by digital experience and mobile browsing behavior for shopping apps. In

Table 3 Descriptive results. [A] Possession Number of shopping apps possessed Group size %age of assigned mobile users Digital experience Online Experience: High Online Experience: Med Mobile Experience: High Mobile Experience: Med Mobile browsing NB: N of non-shopping apps NF: N of sessions per non-shopping app NL: Time per sessions of non-shopping apps

None

One or more

51.03%

48.97%

.212 .399 .119 .379

.376 .429 .213 .403

30.848 38.084 152.140

41.018 30.235 157.013

[B] Purchase Number of shopping apps possessed

None

Number of shopping apps used for purchase

None

Group size %age of assigned mobile users Digital exposure Online Experience: High Online Experience: Med Mobile Experience: High Mobile Experience: Med Mobile browsing SB: N of shopping apps SF: N of sessions per shopping app SL: Time per sessions of shopping apps NB: N of non-shopping apps NF: N of sessions per non-shopping app NL: Time per sessions of non-shopping apps

One or more None

One or more

51.03%

27.84%

21.13%

– – – –

.319 .417 .185 .380

.451 .445 .250 .433

– – – – – –

.926 4.004 81.587 39.310 30.486 154.742

2.122 12.416 162.141 43.268 29.905 160.004

the following section, we test our hypotheses using the empirical models. Models and Findings Model of Possession and Purchase To examine factors motivating mobile users to possess shopping apps, we model the number of shopping apps that mobile user i has. The number of shopping apps possessed, y1,i, is at most 15 as we focus on the top 15 shopping apps. We thus assume that y1,i follows a right-truncated Poisson distribution with the truncation stipulated as y1,i ≤ 15 with the parameter λ1,i (see Lam, Chiang, and Parasuranman (2008) for more explanations). We then model the log-transformed λ1,i as a function of digital experience (i.e., online experience and mobile experience), mobile browsing of non-shopping apps, control variables, and a normally distributed error term. The error term includes factors such as exposure to shopping apps through firms' advertising, positive word-of-mouth, observing friends' usage, and so forth. The right-truncated Poisson model is as follows: " #−1 n 15 λk
 λ1;i1;i 1;i  ∑ ð1Þ P y1;i ¼ n1;i jy1;i ≤15 ¼ n1;i ! k¼0 k!
 log λ1;i ¼ β 1;1  OnlineExperience:Highi þ β 1;2  OnlineExperience:Mediumi þ β 1;3  MobileExperience:Highi þ β 1;4  MobileExperience:Mediumi þ β 1;5  NBi þ β 1;6  NF i þ β 1;7  NLi þ α1 þ γ1  Controlsi þ ε1;i

ð2Þ

where NBi, NFi, and NLi are the breadth, frequency, and length variables of non-shopping apps, and ε1,i is the error term that handles measurement errors and excess of zero observations in addition to unobserved factors discussed above (see Dong, Manchanda, and Chintagunta (2009) for a similar approach). Next, we model the purchase behavior of potential mobile buyers who have at least one shopping app. Note that the number of shopping apps used for purchase, y2,i, is at most the number of shopping apps possessed, y1,i. Thus, we assume y2,i follows a

64

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

right-truncated Poisson distribution with the truncation stipulated as y2,i ≤ y1,i with the parameter λ2,i. The log-transformed λ2,i is modeled as a function of the digital experience, mobile browsing, control variables, and an error term. The right-truncated Poisson model for the purchase stage is:
 P y2;i ¼ n2;i jy2;i ≤y1;i ¼

n λ2;i2;i

n2;i !

"

y1;i

λk2;i

k¼0

k!

 ∑

#−1 ð3Þ


 log λ2;i ¼ β 2;1  OnlineExperience:Highi þ β 2;2  OnlineExperience:Mediumi þ β 2;3  MobileExperience:Highi þ β 2;4  MobileExperience:Mediumi þ β 2;5  SBi þ β 2;6  SF i þ β 2;7  SLi þ β 2;8  NBi þ β 2;9  NF i þ β 2;10  NLi þ α2 þ γ2 Controlsi þ ε2;i ð4Þ As the shopping apps are in use, the mobile browsing variables of the shopping apps are added to the set of variables in Eq. (2). SBi, SFi, and SLi are the breadth, frequency, and length variables of the shopping apps. The error terms ε1,i and ε2,i are assumed to follow a bivariate normal distribution with mean 0, standard deviations σ1, and σ2, and correlation ρ, as the two stages of mobile shopping (i.e., possession and purchase) can be linked. Findings from Hypotheses Testing Table 4 presents the parameter estimates of both models.14 Digital Experience and App Possession (H1) We find that the more experienced online shoppers possess a greater number of mobile shopping apps (β1,1 = .947, p b .001; β1,2 = .622, p b .001), supporting H1(a). As consumers are more accustomed to an online shopping environment, they become more familiar with the mobile app versions of online shopping sites and thus the barrier to try mobile shopping decreases (e.g., Bang et al. 2013). On the other hand, mobile experience is significantly and positively related to the possession of mobile shopping apps (β1,3 = .604, p b .001; β1,4 = .355, p = .004), confirming H1(b). Our finding is consistent with that of online shopping (Emmanouilides and Hammond 2000) in that experienced users of mobile devices may be exposed to a greater number of apps and their benefits. They thus perceive lower risks and greater benefits from downloading apps, and in turn, possess a greater number of shopping apps. Non-shopping Apps and App Possession (H2) Mobile consumers browsing a greater number of non-shopping apps possess a greater number of shopping apps (β1,5 = .013, p b .001). When consumers are exposed to a variety of apps, the hurdle to downloading shopping apps goes down as they become more interested in acquiring shopping apps (Venkatesh and Agarwal 2006). However, heavy users of non-shopping apps 14

We estimate the models using PROC NLMIXED in SAS.

Table 4 Estimation results. Possession Estimate Digital Experience Online Experience: High Online Experience: Med Mobile Experience: High Mobile Experience: Med Mobile Browsing Shopping Apps SB: N of apps SF: N of sessions per app SL: Time per session Non-Shopping Apps NB: N of apps NF: N of sessions per app NL: Time per session Control Variables Intercept Smartphone Hardware LTE Supported: Yes Smartphone Brand: B Smartphone Brand: Others Mobile Service Operator: B Mobile Service Operator: C Socio-demographics Age Age Squared Gender: Female Occupation: Blue collar Occupation: None Income: Below $3,000 Income: $3,000 to $5,000 Region: Capital Browsing of Preloaded Apps PB: N of apps PF: N of sessions per app PL: Time per session Variances Σ ρ

.947⁎ .622⁎ .604⁎ .355⁎

− − −

SE .150 .142 .148 .121

− − −

Purchase Estimate

SE

.775⁎ .764⁎ .699⁎ .266

.343 .311 .264 .225

.280⁎ .036⁎ .001+

.085 .009 .000

.013⁎ −.006⁎ .001

.003 .003 .001

.005 −.006 .001

.005 .005 .002

−2.394⁎

.758

1.303

1.416

.231+ −.303+ −.607⁎ −.053 −.482⁎

.123 .161 .162 .125 .164

−.158 −.259 −.047 .057 .516+

.219 .293 .333 .219 .293

.045 −.001 .131 −.241 .111 .119 .145 −.026

.040 .001 .119 .155 .143 .144 .130 .111

−.137+ .002+ .056 .260 −.342 −.243 −.254 −.321

.074 .001 .217 .271 .242 .252 .231 .199

.004 .001 −.001

.010 .002 .001

−.036+ −.000 −.002

.019 .004 .002

.059 .375

.664⁎

.180

.886⁎ .045

Note: * and + indicate significance at p b .05 and p b .10, respectively.

possess a smaller number of shopping apps (β1,6 = −.006, p = .022). This is possibly due to a resource allocation issue (Chong 2013). Because of the limited attention span, users paying greater attention to using non-shopping apps are more likely to pay less attention to acquiring shopping apps. Interestingly, the time spent per session is not related to the app possession decision (β1,7 = .001, p = .421). This suggests that time per session does not reflect users' overall interest in mobile apps, unlike the numbers of non-shopping apps and sessions per app. Therefore, our findings in general are consistent with H2. Digital Experience and Mobile Purchase (H3) H3(a) is supported as users with high or medium online experience purchase from a greater number of shopping apps than users with low online experience (β2,1 = .775, p = .024; β2,2 = .764, p = .014). Users experienced in online purchasing have lower learning costs when mobile purchasing because they go through similar decision steps and can use information

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

65

they have saved online. The level of mobile experience is also positively correlated to the number of shopping apps users purchase from (β2,3 = .699, p = .008; β2,4 = .266, p = .237), which is consistent with H3(b). High experience with mobile devices lowers concerns (e.g., security and privacy issues) over mobile purchasing, which increases mobile purchases across shopping apps.

possession and mobile purchasing while they exert different degrees of influence across the stages of mobile shopping. Furthermore, the online and mobile channels share some characteristics; however, they also maintain unique characteristics. Our findings on mobile shopping cannot be inferred from the literature on online shopping and thus contribute to the theories of mobile shopping with new insights.

Shopping Apps and Mobile Purchase (H4) It is well known that the stages in the sales funnel are influenced by different factors (Sismeiro and Bucklin 2004; Venkatesh and Agarwal 2006). At the purchase stage, consumers decide to buy via a mobile app when product attributes and site characteristics match their needs. Thus, the browsing of apps unrelated to shopping has no effect (β2,8 = .005, p = .309; β2,9 = −.006, p = .248; β2,10 = .001, p = .699). Unsurprisingly, the browsing of shopping apps significantly relates to the mobile purchase decision. Specifically, users make purchases via a greater number of shopping apps when they use more shopping apps and use them more often and extensively (β2,5 = .280, p = .001; β2,6 = .036, p b .001; β2,7 = .001, p = .057), supporting H4. This supports that the positive relationship between shopping-related browsing and purchasing is the same across online and mobile shopping (e.g., Moe and Fader 2004).

Further Findings from Category-level Analyses

Control Variables Smartphone hardware and socio-demographic variables were selected according to prior studies as mentioned in Data and Measures. We find that all smartphone hardware variables significantly affect possession of shopping apps. Users possess a greater number of shopping apps if their smartphones operate on an LTE network. Since the LTE network is more recent technology than the 3G network, users adopting this technology may be tech-savvier, and thus possess a greater number of shopping apps. Moreover, the possession of shopping apps is salient when users own the brand of smartphone and subscribe to the mobile service operator with the dominant presence in the mobile industry. Surprisingly, while the control variables show significant effects on the possession stage, they have nearly no effect on the purchase stage. The measures for the browsing of preloaded apps are not significantly related to mobile shopping. As preloaded apps are provided by smartphone manufacturers or mobile service operators, it is difficult to infer mobile shopping behaviors from the browsing usage of preloaded apps. Summary We examine the drivers of two stages of mobile shopping, possession of shopping apps and purchase via shopping apps, analyzing the empirical data from multiple mobile retailers. Information regarding digital experience, non-shopping apps, smartphone hardware, and socio-demographics are collectively helpful for understanding the possession of shopping apps. At the purchase stage, digital experience and mobile browsing behavior for shopping apps help understand mobile purchasing. Of note, experience plays a key role in mobile shopping. Both digital experience and mobile browsing help understand app

As mentioned previously (see Data and Measures), shopping apps are classified into two categories: (1) online retailer apps from pure click retailers and (2) traditional retailer apps from offline-based retailers.15 This classification makes sense for two reasons. First, online retailer apps are mainly used to purchase products, but traditional retailer apps are used as showrooms for offline stores (e.g., eMarketer 2014a, b, c). Second, unlike traditional retailer apps, online retailer apps have similar displays, designs, and purchase options on their online sites. We can infer that users perceive lower barriers to trying online rather than traditional apps. Thus, relationships between the key variables (i.e., digital experience and mobile browsing) and the two stages of mobile shopping (i.e., app possession and mobile purchase) can be different across the app categories. Table 5 shows parameter estimates obtained from a categorywise bivariate truncated Poisson-lognormal regression model. Here, the mid-level digital experience leads to possession of online apps but not of traditional ones because users perceive lower barriers to downloading online apps. Moreover, the frequency of non-shopping app usage is not significantly associated with app possession of online apps. This may suggest that online apps do not compete with non-shopping apps for limited resources because the former is used to make purchases whereas the latter focuses on other benefits such as entertainment and information. Interestingly, online experience explains mobile purchasing via online retailer apps while having no effect on traditional retailer apps. It should also be noted that the effect of mobile experience on mobile purchases is the same across both types of retailer apps. Moreover, mobile browsing is influential for within-category purchasing but mostly insignificant for cross-category purchasing. This suggests that it may be worthwhile to define shopping categories at a granular level and decompose browsing data accordingly in order to examine the effect of mobile usage on mobile purchasing. Conclusion Rapid growth of the mobile marketplace offers users many opportunities to be mobile shoppers. However, little is known about the digital experience and mobile browsing, and how they relate to mobile shopping. Inferring the association of digital experience and mobile browsing with mobile shopping 15 We use an online app to refer to an online retailer app hereafter. A traditional retailer app and a traditional app are also used interchangeably.

66

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

Table 5 Category-level estimation results. Possession Onlinea

Purchase

Traditionalb Online Traditional

Digital Experience Online Experience: High 1.032⁎ .790⁎ Online Experience: Med .751⁎ .332 Mobile Experience: High .560⁎ .639⁎ Mobile Experience: Med .354⁎ .301 Mobile Browsing Online Retailer Apps SBon: N of apps − − SFon: N of sessions per app − − SLon: Time per session − − Traditional Retailer Apps SBtr: N of apps − − SFtr: N of sessions per app − − SLtr: Time per session − − Non-Shopping Apps NB: N of apps .012⁎ .017⁎ NF: N of sessions per app −.003 −.020⁎ NL: Time per session .001 .000 Control Variables Intercept −2.218⁎ −6.253⁎ Smartphone Hardware LTE Supported: Yes .382⁎ −.259 Smartphone Brand: B −.366⁎ −.044 Smartphone Brand: Others −.705⁎ −.336 Mobile Service Operator: B −.015 −.142 Mobile Service Operator: C −.440⁎ −.656⁎ Socio-demographics Age .026 .169⁎ −.001 −.002⁎ Age Squared Gender: Female .083 .337+ Occupation: Blue collar −.098 −.652⁎ Occupation: None .123 .063 Income: Below $3,000 .091 .216 Income: $3,000 to $5,000 .054 .383+ Region: Capital −.114 .170 Browsing of Preloaded Apps PB: N of apps .006 .001 PF: N of sessions per app .001 .003 PL: Time per session −.001 .001 Variances Σ .889⁎ ρ .045

1.003⁎ .923⁎ .554⁎ .331

−.074 −.040 1.312⁎ −.318

.673⁎ .031⁎ .000

−.639+ −.018 .001

−.236 −.014 .000

.605⁎ .108⁎ .002⁎

.012⁎ −.007 .002

−.006 .005 −.000

−.768

−1.342

−.087 −.048 .158 −.165 .151

−.638 −.647 −.778 −.276 .931

−.047 .001 −.034 .175 −.150 −.186 −.101 −.305

−.118 .002 .441 −.098 −1.171+ −.070 −.823 .410

−.044⁎ −.000 .001

.008 −.000 −.008

.442⁎

Notes: The label “online” refers to online retailer apps that include seven apps offered by pure click retailers. b The label “traditional” refers to eight traditional retailer apps from offlinebased retailers. * and + indicate significance at p b .05 and p b .10, respectively. a

from the corresponding relationship of online shopping might be misleading (Ghose, Goldfarb, and Han 2013). Thus, we examine this unexplored area of mobile shopping and offer new empirical insights into overall mobile shopping behaviors. First, we show that experienced online shoppers and smartphone users have a greater number of shopping apps on their phones. Second, the browsing breadth and frequency of non-shopping apps help us understand the number of shopping apps mobile users have on their phones — the browsing of preloaded apps, interestingly, is not diagnostic. Third, purchases made via shopping apps are understood by user digital

experience and the browsing information of shopping apps, with all other information being of little value. Fourth, digital experience is more sensitively associated with the possession of online retailer apps than traditional ones. Furthermore, the frequency of non-shopping app usage is not associated with possession of online apps. Finally, purchases via online retailer apps are explained by digital experience and browsing behavior of online apps whereas mobile experience and browsing behavior of traditional apps are most important for traditional retailers. Our key message can be succinctly stated as follows: Experience matters for mobile shopping. Experience includes app usage in the focal category as well as those across distinct categories, and digital experience across the online and mobile channels. Furthermore, experience exerts varying influence across the stages of mobile shopping, such as app possession and mobile purchasing. Our empirical findings offer new and interesting theoretical contributions as well as useful and effective practical implications. Theoretical Implications Our work contributes to the literature on mobile retailing in three ways. First, the empirical findings help us understand the mobile retailing industry as our research uses empirical data obtained through major mobile retailers that collectively account for much of the mobile log data. While previous studies use secondary data or data from a single retailer (e.g., Bang et al. 2013; Kleijnen, De Ruyter, and Wetzels 2007), we analyze empirical data generated by a mobile panel across major shopping apps. We also examine differences across the app categories (i.e., online versus traditional retailer apps). Second, our research furthers the understanding of mobile shopping not only by studying mobile purchasing but also by comparing the purchase stage with the earlier decision stage of possessing shopping apps. Consistent with some of the studies of online retailing (Sismeiro and Bucklin 2004; Venkatesh and Agarwal 2006), the decisions around app possession and mobile purchase are influenced by different factors. This means that different mobile strategies could be developed based on different mobile objectives, for example, increasing downloads or mobile revenues. Third, we study ways in which mobile retailing differs from online retailing. While heavy online browsing of non-shopping sites leads to online shopping, the effectiveness of heavy mobile browsing as a predictor of mobile shopping depends on whether the users browse non-shopping or preloaded apps. In this respect, our findings offer significant implications for research on mobile shopping. Practical Implications The emergence of mobile business has forced both online and offline retailers not only to tailor their website designs, interfaces, and product offerings to smartphones but to also develop and distribute their own exclusive shopping apps for mobile transactions (Bang et al. 2013). All these efforts serve to stay competitive with industry trends and create revenues via

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

mobile apps. Practitioners can use our findings to apply different targeting strategies across multiple shopping stages, and thus improve the effectiveness of marketing programs. The recent trend of app constellation or multi-app strategies shows that companies no longer rely on a single app but launch a collection of apps. For instance, Amazon provides both shopping apps (e.g., Amazon Shopping, Zappos) and non-shopping apps (e.g., Amazon Drive, Audible). Multi-app companies are likely to run multiple online sites collecting both online and mobile log data. Using internal log data, multi-app companies can measure users' digital experience and non-shopping app browsing behavior, which allows them to better target shopping-app downloaders. Companies of a single app however, cannot utilize such internal data. Accordingly, their strategy could be to collaborate with a third party in possession of such mobile browsing and digital experience data, such as Google, and target users using the third party's data. At the purchase stage, multi-app companies could generate mobile sales primarily by focusing their marketing efforts on smartphone users willing to purchase across a greater number of shopping apps. Such users are likely to be heavy browsers of shopping apps with high online shopping experience, and companies can identify such potential mobile buyers by analyzing their own data of shopping apps and shopping sites. Single-app companies can also make use of mobile log data collected within their own shopping apps and try to collect additional information of online shopping experience for example, upon sign-ups or using mobile popup surveys. Furthermore, our research provides useful guidelines for implementing different mobile strategies for online and traditional retailers. To increase the number of apps downloaded and installed, online retailers can target users having at least an average level of digital experience whereas traditional retailers should focus their marketing efforts on highly experienced digital users. At the mobile purchase stage, on the other hand, our findings suggest that online retailers should target users who shop online frequently and use online apps extensively while traditional retailers should target users who have used mobile phones longer and use traditional apps heavily. Mobile retailers agree that they need more mobile data and real-time analysis to enable timely action. Retailers can start with current data to predict mobile shopping patterns; however, these predictions can be improved substantially through meaningful data from a third party. Internal company mobile log data can help produce a solid forecast of mobile purchase opportunities, but the decision stage that precedes purchasing requires greater data. Data from a third party can complement internal data and help create a more comprehensive picture of overall mobile shopping habits. Limitations and Future Research Mobile retailing is the fastest growing retail sector in many countries. It is therefore vital that researchers and practitioners alike build new theories and analyses to understand mobile shopping and identify the mobile users most likely to become mobile shoppers. Our research attempts to contribute to this area

67

but has several limitations. The limitations suggest avenues for future work. First, it would be worthwhile to study more stages in mobile shopping to reflect consumer decision-making behavior more closely. Also additional factors such as firms' marketing efforts and apps' own characteristics will enrich theories of mobile behavior. Second, future research can examine what product categories are being purchased on mobile platforms and what drives these sales. Third, some apps in our empirical data offer non-frequently purchased products such as electronics, home appliances, and home furniture, and thus a longer data window could help better reflect mobile shopping behavior in non-frequently purchased categories or apps. Fourth, a longer data window could also offer an opportunity to examine user loyalty to shopping apps, also referred to as customer lifetime value (CLV). Fifth, the mobile panel consented to install the tracking app that monitors mobile usage and thus might care about privacy issues to a lesser degree. This could bring up the issue of sample representativeness and future research should utilize data collected though less intrusive ways. Sixth, mobile browsing precedes mobile purchases in our empirical setting and endogeneity is less likely to be an issue. Note, however, that unobserved characteristics such as shopping enjoyment may influence both possession and usage of shopping apps. Future research should present empirical findings after addressing potential endogeneity issues by using meaningful instrumental variables. Finally, cross-channel shopping among offline– online–mobile channels is also an area of increasing interest and relevance. We leave these potential areas of research to future endeavors. Acknowledgments The authors are grateful to Yongseob Kim and Sungeun Kwon for generously providing data. References Bang, Youngsok, Kunsoo Han, Animesh Animesh, and Minha Hwang (2013), “From Online to Mobile: Linking Consumers' Online Purchase Behaviors with Mobile Commerce Adoption,” PACIS 2013 Proceedings, paper 128. Bellman, Steven, Robert F. Potter, Shiree Treleaven-Hassard, Jennifer A. Robinson, and Duane Varan (2011), “The Effectiveness of Branded Mobile Phone Apps,” Journal of Interactive Marketing, 25, 4, 191–200. Brengman, Malaika, Maggie Geuens, Bert Weijters, Scott M. Smith, and William R. Swinyard (2005), “Segmenting Internet Shoppers Based on their Web-usage-related Lifestyle: A Cross-cultural Validation,” Journal of Business Research, 58, 1, 79–88. Bucklin, Randolph E. and Catarina Sismeiro (2003), “A Model of Web Site Browsing Behavior Estimated on Clickstream Data,” Journal of Marketing Research, 40, 3, 249–67. Choi, Jeonghye, David R. Bell, and Leonard M. Lodish (2012), “Traditional and IS-enabled Customer Acquisition on the Internet,” Management Science, 58, 4, 754–69. Chong, Alain Yee-Loong (2013), “Mobile Commerce Usage Activities: The Roles of Demographics and Motivation Variables,” Technological Forecasting and Social Change, 80, 7, 1350–9. comScore (2015), “comScore Reports December 2014 U.S. Smartphone Subscriber Market Share,” Retrieved February 9, 2015 from http://www. comscore.com/Insights/Market-Rankings/comScore-Report-December-2014US-Smartphone-Subscriber-Market-Share.

68

M. Kim et al. / Journal of Interactive Marketing 39 (2017) 55–68

De Mel, Suresh, David J. McKenzie, and Christopher Woodruff (2009), “Measuring Microenterprise Profits: Must We Ask How the Sausage is Made?” Journal of Development Economics, 88, 1, 19–31. Dong, Xiaojing, Puneet Manchanda, and Pradeep K. Chintagunta (2009), “Quantifying the Benefits of Individual-level Targeting in the Presence of Firm Strategic Behavior,” Journal of Marketing Research, 46, 2, 207–21. eMarketer (2014a), “The State of Omnichannel Retail: Retailers Playing Catch-up with Consumers,” (Retrieved January, 2014 from) http://www.emarketer.com/ public_media/docs/state_of_omnichannel_retail.pdf. ——— (2014b), “Total US Retail Sales Top $4.5 Trillion in 2013, Outpace GDP Growth,” (Retrieved April 10, 2014 from) http://www.emarketer.com/ Article/Total-US-Retail-Sales-Top-3645-Trillion-2013-Outpace-GDP-Growth/ 1010756. ——— (2014c), “Worldwide Smartphone Usage to Grow 45% in 2014,” (Retrieved June 11, 2014 from) http://www.emarketer.com/Article/ Worldwide-Smartphone-Usage-Grow-25-2014/1010920. Emmanouilides, Christos and Kathy Hammond (2000), “Internet Usage: Predictors of Active Users and Frequency of Use,” Journal of Interactive Marketing, 14, 2, 17–32. Forsythe, Sandra M. and Bo Shi (2003), “Consumer Patronage and Risk Perceptions in Internet Shopping,” Journal of Business Research, 56, 11, 867–75. Ghose, Anindya and Sang Pil Han (2011), “An Empirical Analysis of User Content Generation and Usage Behavior on the Mobile Internet,” Management Science, 57, 9, 1671–91. ———, Avi Goldfarb, and Sang Pil Han (2013), “How Is the Mobile Internet Different? Search Costs and Local Activities,” Information Systems Research, 24, 3, 613–31. Goh, Khim-Yong, Junhong Chu, and Jing Wu (2015), “Mobile Advertising: An Empirical Study of Temporal and Spatial Differences in Search Behavior and Advertising Response,” Journal of Interactive Marketing, 30, 34–45.

Hair Jr., Joseph F., William C. Black, Barry J. Babin, and Rolph E. Anderson (2006), Multivariate Data Analysis, 6th edition. New Jersey: Pearson Prentice Hall. Huang, Peng, Nicholas H. Lurie, and Sabyasachi Mitra (2009), “Searching for Experience on the Web: An Empirical Examination of Consumer Behavior for Search and Experience Goods,” Journal of Marketing, 73, 2, 55–69. Johnson, Eric J., Wendy W. Moe, Peter S. Fader, Steven Bellman, and Gerald L. Lohse (2004), “On the Depth and Dynamics of Online Search Behavior,” Management Science, 50, 3, 299–308. Kleijnen, Mirella, Ko De Ruyter, and Martin Wetzels (2007), “An Assessment of Value Creation in Mobile Service Delivery and the Moderating Role of Time Consciousness,” Journal of Retailing, 83, 1, 33–46. Konus, Umut, Peter C. Verhoef, and Scott A. Neslin (2008), “Multichannel Shopper Segments and Their Covariates,” Journal of Retailing, 84, 4, 398–413. Koyuncu, Cuneyt and Donald Lien (2003), “E-Commerce and Consumer's Purchasing Behavior,” Applied Economics, 35, 6, 721–6. Kraut, Robert, Tridas Mukhopadhyay, Janusz Szczypula, Sara Kiesler, and Bill Scherlis (1999), “Information and Communication: Alternative Uses of the Internet in Households,” Information Systems Research, 10, 4, 287–303. Lam, Shun Yin, Jeongwen Chiang, and A. Parasuranman (2008), “The Effects of the Dimensions of Technology Readiness on Technology Acceptance: An Empirical Analysis,” Journal of Interactive Marketing, 22, 4, 19–39. Moe, Wendy W. and Peter S. Fader (2004), “Capturing Evolving Visit Behavior in Clickstream Data,” Journal of Interactive Marketing, 18, 1, 5–19. Sismeiro, Catarina and Randolph E. Bucklin (2004), “Modeling Purchase at an E-commerce Web Site: A Task-completion Approach,” Journal of Marketing Research, 41, 3, 306–23. Venkatesh, Viswanath and Ritu Agarwal (2006), “Turning Visitors into Customers: A Usability-centric Perspective on Purchase Behavior in Electronic Channels,” Management Science, 52, 3, 367–82.