Service-Oriented Factors Affecting the Adoption of Smartphones

This research investigates the adoption factors of smartphones focusing on the differences of smartphone and feature phone users. We used Technology Acceptance Model (TAM) which incorporates service-oriented and device-oriented functional attributes as exogenous variables for a product-service system such as smartphones. In addition, Decision Tree (DT) and customer surveys were conducted. As a study results, we found that the service-oriented functional attributes ‘wireless internet’ and ‘mobile applications’ affect the adoption of smartphones regardless of users. However, the DT results revealed that the more important factor is ‘mobile applications’ to smartphone users but ‘wireless internet’ for feature phone users. In conclusion, we discovered that a strategy emphasis on the service-oriented attributes is needed for the adoption of smartphones.


Introduction
The market for cellular phones (now more commonly known as mobile phones, or just 'mobiles') experienced rapid growth until 2002 when it can be considered to have reached maturity.Since then, the market focus has been on phone replacement rather than on first-time purchase, and the profiles of buyers -and of their tastes and needs -have changed to a remarkable extent.A new generation of mobiles -generally labeled smartphones -has been released which have excited growing general consumer interest.It is therefore crucial for sellers to understand the diffusion process of these increasingly popular products and to understand what next steps are required to market smartphones to both in current users and potential new customers.
In general, a smartphone is defined as an advanced cellular handset that provides traditional mobile phone features but also such PDA features as Internet access and portable PC functions.Despite the important changeover in the mobile market from earlier, simpler phones (we refer to them as 'feature phones') to smartphones, few studies have addressed issues related to the evolution of the mobile market in terms of user behaviors.At this growth stage, the characteristics of smartphone users will have substantially from those of feature phone users, with the majority of the former being early adopters or early majorities, willing to try products early in their diffusion process, with feature phone users (themselves potential future smartphone buyers) being more cautious about adopting new products -often older and more set in their ways than the early majorities.We argue that the two groups have distinguishing characteristics, and must be separated to analyze the smartphone diffusion process and the factors affecting that process.
Given that the mobile market has evolved rapidly, we aim to analyze the influence of various factors in customers' decisions to buy smartphone.Technology Acceptance Model (TAM) was used in this research to identify and compare the factors affecting the adoption of smartphones by smartphone and feature phone users.For the purpose of analysis, the primary attributes for applying TAM were derived from a survey of 100 mobile phone users, after which 11 hypotheses were established in the model and verified using the structural equation model (SEM) for two groups; smartphone and feature phone users.Finally, the two groups' actual decision-making processes were analyzed using DT.The analysis results show that the smartphone's service-oriented functional attributes significantly affect its adoption by increasing 'the perceived ease of use' and 'the perceived utility' by both groups, though the degree of influence is not equal, signifying that the mobile market has been moving from a 'product-oriented' to a 'service-oriented' market.We expect our research findings to provide valuable information for understanding how the mobile market has evolved and what values the customer wants as the market evolves.

Acceptance factors
The paper consists as follows.The theoretical and methodological background of this research is explained in the literature review.Then, we describe the overall research process and establish our hypotheses.Next, we summarize and discuss the empirical analysis results of SEM and DT in measure, and conclude with notes about our contributions and future research directions.

Literature review
A smartphone can be regarded as a handheld computer integrated with a mobile phone that allows users to run multitask applications: as such, is thus attractive to a wide range of user groups.Smartphones were initially used mainly for specialized purposes such as delivery services and medical services but began to be used as a general-purpose mobile phone when general customer needs were considered and functions such as multimedia and games were included.Leung and Wei (2000) found that motivations for mobile phone use included 'fashion and position', 'emotion and sociality', 'relaxation', 'mobility', 'instantaneity', and 'relief from work'.However, since the characteristics of smartphones differ greatly from those of feature phones (in incorporating more technologies and providing more functionality via wireless Internet) the factors affecting smartphone adoption are likely to differ from those influencing adoption of feature phone.Table 1 groups the factors that previous smartphone studies have found affected their acceptance into three categories.This research focuses on smartphones' functional attributes to investigate if their perceived value differ as factors affecting their adoption by smartphone and by feature phone users.
Figure 1.TAM model for an integrated product-service system

Conceptual model for smartphone acceptance
The Technology Acceptance Model (TAM) was proposed by Davis (1989) specifically to explain and predict consumer acceptance of an information systems and technologies.Especially, it has been frequently used to provide the theoretical foundation for research into mobile commerce (m-commerce) issues (Choi et al., 2008;Lu, Wang and Yu, 2007).A theoretical TAM model assumes that the decision to adopt a particular technology is determined by two key factors: PU and PEOU (see Figure 1).Here, PU is defined as the degree to which a person believes that using a particular technology will enhance their job performance or their life, while PEOU is defined as the degree to which they believe using the technology reasonably easy for them to manage.BI is defined as a measure of an individual's intention to perform a specific behavior and relates to actual behavior, the determinant factor in taking specific action in the future (Ajzen, and Fishbein, 1980).However, most subsequent studies that have used TAM have omitted the attitude variable (Steven, 2003), as have we in this study.This research also includes functional attributes that influence PU and PEOU in the model as exogenous variables.Smartphones can be considered as typical integrated product-service systems, so their functional attributes can be divided into the two categories of service-oriented and device-oriented attributes.Figure 1 describes a TAM for a product-service integrated system, which we suggest in this research.

PLS (Partial Least Square) analysis
An SEM is a technique for analysing causal relations using a combination of statistical data and assumptions about the relations.In estimating its parameters, there are two approaches available: (1) covariance-based and (2) variancebased (Haenlein and Kaplan, 2004).Though the covariancebased approach, which attempts to minimize the difference

Research framework -The overall process
Figure 2 outlines the overall research process.
First, 29 smartphone attributes, identified from literature reviews, are used to develop a research model.Five attributes that are considered the most important in smartphone purchase or replacement decisions are selected via first survey and a TAM with 12 hypotheses is established based on these five.Then a second survey is used to test these hypotheses for smartphone and for feature phone users.Based on the same survey results, DT is used to analyze how users make these decisions.Finally, the analysis results for the two groups are compared to identify common and distinguishing features in their adoption process of smartphones (Basic statistical and DT analysis for this research were conducted using SPSS 19.0, and the TAM was analyzed by SEM, using smartPLS.)

Functional attributes of smartphones
A product consists of functional attributes that meet personal needs (Ferber, 1974): their value will vary according to customer, so some will affect customers' purchasing behavior more than others (Green and Srinivasan, 1990).These functional attributes are critical to the product meeting customer needs and satisfaction, so it is important that they are identified -which has been the focus of previous research.We identified 29 functional attributes of these two technologies from previous studies and industry reports on the mobile phone industry.Then the first survey was designed to identify the factors that most influenced smartphone adoption.For the purpose of this survey, 32-item questionnaires with 5-point Likert scale answers were distributed to individuals in their 20s and 30s: this age-group was judged to be suitable target respondents for this research as they use mobile phones most and are at the center of the shift from feature phones to smartphones in the Korean mobile marketplace.100 responses were collected over a month from the 19 September, 2010, from 70 men and 30 women; 51% of them used feature phones and 49% smartphones.The survey results showed the five most influential attributes were the service-related attributes of 'application' and 'wireless Internet' and the device-related attributes of 'design', 'multimedia' and 'after-service'.between the sample covariance and those predicted by the theoretical model (Chin and Newsted, 1999), has been popular so far, this research adopts the variance-based one, that is, PLS.
The PLS focuses on maximizing the variance of the dependent variables explained by the independent ones instead of reproducing the empirical covariance matrix and thus has several advantages over the co-variance-based approach.First, it requires no assumptions about the population or scale of measurements (Fornell and Bookstein, 1982) and accordingly can be applied without distributional assumptions and with nominal, ordinal and interval scaled variables.Second, the SEM needs a relatively small number of samples compared to the covariance-based approach (Barclay et al., 1995).Since only about 130 samples were used for analysis and the data collected on an interval scale did not fit the normal distribution, the PLS SEM approach was adopted in this research.

Decision tree
DT, one of the most popular data-mining techniques for knowledge discovery, can analyse the information contained in an abundant data source systematically to extract valuable rules and relationships.DT is widely applied in various areas such as analysing customers' decision-making pattern, marketing and market research, and quality management (Quinlan, 1993), usually for classification and prediction purposes (Ganti et al., 1999), as it creates a model that predicts the value of a target variable based on several input variables.The model takes the form of a top-down tree structure in which decisions are made at each node that correspond to one of the input variables.Each leaf (the tree's terminal nodes) represents a value of the target variable based on the values of the input variables as determined by the path from the root to the leaf.If the target variable has a continuous value, a regression tree is developed; on the other hand, if the variable has a discrete value, a classification tree is developed.A tree can be studied by splitting the source set into subsets based on their attributes, and this process can be repeated on each derived subset in until the subset at a node has the same value as the target variable, or when further splitting no longer adds value to the predictions.We selected DT as an appropriate technique in this research for dealing with survey data that had been collected using 7-point Likert scales, as well as using it directly to analyse customers' decision-making processes in the adopting smartphones.Next, multimedia refers to various media functions included in a smartphone's functionality, such as camera, mp3 player, and games.Chang et al. (2009) notes that the integration of so many functions into one mobile device means that multimedia attribute has now become a core smartphone function, diminishing time and space constraints and supporting ease of use.So we can suggest: -H4-1: Multimedia positively influences PU of smartphones.
Finally, the survey results identified after-sales service as a feature that improves product sales and Jan and Hsiao (2004) list various of its elements that increase customer satisfaction, including complaint management and onsite service, which increase convenience and ease of use for customers and thus contribute to smartphone acceptance.We can therefore suggest: -H5-1: After-service positively influences PU of smartphones.
PU can be defined as the degree of users' belief that using the focal technology will help improve their job performance and/or quality of life, and in terms of their productivity in achieving their aims or tasks (Davis, et. al, 1989).PEOU indicates how easy users believe the technology will be to use, as determined by the degree of physical and mental effort in learning how to use it as well as its actual use.Adams et al. (1992) claim that PEOU significantly affects BI, and Igbaria, et. al. (1997) found this effect to be more significant than that of PU on BI.Most existing studies confirm PEOU as a precedent of PU, which means that users prefer to use technologies that are easy to use and which that those technologies help to improve the user's performance (Taylor and Todd, 1995).Therefore, the following hypotheses are developed.
-H6-1: The PEOU of smartphone services positively influence the BI of smartphones.
-H6-2: The PEOU of smartphone services positively influence the PU of smartphones.
-H6-3: The PEOU of smartphone services positively influence the users' acceptance of smartphones.

Measures for smartphone acceptance
We conducted a second survey through an extensive online and offline survey to test our hypotheses: again targeting mobile phone users in their 20s and 30s with a questionnaire developed using a 7-point Likert answer scale, and collecting data over two months from 25 October 2010.Of the 284 responses collected, only 262 were suitable for analysis -206 from men and 56 from women.Of the total 130 used feature phones and 132 smartphones (see Table 2

Research model and hypotheses
We adopted these five primary functional attributes as independent variables in our TAM were assumed to affect the BI of smartphones through PU and PEOU (as shown in Figure 3).Then, based on this model, we developed 12 hypotheses.
In the first -service functionality-group, we can define, wireless Internet as a service that provides digitised information or content to users via a wireless connection, which thus eliminates time or spatial constraints, and so can be expected to affect PU positively (Ng-Kruelle et al., 2002).At the same time, because wireless Internet can be the basis of many other smartphone applications, it can also be expected to influence PEOU positively: Sarker and Wells (2003) found that wireless Internet greatly affected mobile service acceptance.Therefore we can propose that: The other service attribute concerns applications, i.e. software elements that can be executed on a mobile device to deliver content to a user, and Armstrong et al. (2010) report that they have recently been emphasised as being valuable tools for meeting customer needs and increasing their interactions with others.An increasing range of such smartphone applications are being developed to meet various needs and utilities in a convenient manner, and users can purchase and download them on-line from virtual 'App Stores' We expect their availability to affect both PU and PEOU positively, in other words: -H2-1: Application availability positively influences PU of smartphones.-H2-2: Application availability positively influences PEOU of smartphones.
In the second -device oriented -group, smartphone design includes elements such as include the phone's physical layout (shape and size), colour, overall attractiveness of appearance, keypad design, etc.These design elements are expected to affect both PU and PEOU positively.Previous studies have revealed the extent to which mobile phones are regarded as fashion items via which users expressing themselves (Leung and Wei, 2000), implying the significant impact design can have on smartphone adoption, so that: -H3-1: Attractive design positively influences the PU of smartphones.-H3-2: Attractive design positively influences the PEOU of smartphones.

Validity and reliability test for measures
Before verifying the hypotheses, we tested the validity and reliability of the measurements.First, we used Harman's onefactor test for detecting common method bias (Podsakoff and Organ 1986).Eight factors with eigenvalues exceeding 1 were extracted, cumulatively explaining about 75% of the variance, and no single major factor emerged, suggesting that common method bias in not a problem (see Appendix 2).
We conducted confirmatory factor analysis on all multi-item variables to check whether the items measured what they were intended to measure.The analysis results showed that all items except one for feature phone users, which was then removed from our model, loaded appropriately onto eight variables -confirming their reliability and convergent validity (see Appendix 3).The average variance extracted (AVE) values, composite reliability (CR) values, and Cronbach's alpha coefficient values were all over recommended values, 0.5 (Gefen et al., 2000), 0.7 (Werts et al., 1974), and0.6 (Nunnally, 1967) respectively.We also conducted cross-loading analysis and square root AVE analysis, which verified the external validity of our models (see Appendix 4).

Analysis and results -Model validity
The model's validity was tested prior to testing the hypotheses.The PLS structural model is mainly evaluated by Rsquare of endogenous latent variable and redundancy index (obtained from the Stone-Geiser Q-square test) for predictive relevance.The average R-square of endogenous latent for detail.).The potential for non-response bias was investigated by comparing answers from respondents and nonrespondents whom we contacted personally.Also, the online and offline answers were compared to test if there is any difference between the two channels.As a result, these biases were not considered to be a problem (see Appendix 1).

Operational definition and measure of variables
The variables' operational definitions and measurements items were based on previous studies, and measured smartphone users' experience of their devices, and feature phone users' perceptions of smartphones, for each of the five primary functional attributes: -Wireless Internet: the quality of the smartphones' telecommunication services -is measured by four items: speed, range, continuity and quickness.
-Design: users' perceptions of satisfaction with smartphone design -is measured by three items: overall design, color and style.
-Multimedia: the superiority of functionality for playing music or video -consists of five subordinates: compatibility, management, UI (user interface), continuity, and clarity.
-Application: the level of users' satisfaction with application use -is measured by three items: UI, diversity, and quality.
Finally, After-service -which concerns the quality and terms of guarantee of such services -consists of three measurement items: quickness, term, and friendliness.nodes.The DT analysis results confirmed the results of TAM, revealing that the most critical factor in customers' choice to adopt smartphones is 'applications', followed by 'design': we used these factors to develop the following decisionmaking rules for smartphone adoption.
-Where customers' satisfaction with both the applications and design of a smartphone are high, their intention to adopt is very high (90.5%).
-Where customers' satisfaction with smartphone design is high, but the satisfaction with applications is low, their intention to adopt will be high (86.7%). - Where customers' satisfaction with the applications available on a smartphone is high, but their satisfaction with its design is low, their intention to adopt will be medium (73.6%).
-Where customers' satisfaction levels about both these smartphone attributes are low, their intention to adopt will be relatively low (45.5%).

Analysis results for feature phone users
The SEM analysis results for feature phone users are summarised in Table 5.First, 'wireless Internet' and 'applications' positively affect both PU (H1-1 and H1-2 are supported) and PEOU (so H2-1 and H2-2 are supported).Unlike the case of smartphone users, the 'wireless Internet' path coefficient has the greatest value.'Design' has no significant effect on either PEOU or PU (H3-1 and H3-2 are rejected), and neither 'multimedia' nor 'after-service' have any significant impact on PEOU (so H4-1 and H5-1 are also rejected).Finally, PU has a positive effect on BI (supporting H6-1) and PEOU on PU (supporting H6-2) as expected, while PEOU's influence on PU is insignificant (so, again, H6-3 rejected).
variables represents an index for validating the PLS model globally, while the redundancy index measure the predictive relevance of the model by reproducing the observed values by the model itself and its parameter estimates.Table 3 presents the fitness test results: the model's validity can be said to be verified, since all index values for the two groups are at the recommended levels.

Analysis results for smartphone users
The SEM analysis results for smartphone users are summarized in Table 4.They show that the service-oriented functional attributes -'wireless Internet' and 'applications' -have a significant impact on both PU (H1-1 and H2-1 are supported) and PEOU (H1-2 and H2-2 are supported).The 'application' path coefficient has the highest value of all five functional attributes, revealing that this factor has the greatest impact on the choice to use of smartphones.We can find no significant evidence that 'design' has a positive effect on either PU or PEOU (H3-1 and H3-2 are rejected), 'multimedia' positively affects PU and interestingly 'after-service' on PEOU (H4-1 and H5-1 are rejected).Finally, both PU and PEOU are shown to affect BI positively and, as expected, PEOU has a positive effect on PU (H6-1, H6-2 and H6-3 are supported).
DT analysis was then used to investigate detailed decisionmaking process (see Figure 4), when the most representative items for each of the five attributes were selected as independent variables, and BI as the dependent variable.The CHAID algorithm was applied to determine the optimal tree size, and the maximum numbers of parent and children nodes were set to 20 and 10, respectively.Chi-square statistics at the significance level of 10% was used to split the First, 'wireless Internet' has the greater effect on PEOU for feature phone users than for smartphone users, while 'application' is the opposite case, implying that feature phone users believe the wireless Internet function itself (which distinguishes it from the feature phone) adds the most value to smartphones.An interesting finding is that most of the device-oriented factors (including 'after-service') do not have the expected influence on PEOU and PU.It seems that mobile phones' main utility is based on services available, probably, on the Internet -most other device-based attributes are seen as auxiliaries, or as only increasing smartphones' usability.Even 'design' appears not to be a critical factor in customer's purchasing decisions on the contrary to the previous studies.The notion that there are not many options for smartphone designs at this stage perhaps decreases effect on PU or PEOU.
Second, in terms of the relationships between PU, PEOU and BI, PU affects BI directly and PEOU affects BI indirectly via affecting PU.As smartphones are more complex systems than feature phones -so there is more to go wrong and The DT analysis results for feature phone users are shown in Figure 5.The most attractive factor that determines the adoption of smartphones is 'wireless Internet', based on which the following decision-making rules are developed.
-Where the customer's expectations of smartphones' 'wireless Internet' attribute is high, their intention to adopt is relatively high (59.3%).
-If the expectation of 'wireless Internet' on a smartphone is low or average, their intention to adopt is relatively low (37.9%).

Discussion -Comparison analysis
The comparison of the two groups' SEM results shows that the same factors the mobile service-oriented functional attributes -influence both groups of phone users.Because the smartphone provides the wireless Internet function a variety of services have become available, driving the growth of the smartphone market.While both user groups value these mobile service-oriented attributes highly, there are still some differences.ferently.For this analysis, we suggest a TAM incorporating service-oriented and device-oriented functional attributes as influencing factors, which is appropriate for a productservice integrated IT system such as smartphones.Korean mobile phones users in their 20s and 30s were surveyed, after which SEM and DT were used to analyse the survey data.The research findings indicate that, among the various functional attributes of smartphones, the service-oriented factors affect PU and PEOU most significantly, though the effect of these attributes differed for smartphone and feature phone users.On the contrary, the device-oriented functional attributes does not affect PU or PEOU.Both PEOU and PU have a positive a positive influence on BI in the case of smartphone users, while PU affects BI directly but PEOU affects BI indirectly through PU.This study is one of the earliest attempts to investigate the changing factors affecting the adoption of smartphones.Our research results are expected to reflect the characteristics of the market with regard to smartphones and customers who have, and have not yet, adopted them, and so can help understand the evolution of mobile market, and further develop customised R&D and marketing strategies.
Despite these contributions, this paper has two limitations, which point to future research.First, data collection was limited to the Seoul and Kyungki areas of Korea and the ratio of smartphone to feature phone users in our data differs from the overall ratio in the Korean mobile phone market, so our findings cannot be generalised to the wider population.Future research should use more detailed data collection processes, and the hierarchical random sampling method.Second, in-depth analysis of the SEM and DT results is needed.Why some attributes affect PEOU and PU positively or negatively and why some do not should be examined in detail with supporting data and relevant literature; the differences between smartphone and feature phone users should be more fully investigated.

Acknowledgement
This work was supported by the National Research Foundation of Korea Grant funded by the Korean Government (NRF-2011-327-B00189).
PEOU has a positive influence on PU.However, investigating by sector shows that the direct impact of PEOU on BI is significant only for smartphone users, which means that customers may not realize the importance of ease of use until they adopt and use smartphones.
Finally, the DT results reveal that, for smartphone uses, who may be early adopters or early majorities in its life cycle, 'design' as well as 'applications' affect purchasing decisions, and thus determine the patterns of smartphone adoptions.On the other hand, feature phone users decisions to buy smartphones are fuelled by their expectation of using of 'wireless Internet' (a significant distinguishing feature of their practicality compared to feature phones).

Managerial implications
Because the factors affecting the adoption of smartphones differ according to user groups, mobile phone manufacturers and retailers need to develop different strategies to increase smartphone diffusion rate among the two groups.First, product development/marketing strategies that emphasises their wireless Internet function will encourage the adoption of smartphones by those currently using feature phones.
Second, development and marketing applications can help smartphone owners in their use.As the applications factor has the greatest impact on smartphones adoption, user satisfaction about applications will increase users overall satisfaction with their smartphones.A policy or strategy to open the application market should be developed so as to increase their size, quality and applicability.
Third, continuous improvement in smartphone design is also necessary, as DT analysis indicate it has a critical impact on the adoption of smartphones by smartphone users.As many current smartphone users are early adopters and early majority, upgrading product design along with its functionalities will attract such users to adopt the next generation of phones.
Finally, with the changes in the mobile phone market, the market is moving more towards service-oriented business.The co-evolution of product and service attributes should be the focus of attention.And more effort is required to identify service-oriented attributes that affect the diffusion of smartphones.

Conclusion
This study aims to investigate the functional factors that affect the adoption of smartphones by focusing on mobile phone consumers at large, and to identify how these factors influence smartphone and feature phone users dif-

Figure 2 .
Figure 2. The overall analysis process

Figure 4 .
Figure 4. DT results for smartphone users Figure 4. DT results for smartphone users

Table 1 .
Factors affecting smartphone acceptance

Table 2 .
Demographics of second survey respondents

Table 3 .
The results of the fitness for the research models

Table 4 .
Path coefficient analysis results for smartphone users

Table A -
1. Differences between different types of respondents (Smartphone users)

Table A -
4a. Cross loading results for feature phone users Note 1.For all but one (denoted by *) items, the factor loadings on their respective latent variable exceed 0.7 with weak cross factor loadings to other unrelated variables.

Table A -
4b. Cross loading results for smart phone users Note 1.For all items, the factor loadings on their respective latent variable exceed 0.7 with weak cross factor loadings to other unrelated variables.

Table A -
5a. Discriminant validity analysis for feature phone users Note 1.The values in the parenthesis are square root AVEs.Note 2. For every latent variable, its square root AVE value is greater than its correlation coefficient values with other variables, thus verifying discriminant validity.

Table A -
5b. Discriminant validity analysis results for smart phone users Note 1.The values in the parenthesis are square root AVEs.Note 2. For every latent variable, its square root AVE value is greater than its correlation coefficient values with other variables, thus verifying discriminant validity.