Description
Consumer behaviour is the study of individuals, groups, or organizations and the processes they use to select, secure, and dispose of products, services, experiences, or ideas to satisfy needs and the impacts that these processes have on the consumer and society
Effects of apps on consumer behavior of smartphones and telecommunication providers: Feature fatigue vs. mass customization
by D.J.
Agterhuis
Student identity number: 0722219
In partial fulfillment of the requirements for the degree of Master of Science in Innovation Management
First supervisor: Second supervisor: Company Supervisor: Publishing date: University: Faculty: Department:
dr. ing. J.P.M. Wouters dr. J.J.L. Schepers S. Balkenende 9th of August, 2012 Eindhoven University of Technology Industrial Engineering & Innovation Sciences Innovation, Technology, Entrepreneurship and Marketing (ITEM)
Series Master Theses Innovation Management
Subject headings:
Apps, Smartphones, Feature Fatigue, Mass Customization, Customer Satisfaction, Technology Acceptance, Consumer Behavior, Telecommunication
2
Abstract
The use of apps on smartphones raises questions about how apps affect usage and satisfaction with the device and with telecom operators, and which of two concepts – feature fatigue or mass customization – is applicable to model these effects. The number of apps installed on the smartphone is proposed to predict three aspects of value: usefulness, usability, and effort. In turn, this value in turn is proposed to predict usage of both the device itself and of services provided by the operating company, and satisfaction with the device and the operator. The model is tested using survey results of 252 customers of a Dutch telecommunication provider that are combined with objective usage data. Results indicate that mass customization applies to the use of apps on smartphones since the number of apps positively relates to usability. In turn, usability is related to utility, which affects mobile data usage through the daily frequency of use of the smartphone. Satisfaction with the smartphone is directly affected by the number of apps and through value. Implications are that a high number of apps is not related to usability issues and that apps facilitate the transition in the telecom world of voice and SMS services to mobile data.
3
Acknowledgements
Naturally, I would like to thank my first supervisor Joost Wouters of Eindhoven University of Technology for guiding the search towards a suitable topic for graduation, for steering the project into the right direction and for his thoughts on the theories, which were a useful input. I want to thank my second supervisor Jeroen Schepers for his creative contribution to the model, for answering my long e-mails and for his support in the statistical analysis. Much appreciation is also attributed to Sanne, Caspar, Frederieke, Jan Z., Joppe, Michel, Nathalie and Unni for the fantastic and educating experience in the past six months at the “large Dutch telecom company.” In particular, I would like to thank Sanne for her contributions to the research. Also the funding of the market research and the access to the customer database of the company deserves recognition. Last but not least, much is obliged to Esmee, my family, “the sponsors down under”, my friends and Douwe Egberts for the (mental) support and the mere reaching of the graduation thesis, and for the creative discussions which definitely shaped it. Also many thanks go out to Tim and Jan A. for their detailed reviews. I hope you enjoy reading this report, and that you look differently at the palm of your hand after reading it. Dirk
4
Contents
SUMMARY 1 INTRODUCTION 2 THEORY AND HYPOTHESIS DEVELOPMENT 2.1 THEORY AND RECENT RESEARCH 2.1.1 APPS AND SMARTPHONES: MODULES FOR MODULAR PRODUCTS 2.1.2 TOO MANY FEATURES: FEATURE FATIGUE 2.1.3 MASS CUSTOMIZATION: ADAPTING FUNCTIONALITY TO MEET SPECIFIC NEEDS 2.1.4 CUSTOMER SATISFACTION 2.1.5 TECHNOLOGY ACCEPTANCE MODEL (TAM) 2.2 CONCEPTUAL FRAMEWORK AND HYPOTHESES 2.2.1 NUMBER OF APPS ON PERCEIVED VALUE AND CUSTOMER SATISFACTION 2.2.2 PERCEIVED EASE OF USE ON PERCEIVED USEFULNESS 2.2.3 ASPECTS OF VALUE ON USAGE OF THE SMARTPHONE, SERVICES PROVIDED
BY THE OPERATOR AND CUSTOMER SATISFACTION
7 10 11 12 12 12 13 14 14 15 17 19 19 20 21 21 22 22 23 25 25 27 27 28 28 29 29 30 30 33
2.2.4 USAGE OF THE SMARTPHONE ON USAGE OF TELECOMMUNICATION
PROVIDER SERVICES
2.2.5 CUSTOMER SATISFACTION OF THE SMARTPHONE ON CUSTOMER
SATISFACTION OF THE TELECOM PROVIDER
2.2.6 MODERATING VARIABLES 3 METHOD 3.1 3.2 3.3 3.4 DATA SAMPLE MEASUREMENT INSTRUMENT DATA COLLECTION PROCEDURE ANALYTICAL METHOD
4 RESULTS 4.1 SAMPLE AND DATA CHARACTERISTICS 4.1.2 COMMON METHOD BIAS 4.1.3 MISSING DATA, OUTLIERS AND MULTIVARIATE ASSUMPTIONS 4.2 MEASUREMENT MODEL 4.2.1 CONSTRUCT RELIABILITY 4.2.2 CONSTRUCT VALIDITY AND MULTICOLLINEARITY 4.3 STRUCTURAL MODEL 4.3.2 MODERATING VARIABLES
5
4.3.3 CONTROL VARIABLES 4.4 POST-HOC ANALYSES 4.4.2 DIFFERENCES BETWEEN OPERATING SYSTEMS 5 DISCUSSION 5.1 5.2 5.3 5.4 FINDINGS SCHOLARLY IMPLICATIONS MANAGERIAL IMPLICATIONS LIMITATIONS AND AVENUES FOR FURTHER RESEARCH
34 34 36 37 38 41 41 43 45 45 49 50 51 56 58 59
6 REFERENCES 6.1 LITERATURE REFERENCES 6.2 WEB REFERENCES APPENDIX A: MEASUREMENT INSTRUMENT APPENDIX B: QUESTIONNAIRE APPENDIX C: DATA MANIPULATION APPENDIX D: GENERALIZATION OF THE SAMPLE APPENDIX E: DATA EXPLORATION: OUTLIERS AND NORMALITY APPENDIX F: FACTOR LOADINGS AND CROSS LOADINGS OF THE FIRST CFA APPENDIX G: FACTOR LOADINGS AND CROSS LOADINGS OF THE DEFINITE CFA APPENDIX H: STATISTICAL VALIDITY OF THE BASE MODEL APPENDIX I: POST-HOC MEDIATION ANALYSIS APPENDIX J: MANOVA POST-HOC ANALYSIS
61
62 63 64 65
6
Summary
The revolution of apps and smartphones demands insights in how these apps affect customer satisfaction and usage behavior of smartphone consumers with respect to the device itself and usage of services offered by telecommunication providers such as voice, SMS and mobile data. Two concepts compete in modeling these effects: feature fatigue and mass customization. Apps are assumed to be features since the use of apps offers functionality in addition to the hard technical specifications of smartphones. Consumers suffering from feature fatigue have purchased a feature-packed product which is high in utility but low in usability because of that high number of features (Thompson et al., 2005). Because consumers can modulate the functionality of their device by adding or deleting apps, it is interesting to assess whether the phenomenon is suitable to model effects of apps on consumer behavior of smartphones. A competing concept is mass customization, stating that manufacturers meet market needs by allowing consumers to adapt the functionality of products until the latest possible moment in the supply chain (Chase et al., 2006). For smartphones, this can be done with apps even after purchase of the device. This study contributes by adding to the sparse body of research in smartphones and apps. Moreover, because of the potential negative effects of feature fatigue, it is useful for managers to know whether this phenomenon applies to smartphones. Studying the effects of apps on smartphone usage & satisfaction is also deemed useful for practitioners in the telecom industry due to the transition of voice and SMS to mobile data usage. Theory and conceptual framework A literature review was conducted, revealing (recent) research on smartphones and apps, feature fatigue, customer satisfaction and the Technology Acceptance Model (TAM). The basis for the theoretical framework is that apps affect the value of the smartphone as perceived by consumers, which in turn affects usage and satisfaction. For modeling customer satisfaction, a transaction-specific and postpurchase view of customer satisfaction is used in combination with the comparison-standards paradigm in which perceived value is an experience-based standard for comparison (Cadotte et al, 1987). Customers’ perceived value of the smartphone is a tradeoff between benefits and sacrifices perceived by consumers (Eggert & Ulaga, 2002), in which the benefits are adopted from the Technology Acceptance Model: perceived ease of use and perceived usefulness. In the TAM, perceived ease of use and perceived utility are the main variables that indirectly explain the actual use of a product. The sacrifice is consumers’ perceived effort exerted to install, update and to learning to use the apps. These aspects of value were hypothesized to be predicted by the number of apps on the smartphone of a consumer. In turn, these aspects of value predict the usage of a smartphone and of services provided by the telecommunication operator. Usage of the smartphone was broken down into daily frequency of use, daily duration of use and the perceived usage level. Services of the telecom provider that were adopted in the model were voice, SMS and mobile data. The number of apps installed on the smartphone and the aspects of value were also predicted to affect customer satisfaction with both the smartphone and the telecom provider.
7
Moderating effects were also proposed, which were adopted from the consumer decision making process: consumers hedonic or utilitarian attitude towards smartphones, the brand image as perceived by consumers and the subjective knowledge about smartphones were all expected to moderate the relation between the number of apps installed on the device and the aspects of perceived value. Method The model is tested with an online survey among 252 customers of a Dutch telecommunication provider. The customers in this population were postpaid smartphone users in possession of a smartphone of the brand Apple, Blackberry, HTC or Samsung in the consumer segment. Results of the survey are combined with objective data of voice, SMS and mobile data usage. The measurement instrument used to measure the concepts in the hypothesized model was based on prior research. Control variables are age, gender, education, duration of smartphone possession, number of smartphones owned, smartphone brand and type of use (business or consumer. The model was tested using Structural Equation Modeling (SEM) with Partial Least Squares (PLS) as estimation method. Results The data sample appeared to reflect the customer base of the telecom operator. After deletion of a low number of items with insufficient loadings and cross-loading items, the constructs in the model were measured reliably and convergent and discriminant validity were warranted. Most variables or constructs have an R2 which is moderate to high for consumer behavior studies, but the perceived effort, SMS usage and voice usage were weakly explained. Of the 40 hypotheses in the base model, 13 were confirmed. The number of apps on consumers’ smartphones was positively related to the perception of usability of the device and satisfaction with the device, but negatively to the effort perceived to download, update and learning how to use the apps. Perceived usability was positively related to the perception of usefulness of the smartphone. Regarding the aspects of value, all three constructs were positively related to the perceived usage level, but only the perceived usefulness of the smartphone was positively associated to the daily frequency of use of the smartphone. In turn, consumers that use their device more often per day, often have a higher mobile data usage. Moreover, consumers perceiving their smartphone useful, tend to have a lower SMS usage, albeit that only 4% of the variance of SMS usage is explained. Besides the number of apps, customer satisfaction with the smartphone was related negatively to the perceived effort, but positively to the perceived usefulness and the perceived usability. In turn, consumers that are satisfied with their smartphone are disposed to be satisfied with the telecommunication company. Of the moderating effects, only a negative moderating effect of subjective knowledge was confirmed on the relation between the number of apps installed and the perceived usability of the smartphone. The results gave rise to two post-hoc analyses: A one-way MANOVA for assessing differences between operating systems, and the analysis of mediating effects of subjective knowledge, utilitarian attitude, hedonic attitude and the perceived brand image. It appeared
8
that Blackberry users have less apps installed than users of other operating systems, and that Apple users are generally more satisfied than Blackberry or Android users. Finally, consumers’ utilitarian attitude mediates the effect of the number of apps on the perception of usability. Discussion Because the number of apps installed on the smartphone positively related to the perceived usability and negatively to the perceived effort for installing, updating and learning to use apps, the concept of mass customization applies to apps on smartphones and no support is found for feature fatigue. The utility of the smartphone as perceived by consumers plays a key role in predicting the usage: it is directly associated with decreases in SMS usage, and through an increased frequency of daily use it is positively related to mobile data usage. The number of apps and all aspects of value related to satisfaction with the smartphone, the only negative relation being the perceived effort. No direct effects of apps or value of the smartphone predict satisfaction with the operator, albeit that users that are satisfied with their smartphone do tend to be satisfied with their telecom provider. This implies that no crossover effects exist for the value of the smartphone or the number of apps towards the operating company. A main implication for scholars is that they can apply mass customization to model the use of apps on smartphones , and that feature fatigue does not apply to modular products. Integrating customer satisfaction and technology acceptance proved to provide a model with which effects on apps could be adequately predicted. Managerial implications are that managers need not worry about negative effects of a high number of apps. In particular, usability issues are not associated with a high number of apps, on the contrary: mass customization provides opportunities for app developers and for managers in the telecom sector. Moreover, apps can be used as leverage in the transition from voice and SMS to data usage due to the indirect negative associations with SMS usage and positive relations to mobile data usage. To improve satisfaction, managers could consider automating the updating process of apps on the device, and spend more effort on promoting Apple’s high end devices since these customers are generally more satisfied. Finally, since expert users benefit less from increases in perceived usability due to increases in the number of apps installed, self-service programs should aim at usability issues for expert users. Limitations of the study are that no distinction is made between different categories of apps; the aggregation of all apps on the smartphone and satisfaction of the apps to a central, absolute number of apps; the omitting of psychometric differences based on the moment of adoption; a weak theoretical basis for the mediation analysis; and the fact that the study does not differentiate between the number of apps installed and the number of apps actually used. Research design-related issues are the sample size when considering the serious non-normal distribution of the data, using single-item constructs, and the nature of the data which does not allow causal inferences. These issues are directions for further research.
9
1 Introduction
With the convergence of communication technologies and innovative product features in the past decade (Arruda-Filho et al., 2010) the smartphone has arisen: a combination of handheld computers and mobile phones accommodated with a wide variety of features such as camera’s, organizers, web browsers, media players and navigation. The smartphone has “put the world in the palm of your hand” (Madison, 2011 on dailymail.co.uk). In the first quarter of 2012, smartphones sales accounted for 34% of total mobile phone sales (Gartner.com, 2012). In the Netherlands, smartphone penetration has reached 53% in January 2012 (Telecompaper, 2012). It is predicted that smartphone sales will approach one billion units in 2015 (Idc.com, 2011). With this revolution of smartphones, a new way of consuming software applications has surfaced in the form of ‘apps.’ This study assumes that apps are features. In the context of software, features are defined as a distinguishing characteristic of a software item such as performance, portability or functionality (IEEE, 1998). Apps combine hard technical features such as the camera and a connection to the internet in order to provide additional functionality, which is the basis for the assumption. Since apps can be added and deleted, this assumption results in viewing smartphones as modular products – products of which the functionality can be altered (Stone, 2000). The use of apps raises two important questions about the use of smartphones. For one, what are the effects of the use of apps on the behavior of smartphone consumers? Two very different concepts can be used for modeling the use of apps: feature fatigue and mass customization. Consumers that suffer from feature fatigue have purchased a feature-packed product that is complex in use, while they actually only want a simple, easy to use product (Thompson et al., 2005). Concerns of the phenomenon have been even expressed for all-inone portable devices (Taliuaga et al., 2009 on rockresearch.com) and in particular for “app fatigue” (Kendrick, 2011 on Zdnet.com, 2012). On the other hand, mass customization posits that manufacturers can meet specific needs of consumers by providing flexible market offerings (Anderson & Narus, 1998). Apps allow smartphones to be products of which the functionality is customizable after purchase. This thesis provides an answer to the question which of these theories is best applicable to model the effects of apps on behavior of smartphone consumers. Additionally, it is interesting to assess whether external factors influence the effects of apps on behavior of smartphone consumers. A second question stemming from the use of apps is: What is the effect of apps on the use of services offered by telecommunications providers? A transition is in process in which the more traditional services provided by operator companies (or opco’s) such as voice (calling) and SMS (short message service) are moving more towards the consumption of mobile data (Tilson & Lyytinen, 2006). Apps facilitate this transition since apps exist that replace the use of SMS and voice with the use of mobile data, such as ‘Whatsapp’ or ‘Skype’. The aim of this research is to answer these two questions, thereby shedding light on the effect of apps on the behavior of smartphone consumers. The answers to these questions have both theoretical and practical implications. For scholars, this thesis adds to the existing body of empirical research on smartphones and apps, which is sparse according to Peslak et al. (2011). The research also adds smartphones and apps as a case study to the existing
10
research on consumer behavior, such as customer satisfaction research and technology acceptance research, and it is assessed which of two concepts is applicable to modular products such as smartphones and apps: mass customization versus feature fatigue. A final unique aspect of this study is the use of objective data of usage of telecom operator services in order to study effects of apps on consumer behavior of smartphones. Practically, it is useful for managers to know whether feature fatigue is applicable to smartphones since the phenomenon affects consumer behavior, with reductions in customer satisfaction levels and usage levels as a consequence (Thompson et al., 2005). Along with the positive impact of customer satisfaction on market performance indicators (Anderson & Sullivan, 1993; Fornell et al, 1996; Szymanski & Henard, 2001), customer retention through customer satisfaction is especially relevant in the telecommunications industry in which drop-out rates are high (Lee et al., 2001). Therefore, cross-over effects of satisfaction of the smartphone to satisfaction of the opco are also assessed. Moreover, practitioners working in the telecommunications industry will gain insights in the effects of the use of apps on the use of services provided by telecommunication providers, which is important in the context of the transition of voice and SMS towards mobile data. Additionally, findings of this study are relevant for practitioners in the telecommunications industry who want consumers to have an optimal experience with their handset, which is a general strategic trend in telecommunications (Haverila, 2011b; Ojiako & Maguire, 2009). The remainder of this writing is organized as follows: In the following section, theories and prior research of relevant concepts will be discussed, thereby elaborating upon smartphones and apps, feature fatigue, mass customization, customer satisfaction and technology acceptance. This is followed by the formulation of a conceptual framework and the development of hypotheses. Section 3 will discuss the methodology that is used to empirically test the model, followed by Section 4 in which the results of the empirical testing will be presented. Finally, section 5 will discuss the findings, theoretical and practical implications, and the limitations and directions for further research.
2 Theory and hypothesis development
This chapter will present results of a literature review, thereby elaborating on concepts that are relevant to the construction of the conceptual framework with which the use of apps and smartphones can be analyzed. The chapter starts by discussing smartphones and smartphone apps, followed by a brief discussion of feature fatigue, mass customization, customer satisfaction and the technology acceptance model (TAM). Subsequently, the theoretical framework is presented and the hypotheses are developed. The literature review was conducted according to the guidelines of Randolph (2009); the focus was on research outcomes and theories and the goal was to develop a model with which effects of apps on consumer behavior could be modeled. The covered literature was a purposive sample of the available literature, and the review was conceptually organized. Literature was searched by prompting (combinations of) key concepts into several academic search engines such as Google Scholar, Wiley, JSTOR, ABI/Inform, ISI Web of Knowledge and Sciencedirect. Literature was selected based on scanning titles and abstracts for relevance, and was evaluated according to Google Scholar citations and appearance of the publication source in the Harzing Journal Ranking (Harzing, 2011).
11
2.1 Theory and recent research
2.1.1 Apps and smartphones: modules for modular products Various but similar definitions are used for the smartphone. This research uses the following: ‘a smartphone is a mobile phone that includes software that a user is able to modify and update” (Töyssy & Helenius, 2006, p. 110). Smartphone apps are computer programs that can be installed on a smartphone. Different categories of smartphone applications can be identified, such as communication, browsing, media playing, productivity, system, gaming and map apps (Falaki et al., 2010). The apps result in numerous possibilities for smartphone use such as in healthcare (Park & Chen, 2007), mobile commerce (Chang & Chen, 2004), logistic services (Chen et al., 2007), mobile-based corporate intranet systems (Funk, 2006) and sustainable technology (Pitt et al., 2011). Most smartphone users have around 50 applications installed on their smartphone, while they use only 8-12 applications regularly and on a daily basis (Falaki et al., 2010). An important assumption of this study is that apps are viewed as features because apps provide additional functionality in addition to the hard technical features of smartphones. In the context of software, features are defined as a distinguishing characteristic of a software item such as performance, portability or functionality (IEEE, 1998). Smartphones have hard features, such as a camera, which can in turn be used by software (apps) in order to provide additional functionality. An example is the Sleep Cycle Alarm Clock app, which uses the accelerometer and the microphone in order to analyze sleep patterns and to wake you up in the lightest sleep phase within the half hour before you have set the alarm (iTunes.apple.com, 2012). The author believes that this is an adequate example of why apps can be assumed to be features. The assumption leads to the observation that smartphones are modular products. Modular products are machines, assemblies or components that accomplish an overall function through combination of distinct building blocks or modules (Pahl & Beitz, 1998, in: Stone, 2000). The number of hard technological features does not differ greatly between different smartphone types because most smartphones have similar designs. However, in the set of apps that users have installed in their device, variation can be expected, which supports the need to assume that apps are features. 2.1.2 Too many features: Feature fatigue As technology advances, it becomes more feasible to load products with a large number of features or functions. Adding features makes products more appealing for consumers due to increased expected utility of a product, which is in turn appealing for businesses due to the predicted market share increase. However, too many features can make a product overwhelming for consumers and difficult to use, resulting in consumers’ dissatisfaction with their purchase and a state of frustration which is called feature fatigue (Thompson et al., 2005). The existence of this post-purchase phenomenon is underlined by an increase in product returns for products that turn out not to have any technical malfunction (Den Ouden et al., 2005). Feature fatigue can be explained by a shift in consumer preferences before and after the use of a product, illustrated in Figure 1. Before purchase, consumers believe each feature
12
to add to the utility of a product, knowing that more features make products more difficult to use. However, after having used the product, a shift occurs in consumer preferences towards a simple product with basic functionality which is easy to use. This structural change can be explained by different considerations being salient in expected and experienced utility (Hsee et al., 2009; Thompson et al., 2005). Economic theory models consumers’ preferences with an additive utility function (Lancaster, 1971 in: Thompson et al., 2005): adding attributes to a product or service that consumers evaluate as positive increases the perceived utility of that product or service. Market research techniques such as conjoint analysis or discrete choice analysis model products as bundles of attributes in which each attribute has value (Srinivasan et al., 1997 in: Thompson et al., 2005). Hence, products are bundles of features, and each feature adds value to the product’s utility. A tradeoff is postulated between an increase in this utility perception of consumers and a decrease in the perceived usability. Figure 1 Shift in preference for utility to usability. Adapted from Gaigg (2012) on Michealgaigg.com Feature fatigue has been found to affect consumer behavior, or more specific: satisfaction and usage (Thompson et al., 2005). Consumer behavior “encompasses the acquisition, consumption, and disposition of goods, services, time, and ideas by decision making units (e.g. individuals, families, organizations, etc)” (Jacobi, 1978, p. 87). Various theories are used to model consumer behavior, varying from consumer choice theories such as the Black Box model on the consumer decision making process, to models on the adoption of innovations by consumers such as the Technology Acceptance model (Davis, 1989) or other diffusion models and consumer decision making models. Two major streams of research on consumer behavior are discussed later in this chapter: Customer Satisfaction and Technology Acceptance. 2.1.3 Mass customization: adapting functionality to meet specific needs By varying in the set of product features, each consumers’ specific desire in product capability can be adhered to. Suppliers can meet these specific desires by providing flexible market offerings (Anderson & Narus, 1998), which corresponds to mass customization. Originally, mass customization has been interpreted to create value based on customercompany interaction at the manufacturing and assembly stage of operations, thereby creating customized products (Kaplan & Haenlein, 2006). However mass customization is more recently conceptualized as postponing the task of differentiating a product for a specific customer until the latest possible point in the supply network (Chase et al., 2006). This is certainly the case for smartphones for which apps are obtained after purchase of the smartphone. This interpretation is facilitated by the assumption that apps are features, also since modularity is often viewed as a key factor for mass customization since modularity provides the flexibility for quick and inexpensive customization (Feitzinger & Lee, 1997).
13
The concepts of mass customization and feature fatigue lead to two different viewpoints of how consumers use the apps on their smartphone: applying mass customization results in the view that consumers adapt the functionality of their device according to their preferences. On the other hand, feature fatigue indicates that consumers install a large number of apps because of the expected utility, but end up being frustrated because their device is complex in use. In the following sections, customer satisfaction and technology acceptance are described, which are two types of consumer behavior that can be affected by apps on smartphones. 2.1.4 Customer satisfaction Customer satisfaction is an attitude judgment following a purchase act or a series of consumer-product interactions (Yi, 1990 in: Fournier & Mick, 1999) and is often about evaluation as a result of comparison (Cadotte et a., 1987; Oliver, 1993), which is the basis for the comparison standards paradigm. In this paradigm, different norms are used by consumers in order to form standards for comparison. Such norms can be expectations, experiences, desires or equity (Cadotte et al., 1987; Eggert & Ulaga, 2002; Fournier & Mick, 1999; Halstead, 1999). Halstead (1999) offers a typology of customer satisfaction models based on 1) the level of aggregation of the comparison standard; 2) the stage of the comparison process, and 3) the level of abstraction of the comparison. Customer satisfaction can be measured for individuals per transaction, or all individual satisfaction levels of consumers over a series of transactions can be aggregated to a cumulative level (Anderson & Sullivan, 1993). The stage of the comparison process refers to the stages in the consumer decision process such as need recognition or post-purchase evaluation. Finally, for assessing the level of abstraction of the comparison, the hierarchy of value of Gardial et al. (1994) can be used, stating that satisfaction studies focus on 1) satisfaction with product attributes; 2) satisfaction with products overall; or 3) on global satisfaction of consumers. Customer satisfaction is often a performance indicator for companies. Increased satisfaction leads to positive word of mouth and increased customer loyalty, which is the ‘ultimate dependent variable’ of customer satisfaction because of its value for actual customer retention and profitability (Johnson et al., 2001, p. 222). 2.1.5 Technology Acceptance Model (TAM) Originally introduced by Davis (1989), the Technology Acceptance Model posits that external variables influence a technology’s ease of use (usability) and usefulness (utility) as experienced by users of that technology. Both ease of use and usefulness influence consumers’ attitude towards using a technology, which in turn affects the behavioral intention to use it, which finally affects actual use of a technology and the continued use thereof (Davis, 1989). The TAM is based on the Theory of Reasoned Action (TRA). The TRA stems from social psychology and states that one’s behavioral intention depends on the person’s attitude about that behavior and subjective norms, and that if one intends to conduct a certain
14
behavior, it is likely that he or she will do so (Fishbein & Ajzen, 1975). Later variations of the TAM leave out the attitude construct (Venkatesh et al., 2003). The theoretical foundation for using the perceived ease of use and usefulness as predictors for usage behavior can be found in self-efficacy theory and the cost-benefit paradigm. Self-efficacy is defined as “judgments of how well one can execute courses of action required to deal with prospective situations” (Bandura, 1982, p. 122 in Davis, 1989). The cost-benefit paradigm stems from behavioral decision theory and describes decision making strategies in terms of a cognitive tradeoff between the effort required and the quality of the resulting decision (Payne, 1982). Based on the cost-benefit paradigm and self-efficacy theory, Davis (1989) distinguishes between outcomes judgments and self-efficacy judgments, in which the outcomes relate to usefulness of using a technology, and self-efficacy relates to how easy that technology is to use. Both these judgments, or perceptions, of consumers determine the acceptance and usage of a technology by consumers. The Technology Acceptance Model has been extensively tested in research and numerous variations and extensions of the model have been developed. An overview of several variations and extensions is provided by Venkatesh et al. (2003).
2.2 Conceptual framework and hypotheses
This section will first describe the ‘theoretical lens’ with which the questions posed in the introduction were approached, followed by a discussion of the concepts and the relations between them. The conceptual model and the proposed hypotheses can be viewed in Figure 2. It should be noted that, based on the initial literature review, an elaborate model was developed initially. However, only part of that model is discussed here in order to present a simplified model. The framework in Figure 2 can be used to analyze the effects of apps on behavior of smartphone consumers. The number of apps installed on the smartphone predicts the value perceived by consumers. Derived from customer satisfaction research, value is a standard for comparison which consumers use to assess their satisfaction. The majority of customer satisfaction research focuses on expectations as standard for comparison, which is however not relevant for post-purchase processes (Halstead, 1999) and durable goods (Churchill & Surprenant, 1982). Since smartphones are durable goods and apps are acquired after purchase of the smartphone, value as an experience-based norm is used as a standard of comparison. Satisfaction models based on experience-norms use the confirmation/disconfirmation paradigm in combination with brand attribute beliefs and the experience of using a product (Cadotte et al., 1987). Moreover, satisfaction is viewed here as transaction-specific (and not cumulative) since this view is more suitable for assessing individual differences between consumers (Halstead, 1999). Value is based on the exchange theory of marketing (Eggert & Ulaga, 2002), cognitive psychology and economic theory (Thaler, 1985 in: Gallarza & Saura, 2006): perceived value relates to consumer behavior based on the concept of value transaction (Gallarza & Saura, 2006). Value is defined as the benefits customers receive in relation to total costs or ‘sacrifice’ (McDougal & Levesque, 2000). Sacrifices can be monetary and non-monetary (time and effort) costs associated with acquiring and using a product or service (Cronin et al.,
15
Moderators
Usage Opco services
Perceived brand image
Subjective knowledge
Hedonic attitude
Utilitarian attiude
Voice
Mobile data
SMS
Perceived value of the smartphone
Perceived usefulness
H3a-g H6a-i
H8a-c
H9a-c
H10a-c
H11a-c H2(+)
Perceived usability
Usage of the smartphone
H4a-h
Daily frequency of use
Daily usage time
Perceived usage
Number of apps
H1a-c
Perceived effort
H5a-f
Satisfaction
H1d,e(+)
Customer satisfaction smartphone
H7(+)
Telecom provider satisfaction
Figure 2: the hypothesized model
2000). The benefits that a consumer perceives from product features are adopted from the acceptance Model: perceived ease of use and perceived usefulness, which both predict usage (Davis, 1989). In the original TAM, the intention to use is an intermediate variable between usefulness & usability, and usage behavior. However, the intention to use is not adopted in the conceptual framework in order to simplify the model. As in a later extension of the TAM, the Unified Theory of Acceptance and Use of a Technology (Venkatesh & Davis, 2003), also the attitude towards using a technology is not adopted. Perceived usefulness is defined as the belief that using a product or service positively relates to job performance (Davis, 1989), which can be extended to performance in everyday life for consumer products. Perceived ease of use is defined as the degree to which a person believes that using a particular product or service would be free of effort (Davis, 1989), in which effort is conceptualized to be a finite resource that a person may allocate to the various activities that he or she is responsible for (Radner & Rothschild, 1975, in: Davis, 1989). Outcomes of perceived value in this framework are actual usage levels and customer satisfaction. Value and satisfaction are related since they are both evaluative judgments about products or services (Woodruff, 1997). However, they are two distinct constructs due to satisfaction being the result of an affective comparison process and value being the result of a cognitive comparison process. In the model, satisfaction stems from value being an experienced-based standard for comparison which consumers use to assess their satisfaction. Usage is also related to value as it is the result of the value-variables perceived ease of use and perceived usefulness in the Technology Acceptance Model. In the outcome variables usage and satisfaction, distinction is made between the smartphone and the telecommunication operator. For satisfaction, this means that customer satisfaction towards both the telecom provider and the smartphone manufacturer is assessed. Usage is divided into usage of the smartphone, and usage of services provided by the operating company: voice
16
(calling), SMS and mobile data. Usage of the smartphone is broken down into daily usage time, daily frequency of use and the perceived usage level (Al-Gahtani & King, 1999). Recent research has found that the frequency of use and the average duration of usage varies greatly among users of smartphones (Falaki et al., 2010), which is why usage of the smartphone is divided into three separate variables. Finally, variables influencing the relations between the number of apps and the aspects of value are derived from the consumer decision making process (Lamb et al., 2012). Klein (1998) proposes that several characteristics of consumers shape the information-search phase in this process. The following consumer characteristics are adopted in this study: subjective knowledge, attitude towards the smartphone and the perception of the brand’s image in the mind of the consumer. In the following sections, the proposed relations between the concepts are described. 2.2.1 Number of apps on perceived value and customer satisfaction A central predictor in the model for the value of the smartphone as perceived by consumers is the number of apps. Based on the assumption that apps are features, this study focuses on the number of apps installed on the smartphone as the number of features. The number of apps is proposed to predict value and satisfaction. Two competing theories, each leading to a different set of hypotheses, are used to predict the relations between the number of apps and the aspects of value: feature fatigue and mass customization. According to feature fatigue, each additional feature is expected to increase perceived usefulness, to decrease usability, and to increase perceived sacrifice. Due to the added utility of each app, apps are expected to increase perceived usefulness. Moreover, each additional feature is “one more thing to learn, one more thing to possibly misunderstand, and one more thing to search through when looking for the thing you want” (Nielsen, 1993, p. 155 in Thompson et al., 2005). Therefore, a negative relation is expected between the number of apps and the perceived usability. This increase in utility and decrease in usability due to an increase in the number of apps corresponds to feature fatigue. Then there is the cost-aspect of value: applications require resources in the form of time and effort to install and update the applications and to learn to use the applications effectively. Barriers for customer satisfaction of software products that have been found in previous research are effort for installation and maintenance (Kekre et al., 1995) which correspond to the effort of customers required to install and maintain (update) apps. Because a large number of applications is offered free of charge and because the price of apps differs per operating system, monetary costs are not adopted in the model. However, apps are expected to require effort and time since users are to download, install and regularly update the apps and since users are to learn how to use the apps on their smartphone. Therefore, based on the perspective of feature fatigue, smartphone users with a high number of apps are expected to be more likely to perceive a high degree of effort. On the other hand, the concept of mass customization contradicts the expectations that are based on feature fatigue regarding usability and effort. The modular property of smartphones can be considered an anomaly of feature fatigue: consumers can modify the functionality of their device after purchase by adding or deleting apps according to their preferences. Based on this anomaly, mass customization is proposed as a theory competing
17
with feature fatigue as an explanation of the effects of apps on behavior of smartphone consumers. Applying mass customization creates value (Kaplan & Haenlein, 2006) which can be attributed to the increased added utility of each app or feature and also to increased perceptions of usability (Kamis et al., 2008). A positive relation can be proposed between the number of apps and usability based on mass customization: smartphone users with a high number of apps have customized the utility of their smartphone and are also expected to have customized the usability of their smartphone, e.g. by configuring different home-screens and shortcuts that make their device more easy to use. The number of apps installed on consumers’ smartphone may well contribute to a users control over the functionality and usability of their device. Therefore, based on the view of mass customization, a positive association between the number of apps and the perception of usability can be expected. Also, a negative association between the number of apps and the effort perceived by consumers is expected: based on mass customization, it can be expected that users of modular products adapt the functionality of their device according to their preferences. Those users will associate a higher number of apps on their smartphone with a desirable degree of effort exerted into downloading, updating and learning how to use those apps. Moreover, a learning curve can be expected in which consumers that have a high number of apps, perceive less effort for the high number of apps because they are have developed an aptitude for using the apps. This lower perception can be especially expected when they have installed a number of apps conform to their need and have customized their product. Moreover, the number of apps is expected to relate to customer satisfaction directly, which is in line with previous research findings of products attributes and customer satisfaction: according to the hierarchy of value by Gardial et al. (1994), customer satisfaction stems from product attributes. In the context of this research, smartphone users with a high number of apps are expected to be more likely to be satisfied with their smartphone, based on this hierarchy of value. Additionally, cross-over effects are expected between the number of apps on the smartphone and the satisfaction with the telecommunication provider. Mittal et al. (1998) find that users of consumption systems (a combination of products and services) attribute satisfaction of products to service providers over time. Since telecommunication operators deliver smartphones to consumers and provide the services necessary to operate smartphones, they form a consumption system and consumers can be expected to relate their satisfaction with their handset to their satisfaction with the operating company. The following relations are hypothesized between the number of applications that consumers have installed on their smartphone, and the aspects of value and customer satisfaction (note that the original hypotheses reflect feature fatigue and that the alternative hypotheses reflect mass customization):
H1a-e The number of apps installed on the smartphone is (a) positively related to perceived usefulness, (b) negatively related to perceived usability, (c) positively related to perceived effort and is positively related to customer satisfaction of (d)the smartphone and (e) the telecom provider The number of apps installed in the smartphone is (b,alt) positively related to perceived usability and (c,alt) negatively related to perceived effort
H1b,c,alt
18
2.2.2 Perceived ease of use on perceived usefulness Extensive empirical evidence has been generated in previous research for the relation between perceived usability and perceived usefulness, which is why this relation is not widely elaborated. The underlying explanation is that products that are easy to use increase the utility that consumers perceive to enjoy from the product in use (Davis, 1989; Venkatesh & Davis, 2000). Consumers are more likely to perceive a product as useful when they perceive that product to be easy to use, than when they experience difficulties in operating the product. This is especially true for products in the information systems category (Venkatesh et al., 2003), such as smartphones. The following relation is hypothesized:
H2 Perceived usability is positively related to perceived usefulness
2.2.3 Aspects of value on usage of the smartphone, services provided by the operator and customer satisfaction In the Technology Acceptance Model and variants thereof, perceived usefulness and perceived usability are positively related to (attitude towards) usage (e.g. Davis, 1989; Venkatesh & Davis, 2000). Smartphone consumers that believe their device is useful are expected to indicate to 1) use it for more hours per day 2) more often per day, 3) and also to actually say to use it more, when compared to users that perceive their device to be useless. This is expected because they see more purposes for using their smartphone. Additionally, users that perceive their device as easy to use are also expected to use it longer, more often and to perceive to use it more: when the smartphone is easy to handle, there are less barriers to use the device for a longer period of time or more frequent, which also explains why those users believe they have a high degree of usage. Also users that perceive to exert a high degree of effort into their smartphone for installing, updating and learning how to use the apps on their smartphone are expected to use it more often, longer and to perceive a higher degree of usage. This expectation is a direct consequence from the concept of effort, i.e. people that perceive to spend time on updating, installing and learning how to use the apps on their smartphone probably tend to use it more. Besides the positive associations expected between the value aspects of the smartphone and the usage aspects of the smartphone, associations are also expected between the value aspects of the smartphone and the usage of the services provided by telecommunication providers: SMS and voice services and mobile data. Usability and usefulness are expected to relate negatively to SMS and voice usage, but positively to data usage, based on the following rationale: people that perceive their device as easy to use have low barriers to using the apps on their device. Moreover, if consumers believe that their smartphone is useful, they will be expected to use more apps in order to make use of the full potential of their device. Since the barriers to use their smartphone is low and they use more apps, consumers are also expected to use more mobile data since many apps require a connection to the internet. However, with the transition of SMS and voice usage towards mobile data usage in mind, it may be expected that apps which use mobile data are replacements for the more traditional phone services. Therefore, it is expected that consumers that perceive their device as useful, make less use of voice and SMS. The same relation is expected for usability because of the lower barriers for the interaction with the smartphone: consumers that consider that
19
interaction as easy, have more interactions with the apps on their smartphone and consequently often use more mobile data and use less traditional telecom services. Finally, the effort that consumers perceive to have put into the smartphone’s apps are only expected to relate positively to mobile data usage, since this is the only service provided by opco’s that is needed to use apps. The value perceived by consumers is also expected to relate to the degree of satisfaction with both the smartphone and the provider since in the model it is the standard that consumers use to assess their satisfaction. Customer perceived value has been found to directly affect satisfaction (Eggert & Ulaga, 2002; McDougal & Levesque, 2000; Spiteri & Dion, 2004) and especially usability has been found to be an important driver of overall customer satisfaction for software products (Flavian et al, 2005; Kekre et al., 1995). Consumers that perceive their smartphone as useful and easy to use are expected to rate their satisfaction level with the smartphone as higher than consumers who believe otherwise. On the other hand, effort is a negative aspect of value, and therefore effort is expected to be negatively related to satisfaction. Based on the cross-over effects between products and services (Mittal et al., 1998) discussed in section 2.2.1, the aspects of value are also expected to be associated with the satisfaction that customers assign to the provider of the services that are needed to use their smartphones. Based on the above, the following relations are hypothesized:
H3a-c H3d-f H3g H4a-c H4d-f H4g-h H5a, b H5c,d H5e,f Perceived usefulness is negatively related to (a) SMS usage and (b) voice usage but is positively related to (c) mobile data usage Perceived ease of use is negatively related to (d) SMS usage and (e) voice usage but is positively related to (f) mobile data usage Perceived effort is positively related to mobile data usage Perceived usefulness is positively related to (a) daily usage duration, (b) daily frequency of usage and (c) perception of usage Perceived usability is positively related to (d) daily usage duration, (b) daily frequency of usage and (c) perception of usage Perceived effort is positively related to (a) daily usage duration, (b) daily frequency of usage and (c) perception of usage. Perceived usefulness is positively related to customer satisfaction of (a) the smartphone and (b) the telecom provider Perceived ease of use is positively related to customer satisfaction of (a) the smartphone and (b) the telecom provider Perceived effort is negatively related to customer satisfaction of (a) the smartphone and (b) the telecom provider
2.2.4 Usage of the smartphone on usage of telecommunication provider services The aspects of usage of the smartphone are expected to be associated with the usage of the services provided by telecom operators. The top three features that are used on phones are the phone, SMS and internet (Haverila, 2011b). For making calls and for texting messages, smartphone users have to make use of the services provided by telecom operators. However, it is expected that smartphone users make less use of the traditional telecommunication services for texting and calling, which is based on the transition in the telecoms industry from these traditional services to mobile data (Tilson & Lyytinen, 2006). Therefore, users that indicate to use their smartphone more often or longer per day, and/or perceive to have a high level of usage are expected to make more use of mobile data and less use of SMS and voice services. For use of mobile internet services, this relation is less straightforward since heavy
20
users of apps can use a WiFi network, thereby circumventing the costs associated with using mobile data provided by their operator. Nevertheless, consumers that say to be heavy users of smartphones, that use it often and for a long duration on a daily basis, are expected to make more use of mobile data. The following associations are expected:
H6a-c H6d-f H6g-i The daily usage duration of the smartphone is negatively related to (a) SMS usage and (b) voice usage but (c) positively to mobile data usage The daily frequency of use of the smartphone is negatively related to (a) SMS usage and (b) voice usage but (c) positively to mobile data usage The perception of the usage level of the smartphone is positively negatively related to (a) SMS usage and (b) voice usage but (c) positively to mobile data usage
2.2.5 Customer satisfaction of the smartphone on customer satisfaction of the telecom provider The telecommunication provider is a value added reseller of the smartphone because it often provides the handset itself and the services that are necessary to use the smartphone and, such as mobile internet. Again, cross-over effects can be expected between satisfaction that the users of smartphones attribute to their device, and the satisfaction that they attribute to the provider of the device and the services necessary to operate it. For instance, if a customer is in possession of a smartphone that frequently drops calls because of an error in the design of the antenna, it is likely that his or her dissatisfaction with the device is also projected on the provider of the services necessary to make the call. Therefore, it is expected that customer satisfaction with the smartphone and customer satisfaction with the operator are related:
H7 Customer satisfaction of the smartphone is positively related to customer satisfaction of the telecommunication provider
2.2.6 Moderating variables Finally, variables that influence the relation between the number of apps on consumers’ smartphones and the perception of value are discussed. These are subjective knowledge, hedonic attitude or utilitarian attitude towards the smartphone and perceived brand image. However, the directions of the moderating effects are not established in the hypotheses because these are differential, depending on which of the either concepts of feature fatigue or mass customization is supported. Consumers that are knowledgeable about smartphones have been found to be better able to exploit mobile services (Deng et al., 2010). Consumers’ subjective knowledge is defined as “a consumer’s perception of the amount of information they have stored in their memory” (Flynn & Goldsmith, 1999, p. 59), which describes the consumer’s knowledge associated with the product in general. It is plausible to believe that users that consider themselves adequate users of smartphones have a different perception of the usability and utility of their smartphone than novice users. It is also likely that experienced users will benefit more from increases in usefulness and usability, and will suffer less from decreases in perceived usability and increases of sacrifice which are related to the number of apps. The latter can be expected because they take less time and effort to install, learn to use and maintain applications, when compared to average or beginning users of applications. In the TAM, a consumer’s attitude towards a product or technology has been adopted as a moderating variable (Venkatesh & Davis, 2000). Consumers can have different attitudes
21
towards products in which two major categories can be distinguished: a utilitarian and a hedonic attitude (Childers et al., 2001). Consumers with a hedonic attitude find pleasure in using a product, while consumers with a utilitarian attitude use a product to achieve certain outcomes (Batra & Athola, 1990). Based on these different attitudes that consumers can have towards smartphones as a product category, differences in the perception of value are expected because people that experience smartphone as hedonic are more likely to enjoy the range of functionality offered by different applications. On the other hand, consumers with an utilitarian attitude can be expected to be more fatigued by functionality of their smartphone which they do not make use of. It should be noted that consumers could also have both attitudes towards smartphones and that they are not mutually exclusive. The perception of the brand by consumers is expected to affect the extent to which they perceive the aspects of value, e.g. consumers that are biased by a positive brand perception are expected to be positively inclined from functionality of applications and negatively biased from perceived effort increases. Conversely, consumers with a low brand perception could be prone to being irritated from decreases in usability from the high number of apps or perceive more effort required for installing, updating and learning to use apps. The following relations are hypothesized:
H8a-c H9a-d H10a-d H11a-c Consumers’ subjective knowledge moderates the relation between the number of apps and (a) perceived usefulness, (b) perceived usability and (c) perceived effort Consumers’ hedonic attitude towards the smartphone moderates the relation between the number of apps and (a) perceived usefulness (b) perceived usability and (c) perceived effort Consumers’ Utilitarian attitude towards the smartphone moderates the relation between the number of apps (a) perceived usefulness, (b) perceived usability and (c) perceived effort Consumers’ perception of the image of the smartphone brand moderates the relation between the number of apps and (a) perceived usefulness, (b) perceived usability and (c) perceived effort
The model that has been developed in this section should provide an adequate basis for answering the research questions posed in the introduction about the effects of apps on consumer behavior of smartphones and telecommunication provider services. In the following section, the methods used to empirically test the model will be addressed.
3 Method
The model that is developed in the previous section was tested with an online survey. This section will discuss the data sample, the measurement instrument, the data collection procedure and the methods used to analyze the data and to test the model.
3.1 Data sample
Data was collected from customers of a telecommunication provider in the Netherlands. The respondents had to be in possession of a smartphone of the brands Apple, Blackberry, HTC or Samsung. These brands accounted for 90.6% of smartphone sales in March 2012 (GfK, 2012). Only postpaid consumers (i.e. with a monthly subscription) were approached because smartphone apps have a higher penetration rate among postpaid users compared to prepaid users (Telecompaper, 2011). In addition, the respondents were to have a subscription at the operator under focus for at least one month in order to ensure that their smartphone is “out of the box” and in use. Finally, a selection criterion was that the
22
respondents were to have installed at least one app themselves such that they could answer questions about perceived effort of apps.
3.2 Measurement instrument
Several variables in the conceptual model are measured with multiple items. Measuring constructs with multiple items can capture the complexity of a theoretical construct or a latent variable that is not directly measurable (Fornell et al., 1996). To facilitate construct validity and face validity, the constructs were measured with items from previous studies (Flavian et al., 2005; Hair et al., 2010). The variables and constructs in the conceptual model were measured with the measurement instrument in Appendix A. Information on usage of services offered by the telecom provider were extracted from the customer database. The other constructs and variables were measured with an online survey. This survey, which was in Dutch, can be found in Appendix B. The design of the website for the online survey was such that it could easily be completed on desktop or laptop PC’s, smartphones and tablets. The answer options of the questions of the online survey appeared in a random order. The survey and the translation were reviewed by a marketing analyst at the telecom operator and two assistant professors of Marketing at Eindhoven University of Technology. It was also pre-tested by five people. The constructs and variables that are measured by the instrument are the following: Number of apps installed on the smartphone –The number of installed apps was used to measure the number of features that consumer’s have installed on their smartphone. The consumers were asked for the total number of apps on their smartphone, including preinstalled features, in intervals of 10 applications. Above 100 apps, the answer option was open such that the scale could be extended if need be. The pre-installed applications were required because all applications that consumers had installed on their smartphone could provide functionality that pre-installed applications also offer. Perceived usefulness – This measure was adapted from a study assessing a product for business use and was adapted such that the usage situation reflected situations in everyday life. The item “Using this technology improves my job performance” was not adopted, since smartphones were assessed for consumer use and since translating the job performance to performance in everyday life could lead to ambiguity when the other three items are taken into consideration. Perceived ease of use – Is measured by asking respondents for the clearness and understandability of the interaction with the smartphone, the degree of mental effort required for operation, the easiness of use and the degree to which consumers can get their device to do what they want it to do, as in Venkatesh & Davis (2000). Perceived effort and perceived monetary costs – In previous research the sacrifice aspects of value were measured under one construct: perceived sacrifice. In this research however, only the perceived effort or time was measured because the price paid for applications is different for different smartphone operating system, i.e. numerous apps that are freely available in the Android app market are charged for in Apple’s Appstore. Because
23
of possible confusion of effort with the construct perceived ease of use, the effort was measured in perceived time. The measures were adopted from Deng et al. (2010). Usage of the Smartphone – Previous studies have measured usage with the selfreported items daily duration of use, daily frequency of use, the number of applications used and the perceived usage (Al-Gahtani & King, 1999; Kim, 2008). Since the present study focused on usage and satisfaction of smartphones stemming from the number of applications in use, the number of applications were dropped. Because this study expected differential effects for frequency of use, duration of use and perception, these aspects were not used as items but as variables. Previous studies such as Kim (2008) divided the frequency of interaction and hours of use per time period in prefixed time intervals which are probably different for smartphones. Therefore, the daily usage and frequency of interaction were stated as open questions and recoded afterwards. The process of recoding is described in Appendix C. Recoding the frequency and duration of use into ordinal scales should have resulted in more reliable answers since the self-reported measures can be expected to be inaccurate. SMS, Voice and Mobile data usage – Information on SMS, voice and mobile data usage was an objective measure that was extracted from the operator’s customer database and is the average usage per month between December 2011 and February 2012. It should be noted that data usage over WiFi networks cannot be registered by telecom providers. People can use internet on their smartphone solely over their home WiFi network but not have a mobile data subscription, meaning that they can be heavy users of internet on their smartphone without using the mobile data services offered by operators. Customer satisfaction – Almost all studies that use a self-measurement of customer satisfaction show a negatively skewed distribution of this variable (Peterson & Wilson, 1992), i.e. overall, customers are more satisfied than dissatisfied. Turel & Serenko (2006) use a ten-point Likert scale to avoid skewness problems because it enables respondents to make better discriminations (Andrews, 1984 in Turel & Serenko); this approach was adopted in the measurement instrument with an eleven point likert scale in order to allow for a middle answer option, which is coherent with the 7-point Likert scales used for the other items. A single-item was used to measure customer satisfaction directly. Using single-item scales for measuring customer satisfaction may decrease the quality of measurement (LaBarbera & Mazursky, 1983 in: Mittal et al., 1998). Nevertheless, Bergkvist & Rossiter (2007) suggest that single item constructs can be used in marketing for concrete objects and attributes such as customer satisfaction. Using single-item constructs also decrease the completion time of the survey. Subjective knowledge – One item was deleted from Flynn & Goldsmith (1999) since the translation in Dutch language offered too little differentiation from another item measuring the same construct. The measure then consisted of the items knowing much about smartphones, how to operate them, whether the respondent believes that other people think he or she is an expert. The respondent is also asked to compare him/herself to others regarding smartphone knowledge. Utilitarian and hedonic attitude towards smartphones – The two dimensions of attitude towards smartphones are adopted from Voss et al. (2003) who develop measurement scales for the two dimensions of attitude towards product categories. The scales consist of 5
24
items per dimension regarding the hedonic and utilitarian attitude, and are adapted to relate to enjoyment of smartphones and the utility of smartphones in general. Perceived brand image – The perception of the brand’s image by consumers is measured by asking respondents for the attractiveness, reliability and quality of the brand, as in Low & Lamb (2000). Low & Lamb (2000) find support for distinguishing between brand image, brand attitude and perceived quality as three types of brand associations. Control variables – Age, gender, the education level, the number of smartphones owned, the duration of smartphone possession, the smartphone’s brand, type of use (business or consumer) and the device on which the survey was completed were measured as control variables. The motivation for including each of these variables is discussed in the final section of this chapter. It should be noted that all the latent constructs are reflective constructs for which the items are “caused” by the constructs, as opposed to formative constructs which are formed by the items measuring the constructs (Hair et al., 2010, Ch. 12).
3.3 Data collection procedure
The request to complete the online survey was administered via e-mail to 2446 consumer of the Dutch operator. The extraction of the respondents from the customer database of the operator was at random except for the criteria discussed in Section 3.1. Data collection, graphic design of the online survey and the contacting of the consumers was conducted by a marketing research agency. As an incentive to boost the response rate, five vouchers for an online store with a value of €20 were raffled among the respondents. The survey request was issued on April 17th 2012. A reminder was issued after 4 days, and the survey was closed on the 28th of April and hence was available for 15 days. 8 invitations were ‘bounced’ – a non-delivery notification was received due to e.g. a non-existing e-mail address. 18 people accessed the link of the survey, but did not start the survey and 71 potential respondents started the survey without finishing it. A total of 271 surveys were fully completed, a response rate of 11.1%. On average, people took 11:34 minutes to complete the survey. The manipulation of the original dataset is described in Appendix C. Three respondents were deleted from the database because they had a completion time of four minutes or less and because it was suspected that the answering of the questions was conducted at random. Of the remaining 271 respondents, 19 did not meet the criteria of the data sample since four respondents were not customers of the operator and 15 respondents did not download apps. These respondents were deleted from the dataset, which left 252 usable respondents (10.3%).
3.4 Analytical method
The dataset was delivered in Microsoft Excel 2010. SPSS 18 was used for the initial processing of the data. A non-response analysis was conducted using a one-way MANOVA for comparing early and late respondents. To test for common method bias, Harman’s one factor test was applied using principal components analysis in SPSS. Structural equation modeling (SEM) is used to test the model developed in Section 2. SEM consists of two steps: the confirmation of the measurement model and the estimation of
25
the path model or structural model. When the measurement model is analyzed, it is assessed how well the different directly measured items reflect the indirectly measured constructs. Subsequently, in the path analysis, it is assessed whether the predicted relations between the latent constructs exist and to what extent (Hair et al., 2010, Ch. 11). Structural equation models can be estimated by Maximum Likelihood (ML) estimation or by Partial Least Squares (PLS) estimation methods, of which the former is a covariance based approach and well known through software applications such as LISREL or AMOS. The latter is a less-known, variance-based approach (Haenlein & Kaplan, 2004). PLS is used as estimation technique for several reasons. Compared to other multivariate techniques, PLS is robust to violations of distributional assumptions such as normality and smaller sample sizes can be used (Fornell & Bookstein, 1982). In addition, PLS is less sensitive to multicollinearity problems (Hair et al., 2010, Ch. 12), and more information of the data is retained since PLS uses covariance instead of correlation, which is the standardized covariance (Ringle et al., 2005 on smartpls.de). Finally, PLS is more suitable for prediction rather than confirmation of existing theory (Hair et al., 2011). The software package that was used for PLS is SmartPLS 2.0 M3 (Ringle et al., 2005 on Smartpls.de, 2012). An abort criterion of 10-5 was used to assure the PLS algorithm’s convergence while minimizing computational requirements (Hair et al., 2011). A maximum of 300 iterations was permitted. Cutoff values and test approaches for validity and reliability are according to Hair et al. (2011). Bootstrapping with 500 resamples was used as a reactive Monte Carlo resampling strategy in order to assess the significance of the estimates. In bootstrapping, parameters’ values and standard errors are compared to repeated random samples drawn with replacement from the original observed sample data, in order to assess the significance of the estimated parameters (Marcoulides & Saunders, 2006). Effects were deemed significant if the probability of erroneously detecting an effect was smaller than .05 . Moderating effects were analyzed based on the product indicator approach (Henseler & Fassot, 2010). For assessing the moderator effects in SmartPLS, the indicator values were standardized (recomputed to have to a mean of 0 and a standard deviation of 1) before multiplication. Hypotheses regarding moderating effects were significant if the coefficient of the interaction term was significant (Henseler & Fassot, 2010). The strength of the moderating effects was determined by calculating the effect size ƒ2 which is an indicator for the change in R2 due to adding the interaction term of the moderator.1 Effect sizes of respectively .02, .15 and .35 are weak, moderate and strong (Henseler & Fassot, 2010). The moderator analysis was conducted separately per moderating effect, using 500 resamples for the bootstrapping procedure. To ensure correct model estimation, several control variables were included in the model. Preferences for different features and the satisfaction caused by features have been found to differ by demographic variables, such as age, gender and education (Haverila, 2011a, 2011b). In addition, the time of owning a smartphone and the number of smartphones owned are also used because people that have owned multiple smartphones or that have owned a smartphone for a long time may be more experienced with and knowledgeable about
1
ƒ2
(Cohen, 1988 in: Henseler & Fassot, 2010)
26
smartphones. The brand of the device used by the consumer is controlled for as well since the different brands make use of different operating systems, have different apps available to them and have different interfaces, leading to possible differences in e.g. the perception of usability between the users of different smartphone brands. Finally, the type of use of the smartphone (business, pleasure or both) is controlled for, since users that use their smartphone for business purposes do not use it entirely voluntarily, and therefore the evaluations of the device and the apps may differ from the evaluations offered by consumers. Dummy variables were used for categorical variables with more than two categories (i.e. smartphone brand, survey method and type of use). The coding of dummy variables is described in Appendix C. As in prior research using PLS, age, number of smartphones owned, usage time and education were treated as ordinally scaled variables (e.g. Venkatesh & Davis, 2000).
4 Results
This section will present the results of the data collection and the statistical analyses. First, the sample will be described, followed by a non-response analysis and the cleaning of the data. Subsequently, the measurement model will be tested, followed by several tests of the structural model: the base model, moderating effects and control variables. The results induced two post-hoc analyses which are also described: a comparison between the different operating systems and the analysis of mediating variables.
4.1 Sample and data characteristics
Several characteristics of the data sample (N=252) can be viewed in Table 1 (note that this table was reconstructed after the deletion of outliers in section 4.1.3). Corresponding characteristics of other customer groups and the Dutch population can be viewed in Appendix D. The distribution of males and females in the population seems to be relatively even, although the fraction of females is slightly higher than the fraction of males. This is also the case for other customer groups and the Dutch population. The distribution of the different types of education in the sample does not accurately reflect the average Dutch population in the sense that the average Dutch population seems lower educated. Regarding age, the sample does appear to reflect the operators’ customer base, which consists of younger people when compared to the Dutch population. Regarding the representation of the smartphone brands of the sample to the operator’s customer base, Blackberry and Samsung users seem slightly underrepresented as opposed to HTC users, which are overrepresented. Regarding the respondents, an equal proportion of the four brands was present in the 2446 invitees, which is not the case for the actual sample: more HTC and Samsung users responded than Apple and Blackberry users. The majority of the respondents indicated to use their smartphone for consumer purposes only while about one-third uses their smartphone for both business and consumer purposes. Only one respondent indicated to use his or her smartphone solely for business, which makes sense because the survey was issued to consumer users. Finally, the majority of the respondents indicated that their current smartphone was their first, and the majority of the respondents has owned a smartphone for less than two years.
27
Table 1 Descriptive statistics of control variables
Variable Gender Age Label Male Female 15-25 26-35 36-45 46-55 56-65 66-75 75+ None Primary VMBO/MBO1 HAVO/VWO MBO 2-4 HBO WO Other Frequency 123 129 46 46 63 64 26 3 2 1 1 30 29 71 79 34 7 Percentage 48.8 % 51.2 % 18.3 % 18.3 % 25.0 % 25.4 % 10.3 % 1.2 % 0.8 % 0.4 % 0.4 % 11.9 % 29 % 28.2 % 31.3 % 13.5 % 2.8% Variable Smartphone Brand Label Apple Blackberry HTC Samsung Other Consumer Business Both 1 smartphone 2 smartphones 3 smartphones 4 or more smartphones 0-8 months 8-19 months 19-31 months 31-43 months 43-54 months More than 54 months Frequency 47 34 81 86 4 160 1 91 153 44 39 16 52 89 41 27 16 27 Percentage 18.7 % 13.5% 32.1% 34.1 % 1.6 % 63.5 % 0.4 % 36.1 % 60.7% 17.5% 15.5% 6.3% 20.6% 35.3% 16.3% 10.7% 6.3% 10.7%
Type of use
Education
Number of Smartphones owned Usage Time
4.1.1 Non-response analysis Extrapolation is used for assessing non-response. The early and late respondents are compared, since late respondents and non-respondents are both ‘less readily’ in their response (Scott Armstrong & Overton, 1977). The first and last quartile of respondents are compared (n=63). A one-way MANOVA was conducted in SPSS with the response group as the predictor variable and the measured constructs used in the structural model (as calculated by SmartPLS) are used as the dependent variables. In the calculation of these variables, the items that are deleted in section 4.2 are not used. The MANOVA revealed no significant multivariate main effects for response group at the p < .05 level, Wilks’ ? = .868, F(16;105) = 1.000, partial eta squared = .132. Power to detect the effect was .631. Given the low power to detect the effect (under .80), the tests of between subject effects were assessed for further differences between the different response groups on the metric variables, but no significant effects were observed. Based on these findings, it is concluded that there are no significant differences between early and late respondents and that in turn it is not likely that there is a non-response bias. 4.1.2 Common method bias Harman’s one factor test was used to test for common method bias. Using principal component analysis, eight factors were extracted with an eigenvalue exceeding 1.0 (the Kaiser criterion), accounting for 69.0% of the variance. The first and largest factor of the unrotated solution accounted for 30.5% of the variance, indicating that common method bias is probably not an issue since this is not the majority of the variance (Podsakoff et al., 2003). 4.1.3 Missing data, outliers and multivariate assumptions The data did not contain missing values, which is because the online survey tool required the respondents to answer every question. Univariate outliers were detected visually by examining boxplot diagrams and by assessing which variables have a standardized value exceeding ± 4.0 (Hair et al., 2010, Ch. 2). The variables that were measured with a Likert scale or the customer satisfaction variables are not assessed for univariate outliers since univariate outliers are plausible for these variables and were not to be deleted. The outliers
28
can be viewed in Appendix E, as well as the actions following from this analysis. For 11 cases, values were deleted. Multivariate outliers were detected using the Mahalanobis distance, calculated in SPSS. The distance is calculated using all items in the base model (36) in which no values of the D2/df exceeded 2.78. Hence, no multivariate outliers are detected (Hair et al., 2010, Ch. 2). Four multivariate assumptions can be tested on the metric variables which potentially affect every multivariate statistical technique (Hair et al., 2010, Ch. 2): normality, homoscedasticity, linearity and the absence of correlated errors. Regarding the normality assumption, the data is not normally distributed since out of the 40 variables, 31 proved significantly skewed and 20 proved to be significantly platykurtic or leptokurtic (Appendix E). Fortunately, the impact of non-normal distributions of variables in sample sizes exceeding 200 effectively diminishes (Hair et al., 2010, Ch. 2; Tabachnick & Fidell, 2001), while departures from the normal distribution often occur with large sample sizes (Pallant, 2005). Confirmatory factor analysis is also relatively robust to violations of normality under sufficient sample size (Gorsuch, 1983 in: Floyd & Widaman, 1995). However, PLS is sensitive to asymptotic data (Vinzi et al., 2009), which is why a third root transformation was applied to mobile data usage and SMS usage, and a fourth root transformation was applied to voice usage, since these variables has long tails in the distribution. Problems in homoscedasticity are often a result of violation of normality (Hair et al., 2010, Ch. 2). Homoscedasticity and linearity can be assessed by studying scatter plots between two variables. Linearity and homoscedasticity were not assessed because doing so was not pragmatic considering the high number of variables. The absence of correlated errors applies to data collection in groups and time series data (Hair et al., 2010, Ch. 2) due to e.g. a nested data structure. The research design of this study does not encompass these methods, thereby decreasing the probability of correlated errors. Therefore this assumption was not checked.
4.2 Measurement model
4.2.1 Construct reliability Reliability was assessed by examining factor loadings, Cronbach’s alpha and composite reliability. The factor loadings and cross-loadings of the initial measurement model including all directly measured items can be viewed in Appendix F. All individual factor loadings exceeded .70 except for Perceived usability2 (.217) and Utilitarian attitude4 (.575), which were deleted from the model. Cross-loading items were Hedonic atttiude5 with the construct Utilitarian attitude (.719) and Utilitarian Attitude2 with Hedonic attitude (.704). In order to ensure sufficient discriminant validity, these items were deleted from the analysis which can be justified since sufficient items remain to measure hedonic and utilitarian attitude. After these deletions, the model was tested again. The corresponding factor loadings and cross-loadings are illustrated in Appendix G. All items had loadings exceeding .70 on the corresponding constructs and no cross-loadings could be observed. Cronbach’s alpha for the multi-item constructs ranged from .84-92 and composite reliability ranged from .88-.97, as displayed in Table 2. Hence, construct reliability is warranted for the latent constructs. It should be noted that reliability values and the average variance extracted for single item
29
Table 2 Measurement properties
Cronbach's Alpha Composite Reliability R2* .44 .05 .10 .14 .04 .12 .33 .35 .04 .12 .01 R2** .44 .05 .10 .14 .10 .52 .47 .35 .04 .12 .01
Variable
Mean
St. dev.
AVE .74 .91 .72 .86 .76 .80 .84 -
VIF 1.25 2.46 1.61 1.68 2.11 1.32 2.50 1.26 2.54 2.05 1.97 1.61 1.06 1.18 2.13 1.10
Apps 4.08 2.61 Cust. Sat. Smartphone 7.65 2.20 Daily Usage Time 3.77 1.49 Frequency Of Use 2.78 1.38 Hedonic Attitude 5.41 1.06 .88 .92 Mobile Data Usage 5.25 1.89 Perceived Brand Image 5.46 1.34 .95 .97 Perceived Effort 2.93 1.29 .81 .88 Perceived Usability 5.52 1.13 .92 .95 Perceived Usefulness 5.25 1.13 .84 .91 Perceived Usage 5.65 1.24 Subjective Knowledge 4.49 1.32 .92 .94 SMS Usage 3.33 1.30 Telco Cust. Sat. 7.27 1.52 Utilitarian Attitude 5.94 0.80 .91 .94 Voice Usage 2.80 0.70 *R2 of the base model without the direct effect of the moderating variables 2 **R of the model including the direct effects of the moderating variables
constructs equals 1.00 because the items that load on the constructs have a factor loading of 1.00 , which is because the constructs are measured directly. These values are not depicted in Table 2. 4.2.2 Construct validity and multicollinearity Both convergent validity and discriminant validity were assessed. For sufficient convergent validity, values for the average variance extracted (AVE) should exceed .50, which indicates that the latent variable explains more than half of its indicators’ variance (Hair et al., 2011). In Table 2 it can be viewed that the lowest value for the AVE of multiitem constructs was .72 for perceived effort, meaning that convergent validity is sound. Regarding discriminant validity, the Fornell-Larcker criterion was employed and in the previous section, cross-loading items were deleted. When the Fornell & Larcker criterion is met, a latent construct shares more variance with the indicators loading on the construct under scope than with other latent variables in the model (Fornell & Larcker, 1981). In other words, the square root of the AVE should exceed the correlation with any other latent construct. As can be seen in Table 3, this was the case for all multi-item constructs. Therefore, based on the Fornell-Lacker criterion and the absence of cross-loading items, the discriminant validity of the latent constructs in the measurement model is acceptable. Regarding the single-item constructs, no correlations approached 1.00, which is also an indication of the discriminant validity of the single-item constructs. Finally, Multicollinearity was not an issue since no value for the variance inflation factor (VIF) above 5.0 could be detected (Hair et al., 2010), as can be seen in table 2.
4.3 Structural model
For the base model, the significance of the standardized effects sizes are assessed. This is followed by the analysis of moderator variables and control variables.
30
Table 3 Correlation matrix of latent variables and Fornell-Larcker test
Variable 1. 2. 3. 4. 5. 6. 7. 1. Apps 1.00* 2. Cust. Sat. Smartphone .32 1.00 3. Daily Usage Time .06 .04 1.00 4. Frequency Of Use .15 .18 .49 1.00 5. Hedonic Attitude .17 .11 .3 .29 .86 6. Mobile Data Usage .18 .11 .22 .31 .13 1.00 7. Perceived Brand Image .25 .68 .03 .15 .27 .07 .95 8. Perceived Effort -.2 -.32 .06 -.02 -.07 -.08 -.26 9. Perceived Usability .35 .62 .14 .26 .36 .23 .64 10. Perceived Usefulness .27 .49 .19 .3 .37 .25 .58 11. Perceived Usage .23 .34 .41 .42 .39 .3 .4 12. Subjective Knowledge .3 .14 .19 .34 .47 .31 .26 13. SMS Usage -.03 -.02 -.01 .05 0 .08 0 14. Telco Cust. Sat. .03 .31 .02 .01 .14 -.03 .26 15. Utilitarian Attitude .22 .22 .23 .27 .68 .15 .35 16. Voice Usage -.06 .09 -.02 .04 -.04 .12 0 *Diagonal elements in bold are square roots of the AVE values for latent constructs. 8. 9. 10. 11. 12. 13. 14. 15. 16.
.85 -.33 -.15 -.04 -.04 .03 -.16 -.08 -.01
.93 .57 .51 .37 .01 .19 .47 -.01
.87 .52 .41 -.13 .22 .46 -.04
1.00 .4 -.03 .08 .46 .02
.89 .04 0 .37 -.02
1.00 -.02 -.03 .03
1.00 .19 .08
.92 -.05
1.00
4.3.1 Base model The base model includes the main concepts without the direct effects of moderators. The explained variance or the R2 of the dependent variables in the base model can be found in Table 2. The R2 must be evaluated for individual studies (Backhaus et al., 2003 in: Götz et al., 2010). Hair et al. (2010) states that R2 values of respectively .25, .50 and .75 can be described as weak, moderate and substantial as a rule of thumb for marketing research studies. However, the author believes this study could be placed in the category of consumer behavior studies, of which Hair et al. (2011) states that in this specific discipline, R2 values of .20 can be considered as high. A meta-analysis of Arts et al. (2011) on consumer adoption considers that an explained variance of .10 for actual adoption behavior explains variance “to a lesser extent” (Arts et al., 2011). The dependent variables of which the highest amount of variance was explained was customer satisfaction towards the smartphone (.44), perceived usage (.35) and perceived usefulness (.33). Moderately explained variables were mobile data usage (.14), perceived usability (.12), telecommunication company customer satisfaction (.12) and frequency of use (.10). Finally, weakly explained variables were daily usage time (.05), perceived effort (.04), SMS usage (.04) and voice usage (.01). The results of the testing of the hypotheses can be viewed in Table 4. Hypotheses were supported if the predicted direction of the effect was correct and if the effect was significant at the .05 level. Of the 40 hypotheses in the base model, 13 hypotheses were confirmed, 24 effects were non-significant at the .05 level and 2 effect sizes were significant but the effect size was in a direction contradictory to the hypothesis. The following statements can be made about the confirmed hypotheses: Consumers that have more apps are more likely to perceive their smartphone to be more easy to use (H1b,alt, ? = .35, t = 6.41), to perceive less effort (H1c,alt, ? = -.20, t = 2.97) and to be more satisfied about their smartphone (H1d, ? = .09, t = 2.30). In turn, customers who perceive their smartphone as useful are disposed to have a lower degree of SMS usage (H3a, ? = -.22, t = 2.69), to use their smartphone more frequently (H4b, ? = .22, t = 2.55), to perceive to use it more (H4c, ? = .34, t = 3.78) and to be more s atisfied with their smartphone (H5a, ? = .20, t = 2.68). In turn, perceived usability is positively associated with perceived usefulness (H2, ? = .54, t = 10.09), but also with consumers’ perception of their
31
Table 4 Hypothesis testing
Hypothesis Relation ?† T-Value H1a (+) Apps ? Perceived Usefulness .09 1.66* H1b (-) Apps ? Perceived Usability .35 6.41*** H1b,alt (+) Apps ? Perceived Usability .35 6.41*** H1c (+) Apps ? Perceived Effort -.20 2.97*** H1c,alt (-) Apps ? Perceived Effort -.20 2.97*** H1d (+) Apps ? Cust. Sat. Smartphone .09 2.30** H1e (+) Apps ? Telco Cust. Sat. -.10 1.38 H2 (+) Perceived Usability ? Perceived Usefulness .54 10.09*** H3a (-) Perceived Usefulness ? SMS Usage -.22 2.69*** H3b (-) Perceived Usefulness ? Voice Usage -.06 0.87 H3c (+) Perceived Usefulness ? Mobile Data Usage .08 1.04 H3d (-) Perceived Usability ? SMS Usage .12 1.57 H3e (-) Perceived Usability ? Voice Usage -.01 0.21 H3f (+) Perceived Usability ? Mobile Data Usage .04 0.48 H3g (+) Perceived Effort ? Mobile Data Usage -.05 0.85 H4a (+) Perceived Usefulness ? Daily Usage Time .15 1.72* H4b (+) Perceived Usefulness ? Frequency Of Use .22 2.55** H4c (+) Perceived Usefulness ? Perceived Usage Level .34 3.78*** H4d (+) Perceived Usability ? Daily Usage Time .10 1.14 H4e (+) Perceived Usability ? Frequency Of Use .15 1.83* H4f (+) Perceived Usability ? Perceived Usage Level .36 4.08*** H4g (+) Perceived Effort ? Daily Usage Time .12 1.26 H4h (+) Perceived Effort ? Frequency Of Use .06 0.86 H4i (+) Perceived Effort ? Perceived Usage Level .13 2.76*** H5a (+) Perceived Usefulness ? Cust. Sat. Smartphone .20 2.68*** H5b (+) Perceived Usefulness ? Telco Cust. Sat. .12 1.65* H5c (+) Perceived Usability ? Cust. Sat. Smartphone .44 5.86*** H5d (+) Perceived Usability ? Telco Cust. Sat. -.06 0.65 H5e (-) Perceived Effort ? Cust. Sat. Smartphone -.12 2.14** H5f (+) Perceived Effort ? Telco Cust. Sat. -.09 1.29 H6a (-) Daily Usage Time ? SMS Usage -.03 0.42 H6b (-) Daily Usage Time ? Voice Usage -.06 0.78 H6c (+) Daily Usage Time ?Mobile Data Usage .06 0.74 H6d (-) Frequency Of Use ? SMS Usage .11 1.49 H6e (-) Frequency Of Use ? Voice Usage .07 1.10 H6f (+) Frequency Of Use ? Mobile Data Usage .19 2.40** H6g (-) Perceived Usage Level? SMS Usage 0 0.06 H6h (-) Perceived Usage Level? Voice Usage .05 0.72 H6i (+) Perceived Usage Level? Mobile Data Usage .14 1.62 H7 (+) Cust. Sat. Smartphone ? Telco Cust. Sat. .29 3.66*** *
<.1. **<.05, ***<.01 (Two Tailed); † Effect Sizes Are Standardized Supported? No No Yes No Yes Yes No Yes Yes No No No No No No No Yes Yes No No Yes No No Yes Yes No Yes No No No No No No No No Yes No No No Yes
usage level (H4f, ? = .36, t = 4.08) and customer satisfaction of the smartphone (H5c, ? = .44, t = 5.86). As the final aspect of value in the model, perceived effort is positively associated with consumers’ perception of the usage level (H4i, ? = .13, t = 2.76), but negatively with customer satisfaction towards the smartphone (H5e, ? = -.12, t = 2.14). Finally, customers that use their smartphone more often on a daily basis, tend to use more mobile data (H6f, ? = .19, t = 2.40), and customer that are satisfied with their smartphone are more often satisfied with their telecommunication provider (H7, ? = .29, t = 3.66). As in Turel & Serenko (2006), the statistical validity of the significant linkages was assessed by deleting the insignificant effects from the model. Comparison of ?’s, t-values and R2 values can be viewed in Appendix H. The deletion of the insignificant relations in the base model did not result in major changes of the R2’s, the largest change being a small to medium decrease in the R2 of mobile data usage from .14 to .09 (with a small effect size ƒ2 of -.05; see
32
the next section for the calculation of the effect size). The largest change in path coefficients was for frequency of use ? mobile data usage, of ? = .19 to ? = .31 . No relations changed from being significant to being insignificant on the .05 level. Based on these results, the statistical validity of the significant relations was further confirmed. 4.3.2 Moderating variables The results of the moderator analysis can be found in Table 5. Seven direct effects of moderators have been detected: of subjective knowledge about smartphones on the perceived usability (? = .11, t = 1.97) and on perceived usefulness (? = .18, t = 2.78); of consumers’ utilitarian attitude towards smartphones on perceived usability (? = .23, t = 3.41) and on the perceived usefulness of the smartphone (? = .18, t = 2.50). Finally, perceived brand image was directly associated with all three aspects of value: negatively with perceived effort (? = .25, t = 3.03), but positively to perceived usability (? = .50, t = 9.87) and perceived usefulness (? = .34, t = 4.03). Consumers’ hedonic attitude towards smartphones did not have any direct effects. Only one moderating effect was found to be significant: the effect of subjective knowledge on the relation between the number of apps and the perception of usability (H8b, ? = -.12, t = 2.00), which weakly increases the explained variance of perceived ease of use (ƒ2 = .02). The negative value of the coefficient of the interaction term indicates that consumers’ subjective knowledge diminishes the positive effect of the number of apps on the perceived ease of use, i.e. consumers who believe they are more knowledgeable about smartphones believe their smartphone is less easy to use when they have more applications, than consumers who believe they are less knowledgeable about smartphones. The other hypotheses on moderators could not be confirmed.
Table 5 Moderator analysis results
Direct effect Relation Including/excluding moderating effect R
2
ƒ
2
Moderator direct effect ?† .08 .08 .11 .22 .18 .18 -.04 -.25 0 -.01 0 0 .04 .02 .23 .22 .18 .17 -.25 -.24 .5 .49 .34 .34 T-value 1.14 1.09 1.97** 3.28*** 2.78*** 2.85*** 0.49 3*** 0.07 0.2 0.02 0.02 0.39 0.17 3.41*** 3.07*** 2.50** 2.17** 3.03** 2.93*** 9.87*** 8.66*** 4.03*** 4.2***
Interaction term effect ?† .06 -.12 -.02 .01 -.08 .01 -.07 -.04 -.02 .02 -.02 -.03 T-value 0.58 2.00** 0.25 0.09 1.38 0.10 0.69 0.55 0.20 0.20 0.38 0.45
?† Excluding Including Excluding Subjective Knowledge ? (Apps ? Perceived Usability) Including Excluding Subjective Knowledge ? (Apps ? Perceived Usefulness) Including Excluding Hedonic Attitude ? (Apps ? Perceived Effort) Including Excluding Hedonic Attitude ? (Apps ? Perceived Usability) Including Excluding Hedonic Attitude ? (Apps ? Perceived Usefulness) Including Excluding Utilitarian Attitude ? (Apps ? Perceived Effort) Including Excluding Utilitarian Attitude ? (Apps ? Perceived Usability) Including Excluding Utilitarian Attitude ? (Apps ? Perceived Usefulness) Including Excluding Perceived Brand Image ? (Apps ? Perceived Effort) Including Excluding Perceived Brand Image ? (Apps ? Perceived Usability) Including Excluding Perceived Brand Image ? (Apps ? Perceived Usefulness) Including *<.1. **<.05, ***<.01 (two tailed); † effect sizes are standardized Subjective Knowledge ? (Apps ? Perceived Effort) .10 .10 .52 .53 .47 .47 .10 .10 .52 .52 .47 .47 .10 .10 .52 .52 .47 .47 .10 .10 .52 .52 .47 .47 .00 .02 .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 -.17 -.19 .14 .17 .03 .03 -.17 -.17 .14 .16 .03 .03 -.17 -.15 .14 .15 .03 .03 -.17 -.17 .14 .14 .03 .03
T-value 2.47*** 2.82*** 3.13*** 3.41*** 0.57 0.58 2.47** 2.64*** 3.13*** 3.27*** 0.57 0.49 2.47** 2.31** 3.13*** 2.88*** 0.57 0.56 2.47** 2.43** 3.13*** 3.11*** 0.57 0.56
33
4.3.3 Control variables The effects of control variables were tested by entering all control variables in the path model simultaneously. The bootstrap procedure was conducted using 500 resamples. The effects that were found to be significant at the .05 level can be found in Table 6. From the path model including the control variables, it appears that older people tend to be more satisfied with their smartphone (? = .10, t = 1.96), to use it for a shorter period of time (? = -.29, t = 5.27) and less frequent (? = -.5, t = 4.09) on a daily basis, to have a lower usage of mobile data (? = -.38, t = 7.01) and of text messages (? = -.14, t = 2.05), perceive to use their smartphone less (? = -.18, t = 3.71), say to know less about smartphones (? = -.15, t = 2.37) and perceive their smartphone as less useful (? = -.15, t = 2.88). As opposed to male smartphone users, female smartphone users more often indicate to have less apps installed on their device ( = -.16, t = 2.53) and to know less about smartphones (? = -.23, t= 3.61). Higher educated people often use their smartphone for a shorter period of time on a daily basis (? = -.13, t = 1.99). Regarding the smartphone brands, Blackberry users seem to differ from Apple users since they tend to have less apps on their smartphone (? = -.23, t = 4.53), to use less mobile data (? = -.12, t = 2.17), to have a lower perception of the image of their smartphone’s brand (? = -.28, t = 2.40), to spend more effort on installing, maintaining and learning to use the apps on their Blackberry (? = . 19, t = 2.34) but appeared to use it longer on a daily basis (? = .16, t = 2.26) and to perceive to use it more (? = .14, t = 2.07). The only other significant difference between users of different smartphone brands was that Samsung users often have more apps on their device than Apple users (? = .13, t = 2.02). Respondents that completed the survey on a smartphone are more likely to spend more time on their smartphone (? = .15, t = 2.25) and to use it more often (? = .13, t = 2.05) per day, to perceive to use their smartphone more (? = .14, t = 2.78) and to perceive it as easy to use (? = .10, t = 2.45). On the other hand, respondents having completed the survey on a tablet are more likely to be satisfied with their smartphone (? = .08, t = 2.04). Another finding is that consumers that use their smartphone for both business and consumer purposes tend to use more mobile data (? = .15, t = 2.44), say more often that their smartphone is useful (? = .09, t = 2.12) and state more often to know more about smartphones (? = .17, t = 3.29) than consumers using their smartphone only for consumer use. Moreover, business and consumer users tend to have a higher SMS usage (? = .17, t = 2.79). No significant differences were detected for the number of smartphones that people have owned or the time that respondents have owned a smartphone.
4.4 Post-hoc analyses
Based on the findings of the initial results, two post-hoc analyses were conducted. The motivation and findings are discussed of an analysis of mediating variables and a comparison between the different operating systems that consumers use on their device. 4.4.1 Mediating variables Only one weak moderating effect of subjective knowledge was found in the analysis of moderating variables. Neither the hedonic nor utilitarian attitude of consumers towards smartphones, nor the perception of the image of the smartphone brand appeared to influence
34
Table 6 Significant effects of control variables
Relation Of Control Variable ?† T-Value Relation Of Control Variable ?† Age ? Cust. Sat. Smartphone .1 1.96** Samsung ? Apps .13 Age ? Daily Duration Of Use -.29 5.27*** Education ? Daily Duration Of Usage -.13 Age ? Frequency Of Usage -.25 4.09*** Female ? Apps -.16 Age ? Mobile Data Usage -.38 7.01*** Female ? Subjective Knowledge -.23 Age ? Perceived Usefulness -.15 2.88*** Survey On Smartphone ? Daily Duration Of Usage .15 Age ? Perceived Usage Level -.18 3.71*** Survey On Smartphone ? Frequency Of Usage .13 Age ? Subjective Knowledge -.15 2.37** Survey On Smartphone ? Perceived Usability .09 Age ? SMS Usage -.14 2.05** Survey On Smartphone ? Perceived Usage Level .14 Blackberry ? Apps -.23 4.53*** Survey On Tablet ? Cust. Sat. Smartphone .08 Blackberry ? Daily Duration Of Use .16 2.26** Business And Pleasure Use ? Mobile Data Usage .15 Blackberry ? Mobile Data Usage -.12 2.17** Business And Pleasure Use ? Perceived Usefulness .09 Blackberry ? Perceived Brand Image -.28 2.4** Business And Pleasure Use ? Subjective Knowledge .17 Blackberry ? Perceived Effort .19 2.34** Business And Pleasure Use ? SMS Usage .17 Blackberry ? Perceived Usage Level .14 2.07** **<.05, ***<.01 (two tailed); † effect sizes are standardized Reference categories of dummy variables: ‘apple’ for smartphone brand, ‘male’ for gender, ‘pc/laptop’ for survey method and ‘pleasure use only’ for type of use. T-Value 2.02** 1.99** 2.53*** 3.61*** 2.25** 2.05** 2.29** 2.78*** 2.04** 2.44** 2.12** 3.09*** 2.79***
relations between the number of apps installed on the smartphone and the aspects of value. However, direct effects were detected of perceived brand image, utilitarian attitude and subjective knowledge. These direct effects give rise to the possibility that these constructs could play a mediating role in the relation between the number of apps and the aspects of perceived value, i.e. when adding these constructs in the model, they could very well reduce the strength of the effects of installed apps on the aspects of value. Additionally, an analysis of the possible mediating role of these variables could lead to the detection of spurious relations between the number of apps and the aspects of value. In other words, the variables that were initially proposed to moderate the relation between the number of apps and the aspects of value, could also be confounding variables to the relations between the number of apps and the aspects of value. The analysis of mediating effects was conducted according to Baron & Kenny (1986): 1) the independent variable was to be a significant predictor of the dependent variable; 2) the independent variable was to be a significant predictor of the mediator variable; and 3) the mediating variable was to be a significant predictor of the dependent variable, while controlling the mediator with the independent variable. In the last step, the effect of the independent variable was to be smaller in absolute value than in the first step, for partial mediation to exist. The significance of the mediating effect can be assessed with Sobel’s test2 using the unstandardized beta’s.3 In the base model, the number of apps on the smartphone was significantly related to the perceived usability of the smartphone and the perceived effort required for installing, updating and learning to use those apps. Therefore, these two relations qualify for being mediated. The model that was used for testing the mediating effects was the model in which the insignificant effects were deleted (‘the bare model’), described at the end of section 4.3.1.
2
where a and ss refer to the unstandardized coefficient and standard error of the relation
between the independent variable and the mediator, and b refers to those parameters of the relation between the mediator and the dependent variable. 3 Calculated by multiplying the standardized coefficient by the standard deviation of the predicted variable and dividing it by the standard deviation of the predictor variable (Bontis et al., 2007).
35
The R2 value for perceived brand image was .06, for hedonic attitude towards smartphones the value was .03, .05 for utilitarian attitude and .09 for subjective knowledge. As can be viewed in Appendix I, it appears that the number of apps is significantly related to all four postulated mediating variables: hedonic attitude (? = .17, t = 3.01), utilitarian attitude (? = .22, t = 4.01), perceived brand image (? = .25, t = 3.92) and subjective knowledge (? = .30, t = 5.09). However, only three of the mediating variables had significant effects on the predicted variables: perceived brand image on perceived effort (? = -.25, t = 3.22) and on perceived usability (? = .50, t = 10.08) and utilitarian attitude on perceived usability (? = .23, t = 3.43). It also appeared that the coefficient of the relation between the number of apps and perceived effort decreased slightly from ? = -.20 in the bare model to ? = -.17 in the mediator model. Moreover, the coefficient of the relation between the number of apps and the perceived usability of the smartphone decreased from ? = .35 to ? = .14. Hence, the relations between the number of apps, and perceived effort & usability are partially mediated. The assessment of the significance of the partially mediating relations still remained. One significant mediating relation was detected with a Sobel value corresponding to p < .05: the relation between the number of apps and the perceived usability was partially mediated by consumers’ utilitarian attitude towards the smartphone (Sobel z-value = 1.96). See Appendix I for the calculation of these values. 4.4.2 Differences between operating systems Another analysis that was conducted post-hoc was the assessment of differences between the main variables in the model for the different operating systems that the smartphones of different brands run on. Apple’s smartphones run on iOS, HTC & Samsung devices operate with Android and Blackberry handsets use their own operating system, from herewith referred to as BOS. The motivation of this analysis is as follows: in the assessment of the control variables, differences were detected with Apple as the reference category. However, no direct differences between Blackberry, HTC and Samsung were analyzed. To assess these differences with SmartPLS with dummy variables would be time consuming. Therefore the differences are analyzed using a one-way MANOVA in SPSS. Since HTC and Samsung use the same operating system, these two brands have same user interface and the same apps available, which is why no differences between brands are analyzed but differences between operating systems. Moreover, the users the operating systems of the smartphone are easier to target by mass-media advertising campaigns from telecommunication operators than other segments based on the criteria such as age, gender or education. Therefore, a MANOVA is conducted to assess differences between the three operating systems and the main variables in the structural model, i.e. the number of apps, customer satisfaction with the smartphone and the telecom provider, and the 6 usage variables. Before describing the results of the analysis it should be noted that the group sizes are unequal: 47 for iOS, 34 for BOS, and 167 for Android. Also, the assumption of normality is violated for customer satisfaction with the smartphone and the telecom operator, the number of apps, daily usage time, frequency of use and the perceived usage level. However, mainly a platykurtic distribution leads to reduced power (Hair et al., 2010, Ch. 8), which is not the case
36
Table 7 MANOVA results - Differences per operating system
Variable Number Of Apps Cust. Sat. Smartphone Daily Duration Of Use Frequency Of Daily Use Perceived Usage Level Mobile Data Usage SMS Usage Telecom Operator Cust. Sat Voice Usage Degrees of freedom 2 2 2 2 2 2 2 2 2 F-value 15.32 21.67 15.59 11.19 3.26 2.2 0.05 1.22 1.5 Partial eta squared .114 .155 .063 .01 .027 .018 .000 .010 .013 Significance .000 .000 .000 .306 .040 .113 .947 .300 .224 Power .999 1.000 .954 .259 .616 .446 .058 .265 .319
for the variables that are assessed (see Appendix E). Moreover, the results of the analysis of the control variables can be used to support findings of the MANOVA since PLS is robust to violations of normality (Hair et al., 2010, Ch. 8). Additionally, Pillai’s trace was used as test criterion because this statistic is robust to violations of normality and unequal group sizes. Since Box’s M test proved significant at the .01 level, indicating unequal variances across groups, Games-Howell test were used for the post-hoc analyses since this test offers the best performance under unequal variances across groups and unequal sample sizes (Field, 2009). The one-way MANOVA revealed a significant multivariate main effect for operating system. Pillai’s Trace had a value of .375, F(18,460) = 5.90, p< .01, partial eta squared = .188. The observed power to detect the difference was 1.00. For assessing which variables significantly differed for the groups, a Bonferroni-adjustment was made by dividing the .05 significance by 9 since there were 9 independent variables. The significant univariate main effects are illustrated in Table 7. It appeared that the number of apps, satisfaction with the smartphone and the daily usage duration of the smartphone differed per operating system. From post-hoc analyses on these three variables (see Appendix J), several significant differences between the operating systems were observed for these variables. Regarding the number of apps, Blackberry users (mean in the category 11-20 apps) had less apps than Apple or Android users (means in the category 41-50 apps). Customer satisfaction with the device differed significantly between Apple users on one hand, who had a mean score for satisfaction of 9.05, and Android and Blackberry users on the other, which had mean scores for customer satisfaction of 7.62 and 6.06, respectively. The post hoc analysis did not reveal significant differences between the operating systems for daily duration of use. Due to the differences in group sizes, it should be noted that these findings should be interpreted with caution. Differences between brands that were detected in the control variable analysis, but not in the MANOVA, were mainly between Apple and Blackberry users for the variables daily duration of use, mobile data usage and the perceived usage level.
5 Discussion
This chapter will discuss the research questions posed in the introduction. The discussion of the findings is accompanied by a reflection on the results and is followed by implications for scholars and managers. Finally, the limitations of this study are discussed, which offer directions for further research. The results are summarized in Figure 3.
37
Subjective knowledge
Utilitarian attiude (5%)
-.22***
.22***
.23***
SMS usage (4%) Perceived usefulness (33%)
.22*** -.12**
Number of apps .54***
.34*** .35***
Perceived usability (12%)
Daily frequency of use (10%)
.19**
Mobile data usage (14%)
.36*** -.20***
Perceived effort (4%)
Perceived usage (35%)
.13***
.44*** -.12** .09**
.20***
Customer satisfaction smartphone (44%)
.29***
Telecom provider satisfaction (11%)
Figure 3 PLS results. Between brackets within the variables is explained variance. ** significant at the .05 level, ***significant at the .01 level (two-tailed). Note that insignificant effects and effects that were significant at the .10 level were omitted for the sake of overview. The narrow-dashed line represents a moderating effect, the broad-dashed line represents a mediating effect
5.1 Findings
This writing was initiated in order to investigate the effects that the number of apps installed on a smartphone can have on smartphone consumer behavior. Specifically, the effects of apps on usage and satisfaction were assessed with regard to the smartphone, the telecommunication operator and the services provided by the operating company. It was also examined which of two competing concepts is most applicable to smartphone consumption: feature fatigue or mass customization. According to the results of this study, not feature fatigue but mass customization is the appropriate concept to explain the effect of the number of apps on the aspects of value perceived by consumers. Apps were not found to relate to utility, but did appear to be positively related to usability and negatively to the perceived effort required to install, update and learn how to use the apps. Since feature fatigue predicts that more features lead to decreased usability but increased utility, the phenomenon is not applicable to smartphones. The concept of mass customization is applicable though, with the underlying rationale that consumers that alter the functionality of their device and have a high number of apps, often have control over their device and perceive it as easy to use. Another explanation could be that users with a high number of apps, also use more apps that improve the usability of the device. It should be noted that the positive association between the number of apps and usability probably has its limits, since a very high number of apps is likely to decrease the processing speed of the smartphone, thereby retarding the device. Mass customization can also be applied to explain the negative association between the number of apps and the effort that the apps on the smartphone require. Although the degree of explained variance of effort
38
was low, consumers that have a high number of apps seem to customize their device and perceive a lower degree of effort than consumers that have less apps installed. This could be because users can customize their device according to their own preferences. Doing so, they adapt the number of apps to match their demand of effort they want to spend on installing, updating and learning to use those apps and they perceive the degree of effort accordingly. The relation could also be negative because consumers with more apps are more effective at managing the apps on their device, and hence perceive a lower degree of effort. The absence of a relation between the number of apps and perceived utility was unexpected. This can probably lead to the conclusion that not each additional app necessarily adds to the utility of the device, giving rise to the idea that distinction should be made between different types of apps. This is further discussed in the limitations section. One relation among the aspects of value is more or less traditional in technology acceptance literature (e.g. Venkatesh & Davis, 2003): usability positively relates to utility. Since usability is related to apps, utility is indirectly associated with the number of apps on consumers’ smartphones. Through value, apps affect the usage of the smartphone and services offered by telecom operators. All aspects of value are positively related to the perception of the usage level, which was the only aspect of usage of which the hypotheses were all correct and therefore the effects are not surprising. However, none of the aspects of value was related to the time that consumers spend daily on their smartphone, and only the utility of the smartphone was associated to the frequency of daily use. It may be stated that increased usability, usefulness and effort thus relate to the extent to which people say they use their device, but that only the actual utility is related to how often people say to use their device on a daily basis. This could indicate that apps that are actually perceived as useful, are used more often, but not necessarily more in terms of daily duration. The perceived utility of the device also plays a key role, and the only role, in predicting increases or decreases in usage of services provided by operating companies. Increases in utility explain a small degree of SMS usage in the sense that people who believe their smartphone is useful, often have a lower SMS usage. Moreover, the frequency of daily use which is predicted by utility, is the only form of usage that is related to usage of mobile data. The number of calling minutes is not at all affected in the model, and the explained variance was low. Apparently, consumers that believe their device is useful tend to use their device more frequently, which leads to the use of mobile data. The decrease in SMS usage and the increase in data usage, which both (in)directly relate to utility of the smartphone, fits the transition in the telecommunications sector from SMS and voice usage to mobile data (Tilson & Lyytinen, 2006). The absence of relations between usage of the smartphone and voice usage could be explained by the measurement: the variables related to the use of the smartphone that are measured in the survey reflect use without calling behavior, which could be a reason why no relations were found between usage of the device and voice usage. It can be concluded that only the utility of the smartphone plays a central role in actual usage of the smartphone. However, the utility is not related to the number of apps directly, but only through the perceived usability. Effort as a negative aspect of value does not play an important role in the model when it comes to predicting usage. People that perceive to spend a high degree of effort on installing, maintaining and learning to use those
39
apps do also perceive to use their phone more often, but the effort is not associated with selfreported usage or actual usage of telecom services. The finding that effort only affects the perceived usage indicates that the relation between effort and usage is mostly in the mind of the consumers. The discussion of the effects of apps on customer satisfaction remain. The number of apps is directly and positively related to customer satisfaction with the device. Indirect effects are also found: usability and usefulness are positively related to satisfaction, but effort is negatively related. The relation between usability and satisfaction with the device is in line with findings of Choi & Lee (2012). Based on these findings, the conclusion is that smartphone users that customize their smartphone by adding apps, are often more satisfied with their device for a number of reasons. Consumers welcome the utility and usability, but despise the required effort for having apps. Fortunately, this effort perception is often lower when consumers have a high number of apps installed on their handset. Satisfaction with the smartphone is related to numerous variables in the model, but satisfaction with the operating company that allows consumers to use their device anytime, anywhere is not affected. Satisfied smartphone owners are generally more satisfied with their operator, but this satisfaction is not directly affected by the number of apps or the value of the smartphone. Apparently, consumers do not directly relate the value of their device to their satisfaction with the provider of their handset and of the services necessary to operate it. A moderating effect was detected for subjective knowledge about smartphones on the relation between the number of apps and usability: increases in the number of apps installed on the smartphone are associated with increases in the perceived usability of the device, but this is less so for consumers that consider themselves knowledgeable about smartphones. The direction of this moderating effect is contrary to expectations. This could be because consumers that are knowledgeable smartphone users are better able to evaluate the usability of products (Burson, 2007) and can provide a more accurate evaluation of the usability. It was expected that this bias was to be in favor of usability. However, it seems that expert users that can make better judgments on usability believe that their smartphone is not so easy to use when compared to novice users, if they have a high number of apps. Moreover, a partially mediating effect was found for the utilitarian attitude towards smartphones, i.e. the relation between the number of apps and the perception of usability is partly accounted for by consumers that believe that smartphones in general are useful devices, since these consumers both often have more apps and often believe their smartphone is easy to use. This is most likely because people who believe that the smartphone is a useful piece of technology, try to make optimal use of its utility by having a large number of apps installed, and that these consumers are biased to believe that the device is easy to use. The effects of control variables are conform other studies regarding age (Haverila, 2011a) and gender (Haverila, 2011b). The difference of perceptions between Blackberry users and other smartphone users could be because there are less apps for the Blackberry and because apps for this device often appear later in the market, or because the interface of the Blackberry differs greatly from iOS or Android, which are much more alike. This could also explain the differences in satisfaction that exists for consumers using smartphones with different operating systems that were detected with the post-hoc MANOVA, which also supports the findings that Blackberry users have less apps on their device.
40
5.2 Scholarly implications
The findings have several implications for scholars of consumer behavior or marketing. First of all, feature fatigue is not applicable to modular products such as smartphones. The shift in preferences of utility over usability before purchase, to usability over utility after use is no issue for smartphones. Probably this also applies to similar products using apps such as tablets or smart televisions. If consumers think their device is not usable because of the high number of apps, these are easily deleted and consumers no longer have to be frustrated with an unusable product with static functionality and usability. However, this does not necessarily mean that users of devices that use apps cannot be ‘fatigued’ by a high number of apps, only that the number of apps is not negatively related to decreases in usability. For instance, consumers can be overwhelmed by product choice (Kamis et al., 2008), which could harm their usage experience. Scholars wanting to study feature fatigue for smartphones or other devices that use apps, should bear in mind that consumers could still be irritated by apps they don’t use. As opposed to feature fatigue, mass customization is a concept that can be adopted when studying effects of apps on smartphones: scholars are encouraged to predict that a high number of apps is positively related to usability and less to effort, based on the concept of mass customization. The same predictions could be made for other products using apps, based on the findings of this study. Secondly, the effects of the number of apps on usability and effort proved to be relatively stable. Subjective knowledge weakly moderated the relation between the number of apps and usability, and the same relation was partially mediated by consumers’ utilitarian attitude. However, consumers’ hedonic attitude towards smartphones and the image of the brand in consumers’ minds did not moderate or mediate the relations between the number of apps and value. For scholars this implies that they should look for other possible spurious relationships or that the relation between the number of apps and usability is stable except for subjective knowledge and consumers’ utilitarian attitude. Thirdly, the conceptual framework that was used proved useful for modeling the effects of apps on consumer behavior for post-purchase processes. The integration of the Technology Acceptance Model in a model of customer satisfaction with value as the standard for comparison yielded moderate to high degrees of explained variance for satisfaction levels, perceived usage, frequency of use and mobile data usage. Finally, combining traditional survey data with objective usage data of telecommunication services proved useful to study smartphone usage, which is in accordance with scholarly implications of Verkasalo et al. (2010).
5.3 Managerial implications
Several implications can be extracted from this research for managers. First of all, managers need not worry about usability issues for devices that use apps, due to a high number of apps. Apps are a relatively new but successful way of selling and consuming software on smartphones, but also on tablet PC’s and more recently developed televisions. The soon to be released Windows 8 will have an interface that looks much like the one currently used on several smartphones and tablets. This study did not detect negative consequences for usage or satisfaction of consumers that have a high number of apps for issues such as usability. Therefore, managers should not worry that consumers that have
41
dozens of apps suffer from e.g. usability issues when it comes to the perception of value of the smartphone and indirect usage and satisfaction. In fact, it is the other way around: consumers that have a high number of apps tend to perceive their device to be easy to use. Consumers can do almost everything with apps. For example, by placing their smartphone on their mattress, they can use it to track either their sleep cycle, but also their sex life, or even both, with the Sleep Cycle Alarm Clock and the SexTrack app. The same device provides you with the best route through the supermarket when doing groceries with an app of a Dutch supermarket. This example illustrates that consumers with more apps, probably have more applications for using their smartphone, and often believe that their omnipotent device is easy to use despite the diverse range of applications. This effective form of mass customization is an important opportunity for app developers, who should come up with even more creative applications for smartphones. Mass customization is also an opportunity for managers in the telecom industry, who could ask consumers what they want to do with their smartphone, and then could install those prior to them receiving the device as an extension of their service. However, users should be able to delete these apps if they desire to do so, which is in line with mass customization. Secondly, in the context of the current transition in telecommunications from voice and SMS to data (Tilson & Lyytinen, 2006), carriers should encourage the use of apps in order to stimulate the use of mobile data. This is especially so for apps that promote more frequent interaction with the smartphone, rather than longer interaction, since mobile data usage is associated with frequency of use. Examples of such apps are Whatsapp, Facebook or news apps, since these are probably frequently checked by smartphone users when compared to game apps which probably demand longer interaction instead of more frequent interaction. Moreover, marketers and managers should allocate effort and resources to promote the rich collection of apps to consumers because of the positive associations of apps with the value of the device and satisfaction. This is contrary to recent findings of Tan et al. (2012) who state that manufacturers need not do so but should focus more on brand value. Thirdly, managers should bear in mind the antecedents to satisfaction. Since satisfaction affects customer loyalty and word-of-mouth (Szymanski & Henard, 2001) and ultimately, profitability (Hallowell, 1996), managers should be aware that apps, usability and usefulness positively influence satisfaction with the handset. However, the perceived effort by consumers that is required to install, update and learn how to use the apps inhibits satisfaction. Managers should consider to improve usability and usefulness perceptions and to automate the updating of apps in order to facilitate satisfaction with the device. This is a confirmation for Apple that their efforts to monitor the usability of the apps that are distributed via their Appstore are justified and lead to increased satisfaction with the iPhone. On the other hand, predictors to satisfaction with the handset are no predictors for satisfaction with the carrier, who often provides the handset in question. Consumers do not relate the number of apps or the value of the device to their satisfaction with the operating company. However, customers that are satisfied with their device do tend to be more satisfied with their carrier, which is why managers seeking to improve satisfaction with the telecom provider should consider converting existing customers that use Android or Blackberry devices to use Apple’s iPhone, or aim at providing new customers with iPhones. They could also encourage manufacturers of smartphones operating on Android to improve the usability
42
of their devices. Not only do Apple users tend to be more satisfied, they also join Android users in the number of apps they install on their device, when compared to Blackberry users. Marketers should adapt their marketing efforts according to the satisfaction levels experienced by customers of these handsets. Finally, managers should design online self-services that aid expert users in usability issues. This implication stems from the finding that expert users benefit less from increases in usability that are often associated with increases in the number of apps. Several telecom operators in the Netherlands offer start-up programs for new smartphone users in-store, but it is likely that users that consider themselves smartphone experts will not use these services because they already know how to use the device. Therefore, operators can design or improve online self-service programs that solve usability issues, especially for expert users who do not use these services in-store.
5.4 Limitations and avenues for further research
Several limitations impair the findings of this study and can be used to indicate directions for further research. For one, several aspects of the conceptual framework limit the research. A major limitation is the assumption that each app is equal to a feature and the fact that this research has not accounted for different categories of apps. The assumption is open for discussion when it comes to different categories of apps such as game apps. The discussion is whether a category of apps offer functionality, or whether each individual app can be seen as a different feature, e.g. if a user has 50 games installed on his or her smartphone, is each game experienced as a feature or is gaming as a category experienced as a feature of the smartphone? Moreover, previous research has found that different types of features for mobile phones are differently related to customer satisfaction (Haverila, 2011b). In addition, Verkasalo et al. (2009) find that mobile phone services such as apps should be addresses individually, instead of generalizing those services. Future research should differentiate between categories of features or apps. It could be that satisfaction and usage of apps and smartphones can be better explained per individual app and that each app purchase is evaluated separately by consumers. This view allows the use of the expectation/ disconfirmation paradigm of customer satisfaction in which customers expect different levels of value from different applications. Those expectations are then (dis)confirmed, leading to (dis)satisfaction. The satisfaction and usage levels stemming from different apps could then be assessed for apps individually, instead of aggregating the overall increases in satisfaction or usage levels stemming from apps. Another modeling issue is that feature fatigue can perhaps be used as a separate variable, as in Steiner et al. (2009), and to use the number of apps as a predictor and satisfaction and usage as a consequence of feature fatigue. Different models of customer satisfaction can also be explored for assessing feature fatigue such as ambivalent models in which customers can be both satisfied and dissatisfied (e.g. Olsen et al., 2005) or more dynamic models in which temporal interactions of consumers with their smartphone are adopted. Regarding the perception of value, future research could adopt monetary costs as negative aspect of value which was omitted in this research. Likewise, this research has not accounted for psychometric differences based on the adoption of innovations. Moore (2005) posits that adopters of innovations can be segmented
43
based on their psychological profile and the moment of adoption of the innovation. However, with smartphones having achieved about 53% penetration in the Netherlands in January 2012 (Telecompaper, 2012), not all psychological profiles are expected to be represented in proportion in a random sample of smartphone users. When smartphone penetration reaches higher levels, it can be interesting to examine whether adopters with different psychographical profiles differ in value perception, usage and satisfaction. Late adopters might suffer from decreases in usability due to a high number of apps, as opposed to early adopters. A conceptual weakness is also the post-hoc analysis of mediating effects. The analysis of mediating or moderating effects should be motivated by a sound theoretical basis (Baron & Kenny, 1986). This was not the case for the analysis of mediating effects, which was motivated by direct effects of moderators that were detected in the analysis of moderating effects. Future research could provide a sound theoretical basis for the mediating effect of utilitarian attitude between the number of apps installed and the perceived usability. A final conceptual issue is the use of the number of apps installed on consumers smartphones. Most likely, different results would be obtained if the respondents were asked to indicate how many apps they actually used frequently. A discrepancy between the number of apps used frequently and the number of apps installed could then be a predictor for the extent to which consumers are fatigued by apps they don’t use. A second class of limitations is related to the research design. Although PLS is stated to be robust to deviations from normality (Hair et al., 2010, Ch. 12), caution could be advised when interpreting the results from variables which strongly deviate from the normal distribution. When non-normal distributed data is used, a “markedly larger sample size is needed despite the inclusion of highly reliable indicators in the model” (Marcoulides & Saunders, 2006, p. vi). Although no undisputed guidelines exist to determine which violations of normality can be countered with which increases in sample size, the analysis could be carried out again with data that is more manipulated such that it better represents a normal distribution, or with a “markedly” larger sample size. Limiting aspects are also identified in the measurement model. For single item constructs, the measured item directly reflects the construct, which is then not latent (Fuchs & Diamantopoulos, 2009). Therefore, no measurement error is specified in the model. This research assumed that constructs such as satisfaction are straightforward for consumers; an assumption that is debatable. Future research on this topic could consider using multiple-item constructs for the single-item constructs used here in order to allow for a more accurate measurement. Regarding the measurement of the number of apps, it could be that differentiation is needed between pre-installed apps and apps that consumers have downloaded themselves; perhaps consumers experience these apps differently. Another remark on the measurement model is the use of self-reported measures. Although inherent to the use of a survey, the information provided by consumers about daily usage duration of the smartphone and daily frequency of use could be seriously misspecified because those measures are difficult to accurately estimate by respondents. Future research could ask respondents to track their use with an app in order to obtain more accurate numbers. Additionally, the data could also be collected in a cross-sectional and longitudinal manner in order to further examine causality of the effects (Hair et al., 2010), which cannot be done with the current dataset. For instance, the direction of the relations in the model are
44
not always straightforward and require further investigation: do users that believe their device is easy to use install a great number of apps, or do people that have many apps on their device therefore think it is also easy to use? Additionally, evidence of model stability and generalizability can only come from performing the analysis on additional samples and contexts such as other product categories (e.g. tablets), postpaid users, users across different cultures, customers of different carriers and users of other brands of smartphones. The sample seemed to reflect the operator’s population of smartphone users, but additional samples could be used in order to further determine the generalizability of the results.
6 References
6.1 Literature references
Ajzen, I. (1991). “The theory of planned behavior.” Organizational Behavior and Human Decision Processes, 50 (2), pp. 179-211 Al-Gahtani, S.S., King, M. (1999). “Attitudes, satisfaction and usage: contributing to each in the acceptance of information technology.” Behavior & Information Technology, 18 (4), pp. 277-297 Anderson, E.W., Sullivan, M.W. (1993). “The antecedents and consequences of customer satisfaction for firms.” Marketing Science 12 (2), pp. 125-143 Anderson, J.C., Narus, J.A. (1998). “Business marketing: Understand what customers value.” Harvard Business Review, 76 (6), pp. 53-65 Andrews, F.M. (1984). “Construct validity and error components of survey measures: a structur al modeling approach.” The Public Opinion Quarterly, 48 (2), pp. 409-442, in: Turel, O., Serenko, A. (2006). “Satisfaction with mobile services in Canada: an empirical investigation.” Telecommunications Policy, 30, pp. 314-331 Arruda-Filho, E.J.M., Cabusas, J.A., Dholakia, N. (2010). “Social behavior and brand devotion among iPhone innovators.” International Journal of Information Management, 30, pp. 475-480 Arts., J.W.C., Frambach, R.T., Bijmolt, T.H.A. (2011). “Generalizations on consumer innovation adopt ion: A meta-analysis on drivers of intention and behavior.” International Journal of Research in Marketing, 28, pp. 134-144 Athanassopoulos, A.D., Iliakopoulos, A. (2003). “Modeling customer satisfaction in telecommunications: assessing the effets of multiple transaction points on the perceived overall performance of the provider.” Production and Operations Management, 12 (2), pp. 224-245 Backhaus, K., Erichson, B., PLinke, W., Weiber, R. (2003). “Multivariate analysemethoden: eine anwendungsorientierte einführung.” 10thedition, Springer, Berlin in: Götz, O., Liehr-Gobbers, K., Krafft, M. (2010). “Evaluation of structural equation models using the Partial Least Squares (PLS) approach.” Handbook of Partial Least Squares, Springer Handbooks of Computational Statistics, Springer-Verlag Berlin Heidelberg Bandura, A. (1989). “Self-efficacy mechanism in human agency.” American Psychologist, 37 (2), pp. 122-147, in: Davis, F.D. (1989). “Perceived usefulness, perceived ease of use, and user acceptance of information technology.” MIS Quarterly, 13 (3), pp. 319-340 Baron, R.M., Kenny, D.A. (1986). “The moderator -mediator variable distinction in social psychology research: Conceptual, strategic, and statistical considerations.” Journal of Personality and Social Psychology, 51 (6), pp. 1173-1182 Batra, R., Ahtola, O. (1990). “Measuring the hedonic and utilitarian sources of consumer attitudes.” Marketing Letters, 2 (2), pp. 159-170 Bayraktar, E., Tatoglu, E., Turkyilmaz, A., Delen, D., Zaim, S. (2011). “Measuring the ef ficiency of customer satisfaction and loyalty for mobile phone brands with DEA.” Expert Systems with Applications, 39, pp. 99-106 Bergkvist, L., Rossiter, J.R. (2007). “The predictive validity of multiple -item versus single-item measures of the same constructs.” Journal of Marketing Research, 44 (2), pp. 175-184 Bontis, N., Booker, L.D., Serenko, A. (2007). “The mediating effect of organizational reputation on customer loyalty and service recommendation in the banking industry.” Management Decision, 45 (9), pp. 14261445 Brombacher, A.C. (2005). “Reliability in strongly innovative products; a threat or a challenge?” Reliability Engineering and System Safety, 88 (2) , pp. 88-125.
45
Burson, K.A. (2007). “Consumer-Product Skill Matching: The effects of difficulty on relative self-assessment and choice.” Journal of Consumer Research, 34, pp. 104-110 Cadotte, E.R., Woodruff, R.B., Jenkins, R.L. (1987). “Expectations and norms in models of consumer satisfaction.” Journal of Marketing Research, 24 (3), pp. 305-314 Chang, Y., Chen, C.S. (2004). “Smart phone – the choice of client platform for mobile commerce.” Computer Standards & Interfaces, 27, pp. 329-336 Chase, R.B., Jacobs, J.F., Aquilano, N.J. (2006). “Operations management for competitive advantage.” 11th edition, New York: Mcgraw-Hill/Irwin Chen, J.V., Ross, W., Huang, S.F. (2008). “Privacy, trust, and justice considerations for location -based mobile telecommunication services.” Info, 10 (4), pp. 30-46 Chen, J.V., Yen, D.C., Chen, K. (2007). “The acceptance and diffusion of the innovative smart phone use: A case study of a delivery service company in logistics.” Information & Management, 46, pp. 241-248 Childers, T.L., Carr, C.L., Peck, J., Carson, S. (2001). “Hedonic and utilitarian motivations for online retail shopping behavior.” Journal of Retailing, 77 (4), pp. 511-535 Choi, J.H., Lee, H. (2012). “Facets of simplicity for the smartphone interface: a structural model.” International Journal of Human-Computer Studies, 70 (2), pp. 129-142 Churchill, G.A., Surprenant, C. (1982). “An investigation into the determinants of customer satisfaction.” Journal of Marketing Research, 19 (4), pp. 491-504 Cohen, J. (1988). “Statistical power analysis for the behavioral sciences.” Hillsdale, NJ: Lawrence Erlbaum in: Henseler, J., Fassot, G. (2010). “Testing moderating effects in PLS path models: an illustration of available procedures.” Handbook of Partial Least Squares. Springer Handbooks of Computational Statistics. Berlin Heidelberg. Cooper, H. M. (1984). “The integrative research review: A systematic approach.” Applied social research methods series, 2, Beverly Hills, CA: Sage. In: Randolph, J.J. (2009). “A guide to writing the dissertation literature review.” Practical Assessment, Research & Evaluation, 14 (13), pp. 1-13 Cronin, J.J., Brady, M.K., Hult, G.T.M. (2000). “Assessing the effects of quality, value and customer satisfaction on consumer behavioral intentions in service environments.” Journal of Retailing, 76 (2), pp. 193-218 Davis, F.D. (1989). “Perceived usefulness, perceived ease of use, and user acceptance of information technology.” MIS Quarterly, 13 (3), pp. 319-340 Deng, Z., Lu, Y., Wei, K.K., Zhang, J. (2010). “Understanding customer satisfaction and loyalty: an empirical study of mobile instant messages in China.” International Journal of Information Management, 30 , pp. 289-300 Eggert, A., Ulaga, W. (2002). “Customer perceived value: a substitute for satisfaction in business markets?” Journal of Business & Industrial Marketing, 17 (2/3), pp. 107-118 Falaki, H., Mahajan, R., Kandula, S., Lymberopoulos, D., Govindan, R., Estrin, D. (2010). “Diversity in Smartphone usage.” MobiSys 2010: Proceedings of the 8th International Conference on Mobile Systems, Applications and Services, pp. 179-194 Feitzinger, E., Lee, H.L. (1997). “Mass customization at Hewlett-Packard: The power of Postponement.” Harvard Business Review, January-February 1997, pp. 115-122 Field, A. (2009). “Discovering statistics using SPSS. Third edition, SAGE Publication Ltd., London. Chapter 10: “Comparing several means: ANOVA (GLM1).” Fishbein, M., Ajzen, I. (1975). “Belief, attitude, intention and behavior: An introduction to theory and research.” Reading MA Addison-Wesley Flavian, C., Guinalíu, M., Gurrea, R. (2005). “The role played by perceived usability, satisfaction and consumer trust on website loyalty.” Information & Management, 43, pp. 1-14 Flynn, L.R., Goldsmith, R.E. (1999). “A short, reliable measure of subjective knowledge.” Journal of Business Research, 46 (1), pp. 56-66 Fornell, C. (1992). “A national customer satisfaction barometer: The Swedish experience.” Journal of Marketing, 56, pp. 6-21 Fornell, C., Bookstein, F.L. (1982). “Two structural equation models: LISREL and PLS applied to consumer exit-voice theory.” Journal of Marketing Research, 19, pp. 440-452 Fornell, C., Johnson, M.D., Anderson, E.W., Chan, J., Everitt Bryant, B. (1996). “The American Customer Satisfaction Index: nature, purpose and findings.” Journal of Marketing, 60 (4), pp. 7-18 Fornell, C., Larcker, D.F. (1981). “Evaluating structural equation models with unobservable variables and measurement error.” Journal of Marketing Research, 18 (1), pp.39-50 Fournier, S., Mick, D.G. (1999). “Rediscovering satisfaction.” The Journal of Marketing, 63 (4), pp. 5-23 Fuchs, C., Diamantopoulos, A. (2009). “Using single -item measures for construct measurement in management research – conceptual issues and application guidelines. Die Betriebswirtshchaft, 69 (2), pp. 195-210
46
Funk, J.L. (2006). “The future of mobile phone -based internet applications: A view from Japan.” Technovation, 26, pp. 1337-1346 Gallarza, M.G., Saura, I.G. (2006). “Value dimensions, perceived value, satisfaction and loyalty: an investigation of university students’ travel behavior.” Tourism Management, 27, pp. 437-452 Gardial, S.F., Clemons, D.S., Woodruff, R.B., Schumann, D.W., Burns, M.J. (1994). “Comparing consumers’ recall of prepurchase and postpurchase product evaluation experiences.” Journal of Consumer Research, 20 (4), pp. 548-560 GfK (2011). “B2B.” Presentation on sectoral trends of GfK Retail and Technology GfK (2012). “Local Charts Smartphones Netherlands 2012 03.” GfK Retail and Technology Sales Figure Report Gorsuch, R.L. (1983). “Factor analysis.” 2nd edition, Hillsdale, NJ: Erblaum in: Floyd, F.J., Widaman, K.F. (1995). “Factor analsis in the development and refinement of clinical assessment instruments.” Psychological Assessment, 7 (3), pp. 286-299 Haenlein, M., Kaplan, A.M. (2004). “A beginner’s guide to Partial Least Squares analysis.” Understanding statistics, 3 (4), pp. 283-297 Hair, J.F., Black, B., Babin, B., Anderson, R.E. (2010). “Multivariate Data Analysis.” 7th edition, Pearson. Chapter 2: “Cleaning and transforming data.” Chapter 3: “Factor analysis.” Chapter 8: “ANOVA and MANOVA.” Chapter 11: “SEM: An introduction.” Chapter 12: Applications of SEM.” Hair, J.F., Ringle, C.M., Sarstedt, M. (2011). “PLS-SEM: Indeed a silver bullet.” Journal of Marketing Theory and Practice, 19 (2), pp. 139-151 Halstead, D. (1999). “The use of comparison standards in customer satisfaction research and management: a review and proposed typology.” Journal of Marketing Theory and Practice, 7 (3), pp. 13-26 Harzing, A. (2011). “Journal quality list.” 42nd edition, 16 October 2011 Haverila, M. (2011a). “What do we want specifically from the cell phone? An age related study.” Article in press for Telematics and Informatics Haverila, M. (2011b). “Mobile Phone feature preferences, customer satisfaction and repurchase intent among male users.” Australasian Marketing Journal, 19, pp. 238-246 Henseler, J., Fassot, G. (2010). “Testing moderating effects in PLS path models: an illustration of a vailable procedures.” Handbook of Partial Least Squares. Springer Handbooks of Computational Statistics. Berlin Heidelberg. Hirschman, A.O. (1970). “Exit, voice, and loyalty – responses to decline in firms, organizations and states.” Cambridge, MA: Harvard University Press, in: Johnson, M.D., Gustafsson, A., Andreassen, T.W., Lervik, L., Cha, J. (2001). “The evolution and future of national customer satisfaction index models.” Journal of Economic Psychology, 22, pp. 217-245, and in: Fornell, C., Johnson, M.D., Anderson, E.W., Chan, J., Everitt Bryant, B. (1996). “The American Customer Satisfaction Index: nature, purpose and findings.” Journal of Marketing, 60 (4), pp. 7-18 Hsee, C.K., Yang, Y., Gu, Y., Chen, J. (2009). “Specification seeking: how product sp ecifications influence consumer preference.” The Journal of Consumer Research, 35 (6), pp. 952-966 Institute of Electrical and Electronics Engineers (1998). “IEEE Standard for software and system test documentation.” Standard 829-1998 Jacobi, J. (1978). “Consumer research: how valid and useful are all our consumer behavior findings? A state of the art review.” The Journal of Marketing, 42 (2), pp. 87-96 Johnson, M.D., Gustafsson, A., Andreassen, T.W., Lervik, L., Cha, J. (2001). “The evolution and future o f national customer satisfaction index models.” Journal of Economic Psychology, 22, pp. 217-245 Kamis, A., Koufaris, M., Stern, T. (2008). “Using an attribute-based decision support system for usercustomized products online: An experimental investigation.” MIS Quarterly, 32 (1), pp. 159-177 Kaplan, A.M., Haenlein, M. (2006). “Towards a parsimonious definition of traditional and electronic mass customization.” Journal of Product Innovation Management, 23 (2), pp. 168-182 Keijzers, J., Ouden, E. den, Lu, Y. (2008). “Usability benchmark study of commercially available smartphones: Cell Phone type platform, PDA type platform and PC type platform.” Procedings of the 10th International Conference on Human Computer Interaction with Mobile Devices and Services , pp. 265272 Kekre, S., Krishnan, M.S., Srinivasan, K. (1995). “Drivers of customer satisfaction for software products: implications for design and service.” Management Science, 41 (9), pp. 1456-1470 Kim, S.H. (2008). “Moderating effects of job relevance and experience on mobile wireless technology acceptance: adoption of a smartphone by individuals.” Information & Management, 45, pp. 387-393 Klein, L.R. (1998). “Evaluating the potential of interactive media through a new lens: Search versus experience goods.” Journal of Business Research, 41 (3), pp. 195-203 LaBarbera, P.A., Mazursky, D. (1983). “A longitudinal assessment of consumer satisifaction/dissatisfaction: the dynamic aspect of the cognitive process.” Journal of Marketing Research, 20, pp. 393-404, in: Mittal., V., Ross Jr., W.T., Baldasare, P.M. (1998). “The asymmetric impact of negative and positive attribute -
47
level performance on overall satisfaction and repurchase intentions.” The Journal of Marketing, 62 (1), pp. 33-47 Lamb, C.W., Hair, J.F., McDaniel, C. (2012). “Essentials of marketing.” South-Western Cengage Learning, Mason USA Lancaster, K. (1971). “Consumer demand: A new approach.” New York: Columbia University Press, in: Thompson, D.V., Hamilton, R.W., Rust, R.T. (2005). “Feature fatigue: when product capabilities become too much of a good thing.” Journal of Marketing Research, 42 (4), pp. 431-442 Liao, C., Chen, J.L., Yen, D.C. (2007). “Theory of planning behavior (TPB) and customer sa tisfaction in the continued use of e-service: An integrated model.” Computers in Human Behavior, 23, pp. 2804-2822 Low, G.S., Lamb jr., C. W. (2000). “The measurement and dimensionality of brand associations.” Journal of Product & Brand Management, 9 (6), pp. 350-368 Marcoulides, G.A., Saunders, C. (2006). “PLS: a silver bullet?” MIS Quarterly, 30 (2), pp. iii-ix McDougall, G.H.G., Levesque, T. (2000). “Customer satisfaction with services: putting perceived value into the equation.” Journal of Services Marketing, 14 (5), pp. 392-410 Mitchell, A.A., Dacin, P.A. (1996). “The assessment of alternative measures of consumer expertise.” Journal of Consumer Research, 23, pp. 219-239 Mittal., V., Ross Jr., W.T., Baldasare, P.M. (1998). “The asymmetric impact of negative and positive attribute level performance on overall satisfaction and repurchase intentions.” The Journal of Marketing, 62 (1), pp. 33-47 Moore, Geoffrey A. (2006). “Crossing the chasm: Marketing and selling disruptive products to mainstream customers.” Collins Business Essentials Muthukrishnan, A.V., Weitz, B.A. (1991). “Role of product knowledge in evaluation of brand extension.” Advances in Consumer Research, 18, pp. 407-413 Nielsen, J. (1993). “Usability Engineering.” San Diego: Academic Press, in: Thompson, D.V., Hamilton, R.W., Rust, R.T. (2005). “Feature fatigue: when product capabilities become too much of a good thing.” Journal of Marketing Research, 42 (4), pp. 431-442 Ojiako, U., Maguire, S. (2009). “Seeking the perfect customer experience: a cause study of British telecom.” Strategic Change, 18, pp. 179-193 Oliver, R.L. (1993). “Cognitive, affective and attribute bases of the satisfaction response.” Journal of Consumer Research, 20 (3), pp. 418-430 Oliver, R.L. (1999). "Whence Customer Loyalty?" The Journal of Marketing, 63, pp. 33-44 Olsen, S.O., Wilcox, J., Olsson, U. (2005). “Consequences of ambivalence on satisfaction and loyalty.” Psychology & Marketing, 22 (3), pp. 247-269 Ouden, E. den, Yuan, L., Sonnemans, P.J.M., Brombacher, A.C. (2005). “Quality and Reliability problems from a consumer’s perspective: an increasing problem overlooked by business?” Quality and Reliability Engineering International, 22, pp. 821-838 Oulasvirta, A., Wahlström, M. , Ericsson, K.A. (2011). “What does it mean to be good at using a mobile device? An investigation of three levels of experience and skill.” International Journal of Human-Computer Studies, 69, pp. 155-169 Pahl, G., Beitz, W. (1998). “Engineering design: A systematic approach. Springer-Verlag in: Stone, R.S. (2000). “A heuristic method for identifying modules for product architectures.” Design Studies, 21, pp. 5-31 Pallant, J. (2005). “SPSS Survival Manual.” 12th version. Allen & Unwin. Park, Y., Chen, J.V. (2007). “Acceptance and adoption of the innovative use of smartphone.” Industrial Management & Data Systems, 107 (9), pp. 1349-1365 Payne, J.W. (1982). “Contingent decision behavior.” Psychological Bulletin, 92 (2), pp. 382-402 Peslak, A., Shannon, L., Ceccucci, W. (2011). “An empirical study of cell phone and smartphone usage.” Issues in Information Systems, 12 (1), pp. 407-417 Peterson, R.A., Wilson, W.R. (1992). “Measuring customer satisfaction: fact and artifact.” Journal of the Academy of Marketing Science, 20 (1), pp. 61-71 Pitt, L.F., Parent, M., Junglas, I., Chan, A., Spyropoulou, S. (2011). “Integrating the smartphone into a sound environmental information systems strategy: Principles, practices and a research agenda.” Journal of Strategic Information Systems, 20, pp. 27-37 Podsakoff, P.M., Podsakoff, N.P., Mackenzie, S.B., Lee, J. (2003). “Common method biases in behavioral research: A critical review of the literature and recommended remedies.” Journal of Applied Psychology, 88 (5), pp. 879-903 Radner, R., Rothschild, M. (1975). “On the allocation of effort.” Journal of Economic Theory (10), pp. 358-376, in: Davis, F.D. (1989). “Perceived usefulness, perceived ease of use, and user acceptance of information technology.” MIS Quarterly, 13 (3), pp. 319-340 Randolph, J.J. (2009). “A guide to writing the dissertation literature review.” Practical Assessment, Research & Evaluation, 14 (13), pp. 1-13
48
Reinzartz, W.J., Haenlein, M., Henseler, J. (2009). “An empirical comparison of the efficacy of covariancebased and variance-based SEM.” INSEAD Working Paper Scott Armstrong, J., Overton, T.S. (1977). “Estimating nonresponse bias in mail surveys.” Journal of Marketing Research, 14, pp. 396-402 Sobel, M.E. (1982). “Asymptotic confidence intervals for indirect effects in structural equation models.” Social Methodology, 13, pp. 159-186 Spiteri, J.M., Dion, P.A. (2004). “Customer value, overall satisfaction, end -user loyalty, and market performance in detail intensive industries.” Industrial Marketing Management, 33, pp. 675-687 Srinivasan, V., Lovejoy, W.S., Beach, D. (1997). “Integrated product design for marketability and manufacturing.” Journal of Marketing Research, 34, pp. 154-63, in: Thompson, D.V., Hamilton, R.W., Rust, R.T. (2005). “Feature fatigue: when product capabilities become too much of a good thing.” Journal of Marketing Research, 42 (4), pp. 431-442 Steiner, F., Tarman, R.T., Ihl, J.C., Piller, F.T. (2009). “Learning from the customer: Identifying changing user needs during product usage through embedded toolkits for user innovation.” Portland International Conference on Management of Engineering & Technology, August 2-6, 2009, pp. 706-716 Stone, R.S. (2000). “A heuristic method for identifying modules for product architectures.” Design Studies, 21, pp. 5-31 Szymanski, D.M., Henard, D.H. (2001). “Customer satisfaction: a meta-analysis of the empirical evidence.” Academy of Marketing Science Journal, 29 (1), pp. 16-35 Tabachnick, B.G., Fidell, L.S. (2001). “Using multivariate statistics.” 4th edition. HarperCollins Tan, W., Yeh, Y., Chen, S., Lin, Y., Kuo, C. (2012). “How consumers assess product’s features?: A case study of product features of smartphone.” Recent Researches in Applied Mathematics and Economics. Volume unknown, accessible via http://www.wseas.us/elibrary/conferences/2012/Vouliagmeni/MMAS/MMAS-20.pdf Teleompaper (2011). “Dutch mobile apps market.” Telecompaper Research Report 2011, October 2011 Teleompaper (2012). “Dutch mobile apps market.” Telecompaper Research Report 2011, January 2012 Thaler, R. (1985). “Mental accounting and consumer choices.” Marketing Science, 4 (3), pp. 199-214, in: Gallarza, M.G., Saura, I.G. (2006). “Value dimensions, perceived value, satisfaction and loyalty: an investigation of university students’ travel behavior.” Tourism Management, 27, pp. 437-452 Thompson, D.V., Hamilton, R.W., Rust, R.T. (2005). “Feature fatigue: when product capabilities become too much of a good thing.” Journal of Marketing Research, 42 (4), pp. 431-442 Tilson, D., Lyytinen, K. (2006). “The 3G transition: Changes in the US wireless industry.” Telecommunications Policy, 30 (10-11), pp. 569-586 Töyssy, S., Helenius, M. (2006). “About malicious software in smartphones.” Journal in Computer Virology, 2 (2), pp. 109-119 Turel, O., Serenko, A. (2006). “Satisfaction with mobile services in Canada: an empirical investigation.” Telecommunications Policy, 30, pp. 314-331 Venkatesh, V., Davis, F.D. (2000). “A theoretical extension of the Technology Acceptance Model: Four longitudinal field studies.” Management Science, 46 (2), pp. 186-204. Venkatesh, V., Morris, M.G., Davis, G.B., Davis, F.D. (2003). “User acceptance of information technology: Toward a unified view.” MIS Quartetly, 27 (3), pp. 425-478 Verkasalo, H., López-Nicolás, C., Molina-Castillo, F.J., Bouwman, H. (2010). “Analysis of users and non -users of smartphone applications.” Telematics and Informatics, 27, pp. 242-255 Vinzi, V.E., Chin, W.W., Henseler, J., Wang, H. (2009). “Handbook of partial least squares: Concepts, methods and applications.” Springer VerlagBerlin Heidelberg 2010 Voss, K.E., Spangenberg, E.R., Grohmann, B. (2003). “Measuring the hedonic and utilitarian dimensions of consumer attitude.” Journal of Marketing Research, 40 (3), pp. 310-320 Wang, Y., Lo, H., Yang, Y. (2004). “An integrated framework for service quality, customer value, satisfaction: evidence from China’s telecommunication industry.” Information Systems Frontiers, 6 (4), pp. 325-340 Woodruff, R.B. (1997). “Customer value: the next source for competitive advantage.” Academy of Marketing Science Journal, 25 (2), pp. 139-153 Yi, Y. (1990). “A critical review of consumer satisfaction.” Review of Marketing, 1990 in: Fournier, S., Mick, D.G. (1999). “Rediscovering satisfaction.” The Journal of Marketing, 63 (4), pp. 5-23
6.2 Web references
Gaigg, M. (2012). Blog by Micheal Gaigg (publication date unknown). “Feature fatigue: say NO to your clients.” http://www.michaelgaigg.com/blog/tag/feature-fatigue/ , accessed June 20th 2012 Gartner.com (2012): Summary of Quartile 1 2012 market share analysis of mobile devices worldwide, by Gartner – information technology research and advisory. http://www.gartner.com/id=2015916 ; Accessed July 27th 2012
49
IDC.com (2011). Press release on smartphone sales by market intelligence provider IDC. http://www.idc.com/getdoc.jsp?containerId=prUS22871611 , accessed 5th of August, 2011 iTunes.apple.com (2012). Online store where apps can be purchased. http://itunes.apple.com/nl/app/sleep-cycle-alarmclock/id320606217?mt=8 , accessed 7th of August, 2012 Kendrick, J. (2011). Blog by Kendrick, J. (2011). “Are you suffering from app fatigue?” http://www.zdnet.com/blog/mobilenews/are-you-suffering-from-app-fatigue/1305, accessed 27th of June 2012 Madison, J. (2011). “’The world is a better place because of Steve’: the life and times of Apple visionary Steve Jobs.” Appeared on Dailymail.co.uk (2011): Website of newspaper The Daily Mail. http://www.dailymail.co.uk/news/article-2045852/Steve-Jobs-dead-Biography-Apple-visionary.html, accessed 24th of July, 2012 Ringle, C.M., Wende, S., Will, S. (2005). Software package for Partial Least Squares analysis “SmartPLS 2.0 (M3) Beta.” http://www.smartpls.de , accessed May 29th 2012 Statline.cbs.nl (2012). Database of the Dutch Bureau for Satistics, http://statline.cbs.nl , accessed 17th of February 2012 Taliuaga, J., Bates, J., Sumner, J. (2009). “Feature fatigue: A necessary evil in portable devices.” http://www.rockresearch.com/feature-fatigue-a-necessary-evil-in-portable-devices , accessed 27th of June 2012
Appendix A: Measurement instrument
Construct or variable (Cronbach’s ? if applicable) Apps Factor loading (if applicable) Measure (source and Cronbach’s alpha where applicable)
Customer Satisfaction Of The Smartphone** Daily Usage Time
-
How many apps are installed in total on your smartphone, including pre-installed applications? 1-10, 11-20, 21-30, 31-40, 41-50, 51-60, 61-70, 71-80, 81-90, 91-100, more than 100, namely: … I am satisfied with my smartphone (11 point Likert scale: Peterson & Wilson, 1992) On average, how much time do you spend on using your smartphone per day? This concerns use without calling, but including texting, and using internet and applications. Measured in hours, with an interval of 0.25 hours (adapted from Venkatesh & Davis, 2000 and Al-Gahtani & King, 1999) On average, how many times do you use your smartphone per day? This concerns use without calling, but including texting, and using internet and applications (adapted AlGahtani & King, 1999) Mobile data usage in MB/month. This information is extracted from the customer base and is objective. The measure is the monthly data use, averaged over the 3 months before data collection: December 2011 through February 2012 Installing applications on my smartphone requires time (adapted from Wang et al., 2004) Updating applications on my smartphone requires time (adapted from Wang et al., 2004) Learning to use the applications on my smartphone requires time (adapted from Wang et al., 2004) My interaction with my smartphone is clear and understandable (adapted from Venkatesh & Davis, 2000, ? = .86-.98) Interacting with my smartphone does not require a lot of my mental effort (adapted from Venkatesh & Davis, 2000, ? = .86-.98) [is reversed in the survey] I find my smartphone to be easy to use (adapted from Venkat esh & Davis, 2000, ? = .86.98) I find it easy to get my smartphone to do what I want it to do (adapted from Venkatesh & Davis, 2000, ? = .86-.98) I often use my smartphone (adapted from Al-Gahtani & King, 1999). Using my smartphone increases my productivity in everyday activities (adapted from Venkatesh & Davis, 2000, ? = .87-.98) Using my smartphone enhances my effectiveness of everyday activities (adapted from Venkatesh & Davis, 2000, ? = .87-.98) I find my smartphone to be useful in everyday life (adapted from Venkatesh & Davis, 2000, ? = .87-.98) SMS usage in number of SMS text messages per month, averaged over December 2011through February 2012. This objective information is extracted from the customer base of the Telco I am satisfied with my telecom provider (11 point Likert scale: Peterson & Wilson, 1992)
Frequency Of Use
-
Mobile Data Usage (Objective Measure – Not In Survey) Perceived Effort* (? = .81)
-
.882 .905 .740
Perceived Usability/Ease Of Use* (? = .92)
.940 Deleted .922 .915
Perceived Usage Perceived Usefulness* (? = .84)
.885 .882 .848
SMS Usage (Objective Measure – Not In Survey) Telecommunication Provider /Company (Telco) Customer Satisfaction** Voice Usage (Objective Measure – Not In Survey)
-
-
-
Voice usage in number of minutes per month, averaged over December 2011 through February 2012. This objective information is extracted from the customer base of the Telco Smartphones are fun (adapted from Voss et al., 2003, ? = .92- .95) Smartphones are exciting (adapted from Voss et al., 2003, ? = .92- .95)
Moderating Variables Hedonic Attitude Towards The Smartphone* (? = .88)
.831 .894
50
Perceived Brand Image* (? = .95) Subjective Knowledge* (? = .92)
.882 .830 Deleted .938 .961 .964 .932 .830 .898 .910
Smartphones are delightful (adapted from Voss et al., 2003, ? = .92- .95) Smartphones are thrilling (adapted from Voss et al., 2003, ? = .92- .95) Smartphones are enjoyable (adapted from Voss et al., 2003, ? = .92- .95) The brand of my smartphone is attractive (Low & Lamb, 2000, ? = .85) The brand of my smartphone is reliable (Low & Lamb, 2000, ? = .85) The brand of my smartphone is of high quality (Low & Lamb, 2000, ? = .85) I know much about smartphones (Adapted from Flynn & Goldsmith, 1999, ? = .87 - .94) I know how to operate smartphones (Adapted from Flynn & Goldsmith, 1999, ? = .87 .94) Among my circle of friends, I’m one of the experts on smartphones (Flynn & Goldsmith, 1999, ? = .87 - .94) Compared to most other people, I know more about smartphones (Flynn & Goldsmith, 1999, ? = .87 - .94) Smartphones are effective (adapted from Voss et al., 2003, ? = .95) Smartphones are helpful (adapted from Voss et al., 2003, ? = .95) Smartphones are functional (adapted from Voss et al., 2003, ? = .95) Smartphones are necessary (adapted from Voss et al., 2003, ? = .95) Smartphones are practical (adapted from Voss et al., 2003, ? = .95)
Utilitarian Attitude Towards The Smartphone* (? = .94)
.904 Deleted .934 Deleted .917
Control Variables Age Gender Education Level Usage Time
-
Smartphone Brand Type Of Use Survey Method
-
What is your age? What is your gender (M/F)? What is your education level: primary education, vmbo/mbo 1, havo/vwo, mbo 2-4, hbo, wo, other (Statline.cbs.nl, 2012) How many smartphones have you owned in total, including your current smartphone? (adapted from Mitchell & Dacin, 1996) When did you purchase/receive your current smartphone? (which month and year, adapted from Turel & Serenko, 2006) ; or if this is not your first smartphone: when did you receive/purchase your first smartphone? Which brand is your current smartphone? Apple, Blackberry, Samsung, HTC , other: … (GfK, 2011) Do you use your smartphone for business, private, or both? On which device did you fill in this survey? On a PC/laptop, smartphone, tablet or other
All items of the constructs marked with an asterisk are measured on a 7 point Likert scale with 1: strongly disagree, 2: disagree, 3: somewhat disagree, 4: neutral (neither disagree nor agree), 5: somewhat agree, 6: agree, 7: strongly agree (adapted from Venktatesh & Davis, 2000). The constructs marked with a double asterisk are measured with an 11-point Likert scale. In Appendix B, the measurement instrument is integrated into a Dutch questionnaire.
Appendix B: Questionnaire
This Appendix contains the measurement instrument translated in Dutch. The questionnaire contains 35 questions in total. Some elements were added due to the interest of the telecom operator and in order to filter irrelevant respondents. These elements are marked with an “X” in the item indication. Additionally, every item is guided by a Dutch text in which the questions corresponding to the items are explained. Abbreviations: APP = number of apps, CS = customer satisfaction with the smartphone, TPCS = customer satisfaction with the telecommunication operator, PU = perceived usefulness, PEU = perceived ease of use/usability, DU = daily duration of use, FREQ = daily frequency of use, PULS = perceived usage level, LS = loyalty towards the smartphone, LTC = loyalty towards the telecommunication operator, PBI = perceived image of the smartphone brand, SK = subjective knowledge, UA/HA = utilitarian/hedonic attitude, PE = perceived effort put into apps, PMC = perceived monetary costs, PO = perceived overlap, CON = conscientiousness, UT = usage time, A = age, G = gender, E = education, SB = smartphone brand, SM = survey method, TU = type of use.
51
********* Start of the questionnaire ********** Deze enquête gaat over smartphones en het gebruik van apps. Als je meerdere smartphones in gebruik hebt, neem dan voor het invullen van deze vragenlijst de smartphone in gedachten die je het meest gebruikt. XA1. Volgens onze gegevens maak je gebruik van een smartphone en heb je een abonnement bij [telecom provider] Klopt dit? ? Ja (“Yes”) (1) ? Nee (“No”) (2) einde enquête: “Helaas val je buiten de doelgroep van het onderzoek. Hartelijk dank voor je medewerking.” De volgende twee stellingen gaan over jouw tevredenheid met je smartphone en met [telecom provider]. Geef aan in hoeverre je het eens bent met de volgende stellingen [1 (helemaal mee eens)…(11 helemaal mee oneens)]: CS Ik ben tevreden met mijn smartphone TPCS Ik ben tevreden met [telecom provider] De volgende stellingen gaan over het gebruik van jouw smartphone. Geef aan in hoeverre je het eens bent met de volgende stellingen: PU1 Door mijn smartphone kan ik meer doen in dezelfde tijd PU2 Door mijn smartphone kan ik alledaagse activiteiten beter uitvoeren PU3 Mijn smartphone is nuttig in het alledaagse leven PEU1 Mijn smartphone is duidelijk en begrijpelijk PEU2 Ik moet mij concentreren als ik mijn smartphone gebruik [REVERSED] PEU3 Mijn smartphone is eenvoudig in gebruik PEU4 Het is gemakkelijk om mijn smartphone te laten doen wat ik wil DU Hoeveel tijd besteed je gemiddeld per dag aan het gebruiken van je smartphone?Reken bellen hier niet onder en rond je antwoord af in kwartieren. 1 uur en 40 minuten per dag rond je bijvoorbeeld af naar 1:45 ? [in uren met intervallen van een kwartier; bij voorkeur twee uitklapvensters, links voor # uren 1 t/m 24, rechts voor het aantal kwartieren: 0, 15, 30, 45] FREQ Hoe vaak gebruik jij je smartphone gemiddeld op een dag? Reken bellen hier niet onder en denk hierbij dus aan het volgende gebruik: SMS-en en het gebruiken van internet en apps ? [Aantal (open antwoord) ] keer per dag PULS Ik maak veel gebruik van mijn smartphone (reken belgedrag hier niet onder) [Monthly mobile data, voice, and SMS usage is extracted from customer database in Megabytes/ minutes/SMS per month averaged over 3 months] De volgende vragen gaan over toekomstig gebruik van jouw smartphone. Geef aan in hoeverre je het eens bent met de volgende stellingen: XLS0 Mijn volgende telefoon wordt een smartphone LS1 Mijn volgende smartphone wordt van hetzelfde merk LS2 Als ik op dit moment zou kunnen kiezen zonder dat het geld of moeite kost om over te stappen, zou ik een smartphone van een ander merk nemen [Reversed]
52
LS3 Als mensen het mij zouden vragen dan zou ik het merk van mijn smartphone aanraden De volgende vragen gaan over het toekomstig gebruik van [telecom provider]. Geef aan in hoeverre je het eens bent met de volgende stellingen: LTC1 Mijn volgende mobiele abonnement wordt van [telecom provider] LTC2 Als ik op dit moment zou kunnen kiezen, zonder dat het geld of moeite kost om over te stappen, zou ik voor een abonnement van een andere aanbieder kiezen [Reversed] LTC3 Als mensen het mij zouden vragen dan zou ik [telecom provider] aanraden als mobiele aanbieder De volgende stellingen gaan over jouw beleving van het merk van jouw smartphone. Geef aan in hoeverre je het eens bent met de volgende stellingen: PBI1 Het merk van mijn smartphone is aantrekkelijk PBI2 Het merk van mijn smartphone is betrouwbaar PBI3 Het merk van mijn smartphone staat voor hoge kwaliteit De volgende vragen gaan over jouw kennis over smartphones. Geef aan in hoeverre je het eens bent met de volgende stellingen: SK1 Ik weet veel van smartphones SK2 Ik kan goed met smartphones overweg SK3 In mijn vriendenkring ben ik een van de smartphone-experts SK4 In vergelijking met anderen weet ik veel van smartphones De volgende 10 stellingen gaan over jouw houding ten opzichte van smartphones. Geef aan in hoeverre je het eens bent met de volgende stellingen: HA1 Smartphones zijn leuk HA2 Smartphones zijn boeiend HA3 Smartphones geven voldoening HA4 Smartphones zijn spannend HA5 Smartphones zijn plezierig UA1 Smartphones zijn effectief UA2 Smartphones zijn behulpzaam UA3 Smartphones zijn functioneel UA4 Smartphones zijn noodzakelijk UA5 Smartphones zijn praktisch De volgende vragen gaan over de apps die op jouw smartphone staan geïnstalleerd en over het gebruik van deze apps. APP Hoeveel apps heb jij in totaal op jouw smartphone geïnstalleerd? Reken hiertoe ook vooraf geïnstalleerde apps mee, zoals de SMS-app of agenda-apps. ? 1-10 (1) ? 11-20 (2) ? 21-30 (3) ? 31-40 (4) ? 41-50 (5) ? 51-60 (6) ? 61-70 (7) ? 71-80 (8)
53
? ? ?
81-90 (9) 91-100 (10) Meer dan honderd, namelijk: [open antwoord met waarde >100]
(11)
[It should be noted that an error occurred in the design of the web survey: the first 33 respondents were not able to provide a 15 minute interval in their answer to the question how much time they spend on an average day on using their smartphone, when they selected ‘1 hour’ as an answer option. As a result, the following answer options ‘15’, ‘30’ and ‘45’ minutes were not available to these 33 respondents. Fortunately, the lowest answer to this question from this batch of respondents was 1:45 hours, which is an indication that the leavingout of the answer options did not affect the results because it could be expected that respondents that wanted to fill in the answer options that were not available would have indicated the lowest possible answer, which is 0 hours or 1 hour. After one day the answer options were adapted such that the left-out answer options are made available to respondents.] XOWNAPPS Heb je naast de apps die reeds op je telefoon stonden, ook zelf apps geïnstalleerd? ? Ja (1) ? Nee (2) Niet vragen (“Skip”): PMC1, PMC2, PMC3, PE1 Geef aan in hoeverre je het eens bent met de volgende stellingen: PMC1 De apps die ik op mijn smartphone heb geïnstalleerd zijn redelijk geprijsd [Reversed] PMC2 De verkoopprijs van apps die ik heb geïnstalleerd op mijn smartphone is voordelig[Reversed] PMC3 De apps op mijn smartphone zijn hun geld waard [Reversed] PE1 Het heeft veel tijd gekost om de apps op mijn smartphone te installeren PE2 Het updaten van de apps op mijn smartphone kost veel tijd PE3 Ik heb veel tijd geïnvesteerd om de apps op mijn smartphone te leren gebruiken De volgende stellingen gaan over jouw beleving van de onderlinge verschillen tussen de apps op jouw smartphone. Geef aan in hoeverre je het eens bent met de volgende stellingen: PO1 Er is veel overlap tussen de functies van de apps op mijn smartphone PO2 Ik heb meerdere apps voor hetzelfde doeleind (Bijvoorbeeld meerdere apps voor reizen met het openbaar vervoer of voor weersvoorspellingen) PO3 De verschillende apps op mijn smartphone voorzien in dezelfde behoeften De volgende stellingen gaan over het beheren van apps op jouw smartphone, welke afwegingen er plaatsvinden bij het downloaden van apps en de functionaliteit van jouw smartphone. Geef aan in hoeverre je het eens bent met de volgende stellingen [Measured on a five-point Likert scale] : XSTR Ik structureer de apps op mijn smartphone (denk hierbij bijvoorbeeld aan het beheren van mappen waarin je de apps hebt opgeslagen (voor iPhone gebruikers) of het plaatsen van widgets en snelkoppelingen (voor Android gebruikers) XDEL Ik verwijder apps van mijn smartphone als ik ze niet gebruik XCOST Ik let op de kosten van een app als ik overweeg om een app te downloaden XMEM Ik let op de hoeveelheid geheugen die een app inneemt op mijn smartphone als ik overweeg om een app te downloaden XFUNC Ik vind het belangrijk dat mijn smartphone andere functionaliteit biedt naast bellen XFF Ik erger mij aan apps die ik niet gebruik XOA8 Geef een top 3 van apps die je het meest gebruik [Open vraag] Tot slot volgen er nu nog enkele algemene vragen over jou en jouw smartphone. Geef aan in hoeverre je het eens bent met de volgende stellingen:
54
Con1 Ik let op details Con2 Ik begin meteen aan taken Con3 Ik vind het prettig als dingen geordend zijn Con4 Ik ben precies in mijn werk Con5 Ik laat mijn spullen rondslingeren XSEGMENT1 Welk van de volgende antwoorden associeer jij het meest met het gebruik van apps? ? Ontdekken, proberen, testen (1) ? Delen, tonen, mode rn (2) ? Socialiseren, relaties, vrienden (3) ? Functioneel, praktisch, oplossing (4) XSEGMENT2 Apps zijn voor mij vooral ? Nieuw, onbekend, vreemd ? Praktisch, eenvoudig, bekend ? Vertrouwd, handig, leuk, ? Intens, actief, onmisbaar
(1) (2) (3) (4)
XSEGMENT3 Hoe zou je je leven over het algemeen omschrijven? ? Gestructureerd, duidelijk rechtlijnig (1) ? Druk, spanning, snel (2) ? Complex, verwarrend, veranderlijk (3) ? Groei, vooruitgang, opbouwen (4) ? Geleidelijk, rustig, uitgebalanceerd (5) A Hoe oud ben je? ? [leeftijd] G Ik ben een ? Man ? Vrouw
(1) (2)
E Wat is het niveau van je hoogst genoten of je huidige opleiding? ? Geen (1) ? Primair onderwijs of vergelijkbaar (2) ? vmbo/mbo 1 of vergelijkbaar (3) ? havo/vwo of vergelijkbaar (4) ? mbo 2-4 of vergelijkbaar (5) ? hbo of vergelijkbaar (6) ? wo of vergelijkbaar (7) ? anders (8) UT1 Hoeveel smartphones heb je gehad (inclusief je huidige smartphone)? ? Dit is mijn eerste smartphone (1) ? [aantal] smartphones (2) UT2 Wanneer heb je je huidige smartphone gekregen? ? [Maand en jaar] UT3 Wanneer heb je je eerste smartphone gekregen? ? [maand en jaar]
UT3 niet vragen
55
SB Van welk merk is je smartphone? ? Apple ? Blackberry ? HTC ? Samsung ? Anders
(1) (2) (3) (4) (5)
TU Gebruik je je smartphone zakelijk, privé of voor beiden? ? Privé (1) ? Zakelijk (2) ? Beiden (3) SM Met welk apparaat heb je deze enquête ingevuld? ? Op een PC/laptop (1) ? Op een smartphone (2) ? Op een tablet (3) Hartelijk dank voor het invullen van deze enquête! *********** End of the questionnaire ***********
Appendix C: Data manipulation
This appendix describes the manipulation of the dataset that was delivered by the market research agency. First of all, the e-mail addresses and phone numbers of the respondents were deleted in order to ensure animosity of the respondents. The labels added by the market research agency were replaced by the original labels as used in the measurement instrument. The information containing the gender and handset brand in use that stemmed from the customer database was deleted because it is expected that this information is more reliable when directly provided by respondents. The two variables for duration of daily usage (hours and minutes) were aggregated to from 1 variable: hours of use. 3 respondents indicated that they filled in the survey on an ‘other’ device than a smartphone, tablet or desktop/laptop PC. However, these cases concerned an “I MAC” an “Ipad” and a “MAC” of which the MAC and I MAC can be categorized under desktops and laptop PC’s and the iPad can be categorized as a tablet. Therefore, these cases were recoded. The variables for the duration of owning a smartphone were joined such that the months and years of owning a smartphone were combined in a single variable. The different variables for the different paths in the survey (for people having owned multiple smartphones and people owing their first smartphone) were then joined to reflect the time that people have owned a smartphone in total, in months. Subsequently, the data was saved in SPSS and the corresponding labels were added. The item perceived ease of use2 was reversed in the questionnaire and therefore they were reverse recoded in order to reflect the original variable. The brand of the smartphone as provided by the respondent matched the brand of the handset of the customer database for every respondent, except for the 4 cases for which this question was answered with ‘other’; these answers were assigned as missing.
56
For the number of apps installed on the smartphone, the answer option above 100 was “other, namely:”. The maximum value for this item was 188 apps. The answers above 100 were recoded such that they corresponded to the value of the last group on the scale. This last group on the scale, consisting of respondents with more than 101 apps on their smartphone, consisted of 8 respondents. One respondent indicated to be of an age of 17.5 years. This case was rounded to 18 years. The data on SMS and voice usage was delivered per month for December 2011February 2012. This data was averaged and rounded with two digits. Finally, because the distribution of mobile data, SMS and voice usage appeared to be asymptotic, i.e. the distributions had “long tails”. A third root transformation was applied to the mobile data usage and SMS usage, and a fourth root transformation was applied to voice usage. These specific transformations were selected because they yielded distributions of the corresponding variables that best approached the normal distribution. Recoding of daily usage duration and frequency of daily interaction Scale increments for the variables ‘daily interaction duration’ and ‘daily frequency of use’ were created such that the mean was in the middle-answer option and the increments of the scale corresponded to 0.5 standard deviations, in an attempt to create a 7-point scale which corresponds to the scale-size of other measured items and to allow for sufficient discrimination between the categories. In Table 8, the calculations of the scales are displayed. Frequency of use was now measured on a 6-point scale and daily duration of use on a 7-point scale. The scale boundaries are not strict; i.e. if scale value 1 ranges from 0.00 to 5.00; 5.00 is assigned value 1 but 5.01 is assigned value 2. The values of both variables were recoded to reflect the values of the new scales that were developed.
Table 8 Recoding of daily duration of use and dialy frequency of use
Value: 1 2 3 4 5 6 7 mean S.D. min value max value Daily duration of use lower bound upper bound 0.00 1.33 1.33 2.42 2.42 3.51 3.51 4.59 4.59 5.68 5.68 6.77 6.77 or more 4.05 2.17 1.25 17.25 Daily frequency of use lower bound upper bound 0 5 5 17 17 29 29 41 41 54 54 or more mean S.D. min value max value 23.08 24.50 0.00 200.00
Coding of dummy variables for survey method and type of use In the Table 9, the values for the dummy variables and the corresponding categories can be viewed. Since only 1 respondent indicated to use his smartphone for business uses, this category was not included in the model and was treated as missing data.
57
Table 9 Coding of dummy variables
Variable Age (original code) Male (1) Female (2) Type of use categories (original code) Consumer (1) Business (2) (not included into analysis due to only 1 respondent) Consumer and Business (3) Survey method categories (original code) PC/laptop (1) Smartphone (2) Tablet (3) Smartphone brand categories (original code) Apple (1) Blackberry (2) HTC (3) Samsung (4) Other (5) is not controlled for Recoded into variables Original variable 0 1 Original variable 0 1 SMDummy1 0 1 0 SBDummy1 0 1 0 0 SMDummy2 0 0 1 SBDummy2 0 0 1 0 -
SBDummy3 0 0 0 1 -
Appendix D: Generalization of the sample
Table 10 General characteristics of the sample and other customer groups
Consumer postpaid base (smartphone, 4 brands) 47,33% 52,67% 3,17% 31,33% 21,78% 24,75% 14,40% 3,92% 0,66% Consumer postpaid + prepaid base (smartphone, 4 brands) 47,53% 52,47% 8,37% 30,17% 20,21% 23,04% 13,44% 3,88% 0,89% Consumer + business, postpaid + prepaid (Smartphone, 4 brands)
Variable
Category
Respondents
Dutch population 49.51%* 50.49% 13,49%* 12,03% 14,44% 14,87% 13,02% 8,39% 7,13% 5.13%** 18.06% 9.05% 33.20% 21.79% 11.80% 0.97%
Gender Age
Education
Smartphone brand
Male Female 15-25 26-35 36-45 46-55 56-65 66-75 75+ None Primary VMBO/MBO1 HAVO/VWO MBO 2-4 HBO WO other Apple Blackberry HTC Samsung
48.8% 51.2 % 18.3 % 18.3 % 25.0 % 25.4 % 10.3 % 1.2 % 0.8 % 0.4 % 0.4 % 11.9 % 29% 28.2 % 31.3 % 13.5 % 2.8% 18.7 % 13.5% 32.1% 34.1 %
8,25% 29,97% 20,31% 23,18% 13,49% 3,91% 0,89%
21,60% 17,76% 17,92% 42,72%
19,87% 19,65% 16,92% 43,55%
20,06% 19,74% 16,87% 43,32%
*As of January 1st, 2012; **Over 2011. Source: Telecommunication operator’s customer base data and Statline.cbs.nl (2012).
58
Appendix E: Data exploration: outliers and normality
Table 11 Univariate outliers
Case ID 528067 521834 523001 523188 521892 527015 527748 522188 522723 5226821 522188 522473 525776 526767 527078 526776 521772 522778 521768 522216 Variable ULS1: Smartphone Use Per Day ULS1: Smartphone Use Per Day ULS1: Smartphone Use Per Day ULS2: Frequency Of Smartphone Use Per Day ULS2: Frequency Of Smartphone Use Per Day ULS2: Frequency Of Smartphone Use Per Day MDU UT1: Number Of Smartphones Owned UT1: Number Of Smartphones Owned UT1: Number Of Smartphones Owned UT2: Months Owning A Smartphone UT2: Months Owning A Smartphone Age Age SMS SMS SMS Voice Voice Voice Z-Value 7.2 6.1 4.2 14.7 7.22 5.18 6.3 6.6 5.0 4.2 4.7 4.5 13.8 4.4 5.19 6.20 6.83. 4.60 4.92 8.20 Variable value 21.5 hours/day 17.25 hours/day 13.25 hours/day 1000 times per day 200 150 1503 Mb/month 10 smartphones 8 smartphones 7 smartphones 135 months 131 months 500 190 475.67 556.33 607.33 597.05 631.17 991.08 Outlier category Procedural error Extraordinary observation Extraordinary observation Procedural error Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Procedural error Procedural error Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Action Delete value Retain Retain Delete value Retain Retain Delete value Retain Retain Retain Retain Retain Delete value Delete value Delete value Delete value Delete value Delete value Delete value Delete value
The outliers are displayed in Table 11. The outliers are categorized according to 3 categories provided by Hair et al. (2010): procedural errors, extraordinary events or extraordinary observations. The following is observed about the outliers: Outliers 528067, 523188, 525776 and 526767 are classified as procedural errors because of their unrealistic values. Respondent 528067 indicates to use his/her smartphone for 21.25 hours per day, with a frequency of 25 times per day, meaning an average of 51 minutes per interaction. Due to this improbable value, which indicates either use of a smartphone during sleep or 2.75 hours sleep per day, the value was deleted. Respondent 523188 indicates a frequency of 1000 interactions with his/her smartphone per day, with the next runner up in the dataset indicating a value of 200. The respondent indicates a use of 9.25 hours per day and uses 607 Mb of data per month and the rest of the values seem normal. It is assumed that the value of 1000 is a typing error, and the value is deleted. The same is assumed for respondents 525776 and 526767, indicating an age of respectively 500 and 190 years. These values are deleted. The other univariate outliers are classified as extraordinary observations because there are no obvious explanations for them but the values are realistic and plausible. The value of 17.25 hours smartphone use per day is high, but possible for someone using his/her smartphone for both business and pleasure. For the objective data on voice, SMS and mobile data usage, the outlying values are deleted since these are central variables and in order to prevent asymptotic data, could lead to unreliable estimation of parameters in PLS (Vinzi et al., 2009). The outliers of the control variables are retained.
The distribution of the variables is displayed in Table 12.
59
Table 12 Data distribution
Shape descriptions Variable Zvalue* Age -0.32 -1.02 Perceived Usability3 Apps 0.43 1.39 Perceived Usability4 Daily Usage Time -0.24 -0.77 Perceived Usage Level Duration Of Ownership 4.35 Perceived Usefulness1 14.11 Education -0.21 -0.68 Perceived Usefulness2 Frequency of use -0.10 -0.33 Perceived Usefulness3 Hedonic Attitude1 4.36 Smartphone Cust Sat 14.13 Hedonic Attitude2 1.51 Smartphones Owned 4.91 Hedonic Attitude3 0.38 1.22 SMS Usage** Hedonic Attitude4 0.12 0.38 Subjective Knowledge1 Hedonic Attitude5 1.86 Subjective Knowledge2 6.04 Mobile Data Usage** -0.72 -2.32 Subjective Knowledge3 Perceived Brand Image1 1.42 Subjective Knowledge4 4.61 Perceived Brand Image2 1.66 Telco Cust Sat 5.39 Perceived Brand Image3 1.89 Utilitarian Attiude1 6.12 Perceived Effort1 0.37 1.18 Utilitarian Attitude2 Perceived Effort2 -0.53 -1.73 Utilitarian Attitude3 Perceived Effort3 -0.16 -0.52 Utilitarian Attitude4 Perceived Usability1 1.89 Utilitarian Attitude5 6.13 Perceived Usability2 -1.09 Voice Usage** -3.52 *Values exceeding ± 2.58 (significance p<.01) have been highlighted in bold. statistic statistic Skewness Zvalue* 0.02 0.14 0.98 6.37 0.83 5.37 1.91 12.39 -0.35 -2.26 0.80 5.05 -1.58 -10.26 -1.02 -6.62 -0.74 -4.79 -0.50 -3.22 -1.09 -7.07 -0.13 0.84 -1.27 -8.21 -1.29 -8.38 -1.33 -8.64 0.94 6.07 0.57 3.69 0.60 3.87 -1.29 -8.37 0.10 0.62 Kurtosis Variable Shape descriptions Skewness Zvalue* -1.39 -8.98 -1.10 -7.15 -1.12 -7.27 -0.69 -4.44 -0.71 -4.63 -1.47 -9.51 -1.39 -9.01 2.64 17.12 0.25 1.64 -0.44 -2.87 -1.00 -6.51 -0.01 -0.05 -0.31 -1.98 -0.42 -2.70 -0.95 -6.19 -1.17 -7.58 -1.35 -8.73 -0.22 -1.46 -1.26 -8.14 0.01 0.09 Kurtosis statistic 2.85 0.62 1.13 0.11 0.29 3.01 1.99 10.34 0.05 -0.52 1.11 -0.85 -0.54 0.26 2.39 2.33 4.59 -0.53 4.06 0.04 Zvalue* 9.22 2.00 3.67 0.35 0.95 9.75 6.46 33.51 0.15 -1.70 3.61 -2.77 -1.76 0.84 7.74 7.55 14.89 -1.70 13.16 0.12 with N
statistic
and
= sample size = 252. Note that the distribution is reflected by the data of which the outliers have been deleted. ** these variables have been transformed, see section 4.1.3
60
Appendix F: Factor loadings and cross loadings of the first CFA
Table 13 Factor loadings and cross loadings of first confirmatory factor analysis
Telco cust sat
Perceived brand image
Apps 1 .32 .057 Cust sat .32 1 .039 Daily usage time .057 .039 1 Frequency .152 .178 .489 Hedonic atttiude1 .193 .165 .244 Hedonic attitude2 .158 .063 .262 Hedonic attitude3 .091 .083 .263 Hedonic attitude4 .12 .055 .266 Hedonic attitude5 .174 .146 .186 Mobile data usage .178 .107 .222 Perceived brand image1 .23 .629 .052 Perceived brand image2 .225 .657 .032 Perceived brand image3 .249 .67 .008 Perceived effort1 -.21 -.253 .084 Perceived effort2 -.19 -.343 .058 Perceived effort3 -.089 -.165 -.003 Perceived usability1 .334 .576 .162 Perceived usability2 .085 .124 .055 Perceived usability3 .314 .542 .145 Perceived usability4 .315 .609 .093 Perceived usefulness1 .163 .402 .155 Perceived usefulness2 .287 .39 .143 Perceived usefulness3 .26 .479 .184 Perceived usage .226 .339 .408 Subjective knowledge1 .216 .117 .211 Subjective knowledge2 .276 .157 .142 Subjective knowledge3 .281 .104 .134 Subjective knowledge4 .289 .122 .195 SMS usage -.027 -.02 -.01 Telco cust sat .026 .312 .02 Utilitarian attitude1 .212 .145 .213 Utilitarian attitude2 .248 .207 .221 Utilitarian attitude3 .218 .214 .208 Utilitarian attitude4 .128 .04 .191 Utilitarian attitude5 .188 .255 .204 Voice usage -.062 .091 -.02 Note: item loadings above .70 are highlighted
.152 .178 .489 1 .299 .257 .241 .172 .235 .307 .106 .171 .151 -.028 .007 -.061 .264 .114 .231 .219 .244 .22 .309 .42 .352 .269 .317 .287 .05 .01 .274 .244 .263 .072 .215 .04
.175 .125 .285 .288 .842 .866 .867 .814 .862 .125 .264 .251 .294 -.061 -.099 -.036 .376 .03 .34 .314 .305 .296 .354 .399 .426 .479 .324 .345 -.008 .16 .667 .704 .679 .459 .598 -.036
.178 .107 .222 .307 .185 .121 .1 .019 .08 1 .038 .084 .08 -.069 -.059 -.087 .183 .092 .267 .181 .2 .207 .231 .303 .251 .258 .285 .303 .082 -.03 .185 .148 .1 .117 .124 .116
.246 .683 .031 .15 .28 .285 .21 .141 .266 .071 .938 .962 .964 -.242 -.262 -.127 .63 .016 .557 .589 .523 .457 .528 .399 .232 .297 .185 .197 -.001 .264 .293 .298 .35 .145 .322 .004
-.205 -.318 .064 -.023 -.08 -.068 -.053 -.037 -.11 -.079 -.214 -.287 -.254 .882 .905 .74 -.272 -.245 -.334 -.323 -.139 -.169 -.088 -.036 -.023 -.13 -.012 .057 .029 -.158 -.079 -.109 -.093 .035 -.059 -.009
.348 .622 .146 .261 .378 .27 .316 .247 .337 .228 .586 .605 .629 -.278 -.339 -.23 .939 .217 .921 .915 .475 .437 .543 .505 .31 .441 .238 .276 .016 .187 .429 .393 .442 .145 .428 -.014
.272 .49 .186 .299 .324 .32 .351 .25 .305 .246 .575 .528 .556 -.118 -.154 -.093 .567 -.025 .475 .53 .885 .882 .848 .52 .389 .361 .338 .376 -.134 .22 .42 .422 .385 .333 .453 -.035
.226 .339 .408 .42 .39 .344 .332 .268 .347 .303 .392 .383 .367 -.061 -.009 -.025 .497 .072 .483 .427 .423 .413 .513 1 .351 .459 .26 .331 -.032 .077 .404 .384 .431 .208 .42 .016
.297 .143 .192 .343 .355 .456 .399 .401 .321 .307 .269 .231 .249 -.047 -.005 -.064 .358 .041 .302 .358 .365 .359 .354 .403 .931 .831 .897 .909 .039 .001 .358 .346 .34 .323 .333 -.024
-.027 -.02 -.01 .05 .002 -.001 0 -.023 -.017 .082 -.021 -.001 .019 .055 -.001 .025 .061 .138 0 -.037 -.139 -.127 -.089 -.032 .075 .067 -.012 -.007 1 -.017 .033 .018 -.012 -.059 -.088 .034
.026 .312 .02 .01 .126 .054 .134 .16 .208 -.03 .278 .225 .253 -.124 -.176 -.075 .18 -.011 .164 .182 .234 .17 .172 .077 .031 .055 -.078 -.027 -.017 1 .142 .21 .145 .162 .224 .081
.241 .218 .246 .267 .676 .594 .625 .541 .719 .161 .328 .319 .344 -.07 -.089 -.036 .461 .02 .415 .402 .386 .377 .484 .453 .37 .501 .225 .282 -.022 .209 .883 .867 .923 .575 .884 -.049
-.062 .091 -.02 .04 -.016 -.021 -.058 -.051 -.012 .116 -.005 .017 .001 .045 -.085 .06 -.019 .01 .01 -.032 .016 -.045 -.06 .016 .011 -.02 -.02 -.058 .034 .081 -.03 .014 -.045 -.1 -.073 1
61
Voice usage
Mobile data usage
Daily usage time
SMS usage
Subjective knowledge
Frequency
Perceived usefulness
Utilitarian attitude
Perceived effort
Perceived usability
Perceived usage
Henodic attitude
Cust sat
Apps
Appendix G: Factor loadings and cross loadings of the definite CFA
Table 14 Factor loadings and cross loadings of definite confirmatory factor analysis
Telco cust sat Perceived brand image Voice usage -.062 .091 -.02 .04 -.016 -.021 -.058 -.051 .116 -.005 .017 .001 .045 -.085 .06 -.019 .01 -.032 .016 -.045 -.06 .016 .011 -.02 -.02 -.058 .034 .081 -.03 -.045 -.073 1 Mobile data usage Daily usage time
SMS usage
Subjective knowledge
Frequency
Perceived usefulness
Apps Cust sat Daily usage time Frequency Hedonic atttiude1 Hedonic attitude2 Hedonic attitude3 Hedonic attitude4 Mobile data usage Perceived brand image1 Perceived brand image2 Perceived brand image3 Perceived effort1 Perceived effort2 Perceived effort3 Perceived usability1 Perceived usability3 Perceived usability4 Perceived usefulness1 Perceived usefulness2 Perceived usefulness3 Perceived usage Subjective knowledge1 Subjective knowledge2 Subjective knowledge3 Subjective knowledge4 SMS usage Telco cust sat Utilitarian attitude1 Utilitarian attitude3 Utilitarian attitude5 Voice usage
1 .32 .057 .152 .193 .158 .091 .12 .178 .23 .225 .249 -.21 -.19 -.089 .334 .314 .315 .163 .287 .26 .226 .216 .276 .281 .289 -.027 .026 .212 .218 .188 -.062
.32 1 .039 .178 .165 .063 .083 .055 .107 .629 .657 .67 -.253 -.343 -.165 .576 .542 .609 .402 .39 .479 .339 .117 .157 .104 .122 -.02 .312 .145 .214 .255 .091
.057 .039 1 .489 .244 .262 .263 .266 .222 .052 .032 .008 .084 .058 -.003 .162 .145 .093 .155 .143 .184 .408 .211 .142 .134 .195 -.01 .02 .213 .208 .204 -.02
.152 .178 .489 1 .299 .257 .241 .172 .307 .106 .171 .151 -.028 .007 -.061 .264 .231 .219 .244 .22 .309 .42 .352 .269 .317 .287 .05 .01 .274 .263 .215 .04
.166 .112 .3 .288 .831 .894 .882 .83 .132 .257 .237 .285 -.047 -.092 -.022 .367 .327 .307 .298 .302 .353 .394 .442 .475 .35 .366 -.005 .136 .638 .648 .576 -.042
.178 .107 .222 .307 .185 .121 .1 .019 1 .038 .084 .08 -.069 -.059 -.087 .183 .267 .181 .2 .207 .231 .303 .251 .258 .285 .303 .082 -.03 .185 .1 .124 .116
.246 .683 .031 .15 .28 .285 .21 .141 .071 .938 .961 .964 -.242 -.262 -.127 .63 .557 .589 .523 .457 .528 .399 .232 .297 .185 .197 -.001 .264 .293 .35 .322 .004
-.205 -.318 .064 -.023 -.08 -.068 -.053 -.037 -.079 -.214 -.287 -.254 .882 .905 .74 -.272 -.334 -.323 -.139 -.169 -.088 -.036 -.023 -.13 -.012 .057 .029 -.158 -.079 -.093 -.059 -.009
.347 .622 .145 .258 .38 .271 .318 .247 .226 .591 .608 .633 -.272 -.333 -.22 .94 .922 .915 .482 .441 .548 .506 .313 .439 .241 .279 .01 .189 .432 .443 .43 -.015
.272 .49 .186 .299 .324 .32 .351 .25 .246 .575 .528 .557 -.118 -.154 -.093 .567 .475 .53 .885 .882 .848 .52 .389 .361 .338 .376 -.134 .22 .42 .385 .454 -.035
.226 .339 .408 .42 .39 .344 .332 .268 .303 .392 .383 .367 -.061 -.009 -.025 .497 .483 .427 .423 .413 .513 1 .351 .459 .26 .331 -.032 .077 .404 .431 .42 .016
.297 .143 .192 .343 .355 .455 .399 .401 .307 .269 .231 .249 -.046 -.005 -.064 .358 .301 .358 .365 .359 .354 .403 .932 .83 .898 .91 .039 .001 .357 .34 .333 -.024
-.027 -.02 -.01 .05 .002 -.001 0 -.023 .082 -.021 -.001 .019 .055 -.001 .025 .061 0 -.037 -.139 -.127 -.089 -.032 .075 .067 -.012 -.007 1 -.017 .033 -.012 -.088 .034
.026 .312 .02 .01 .126 .054 .134 .16 -.03 .278 .225 .253 -.124 -.176 -.075 .18 .164 .182 .234 .17 .172 .077 .031 .055 -.078 -.027 -.017 1 .142 .145 .224 .081
.224 .223 .227 .273 .667 .561 .579 .484 .149 .331 .323 .347 -.071 -.093 -.029 .471 .444 .399 .36 .329 .489 .456 .344 .491 .193 .248 -.025 .186 .904 .934 .917 -.054
Note: item loadings above .70 are highlighted
62
Utilitarian attitude
Perceived effort
Perceived usability
Perceived usage
Henodic attitude
Cust sat
Apps
Appendix H: Statistical validity of the base model
For the estimation of the significance of the parameters, bootstrapping was applied with a resample of 500. Displayed here are the supported hypotheses of the base model and the effect sizes of changes in R2 due to the deletion of insignificant effects.
Table 15 Hypotheses in the base model
Hypothesis H1b,alt (+) H1c,alt (-) H1d (+) H2 (+) H3a (-) H4b (+) H4c (+) H4f (+) H4f (+) H5a (+) H5c (+) H5e (-) H6f (+) H7 (+) Relation Apps ? Perceived Usability Apps ? Perceived Effort Apps ? Cust. Sat. Smartphone Perceived Usability ? Perceived Usefulness Perceived Usefulness ? SMS Usage Perceived Usefulness ? Frequency Of Use Perceived Usefulness ? Perceived Usage Level Perceived Usability ? Perceived Usage Level Perceived Usability ? Perceived Usage Level Perceived Usefulness ? Cust. Sat. Smartphone Perceived Usability ? Cust. Sat. Smartphone Perceived Effort ? Cust. Sat. Smartphone Frequency Of Use ? Mobile Data Usage Cust. Sat. Smartphone ? Telco Cust. Sat. ?† base model .35 -.2 .09 .54 -.22 .22 .34 .36 .36 .2 .44 -.12 .19 .29 T-value base model 6.41*** 2.97*** 2.3** 10.09*** 2.69*** 2.55** 3.78*** 4.08*** 4.08*** 2.68*** 5.86*** 2.14** 2.4** 3.66*** ?† reduced model .35 -.2 .09 .57 -.13 .3 .34 .13 .36 .2 .44 -.12 .31 .31 T-value reduced model 6.67*** 3.38*** 2.26** 11.35*** 2.34** 4.97*** 4.03*** 2.87*** 4.61*** 2.8*** 5.8*** 2.12** 5.39*** 4.77***
*<.1. **<.05, ***<.01 (Two Tailed); † Effect Sizes Are Standardized
Table 16 Effect sizes of changes in R2
Variable Cust sat smartphone Daily usage time Frequency of use Mobile data usage Perceived Effort Perceived usability Perceived usefulness Perceived usage SMS usage Telco cust sat Voice usage R2 base model .44 .05 .10 .14 .04 .12 .33 .35 .04 .12 .01 R2 reduced model .44 Variable deleted .09 .09 .04 .12 .32 .35 .02 .10 Variable deleted ƒ2 0 -.01 -.05 0 0 -.01 0 -.02 -.02 -
63
Appendix I: Post-hoc Mediation analysis
Table 17 Coefficients and significance of the effects
Base model without insignificant effects Relation Apps ? Cust. Sat. Smartphone Apps ? Perceived Effort Apps ? Perceived Usability Cust. Sat. Smartphone ? Telco cust. Sat. Frequency Of Use ? Mobile Data Usage Perceived Effort ? Cust. Sat. Smartphone Perceived Effort ? Perceived Usage Level Perceived Usability ? Cust. Sat. Smartphone Perceived Usability ? Perceived Usefulness Perceived Usability ? Perceived Usage Level Perceived Usefulness ? Cust. Sat. Smartphone Perceived Usefulness ? Frequency Of Use Perceived Usefulness ? Perceived Usage Level Perceived Usefulness ? Sms Usage ?
†
Mediator model T-value 2.23** 3.19** 6.55*** 4.72*** 5.39*** 2.2** 2.75*** 5.99*** 11.83*** 4.29*** 2.93*** 4.88*** 4.04*** 2.14** Relation Apps ? Cust. Sat. Smartphone Apps ? Hedonic Attitude Apps ? Perceived Brand Image Apps ? Perceived Effort Apps ? Perceived Usability Apps ? Subjective Knowledge Apps ? Utilitarian Attitude Cust. Sat. Smartphone ? Telco cust. Sat. Frequency Of Use ? Mobile Data Usage Hedonic Attitude ? Perceived Effort Hedonic Attitude ? Perceived Usability Perceived Brand Image ? Perceived Effort Perceived Brand Image ? Perceived Usability Perceived Effort ? Cust. Sat. Smartphone Perceived Effort ? Perceived Usage Level Perceived Usability ? Cust. Sat. Smartphone Perceived Usability ? Perceived Usefulness Perceived Usability ? Perceived Usage Level Perceived Usefulness ? Cust. Sat. Smartphone Perceived Usefulness ? Frequency Of Use Perceived Usefulness ? Perceived Usage Level Perceived Usefulness ? Sms Usage Subjective Knowledge ? Perceived Effort Subjective Knowledge ? Perceived Usability Utilitarian Attitude ? Perceived Effort Utilitarian Attitude ? Perceived Usability ?† .09 .17 .25 -.17 .14 .30 .22 .31 .31 -.04 .00 -.25 .50 -.12 .13 .44 .57 .36 .20 .30 .34 -.13 .08 .11 .03 .23 T-value 2.25** 3.01*** 3.92*** 2.53** 3.12*** 5.09*** 4.01*** 4.76*** 5.37*** 0.48 0.07 3.22*** 10.08*** 2.16** 2.67*** 6.01*** 11.5*** 4.25*** 2.63*** 5.02*** 3.91*** 2.17** 1.12 1.79 0.37 3.43***
.09 -.2 .35 .31 .31 -.12 .13 .44 .57 .36 .20 .30 .34 -.13
**: p < .05; ***: p< .01 (two-tailed). † Effect Sizes Are Standardized
Table 18 Calculation of the significance of the mediating effects
? Apps ? Mediator Mediating Relation Apps ? Perceived Brand Image ? Perceived Effort Apps ? Perceived Brand Image ? Perceived Usability Apps ? Utilitarian Attitude ? Perceived Usability **p< .05 Standardized .246 .246 .224 Unstandardized .126 .126 .069 Std. Dev. 2.61 1.34 2.93 1.13 0.80 Std. Error .0628 .0628 .056 ? Mediator ? Dependent Variable Standardized -.246 .498 .228 Unstandardized -.537 .420 .322 Std. Error .0764 .0494 .0665 Sobel ZValue -1.93 1.96** 1.19
Variable Apps Perceived Brand Image Perceived Effort Perceived Usability Utilitarian Attitude
See footnotes 2 and 3 for the formulae for the unstandardized coefficients and Sobel value.
64
Appendix J: MANOVA Post-hoc analysis
For the post-hoc analysis, Games-Howell tests were used. Only the tests for the variables that proved to contain significant differences between the operating systems, based on the between-subject effects, are displayed.
Table 19 Post-hoc comparisons between groups
Dependent Variable Apps Reference Category iOS Comparing Category Mean Difference Std. Error Sig.
BOS 2.79 0.46 .0000* Android 0.46 0.42 .5118 BOS iOS -2.79 0.46 .0000* Android -2.32 0.35 .0000* Android iOS -0.46 0.42 .5118 BOS 2.32 0.35 .0000* Cust. Sat. iOS BOS 2.99 0.52 .0000* Smartphone Android 1.42 0.21 .0000* BOS iOS -2.99 0.52 .0000* Android -1.56 0.52 .0119 Android iOS -1.42 0.21 .0000* BOS 1.56 0.52 .0119 Duration of daily iOS BOS -0.93 0.36 .0302 use Android 0.12 0.25 .8752 BOS iOS 0.93 0.36 .0302 Android 1.05 0.30 .0027 Android iOS 0.12 0.25 .8752 BOS -1.05 0.30 .0027 *Significant at the .05/54 = .0009 level. Such a strict level is used since there are 3 categories being compared across 9 variables (displayed here are three of the nine compared variables), leading to 3*2*1*9=54 comparisons.
65
doc_387751616.pdf
Consumer behaviour is the study of individuals, groups, or organizations and the processes they use to select, secure, and dispose of products, services, experiences, or ideas to satisfy needs and the impacts that these processes have on the consumer and society
Effects of apps on consumer behavior of smartphones and telecommunication providers: Feature fatigue vs. mass customization
by D.J.
Agterhuis
Student identity number: 0722219
In partial fulfillment of the requirements for the degree of Master of Science in Innovation Management
First supervisor: Second supervisor: Company Supervisor: Publishing date: University: Faculty: Department:
dr. ing. J.P.M. Wouters dr. J.J.L. Schepers S. Balkenende 9th of August, 2012 Eindhoven University of Technology Industrial Engineering & Innovation Sciences Innovation, Technology, Entrepreneurship and Marketing (ITEM)
Series Master Theses Innovation Management
Subject headings:
Apps, Smartphones, Feature Fatigue, Mass Customization, Customer Satisfaction, Technology Acceptance, Consumer Behavior, Telecommunication
2
Abstract
The use of apps on smartphones raises questions about how apps affect usage and satisfaction with the device and with telecom operators, and which of two concepts – feature fatigue or mass customization – is applicable to model these effects. The number of apps installed on the smartphone is proposed to predict three aspects of value: usefulness, usability, and effort. In turn, this value in turn is proposed to predict usage of both the device itself and of services provided by the operating company, and satisfaction with the device and the operator. The model is tested using survey results of 252 customers of a Dutch telecommunication provider that are combined with objective usage data. Results indicate that mass customization applies to the use of apps on smartphones since the number of apps positively relates to usability. In turn, usability is related to utility, which affects mobile data usage through the daily frequency of use of the smartphone. Satisfaction with the smartphone is directly affected by the number of apps and through value. Implications are that a high number of apps is not related to usability issues and that apps facilitate the transition in the telecom world of voice and SMS services to mobile data.
3
Acknowledgements
Naturally, I would like to thank my first supervisor Joost Wouters of Eindhoven University of Technology for guiding the search towards a suitable topic for graduation, for steering the project into the right direction and for his thoughts on the theories, which were a useful input. I want to thank my second supervisor Jeroen Schepers for his creative contribution to the model, for answering my long e-mails and for his support in the statistical analysis. Much appreciation is also attributed to Sanne, Caspar, Frederieke, Jan Z., Joppe, Michel, Nathalie and Unni for the fantastic and educating experience in the past six months at the “large Dutch telecom company.” In particular, I would like to thank Sanne for her contributions to the research. Also the funding of the market research and the access to the customer database of the company deserves recognition. Last but not least, much is obliged to Esmee, my family, “the sponsors down under”, my friends and Douwe Egberts for the (mental) support and the mere reaching of the graduation thesis, and for the creative discussions which definitely shaped it. Also many thanks go out to Tim and Jan A. for their detailed reviews. I hope you enjoy reading this report, and that you look differently at the palm of your hand after reading it. Dirk
4
Contents
SUMMARY 1 INTRODUCTION 2 THEORY AND HYPOTHESIS DEVELOPMENT 2.1 THEORY AND RECENT RESEARCH 2.1.1 APPS AND SMARTPHONES: MODULES FOR MODULAR PRODUCTS 2.1.2 TOO MANY FEATURES: FEATURE FATIGUE 2.1.3 MASS CUSTOMIZATION: ADAPTING FUNCTIONALITY TO MEET SPECIFIC NEEDS 2.1.4 CUSTOMER SATISFACTION 2.1.5 TECHNOLOGY ACCEPTANCE MODEL (TAM) 2.2 CONCEPTUAL FRAMEWORK AND HYPOTHESES 2.2.1 NUMBER OF APPS ON PERCEIVED VALUE AND CUSTOMER SATISFACTION 2.2.2 PERCEIVED EASE OF USE ON PERCEIVED USEFULNESS 2.2.3 ASPECTS OF VALUE ON USAGE OF THE SMARTPHONE, SERVICES PROVIDED
BY THE OPERATOR AND CUSTOMER SATISFACTION
7 10 11 12 12 12 13 14 14 15 17 19 19 20 21 21 22 22 23 25 25 27 27 28 28 29 29 30 30 33
2.2.4 USAGE OF THE SMARTPHONE ON USAGE OF TELECOMMUNICATION
PROVIDER SERVICES
2.2.5 CUSTOMER SATISFACTION OF THE SMARTPHONE ON CUSTOMER
SATISFACTION OF THE TELECOM PROVIDER
2.2.6 MODERATING VARIABLES 3 METHOD 3.1 3.2 3.3 3.4 DATA SAMPLE MEASUREMENT INSTRUMENT DATA COLLECTION PROCEDURE ANALYTICAL METHOD
4 RESULTS 4.1 SAMPLE AND DATA CHARACTERISTICS 4.1.2 COMMON METHOD BIAS 4.1.3 MISSING DATA, OUTLIERS AND MULTIVARIATE ASSUMPTIONS 4.2 MEASUREMENT MODEL 4.2.1 CONSTRUCT RELIABILITY 4.2.2 CONSTRUCT VALIDITY AND MULTICOLLINEARITY 4.3 STRUCTURAL MODEL 4.3.2 MODERATING VARIABLES
5
4.3.3 CONTROL VARIABLES 4.4 POST-HOC ANALYSES 4.4.2 DIFFERENCES BETWEEN OPERATING SYSTEMS 5 DISCUSSION 5.1 5.2 5.3 5.4 FINDINGS SCHOLARLY IMPLICATIONS MANAGERIAL IMPLICATIONS LIMITATIONS AND AVENUES FOR FURTHER RESEARCH
34 34 36 37 38 41 41 43 45 45 49 50 51 56 58 59
6 REFERENCES 6.1 LITERATURE REFERENCES 6.2 WEB REFERENCES APPENDIX A: MEASUREMENT INSTRUMENT APPENDIX B: QUESTIONNAIRE APPENDIX C: DATA MANIPULATION APPENDIX D: GENERALIZATION OF THE SAMPLE APPENDIX E: DATA EXPLORATION: OUTLIERS AND NORMALITY APPENDIX F: FACTOR LOADINGS AND CROSS LOADINGS OF THE FIRST CFA APPENDIX G: FACTOR LOADINGS AND CROSS LOADINGS OF THE DEFINITE CFA APPENDIX H: STATISTICAL VALIDITY OF THE BASE MODEL APPENDIX I: POST-HOC MEDIATION ANALYSIS APPENDIX J: MANOVA POST-HOC ANALYSIS
61
62 63 64 65
6
Summary
The revolution of apps and smartphones demands insights in how these apps affect customer satisfaction and usage behavior of smartphone consumers with respect to the device itself and usage of services offered by telecommunication providers such as voice, SMS and mobile data. Two concepts compete in modeling these effects: feature fatigue and mass customization. Apps are assumed to be features since the use of apps offers functionality in addition to the hard technical specifications of smartphones. Consumers suffering from feature fatigue have purchased a feature-packed product which is high in utility but low in usability because of that high number of features (Thompson et al., 2005). Because consumers can modulate the functionality of their device by adding or deleting apps, it is interesting to assess whether the phenomenon is suitable to model effects of apps on consumer behavior of smartphones. A competing concept is mass customization, stating that manufacturers meet market needs by allowing consumers to adapt the functionality of products until the latest possible moment in the supply chain (Chase et al., 2006). For smartphones, this can be done with apps even after purchase of the device. This study contributes by adding to the sparse body of research in smartphones and apps. Moreover, because of the potential negative effects of feature fatigue, it is useful for managers to know whether this phenomenon applies to smartphones. Studying the effects of apps on smartphone usage & satisfaction is also deemed useful for practitioners in the telecom industry due to the transition of voice and SMS to mobile data usage. Theory and conceptual framework A literature review was conducted, revealing (recent) research on smartphones and apps, feature fatigue, customer satisfaction and the Technology Acceptance Model (TAM). The basis for the theoretical framework is that apps affect the value of the smartphone as perceived by consumers, which in turn affects usage and satisfaction. For modeling customer satisfaction, a transaction-specific and postpurchase view of customer satisfaction is used in combination with the comparison-standards paradigm in which perceived value is an experience-based standard for comparison (Cadotte et al, 1987). Customers’ perceived value of the smartphone is a tradeoff between benefits and sacrifices perceived by consumers (Eggert & Ulaga, 2002), in which the benefits are adopted from the Technology Acceptance Model: perceived ease of use and perceived usefulness. In the TAM, perceived ease of use and perceived utility are the main variables that indirectly explain the actual use of a product. The sacrifice is consumers’ perceived effort exerted to install, update and to learning to use the apps. These aspects of value were hypothesized to be predicted by the number of apps on the smartphone of a consumer. In turn, these aspects of value predict the usage of a smartphone and of services provided by the telecommunication operator. Usage of the smartphone was broken down into daily frequency of use, daily duration of use and the perceived usage level. Services of the telecom provider that were adopted in the model were voice, SMS and mobile data. The number of apps installed on the smartphone and the aspects of value were also predicted to affect customer satisfaction with both the smartphone and the telecom provider.
7
Moderating effects were also proposed, which were adopted from the consumer decision making process: consumers hedonic or utilitarian attitude towards smartphones, the brand image as perceived by consumers and the subjective knowledge about smartphones were all expected to moderate the relation between the number of apps installed on the device and the aspects of perceived value. Method The model is tested with an online survey among 252 customers of a Dutch telecommunication provider. The customers in this population were postpaid smartphone users in possession of a smartphone of the brand Apple, Blackberry, HTC or Samsung in the consumer segment. Results of the survey are combined with objective data of voice, SMS and mobile data usage. The measurement instrument used to measure the concepts in the hypothesized model was based on prior research. Control variables are age, gender, education, duration of smartphone possession, number of smartphones owned, smartphone brand and type of use (business or consumer. The model was tested using Structural Equation Modeling (SEM) with Partial Least Squares (PLS) as estimation method. Results The data sample appeared to reflect the customer base of the telecom operator. After deletion of a low number of items with insufficient loadings and cross-loading items, the constructs in the model were measured reliably and convergent and discriminant validity were warranted. Most variables or constructs have an R2 which is moderate to high for consumer behavior studies, but the perceived effort, SMS usage and voice usage were weakly explained. Of the 40 hypotheses in the base model, 13 were confirmed. The number of apps on consumers’ smartphones was positively related to the perception of usability of the device and satisfaction with the device, but negatively to the effort perceived to download, update and learning how to use the apps. Perceived usability was positively related to the perception of usefulness of the smartphone. Regarding the aspects of value, all three constructs were positively related to the perceived usage level, but only the perceived usefulness of the smartphone was positively associated to the daily frequency of use of the smartphone. In turn, consumers that use their device more often per day, often have a higher mobile data usage. Moreover, consumers perceiving their smartphone useful, tend to have a lower SMS usage, albeit that only 4% of the variance of SMS usage is explained. Besides the number of apps, customer satisfaction with the smartphone was related negatively to the perceived effort, but positively to the perceived usefulness and the perceived usability. In turn, consumers that are satisfied with their smartphone are disposed to be satisfied with the telecommunication company. Of the moderating effects, only a negative moderating effect of subjective knowledge was confirmed on the relation between the number of apps installed and the perceived usability of the smartphone. The results gave rise to two post-hoc analyses: A one-way MANOVA for assessing differences between operating systems, and the analysis of mediating effects of subjective knowledge, utilitarian attitude, hedonic attitude and the perceived brand image. It appeared
8
that Blackberry users have less apps installed than users of other operating systems, and that Apple users are generally more satisfied than Blackberry or Android users. Finally, consumers’ utilitarian attitude mediates the effect of the number of apps on the perception of usability. Discussion Because the number of apps installed on the smartphone positively related to the perceived usability and negatively to the perceived effort for installing, updating and learning to use apps, the concept of mass customization applies to apps on smartphones and no support is found for feature fatigue. The utility of the smartphone as perceived by consumers plays a key role in predicting the usage: it is directly associated with decreases in SMS usage, and through an increased frequency of daily use it is positively related to mobile data usage. The number of apps and all aspects of value related to satisfaction with the smartphone, the only negative relation being the perceived effort. No direct effects of apps or value of the smartphone predict satisfaction with the operator, albeit that users that are satisfied with their smartphone do tend to be satisfied with their telecom provider. This implies that no crossover effects exist for the value of the smartphone or the number of apps towards the operating company. A main implication for scholars is that they can apply mass customization to model the use of apps on smartphones , and that feature fatigue does not apply to modular products. Integrating customer satisfaction and technology acceptance proved to provide a model with which effects on apps could be adequately predicted. Managerial implications are that managers need not worry about negative effects of a high number of apps. In particular, usability issues are not associated with a high number of apps, on the contrary: mass customization provides opportunities for app developers and for managers in the telecom sector. Moreover, apps can be used as leverage in the transition from voice and SMS to data usage due to the indirect negative associations with SMS usage and positive relations to mobile data usage. To improve satisfaction, managers could consider automating the updating process of apps on the device, and spend more effort on promoting Apple’s high end devices since these customers are generally more satisfied. Finally, since expert users benefit less from increases in perceived usability due to increases in the number of apps installed, self-service programs should aim at usability issues for expert users. Limitations of the study are that no distinction is made between different categories of apps; the aggregation of all apps on the smartphone and satisfaction of the apps to a central, absolute number of apps; the omitting of psychometric differences based on the moment of adoption; a weak theoretical basis for the mediation analysis; and the fact that the study does not differentiate between the number of apps installed and the number of apps actually used. Research design-related issues are the sample size when considering the serious non-normal distribution of the data, using single-item constructs, and the nature of the data which does not allow causal inferences. These issues are directions for further research.
9
1 Introduction
With the convergence of communication technologies and innovative product features in the past decade (Arruda-Filho et al., 2010) the smartphone has arisen: a combination of handheld computers and mobile phones accommodated with a wide variety of features such as camera’s, organizers, web browsers, media players and navigation. The smartphone has “put the world in the palm of your hand” (Madison, 2011 on dailymail.co.uk). In the first quarter of 2012, smartphones sales accounted for 34% of total mobile phone sales (Gartner.com, 2012). In the Netherlands, smartphone penetration has reached 53% in January 2012 (Telecompaper, 2012). It is predicted that smartphone sales will approach one billion units in 2015 (Idc.com, 2011). With this revolution of smartphones, a new way of consuming software applications has surfaced in the form of ‘apps.’ This study assumes that apps are features. In the context of software, features are defined as a distinguishing characteristic of a software item such as performance, portability or functionality (IEEE, 1998). Apps combine hard technical features such as the camera and a connection to the internet in order to provide additional functionality, which is the basis for the assumption. Since apps can be added and deleted, this assumption results in viewing smartphones as modular products – products of which the functionality can be altered (Stone, 2000). The use of apps raises two important questions about the use of smartphones. For one, what are the effects of the use of apps on the behavior of smartphone consumers? Two very different concepts can be used for modeling the use of apps: feature fatigue and mass customization. Consumers that suffer from feature fatigue have purchased a feature-packed product that is complex in use, while they actually only want a simple, easy to use product (Thompson et al., 2005). Concerns of the phenomenon have been even expressed for all-inone portable devices (Taliuaga et al., 2009 on rockresearch.com) and in particular for “app fatigue” (Kendrick, 2011 on Zdnet.com, 2012). On the other hand, mass customization posits that manufacturers can meet specific needs of consumers by providing flexible market offerings (Anderson & Narus, 1998). Apps allow smartphones to be products of which the functionality is customizable after purchase. This thesis provides an answer to the question which of these theories is best applicable to model the effects of apps on behavior of smartphone consumers. Additionally, it is interesting to assess whether external factors influence the effects of apps on behavior of smartphone consumers. A second question stemming from the use of apps is: What is the effect of apps on the use of services offered by telecommunications providers? A transition is in process in which the more traditional services provided by operator companies (or opco’s) such as voice (calling) and SMS (short message service) are moving more towards the consumption of mobile data (Tilson & Lyytinen, 2006). Apps facilitate this transition since apps exist that replace the use of SMS and voice with the use of mobile data, such as ‘Whatsapp’ or ‘Skype’. The aim of this research is to answer these two questions, thereby shedding light on the effect of apps on the behavior of smartphone consumers. The answers to these questions have both theoretical and practical implications. For scholars, this thesis adds to the existing body of empirical research on smartphones and apps, which is sparse according to Peslak et al. (2011). The research also adds smartphones and apps as a case study to the existing
10
research on consumer behavior, such as customer satisfaction research and technology acceptance research, and it is assessed which of two concepts is applicable to modular products such as smartphones and apps: mass customization versus feature fatigue. A final unique aspect of this study is the use of objective data of usage of telecom operator services in order to study effects of apps on consumer behavior of smartphones. Practically, it is useful for managers to know whether feature fatigue is applicable to smartphones since the phenomenon affects consumer behavior, with reductions in customer satisfaction levels and usage levels as a consequence (Thompson et al., 2005). Along with the positive impact of customer satisfaction on market performance indicators (Anderson & Sullivan, 1993; Fornell et al, 1996; Szymanski & Henard, 2001), customer retention through customer satisfaction is especially relevant in the telecommunications industry in which drop-out rates are high (Lee et al., 2001). Therefore, cross-over effects of satisfaction of the smartphone to satisfaction of the opco are also assessed. Moreover, practitioners working in the telecommunications industry will gain insights in the effects of the use of apps on the use of services provided by telecommunication providers, which is important in the context of the transition of voice and SMS towards mobile data. Additionally, findings of this study are relevant for practitioners in the telecommunications industry who want consumers to have an optimal experience with their handset, which is a general strategic trend in telecommunications (Haverila, 2011b; Ojiako & Maguire, 2009). The remainder of this writing is organized as follows: In the following section, theories and prior research of relevant concepts will be discussed, thereby elaborating upon smartphones and apps, feature fatigue, mass customization, customer satisfaction and technology acceptance. This is followed by the formulation of a conceptual framework and the development of hypotheses. Section 3 will discuss the methodology that is used to empirically test the model, followed by Section 4 in which the results of the empirical testing will be presented. Finally, section 5 will discuss the findings, theoretical and practical implications, and the limitations and directions for further research.
2 Theory and hypothesis development
This chapter will present results of a literature review, thereby elaborating on concepts that are relevant to the construction of the conceptual framework with which the use of apps and smartphones can be analyzed. The chapter starts by discussing smartphones and smartphone apps, followed by a brief discussion of feature fatigue, mass customization, customer satisfaction and the technology acceptance model (TAM). Subsequently, the theoretical framework is presented and the hypotheses are developed. The literature review was conducted according to the guidelines of Randolph (2009); the focus was on research outcomes and theories and the goal was to develop a model with which effects of apps on consumer behavior could be modeled. The covered literature was a purposive sample of the available literature, and the review was conceptually organized. Literature was searched by prompting (combinations of) key concepts into several academic search engines such as Google Scholar, Wiley, JSTOR, ABI/Inform, ISI Web of Knowledge and Sciencedirect. Literature was selected based on scanning titles and abstracts for relevance, and was evaluated according to Google Scholar citations and appearance of the publication source in the Harzing Journal Ranking (Harzing, 2011).
11
2.1 Theory and recent research
2.1.1 Apps and smartphones: modules for modular products Various but similar definitions are used for the smartphone. This research uses the following: ‘a smartphone is a mobile phone that includes software that a user is able to modify and update” (Töyssy & Helenius, 2006, p. 110). Smartphone apps are computer programs that can be installed on a smartphone. Different categories of smartphone applications can be identified, such as communication, browsing, media playing, productivity, system, gaming and map apps (Falaki et al., 2010). The apps result in numerous possibilities for smartphone use such as in healthcare (Park & Chen, 2007), mobile commerce (Chang & Chen, 2004), logistic services (Chen et al., 2007), mobile-based corporate intranet systems (Funk, 2006) and sustainable technology (Pitt et al., 2011). Most smartphone users have around 50 applications installed on their smartphone, while they use only 8-12 applications regularly and on a daily basis (Falaki et al., 2010). An important assumption of this study is that apps are viewed as features because apps provide additional functionality in addition to the hard technical features of smartphones. In the context of software, features are defined as a distinguishing characteristic of a software item such as performance, portability or functionality (IEEE, 1998). Smartphones have hard features, such as a camera, which can in turn be used by software (apps) in order to provide additional functionality. An example is the Sleep Cycle Alarm Clock app, which uses the accelerometer and the microphone in order to analyze sleep patterns and to wake you up in the lightest sleep phase within the half hour before you have set the alarm (iTunes.apple.com, 2012). The author believes that this is an adequate example of why apps can be assumed to be features. The assumption leads to the observation that smartphones are modular products. Modular products are machines, assemblies or components that accomplish an overall function through combination of distinct building blocks or modules (Pahl & Beitz, 1998, in: Stone, 2000). The number of hard technological features does not differ greatly between different smartphone types because most smartphones have similar designs. However, in the set of apps that users have installed in their device, variation can be expected, which supports the need to assume that apps are features. 2.1.2 Too many features: Feature fatigue As technology advances, it becomes more feasible to load products with a large number of features or functions. Adding features makes products more appealing for consumers due to increased expected utility of a product, which is in turn appealing for businesses due to the predicted market share increase. However, too many features can make a product overwhelming for consumers and difficult to use, resulting in consumers’ dissatisfaction with their purchase and a state of frustration which is called feature fatigue (Thompson et al., 2005). The existence of this post-purchase phenomenon is underlined by an increase in product returns for products that turn out not to have any technical malfunction (Den Ouden et al., 2005). Feature fatigue can be explained by a shift in consumer preferences before and after the use of a product, illustrated in Figure 1. Before purchase, consumers believe each feature
12
to add to the utility of a product, knowing that more features make products more difficult to use. However, after having used the product, a shift occurs in consumer preferences towards a simple product with basic functionality which is easy to use. This structural change can be explained by different considerations being salient in expected and experienced utility (Hsee et al., 2009; Thompson et al., 2005). Economic theory models consumers’ preferences with an additive utility function (Lancaster, 1971 in: Thompson et al., 2005): adding attributes to a product or service that consumers evaluate as positive increases the perceived utility of that product or service. Market research techniques such as conjoint analysis or discrete choice analysis model products as bundles of attributes in which each attribute has value (Srinivasan et al., 1997 in: Thompson et al., 2005). Hence, products are bundles of features, and each feature adds value to the product’s utility. A tradeoff is postulated between an increase in this utility perception of consumers and a decrease in the perceived usability. Figure 1 Shift in preference for utility to usability. Adapted from Gaigg (2012) on Michealgaigg.com Feature fatigue has been found to affect consumer behavior, or more specific: satisfaction and usage (Thompson et al., 2005). Consumer behavior “encompasses the acquisition, consumption, and disposition of goods, services, time, and ideas by decision making units (e.g. individuals, families, organizations, etc)” (Jacobi, 1978, p. 87). Various theories are used to model consumer behavior, varying from consumer choice theories such as the Black Box model on the consumer decision making process, to models on the adoption of innovations by consumers such as the Technology Acceptance model (Davis, 1989) or other diffusion models and consumer decision making models. Two major streams of research on consumer behavior are discussed later in this chapter: Customer Satisfaction and Technology Acceptance. 2.1.3 Mass customization: adapting functionality to meet specific needs By varying in the set of product features, each consumers’ specific desire in product capability can be adhered to. Suppliers can meet these specific desires by providing flexible market offerings (Anderson & Narus, 1998), which corresponds to mass customization. Originally, mass customization has been interpreted to create value based on customercompany interaction at the manufacturing and assembly stage of operations, thereby creating customized products (Kaplan & Haenlein, 2006). However mass customization is more recently conceptualized as postponing the task of differentiating a product for a specific customer until the latest possible point in the supply network (Chase et al., 2006). This is certainly the case for smartphones for which apps are obtained after purchase of the smartphone. This interpretation is facilitated by the assumption that apps are features, also since modularity is often viewed as a key factor for mass customization since modularity provides the flexibility for quick and inexpensive customization (Feitzinger & Lee, 1997).
13
The concepts of mass customization and feature fatigue lead to two different viewpoints of how consumers use the apps on their smartphone: applying mass customization results in the view that consumers adapt the functionality of their device according to their preferences. On the other hand, feature fatigue indicates that consumers install a large number of apps because of the expected utility, but end up being frustrated because their device is complex in use. In the following sections, customer satisfaction and technology acceptance are described, which are two types of consumer behavior that can be affected by apps on smartphones. 2.1.4 Customer satisfaction Customer satisfaction is an attitude judgment following a purchase act or a series of consumer-product interactions (Yi, 1990 in: Fournier & Mick, 1999) and is often about evaluation as a result of comparison (Cadotte et a., 1987; Oliver, 1993), which is the basis for the comparison standards paradigm. In this paradigm, different norms are used by consumers in order to form standards for comparison. Such norms can be expectations, experiences, desires or equity (Cadotte et al., 1987; Eggert & Ulaga, 2002; Fournier & Mick, 1999; Halstead, 1999). Halstead (1999) offers a typology of customer satisfaction models based on 1) the level of aggregation of the comparison standard; 2) the stage of the comparison process, and 3) the level of abstraction of the comparison. Customer satisfaction can be measured for individuals per transaction, or all individual satisfaction levels of consumers over a series of transactions can be aggregated to a cumulative level (Anderson & Sullivan, 1993). The stage of the comparison process refers to the stages in the consumer decision process such as need recognition or post-purchase evaluation. Finally, for assessing the level of abstraction of the comparison, the hierarchy of value of Gardial et al. (1994) can be used, stating that satisfaction studies focus on 1) satisfaction with product attributes; 2) satisfaction with products overall; or 3) on global satisfaction of consumers. Customer satisfaction is often a performance indicator for companies. Increased satisfaction leads to positive word of mouth and increased customer loyalty, which is the ‘ultimate dependent variable’ of customer satisfaction because of its value for actual customer retention and profitability (Johnson et al., 2001, p. 222). 2.1.5 Technology Acceptance Model (TAM) Originally introduced by Davis (1989), the Technology Acceptance Model posits that external variables influence a technology’s ease of use (usability) and usefulness (utility) as experienced by users of that technology. Both ease of use and usefulness influence consumers’ attitude towards using a technology, which in turn affects the behavioral intention to use it, which finally affects actual use of a technology and the continued use thereof (Davis, 1989). The TAM is based on the Theory of Reasoned Action (TRA). The TRA stems from social psychology and states that one’s behavioral intention depends on the person’s attitude about that behavior and subjective norms, and that if one intends to conduct a certain
14
behavior, it is likely that he or she will do so (Fishbein & Ajzen, 1975). Later variations of the TAM leave out the attitude construct (Venkatesh et al., 2003). The theoretical foundation for using the perceived ease of use and usefulness as predictors for usage behavior can be found in self-efficacy theory and the cost-benefit paradigm. Self-efficacy is defined as “judgments of how well one can execute courses of action required to deal with prospective situations” (Bandura, 1982, p. 122 in Davis, 1989). The cost-benefit paradigm stems from behavioral decision theory and describes decision making strategies in terms of a cognitive tradeoff between the effort required and the quality of the resulting decision (Payne, 1982). Based on the cost-benefit paradigm and self-efficacy theory, Davis (1989) distinguishes between outcomes judgments and self-efficacy judgments, in which the outcomes relate to usefulness of using a technology, and self-efficacy relates to how easy that technology is to use. Both these judgments, or perceptions, of consumers determine the acceptance and usage of a technology by consumers. The Technology Acceptance Model has been extensively tested in research and numerous variations and extensions of the model have been developed. An overview of several variations and extensions is provided by Venkatesh et al. (2003).
2.2 Conceptual framework and hypotheses
This section will first describe the ‘theoretical lens’ with which the questions posed in the introduction were approached, followed by a discussion of the concepts and the relations between them. The conceptual model and the proposed hypotheses can be viewed in Figure 2. It should be noted that, based on the initial literature review, an elaborate model was developed initially. However, only part of that model is discussed here in order to present a simplified model. The framework in Figure 2 can be used to analyze the effects of apps on behavior of smartphone consumers. The number of apps installed on the smartphone predicts the value perceived by consumers. Derived from customer satisfaction research, value is a standard for comparison which consumers use to assess their satisfaction. The majority of customer satisfaction research focuses on expectations as standard for comparison, which is however not relevant for post-purchase processes (Halstead, 1999) and durable goods (Churchill & Surprenant, 1982). Since smartphones are durable goods and apps are acquired after purchase of the smartphone, value as an experience-based norm is used as a standard of comparison. Satisfaction models based on experience-norms use the confirmation/disconfirmation paradigm in combination with brand attribute beliefs and the experience of using a product (Cadotte et al., 1987). Moreover, satisfaction is viewed here as transaction-specific (and not cumulative) since this view is more suitable for assessing individual differences between consumers (Halstead, 1999). Value is based on the exchange theory of marketing (Eggert & Ulaga, 2002), cognitive psychology and economic theory (Thaler, 1985 in: Gallarza & Saura, 2006): perceived value relates to consumer behavior based on the concept of value transaction (Gallarza & Saura, 2006). Value is defined as the benefits customers receive in relation to total costs or ‘sacrifice’ (McDougal & Levesque, 2000). Sacrifices can be monetary and non-monetary (time and effort) costs associated with acquiring and using a product or service (Cronin et al.,
15
Moderators
Usage Opco services
Perceived brand image
Subjective knowledge
Hedonic attitude
Utilitarian attiude
Voice
Mobile data
SMS
Perceived value of the smartphone
Perceived usefulness
H3a-g H6a-i
H8a-c
H9a-c
H10a-c
H11a-c H2(+)
Perceived usability
Usage of the smartphone
H4a-h
Daily frequency of use
Daily usage time
Perceived usage
Number of apps
H1a-c
Perceived effort
H5a-f
Satisfaction
H1d,e(+)
Customer satisfaction smartphone
H7(+)
Telecom provider satisfaction
Figure 2: the hypothesized model
2000). The benefits that a consumer perceives from product features are adopted from the acceptance Model: perceived ease of use and perceived usefulness, which both predict usage (Davis, 1989). In the original TAM, the intention to use is an intermediate variable between usefulness & usability, and usage behavior. However, the intention to use is not adopted in the conceptual framework in order to simplify the model. As in a later extension of the TAM, the Unified Theory of Acceptance and Use of a Technology (Venkatesh & Davis, 2003), also the attitude towards using a technology is not adopted. Perceived usefulness is defined as the belief that using a product or service positively relates to job performance (Davis, 1989), which can be extended to performance in everyday life for consumer products. Perceived ease of use is defined as the degree to which a person believes that using a particular product or service would be free of effort (Davis, 1989), in which effort is conceptualized to be a finite resource that a person may allocate to the various activities that he or she is responsible for (Radner & Rothschild, 1975, in: Davis, 1989). Outcomes of perceived value in this framework are actual usage levels and customer satisfaction. Value and satisfaction are related since they are both evaluative judgments about products or services (Woodruff, 1997). However, they are two distinct constructs due to satisfaction being the result of an affective comparison process and value being the result of a cognitive comparison process. In the model, satisfaction stems from value being an experienced-based standard for comparison which consumers use to assess their satisfaction. Usage is also related to value as it is the result of the value-variables perceived ease of use and perceived usefulness in the Technology Acceptance Model. In the outcome variables usage and satisfaction, distinction is made between the smartphone and the telecommunication operator. For satisfaction, this means that customer satisfaction towards both the telecom provider and the smartphone manufacturer is assessed. Usage is divided into usage of the smartphone, and usage of services provided by the operating company: voice
16
(calling), SMS and mobile data. Usage of the smartphone is broken down into daily usage time, daily frequency of use and the perceived usage level (Al-Gahtani & King, 1999). Recent research has found that the frequency of use and the average duration of usage varies greatly among users of smartphones (Falaki et al., 2010), which is why usage of the smartphone is divided into three separate variables. Finally, variables influencing the relations between the number of apps and the aspects of value are derived from the consumer decision making process (Lamb et al., 2012). Klein (1998) proposes that several characteristics of consumers shape the information-search phase in this process. The following consumer characteristics are adopted in this study: subjective knowledge, attitude towards the smartphone and the perception of the brand’s image in the mind of the consumer. In the following sections, the proposed relations between the concepts are described. 2.2.1 Number of apps on perceived value and customer satisfaction A central predictor in the model for the value of the smartphone as perceived by consumers is the number of apps. Based on the assumption that apps are features, this study focuses on the number of apps installed on the smartphone as the number of features. The number of apps is proposed to predict value and satisfaction. Two competing theories, each leading to a different set of hypotheses, are used to predict the relations between the number of apps and the aspects of value: feature fatigue and mass customization. According to feature fatigue, each additional feature is expected to increase perceived usefulness, to decrease usability, and to increase perceived sacrifice. Due to the added utility of each app, apps are expected to increase perceived usefulness. Moreover, each additional feature is “one more thing to learn, one more thing to possibly misunderstand, and one more thing to search through when looking for the thing you want” (Nielsen, 1993, p. 155 in Thompson et al., 2005). Therefore, a negative relation is expected between the number of apps and the perceived usability. This increase in utility and decrease in usability due to an increase in the number of apps corresponds to feature fatigue. Then there is the cost-aspect of value: applications require resources in the form of time and effort to install and update the applications and to learn to use the applications effectively. Barriers for customer satisfaction of software products that have been found in previous research are effort for installation and maintenance (Kekre et al., 1995) which correspond to the effort of customers required to install and maintain (update) apps. Because a large number of applications is offered free of charge and because the price of apps differs per operating system, monetary costs are not adopted in the model. However, apps are expected to require effort and time since users are to download, install and regularly update the apps and since users are to learn how to use the apps on their smartphone. Therefore, based on the perspective of feature fatigue, smartphone users with a high number of apps are expected to be more likely to perceive a high degree of effort. On the other hand, the concept of mass customization contradicts the expectations that are based on feature fatigue regarding usability and effort. The modular property of smartphones can be considered an anomaly of feature fatigue: consumers can modify the functionality of their device after purchase by adding or deleting apps according to their preferences. Based on this anomaly, mass customization is proposed as a theory competing
17
with feature fatigue as an explanation of the effects of apps on behavior of smartphone consumers. Applying mass customization creates value (Kaplan & Haenlein, 2006) which can be attributed to the increased added utility of each app or feature and also to increased perceptions of usability (Kamis et al., 2008). A positive relation can be proposed between the number of apps and usability based on mass customization: smartphone users with a high number of apps have customized the utility of their smartphone and are also expected to have customized the usability of their smartphone, e.g. by configuring different home-screens and shortcuts that make their device more easy to use. The number of apps installed on consumers’ smartphone may well contribute to a users control over the functionality and usability of their device. Therefore, based on the view of mass customization, a positive association between the number of apps and the perception of usability can be expected. Also, a negative association between the number of apps and the effort perceived by consumers is expected: based on mass customization, it can be expected that users of modular products adapt the functionality of their device according to their preferences. Those users will associate a higher number of apps on their smartphone with a desirable degree of effort exerted into downloading, updating and learning how to use those apps. Moreover, a learning curve can be expected in which consumers that have a high number of apps, perceive less effort for the high number of apps because they are have developed an aptitude for using the apps. This lower perception can be especially expected when they have installed a number of apps conform to their need and have customized their product. Moreover, the number of apps is expected to relate to customer satisfaction directly, which is in line with previous research findings of products attributes and customer satisfaction: according to the hierarchy of value by Gardial et al. (1994), customer satisfaction stems from product attributes. In the context of this research, smartphone users with a high number of apps are expected to be more likely to be satisfied with their smartphone, based on this hierarchy of value. Additionally, cross-over effects are expected between the number of apps on the smartphone and the satisfaction with the telecommunication provider. Mittal et al. (1998) find that users of consumption systems (a combination of products and services) attribute satisfaction of products to service providers over time. Since telecommunication operators deliver smartphones to consumers and provide the services necessary to operate smartphones, they form a consumption system and consumers can be expected to relate their satisfaction with their handset to their satisfaction with the operating company. The following relations are hypothesized between the number of applications that consumers have installed on their smartphone, and the aspects of value and customer satisfaction (note that the original hypotheses reflect feature fatigue and that the alternative hypotheses reflect mass customization):
H1a-e The number of apps installed on the smartphone is (a) positively related to perceived usefulness, (b) negatively related to perceived usability, (c) positively related to perceived effort and is positively related to customer satisfaction of (d)the smartphone and (e) the telecom provider The number of apps installed in the smartphone is (b,alt) positively related to perceived usability and (c,alt) negatively related to perceived effort
H1b,c,alt
18
2.2.2 Perceived ease of use on perceived usefulness Extensive empirical evidence has been generated in previous research for the relation between perceived usability and perceived usefulness, which is why this relation is not widely elaborated. The underlying explanation is that products that are easy to use increase the utility that consumers perceive to enjoy from the product in use (Davis, 1989; Venkatesh & Davis, 2000). Consumers are more likely to perceive a product as useful when they perceive that product to be easy to use, than when they experience difficulties in operating the product. This is especially true for products in the information systems category (Venkatesh et al., 2003), such as smartphones. The following relation is hypothesized:
H2 Perceived usability is positively related to perceived usefulness
2.2.3 Aspects of value on usage of the smartphone, services provided by the operator and customer satisfaction In the Technology Acceptance Model and variants thereof, perceived usefulness and perceived usability are positively related to (attitude towards) usage (e.g. Davis, 1989; Venkatesh & Davis, 2000). Smartphone consumers that believe their device is useful are expected to indicate to 1) use it for more hours per day 2) more often per day, 3) and also to actually say to use it more, when compared to users that perceive their device to be useless. This is expected because they see more purposes for using their smartphone. Additionally, users that perceive their device as easy to use are also expected to use it longer, more often and to perceive to use it more: when the smartphone is easy to handle, there are less barriers to use the device for a longer period of time or more frequent, which also explains why those users believe they have a high degree of usage. Also users that perceive to exert a high degree of effort into their smartphone for installing, updating and learning how to use the apps on their smartphone are expected to use it more often, longer and to perceive a higher degree of usage. This expectation is a direct consequence from the concept of effort, i.e. people that perceive to spend time on updating, installing and learning how to use the apps on their smartphone probably tend to use it more. Besides the positive associations expected between the value aspects of the smartphone and the usage aspects of the smartphone, associations are also expected between the value aspects of the smartphone and the usage of the services provided by telecommunication providers: SMS and voice services and mobile data. Usability and usefulness are expected to relate negatively to SMS and voice usage, but positively to data usage, based on the following rationale: people that perceive their device as easy to use have low barriers to using the apps on their device. Moreover, if consumers believe that their smartphone is useful, they will be expected to use more apps in order to make use of the full potential of their device. Since the barriers to use their smartphone is low and they use more apps, consumers are also expected to use more mobile data since many apps require a connection to the internet. However, with the transition of SMS and voice usage towards mobile data usage in mind, it may be expected that apps which use mobile data are replacements for the more traditional phone services. Therefore, it is expected that consumers that perceive their device as useful, make less use of voice and SMS. The same relation is expected for usability because of the lower barriers for the interaction with the smartphone: consumers that consider that
19
interaction as easy, have more interactions with the apps on their smartphone and consequently often use more mobile data and use less traditional telecom services. Finally, the effort that consumers perceive to have put into the smartphone’s apps are only expected to relate positively to mobile data usage, since this is the only service provided by opco’s that is needed to use apps. The value perceived by consumers is also expected to relate to the degree of satisfaction with both the smartphone and the provider since in the model it is the standard that consumers use to assess their satisfaction. Customer perceived value has been found to directly affect satisfaction (Eggert & Ulaga, 2002; McDougal & Levesque, 2000; Spiteri & Dion, 2004) and especially usability has been found to be an important driver of overall customer satisfaction for software products (Flavian et al, 2005; Kekre et al., 1995). Consumers that perceive their smartphone as useful and easy to use are expected to rate their satisfaction level with the smartphone as higher than consumers who believe otherwise. On the other hand, effort is a negative aspect of value, and therefore effort is expected to be negatively related to satisfaction. Based on the cross-over effects between products and services (Mittal et al., 1998) discussed in section 2.2.1, the aspects of value are also expected to be associated with the satisfaction that customers assign to the provider of the services that are needed to use their smartphones. Based on the above, the following relations are hypothesized:
H3a-c H3d-f H3g H4a-c H4d-f H4g-h H5a, b H5c,d H5e,f Perceived usefulness is negatively related to (a) SMS usage and (b) voice usage but is positively related to (c) mobile data usage Perceived ease of use is negatively related to (d) SMS usage and (e) voice usage but is positively related to (f) mobile data usage Perceived effort is positively related to mobile data usage Perceived usefulness is positively related to (a) daily usage duration, (b) daily frequency of usage and (c) perception of usage Perceived usability is positively related to (d) daily usage duration, (b) daily frequency of usage and (c) perception of usage Perceived effort is positively related to (a) daily usage duration, (b) daily frequency of usage and (c) perception of usage. Perceived usefulness is positively related to customer satisfaction of (a) the smartphone and (b) the telecom provider Perceived ease of use is positively related to customer satisfaction of (a) the smartphone and (b) the telecom provider Perceived effort is negatively related to customer satisfaction of (a) the smartphone and (b) the telecom provider
2.2.4 Usage of the smartphone on usage of telecommunication provider services The aspects of usage of the smartphone are expected to be associated with the usage of the services provided by telecom operators. The top three features that are used on phones are the phone, SMS and internet (Haverila, 2011b). For making calls and for texting messages, smartphone users have to make use of the services provided by telecom operators. However, it is expected that smartphone users make less use of the traditional telecommunication services for texting and calling, which is based on the transition in the telecoms industry from these traditional services to mobile data (Tilson & Lyytinen, 2006). Therefore, users that indicate to use their smartphone more often or longer per day, and/or perceive to have a high level of usage are expected to make more use of mobile data and less use of SMS and voice services. For use of mobile internet services, this relation is less straightforward since heavy
20
users of apps can use a WiFi network, thereby circumventing the costs associated with using mobile data provided by their operator. Nevertheless, consumers that say to be heavy users of smartphones, that use it often and for a long duration on a daily basis, are expected to make more use of mobile data. The following associations are expected:
H6a-c H6d-f H6g-i The daily usage duration of the smartphone is negatively related to (a) SMS usage and (b) voice usage but (c) positively to mobile data usage The daily frequency of use of the smartphone is negatively related to (a) SMS usage and (b) voice usage but (c) positively to mobile data usage The perception of the usage level of the smartphone is positively negatively related to (a) SMS usage and (b) voice usage but (c) positively to mobile data usage
2.2.5 Customer satisfaction of the smartphone on customer satisfaction of the telecom provider The telecommunication provider is a value added reseller of the smartphone because it often provides the handset itself and the services that are necessary to use the smartphone and, such as mobile internet. Again, cross-over effects can be expected between satisfaction that the users of smartphones attribute to their device, and the satisfaction that they attribute to the provider of the device and the services necessary to operate it. For instance, if a customer is in possession of a smartphone that frequently drops calls because of an error in the design of the antenna, it is likely that his or her dissatisfaction with the device is also projected on the provider of the services necessary to make the call. Therefore, it is expected that customer satisfaction with the smartphone and customer satisfaction with the operator are related:
H7 Customer satisfaction of the smartphone is positively related to customer satisfaction of the telecommunication provider
2.2.6 Moderating variables Finally, variables that influence the relation between the number of apps on consumers’ smartphones and the perception of value are discussed. These are subjective knowledge, hedonic attitude or utilitarian attitude towards the smartphone and perceived brand image. However, the directions of the moderating effects are not established in the hypotheses because these are differential, depending on which of the either concepts of feature fatigue or mass customization is supported. Consumers that are knowledgeable about smartphones have been found to be better able to exploit mobile services (Deng et al., 2010). Consumers’ subjective knowledge is defined as “a consumer’s perception of the amount of information they have stored in their memory” (Flynn & Goldsmith, 1999, p. 59), which describes the consumer’s knowledge associated with the product in general. It is plausible to believe that users that consider themselves adequate users of smartphones have a different perception of the usability and utility of their smartphone than novice users. It is also likely that experienced users will benefit more from increases in usefulness and usability, and will suffer less from decreases in perceived usability and increases of sacrifice which are related to the number of apps. The latter can be expected because they take less time and effort to install, learn to use and maintain applications, when compared to average or beginning users of applications. In the TAM, a consumer’s attitude towards a product or technology has been adopted as a moderating variable (Venkatesh & Davis, 2000). Consumers can have different attitudes
21
towards products in which two major categories can be distinguished: a utilitarian and a hedonic attitude (Childers et al., 2001). Consumers with a hedonic attitude find pleasure in using a product, while consumers with a utilitarian attitude use a product to achieve certain outcomes (Batra & Athola, 1990). Based on these different attitudes that consumers can have towards smartphones as a product category, differences in the perception of value are expected because people that experience smartphone as hedonic are more likely to enjoy the range of functionality offered by different applications. On the other hand, consumers with an utilitarian attitude can be expected to be more fatigued by functionality of their smartphone which they do not make use of. It should be noted that consumers could also have both attitudes towards smartphones and that they are not mutually exclusive. The perception of the brand by consumers is expected to affect the extent to which they perceive the aspects of value, e.g. consumers that are biased by a positive brand perception are expected to be positively inclined from functionality of applications and negatively biased from perceived effort increases. Conversely, consumers with a low brand perception could be prone to being irritated from decreases in usability from the high number of apps or perceive more effort required for installing, updating and learning to use apps. The following relations are hypothesized:
H8a-c H9a-d H10a-d H11a-c Consumers’ subjective knowledge moderates the relation between the number of apps and (a) perceived usefulness, (b) perceived usability and (c) perceived effort Consumers’ hedonic attitude towards the smartphone moderates the relation between the number of apps and (a) perceived usefulness (b) perceived usability and (c) perceived effort Consumers’ Utilitarian attitude towards the smartphone moderates the relation between the number of apps (a) perceived usefulness, (b) perceived usability and (c) perceived effort Consumers’ perception of the image of the smartphone brand moderates the relation between the number of apps and (a) perceived usefulness, (b) perceived usability and (c) perceived effort
The model that has been developed in this section should provide an adequate basis for answering the research questions posed in the introduction about the effects of apps on consumer behavior of smartphones and telecommunication provider services. In the following section, the methods used to empirically test the model will be addressed.
3 Method
The model that is developed in the previous section was tested with an online survey. This section will discuss the data sample, the measurement instrument, the data collection procedure and the methods used to analyze the data and to test the model.
3.1 Data sample
Data was collected from customers of a telecommunication provider in the Netherlands. The respondents had to be in possession of a smartphone of the brands Apple, Blackberry, HTC or Samsung. These brands accounted for 90.6% of smartphone sales in March 2012 (GfK, 2012). Only postpaid consumers (i.e. with a monthly subscription) were approached because smartphone apps have a higher penetration rate among postpaid users compared to prepaid users (Telecompaper, 2011). In addition, the respondents were to have a subscription at the operator under focus for at least one month in order to ensure that their smartphone is “out of the box” and in use. Finally, a selection criterion was that the
22
respondents were to have installed at least one app themselves such that they could answer questions about perceived effort of apps.
3.2 Measurement instrument
Several variables in the conceptual model are measured with multiple items. Measuring constructs with multiple items can capture the complexity of a theoretical construct or a latent variable that is not directly measurable (Fornell et al., 1996). To facilitate construct validity and face validity, the constructs were measured with items from previous studies (Flavian et al., 2005; Hair et al., 2010). The variables and constructs in the conceptual model were measured with the measurement instrument in Appendix A. Information on usage of services offered by the telecom provider were extracted from the customer database. The other constructs and variables were measured with an online survey. This survey, which was in Dutch, can be found in Appendix B. The design of the website for the online survey was such that it could easily be completed on desktop or laptop PC’s, smartphones and tablets. The answer options of the questions of the online survey appeared in a random order. The survey and the translation were reviewed by a marketing analyst at the telecom operator and two assistant professors of Marketing at Eindhoven University of Technology. It was also pre-tested by five people. The constructs and variables that are measured by the instrument are the following: Number of apps installed on the smartphone –The number of installed apps was used to measure the number of features that consumer’s have installed on their smartphone. The consumers were asked for the total number of apps on their smartphone, including preinstalled features, in intervals of 10 applications. Above 100 apps, the answer option was open such that the scale could be extended if need be. The pre-installed applications were required because all applications that consumers had installed on their smartphone could provide functionality that pre-installed applications also offer. Perceived usefulness – This measure was adapted from a study assessing a product for business use and was adapted such that the usage situation reflected situations in everyday life. The item “Using this technology improves my job performance” was not adopted, since smartphones were assessed for consumer use and since translating the job performance to performance in everyday life could lead to ambiguity when the other three items are taken into consideration. Perceived ease of use – Is measured by asking respondents for the clearness and understandability of the interaction with the smartphone, the degree of mental effort required for operation, the easiness of use and the degree to which consumers can get their device to do what they want it to do, as in Venkatesh & Davis (2000). Perceived effort and perceived monetary costs – In previous research the sacrifice aspects of value were measured under one construct: perceived sacrifice. In this research however, only the perceived effort or time was measured because the price paid for applications is different for different smartphone operating system, i.e. numerous apps that are freely available in the Android app market are charged for in Apple’s Appstore. Because
23
of possible confusion of effort with the construct perceived ease of use, the effort was measured in perceived time. The measures were adopted from Deng et al. (2010). Usage of the Smartphone – Previous studies have measured usage with the selfreported items daily duration of use, daily frequency of use, the number of applications used and the perceived usage (Al-Gahtani & King, 1999; Kim, 2008). Since the present study focused on usage and satisfaction of smartphones stemming from the number of applications in use, the number of applications were dropped. Because this study expected differential effects for frequency of use, duration of use and perception, these aspects were not used as items but as variables. Previous studies such as Kim (2008) divided the frequency of interaction and hours of use per time period in prefixed time intervals which are probably different for smartphones. Therefore, the daily usage and frequency of interaction were stated as open questions and recoded afterwards. The process of recoding is described in Appendix C. Recoding the frequency and duration of use into ordinal scales should have resulted in more reliable answers since the self-reported measures can be expected to be inaccurate. SMS, Voice and Mobile data usage – Information on SMS, voice and mobile data usage was an objective measure that was extracted from the operator’s customer database and is the average usage per month between December 2011 and February 2012. It should be noted that data usage over WiFi networks cannot be registered by telecom providers. People can use internet on their smartphone solely over their home WiFi network but not have a mobile data subscription, meaning that they can be heavy users of internet on their smartphone without using the mobile data services offered by operators. Customer satisfaction – Almost all studies that use a self-measurement of customer satisfaction show a negatively skewed distribution of this variable (Peterson & Wilson, 1992), i.e. overall, customers are more satisfied than dissatisfied. Turel & Serenko (2006) use a ten-point Likert scale to avoid skewness problems because it enables respondents to make better discriminations (Andrews, 1984 in Turel & Serenko); this approach was adopted in the measurement instrument with an eleven point likert scale in order to allow for a middle answer option, which is coherent with the 7-point Likert scales used for the other items. A single-item was used to measure customer satisfaction directly. Using single-item scales for measuring customer satisfaction may decrease the quality of measurement (LaBarbera & Mazursky, 1983 in: Mittal et al., 1998). Nevertheless, Bergkvist & Rossiter (2007) suggest that single item constructs can be used in marketing for concrete objects and attributes such as customer satisfaction. Using single-item constructs also decrease the completion time of the survey. Subjective knowledge – One item was deleted from Flynn & Goldsmith (1999) since the translation in Dutch language offered too little differentiation from another item measuring the same construct. The measure then consisted of the items knowing much about smartphones, how to operate them, whether the respondent believes that other people think he or she is an expert. The respondent is also asked to compare him/herself to others regarding smartphone knowledge. Utilitarian and hedonic attitude towards smartphones – The two dimensions of attitude towards smartphones are adopted from Voss et al. (2003) who develop measurement scales for the two dimensions of attitude towards product categories. The scales consist of 5
24
items per dimension regarding the hedonic and utilitarian attitude, and are adapted to relate to enjoyment of smartphones and the utility of smartphones in general. Perceived brand image – The perception of the brand’s image by consumers is measured by asking respondents for the attractiveness, reliability and quality of the brand, as in Low & Lamb (2000). Low & Lamb (2000) find support for distinguishing between brand image, brand attitude and perceived quality as three types of brand associations. Control variables – Age, gender, the education level, the number of smartphones owned, the duration of smartphone possession, the smartphone’s brand, type of use (business or consumer) and the device on which the survey was completed were measured as control variables. The motivation for including each of these variables is discussed in the final section of this chapter. It should be noted that all the latent constructs are reflective constructs for which the items are “caused” by the constructs, as opposed to formative constructs which are formed by the items measuring the constructs (Hair et al., 2010, Ch. 12).
3.3 Data collection procedure
The request to complete the online survey was administered via e-mail to 2446 consumer of the Dutch operator. The extraction of the respondents from the customer database of the operator was at random except for the criteria discussed in Section 3.1. Data collection, graphic design of the online survey and the contacting of the consumers was conducted by a marketing research agency. As an incentive to boost the response rate, five vouchers for an online store with a value of €20 were raffled among the respondents. The survey request was issued on April 17th 2012. A reminder was issued after 4 days, and the survey was closed on the 28th of April and hence was available for 15 days. 8 invitations were ‘bounced’ – a non-delivery notification was received due to e.g. a non-existing e-mail address. 18 people accessed the link of the survey, but did not start the survey and 71 potential respondents started the survey without finishing it. A total of 271 surveys were fully completed, a response rate of 11.1%. On average, people took 11:34 minutes to complete the survey. The manipulation of the original dataset is described in Appendix C. Three respondents were deleted from the database because they had a completion time of four minutes or less and because it was suspected that the answering of the questions was conducted at random. Of the remaining 271 respondents, 19 did not meet the criteria of the data sample since four respondents were not customers of the operator and 15 respondents did not download apps. These respondents were deleted from the dataset, which left 252 usable respondents (10.3%).
3.4 Analytical method
The dataset was delivered in Microsoft Excel 2010. SPSS 18 was used for the initial processing of the data. A non-response analysis was conducted using a one-way MANOVA for comparing early and late respondents. To test for common method bias, Harman’s one factor test was applied using principal components analysis in SPSS. Structural equation modeling (SEM) is used to test the model developed in Section 2. SEM consists of two steps: the confirmation of the measurement model and the estimation of
25
the path model or structural model. When the measurement model is analyzed, it is assessed how well the different directly measured items reflect the indirectly measured constructs. Subsequently, in the path analysis, it is assessed whether the predicted relations between the latent constructs exist and to what extent (Hair et al., 2010, Ch. 11). Structural equation models can be estimated by Maximum Likelihood (ML) estimation or by Partial Least Squares (PLS) estimation methods, of which the former is a covariance based approach and well known through software applications such as LISREL or AMOS. The latter is a less-known, variance-based approach (Haenlein & Kaplan, 2004). PLS is used as estimation technique for several reasons. Compared to other multivariate techniques, PLS is robust to violations of distributional assumptions such as normality and smaller sample sizes can be used (Fornell & Bookstein, 1982). In addition, PLS is less sensitive to multicollinearity problems (Hair et al., 2010, Ch. 12), and more information of the data is retained since PLS uses covariance instead of correlation, which is the standardized covariance (Ringle et al., 2005 on smartpls.de). Finally, PLS is more suitable for prediction rather than confirmation of existing theory (Hair et al., 2011). The software package that was used for PLS is SmartPLS 2.0 M3 (Ringle et al., 2005 on Smartpls.de, 2012). An abort criterion of 10-5 was used to assure the PLS algorithm’s convergence while minimizing computational requirements (Hair et al., 2011). A maximum of 300 iterations was permitted. Cutoff values and test approaches for validity and reliability are according to Hair et al. (2011). Bootstrapping with 500 resamples was used as a reactive Monte Carlo resampling strategy in order to assess the significance of the estimates. In bootstrapping, parameters’ values and standard errors are compared to repeated random samples drawn with replacement from the original observed sample data, in order to assess the significance of the estimated parameters (Marcoulides & Saunders, 2006). Effects were deemed significant if the probability of erroneously detecting an effect was smaller than .05 . Moderating effects were analyzed based on the product indicator approach (Henseler & Fassot, 2010). For assessing the moderator effects in SmartPLS, the indicator values were standardized (recomputed to have to a mean of 0 and a standard deviation of 1) before multiplication. Hypotheses regarding moderating effects were significant if the coefficient of the interaction term was significant (Henseler & Fassot, 2010). The strength of the moderating effects was determined by calculating the effect size ƒ2 which is an indicator for the change in R2 due to adding the interaction term of the moderator.1 Effect sizes of respectively .02, .15 and .35 are weak, moderate and strong (Henseler & Fassot, 2010). The moderator analysis was conducted separately per moderating effect, using 500 resamples for the bootstrapping procedure. To ensure correct model estimation, several control variables were included in the model. Preferences for different features and the satisfaction caused by features have been found to differ by demographic variables, such as age, gender and education (Haverila, 2011a, 2011b). In addition, the time of owning a smartphone and the number of smartphones owned are also used because people that have owned multiple smartphones or that have owned a smartphone for a long time may be more experienced with and knowledgeable about
1
ƒ2
(Cohen, 1988 in: Henseler & Fassot, 2010)
26
smartphones. The brand of the device used by the consumer is controlled for as well since the different brands make use of different operating systems, have different apps available to them and have different interfaces, leading to possible differences in e.g. the perception of usability between the users of different smartphone brands. Finally, the type of use of the smartphone (business, pleasure or both) is controlled for, since users that use their smartphone for business purposes do not use it entirely voluntarily, and therefore the evaluations of the device and the apps may differ from the evaluations offered by consumers. Dummy variables were used for categorical variables with more than two categories (i.e. smartphone brand, survey method and type of use). The coding of dummy variables is described in Appendix C. As in prior research using PLS, age, number of smartphones owned, usage time and education were treated as ordinally scaled variables (e.g. Venkatesh & Davis, 2000).
4 Results
This section will present the results of the data collection and the statistical analyses. First, the sample will be described, followed by a non-response analysis and the cleaning of the data. Subsequently, the measurement model will be tested, followed by several tests of the structural model: the base model, moderating effects and control variables. The results induced two post-hoc analyses which are also described: a comparison between the different operating systems and the analysis of mediating variables.
4.1 Sample and data characteristics
Several characteristics of the data sample (N=252) can be viewed in Table 1 (note that this table was reconstructed after the deletion of outliers in section 4.1.3). Corresponding characteristics of other customer groups and the Dutch population can be viewed in Appendix D. The distribution of males and females in the population seems to be relatively even, although the fraction of females is slightly higher than the fraction of males. This is also the case for other customer groups and the Dutch population. The distribution of the different types of education in the sample does not accurately reflect the average Dutch population in the sense that the average Dutch population seems lower educated. Regarding age, the sample does appear to reflect the operators’ customer base, which consists of younger people when compared to the Dutch population. Regarding the representation of the smartphone brands of the sample to the operator’s customer base, Blackberry and Samsung users seem slightly underrepresented as opposed to HTC users, which are overrepresented. Regarding the respondents, an equal proportion of the four brands was present in the 2446 invitees, which is not the case for the actual sample: more HTC and Samsung users responded than Apple and Blackberry users. The majority of the respondents indicated to use their smartphone for consumer purposes only while about one-third uses their smartphone for both business and consumer purposes. Only one respondent indicated to use his or her smartphone solely for business, which makes sense because the survey was issued to consumer users. Finally, the majority of the respondents indicated that their current smartphone was their first, and the majority of the respondents has owned a smartphone for less than two years.
27
Table 1 Descriptive statistics of control variables
Variable Gender Age Label Male Female 15-25 26-35 36-45 46-55 56-65 66-75 75+ None Primary VMBO/MBO1 HAVO/VWO MBO 2-4 HBO WO Other Frequency 123 129 46 46 63 64 26 3 2 1 1 30 29 71 79 34 7 Percentage 48.8 % 51.2 % 18.3 % 18.3 % 25.0 % 25.4 % 10.3 % 1.2 % 0.8 % 0.4 % 0.4 % 11.9 % 29 % 28.2 % 31.3 % 13.5 % 2.8% Variable Smartphone Brand Label Apple Blackberry HTC Samsung Other Consumer Business Both 1 smartphone 2 smartphones 3 smartphones 4 or more smartphones 0-8 months 8-19 months 19-31 months 31-43 months 43-54 months More than 54 months Frequency 47 34 81 86 4 160 1 91 153 44 39 16 52 89 41 27 16 27 Percentage 18.7 % 13.5% 32.1% 34.1 % 1.6 % 63.5 % 0.4 % 36.1 % 60.7% 17.5% 15.5% 6.3% 20.6% 35.3% 16.3% 10.7% 6.3% 10.7%
Type of use
Education
Number of Smartphones owned Usage Time
4.1.1 Non-response analysis Extrapolation is used for assessing non-response. The early and late respondents are compared, since late respondents and non-respondents are both ‘less readily’ in their response (Scott Armstrong & Overton, 1977). The first and last quartile of respondents are compared (n=63). A one-way MANOVA was conducted in SPSS with the response group as the predictor variable and the measured constructs used in the structural model (as calculated by SmartPLS) are used as the dependent variables. In the calculation of these variables, the items that are deleted in section 4.2 are not used. The MANOVA revealed no significant multivariate main effects for response group at the p < .05 level, Wilks’ ? = .868, F(16;105) = 1.000, partial eta squared = .132. Power to detect the effect was .631. Given the low power to detect the effect (under .80), the tests of between subject effects were assessed for further differences between the different response groups on the metric variables, but no significant effects were observed. Based on these findings, it is concluded that there are no significant differences between early and late respondents and that in turn it is not likely that there is a non-response bias. 4.1.2 Common method bias Harman’s one factor test was used to test for common method bias. Using principal component analysis, eight factors were extracted with an eigenvalue exceeding 1.0 (the Kaiser criterion), accounting for 69.0% of the variance. The first and largest factor of the unrotated solution accounted for 30.5% of the variance, indicating that common method bias is probably not an issue since this is not the majority of the variance (Podsakoff et al., 2003). 4.1.3 Missing data, outliers and multivariate assumptions The data did not contain missing values, which is because the online survey tool required the respondents to answer every question. Univariate outliers were detected visually by examining boxplot diagrams and by assessing which variables have a standardized value exceeding ± 4.0 (Hair et al., 2010, Ch. 2). The variables that were measured with a Likert scale or the customer satisfaction variables are not assessed for univariate outliers since univariate outliers are plausible for these variables and were not to be deleted. The outliers
28
can be viewed in Appendix E, as well as the actions following from this analysis. For 11 cases, values were deleted. Multivariate outliers were detected using the Mahalanobis distance, calculated in SPSS. The distance is calculated using all items in the base model (36) in which no values of the D2/df exceeded 2.78. Hence, no multivariate outliers are detected (Hair et al., 2010, Ch. 2). Four multivariate assumptions can be tested on the metric variables which potentially affect every multivariate statistical technique (Hair et al., 2010, Ch. 2): normality, homoscedasticity, linearity and the absence of correlated errors. Regarding the normality assumption, the data is not normally distributed since out of the 40 variables, 31 proved significantly skewed and 20 proved to be significantly platykurtic or leptokurtic (Appendix E). Fortunately, the impact of non-normal distributions of variables in sample sizes exceeding 200 effectively diminishes (Hair et al., 2010, Ch. 2; Tabachnick & Fidell, 2001), while departures from the normal distribution often occur with large sample sizes (Pallant, 2005). Confirmatory factor analysis is also relatively robust to violations of normality under sufficient sample size (Gorsuch, 1983 in: Floyd & Widaman, 1995). However, PLS is sensitive to asymptotic data (Vinzi et al., 2009), which is why a third root transformation was applied to mobile data usage and SMS usage, and a fourth root transformation was applied to voice usage, since these variables has long tails in the distribution. Problems in homoscedasticity are often a result of violation of normality (Hair et al., 2010, Ch. 2). Homoscedasticity and linearity can be assessed by studying scatter plots between two variables. Linearity and homoscedasticity were not assessed because doing so was not pragmatic considering the high number of variables. The absence of correlated errors applies to data collection in groups and time series data (Hair et al., 2010, Ch. 2) due to e.g. a nested data structure. The research design of this study does not encompass these methods, thereby decreasing the probability of correlated errors. Therefore this assumption was not checked.
4.2 Measurement model
4.2.1 Construct reliability Reliability was assessed by examining factor loadings, Cronbach’s alpha and composite reliability. The factor loadings and cross-loadings of the initial measurement model including all directly measured items can be viewed in Appendix F. All individual factor loadings exceeded .70 except for Perceived usability2 (.217) and Utilitarian attitude4 (.575), which were deleted from the model. Cross-loading items were Hedonic atttiude5 with the construct Utilitarian attitude (.719) and Utilitarian Attitude2 with Hedonic attitude (.704). In order to ensure sufficient discriminant validity, these items were deleted from the analysis which can be justified since sufficient items remain to measure hedonic and utilitarian attitude. After these deletions, the model was tested again. The corresponding factor loadings and cross-loadings are illustrated in Appendix G. All items had loadings exceeding .70 on the corresponding constructs and no cross-loadings could be observed. Cronbach’s alpha for the multi-item constructs ranged from .84-92 and composite reliability ranged from .88-.97, as displayed in Table 2. Hence, construct reliability is warranted for the latent constructs. It should be noted that reliability values and the average variance extracted for single item
29
Table 2 Measurement properties
Cronbach's Alpha Composite Reliability R2* .44 .05 .10 .14 .04 .12 .33 .35 .04 .12 .01 R2** .44 .05 .10 .14 .10 .52 .47 .35 .04 .12 .01
Variable
Mean
St. dev.
AVE .74 .91 .72 .86 .76 .80 .84 -
VIF 1.25 2.46 1.61 1.68 2.11 1.32 2.50 1.26 2.54 2.05 1.97 1.61 1.06 1.18 2.13 1.10
Apps 4.08 2.61 Cust. Sat. Smartphone 7.65 2.20 Daily Usage Time 3.77 1.49 Frequency Of Use 2.78 1.38 Hedonic Attitude 5.41 1.06 .88 .92 Mobile Data Usage 5.25 1.89 Perceived Brand Image 5.46 1.34 .95 .97 Perceived Effort 2.93 1.29 .81 .88 Perceived Usability 5.52 1.13 .92 .95 Perceived Usefulness 5.25 1.13 .84 .91 Perceived Usage 5.65 1.24 Subjective Knowledge 4.49 1.32 .92 .94 SMS Usage 3.33 1.30 Telco Cust. Sat. 7.27 1.52 Utilitarian Attitude 5.94 0.80 .91 .94 Voice Usage 2.80 0.70 *R2 of the base model without the direct effect of the moderating variables 2 **R of the model including the direct effects of the moderating variables
constructs equals 1.00 because the items that load on the constructs have a factor loading of 1.00 , which is because the constructs are measured directly. These values are not depicted in Table 2. 4.2.2 Construct validity and multicollinearity Both convergent validity and discriminant validity were assessed. For sufficient convergent validity, values for the average variance extracted (AVE) should exceed .50, which indicates that the latent variable explains more than half of its indicators’ variance (Hair et al., 2011). In Table 2 it can be viewed that the lowest value for the AVE of multiitem constructs was .72 for perceived effort, meaning that convergent validity is sound. Regarding discriminant validity, the Fornell-Larcker criterion was employed and in the previous section, cross-loading items were deleted. When the Fornell & Larcker criterion is met, a latent construct shares more variance with the indicators loading on the construct under scope than with other latent variables in the model (Fornell & Larcker, 1981). In other words, the square root of the AVE should exceed the correlation with any other latent construct. As can be seen in Table 3, this was the case for all multi-item constructs. Therefore, based on the Fornell-Lacker criterion and the absence of cross-loading items, the discriminant validity of the latent constructs in the measurement model is acceptable. Regarding the single-item constructs, no correlations approached 1.00, which is also an indication of the discriminant validity of the single-item constructs. Finally, Multicollinearity was not an issue since no value for the variance inflation factor (VIF) above 5.0 could be detected (Hair et al., 2010), as can be seen in table 2.
4.3 Structural model
For the base model, the significance of the standardized effects sizes are assessed. This is followed by the analysis of moderator variables and control variables.
30
Table 3 Correlation matrix of latent variables and Fornell-Larcker test
Variable 1. 2. 3. 4. 5. 6. 7. 1. Apps 1.00* 2. Cust. Sat. Smartphone .32 1.00 3. Daily Usage Time .06 .04 1.00 4. Frequency Of Use .15 .18 .49 1.00 5. Hedonic Attitude .17 .11 .3 .29 .86 6. Mobile Data Usage .18 .11 .22 .31 .13 1.00 7. Perceived Brand Image .25 .68 .03 .15 .27 .07 .95 8. Perceived Effort -.2 -.32 .06 -.02 -.07 -.08 -.26 9. Perceived Usability .35 .62 .14 .26 .36 .23 .64 10. Perceived Usefulness .27 .49 .19 .3 .37 .25 .58 11. Perceived Usage .23 .34 .41 .42 .39 .3 .4 12. Subjective Knowledge .3 .14 .19 .34 .47 .31 .26 13. SMS Usage -.03 -.02 -.01 .05 0 .08 0 14. Telco Cust. Sat. .03 .31 .02 .01 .14 -.03 .26 15. Utilitarian Attitude .22 .22 .23 .27 .68 .15 .35 16. Voice Usage -.06 .09 -.02 .04 -.04 .12 0 *Diagonal elements in bold are square roots of the AVE values for latent constructs. 8. 9. 10. 11. 12. 13. 14. 15. 16.
.85 -.33 -.15 -.04 -.04 .03 -.16 -.08 -.01
.93 .57 .51 .37 .01 .19 .47 -.01
.87 .52 .41 -.13 .22 .46 -.04
1.00 .4 -.03 .08 .46 .02
.89 .04 0 .37 -.02
1.00 -.02 -.03 .03
1.00 .19 .08
.92 -.05
1.00
4.3.1 Base model The base model includes the main concepts without the direct effects of moderators. The explained variance or the R2 of the dependent variables in the base model can be found in Table 2. The R2 must be evaluated for individual studies (Backhaus et al., 2003 in: Götz et al., 2010). Hair et al. (2010) states that R2 values of respectively .25, .50 and .75 can be described as weak, moderate and substantial as a rule of thumb for marketing research studies. However, the author believes this study could be placed in the category of consumer behavior studies, of which Hair et al. (2011) states that in this specific discipline, R2 values of .20 can be considered as high. A meta-analysis of Arts et al. (2011) on consumer adoption considers that an explained variance of .10 for actual adoption behavior explains variance “to a lesser extent” (Arts et al., 2011). The dependent variables of which the highest amount of variance was explained was customer satisfaction towards the smartphone (.44), perceived usage (.35) and perceived usefulness (.33). Moderately explained variables were mobile data usage (.14), perceived usability (.12), telecommunication company customer satisfaction (.12) and frequency of use (.10). Finally, weakly explained variables were daily usage time (.05), perceived effort (.04), SMS usage (.04) and voice usage (.01). The results of the testing of the hypotheses can be viewed in Table 4. Hypotheses were supported if the predicted direction of the effect was correct and if the effect was significant at the .05 level. Of the 40 hypotheses in the base model, 13 hypotheses were confirmed, 24 effects were non-significant at the .05 level and 2 effect sizes were significant but the effect size was in a direction contradictory to the hypothesis. The following statements can be made about the confirmed hypotheses: Consumers that have more apps are more likely to perceive their smartphone to be more easy to use (H1b,alt, ? = .35, t = 6.41), to perceive less effort (H1c,alt, ? = -.20, t = 2.97) and to be more satisfied about their smartphone (H1d, ? = .09, t = 2.30). In turn, customers who perceive their smartphone as useful are disposed to have a lower degree of SMS usage (H3a, ? = -.22, t = 2.69), to use their smartphone more frequently (H4b, ? = .22, t = 2.55), to perceive to use it more (H4c, ? = .34, t = 3.78) and to be more s atisfied with their smartphone (H5a, ? = .20, t = 2.68). In turn, perceived usability is positively associated with perceived usefulness (H2, ? = .54, t = 10.09), but also with consumers’ perception of their
31
Table 4 Hypothesis testing
Hypothesis Relation ?† T-Value H1a (+) Apps ? Perceived Usefulness .09 1.66* H1b (-) Apps ? Perceived Usability .35 6.41*** H1b,alt (+) Apps ? Perceived Usability .35 6.41*** H1c (+) Apps ? Perceived Effort -.20 2.97*** H1c,alt (-) Apps ? Perceived Effort -.20 2.97*** H1d (+) Apps ? Cust. Sat. Smartphone .09 2.30** H1e (+) Apps ? Telco Cust. Sat. -.10 1.38 H2 (+) Perceived Usability ? Perceived Usefulness .54 10.09*** H3a (-) Perceived Usefulness ? SMS Usage -.22 2.69*** H3b (-) Perceived Usefulness ? Voice Usage -.06 0.87 H3c (+) Perceived Usefulness ? Mobile Data Usage .08 1.04 H3d (-) Perceived Usability ? SMS Usage .12 1.57 H3e (-) Perceived Usability ? Voice Usage -.01 0.21 H3f (+) Perceived Usability ? Mobile Data Usage .04 0.48 H3g (+) Perceived Effort ? Mobile Data Usage -.05 0.85 H4a (+) Perceived Usefulness ? Daily Usage Time .15 1.72* H4b (+) Perceived Usefulness ? Frequency Of Use .22 2.55** H4c (+) Perceived Usefulness ? Perceived Usage Level .34 3.78*** H4d (+) Perceived Usability ? Daily Usage Time .10 1.14 H4e (+) Perceived Usability ? Frequency Of Use .15 1.83* H4f (+) Perceived Usability ? Perceived Usage Level .36 4.08*** H4g (+) Perceived Effort ? Daily Usage Time .12 1.26 H4h (+) Perceived Effort ? Frequency Of Use .06 0.86 H4i (+) Perceived Effort ? Perceived Usage Level .13 2.76*** H5a (+) Perceived Usefulness ? Cust. Sat. Smartphone .20 2.68*** H5b (+) Perceived Usefulness ? Telco Cust. Sat. .12 1.65* H5c (+) Perceived Usability ? Cust. Sat. Smartphone .44 5.86*** H5d (+) Perceived Usability ? Telco Cust. Sat. -.06 0.65 H5e (-) Perceived Effort ? Cust. Sat. Smartphone -.12 2.14** H5f (+) Perceived Effort ? Telco Cust. Sat. -.09 1.29 H6a (-) Daily Usage Time ? SMS Usage -.03 0.42 H6b (-) Daily Usage Time ? Voice Usage -.06 0.78 H6c (+) Daily Usage Time ?Mobile Data Usage .06 0.74 H6d (-) Frequency Of Use ? SMS Usage .11 1.49 H6e (-) Frequency Of Use ? Voice Usage .07 1.10 H6f (+) Frequency Of Use ? Mobile Data Usage .19 2.40** H6g (-) Perceived Usage Level? SMS Usage 0 0.06 H6h (-) Perceived Usage Level? Voice Usage .05 0.72 H6i (+) Perceived Usage Level? Mobile Data Usage .14 1.62 H7 (+) Cust. Sat. Smartphone ? Telco Cust. Sat. .29 3.66*** *
<.1. **<.05, ***<.01 (Two Tailed); † Effect Sizes Are Standardized Supported? No No Yes No Yes Yes No Yes Yes No No No No No No No Yes Yes No No Yes No No Yes Yes No Yes No No No No No No No No Yes No No No Yesusage level (H4f, ? = .36, t = 4.08) and customer satisfaction of the smartphone (H5c, ? = .44, t = 5.86). As the final aspect of value in the model, perceived effort is positively associated with consumers’ perception of the usage level (H4i, ? = .13, t = 2.76), but negatively with customer satisfaction towards the smartphone (H5e, ? = -.12, t = 2.14). Finally, customers that use their smartphone more often on a daily basis, tend to use more mobile data (H6f, ? = .19, t = 2.40), and customer that are satisfied with their smartphone are more often satisfied with their telecommunication provider (H7, ? = .29, t = 3.66). As in Turel & Serenko (2006), the statistical validity of the significant linkages was assessed by deleting the insignificant effects from the model. Comparison of ?’s, t-values and R2 values can be viewed in Appendix H. The deletion of the insignificant relations in the base model did not result in major changes of the R2’s, the largest change being a small to medium decrease in the R2 of mobile data usage from .14 to .09 (with a small effect size ƒ2 of -.05; see
32
the next section for the calculation of the effect size). The largest change in path coefficients was for frequency of use ? mobile data usage, of ? = .19 to ? = .31 . No relations changed from being significant to being insignificant on the .05 level. Based on these results, the statistical validity of the significant relations was further confirmed. 4.3.2 Moderating variables The results of the moderator analysis can be found in Table 5. Seven direct effects of moderators have been detected: of subjective knowledge about smartphones on the perceived usability (? = .11, t = 1.97) and on perceived usefulness (? = .18, t = 2.78); of consumers’ utilitarian attitude towards smartphones on perceived usability (? = .23, t = 3.41) and on the perceived usefulness of the smartphone (? = .18, t = 2.50). Finally, perceived brand image was directly associated with all three aspects of value: negatively with perceived effort (? = .25, t = 3.03), but positively to perceived usability (? = .50, t = 9.87) and perceived usefulness (? = .34, t = 4.03). Consumers’ hedonic attitude towards smartphones did not have any direct effects. Only one moderating effect was found to be significant: the effect of subjective knowledge on the relation between the number of apps and the perception of usability (H8b, ? = -.12, t = 2.00), which weakly increases the explained variance of perceived ease of use (ƒ2 = .02). The negative value of the coefficient of the interaction term indicates that consumers’ subjective knowledge diminishes the positive effect of the number of apps on the perceived ease of use, i.e. consumers who believe they are more knowledgeable about smartphones believe their smartphone is less easy to use when they have more applications, than consumers who believe they are less knowledgeable about smartphones. The other hypotheses on moderators could not be confirmed.
Table 5 Moderator analysis results
Direct effect Relation Including/excluding moderating effect R
2
ƒ
2
Moderator direct effect ?† .08 .08 .11 .22 .18 .18 -.04 -.25 0 -.01 0 0 .04 .02 .23 .22 .18 .17 -.25 -.24 .5 .49 .34 .34 T-value 1.14 1.09 1.97** 3.28*** 2.78*** 2.85*** 0.49 3*** 0.07 0.2 0.02 0.02 0.39 0.17 3.41*** 3.07*** 2.50** 2.17** 3.03** 2.93*** 9.87*** 8.66*** 4.03*** 4.2***
Interaction term effect ?† .06 -.12 -.02 .01 -.08 .01 -.07 -.04 -.02 .02 -.02 -.03 T-value 0.58 2.00** 0.25 0.09 1.38 0.10 0.69 0.55 0.20 0.20 0.38 0.45
?† Excluding Including Excluding Subjective Knowledge ? (Apps ? Perceived Usability) Including Excluding Subjective Knowledge ? (Apps ? Perceived Usefulness) Including Excluding Hedonic Attitude ? (Apps ? Perceived Effort) Including Excluding Hedonic Attitude ? (Apps ? Perceived Usability) Including Excluding Hedonic Attitude ? (Apps ? Perceived Usefulness) Including Excluding Utilitarian Attitude ? (Apps ? Perceived Effort) Including Excluding Utilitarian Attitude ? (Apps ? Perceived Usability) Including Excluding Utilitarian Attitude ? (Apps ? Perceived Usefulness) Including Excluding Perceived Brand Image ? (Apps ? Perceived Effort) Including Excluding Perceived Brand Image ? (Apps ? Perceived Usability) Including Excluding Perceived Brand Image ? (Apps ? Perceived Usefulness) Including *
<.1. **<.05, ***<.01 (two tailed); † effect sizes are standardized Subjective Knowledge ? (Apps ? Perceived Effort) .10 .10 .52 .53 .47 .47 .10 .10 .52 .52 .47 .47 .10 .10 .52 .52 .47 .47 .10 .10 .52 .52 .47 .47 .00 .02 .00 .00 .00 .00 .00 .00 .00 .00 .00 .00 -.17 -.19 .14 .17 .03 .03 -.17 -.17 .14 .16 .03 .03 -.17 -.15 .14 .15 .03 .03 -.17 -.17 .14 .14 .03 .03T-value 2.47*** 2.82*** 3.13*** 3.41*** 0.57 0.58 2.47** 2.64*** 3.13*** 3.27*** 0.57 0.49 2.47** 2.31** 3.13*** 2.88*** 0.57 0.56 2.47** 2.43** 3.13*** 3.11*** 0.57 0.56
33
4.3.3 Control variables The effects of control variables were tested by entering all control variables in the path model simultaneously. The bootstrap procedure was conducted using 500 resamples. The effects that were found to be significant at the .05 level can be found in Table 6. From the path model including the control variables, it appears that older people tend to be more satisfied with their smartphone (? = .10, t = 1.96), to use it for a shorter period of time (? = -.29, t = 5.27) and less frequent (? = -.5, t = 4.09) on a daily basis, to have a lower usage of mobile data (? = -.38, t = 7.01) and of text messages (? = -.14, t = 2.05), perceive to use their smartphone less (? = -.18, t = 3.71), say to know less about smartphones (? = -.15, t = 2.37) and perceive their smartphone as less useful (? = -.15, t = 2.88). As opposed to male smartphone users, female smartphone users more often indicate to have less apps installed on their device ( = -.16, t = 2.53) and to know less about smartphones (? = -.23, t= 3.61). Higher educated people often use their smartphone for a shorter period of time on a daily basis (? = -.13, t = 1.99). Regarding the smartphone brands, Blackberry users seem to differ from Apple users since they tend to have less apps on their smartphone (? = -.23, t = 4.53), to use less mobile data (? = -.12, t = 2.17), to have a lower perception of the image of their smartphone’s brand (? = -.28, t = 2.40), to spend more effort on installing, maintaining and learning to use the apps on their Blackberry (? = . 19, t = 2.34) but appeared to use it longer on a daily basis (? = .16, t = 2.26) and to perceive to use it more (? = .14, t = 2.07). The only other significant difference between users of different smartphone brands was that Samsung users often have more apps on their device than Apple users (? = .13, t = 2.02). Respondents that completed the survey on a smartphone are more likely to spend more time on their smartphone (? = .15, t = 2.25) and to use it more often (? = .13, t = 2.05) per day, to perceive to use their smartphone more (? = .14, t = 2.78) and to perceive it as easy to use (? = .10, t = 2.45). On the other hand, respondents having completed the survey on a tablet are more likely to be satisfied with their smartphone (? = .08, t = 2.04). Another finding is that consumers that use their smartphone for both business and consumer purposes tend to use more mobile data (? = .15, t = 2.44), say more often that their smartphone is useful (? = .09, t = 2.12) and state more often to know more about smartphones (? = .17, t = 3.29) than consumers using their smartphone only for consumer use. Moreover, business and consumer users tend to have a higher SMS usage (? = .17, t = 2.79). No significant differences were detected for the number of smartphones that people have owned or the time that respondents have owned a smartphone.
4.4 Post-hoc analyses
Based on the findings of the initial results, two post-hoc analyses were conducted. The motivation and findings are discussed of an analysis of mediating variables and a comparison between the different operating systems that consumers use on their device. 4.4.1 Mediating variables Only one weak moderating effect of subjective knowledge was found in the analysis of moderating variables. Neither the hedonic nor utilitarian attitude of consumers towards smartphones, nor the perception of the image of the smartphone brand appeared to influence
34
Table 6 Significant effects of control variables
Relation Of Control Variable ?† T-Value Relation Of Control Variable ?† Age ? Cust. Sat. Smartphone .1 1.96** Samsung ? Apps .13 Age ? Daily Duration Of Use -.29 5.27*** Education ? Daily Duration Of Usage -.13 Age ? Frequency Of Usage -.25 4.09*** Female ? Apps -.16 Age ? Mobile Data Usage -.38 7.01*** Female ? Subjective Knowledge -.23 Age ? Perceived Usefulness -.15 2.88*** Survey On Smartphone ? Daily Duration Of Usage .15 Age ? Perceived Usage Level -.18 3.71*** Survey On Smartphone ? Frequency Of Usage .13 Age ? Subjective Knowledge -.15 2.37** Survey On Smartphone ? Perceived Usability .09 Age ? SMS Usage -.14 2.05** Survey On Smartphone ? Perceived Usage Level .14 Blackberry ? Apps -.23 4.53*** Survey On Tablet ? Cust. Sat. Smartphone .08 Blackberry ? Daily Duration Of Use .16 2.26** Business And Pleasure Use ? Mobile Data Usage .15 Blackberry ? Mobile Data Usage -.12 2.17** Business And Pleasure Use ? Perceived Usefulness .09 Blackberry ? Perceived Brand Image -.28 2.4** Business And Pleasure Use ? Subjective Knowledge .17 Blackberry ? Perceived Effort .19 2.34** Business And Pleasure Use ? SMS Usage .17 Blackberry ? Perceived Usage Level .14 2.07** **
<.05, ***<.01 (two tailed); † effect sizes are standardized Reference categories of dummy variables: ‘apple’ for smartphone brand, ‘male’ for gender, ‘pc/laptop’ for survey method and ‘pleasure use only’ for type of use. T-Value 2.02** 1.99** 2.53*** 3.61*** 2.25** 2.05** 2.29** 2.78*** 2.04** 2.44** 2.12** 3.09*** 2.79***relations between the number of apps installed on the smartphone and the aspects of value. However, direct effects were detected of perceived brand image, utilitarian attitude and subjective knowledge. These direct effects give rise to the possibility that these constructs could play a mediating role in the relation between the number of apps and the aspects of perceived value, i.e. when adding these constructs in the model, they could very well reduce the strength of the effects of installed apps on the aspects of value. Additionally, an analysis of the possible mediating role of these variables could lead to the detection of spurious relations between the number of apps and the aspects of value. In other words, the variables that were initially proposed to moderate the relation between the number of apps and the aspects of value, could also be confounding variables to the relations between the number of apps and the aspects of value. The analysis of mediating effects was conducted according to Baron & Kenny (1986): 1) the independent variable was to be a significant predictor of the dependent variable; 2) the independent variable was to be a significant predictor of the mediator variable; and 3) the mediating variable was to be a significant predictor of the dependent variable, while controlling the mediator with the independent variable. In the last step, the effect of the independent variable was to be smaller in absolute value than in the first step, for partial mediation to exist. The significance of the mediating effect can be assessed with Sobel’s test2 using the unstandardized beta’s.3 In the base model, the number of apps on the smartphone was significantly related to the perceived usability of the smartphone and the perceived effort required for installing, updating and learning to use those apps. Therefore, these two relations qualify for being mediated. The model that was used for testing the mediating effects was the model in which the insignificant effects were deleted (‘the bare model’), described at the end of section 4.3.1.
2
where a and ss refer to the unstandardized coefficient and standard error of the relation
between the independent variable and the mediator, and b refers to those parameters of the relation between the mediator and the dependent variable. 3 Calculated by multiplying the standardized coefficient by the standard deviation of the predicted variable and dividing it by the standard deviation of the predictor variable (Bontis et al., 2007).
35
The R2 value for perceived brand image was .06, for hedonic attitude towards smartphones the value was .03, .05 for utilitarian attitude and .09 for subjective knowledge. As can be viewed in Appendix I, it appears that the number of apps is significantly related to all four postulated mediating variables: hedonic attitude (? = .17, t = 3.01), utilitarian attitude (? = .22, t = 4.01), perceived brand image (? = .25, t = 3.92) and subjective knowledge (? = .30, t = 5.09). However, only three of the mediating variables had significant effects on the predicted variables: perceived brand image on perceived effort (? = -.25, t = 3.22) and on perceived usability (? = .50, t = 10.08) and utilitarian attitude on perceived usability (? = .23, t = 3.43). It also appeared that the coefficient of the relation between the number of apps and perceived effort decreased slightly from ? = -.20 in the bare model to ? = -.17 in the mediator model. Moreover, the coefficient of the relation between the number of apps and the perceived usability of the smartphone decreased from ? = .35 to ? = .14. Hence, the relations between the number of apps, and perceived effort & usability are partially mediated. The assessment of the significance of the partially mediating relations still remained. One significant mediating relation was detected with a Sobel value corresponding to p < .05: the relation between the number of apps and the perceived usability was partially mediated by consumers’ utilitarian attitude towards the smartphone (Sobel z-value = 1.96). See Appendix I for the calculation of these values. 4.4.2 Differences between operating systems Another analysis that was conducted post-hoc was the assessment of differences between the main variables in the model for the different operating systems that the smartphones of different brands run on. Apple’s smartphones run on iOS, HTC & Samsung devices operate with Android and Blackberry handsets use their own operating system, from herewith referred to as BOS. The motivation of this analysis is as follows: in the assessment of the control variables, differences were detected with Apple as the reference category. However, no direct differences between Blackberry, HTC and Samsung were analyzed. To assess these differences with SmartPLS with dummy variables would be time consuming. Therefore the differences are analyzed using a one-way MANOVA in SPSS. Since HTC and Samsung use the same operating system, these two brands have same user interface and the same apps available, which is why no differences between brands are analyzed but differences between operating systems. Moreover, the users the operating systems of the smartphone are easier to target by mass-media advertising campaigns from telecommunication operators than other segments based on the criteria such as age, gender or education. Therefore, a MANOVA is conducted to assess differences between the three operating systems and the main variables in the structural model, i.e. the number of apps, customer satisfaction with the smartphone and the telecom provider, and the 6 usage variables. Before describing the results of the analysis it should be noted that the group sizes are unequal: 47 for iOS, 34 for BOS, and 167 for Android. Also, the assumption of normality is violated for customer satisfaction with the smartphone and the telecom operator, the number of apps, daily usage time, frequency of use and the perceived usage level. However, mainly a platykurtic distribution leads to reduced power (Hair et al., 2010, Ch. 8), which is not the case
36
Table 7 MANOVA results - Differences per operating system
Variable Number Of Apps Cust. Sat. Smartphone Daily Duration Of Use Frequency Of Daily Use Perceived Usage Level Mobile Data Usage SMS Usage Telecom Operator Cust. Sat Voice Usage Degrees of freedom 2 2 2 2 2 2 2 2 2 F-value 15.32 21.67 15.59 11.19 3.26 2.2 0.05 1.22 1.5 Partial eta squared .114 .155 .063 .01 .027 .018 .000 .010 .013 Significance .000 .000 .000 .306 .040 .113 .947 .300 .224 Power .999 1.000 .954 .259 .616 .446 .058 .265 .319
for the variables that are assessed (see Appendix E). Moreover, the results of the analysis of the control variables can be used to support findings of the MANOVA since PLS is robust to violations of normality (Hair et al., 2010, Ch. 8). Additionally, Pillai’s trace was used as test criterion because this statistic is robust to violations of normality and unequal group sizes. Since Box’s M test proved significant at the .01 level, indicating unequal variances across groups, Games-Howell test were used for the post-hoc analyses since this test offers the best performance under unequal variances across groups and unequal sample sizes (Field, 2009). The one-way MANOVA revealed a significant multivariate main effect for operating system. Pillai’s Trace had a value of .375, F(18,460) = 5.90, p< .01, partial eta squared = .188. The observed power to detect the difference was 1.00. For assessing which variables significantly differed for the groups, a Bonferroni-adjustment was made by dividing the .05 significance by 9 since there were 9 independent variables. The significant univariate main effects are illustrated in Table 7. It appeared that the number of apps, satisfaction with the smartphone and the daily usage duration of the smartphone differed per operating system. From post-hoc analyses on these three variables (see Appendix J), several significant differences between the operating systems were observed for these variables. Regarding the number of apps, Blackberry users (mean in the category 11-20 apps) had less apps than Apple or Android users (means in the category 41-50 apps). Customer satisfaction with the device differed significantly between Apple users on one hand, who had a mean score for satisfaction of 9.05, and Android and Blackberry users on the other, which had mean scores for customer satisfaction of 7.62 and 6.06, respectively. The post hoc analysis did not reveal significant differences between the operating systems for daily duration of use. Due to the differences in group sizes, it should be noted that these findings should be interpreted with caution. Differences between brands that were detected in the control variable analysis, but not in the MANOVA, were mainly between Apple and Blackberry users for the variables daily duration of use, mobile data usage and the perceived usage level.
5 Discussion
This chapter will discuss the research questions posed in the introduction. The discussion of the findings is accompanied by a reflection on the results and is followed by implications for scholars and managers. Finally, the limitations of this study are discussed, which offer directions for further research. The results are summarized in Figure 3.
37
Subjective knowledge
Utilitarian attiude (5%)
-.22***
.22***
.23***
SMS usage (4%) Perceived usefulness (33%)
.22*** -.12**
Number of apps .54***
.34*** .35***
Perceived usability (12%)
Daily frequency of use (10%)
.19**
Mobile data usage (14%)
.36*** -.20***
Perceived effort (4%)
Perceived usage (35%)
.13***
.44*** -.12** .09**
.20***
Customer satisfaction smartphone (44%)
.29***
Telecom provider satisfaction (11%)
Figure 3 PLS results. Between brackets within the variables is explained variance. ** significant at the .05 level, ***significant at the .01 level (two-tailed). Note that insignificant effects and effects that were significant at the .10 level were omitted for the sake of overview. The narrow-dashed line represents a moderating effect, the broad-dashed line represents a mediating effect
5.1 Findings
This writing was initiated in order to investigate the effects that the number of apps installed on a smartphone can have on smartphone consumer behavior. Specifically, the effects of apps on usage and satisfaction were assessed with regard to the smartphone, the telecommunication operator and the services provided by the operating company. It was also examined which of two competing concepts is most applicable to smartphone consumption: feature fatigue or mass customization. According to the results of this study, not feature fatigue but mass customization is the appropriate concept to explain the effect of the number of apps on the aspects of value perceived by consumers. Apps were not found to relate to utility, but did appear to be positively related to usability and negatively to the perceived effort required to install, update and learn how to use the apps. Since feature fatigue predicts that more features lead to decreased usability but increased utility, the phenomenon is not applicable to smartphones. The concept of mass customization is applicable though, with the underlying rationale that consumers that alter the functionality of their device and have a high number of apps, often have control over their device and perceive it as easy to use. Another explanation could be that users with a high number of apps, also use more apps that improve the usability of the device. It should be noted that the positive association between the number of apps and usability probably has its limits, since a very high number of apps is likely to decrease the processing speed of the smartphone, thereby retarding the device. Mass customization can also be applied to explain the negative association between the number of apps and the effort that the apps on the smartphone require. Although the degree of explained variance of effort
38
was low, consumers that have a high number of apps seem to customize their device and perceive a lower degree of effort than consumers that have less apps installed. This could be because users can customize their device according to their own preferences. Doing so, they adapt the number of apps to match their demand of effort they want to spend on installing, updating and learning to use those apps and they perceive the degree of effort accordingly. The relation could also be negative because consumers with more apps are more effective at managing the apps on their device, and hence perceive a lower degree of effort. The absence of a relation between the number of apps and perceived utility was unexpected. This can probably lead to the conclusion that not each additional app necessarily adds to the utility of the device, giving rise to the idea that distinction should be made between different types of apps. This is further discussed in the limitations section. One relation among the aspects of value is more or less traditional in technology acceptance literature (e.g. Venkatesh & Davis, 2003): usability positively relates to utility. Since usability is related to apps, utility is indirectly associated with the number of apps on consumers’ smartphones. Through value, apps affect the usage of the smartphone and services offered by telecom operators. All aspects of value are positively related to the perception of the usage level, which was the only aspect of usage of which the hypotheses were all correct and therefore the effects are not surprising. However, none of the aspects of value was related to the time that consumers spend daily on their smartphone, and only the utility of the smartphone was associated to the frequency of daily use. It may be stated that increased usability, usefulness and effort thus relate to the extent to which people say they use their device, but that only the actual utility is related to how often people say to use their device on a daily basis. This could indicate that apps that are actually perceived as useful, are used more often, but not necessarily more in terms of daily duration. The perceived utility of the device also plays a key role, and the only role, in predicting increases or decreases in usage of services provided by operating companies. Increases in utility explain a small degree of SMS usage in the sense that people who believe their smartphone is useful, often have a lower SMS usage. Moreover, the frequency of daily use which is predicted by utility, is the only form of usage that is related to usage of mobile data. The number of calling minutes is not at all affected in the model, and the explained variance was low. Apparently, consumers that believe their device is useful tend to use their device more frequently, which leads to the use of mobile data. The decrease in SMS usage and the increase in data usage, which both (in)directly relate to utility of the smartphone, fits the transition in the telecommunications sector from SMS and voice usage to mobile data (Tilson & Lyytinen, 2006). The absence of relations between usage of the smartphone and voice usage could be explained by the measurement: the variables related to the use of the smartphone that are measured in the survey reflect use without calling behavior, which could be a reason why no relations were found between usage of the device and voice usage. It can be concluded that only the utility of the smartphone plays a central role in actual usage of the smartphone. However, the utility is not related to the number of apps directly, but only through the perceived usability. Effort as a negative aspect of value does not play an important role in the model when it comes to predicting usage. People that perceive to spend a high degree of effort on installing, maintaining and learning to use those
39
apps do also perceive to use their phone more often, but the effort is not associated with selfreported usage or actual usage of telecom services. The finding that effort only affects the perceived usage indicates that the relation between effort and usage is mostly in the mind of the consumers. The discussion of the effects of apps on customer satisfaction remain. The number of apps is directly and positively related to customer satisfaction with the device. Indirect effects are also found: usability and usefulness are positively related to satisfaction, but effort is negatively related. The relation between usability and satisfaction with the device is in line with findings of Choi & Lee (2012). Based on these findings, the conclusion is that smartphone users that customize their smartphone by adding apps, are often more satisfied with their device for a number of reasons. Consumers welcome the utility and usability, but despise the required effort for having apps. Fortunately, this effort perception is often lower when consumers have a high number of apps installed on their handset. Satisfaction with the smartphone is related to numerous variables in the model, but satisfaction with the operating company that allows consumers to use their device anytime, anywhere is not affected. Satisfied smartphone owners are generally more satisfied with their operator, but this satisfaction is not directly affected by the number of apps or the value of the smartphone. Apparently, consumers do not directly relate the value of their device to their satisfaction with the provider of their handset and of the services necessary to operate it. A moderating effect was detected for subjective knowledge about smartphones on the relation between the number of apps and usability: increases in the number of apps installed on the smartphone are associated with increases in the perceived usability of the device, but this is less so for consumers that consider themselves knowledgeable about smartphones. The direction of this moderating effect is contrary to expectations. This could be because consumers that are knowledgeable smartphone users are better able to evaluate the usability of products (Burson, 2007) and can provide a more accurate evaluation of the usability. It was expected that this bias was to be in favor of usability. However, it seems that expert users that can make better judgments on usability believe that their smartphone is not so easy to use when compared to novice users, if they have a high number of apps. Moreover, a partially mediating effect was found for the utilitarian attitude towards smartphones, i.e. the relation between the number of apps and the perception of usability is partly accounted for by consumers that believe that smartphones in general are useful devices, since these consumers both often have more apps and often believe their smartphone is easy to use. This is most likely because people who believe that the smartphone is a useful piece of technology, try to make optimal use of its utility by having a large number of apps installed, and that these consumers are biased to believe that the device is easy to use. The effects of control variables are conform other studies regarding age (Haverila, 2011a) and gender (Haverila, 2011b). The difference of perceptions between Blackberry users and other smartphone users could be because there are less apps for the Blackberry and because apps for this device often appear later in the market, or because the interface of the Blackberry differs greatly from iOS or Android, which are much more alike. This could also explain the differences in satisfaction that exists for consumers using smartphones with different operating systems that were detected with the post-hoc MANOVA, which also supports the findings that Blackberry users have less apps on their device.
40
5.2 Scholarly implications
The findings have several implications for scholars of consumer behavior or marketing. First of all, feature fatigue is not applicable to modular products such as smartphones. The shift in preferences of utility over usability before purchase, to usability over utility after use is no issue for smartphones. Probably this also applies to similar products using apps such as tablets or smart televisions. If consumers think their device is not usable because of the high number of apps, these are easily deleted and consumers no longer have to be frustrated with an unusable product with static functionality and usability. However, this does not necessarily mean that users of devices that use apps cannot be ‘fatigued’ by a high number of apps, only that the number of apps is not negatively related to decreases in usability. For instance, consumers can be overwhelmed by product choice (Kamis et al., 2008), which could harm their usage experience. Scholars wanting to study feature fatigue for smartphones or other devices that use apps, should bear in mind that consumers could still be irritated by apps they don’t use. As opposed to feature fatigue, mass customization is a concept that can be adopted when studying effects of apps on smartphones: scholars are encouraged to predict that a high number of apps is positively related to usability and less to effort, based on the concept of mass customization. The same predictions could be made for other products using apps, based on the findings of this study. Secondly, the effects of the number of apps on usability and effort proved to be relatively stable. Subjective knowledge weakly moderated the relation between the number of apps and usability, and the same relation was partially mediated by consumers’ utilitarian attitude. However, consumers’ hedonic attitude towards smartphones and the image of the brand in consumers’ minds did not moderate or mediate the relations between the number of apps and value. For scholars this implies that they should look for other possible spurious relationships or that the relation between the number of apps and usability is stable except for subjective knowledge and consumers’ utilitarian attitude. Thirdly, the conceptual framework that was used proved useful for modeling the effects of apps on consumer behavior for post-purchase processes. The integration of the Technology Acceptance Model in a model of customer satisfaction with value as the standard for comparison yielded moderate to high degrees of explained variance for satisfaction levels, perceived usage, frequency of use and mobile data usage. Finally, combining traditional survey data with objective usage data of telecommunication services proved useful to study smartphone usage, which is in accordance with scholarly implications of Verkasalo et al. (2010).
5.3 Managerial implications
Several implications can be extracted from this research for managers. First of all, managers need not worry about usability issues for devices that use apps, due to a high number of apps. Apps are a relatively new but successful way of selling and consuming software on smartphones, but also on tablet PC’s and more recently developed televisions. The soon to be released Windows 8 will have an interface that looks much like the one currently used on several smartphones and tablets. This study did not detect negative consequences for usage or satisfaction of consumers that have a high number of apps for issues such as usability. Therefore, managers should not worry that consumers that have
41
dozens of apps suffer from e.g. usability issues when it comes to the perception of value of the smartphone and indirect usage and satisfaction. In fact, it is the other way around: consumers that have a high number of apps tend to perceive their device to be easy to use. Consumers can do almost everything with apps. For example, by placing their smartphone on their mattress, they can use it to track either their sleep cycle, but also their sex life, or even both, with the Sleep Cycle Alarm Clock and the SexTrack app. The same device provides you with the best route through the supermarket when doing groceries with an app of a Dutch supermarket. This example illustrates that consumers with more apps, probably have more applications for using their smartphone, and often believe that their omnipotent device is easy to use despite the diverse range of applications. This effective form of mass customization is an important opportunity for app developers, who should come up with even more creative applications for smartphones. Mass customization is also an opportunity for managers in the telecom industry, who could ask consumers what they want to do with their smartphone, and then could install those prior to them receiving the device as an extension of their service. However, users should be able to delete these apps if they desire to do so, which is in line with mass customization. Secondly, in the context of the current transition in telecommunications from voice and SMS to data (Tilson & Lyytinen, 2006), carriers should encourage the use of apps in order to stimulate the use of mobile data. This is especially so for apps that promote more frequent interaction with the smartphone, rather than longer interaction, since mobile data usage is associated with frequency of use. Examples of such apps are Whatsapp, Facebook or news apps, since these are probably frequently checked by smartphone users when compared to game apps which probably demand longer interaction instead of more frequent interaction. Moreover, marketers and managers should allocate effort and resources to promote the rich collection of apps to consumers because of the positive associations of apps with the value of the device and satisfaction. This is contrary to recent findings of Tan et al. (2012) who state that manufacturers need not do so but should focus more on brand value. Thirdly, managers should bear in mind the antecedents to satisfaction. Since satisfaction affects customer loyalty and word-of-mouth (Szymanski & Henard, 2001) and ultimately, profitability (Hallowell, 1996), managers should be aware that apps, usability and usefulness positively influence satisfaction with the handset. However, the perceived effort by consumers that is required to install, update and learn how to use the apps inhibits satisfaction. Managers should consider to improve usability and usefulness perceptions and to automate the updating of apps in order to facilitate satisfaction with the device. This is a confirmation for Apple that their efforts to monitor the usability of the apps that are distributed via their Appstore are justified and lead to increased satisfaction with the iPhone. On the other hand, predictors to satisfaction with the handset are no predictors for satisfaction with the carrier, who often provides the handset in question. Consumers do not relate the number of apps or the value of the device to their satisfaction with the operating company. However, customers that are satisfied with their device do tend to be more satisfied with their carrier, which is why managers seeking to improve satisfaction with the telecom provider should consider converting existing customers that use Android or Blackberry devices to use Apple’s iPhone, or aim at providing new customers with iPhones. They could also encourage manufacturers of smartphones operating on Android to improve the usability
42
of their devices. Not only do Apple users tend to be more satisfied, they also join Android users in the number of apps they install on their device, when compared to Blackberry users. Marketers should adapt their marketing efforts according to the satisfaction levels experienced by customers of these handsets. Finally, managers should design online self-services that aid expert users in usability issues. This implication stems from the finding that expert users benefit less from increases in usability that are often associated with increases in the number of apps. Several telecom operators in the Netherlands offer start-up programs for new smartphone users in-store, but it is likely that users that consider themselves smartphone experts will not use these services because they already know how to use the device. Therefore, operators can design or improve online self-service programs that solve usability issues, especially for expert users who do not use these services in-store.
5.4 Limitations and avenues for further research
Several limitations impair the findings of this study and can be used to indicate directions for further research. For one, several aspects of the conceptual framework limit the research. A major limitation is the assumption that each app is equal to a feature and the fact that this research has not accounted for different categories of apps. The assumption is open for discussion when it comes to different categories of apps such as game apps. The discussion is whether a category of apps offer functionality, or whether each individual app can be seen as a different feature, e.g. if a user has 50 games installed on his or her smartphone, is each game experienced as a feature or is gaming as a category experienced as a feature of the smartphone? Moreover, previous research has found that different types of features for mobile phones are differently related to customer satisfaction (Haverila, 2011b). In addition, Verkasalo et al. (2009) find that mobile phone services such as apps should be addresses individually, instead of generalizing those services. Future research should differentiate between categories of features or apps. It could be that satisfaction and usage of apps and smartphones can be better explained per individual app and that each app purchase is evaluated separately by consumers. This view allows the use of the expectation/ disconfirmation paradigm of customer satisfaction in which customers expect different levels of value from different applications. Those expectations are then (dis)confirmed, leading to (dis)satisfaction. The satisfaction and usage levels stemming from different apps could then be assessed for apps individually, instead of aggregating the overall increases in satisfaction or usage levels stemming from apps. Another modeling issue is that feature fatigue can perhaps be used as a separate variable, as in Steiner et al. (2009), and to use the number of apps as a predictor and satisfaction and usage as a consequence of feature fatigue. Different models of customer satisfaction can also be explored for assessing feature fatigue such as ambivalent models in which customers can be both satisfied and dissatisfied (e.g. Olsen et al., 2005) or more dynamic models in which temporal interactions of consumers with their smartphone are adopted. Regarding the perception of value, future research could adopt monetary costs as negative aspect of value which was omitted in this research. Likewise, this research has not accounted for psychometric differences based on the adoption of innovations. Moore (2005) posits that adopters of innovations can be segmented
43
based on their psychological profile and the moment of adoption of the innovation. However, with smartphones having achieved about 53% penetration in the Netherlands in January 2012 (Telecompaper, 2012), not all psychological profiles are expected to be represented in proportion in a random sample of smartphone users. When smartphone penetration reaches higher levels, it can be interesting to examine whether adopters with different psychographical profiles differ in value perception, usage and satisfaction. Late adopters might suffer from decreases in usability due to a high number of apps, as opposed to early adopters. A conceptual weakness is also the post-hoc analysis of mediating effects. The analysis of mediating or moderating effects should be motivated by a sound theoretical basis (Baron & Kenny, 1986). This was not the case for the analysis of mediating effects, which was motivated by direct effects of moderators that were detected in the analysis of moderating effects. Future research could provide a sound theoretical basis for the mediating effect of utilitarian attitude between the number of apps installed and the perceived usability. A final conceptual issue is the use of the number of apps installed on consumers smartphones. Most likely, different results would be obtained if the respondents were asked to indicate how many apps they actually used frequently. A discrepancy between the number of apps used frequently and the number of apps installed could then be a predictor for the extent to which consumers are fatigued by apps they don’t use. A second class of limitations is related to the research design. Although PLS is stated to be robust to deviations from normality (Hair et al., 2010, Ch. 12), caution could be advised when interpreting the results from variables which strongly deviate from the normal distribution. When non-normal distributed data is used, a “markedly larger sample size is needed despite the inclusion of highly reliable indicators in the model” (Marcoulides & Saunders, 2006, p. vi). Although no undisputed guidelines exist to determine which violations of normality can be countered with which increases in sample size, the analysis could be carried out again with data that is more manipulated such that it better represents a normal distribution, or with a “markedly” larger sample size. Limiting aspects are also identified in the measurement model. For single item constructs, the measured item directly reflects the construct, which is then not latent (Fuchs & Diamantopoulos, 2009). Therefore, no measurement error is specified in the model. This research assumed that constructs such as satisfaction are straightforward for consumers; an assumption that is debatable. Future research on this topic could consider using multiple-item constructs for the single-item constructs used here in order to allow for a more accurate measurement. Regarding the measurement of the number of apps, it could be that differentiation is needed between pre-installed apps and apps that consumers have downloaded themselves; perhaps consumers experience these apps differently. Another remark on the measurement model is the use of self-reported measures. Although inherent to the use of a survey, the information provided by consumers about daily usage duration of the smartphone and daily frequency of use could be seriously misspecified because those measures are difficult to accurately estimate by respondents. Future research could ask respondents to track their use with an app in order to obtain more accurate numbers. Additionally, the data could also be collected in a cross-sectional and longitudinal manner in order to further examine causality of the effects (Hair et al., 2010), which cannot be done with the current dataset. For instance, the direction of the relations in the model are
44
not always straightforward and require further investigation: do users that believe their device is easy to use install a great number of apps, or do people that have many apps on their device therefore think it is also easy to use? Additionally, evidence of model stability and generalizability can only come from performing the analysis on additional samples and contexts such as other product categories (e.g. tablets), postpaid users, users across different cultures, customers of different carriers and users of other brands of smartphones. The sample seemed to reflect the operator’s population of smartphone users, but additional samples could be used in order to further determine the generalizability of the results.
6 References
6.1 Literature references
Ajzen, I. (1991). “The theory of planned behavior.” Organizational Behavior and Human Decision Processes, 50 (2), pp. 179-211 Al-Gahtani, S.S., King, M. (1999). “Attitudes, satisfaction and usage: contributing to each in the acceptance of information technology.” Behavior & Information Technology, 18 (4), pp. 277-297 Anderson, E.W., Sullivan, M.W. (1993). “The antecedents and consequences of customer satisfaction for firms.” Marketing Science 12 (2), pp. 125-143 Anderson, J.C., Narus, J.A. (1998). “Business marketing: Understand what customers value.” Harvard Business Review, 76 (6), pp. 53-65 Andrews, F.M. (1984). “Construct validity and error components of survey measures: a structur al modeling approach.” The Public Opinion Quarterly, 48 (2), pp. 409-442, in: Turel, O., Serenko, A. (2006). “Satisfaction with mobile services in Canada: an empirical investigation.” Telecommunications Policy, 30, pp. 314-331 Arruda-Filho, E.J.M., Cabusas, J.A., Dholakia, N. (2010). “Social behavior and brand devotion among iPhone innovators.” International Journal of Information Management, 30, pp. 475-480 Arts., J.W.C., Frambach, R.T., Bijmolt, T.H.A. (2011). “Generalizations on consumer innovation adopt ion: A meta-analysis on drivers of intention and behavior.” International Journal of Research in Marketing, 28, pp. 134-144 Athanassopoulos, A.D., Iliakopoulos, A. (2003). “Modeling customer satisfaction in telecommunications: assessing the effets of multiple transaction points on the perceived overall performance of the provider.” Production and Operations Management, 12 (2), pp. 224-245 Backhaus, K., Erichson, B., PLinke, W., Weiber, R. (2003). “Multivariate analysemethoden: eine anwendungsorientierte einführung.” 10thedition, Springer, Berlin in: Götz, O., Liehr-Gobbers, K., Krafft, M. (2010). “Evaluation of structural equation models using the Partial Least Squares (PLS) approach.” Handbook of Partial Least Squares, Springer Handbooks of Computational Statistics, Springer-Verlag Berlin Heidelberg Bandura, A. (1989). “Self-efficacy mechanism in human agency.” American Psychologist, 37 (2), pp. 122-147, in: Davis, F.D. (1989). “Perceived usefulness, perceived ease of use, and user acceptance of information technology.” MIS Quarterly, 13 (3), pp. 319-340 Baron, R.M., Kenny, D.A. (1986). “The moderator -mediator variable distinction in social psychology research: Conceptual, strategic, and statistical considerations.” Journal of Personality and Social Psychology, 51 (6), pp. 1173-1182 Batra, R., Ahtola, O. (1990). “Measuring the hedonic and utilitarian sources of consumer attitudes.” Marketing Letters, 2 (2), pp. 159-170 Bayraktar, E., Tatoglu, E., Turkyilmaz, A., Delen, D., Zaim, S. (2011). “Measuring the ef ficiency of customer satisfaction and loyalty for mobile phone brands with DEA.” Expert Systems with Applications, 39, pp. 99-106 Bergkvist, L., Rossiter, J.R. (2007). “The predictive validity of multiple -item versus single-item measures of the same constructs.” Journal of Marketing Research, 44 (2), pp. 175-184 Bontis, N., Booker, L.D., Serenko, A. (2007). “The mediating effect of organizational reputation on customer loyalty and service recommendation in the banking industry.” Management Decision, 45 (9), pp. 14261445 Brombacher, A.C. (2005). “Reliability in strongly innovative products; a threat or a challenge?” Reliability Engineering and System Safety, 88 (2) , pp. 88-125.
45
Burson, K.A. (2007). “Consumer-Product Skill Matching: The effects of difficulty on relative self-assessment and choice.” Journal of Consumer Research, 34, pp. 104-110 Cadotte, E.R., Woodruff, R.B., Jenkins, R.L. (1987). “Expectations and norms in models of consumer satisfaction.” Journal of Marketing Research, 24 (3), pp. 305-314 Chang, Y., Chen, C.S. (2004). “Smart phone – the choice of client platform for mobile commerce.” Computer Standards & Interfaces, 27, pp. 329-336 Chase, R.B., Jacobs, J.F., Aquilano, N.J. (2006). “Operations management for competitive advantage.” 11th edition, New York: Mcgraw-Hill/Irwin Chen, J.V., Ross, W., Huang, S.F. (2008). “Privacy, trust, and justice considerations for location -based mobile telecommunication services.” Info, 10 (4), pp. 30-46 Chen, J.V., Yen, D.C., Chen, K. (2007). “The acceptance and diffusion of the innovative smart phone use: A case study of a delivery service company in logistics.” Information & Management, 46, pp. 241-248 Childers, T.L., Carr, C.L., Peck, J., Carson, S. (2001). “Hedonic and utilitarian motivations for online retail shopping behavior.” Journal of Retailing, 77 (4), pp. 511-535 Choi, J.H., Lee, H. (2012). “Facets of simplicity for the smartphone interface: a structural model.” International Journal of Human-Computer Studies, 70 (2), pp. 129-142 Churchill, G.A., Surprenant, C. (1982). “An investigation into the determinants of customer satisfaction.” Journal of Marketing Research, 19 (4), pp. 491-504 Cohen, J. (1988). “Statistical power analysis for the behavioral sciences.” Hillsdale, NJ: Lawrence Erlbaum in: Henseler, J., Fassot, G. (2010). “Testing moderating effects in PLS path models: an illustration of available procedures.” Handbook of Partial Least Squares. Springer Handbooks of Computational Statistics. Berlin Heidelberg. Cooper, H. M. (1984). “The integrative research review: A systematic approach.” Applied social research methods series, 2, Beverly Hills, CA: Sage. In: Randolph, J.J. (2009). “A guide to writing the dissertation literature review.” Practical Assessment, Research & Evaluation, 14 (13), pp. 1-13 Cronin, J.J., Brady, M.K., Hult, G.T.M. (2000). “Assessing the effects of quality, value and customer satisfaction on consumer behavioral intentions in service environments.” Journal of Retailing, 76 (2), pp. 193-218 Davis, F.D. (1989). “Perceived usefulness, perceived ease of use, and user acceptance of information technology.” MIS Quarterly, 13 (3), pp. 319-340 Deng, Z., Lu, Y., Wei, K.K., Zhang, J. (2010). “Understanding customer satisfaction and loyalty: an empirical study of mobile instant messages in China.” International Journal of Information Management, 30 , pp. 289-300 Eggert, A., Ulaga, W. (2002). “Customer perceived value: a substitute for satisfaction in business markets?” Journal of Business & Industrial Marketing, 17 (2/3), pp. 107-118 Falaki, H., Mahajan, R., Kandula, S., Lymberopoulos, D., Govindan, R., Estrin, D. (2010). “Diversity in Smartphone usage.” MobiSys 2010: Proceedings of the 8th International Conference on Mobile Systems, Applications and Services, pp. 179-194 Feitzinger, E., Lee, H.L. (1997). “Mass customization at Hewlett-Packard: The power of Postponement.” Harvard Business Review, January-February 1997, pp. 115-122 Field, A. (2009). “Discovering statistics using SPSS. Third edition, SAGE Publication Ltd., London. Chapter 10: “Comparing several means: ANOVA (GLM1).” Fishbein, M., Ajzen, I. (1975). “Belief, attitude, intention and behavior: An introduction to theory and research.” Reading MA Addison-Wesley Flavian, C., Guinalíu, M., Gurrea, R. (2005). “The role played by perceived usability, satisfaction and consumer trust on website loyalty.” Information & Management, 43, pp. 1-14 Flynn, L.R., Goldsmith, R.E. (1999). “A short, reliable measure of subjective knowledge.” Journal of Business Research, 46 (1), pp. 56-66 Fornell, C. (1992). “A national customer satisfaction barometer: The Swedish experience.” Journal of Marketing, 56, pp. 6-21 Fornell, C., Bookstein, F.L. (1982). “Two structural equation models: LISREL and PLS applied to consumer exit-voice theory.” Journal of Marketing Research, 19, pp. 440-452 Fornell, C., Johnson, M.D., Anderson, E.W., Chan, J., Everitt Bryant, B. (1996). “The American Customer Satisfaction Index: nature, purpose and findings.” Journal of Marketing, 60 (4), pp. 7-18 Fornell, C., Larcker, D.F. (1981). “Evaluating structural equation models with unobservable variables and measurement error.” Journal of Marketing Research, 18 (1), pp.39-50 Fournier, S., Mick, D.G. (1999). “Rediscovering satisfaction.” The Journal of Marketing, 63 (4), pp. 5-23 Fuchs, C., Diamantopoulos, A. (2009). “Using single -item measures for construct measurement in management research – conceptual issues and application guidelines. Die Betriebswirtshchaft, 69 (2), pp. 195-210
46
Funk, J.L. (2006). “The future of mobile phone -based internet applications: A view from Japan.” Technovation, 26, pp. 1337-1346 Gallarza, M.G., Saura, I.G. (2006). “Value dimensions, perceived value, satisfaction and loyalty: an investigation of university students’ travel behavior.” Tourism Management, 27, pp. 437-452 Gardial, S.F., Clemons, D.S., Woodruff, R.B., Schumann, D.W., Burns, M.J. (1994). “Comparing consumers’ recall of prepurchase and postpurchase product evaluation experiences.” Journal of Consumer Research, 20 (4), pp. 548-560 GfK (2011). “B2B.” Presentation on sectoral trends of GfK Retail and Technology GfK (2012). “Local Charts Smartphones Netherlands 2012 03.” GfK Retail and Technology Sales Figure Report Gorsuch, R.L. (1983). “Factor analysis.” 2nd edition, Hillsdale, NJ: Erblaum in: Floyd, F.J., Widaman, K.F. (1995). “Factor analsis in the development and refinement of clinical assessment instruments.” Psychological Assessment, 7 (3), pp. 286-299 Haenlein, M., Kaplan, A.M. (2004). “A beginner’s guide to Partial Least Squares analysis.” Understanding statistics, 3 (4), pp. 283-297 Hair, J.F., Black, B., Babin, B., Anderson, R.E. (2010). “Multivariate Data Analysis.” 7th edition, Pearson. Chapter 2: “Cleaning and transforming data.” Chapter 3: “Factor analysis.” Chapter 8: “ANOVA and MANOVA.” Chapter 11: “SEM: An introduction.” Chapter 12: Applications of SEM.” Hair, J.F., Ringle, C.M., Sarstedt, M. (2011). “PLS-SEM: Indeed a silver bullet.” Journal of Marketing Theory and Practice, 19 (2), pp. 139-151 Halstead, D. (1999). “The use of comparison standards in customer satisfaction research and management: a review and proposed typology.” Journal of Marketing Theory and Practice, 7 (3), pp. 13-26 Harzing, A. (2011). “Journal quality list.” 42nd edition, 16 October 2011 Haverila, M. (2011a). “What do we want specifically from the cell phone? An age related study.” Article in press for Telematics and Informatics Haverila, M. (2011b). “Mobile Phone feature preferences, customer satisfaction and repurchase intent among male users.” Australasian Marketing Journal, 19, pp. 238-246 Henseler, J., Fassot, G. (2010). “Testing moderating effects in PLS path models: an illustration of a vailable procedures.” Handbook of Partial Least Squares. Springer Handbooks of Computational Statistics. Berlin Heidelberg. Hirschman, A.O. (1970). “Exit, voice, and loyalty – responses to decline in firms, organizations and states.” Cambridge, MA: Harvard University Press, in: Johnson, M.D., Gustafsson, A., Andreassen, T.W., Lervik, L., Cha, J. (2001). “The evolution and future of national customer satisfaction index models.” Journal of Economic Psychology, 22, pp. 217-245, and in: Fornell, C., Johnson, M.D., Anderson, E.W., Chan, J., Everitt Bryant, B. (1996). “The American Customer Satisfaction Index: nature, purpose and findings.” Journal of Marketing, 60 (4), pp. 7-18 Hsee, C.K., Yang, Y., Gu, Y., Chen, J. (2009). “Specification seeking: how product sp ecifications influence consumer preference.” The Journal of Consumer Research, 35 (6), pp. 952-966 Institute of Electrical and Electronics Engineers (1998). “IEEE Standard for software and system test documentation.” Standard 829-1998 Jacobi, J. (1978). “Consumer research: how valid and useful are all our consumer behavior findings? A state of the art review.” The Journal of Marketing, 42 (2), pp. 87-96 Johnson, M.D., Gustafsson, A., Andreassen, T.W., Lervik, L., Cha, J. (2001). “The evolution and future o f national customer satisfaction index models.” Journal of Economic Psychology, 22, pp. 217-245 Kamis, A., Koufaris, M., Stern, T. (2008). “Using an attribute-based decision support system for usercustomized products online: An experimental investigation.” MIS Quarterly, 32 (1), pp. 159-177 Kaplan, A.M., Haenlein, M. (2006). “Towards a parsimonious definition of traditional and electronic mass customization.” Journal of Product Innovation Management, 23 (2), pp. 168-182 Keijzers, J., Ouden, E. den, Lu, Y. (2008). “Usability benchmark study of commercially available smartphones: Cell Phone type platform, PDA type platform and PC type platform.” Procedings of the 10th International Conference on Human Computer Interaction with Mobile Devices and Services , pp. 265272 Kekre, S., Krishnan, M.S., Srinivasan, K. (1995). “Drivers of customer satisfaction for software products: implications for design and service.” Management Science, 41 (9), pp. 1456-1470 Kim, S.H. (2008). “Moderating effects of job relevance and experience on mobile wireless technology acceptance: adoption of a smartphone by individuals.” Information & Management, 45, pp. 387-393 Klein, L.R. (1998). “Evaluating the potential of interactive media through a new lens: Search versus experience goods.” Journal of Business Research, 41 (3), pp. 195-203 LaBarbera, P.A., Mazursky, D. (1983). “A longitudinal assessment of consumer satisifaction/dissatisfaction: the dynamic aspect of the cognitive process.” Journal of Marketing Research, 20, pp. 393-404, in: Mittal., V., Ross Jr., W.T., Baldasare, P.M. (1998). “The asymmetric impact of negative and positive attribute -
47
level performance on overall satisfaction and repurchase intentions.” The Journal of Marketing, 62 (1), pp. 33-47 Lamb, C.W., Hair, J.F., McDaniel, C. (2012). “Essentials of marketing.” South-Western Cengage Learning, Mason USA Lancaster, K. (1971). “Consumer demand: A new approach.” New York: Columbia University Press, in: Thompson, D.V., Hamilton, R.W., Rust, R.T. (2005). “Feature fatigue: when product capabilities become too much of a good thing.” Journal of Marketing Research, 42 (4), pp. 431-442 Liao, C., Chen, J.L., Yen, D.C. (2007). “Theory of planning behavior (TPB) and customer sa tisfaction in the continued use of e-service: An integrated model.” Computers in Human Behavior, 23, pp. 2804-2822 Low, G.S., Lamb jr., C. W. (2000). “The measurement and dimensionality of brand associations.” Journal of Product & Brand Management, 9 (6), pp. 350-368 Marcoulides, G.A., Saunders, C. (2006). “PLS: a silver bullet?” MIS Quarterly, 30 (2), pp. iii-ix McDougall, G.H.G., Levesque, T. (2000). “Customer satisfaction with services: putting perceived value into the equation.” Journal of Services Marketing, 14 (5), pp. 392-410 Mitchell, A.A., Dacin, P.A. (1996). “The assessment of alternative measures of consumer expertise.” Journal of Consumer Research, 23, pp. 219-239 Mittal., V., Ross Jr., W.T., Baldasare, P.M. (1998). “The asymmetric impact of negative and positive attribute level performance on overall satisfaction and repurchase intentions.” The Journal of Marketing, 62 (1), pp. 33-47 Moore, Geoffrey A. (2006). “Crossing the chasm: Marketing and selling disruptive products to mainstream customers.” Collins Business Essentials Muthukrishnan, A.V., Weitz, B.A. (1991). “Role of product knowledge in evaluation of brand extension.” Advances in Consumer Research, 18, pp. 407-413 Nielsen, J. (1993). “Usability Engineering.” San Diego: Academic Press, in: Thompson, D.V., Hamilton, R.W., Rust, R.T. (2005). “Feature fatigue: when product capabilities become too much of a good thing.” Journal of Marketing Research, 42 (4), pp. 431-442 Ojiako, U., Maguire, S. (2009). “Seeking the perfect customer experience: a cause study of British telecom.” Strategic Change, 18, pp. 179-193 Oliver, R.L. (1993). “Cognitive, affective and attribute bases of the satisfaction response.” Journal of Consumer Research, 20 (3), pp. 418-430 Oliver, R.L. (1999). "Whence Customer Loyalty?" The Journal of Marketing, 63, pp. 33-44 Olsen, S.O., Wilcox, J., Olsson, U. (2005). “Consequences of ambivalence on satisfaction and loyalty.” Psychology & Marketing, 22 (3), pp. 247-269 Ouden, E. den, Yuan, L., Sonnemans, P.J.M., Brombacher, A.C. (2005). “Quality and Reliability problems from a consumer’s perspective: an increasing problem overlooked by business?” Quality and Reliability Engineering International, 22, pp. 821-838 Oulasvirta, A., Wahlström, M. , Ericsson, K.A. (2011). “What does it mean to be good at using a mobile device? An investigation of three levels of experience and skill.” International Journal of Human-Computer Studies, 69, pp. 155-169 Pahl, G., Beitz, W. (1998). “Engineering design: A systematic approach. Springer-Verlag in: Stone, R.S. (2000). “A heuristic method for identifying modules for product architectures.” Design Studies, 21, pp. 5-31 Pallant, J. (2005). “SPSS Survival Manual.” 12th version. Allen & Unwin. Park, Y., Chen, J.V. (2007). “Acceptance and adoption of the innovative use of smartphone.” Industrial Management & Data Systems, 107 (9), pp. 1349-1365 Payne, J.W. (1982). “Contingent decision behavior.” Psychological Bulletin, 92 (2), pp. 382-402 Peslak, A., Shannon, L., Ceccucci, W. (2011). “An empirical study of cell phone and smartphone usage.” Issues in Information Systems, 12 (1), pp. 407-417 Peterson, R.A., Wilson, W.R. (1992). “Measuring customer satisfaction: fact and artifact.” Journal of the Academy of Marketing Science, 20 (1), pp. 61-71 Pitt, L.F., Parent, M., Junglas, I., Chan, A., Spyropoulou, S. (2011). “Integrating the smartphone into a sound environmental information systems strategy: Principles, practices and a research agenda.” Journal of Strategic Information Systems, 20, pp. 27-37 Podsakoff, P.M., Podsakoff, N.P., Mackenzie, S.B., Lee, J. (2003). “Common method biases in behavioral research: A critical review of the literature and recommended remedies.” Journal of Applied Psychology, 88 (5), pp. 879-903 Radner, R., Rothschild, M. (1975). “On the allocation of effort.” Journal of Economic Theory (10), pp. 358-376, in: Davis, F.D. (1989). “Perceived usefulness, perceived ease of use, and user acceptance of information technology.” MIS Quarterly, 13 (3), pp. 319-340 Randolph, J.J. (2009). “A guide to writing the dissertation literature review.” Practical Assessment, Research & Evaluation, 14 (13), pp. 1-13
48
Reinzartz, W.J., Haenlein, M., Henseler, J. (2009). “An empirical comparison of the efficacy of covariancebased and variance-based SEM.” INSEAD Working Paper Scott Armstrong, J., Overton, T.S. (1977). “Estimating nonresponse bias in mail surveys.” Journal of Marketing Research, 14, pp. 396-402 Sobel, M.E. (1982). “Asymptotic confidence intervals for indirect effects in structural equation models.” Social Methodology, 13, pp. 159-186 Spiteri, J.M., Dion, P.A. (2004). “Customer value, overall satisfaction, end -user loyalty, and market performance in detail intensive industries.” Industrial Marketing Management, 33, pp. 675-687 Srinivasan, V., Lovejoy, W.S., Beach, D. (1997). “Integrated product design for marketability and manufacturing.” Journal of Marketing Research, 34, pp. 154-63, in: Thompson, D.V., Hamilton, R.W., Rust, R.T. (2005). “Feature fatigue: when product capabilities become too much of a good thing.” Journal of Marketing Research, 42 (4), pp. 431-442 Steiner, F., Tarman, R.T., Ihl, J.C., Piller, F.T. (2009). “Learning from the customer: Identifying changing user needs during product usage through embedded toolkits for user innovation.” Portland International Conference on Management of Engineering & Technology, August 2-6, 2009, pp. 706-716 Stone, R.S. (2000). “A heuristic method for identifying modules for product architectures.” Design Studies, 21, pp. 5-31 Szymanski, D.M., Henard, D.H. (2001). “Customer satisfaction: a meta-analysis of the empirical evidence.” Academy of Marketing Science Journal, 29 (1), pp. 16-35 Tabachnick, B.G., Fidell, L.S. (2001). “Using multivariate statistics.” 4th edition. HarperCollins Tan, W., Yeh, Y., Chen, S., Lin, Y., Kuo, C. (2012). “How consumers assess product’s features?: A case study of product features of smartphone.” Recent Researches in Applied Mathematics and Economics. Volume unknown, accessible via http://www.wseas.us/elibrary/conferences/2012/Vouliagmeni/MMAS/MMAS-20.pdf Teleompaper (2011). “Dutch mobile apps market.” Telecompaper Research Report 2011, October 2011 Teleompaper (2012). “Dutch mobile apps market.” Telecompaper Research Report 2011, January 2012 Thaler, R. (1985). “Mental accounting and consumer choices.” Marketing Science, 4 (3), pp. 199-214, in: Gallarza, M.G., Saura, I.G. (2006). “Value dimensions, perceived value, satisfaction and loyalty: an investigation of university students’ travel behavior.” Tourism Management, 27, pp. 437-452 Thompson, D.V., Hamilton, R.W., Rust, R.T. (2005). “Feature fatigue: when product capabilities become too much of a good thing.” Journal of Marketing Research, 42 (4), pp. 431-442 Tilson, D., Lyytinen, K. (2006). “The 3G transition: Changes in the US wireless industry.” Telecommunications Policy, 30 (10-11), pp. 569-586 Töyssy, S., Helenius, M. (2006). “About malicious software in smartphones.” Journal in Computer Virology, 2 (2), pp. 109-119 Turel, O., Serenko, A. (2006). “Satisfaction with mobile services in Canada: an empirical investigation.” Telecommunications Policy, 30, pp. 314-331 Venkatesh, V., Davis, F.D. (2000). “A theoretical extension of the Technology Acceptance Model: Four longitudinal field studies.” Management Science, 46 (2), pp. 186-204. Venkatesh, V., Morris, M.G., Davis, G.B., Davis, F.D. (2003). “User acceptance of information technology: Toward a unified view.” MIS Quartetly, 27 (3), pp. 425-478 Verkasalo, H., López-Nicolás, C., Molina-Castillo, F.J., Bouwman, H. (2010). “Analysis of users and non -users of smartphone applications.” Telematics and Informatics, 27, pp. 242-255 Vinzi, V.E., Chin, W.W., Henseler, J., Wang, H. (2009). “Handbook of partial least squares: Concepts, methods and applications.” Springer VerlagBerlin Heidelberg 2010 Voss, K.E., Spangenberg, E.R., Grohmann, B. (2003). “Measuring the hedonic and utilitarian dimensions of consumer attitude.” Journal of Marketing Research, 40 (3), pp. 310-320 Wang, Y., Lo, H., Yang, Y. (2004). “An integrated framework for service quality, customer value, satisfaction: evidence from China’s telecommunication industry.” Information Systems Frontiers, 6 (4), pp. 325-340 Woodruff, R.B. (1997). “Customer value: the next source for competitive advantage.” Academy of Marketing Science Journal, 25 (2), pp. 139-153 Yi, Y. (1990). “A critical review of consumer satisfaction.” Review of Marketing, 1990 in: Fournier, S., Mick, D.G. (1999). “Rediscovering satisfaction.” The Journal of Marketing, 63 (4), pp. 5-23
6.2 Web references
Gaigg, M. (2012). Blog by Micheal Gaigg (publication date unknown). “Feature fatigue: say NO to your clients.” http://www.michaelgaigg.com/blog/tag/feature-fatigue/ , accessed June 20th 2012 Gartner.com (2012): Summary of Quartile 1 2012 market share analysis of mobile devices worldwide, by Gartner – information technology research and advisory. http://www.gartner.com/id=2015916 ; Accessed July 27th 2012
49
IDC.com (2011). Press release on smartphone sales by market intelligence provider IDC. http://www.idc.com/getdoc.jsp?containerId=prUS22871611 , accessed 5th of August, 2011 iTunes.apple.com (2012). Online store where apps can be purchased. http://itunes.apple.com/nl/app/sleep-cycle-alarmclock/id320606217?mt=8 , accessed 7th of August, 2012 Kendrick, J. (2011). Blog by Kendrick, J. (2011). “Are you suffering from app fatigue?” http://www.zdnet.com/blog/mobilenews/are-you-suffering-from-app-fatigue/1305, accessed 27th of June 2012 Madison, J. (2011). “’The world is a better place because of Steve’: the life and times of Apple visionary Steve Jobs.” Appeared on Dailymail.co.uk (2011): Website of newspaper The Daily Mail. http://www.dailymail.co.uk/news/article-2045852/Steve-Jobs-dead-Biography-Apple-visionary.html, accessed 24th of July, 2012 Ringle, C.M., Wende, S., Will, S. (2005). Software package for Partial Least Squares analysis “SmartPLS 2.0 (M3) Beta.” http://www.smartpls.de , accessed May 29th 2012 Statline.cbs.nl (2012). Database of the Dutch Bureau for Satistics, http://statline.cbs.nl , accessed 17th of February 2012 Taliuaga, J., Bates, J., Sumner, J. (2009). “Feature fatigue: A necessary evil in portable devices.” http://www.rockresearch.com/feature-fatigue-a-necessary-evil-in-portable-devices , accessed 27th of June 2012
Appendix A: Measurement instrument
Construct or variable (Cronbach’s ? if applicable) Apps Factor loading (if applicable) Measure (source and Cronbach’s alpha where applicable)
Customer Satisfaction Of The Smartphone** Daily Usage Time
-
How many apps are installed in total on your smartphone, including pre-installed applications? 1-10, 11-20, 21-30, 31-40, 41-50, 51-60, 61-70, 71-80, 81-90, 91-100, more than 100, namely: … I am satisfied with my smartphone (11 point Likert scale: Peterson & Wilson, 1992) On average, how much time do you spend on using your smartphone per day? This concerns use without calling, but including texting, and using internet and applications. Measured in hours, with an interval of 0.25 hours (adapted from Venkatesh & Davis, 2000 and Al-Gahtani & King, 1999) On average, how many times do you use your smartphone per day? This concerns use without calling, but including texting, and using internet and applications (adapted AlGahtani & King, 1999) Mobile data usage in MB/month. This information is extracted from the customer base and is objective. The measure is the monthly data use, averaged over the 3 months before data collection: December 2011 through February 2012 Installing applications on my smartphone requires time (adapted from Wang et al., 2004) Updating applications on my smartphone requires time (adapted from Wang et al., 2004) Learning to use the applications on my smartphone requires time (adapted from Wang et al., 2004) My interaction with my smartphone is clear and understandable (adapted from Venkatesh & Davis, 2000, ? = .86-.98) Interacting with my smartphone does not require a lot of my mental effort (adapted from Venkatesh & Davis, 2000, ? = .86-.98) [is reversed in the survey] I find my smartphone to be easy to use (adapted from Venkat esh & Davis, 2000, ? = .86.98) I find it easy to get my smartphone to do what I want it to do (adapted from Venkatesh & Davis, 2000, ? = .86-.98) I often use my smartphone (adapted from Al-Gahtani & King, 1999). Using my smartphone increases my productivity in everyday activities (adapted from Venkatesh & Davis, 2000, ? = .87-.98) Using my smartphone enhances my effectiveness of everyday activities (adapted from Venkatesh & Davis, 2000, ? = .87-.98) I find my smartphone to be useful in everyday life (adapted from Venkatesh & Davis, 2000, ? = .87-.98) SMS usage in number of SMS text messages per month, averaged over December 2011through February 2012. This objective information is extracted from the customer base of the Telco I am satisfied with my telecom provider (11 point Likert scale: Peterson & Wilson, 1992)
Frequency Of Use
-
Mobile Data Usage (Objective Measure – Not In Survey) Perceived Effort* (? = .81)
-
.882 .905 .740
Perceived Usability/Ease Of Use* (? = .92)
.940 Deleted .922 .915
Perceived Usage Perceived Usefulness* (? = .84)
.885 .882 .848
SMS Usage (Objective Measure – Not In Survey) Telecommunication Provider /Company (Telco) Customer Satisfaction** Voice Usage (Objective Measure – Not In Survey)
-
-
-
Voice usage in number of minutes per month, averaged over December 2011 through February 2012. This objective information is extracted from the customer base of the Telco Smartphones are fun (adapted from Voss et al., 2003, ? = .92- .95) Smartphones are exciting (adapted from Voss et al., 2003, ? = .92- .95)
Moderating Variables Hedonic Attitude Towards The Smartphone* (? = .88)
.831 .894
50
Perceived Brand Image* (? = .95) Subjective Knowledge* (? = .92)
.882 .830 Deleted .938 .961 .964 .932 .830 .898 .910
Smartphones are delightful (adapted from Voss et al., 2003, ? = .92- .95) Smartphones are thrilling (adapted from Voss et al., 2003, ? = .92- .95) Smartphones are enjoyable (adapted from Voss et al., 2003, ? = .92- .95) The brand of my smartphone is attractive (Low & Lamb, 2000, ? = .85) The brand of my smartphone is reliable (Low & Lamb, 2000, ? = .85) The brand of my smartphone is of high quality (Low & Lamb, 2000, ? = .85) I know much about smartphones (Adapted from Flynn & Goldsmith, 1999, ? = .87 - .94) I know how to operate smartphones (Adapted from Flynn & Goldsmith, 1999, ? = .87 .94) Among my circle of friends, I’m one of the experts on smartphones (Flynn & Goldsmith, 1999, ? = .87 - .94) Compared to most other people, I know more about smartphones (Flynn & Goldsmith, 1999, ? = .87 - .94) Smartphones are effective (adapted from Voss et al., 2003, ? = .95) Smartphones are helpful (adapted from Voss et al., 2003, ? = .95) Smartphones are functional (adapted from Voss et al., 2003, ? = .95) Smartphones are necessary (adapted from Voss et al., 2003, ? = .95) Smartphones are practical (adapted from Voss et al., 2003, ? = .95)
Utilitarian Attitude Towards The Smartphone* (? = .94)
.904 Deleted .934 Deleted .917
Control Variables Age Gender Education Level Usage Time
-
Smartphone Brand Type Of Use Survey Method
-
What is your age? What is your gender (M/F)? What is your education level: primary education, vmbo/mbo 1, havo/vwo, mbo 2-4, hbo, wo, other (Statline.cbs.nl, 2012) How many smartphones have you owned in total, including your current smartphone? (adapted from Mitchell & Dacin, 1996) When did you purchase/receive your current smartphone? (which month and year, adapted from Turel & Serenko, 2006) ; or if this is not your first smartphone: when did you receive/purchase your first smartphone? Which brand is your current smartphone? Apple, Blackberry, Samsung, HTC , other: … (GfK, 2011) Do you use your smartphone for business, private, or both? On which device did you fill in this survey? On a PC/laptop, smartphone, tablet or other
All items of the constructs marked with an asterisk are measured on a 7 point Likert scale with 1: strongly disagree, 2: disagree, 3: somewhat disagree, 4: neutral (neither disagree nor agree), 5: somewhat agree, 6: agree, 7: strongly agree (adapted from Venktatesh & Davis, 2000). The constructs marked with a double asterisk are measured with an 11-point Likert scale. In Appendix B, the measurement instrument is integrated into a Dutch questionnaire.
Appendix B: Questionnaire
This Appendix contains the measurement instrument translated in Dutch. The questionnaire contains 35 questions in total. Some elements were added due to the interest of the telecom operator and in order to filter irrelevant respondents. These elements are marked with an “X” in the item indication. Additionally, every item is guided by a Dutch text in which the questions corresponding to the items are explained. Abbreviations: APP = number of apps, CS = customer satisfaction with the smartphone, TPCS = customer satisfaction with the telecommunication operator, PU = perceived usefulness, PEU = perceived ease of use/usability, DU = daily duration of use, FREQ = daily frequency of use, PULS = perceived usage level, LS = loyalty towards the smartphone, LTC = loyalty towards the telecommunication operator, PBI = perceived image of the smartphone brand, SK = subjective knowledge, UA/HA = utilitarian/hedonic attitude, PE = perceived effort put into apps, PMC = perceived monetary costs, PO = perceived overlap, CON = conscientiousness, UT = usage time, A = age, G = gender, E = education, SB = smartphone brand, SM = survey method, TU = type of use.
51
********* Start of the questionnaire ********** Deze enquête gaat over smartphones en het gebruik van apps. Als je meerdere smartphones in gebruik hebt, neem dan voor het invullen van deze vragenlijst de smartphone in gedachten die je het meest gebruikt. XA1. Volgens onze gegevens maak je gebruik van een smartphone en heb je een abonnement bij [telecom provider] Klopt dit? ? Ja (“Yes”) (1) ? Nee (“No”) (2) einde enquête: “Helaas val je buiten de doelgroep van het onderzoek. Hartelijk dank voor je medewerking.” De volgende twee stellingen gaan over jouw tevredenheid met je smartphone en met [telecom provider]. Geef aan in hoeverre je het eens bent met de volgende stellingen [1 (helemaal mee eens)…(11 helemaal mee oneens)]: CS Ik ben tevreden met mijn smartphone TPCS Ik ben tevreden met [telecom provider] De volgende stellingen gaan over het gebruik van jouw smartphone. Geef aan in hoeverre je het eens bent met de volgende stellingen: PU1 Door mijn smartphone kan ik meer doen in dezelfde tijd PU2 Door mijn smartphone kan ik alledaagse activiteiten beter uitvoeren PU3 Mijn smartphone is nuttig in het alledaagse leven PEU1 Mijn smartphone is duidelijk en begrijpelijk PEU2 Ik moet mij concentreren als ik mijn smartphone gebruik [REVERSED] PEU3 Mijn smartphone is eenvoudig in gebruik PEU4 Het is gemakkelijk om mijn smartphone te laten doen wat ik wil DU Hoeveel tijd besteed je gemiddeld per dag aan het gebruiken van je smartphone?Reken bellen hier niet onder en rond je antwoord af in kwartieren. 1 uur en 40 minuten per dag rond je bijvoorbeeld af naar 1:45 ? [in uren met intervallen van een kwartier; bij voorkeur twee uitklapvensters, links voor # uren 1 t/m 24, rechts voor het aantal kwartieren: 0, 15, 30, 45] FREQ Hoe vaak gebruik jij je smartphone gemiddeld op een dag? Reken bellen hier niet onder en denk hierbij dus aan het volgende gebruik: SMS-en en het gebruiken van internet en apps ? [Aantal (open antwoord) ] keer per dag PULS Ik maak veel gebruik van mijn smartphone (reken belgedrag hier niet onder) [Monthly mobile data, voice, and SMS usage is extracted from customer database in Megabytes/ minutes/SMS per month averaged over 3 months] De volgende vragen gaan over toekomstig gebruik van jouw smartphone. Geef aan in hoeverre je het eens bent met de volgende stellingen: XLS0 Mijn volgende telefoon wordt een smartphone LS1 Mijn volgende smartphone wordt van hetzelfde merk LS2 Als ik op dit moment zou kunnen kiezen zonder dat het geld of moeite kost om over te stappen, zou ik een smartphone van een ander merk nemen [Reversed]
52
LS3 Als mensen het mij zouden vragen dan zou ik het merk van mijn smartphone aanraden De volgende vragen gaan over het toekomstig gebruik van [telecom provider]. Geef aan in hoeverre je het eens bent met de volgende stellingen: LTC1 Mijn volgende mobiele abonnement wordt van [telecom provider] LTC2 Als ik op dit moment zou kunnen kiezen, zonder dat het geld of moeite kost om over te stappen, zou ik voor een abonnement van een andere aanbieder kiezen [Reversed] LTC3 Als mensen het mij zouden vragen dan zou ik [telecom provider] aanraden als mobiele aanbieder De volgende stellingen gaan over jouw beleving van het merk van jouw smartphone. Geef aan in hoeverre je het eens bent met de volgende stellingen: PBI1 Het merk van mijn smartphone is aantrekkelijk PBI2 Het merk van mijn smartphone is betrouwbaar PBI3 Het merk van mijn smartphone staat voor hoge kwaliteit De volgende vragen gaan over jouw kennis over smartphones. Geef aan in hoeverre je het eens bent met de volgende stellingen: SK1 Ik weet veel van smartphones SK2 Ik kan goed met smartphones overweg SK3 In mijn vriendenkring ben ik een van de smartphone-experts SK4 In vergelijking met anderen weet ik veel van smartphones De volgende 10 stellingen gaan over jouw houding ten opzichte van smartphones. Geef aan in hoeverre je het eens bent met de volgende stellingen: HA1 Smartphones zijn leuk HA2 Smartphones zijn boeiend HA3 Smartphones geven voldoening HA4 Smartphones zijn spannend HA5 Smartphones zijn plezierig UA1 Smartphones zijn effectief UA2 Smartphones zijn behulpzaam UA3 Smartphones zijn functioneel UA4 Smartphones zijn noodzakelijk UA5 Smartphones zijn praktisch De volgende vragen gaan over de apps die op jouw smartphone staan geïnstalleerd en over het gebruik van deze apps. APP Hoeveel apps heb jij in totaal op jouw smartphone geïnstalleerd? Reken hiertoe ook vooraf geïnstalleerde apps mee, zoals de SMS-app of agenda-apps. ? 1-10 (1) ? 11-20 (2) ? 21-30 (3) ? 31-40 (4) ? 41-50 (5) ? 51-60 (6) ? 61-70 (7) ? 71-80 (8)
53
? ? ?
81-90 (9) 91-100 (10) Meer dan honderd, namelijk: [open antwoord met waarde >100]
(11)
[It should be noted that an error occurred in the design of the web survey: the first 33 respondents were not able to provide a 15 minute interval in their answer to the question how much time they spend on an average day on using their smartphone, when they selected ‘1 hour’ as an answer option. As a result, the following answer options ‘15’, ‘30’ and ‘45’ minutes were not available to these 33 respondents. Fortunately, the lowest answer to this question from this batch of respondents was 1:45 hours, which is an indication that the leavingout of the answer options did not affect the results because it could be expected that respondents that wanted to fill in the answer options that were not available would have indicated the lowest possible answer, which is 0 hours or 1 hour. After one day the answer options were adapted such that the left-out answer options are made available to respondents.] XOWNAPPS Heb je naast de apps die reeds op je telefoon stonden, ook zelf apps geïnstalleerd? ? Ja (1) ? Nee (2) Niet vragen (“Skip”): PMC1, PMC2, PMC3, PE1 Geef aan in hoeverre je het eens bent met de volgende stellingen: PMC1 De apps die ik op mijn smartphone heb geïnstalleerd zijn redelijk geprijsd [Reversed] PMC2 De verkoopprijs van apps die ik heb geïnstalleerd op mijn smartphone is voordelig[Reversed] PMC3 De apps op mijn smartphone zijn hun geld waard [Reversed] PE1 Het heeft veel tijd gekost om de apps op mijn smartphone te installeren PE2 Het updaten van de apps op mijn smartphone kost veel tijd PE3 Ik heb veel tijd geïnvesteerd om de apps op mijn smartphone te leren gebruiken De volgende stellingen gaan over jouw beleving van de onderlinge verschillen tussen de apps op jouw smartphone. Geef aan in hoeverre je het eens bent met de volgende stellingen: PO1 Er is veel overlap tussen de functies van de apps op mijn smartphone PO2 Ik heb meerdere apps voor hetzelfde doeleind (Bijvoorbeeld meerdere apps voor reizen met het openbaar vervoer of voor weersvoorspellingen) PO3 De verschillende apps op mijn smartphone voorzien in dezelfde behoeften De volgende stellingen gaan over het beheren van apps op jouw smartphone, welke afwegingen er plaatsvinden bij het downloaden van apps en de functionaliteit van jouw smartphone. Geef aan in hoeverre je het eens bent met de volgende stellingen [Measured on a five-point Likert scale] : XSTR Ik structureer de apps op mijn smartphone (denk hierbij bijvoorbeeld aan het beheren van mappen waarin je de apps hebt opgeslagen (voor iPhone gebruikers) of het plaatsen van widgets en snelkoppelingen (voor Android gebruikers) XDEL Ik verwijder apps van mijn smartphone als ik ze niet gebruik XCOST Ik let op de kosten van een app als ik overweeg om een app te downloaden XMEM Ik let op de hoeveelheid geheugen die een app inneemt op mijn smartphone als ik overweeg om een app te downloaden XFUNC Ik vind het belangrijk dat mijn smartphone andere functionaliteit biedt naast bellen XFF Ik erger mij aan apps die ik niet gebruik XOA8 Geef een top 3 van apps die je het meest gebruik [Open vraag] Tot slot volgen er nu nog enkele algemene vragen over jou en jouw smartphone. Geef aan in hoeverre je het eens bent met de volgende stellingen:
54
Con1 Ik let op details Con2 Ik begin meteen aan taken Con3 Ik vind het prettig als dingen geordend zijn Con4 Ik ben precies in mijn werk Con5 Ik laat mijn spullen rondslingeren XSEGMENT1 Welk van de volgende antwoorden associeer jij het meest met het gebruik van apps? ? Ontdekken, proberen, testen (1) ? Delen, tonen, mode rn (2) ? Socialiseren, relaties, vrienden (3) ? Functioneel, praktisch, oplossing (4) XSEGMENT2 Apps zijn voor mij vooral ? Nieuw, onbekend, vreemd ? Praktisch, eenvoudig, bekend ? Vertrouwd, handig, leuk, ? Intens, actief, onmisbaar
(1) (2) (3) (4)
XSEGMENT3 Hoe zou je je leven over het algemeen omschrijven? ? Gestructureerd, duidelijk rechtlijnig (1) ? Druk, spanning, snel (2) ? Complex, verwarrend, veranderlijk (3) ? Groei, vooruitgang, opbouwen (4) ? Geleidelijk, rustig, uitgebalanceerd (5) A Hoe oud ben je? ? [leeftijd] G Ik ben een ? Man ? Vrouw
(1) (2)
E Wat is het niveau van je hoogst genoten of je huidige opleiding? ? Geen (1) ? Primair onderwijs of vergelijkbaar (2) ? vmbo/mbo 1 of vergelijkbaar (3) ? havo/vwo of vergelijkbaar (4) ? mbo 2-4 of vergelijkbaar (5) ? hbo of vergelijkbaar (6) ? wo of vergelijkbaar (7) ? anders (8) UT1 Hoeveel smartphones heb je gehad (inclusief je huidige smartphone)? ? Dit is mijn eerste smartphone (1) ? [aantal] smartphones (2) UT2 Wanneer heb je je huidige smartphone gekregen? ? [Maand en jaar] UT3 Wanneer heb je je eerste smartphone gekregen? ? [maand en jaar]
UT3 niet vragen
55
SB Van welk merk is je smartphone? ? Apple ? Blackberry ? HTC ? Samsung ? Anders
(1) (2) (3) (4) (5)
TU Gebruik je je smartphone zakelijk, privé of voor beiden? ? Privé (1) ? Zakelijk (2) ? Beiden (3) SM Met welk apparaat heb je deze enquête ingevuld? ? Op een PC/laptop (1) ? Op een smartphone (2) ? Op een tablet (3) Hartelijk dank voor het invullen van deze enquête! *********** End of the questionnaire ***********
Appendix C: Data manipulation
This appendix describes the manipulation of the dataset that was delivered by the market research agency. First of all, the e-mail addresses and phone numbers of the respondents were deleted in order to ensure animosity of the respondents. The labels added by the market research agency were replaced by the original labels as used in the measurement instrument. The information containing the gender and handset brand in use that stemmed from the customer database was deleted because it is expected that this information is more reliable when directly provided by respondents. The two variables for duration of daily usage (hours and minutes) were aggregated to from 1 variable: hours of use. 3 respondents indicated that they filled in the survey on an ‘other’ device than a smartphone, tablet or desktop/laptop PC. However, these cases concerned an “I MAC” an “Ipad” and a “MAC” of which the MAC and I MAC can be categorized under desktops and laptop PC’s and the iPad can be categorized as a tablet. Therefore, these cases were recoded. The variables for the duration of owning a smartphone were joined such that the months and years of owning a smartphone were combined in a single variable. The different variables for the different paths in the survey (for people having owned multiple smartphones and people owing their first smartphone) were then joined to reflect the time that people have owned a smartphone in total, in months. Subsequently, the data was saved in SPSS and the corresponding labels were added. The item perceived ease of use2 was reversed in the questionnaire and therefore they were reverse recoded in order to reflect the original variable. The brand of the smartphone as provided by the respondent matched the brand of the handset of the customer database for every respondent, except for the 4 cases for which this question was answered with ‘other’; these answers were assigned as missing.
56
For the number of apps installed on the smartphone, the answer option above 100 was “other, namely:”. The maximum value for this item was 188 apps. The answers above 100 were recoded such that they corresponded to the value of the last group on the scale. This last group on the scale, consisting of respondents with more than 101 apps on their smartphone, consisted of 8 respondents. One respondent indicated to be of an age of 17.5 years. This case was rounded to 18 years. The data on SMS and voice usage was delivered per month for December 2011February 2012. This data was averaged and rounded with two digits. Finally, because the distribution of mobile data, SMS and voice usage appeared to be asymptotic, i.e. the distributions had “long tails”. A third root transformation was applied to the mobile data usage and SMS usage, and a fourth root transformation was applied to voice usage. These specific transformations were selected because they yielded distributions of the corresponding variables that best approached the normal distribution. Recoding of daily usage duration and frequency of daily interaction Scale increments for the variables ‘daily interaction duration’ and ‘daily frequency of use’ were created such that the mean was in the middle-answer option and the increments of the scale corresponded to 0.5 standard deviations, in an attempt to create a 7-point scale which corresponds to the scale-size of other measured items and to allow for sufficient discrimination between the categories. In Table 8, the calculations of the scales are displayed. Frequency of use was now measured on a 6-point scale and daily duration of use on a 7-point scale. The scale boundaries are not strict; i.e. if scale value 1 ranges from 0.00 to 5.00; 5.00 is assigned value 1 but 5.01 is assigned value 2. The values of both variables were recoded to reflect the values of the new scales that were developed.
Table 8 Recoding of daily duration of use and dialy frequency of use
Value: 1 2 3 4 5 6 7 mean S.D. min value max value Daily duration of use lower bound upper bound 0.00 1.33 1.33 2.42 2.42 3.51 3.51 4.59 4.59 5.68 5.68 6.77 6.77 or more 4.05 2.17 1.25 17.25 Daily frequency of use lower bound upper bound 0 5 5 17 17 29 29 41 41 54 54 or more mean S.D. min value max value 23.08 24.50 0.00 200.00
Coding of dummy variables for survey method and type of use In the Table 9, the values for the dummy variables and the corresponding categories can be viewed. Since only 1 respondent indicated to use his smartphone for business uses, this category was not included in the model and was treated as missing data.
57
Table 9 Coding of dummy variables
Variable Age (original code) Male (1) Female (2) Type of use categories (original code) Consumer (1) Business (2) (not included into analysis due to only 1 respondent) Consumer and Business (3) Survey method categories (original code) PC/laptop (1) Smartphone (2) Tablet (3) Smartphone brand categories (original code) Apple (1) Blackberry (2) HTC (3) Samsung (4) Other (5) is not controlled for Recoded into variables Original variable 0 1 Original variable 0 1 SMDummy1 0 1 0 SBDummy1 0 1 0 0 SMDummy2 0 0 1 SBDummy2 0 0 1 0 -
SBDummy3 0 0 0 1 -
Appendix D: Generalization of the sample
Table 10 General characteristics of the sample and other customer groups
Consumer postpaid base (smartphone, 4 brands) 47,33% 52,67% 3,17% 31,33% 21,78% 24,75% 14,40% 3,92% 0,66% Consumer postpaid + prepaid base (smartphone, 4 brands) 47,53% 52,47% 8,37% 30,17% 20,21% 23,04% 13,44% 3,88% 0,89% Consumer + business, postpaid + prepaid (Smartphone, 4 brands)
Variable
Category
Respondents
Dutch population 49.51%* 50.49% 13,49%* 12,03% 14,44% 14,87% 13,02% 8,39% 7,13% 5.13%** 18.06% 9.05% 33.20% 21.79% 11.80% 0.97%
Gender Age
Education
Smartphone brand
Male Female 15-25 26-35 36-45 46-55 56-65 66-75 75+ None Primary VMBO/MBO1 HAVO/VWO MBO 2-4 HBO WO other Apple Blackberry HTC Samsung
48.8% 51.2 % 18.3 % 18.3 % 25.0 % 25.4 % 10.3 % 1.2 % 0.8 % 0.4 % 0.4 % 11.9 % 29% 28.2 % 31.3 % 13.5 % 2.8% 18.7 % 13.5% 32.1% 34.1 %
8,25% 29,97% 20,31% 23,18% 13,49% 3,91% 0,89%
21,60% 17,76% 17,92% 42,72%
19,87% 19,65% 16,92% 43,55%
20,06% 19,74% 16,87% 43,32%
*As of January 1st, 2012; **Over 2011. Source: Telecommunication operator’s customer base data and Statline.cbs.nl (2012).
58
Appendix E: Data exploration: outliers and normality
Table 11 Univariate outliers
Case ID 528067 521834 523001 523188 521892 527015 527748 522188 522723 5226821 522188 522473 525776 526767 527078 526776 521772 522778 521768 522216 Variable ULS1: Smartphone Use Per Day ULS1: Smartphone Use Per Day ULS1: Smartphone Use Per Day ULS2: Frequency Of Smartphone Use Per Day ULS2: Frequency Of Smartphone Use Per Day ULS2: Frequency Of Smartphone Use Per Day MDU UT1: Number Of Smartphones Owned UT1: Number Of Smartphones Owned UT1: Number Of Smartphones Owned UT2: Months Owning A Smartphone UT2: Months Owning A Smartphone Age Age SMS SMS SMS Voice Voice Voice Z-Value 7.2 6.1 4.2 14.7 7.22 5.18 6.3 6.6 5.0 4.2 4.7 4.5 13.8 4.4 5.19 6.20 6.83. 4.60 4.92 8.20 Variable value 21.5 hours/day 17.25 hours/day 13.25 hours/day 1000 times per day 200 150 1503 Mb/month 10 smartphones 8 smartphones 7 smartphones 135 months 131 months 500 190 475.67 556.33 607.33 597.05 631.17 991.08 Outlier category Procedural error Extraordinary observation Extraordinary observation Procedural error Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Procedural error Procedural error Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Extraordinary observation Action Delete value Retain Retain Delete value Retain Retain Delete value Retain Retain Retain Retain Retain Delete value Delete value Delete value Delete value Delete value Delete value Delete value Delete value
The outliers are displayed in Table 11. The outliers are categorized according to 3 categories provided by Hair et al. (2010): procedural errors, extraordinary events or extraordinary observations. The following is observed about the outliers: Outliers 528067, 523188, 525776 and 526767 are classified as procedural errors because of their unrealistic values. Respondent 528067 indicates to use his/her smartphone for 21.25 hours per day, with a frequency of 25 times per day, meaning an average of 51 minutes per interaction. Due to this improbable value, which indicates either use of a smartphone during sleep or 2.75 hours sleep per day, the value was deleted. Respondent 523188 indicates a frequency of 1000 interactions with his/her smartphone per day, with the next runner up in the dataset indicating a value of 200. The respondent indicates a use of 9.25 hours per day and uses 607 Mb of data per month and the rest of the values seem normal. It is assumed that the value of 1000 is a typing error, and the value is deleted. The same is assumed for respondents 525776 and 526767, indicating an age of respectively 500 and 190 years. These values are deleted. The other univariate outliers are classified as extraordinary observations because there are no obvious explanations for them but the values are realistic and plausible. The value of 17.25 hours smartphone use per day is high, but possible for someone using his/her smartphone for both business and pleasure. For the objective data on voice, SMS and mobile data usage, the outlying values are deleted since these are central variables and in order to prevent asymptotic data, could lead to unreliable estimation of parameters in PLS (Vinzi et al., 2009). The outliers of the control variables are retained.
The distribution of the variables is displayed in Table 12.
59
Table 12 Data distribution
Shape descriptions Variable Zvalue* Age -0.32 -1.02 Perceived Usability3 Apps 0.43 1.39 Perceived Usability4 Daily Usage Time -0.24 -0.77 Perceived Usage Level Duration Of Ownership 4.35 Perceived Usefulness1 14.11 Education -0.21 -0.68 Perceived Usefulness2 Frequency of use -0.10 -0.33 Perceived Usefulness3 Hedonic Attitude1 4.36 Smartphone Cust Sat 14.13 Hedonic Attitude2 1.51 Smartphones Owned 4.91 Hedonic Attitude3 0.38 1.22 SMS Usage** Hedonic Attitude4 0.12 0.38 Subjective Knowledge1 Hedonic Attitude5 1.86 Subjective Knowledge2 6.04 Mobile Data Usage** -0.72 -2.32 Subjective Knowledge3 Perceived Brand Image1 1.42 Subjective Knowledge4 4.61 Perceived Brand Image2 1.66 Telco Cust Sat 5.39 Perceived Brand Image3 1.89 Utilitarian Attiude1 6.12 Perceived Effort1 0.37 1.18 Utilitarian Attitude2 Perceived Effort2 -0.53 -1.73 Utilitarian Attitude3 Perceived Effort3 -0.16 -0.52 Utilitarian Attitude4 Perceived Usability1 1.89 Utilitarian Attitude5 6.13 Perceived Usability2 -1.09 Voice Usage** -3.52 *Values exceeding ± 2.58 (significance p<.01) have been highlighted in bold. statistic statistic Skewness Zvalue* 0.02 0.14 0.98 6.37 0.83 5.37 1.91 12.39 -0.35 -2.26 0.80 5.05 -1.58 -10.26 -1.02 -6.62 -0.74 -4.79 -0.50 -3.22 -1.09 -7.07 -0.13 0.84 -1.27 -8.21 -1.29 -8.38 -1.33 -8.64 0.94 6.07 0.57 3.69 0.60 3.87 -1.29 -8.37 0.10 0.62 Kurtosis Variable Shape descriptions Skewness Zvalue* -1.39 -8.98 -1.10 -7.15 -1.12 -7.27 -0.69 -4.44 -0.71 -4.63 -1.47 -9.51 -1.39 -9.01 2.64 17.12 0.25 1.64 -0.44 -2.87 -1.00 -6.51 -0.01 -0.05 -0.31 -1.98 -0.42 -2.70 -0.95 -6.19 -1.17 -7.58 -1.35 -8.73 -0.22 -1.46 -1.26 -8.14 0.01 0.09 Kurtosis statistic 2.85 0.62 1.13 0.11 0.29 3.01 1.99 10.34 0.05 -0.52 1.11 -0.85 -0.54 0.26 2.39 2.33 4.59 -0.53 4.06 0.04 Zvalue* 9.22 2.00 3.67 0.35 0.95 9.75 6.46 33.51 0.15 -1.70 3.61 -2.77 -1.76 0.84 7.74 7.55 14.89 -1.70 13.16 0.12 with N
statistic
and
= sample size = 252. Note that the distribution is reflected by the data of which the outliers have been deleted. ** these variables have been transformed, see section 4.1.3
60
Appendix F: Factor loadings and cross loadings of the first CFA
Table 13 Factor loadings and cross loadings of first confirmatory factor analysis
Telco cust sat
Perceived brand image
Apps 1 .32 .057 Cust sat .32 1 .039 Daily usage time .057 .039 1 Frequency .152 .178 .489 Hedonic atttiude1 .193 .165 .244 Hedonic attitude2 .158 .063 .262 Hedonic attitude3 .091 .083 .263 Hedonic attitude4 .12 .055 .266 Hedonic attitude5 .174 .146 .186 Mobile data usage .178 .107 .222 Perceived brand image1 .23 .629 .052 Perceived brand image2 .225 .657 .032 Perceived brand image3 .249 .67 .008 Perceived effort1 -.21 -.253 .084 Perceived effort2 -.19 -.343 .058 Perceived effort3 -.089 -.165 -.003 Perceived usability1 .334 .576 .162 Perceived usability2 .085 .124 .055 Perceived usability3 .314 .542 .145 Perceived usability4 .315 .609 .093 Perceived usefulness1 .163 .402 .155 Perceived usefulness2 .287 .39 .143 Perceived usefulness3 .26 .479 .184 Perceived usage .226 .339 .408 Subjective knowledge1 .216 .117 .211 Subjective knowledge2 .276 .157 .142 Subjective knowledge3 .281 .104 .134 Subjective knowledge4 .289 .122 .195 SMS usage -.027 -.02 -.01 Telco cust sat .026 .312 .02 Utilitarian attitude1 .212 .145 .213 Utilitarian attitude2 .248 .207 .221 Utilitarian attitude3 .218 .214 .208 Utilitarian attitude4 .128 .04 .191 Utilitarian attitude5 .188 .255 .204 Voice usage -.062 .091 -.02 Note: item loadings above .70 are highlighted
.152 .178 .489 1 .299 .257 .241 .172 .235 .307 .106 .171 .151 -.028 .007 -.061 .264 .114 .231 .219 .244 .22 .309 .42 .352 .269 .317 .287 .05 .01 .274 .244 .263 .072 .215 .04
.175 .125 .285 .288 .842 .866 .867 .814 .862 .125 .264 .251 .294 -.061 -.099 -.036 .376 .03 .34 .314 .305 .296 .354 .399 .426 .479 .324 .345 -.008 .16 .667 .704 .679 .459 .598 -.036
.178 .107 .222 .307 .185 .121 .1 .019 .08 1 .038 .084 .08 -.069 -.059 -.087 .183 .092 .267 .181 .2 .207 .231 .303 .251 .258 .285 .303 .082 -.03 .185 .148 .1 .117 .124 .116
.246 .683 .031 .15 .28 .285 .21 .141 .266 .071 .938 .962 .964 -.242 -.262 -.127 .63 .016 .557 .589 .523 .457 .528 .399 .232 .297 .185 .197 -.001 .264 .293 .298 .35 .145 .322 .004
-.205 -.318 .064 -.023 -.08 -.068 -.053 -.037 -.11 -.079 -.214 -.287 -.254 .882 .905 .74 -.272 -.245 -.334 -.323 -.139 -.169 -.088 -.036 -.023 -.13 -.012 .057 .029 -.158 -.079 -.109 -.093 .035 -.059 -.009
.348 .622 .146 .261 .378 .27 .316 .247 .337 .228 .586 .605 .629 -.278 -.339 -.23 .939 .217 .921 .915 .475 .437 .543 .505 .31 .441 .238 .276 .016 .187 .429 .393 .442 .145 .428 -.014
.272 .49 .186 .299 .324 .32 .351 .25 .305 .246 .575 .528 .556 -.118 -.154 -.093 .567 -.025 .475 .53 .885 .882 .848 .52 .389 .361 .338 .376 -.134 .22 .42 .422 .385 .333 .453 -.035
.226 .339 .408 .42 .39 .344 .332 .268 .347 .303 .392 .383 .367 -.061 -.009 -.025 .497 .072 .483 .427 .423 .413 .513 1 .351 .459 .26 .331 -.032 .077 .404 .384 .431 .208 .42 .016
.297 .143 .192 .343 .355 .456 .399 .401 .321 .307 .269 .231 .249 -.047 -.005 -.064 .358 .041 .302 .358 .365 .359 .354 .403 .931 .831 .897 .909 .039 .001 .358 .346 .34 .323 .333 -.024
-.027 -.02 -.01 .05 .002 -.001 0 -.023 -.017 .082 -.021 -.001 .019 .055 -.001 .025 .061 .138 0 -.037 -.139 -.127 -.089 -.032 .075 .067 -.012 -.007 1 -.017 .033 .018 -.012 -.059 -.088 .034
.026 .312 .02 .01 .126 .054 .134 .16 .208 -.03 .278 .225 .253 -.124 -.176 -.075 .18 -.011 .164 .182 .234 .17 .172 .077 .031 .055 -.078 -.027 -.017 1 .142 .21 .145 .162 .224 .081
.241 .218 .246 .267 .676 .594 .625 .541 .719 .161 .328 .319 .344 -.07 -.089 -.036 .461 .02 .415 .402 .386 .377 .484 .453 .37 .501 .225 .282 -.022 .209 .883 .867 .923 .575 .884 -.049
-.062 .091 -.02 .04 -.016 -.021 -.058 -.051 -.012 .116 -.005 .017 .001 .045 -.085 .06 -.019 .01 .01 -.032 .016 -.045 -.06 .016 .011 -.02 -.02 -.058 .034 .081 -.03 .014 -.045 -.1 -.073 1
61
Voice usage
Mobile data usage
Daily usage time
SMS usage
Subjective knowledge
Frequency
Perceived usefulness
Utilitarian attitude
Perceived effort
Perceived usability
Perceived usage
Henodic attitude
Cust sat
Apps
Appendix G: Factor loadings and cross loadings of the definite CFA
Table 14 Factor loadings and cross loadings of definite confirmatory factor analysis
Telco cust sat Perceived brand image Voice usage -.062 .091 -.02 .04 -.016 -.021 -.058 -.051 .116 -.005 .017 .001 .045 -.085 .06 -.019 .01 -.032 .016 -.045 -.06 .016 .011 -.02 -.02 -.058 .034 .081 -.03 -.045 -.073 1 Mobile data usage Daily usage time
SMS usage
Subjective knowledge
Frequency
Perceived usefulness
Apps Cust sat Daily usage time Frequency Hedonic atttiude1 Hedonic attitude2 Hedonic attitude3 Hedonic attitude4 Mobile data usage Perceived brand image1 Perceived brand image2 Perceived brand image3 Perceived effort1 Perceived effort2 Perceived effort3 Perceived usability1 Perceived usability3 Perceived usability4 Perceived usefulness1 Perceived usefulness2 Perceived usefulness3 Perceived usage Subjective knowledge1 Subjective knowledge2 Subjective knowledge3 Subjective knowledge4 SMS usage Telco cust sat Utilitarian attitude1 Utilitarian attitude3 Utilitarian attitude5 Voice usage
1 .32 .057 .152 .193 .158 .091 .12 .178 .23 .225 .249 -.21 -.19 -.089 .334 .314 .315 .163 .287 .26 .226 .216 .276 .281 .289 -.027 .026 .212 .218 .188 -.062
.32 1 .039 .178 .165 .063 .083 .055 .107 .629 .657 .67 -.253 -.343 -.165 .576 .542 .609 .402 .39 .479 .339 .117 .157 .104 .122 -.02 .312 .145 .214 .255 .091
.057 .039 1 .489 .244 .262 .263 .266 .222 .052 .032 .008 .084 .058 -.003 .162 .145 .093 .155 .143 .184 .408 .211 .142 .134 .195 -.01 .02 .213 .208 .204 -.02
.152 .178 .489 1 .299 .257 .241 .172 .307 .106 .171 .151 -.028 .007 -.061 .264 .231 .219 .244 .22 .309 .42 .352 .269 .317 .287 .05 .01 .274 .263 .215 .04
.166 .112 .3 .288 .831 .894 .882 .83 .132 .257 .237 .285 -.047 -.092 -.022 .367 .327 .307 .298 .302 .353 .394 .442 .475 .35 .366 -.005 .136 .638 .648 .576 -.042
.178 .107 .222 .307 .185 .121 .1 .019 1 .038 .084 .08 -.069 -.059 -.087 .183 .267 .181 .2 .207 .231 .303 .251 .258 .285 .303 .082 -.03 .185 .1 .124 .116
.246 .683 .031 .15 .28 .285 .21 .141 .071 .938 .961 .964 -.242 -.262 -.127 .63 .557 .589 .523 .457 .528 .399 .232 .297 .185 .197 -.001 .264 .293 .35 .322 .004
-.205 -.318 .064 -.023 -.08 -.068 -.053 -.037 -.079 -.214 -.287 -.254 .882 .905 .74 -.272 -.334 -.323 -.139 -.169 -.088 -.036 -.023 -.13 -.012 .057 .029 -.158 -.079 -.093 -.059 -.009
.347 .622 .145 .258 .38 .271 .318 .247 .226 .591 .608 .633 -.272 -.333 -.22 .94 .922 .915 .482 .441 .548 .506 .313 .439 .241 .279 .01 .189 .432 .443 .43 -.015
.272 .49 .186 .299 .324 .32 .351 .25 .246 .575 .528 .557 -.118 -.154 -.093 .567 .475 .53 .885 .882 .848 .52 .389 .361 .338 .376 -.134 .22 .42 .385 .454 -.035
.226 .339 .408 .42 .39 .344 .332 .268 .303 .392 .383 .367 -.061 -.009 -.025 .497 .483 .427 .423 .413 .513 1 .351 .459 .26 .331 -.032 .077 .404 .431 .42 .016
.297 .143 .192 .343 .355 .455 .399 .401 .307 .269 .231 .249 -.046 -.005 -.064 .358 .301 .358 .365 .359 .354 .403 .932 .83 .898 .91 .039 .001 .357 .34 .333 -.024
-.027 -.02 -.01 .05 .002 -.001 0 -.023 .082 -.021 -.001 .019 .055 -.001 .025 .061 0 -.037 -.139 -.127 -.089 -.032 .075 .067 -.012 -.007 1 -.017 .033 -.012 -.088 .034
.026 .312 .02 .01 .126 .054 .134 .16 -.03 .278 .225 .253 -.124 -.176 -.075 .18 .164 .182 .234 .17 .172 .077 .031 .055 -.078 -.027 -.017 1 .142 .145 .224 .081
.224 .223 .227 .273 .667 .561 .579 .484 .149 .331 .323 .347 -.071 -.093 -.029 .471 .444 .399 .36 .329 .489 .456 .344 .491 .193 .248 -.025 .186 .904 .934 .917 -.054
Note: item loadings above .70 are highlighted
62
Utilitarian attitude
Perceived effort
Perceived usability
Perceived usage
Henodic attitude
Cust sat
Apps
Appendix H: Statistical validity of the base model
For the estimation of the significance of the parameters, bootstrapping was applied with a resample of 500. Displayed here are the supported hypotheses of the base model and the effect sizes of changes in R2 due to the deletion of insignificant effects.
Table 15 Hypotheses in the base model
Hypothesis H1b,alt (+) H1c,alt (-) H1d (+) H2 (+) H3a (-) H4b (+) H4c (+) H4f (+) H4f (+) H5a (+) H5c (+) H5e (-) H6f (+) H7 (+) Relation Apps ? Perceived Usability Apps ? Perceived Effort Apps ? Cust. Sat. Smartphone Perceived Usability ? Perceived Usefulness Perceived Usefulness ? SMS Usage Perceived Usefulness ? Frequency Of Use Perceived Usefulness ? Perceived Usage Level Perceived Usability ? Perceived Usage Level Perceived Usability ? Perceived Usage Level Perceived Usefulness ? Cust. Sat. Smartphone Perceived Usability ? Cust. Sat. Smartphone Perceived Effort ? Cust. Sat. Smartphone Frequency Of Use ? Mobile Data Usage Cust. Sat. Smartphone ? Telco Cust. Sat. ?† base model .35 -.2 .09 .54 -.22 .22 .34 .36 .36 .2 .44 -.12 .19 .29 T-value base model 6.41*** 2.97*** 2.3** 10.09*** 2.69*** 2.55** 3.78*** 4.08*** 4.08*** 2.68*** 5.86*** 2.14** 2.4** 3.66*** ?† reduced model .35 -.2 .09 .57 -.13 .3 .34 .13 .36 .2 .44 -.12 .31 .31 T-value reduced model 6.67*** 3.38*** 2.26** 11.35*** 2.34** 4.97*** 4.03*** 2.87*** 4.61*** 2.8*** 5.8*** 2.12** 5.39*** 4.77***
*
<.1. **<.05, ***<.01 (Two Tailed); † Effect Sizes Are StandardizedTable 16 Effect sizes of changes in R2
Variable Cust sat smartphone Daily usage time Frequency of use Mobile data usage Perceived Effort Perceived usability Perceived usefulness Perceived usage SMS usage Telco cust sat Voice usage R2 base model .44 .05 .10 .14 .04 .12 .33 .35 .04 .12 .01 R2 reduced model .44 Variable deleted .09 .09 .04 .12 .32 .35 .02 .10 Variable deleted ƒ2 0 -.01 -.05 0 0 -.01 0 -.02 -.02 -
63
Appendix I: Post-hoc Mediation analysis
Table 17 Coefficients and significance of the effects
Base model without insignificant effects Relation Apps ? Cust. Sat. Smartphone Apps ? Perceived Effort Apps ? Perceived Usability Cust. Sat. Smartphone ? Telco cust. Sat. Frequency Of Use ? Mobile Data Usage Perceived Effort ? Cust. Sat. Smartphone Perceived Effort ? Perceived Usage Level Perceived Usability ? Cust. Sat. Smartphone Perceived Usability ? Perceived Usefulness Perceived Usability ? Perceived Usage Level Perceived Usefulness ? Cust. Sat. Smartphone Perceived Usefulness ? Frequency Of Use Perceived Usefulness ? Perceived Usage Level Perceived Usefulness ? Sms Usage ?
†
Mediator model T-value 2.23** 3.19** 6.55*** 4.72*** 5.39*** 2.2** 2.75*** 5.99*** 11.83*** 4.29*** 2.93*** 4.88*** 4.04*** 2.14** Relation Apps ? Cust. Sat. Smartphone Apps ? Hedonic Attitude Apps ? Perceived Brand Image Apps ? Perceived Effort Apps ? Perceived Usability Apps ? Subjective Knowledge Apps ? Utilitarian Attitude Cust. Sat. Smartphone ? Telco cust. Sat. Frequency Of Use ? Mobile Data Usage Hedonic Attitude ? Perceived Effort Hedonic Attitude ? Perceived Usability Perceived Brand Image ? Perceived Effort Perceived Brand Image ? Perceived Usability Perceived Effort ? Cust. Sat. Smartphone Perceived Effort ? Perceived Usage Level Perceived Usability ? Cust. Sat. Smartphone Perceived Usability ? Perceived Usefulness Perceived Usability ? Perceived Usage Level Perceived Usefulness ? Cust. Sat. Smartphone Perceived Usefulness ? Frequency Of Use Perceived Usefulness ? Perceived Usage Level Perceived Usefulness ? Sms Usage Subjective Knowledge ? Perceived Effort Subjective Knowledge ? Perceived Usability Utilitarian Attitude ? Perceived Effort Utilitarian Attitude ? Perceived Usability ?† .09 .17 .25 -.17 .14 .30 .22 .31 .31 -.04 .00 -.25 .50 -.12 .13 .44 .57 .36 .20 .30 .34 -.13 .08 .11 .03 .23 T-value 2.25** 3.01*** 3.92*** 2.53** 3.12*** 5.09*** 4.01*** 4.76*** 5.37*** 0.48 0.07 3.22*** 10.08*** 2.16** 2.67*** 6.01*** 11.5*** 4.25*** 2.63*** 5.02*** 3.91*** 2.17** 1.12 1.79 0.37 3.43***
.09 -.2 .35 .31 .31 -.12 .13 .44 .57 .36 .20 .30 .34 -.13
**: p < .05; ***: p< .01 (two-tailed). † Effect Sizes Are Standardized
Table 18 Calculation of the significance of the mediating effects
? Apps ? Mediator Mediating Relation Apps ? Perceived Brand Image ? Perceived Effort Apps ? Perceived Brand Image ? Perceived Usability Apps ? Utilitarian Attitude ? Perceived Usability **p< .05 Standardized .246 .246 .224 Unstandardized .126 .126 .069 Std. Dev. 2.61 1.34 2.93 1.13 0.80 Std. Error .0628 .0628 .056 ? Mediator ? Dependent Variable Standardized -.246 .498 .228 Unstandardized -.537 .420 .322 Std. Error .0764 .0494 .0665 Sobel ZValue -1.93 1.96** 1.19
Variable Apps Perceived Brand Image Perceived Effort Perceived Usability Utilitarian Attitude
See footnotes 2 and 3 for the formulae for the unstandardized coefficients and Sobel value.
64
Appendix J: MANOVA Post-hoc analysis
For the post-hoc analysis, Games-Howell tests were used. Only the tests for the variables that proved to contain significant differences between the operating systems, based on the between-subject effects, are displayed.
Table 19 Post-hoc comparisons between groups
Dependent Variable Apps Reference Category iOS Comparing Category Mean Difference Std. Error Sig.
BOS 2.79 0.46 .0000* Android 0.46 0.42 .5118 BOS iOS -2.79 0.46 .0000* Android -2.32 0.35 .0000* Android iOS -0.46 0.42 .5118 BOS 2.32 0.35 .0000* Cust. Sat. iOS BOS 2.99 0.52 .0000* Smartphone Android 1.42 0.21 .0000* BOS iOS -2.99 0.52 .0000* Android -1.56 0.52 .0119 Android iOS -1.42 0.21 .0000* BOS 1.56 0.52 .0119 Duration of daily iOS BOS -0.93 0.36 .0302 use Android 0.12 0.25 .8752 BOS iOS 0.93 0.36 .0302 Android 1.05 0.30 .0027 Android iOS 0.12 0.25 .8752 BOS -1.05 0.30 .0027 *Significant at the .05/54 = .0009 level. Such a strict level is used since there are 3 categories being compared across 9 variables (displayed here are three of the nine compared variables), leading to 3*2*1*9=54 comparisons.
65
doc_387751616.pdf