Exploring consumer preferences in cloud archiving ...

Behaviour & Information Technology, 2015 http://dx.doi.org/10.1080/0144929X.2015.1012650

Exploring consumer preferences in cloud archiving – a student’s perspective Daniel Burda ∗ and Frank Teuteberg Institute of Information Management and Information Systems, University of Osnabrueck, Katharinenstrasse 1, 49069 Osnabrueck, Germany

Downloaded by [Daniel Burda] at 23:01 02 March 2015

(Received 13 September 2014; accepted 7 January 2015 ) Cloud storage has seen an increasing rise in demand and diffusion. Consequently, the cloud storage market is also becoming an increasingly commoditised market. That is, homogenous products are offered at equal prices, and this offer makes it more difficult for cloud storage providers to generate revenue and differentiate themselves from their competitors. Therefore, it is vital for providers to precisely understand customer preferences so that these can be targeted with appropriate services. To examine these preferences, we conduct a choice experiment and analyse choice decisions gathered from 340 German students by means of a conjoint analysis. We perform an individual-level analysis of preferences, which reveals significant differences and heterogeneity within the sample. By using a subsequent cluster analysis, we identify three distinct customer segments that also show significant differences in, for example, the perceptions of information privacy and risks. Our findings contribute to the literature by uncovering the preference structure and trade-offs that users make in their choice of storage services when employed for the purpose of archiving. We conclude the study with a discussion of practical implications that can aid cloud storage providers in service design decisions, and highlight the limitations associated with our research approach and drawn sample. Keywords: cloud storage; cloud archiving; conjoint analysis; discrete choice experiment; consumer preferences

1. Introduction Preserving digital data for the long-term is a challenging task in the light of rapidly changing technologies and the associated risk of media degradation and obsolete soft/hardware (Burda and Teuteberg 2013). In the private domain, valuable personal files, such as photographs, documents, music or any other file, are still mainly archived on conventional media such as local, external hard disks or DVDs (Ion et al. 2011). From a user perspective, these files are often irreplaceable memories which, in case of loss or failure, cannot be bought back. Nevertheless, hard drives will fail eventually, usually happening at random and resulting in a loss of files (Top 2013). To encounter some of these threats, consumer cloud storage solutions provide adequate means and have seen an increasing rise in demand and diffusion (Marshall and Tang 2012) that is still expected to grow rapidly in future according to recent market research (Gillett et al. 2011; Verma 2012). By using on-demand storage services from a shared pool of highly available and reliable computing resources, users can remotely archive their data in an easy-to-use manner without the burden of local data storage (Armbrust et al. 2010; Wang et al. 2013). Archiving in the cloud, referred to as cloud archiving in this study, offers several advantages compared to archiving on conventional media. It provides

*Corresponding author. Email: [email protected]

© 2015 Taylor & Francis

central and continuous availability of archived data that can be accessed simultaneously from various devices such as laptops or mobile devices anytime and anywhere. Moreover, long-term data access can be preserved without the threat of media obsolescence that often forces users to periodically replicate their data onto newer storage media. However, despite the advantages of cloud storage compared to conventional storage media, prior research indicates that users do not take full advantage of cloud storage services. Moreover, there is a general tendency of end-users to prefer local storage over cloud storage, particularly for data considered sensitive (Ion et al. 2011; Marshall and Tang 2012). In an effort to investigate this phenomena, other works concentrate on examining users’ continuance intention of cloud services including factors such as trust, information security/privacy and risk perceptions (Trenz, Huntgeburth, and Veit 2013; Loske et al. 2014; Walterbusch and Teuteberg 2014). However, existing research has failed to provide an understanding of the preferences end-users have in their choice of cloud storage services. In an overall effort to attract new customers and increase market shares, understanding these preferences and the underlying value structure is arguably imperative for cloud storage providers to be able to respond to these needs more effectively than their competitors (Allenby and


2

D. Burda and F. Teuteberg

Ginter 1995). With such understanding, providers could, for instance, design services and direct these to the individuals who are most likely to respond favourably. This in turn could cause an immediate and strong market response (Allenby and Rossi 2006), foster the conversion of users into paying customers and contribute to coping with the increasing degree of commoditisation in the cloud storage market, a market in which services are offered with almost identical features at equal prices and low switching costs for end-users (Durkee 2010; Nielsen 2012). To fill this research gap, this study examines the preferences of end-users when choosing cloud storage services for the purpose of personal archiving. Based on a literature review and interviews, we identify five important product attributes: encryption, customer support, accessibility, storage capacity and price. We collect data using a discrete choice experiment and conduct a conjoint analysis, which is an accepted approach to measure customer preferences for specific product attributes and to elicit their relative importance (RI) (Green and Srinivasan 1990; Vriens 1994). Based on the conjoint results, we segment participants into three clusters and elaborate their unique characteristics. The findings of this study can provide guidance for cloud storage providers as to which service attributes to place particular focus on when implementing new features or devising marketing campaigns. This paper is structured as follows. First, we review previous research. In Section 3, we elaborate our research design before we present the results of the data analysis in Section 4. Subsequently, in Section 5, we discuss important findings, theoretical contributions, practical implications, limitations of this research as well as future research opportunities. 2.

Previous research

At the beginning of this study, we examined existing research to determine factors that are considered relevant in the users’ cloud provider selection decisions. Therefore, we primarily focused on literature that studies cloud adoption from an individual end-user perspective and that explores factors influencing users’ decision-making in cloud computing. Many previous studies, examining cloud computing from an individual user perspective, focus on important determiners and inhibitors of cloud computing by applying commonly used IT adoption theories such as the technology acceptance model. For instance, Bhattacherjee and Park (2013) investigate the motivation of end-users to migrate from client-hosted computing to cloud computing. They conclude that despite the benefits cloud services offer, users are averse to migrate to the cloud because of concerns about security and privacy. Ion et al. (2011) empirically investigate users’ perceptions and privacy concerns with cloud storage providers. Based on interview and survey data collected from 402 participants, Ion et al.

observe that 69% of all respondents prefer local storage over cloud storage and that they do not use cloud storage as their main storage medium for security and privacy reasons. Li and Chang (2012) examine user acceptance of cloud applications and find that privacy concerns negatively affect risk perceptions and in turn attitude towards cloud applications (see also, Ambrose and Chiravuri 2010; Sturm, Lansing, and Sunyaev 2014). The authors recommend that data encryption mechanisms may be implemented by cloud providers to adequately address these concerns (see also, Itani, Kayssi, and Chehab 2009; Armbrust et al. 2010; Zhiyun, Meina, and Junde 2010; Trenz and Huntgeburth 2014; Walterbusch and Teuteberg 2014). In this line of thinking, Benlian, Koufaris, and Hess (2011) conceptualise security including data encryption as one of six constituents of their software as a service quality (SaaS-Qual) measurement instrument. In summary, information privacy/security concerns, uncertainty and risk perceptions are found to be major deterrents to cloud adoption and received significant attention in recent research (Benlian, Hess, and Buxmann 2009; Martens, Poeppelbuss, and Teuteberg 2011; Lee, Tang, and Sugumaran 2012; see, e.g. Ackermann et al. 2013; Trenz, Huntgeburth, and Veit 2013; Amiri, Cavusoglu, and Benbasat 2014; Ermakova, Fabian, and Zarnekow 2014). For instance, Loske et al. (2014) find that perceived IT security risks increase the perception of risks regarding a cloud service, which in turn negatively affects cloud adoption. Another factor of the SaaS-Qual instrument refers to the provider’s ability to offer knowledgeable, caring and courteous customer support which significantly influences a user’s perception of overall service quality. Mirroring this finding, Koehler et al. (2010) identify customer support as an important attribute of cloud services using a conjoint analysis approach, and Martens, Teuteberg, and Gräuler (2011) present a cloud computing service maturity model including customer support as an important requirement (see also, Repschlaeger et al. 2012; Lansing, Schneider, and Sunyaev 2013). In another conjoint study, Giessmann and Stanoevska (2012) study end-user preferences in platform as a service (PaaS) solutions. The authors find that consumers clearly value having mobile device access to a cloud platform. Bhattacherjee and Park (2013) conclude that universal accessibility (termed omnipresence in their study) is more salient than other traditional motivations of IT acceptance (such as usefulness of IT usage) in cloud computing. In fact, universal accessibility constitutes one of the main benefits of cloud computing allowing to access data independent of geographical locations and devices (Wang et al. 2013). Besides the above factors, additional drivers and inhibitors in cloud computing are examined, for example, the role of trust, social agents and cloud service certifications (Lee, Tang, and Sugumaran 2012; Lansing, Schneider, and Sunyaev 2013; Walterbusch, Martens, and Teuteberg 2013; Sturm, Lansing, and Sunyaev 2014; Walter et al. 2014),


Behaviour & Information Technology pre-existing attitude towards SaaS-adoption (Benlian, Hess, and Buxmann 2009) or personal IT innovativeness (Behrend et al. 2011). Despite the vast amount of previous work, there is still a lack of research investigating consumer preferences in the context of cloud storage employed as a means of personal archiving. In an effort to address this gap, we conduct a choice experiment to explore the underlying benefit structure of end-users in cloud storage adoption decisions. This approach allows us to uncover end-user preferences and to analyse the RI of various service attributes in the selection of cloud storage providers. We note that this paper presents a substantial extension of previous work by the authors that has been presented at the European Conference on Information Systems (see Burda and Teuteberg 2014). Based on the conference feedback gained and additional analyses conducted, this paper provides a significant enhancement of the previous publication including a more detailed elaboration of our research approach, several additional analyses regarding the cluster’s characteristics and graphical presentations of our results. For instance, we incorporated results of a detailed measurement model validation (Table 7) as well as additional inter-cluster analyses (see Table 6 or Table 8) in terms of the utility-level changes and latent variables per cluster. Based on these additional analyses and findings, we extended the discussion of our results in Section 5 providing more insights and highlighting important practical implications of our work. 3. Research approach In this study, we employ a choice-based conjoint analysis which is a general approach for the analysis of consumer preferences and often used by marketing managers to gain insights into consumers’ preferences for multi-attributed products and services (Green and Srinivasan 1990; Sattler and Hartmann 2008). In conjoint analyses, products and services are viewed as a bundle of specific characteristics (attributes) which are defined by a number of different values (levels). Based on this set of attributes and respective levels, different product alternatives (stimuli) can be defined. Respondents are then asked to evaluate these alternatives in terms of preference or attraction in an experimental setting (Vriens 1994). It is expected that participants pick the choice option that provides them with the highest utility. Based on the respondents’ choices, the overall utilities of the different product alternatives can be decomposed so that the underlying value system and RI of product attributes become obvious. 3.1.

Identification of product attributes and attribute levels To identify relevant attributes in cloud storage adoption decisions and their respective levels, we first reviewed extant research about cloud computing and particularly cloud storage adoption. Subsequently, following Green

3

and Srinivasan (1978), we conducted nine semi-structured interviews with current cloud storage users to determine those attributes that are most frequently regarded as relevant and thus may be considered as important drivers of utility of cloud storage services. In the interviews, seven out of nine participants mentioned their information security and privacy concerns and stressed the importance of encryption and security mechanisms employed by the cloud storage provider as pivotal drivers of utility. As described above, this finding is consistent with extant cloud computing research that considers information privacy concerns as a significant inhibitor of cloud adoption (Ambrose and Chiravuri 2010; Trenz, Huntgeburth, and Veit 2013; Ermakova, Fabian, and Zarnekow 2014). Thus, we included encryption as attribute in this study. Mirroring the empirical findings from Walterbusch and Teuteberg (2014), the price for using the storage service has been mentioned by eight out of nine interviewees as an important attribute influencing their choice of cloud storage services. As a result, we included price as an attribute in this study and conceptualised it as a monthly flat rate. To date, flat rates can be conceived as the prevailing pricing model in the cloud storage market (Top 2013). Moreover, the participants mentioned storage space as an important attribute of cloud storage offers. For instance, a respondent stated: ‘The more space you get, the better it is as my data collection is constantly growing’. In line with the findings of recent studies on cloud adoption (Bhattacherjee and Park 2013), accessibility has been mentioned by seven participants who emphasised their desire to access cloud storage seamlessly and easily with all their devices (e.g. desktop computer, tablet, smartphone). Therefore, storage space and accessibility were included as attributes in this study. In addition to these four attributes, we included customer support. While customer support received some attention in the cloud decisionmaking context (Benlian, Koufaris, and Hess 2011; Lansing, Schneider, and Sunyaev 2013), it was not mentioned by our interviewees. However, previous studies suggest to include all attributes that might be important to the respondent (customer) and relevant to the cloud storage provider (Bridges et al. 2011). Even if customers do not value customer support, the knowledge of that fact is of importance for the provider since there is no need for him to place focus on. For the definition of the attribute levels, we used the interview data and analysed current cloud storage services concerning the selected attributes and their specific offerings. Therefore, we collected information from provider websites and websites providing reviews and comparisons of cloud storage offerings (see, e.g. Cloud Storage Reviews 2013; Top 2013). All levels were selected based on their prevalent use so that the level-specific utility of commonly provided storage service configurations can be assessed. By this means, we derived the numerous

4

D. Burda and F. Teuteberg Table 1.

Attributes and attribute levels used.

Attribute

Attribute description

Attribute levels

Price

Monthly price for using the cloud storage Offered storage space The technological possibilities of accessing the cloud storage

(1) 0.00 Euro per month; (2) 5.00 Euro per month; (3) 10.00 Euro per month


Storage space Accessibility

(1) 5 GB; (2) 50 GB; (3) 100 GB (1) Website only: The cloud storage can only be accessed after successful login on the provider’s website (2) Website + software (PC/Mac, Linux, etc.): In addition to the website access, software is provided with which the cloud storage can be accessed in a way that is comparable to accessing local directories/files on the computer. Once directories/files are added/changed, they are automatically synchronised with the cloud storage (3) Website + software + mobile: In addition to the access provided via website and software, an app for mobile devices such as tablets and smartphones is provided. All features of the cloud storage service can be accessed using this app Encryption The encryption methods (1) Server-side encryption: Before your data are uploaded, a secure connection (e.g. employed by the cloud secure sockets layer (SSL)) between your computer and the storage provider storage provider is established so that no one can wiretap the transfer. The transferred data are stored on encrypted disks so that unauthorised access through third parties is not possible. However, your data are visible to the provider and its staff (2) Client-side encryption: Before your data are uploaded to the cloud storage via a secure connection (e.g. SSL), it is encrypted locally on your device. No one, including the provider’s staff, unless explicitly authorised by you, can see your data Customer support The way customer service (1) Basic: Customer support is exclusively provided in the form of comprehensive is provided by the cloud FAQs, documentations and video tutorials. There is no support staff who can be storage provider contacted by the customer (2) Enhanced: Customer support is e-mail oriented. Customers can contact support staff via e-mail or contact forms provided on the website (3) Live: Customers get live phone or online chat support that is provided 24/7

attribute levels which should not, like the attributes themselves, be considered to be exhaustive. Table 1 describes the five selected attributes including their corresponding levels.

3.2. Experiment design The survey contained major parts. In the first part, respondents were asked to assess product alternatives. In the second part, they were asked to answer a set of questions measuring various latent variables and to provide their demographic information. Before the assessment of the product alternatives, we instructed the participants accurately; all participants were briefed that they would run through several choice scenarios with each of the scenarios showing two cloud storage services. They were further instructed that each presented service may vary in all of the five attributes, which we then described including the respective levels as shown in Table 1. Finally, we requested the participants to imagine the following while assessing the choice sets: (1) they were currently not using cloud storage and (2) they were currently in the process of finding an adequate cloud storage provider who would archive their personal files (e.g. photographs, music or documents) most appropriately. After the descriptions were shown, the respondents were asked to take decisions in 17

choice scenarios that followed a full profile design (Wittink and Cattin 1989). All of the detailed descriptions of the attributes provided in Table 1 could be displayed to the participants throughout the survey upon their request. In order to make the choices more realistic, we included the ‘nochoice’ option in each scenario (Haaijer, Kamakura, and Wedel 2001). The choice scenarios were created using the software environment for statistical computing R (R Core Team 2013). In R, we used the package Algorithmic Experimental Design (Wheeler 2011) to compute a fractional factorial design from our full factorial design with 162 possible stimuli (3 × 3 × 3 × 3 × 2). To this end, we followed a five-step procedure presented by Aizaki and Nishimura (2008) and generated a fractional factorial design comprising 15 different stimuli which were arbitrarily pooled to form the choice scenarios. Moreover, we created two additional choice scenarios which were used as hold-out sets. These hold-out sets were answered by all respondents, but not used for estimating the utilities. Instead, we used them to evaluate the predictive validity and quality of the responses by comparing how accurately the estimated utilities predict choices from the hold-out tasks (Orme, Alpert, and Christensen 1997). To convert this final design for data analysis, we used effects coding (Hensher, Rose, and Greene 2005).


Behaviour & Information Technology It should be noted that there is no clear guidance about how many choice sets should be presented to each respondent in choice experiments. Whereas some authors argue that there is a decline in choice consistency and rise in error variance along with an increasing number of choice sets displayed per respondent (Bradley and Daly 1994), others did not find such a tendency and refer to learning effects that occur with a growing number of choice sets (for a detailed discussion, see, Savage and Waldman 2008; Hess, Hensher, and Daly 2012). That is, there is no generally accepted threshold for the maximum number of choice sets that should be displayed. However, Bech, Kjaer, and Lauridsen (2011) found that respondents are capable of managing 17 choice sets without problems, and a metastudy on the commercial use of conjoint analysis found a median value of 16 choice sets in a typical conjoint design (Wittink and Cattin 1989). The resulting individual attribute utility values from conjoint analysis are often subjected to a consecutive cluster analysis which provides an ancillary analysis technique that aims to segment respondents into groups with comparable utility structures (Green, Krieger, and Wind 2001). To better describe and discriminate those segments and yield a more comprehensive picture of the segments and their underlying differences, additional variables should be explored (Allenby and Ginter 1995). We consequently gathered data on user characteristics thereby concentrating on variables that are deemed important in the context of cloud acceptance according to existing research (see Section 2) and/or that saliently emerged from our interviews (e.g. innovativeness). The selected constructs are presented in Table 2. All constructs in our model are operationalised as reflective and refer to a person’s perception, thought or feeling regarding the specific properties

5

under investigation (MacKenzie, Podsakoff, and Podsakoff 2011). All constructs were measured with multiple items on five-point Likert scales and extant psychometric instruments which we adapted to our research context. 3.3. Data collection The survey was implemented with the survey application Limesurvey (Limesurvey 2013) based on the choice sets created with R in the previous step. An example of 1 of the 17 choice scenarios all participants had to take is shown in Figure 1. The online survey was reviewed by three research colleagues and revised accordingly before it was subjected to a small-scale pretest which was conducted with a convenience sample of 12 respondents randomly drawn from faculty members and graduate students. The respondents completed the questionnaire and reported their feedback on the wording, length and concerns if they had any. Based on their feedback, we further adjusted some of the attribute descriptions and measurement items. Following a convenience sampling strategy, we invited students from our university via e-mail to participate in the study and posted an announcement in the university’s internal online portal accessible to all university members. The university is located in northern Germany and combines different scientific disciplines in teaching and research, such as law, business administration/economics and social sciences. In an effort to increase the response rate, we decided to employ a lottery which is deemed to exert positive influence on web survey response rates (Heerwegh 2006). In the invitation, we informed the potential participants that, if they responded to the survey, they would be entered in a prize draw of five 50 Euro Amazon gift certificates.

Table 2. Operationalisation of constructs and measurement items. Construct

Operationalisation

Items

Disposition to trust

Disposition to trust is a general inclination to display faith in humanity and to adopt a trusting stance towards others (Gefen 2000) Refers to the degree to which the consumer focuses exclusively on paying low prices (Lichtenstein, Ridgway, and Netemeyer 1993) Refers to an individual’s subjective views of fairness within the context of information privacy (Malhotra, Kim, and Agarwal 2004) Refers to the expectation that a high potential for loss is associated with the usage of cloud storage (Malhotra, Kim, and Agarwal 2004) The willingness of an individual to try out any new information technology (Agarwal and Prasad 1998) The degree to which a user has a favourable evaluation of using cloud storage as a means of personal archiving (Ajzen 1991) The readiness to engage in a particular behaviour (Fishbein 2008)

Gefen (2000)

Price consciousness Information privacy concern Risk beliefs Personal IT innovativeness Attitude towards cloud archiving Intention to use

Lichtenstein, Ridgway, and Netemeyer (1993) Malhotra, Kim, and Agarwal (2004) Malhotra, Kim, and Agarwal (2004) Agarwal and Prasad (1998) Jarvenpaa, Tractinsky, and Vitale (2000) Gefen, Karahanna, and Straub (2003)


6


Figure 1. Example of a choice scenario.

Although the use of student samples is not free of criticism, we believe that it is an adequate target sample for our study due to the following reasons. First, prior studies indicate a high degree of cloud storage adoption among students (Ion et al. 2011). Thus, it is reasonable to expect that students constitute a significant part of the target population (Compeau et al. 2012). Second, prior online behaviour research found that a student sample can anticipate the direction in which the general population is moving since it usually represents experienced users and early adopters of an innovation like cloud computing (Gallagher, Parsons, and Foster 2001). Third, we deem students to benefit from universal data accessibility provided by cloud storage more than others. This is because students usually require access to their data from different devices and places such as the university campus, their residence during the semester and semester break. The data collection took place in October 2013 and the survey was made available online for 10 days. A reminder

e-mail was sent out five days after the initial invitation e-mail. We collected 357 completed questionnaires out of which we excluded 17 respondents from our analysis due to extremely low completion times and hit rate. Eventually, a sample of 340 usable and completed questionnaires was used in the data analysis. On average, it took the respondents 12.43 minutes to complete the survey. Moreover, a possible non-response bias was addressed by adopting the procedure recommended by Armstrong and Overton (1977). We tested for mean differences in utility estimates and construct items between the respondents’ data gained prior to and after the reminder that was sent out after five days. The test revealed no significant differences, so we concluded that non-response bias was not an issue in this study. Table 3 provides an overview of the final sample that was used in subsequent analysis. The table shows that 74.1% of the respondents were in the 18–24 age range, 90.3% were students and 49.8% rated their computer

Table 3. Profile of respondents (n = 340). Gender Age Level of education Occupation

Female: 109 (32.1%) 18–24 years: 252 (74.1%) Some college degree: 10 (2.9%) Student: 307 (90.3%)

Male: 231 (67.9%) 25–34 years: 81 (23.8%)

High school degree: 231 (68.1%)

35–44 years: 6 (1.8%)

45–54 years: 1 (0.3%)

University degree: 88 (25.9%)

Other: 11 (3.2%)

Official: 2 (0.6%)

Other: 3 (0.9%)

Employee: 28 (8.2%)

Number of devices used

1 device: 12 (3.5%)

2 devices: 103 (30.3%)

3 devices: 140 (41.2%)

4 devices: 56 (16.5%)

>4 devices: 29 (8.5%)

Computer proficiency (self-reported measure)

Very poor: 0 (0.0%)

Poor: 5 (1.5%)

Intermediate: 56 (16.5%)

Good: 169 (49.8%)

Excellent: 110 (32.4%)

Behaviour & Information Technology


proficiency (self-reported) as good. Furthermore, 71.5% of the respondents were cloud storage users out of which 4.5% paid 4.51 Euro per month on average for cloud storage services such as Google Drive or Dropbox. 4. Data analysis We used the R package bayesm (Rossi 2012) for data analysis which provides methods for analysing choice decisions with hierarchical Bayes (HB) models.1 Using HB estimation is the generally preferred method for analysing choice decisions as it accounts for the fact that consumers have heterogeneous preferences regarding product-specific attributes and therefore should not be treated alike (Rossi and Allenby 2003). In contrast to other approaches, HB methods facilitate the calculation of parameter estimates associated with specific respondents, that is, individuallevel part-worth utilities. These individual-level utilities lead to a better understanding of market structure and provide a basis for subsequent clustering of customers into distinct segments (Vriens 1994; Allenby and Rossi 2006). The analysis procedure in this study involved three major stages and is inspired by the approach described by Krasnova, Hildebrand, and Guenther (2009). In a first step, we calculated the individual and aggregated utilities as well as RI which denote the attractiveness of a specific attribute level and the weight each attribute carries in a user’s provider selection decision. Based on the aggregate part-worth utilities, we calculated the utility change between the various attribute levels as well as corresponding Euro values for each attribute-level change. These steps provide us with a deeper understanding of the underlying value structure and trade-offs users may consider as well as how preferences can be translated into monetary value (Krasnova, Hildebrand, and Guenther 2009). In a second step, we performed a cluster analysis based on the individual part-worth utilities to detect similar utility structures across users. Finally, we analysed these clusters regarding differences in utilities, utility-level changes, RI of attributes and additional variables collected. 4.1. Conjoint results To assess the validity of our estimates and the goodness of fit of the estimated model, we conducted two tests. First, we performed a likelihood ratio (LR) test that measures how well the model and its estimated parameters perform compared with a model in which all the parameters are zero, which is equivalent to having no model (Train 2003). The test indicates that the estimated model is statistically valid with LR = 28.27 (df = 10, p < .01) exceeding the critical value of χ 2 . That is, the null hypothesis that the estimated model and zero model are equal can be rejected. This finding is also supported by a hit rate of 95.2% on the 15 choice sets compared to a 33% hit rate in the case of pure chance (no model). Second, we calculated the hit rate

7

on the hold-out sets to assess the predictive validity of our model by identifying the alternative with the highest probability in the two hold-out sets for each respondent (for a discussion on the use of hold-out tasks see, e.g. Orme, Alpert, and Christensen 1997). This was done on the basis of the estimated model and by determining whether or not this was the alternative the respondent actually chose (Train 2003). We find that our model correctly predicted 92.1% of the actual choices, thereby indicating an adequate degree of predictive validity (Gensler 2006). Table 4 shows the estimated part-worth utilities for the respective attribute levels. Considering the results in Table 4, we see that the estimated parameter relations agree with our a priori expectations and hence indicate face validity (Green and Srinivasan 1978). Considering the estimated part-worth utilities, we find that the optimal cloud storage service is free (max(4.865, − 0.929, − 3.936) = 4.865), offers 100 GB storage (max( − 2.679, 0.718, 1.961) = 1.961) and cloud storage access through website, software and mobile applications (max( − 1.316, 0.194, 1.122) = 1.122). Furthermore, it provides client-side encryption mechanisms (max( − 0.498, 0.498) = 0.498) and enhanced support (max( − 0.466, 0.251, 0.216) = 0.251). Based on the estimated part-worth utilities, we calculated the RI of each attribute, which reflects the ranking of an attribute in cloud storage selection decisions. The results are shown in Figure 2 and indicate that price is the most important factor with an RI of 44.45% followed by storage space which is the second most important attribute with an RI of 23.60%. Accessibility is the third most important variable with an RI of 15.14% followed by encryption and customer support with a RI of 10.23% and 6.59%, respectively. To better understand the users’ value system and the trade-offs they make, we further calculated the utility change between the various attribute levels as can be seen

Table 4.

Aggregated part-worth utilities of attribute levels.

Product attribute Price Storage space Accessibility Encryption Customer support

Product attribute level 0 Euro 5 Euro 10 Euro 5 GB 50 GB 100 GB Website only Website + software Website + software + mobile Server-side encryption Client-side encryption Basic Enhanced Live

Part-worth utility 4.865 − 0.929 − 3.936 − 2.679 0.718 1.961 − 1.316 0.194 1.122 − 0.498 0.498 − 0.466 0.251 0.216

8



Figure 2. Aggregated RI of product attributes.

from Table 5. The table illustrates the change in utility when the cloud storage provider modifies his offering from one level of an attribute to another one. The fourth column of Table 5 shows the p-value of the dependent samples t-test, which tests whether the compared levels provide different utility for the cloud storage users while the last column depicts the Euro equivalent of the level change. Acknowledging the results in Table 5, it becomes obvious that users perceive a significant (p-value in column 4) utility drop for a change in the price from 0 to 5 Euro which is almost twice as high as a change from 5 to 10 Euro. In terms of storage space, we find that the 100 GB option provides users with the highest utility. However, the results also indicate a diminishing marginal utility: a change from 5 to 50 GB storage exhibits a bigger utility increase (u = 3.396) than a change from 50 to 100 GB (u = 1.243). Although being the second most important attribute in cloud storage choice, the dominant role of price becomes obvious when comparing the utility changes between different price and storage space levels. For example, if a provider decided to increase the price from 0 to 5 Euro per month, this drop (u = − 5.794 utility units) could not be compensated by increasing the provided storage space from 5 to 100 GB (u = 4.639 utility units) with otherwise unchanged conditions. The estimates for the attribute accessibility show that all attribute levels provide

significantly different utilities to the users (as indicated by the p-values in Table 5) and that the utility is higher, the more ways of access to the cloud storage are provided. As shown in Table 5, users attach, although only little, value to client-side encryption compared to server-side encryption. Regarding customer support, utility is higher when enhanced support is provided instead of basic support. In contrast, users do not perceive a significant difference in utility between enhanced and live support as indicated by the p-value of .29. Based on the estimated utility changes for the price attribute, the Euro equivalent of a change in the levels of all other attributes under study can be derived (Krasnova, Hildebrand, and Guenther 2009). Table 5 exhibits that a change in utility from 5 to 0 Euro corresponds to a change of 5.794/5 = 1.159 units of final utility per Euro while a change from 10 to 5 Euro corresponds to 3.006/5 = 0.601 units. These values reflect an upper and a lower bound for the utility change per Euro, which may be used to calculate the Euro equivalent of a change in the levels of the other attributes as shown in Table 5 (last column). We find that, for example, users are willing to pay between 2.93 and 5.65 Euros for a shift from 5 to 50 GB storage and between 1.07 and 2.07 Euros for a change from 50 to 100 GB storage. Likewise, providing access to the cloud storage via website and additional software is worth between 1.30 and 2.51 Euros per month while user would pay between 0.80 and 1.54 Euros for additional mobile access. Clientside encryption is valued the equivalent of between 0.86 and 1.66 Euros per month. Finally, in terms of customer support, users would pay between 0.62 and 1.19 Euros on average for a change from basic to enhanced support, whereas a change from enhanced to live support is even negatively valued by the respondents between − 0.06 and − 0.03 Euros.

4.2. Cluster analysis To segment respondents into groups based on the similarity of their preferences, we conducted a hierarchical

Table 5. Utility change in attribute levels and equivalent euro value of change.

Attribute

Attribute-level change

Price

0 Euro → 5 Euro 5 Euro → 10 Euro 5 GB → 50 GB 50 GB → 100 GB Website only → website + software Website + software → website + software + mobile Server-side → client-side Basic → enhanced Enhanced → live

Storage space Accessibility Encryption Customer support

Utility delta (u)

p-Value (t-test)

Euro equivalent of level change (Bound 1) – (Bound 2)

− 5.794 − 3.006 3.396 1.243 1.510 0.928 0.996 0.717 − 0.035

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.29

(2.93) – (5.65) (1.07) – (2.07) (1.30) – (2.51) (0.80) – (1.54) (0.86) – (1.66) (0.62) – (1.19) ( − 0.03) – ( − 0.06)

Behaviour & Information Technology agglomerative cluster analysis based on the individual part-worth utilities. Towards this end, we applied Ward’s method which aims at identifying compact clusters by minimising the within-cluster variance (Punj and Stewart 1983). Since cluster analysis is not a statistical inference method that provides one single correct number of clusters, we tested several alternative solutions (Rokka and Uusitalo 2008) and finally found a three cluster solution with C1 = 57 (16.8%), C2 = 101 (29.7%) and C3 = 182 (53.5%) users in each cluster. The following two subsections describe our analysis results regarding the RI of attributes, attribute-level part-worth utilities, part-worth utilities changes per attribute level (Section 4.2.1) and additional latent variables for each of the three clusters C1, C2 and C3 (Section 4.2.2).


4.2.1.

Part-worth utilities, RI and attribute-level changes Figure 3 and Table 6 present the clusters and show the RI of attributes, attribute-level part-worth utilities and partworth utility changes per attribute level. Furthermore, we report the p-values resulting from pairwise t-tests that were used to assess whether the means of the variables are significantly different from each other. In line with the aggregated findings described in Figure 2, we find that price and storage capacity are the most important attributes and customer support, the least important attribute across all clusters (see Figure 3). However, taking a closer look at the estimates provided in Figure 3 and Table 6, we find some important differences between the identified clusters. Examining the RI in cluster C1, we find that the price has an RI of 33.8% which is similar to C2, but significantly different from C3. Accessibility is considered more than twice as important compared to the cluster C2/C3 with

Figure 3. RI of product attributes per cluster.

9

an RI of 24.7%. Consistently, these respondents draw substantial utility along with increasing means of access to the cloud storage (u = 2.92/u = 2.28) which is significantly higher than in all other clusters. Likewise, these respondents place more importance on storage capacity and also perceive increases in storage capacity to be more beneficial than respondents in C2 and C3, particularly for an increase in storage capacity from 50 to 100 GB (u = 1.78). The importance of encryption (RI of 8.4%) in this cluster is not significantly different from C3, but significantly lower compared to C2. However, a change from server-side encryption to client-side encryption is positively valued by these respondents with 1.45 utility units. This increase is significantly higher compared to C3 and lower compared to C2. Customer support plays the least important role in this cluster with an RI of 5.4% and a change from enhanced to live support is even perceived as marginally negative (u = − 0.30, p < .01). In summary, respondents in C1 seem to be ready to make trade-offs and, for example, spend 10 Euros for a storage service that offers 100 GB and access through website and software (3.97 + 1.78 + 2.92 > 4.41 + 2.73). While price (RI = 34.9%) and storage capacity (RI = 23.5%) are the most important attributes in C2 as well, the analysis reveals some unique characteristics that distinguish C2 from C1 and C3: respondents in C2 attach significantly more value to encryption than respondents from C1/C3, rendering it the third most important attribute in this cluster with an RI of 20.8% compared to an RI of 8.4% and 7.8% in clusters C1 and C3. Compared to respondents in C1 and the C3, respondents in C2 gain most value from a change from server-side to client-side encryption (u = 3.82). Notwithstanding that customer support also plays the least important role in cloud storage choice with a RI of 9.4%, it carries significantly more weight in C2

10


Table 6. RI, part-worth utilities and attribute-level changes per cluster. Cluster Variable


Part-worth utility and attribute-level change (u)

t-Tests (p-value)

C1

C2

C3

(1;2)

(2;3)

(1;3)

Price 0 Euro 5 Euro 10 Euro u (0 Euro → 5 Euro) u (5 Euro → 10 Euro)

3.85 − 0.56 − 3.29 − 4.41 − 2.73

3.51 − 0.69 − 2.81 − 4.20 − 2.12

5.85 − 1.21 − 4.65 − 7.06 − 3.44

0.05a 0.16 0.00 0.38 0.00b

0.00b 0.00b 0.00b 0.00b 0.00b

0.00b 0.00b 0.00b 0.00b 0.00b

Storage 5 GB 50 GB 100 GB u (5 GB → 50 GB) u (50 GB → 100 GB)

− 3.24 0.73 2.51 3.97 1.78

− 2.46 0.53 1.93 3.00 1.40

− 2.44 0.77 1.67 3.21 0.90

0.00b 0.02a 0.00b 0.00b 0.01b

0.89 0.00b 0.09 0.42 0.00b

0.00b 0.49 0.00b 0.00b 0.00b

− 2.71 0.21 2.50 2.92

− 0.80 0.61 0.18 1.41

− 0.71 0.05 0.65 0.76

0.00b 0.00b 0.00b 0.00b

0.54 0.00b 0.00b 0.00b

0.00b 0.06 0.00b 0.00b

2.28

− 0.43

0.60

0.00b

0.00b

0.00b

Encryption Server-side encryption Client-side encryption u (server-side → client-side)

− 0.73 0.73 1.45

− 1.91 1.91 3.82

0.07 − 0.07 − 0.14

0.00b 0.00b 0.00b

0.00b 0.00b 0.00b

0.00b 0.00b 0.00b

Customer support Basic Enhanced Live u (basic → enhanced) u (enhanced → live)

− 0.33 0.32 0.01 0.65 − 0.30

− 0.78 0.41 0.37 1.19 − 0.04

− 0.44 0.16 0.28 0.61 0.11

0.00b 0.22 0.00b 0.00b 0.01a

0.00b 0.00b 0.26 0.00b 0.12

0.16 0.00b 0.00b 0.71 0.00b

Accessibility Website only Website + software Website + software + mobile u (website only → website + software) u (website + software → website website + software + mobile)

a Significance level: 5%. b Significance level: 1%.

than in C1 and C3. Consistently, a change from basic to enhanced support is perceived as more beneficial than in any other cluster (u = 1.19). Acknowledging the results for the attributes encryption and customer support, we find that providing respondents in C2 with client-side encryption and enhanced customer support would exceed the drop in utility resulting from a price change from 0 to 5 Euros. This indicates that they are likely to make trade-offs and spend 5 Euros in an effort to increase their privacy and receive better support. Considering the results for cluster C3, we find that these respondents perceive price (RI = 54.1%) as significantly more important than respondents in C1 and C2. They also suffer the highest drop in utility when the cloud storage provider decides to change the monthly price from 0 to 5 Euros (u = − 7.06) and from 5 to 10 Euros (u = − 3.44) respectively. We can further observe that they perceive the lowest utility gain with increasing accessibility. Neither do they gain any utility from a change from client-side encryption to server-side

encryption (u = − 0.14) which presents a significant difference from clusters C1 and C2. Taking into account all results from Table 6, we see that even when offering respondents in C3 a service with 100 GB storage capacity, full accessibility (website, software and mobile), clientside encryption and live support, it would not provide them with sufficient utility to make a trade-off and spend 5 Euros per month.

4.2.2. Additional variables In an effort to validate our measurement of the additional latent variables described in Table 2, we evaluated the individual item reliability, convergent validity, discriminant validity and scale reliability of our constructs. Table 7 summarises the results of our assessment. The scale reliability and internal consistency were examined by calculating composite reliability and Cronbach’s alpha values. The results in Table 7 show that all values for all of the

Behaviour & Information Technology

11

Table 7. Reliabilities, AVE and latent variable correlations. Construct

CA

CR

AVE

1

2

3

4

5

6

7

1. Attitude 2. Personal IT innovativeness 3. Price consciousness 4. Information privacy concern 5. Risk beliefs 6. Intention to use 7. Disposition to trust

0.92 0.90 0.81 0.88 0.90 0.97 0.74

0.84 0.79 0.54 0.72 0.84 0.95 0.88

0.86 0.83 0.68 0.78 0.76 0.95 0.78

0.93 0.43 0.03 − 0.17 − 0.46 0.75 0.12

0.91 0.06 − 0.02 − 0.22 0.40 0.03

0.82 0.05 − 0.04 − 0.01 − 0.02

0.88 0.31 − 0.19 − 0.04

0.87 − 0.49 − 0.19

0.97 0.12

0.88

Note: CA, Cronbach’s alpha; CR, composite reliability; shaded cells, square root of AVE. Table 8.

Analysis of additional variables per cluster. Cluster Variable


Index values of additional variables

Attitude Personal IT innovativeness Price consciousness Information privacy concern Risk beliefs Intention to use Disposition to trust Computer proficiency Number of devices used % of actual cloud storage user

t-Tests (p-value)

C1

C2

C3

(1;2)

(2;3)

(1;3)

4.04 3.69 3.87 3.68 2.65 3.75 3.58 4.33 3.28 76.2%

3.46 3.33 4.05 4.25 3.13 3.11 3.43 4.11 2.82 56.14%

3.74 3.39 4.00 3.66 2.82 3.44 3.56 4.03 2.83 73.6%

0.00a 0.03c 0.12 0.00a 0.00a 0.00a 0.34 0.05c 0.01a 0.01a

0.08b 0.71 0.61 0.00a 0.03c 0.09b 0.38 0.45 0.97 0.01c

0.00a 0.02c 0.22 0.85 0.13 0.02c 0.84 0.00a 0.00a 0.63

a Significance level: 1%. b Significance level: 10%. c Significance level: 5%.

constructs in our model are above the commonly accepted minimum threshold of 0.7 (Gefen, Straub, and Boudreau 2000). All of the measurement items exhibit loadings that are significant at the .01 level on the hypothesised constructs and above the recommended minimum value of 0.707. In addition, each AVE value exceeds the accepted minimum of 0.50 and the square root of the AVE for each of the constructs is larger than all other inter-construct correlations (Fornell and Larcker 1981). We also performed a confirmatory factor analysis and assessed the cross loadings of the individual items. The test showed that each item loading is above 0.75 on the assigned target construct and at least 0.1 less on other constructs except for two items (Gefen and Straub 2005). After the validation of the measurement instruments, we analysed the averages of the latent variables (index values), the number of devices used and percentage of actual cloud storage user for each cluster as shown in Table 8. Again, we conducted pairwise t-tests to assess whether the index values of the variables are significantly different in each cluster; the p-values of these tests are also reported in Table 8. As can be seen from Table 8, the constructs disposition to trust and price consciousness do not denote any differences across the three clusters and therefore do

not lend themselves as discriminating variables with which we could better understand the various clusters and their inherent characteristics. Analysing the values in Table 8, we find that respondents in C1 exhibit the highest ratings on attitude towards cloud archiving, intention to use, personal IT innovativeness and computer proficiency which are, on average, significantly higher (p < .05) than in C2 and C3. Respondents in this cluster also use significantly more devices (μ = 3.28) to access cloud storage compared to C2 and C3. Furthermore, we can observe that people in this cluster show significantly lower ratings on the information privacy concern as well as risk beliefs constructs than in cluster C2. In addition, cluster C1 captures the highest share of actual cloud storage users that is significantly higher than in C2. Turning to the analysis of C2, we find that respondents in this cluster exhibit the strongest information privacy concern and risk beliefs with significantly higher index values than respondents in C1 and C3. These respondents also show the lowest values on the attitude and intention to use constructs, which indicates a less positive attitude towards cloud archiving and consistently less readiness to use cloud archiving. Consistent with this observation, this cluster also has the lowest share of actual cloud storage

12



users with a percentage of 56.14% which is significantly lower than in clusters C1 and C3. Acknowledging the results for cluster C3, we do not find any unique characteristics that would saliently discriminate these respondents from the remaining clusters. While Table 6 clearly indicates the dominant role of the price in this cluster, we do not find any differences on the price consciousness construct in this cluster as one could have expected. Considering the index values of attitude and intention that clearly separate cluster C1 from C2, the values in cluster C3 are somewhat in between, which indicates a significantly less positive attitude and less intention to use cloud archiving than respondents in C1, although there is no significant difference compared to C2.

5. Discussion 5.1. Findings and theoretical contribution Based on a sample of 340 respondents and data gathered through a discrete choice experiment, we employ a conjoint analysis to better understand user preferences in cloud storage choices in a more realistic setting where users have to balance their preferences over different cloud storage product attributes. We use the HB estimation method to calculate the individual utilities of attribute levels which are subsequently subjected to a cluster analysis in an effort to segment users into distinct groups with similar preferences. Acknowledging the aggregated results, we observe that price and storage capacity are the most important factors in the choice of a cloud storage service across all clusters. This result confirms the extant pricing pressure in the cloud storage market and is indicative of the increasing degree of commoditisation of cloud storage services (Durkee 2010; Nielsen 2012). Commoditisation is characterised by both increasingly homogenous products, which makes it more challenging for providers to differentiate themselves and sustain a competitive advantage, and price-sensitive customers who face relatively low costs in changing a provider (Manning et al. 2010; Reimann, Schilke, and Thomas 2010). Also consistent across all clusters, we can observe that the drop in utility caused by an increase in price from 0 to 5 Euros cannot solely be compensated by a change in any other attribute. One possible reason for the relatively strong aversion to shift from a free to a charged storage service may be attributed to the existence of a status quo bias (Samuelson and Zeckhauser 1988). Most of the cloud storage providers employ a freemium business model that seeks to attract customers with free products and then subsequently convert them into paying customers by selling complementary or enhanced features and services (Teece 2010). Nevertheless, if the retention of the status quo, that is, using a free cloud storage service, presents a viable option, a strong bias in favour of the status quo exists since shortcomings of possible alternatives,

that is, paying a monthly fee, to the status quo are weighted more heavily than its benefits, for example, receiving more storage capacity (Samuelson and Zeckhauser 1988; Kahneman 1992). In this study, the status quo is likely to be related to the free service, therefore biasing respondents towards a price of 0 Euro. This point towards a particular challenge for cloud storage providers operating a freemium business model since converting free customers into paying ones is both difficult and imperative for the existence and success of the provider (Needleman and Loten 2012). In view of the utility drops for the price in Table 5, we can derive that converting a customer into a paying customer who spends 5 instead of 0 Euros is almost twice as difficult on average as getting an already paying customer to spend additional 5 Euros. This finding also points towards the importance of customer retention which is, however, difficult to achieve when customers face relatively low transaction costs in switching the provider. Confirming this observation, Trenz and Huntgeburth (2014) find a positive influence of customer loyalty on willingness to pay, and conclude that cloud providers should focus on creating a loyal customer base instead of growing their base of free service users to eventually generate sustainable revenue streams. However, we note that our observation and the comparatively large size of C3 might also be attributed to our sample which is dominated by students who usually have limited disposable income. Nonetheless, an analogous tendency is reported in a study about customer conversion in the context of music as a service (Wagner, Benlian, and Hess 2013). Wager et al. conclude that users who prefer the gratis service build a negative attitude towards the premium version which might lead to a situation often referred to as the freemium trap (Farr 2013). Our results show that customer support is perceived as the least important attribute across all clusters with a RI ranging from 5.4% to 9.4%. The results show that, whereas a change from basic support (FAQs, tutorials) to enhanced support (e-mail support) is somewhat valued by the respondents, a change from enhanced support to live support (24/7 hotline, live chat) does not lead to an increase in utility. This finding indicates a low demand for costly live support capacities which may be accounted for by the easy-to-use user interfaces and high availability/reliability of cloud storage services (Gracia-Tinedo et al. 2013). Yet, most of the cloud storage providers promise and promote a 99.9% or better up-time for their services (Cloud Storage Reviews 2013). Moreover, we discovered structural differences between user preferences by means of a cluster analysis. As a result of this analysis, we derived three clusters that distinguish users into subcategories with similar preferences which vary in terms of the weights that the analysed product attributes carry in the choice of cloud storage services. Summarising the findings from this data analysis, we can characterise the first cluster (C1) as follows.


Behaviour & Information Technology Compared to the other clusters, respondents in C1 derive more value from storage capacity and increasing accessibility while they do care least about customer support. Cluster C1 captures the highest share of actual cloud users who also use significantly more computing devices to connect with the cloud service than any other cluster. Evaluating the additional variables collected for this cluster, we can observe that these respondents have a more positive attitude towards cloud archiving and correspondingly a stronger intention to use cloud archiving. Furthermore, they seem to be more IT innovative and computer proficient compared to C2 and C3, which indicates a higher degree of technological affinity and literacy. We might therefore label this cluster with the bold but simple name ‘technophiles’. Technophiles seem to be willing to make trade-offs and accept a price increase from 0 to 5 EUR in exchange for more storage capacity or more ways of access. Cluster 2 also exhibits unique characteristics that distinguish these respondents from all others. This cluster seems to capture the ‘cautious’ and has the lowest share of actual cloud storage users. These cautious respondents attach significantly more value to encryption and customer support and also perceive the highest increase in utility when an enhanced support instead of a basic support is provided. The cautious also seem to be ready to make trade-offs and accept a price increase from 0 to 5 EUR in exchange for, for example, an increased level of encryption and customer support. Correspondingly, the cautious show more information privacy concerns and stronger risk beliefs as well as less positive attitude towards cloud archiving and less intention to use cloud archiving compared to any other cluster. These findings correspond with the results of recent studies on cloud and online service adoption. For instance, Walter et al. (2014) find that enhancing a cloud provider’s privacy policies positively influences perceived trustworthiness, while Mou and Cohen (2014) identify trust and perceived risk as important factors in the context of online health services acceptance. Cluster 3, with 182 respondents (53.5%) the largest cluster in this study, captures respondents who are most focused on the price. These respondents could therefore be termed ‘price hunters’. These price hunters suffer the highest drop in utility due to price increase from 0 to 5 Euros and are very sensitive to price changes. Although price hunters do not exhibit a significant difference on the price consciousness construct compared to C1 and C2, they seem to follow a price-aversion strategy (Teltis and Gaeth 1990), that is, to decide only on the price of a cloud storage service. In contrast to the technophiles and cautious, the price hunters’ perceived drop in utility resulting from a price increase from 0 to 5 EUR could not be compensated by any adjustment of other features (see Table 6). From a theoretical perspective, this study makes some contributions to the body of research about the adoption of cloud storage and cloud archiving from an end-user

13

perspective. First, we identified five vital factors and their RI in the choice of cloud storage services when employed as a means of personal archiving. To the best of our knowledge, this study is the first to empirically investigate these factors and their relative ranking by using a conjoint analysis and individual-level estimates rather than aggregated values. Second, we uncovered three groups of users with different preference structures. We extracted some of the clusters’ unique characteristics in an effort to better describe and understand not only the preferences of the users but also their perceptions regarding various aspects such as information privacy concerns, risk beliefs and IT innovativeness. Third, this study makes a methodological contribution in which we make the used R script available to the research community. The script can be downloaded and easily adopted by other IS scholars who seek to conduct conjoint analyses using HB estimation. While most of the previous conjoint studies in the research field relied on the application of commercial software tools (Krasnova, Hildebrand, and Guenther 2009; see, e.g. Hu, Moore, and Hu 2012), such as Sawtooth (Sawtooth 2013), we present an alternative free of charge. Furthermore, the script can support other scholars in the creation of the discrete choice experiment design and the subsequent data analysis; and in addition, it may as such foster the exploration of other research areas. 5.2. Practical implications The results of this study also entail important implications for practitioners and cloud storage providers by allowing for both a deeper understanding of the benefit structure and insights into their perceptions of some constructs that are considered important in cloud adoption decisions such as information privacy concerns. We believe that cloud storage providers could make investment and service design decisions more effectively based on our findings. For instance, acknowledging the RI of customer support, providers could decide to abstain from offering expensive live support and, instead, provide basic or enhanced support. Resulting savings could alternatively be invested into features that are more valuable to their customers or specific customer groups. Cloud storage providers could also offer different service configurations that are tailored to the specific needs of the identified customer segments. For instance, to explicitly address the needs of the ‘cautious’, a provider could offer a storage service with client-side or other enhanced encryption mechanism while another service could be offered with a maximum of storage capacity and other state-of-the-art features to attract the technophiles. These different service configurations could then be offered at different prices acknowledging the specific customer’s willingness to bear the corresponding costs. That means, the cluster-specific part-worth utilities reported in this study may support cloud storage providers to more precisely address the needs of


14


their users and potential customers. And in an overall effort to grow revenues, fostering the conversion of free users into paying users is critical to the long-term success of a provider. Towards this end, design effort should be focused on service attributes that are considered important by the users in order to eventually offer a product where the perceived utility compensates the disutility resulting from the price. That being said, providers should be able to increase users’ perceived usefulness of the services itself which was also found to exert a positive influence on willingness to pay (Trenz and Huntgeburth 2014). Moreover, the findings shed some light on important psychological aspects of the identified user groups that could be addressed by accompanying measures. For example, the cautious exhibit significantly higher information privacy concerns and risk beliefs. Addressing these concerns with trust building measures can help to overcome these adoption inhibitors. For instance, in a study about health clouds acceptance, Ermakova, Fabian, and Zarnekow (2014) found that individuals’ concerns for information privacy can be mitigated by building trust in the provider itself, trust in privacy-preserving regulatory and technological mechanisms as well as by convincing individuals of the service benefits. Actively addressing these concerns, for example, by means of public relations or marketing campaigns, might help providers to foster trust. In consequence, such effort should help to positively influence individuals’ attitudes towards cloud archiving and, in turn, their use and continuance intention. 5.3. Limitations and future research directions There are limitations to be noted when interpreting the results of this study. One limitation of the paper is related to the employed research method. Previous empirical studies about discrete choice experiments indicate the existence of various order effects as well as attribute range (Beattie and Baron 1991) and attribute-level effects that can influence the way participants respond and, consequently, the resulting utility estimates. For example, Chrzan (1994) found that statistically significant choice set order, profile order and attribute order effects occur, and Wittink and Seetharaman (1999) demonstrated the effect of the number of defined attribute levels on the derived RI weights and predicted choices. While the theoretical support for the existence and in particular causes of these effects are rather mixed (for a detailed discussion see, e.g. Melles 2001), some potential remedies are suggested to encounter these effect. For instance, Chrzan (1994) suggests rotating the order of respondents, stimuli and attributes to offset the order biases on an aggregate level. However, adopting such remedies requires a different research design and thus points towards an opportunity for future research where these effects and their impacts could be controlled. Another limitation relates to the selected attributes which should not be considered exhaustive. There may be

other salient attributes that are important in the provider choice. However, in a trade-off between the exhaustiveness of attributes and a threatening participant exhaustion, which is deemed to increase with the number of included attributes (Green and Srinivasan 1978; Cattin and Wittink 1982), we decided to examine only the previously described attributes. For instance, sharing capabilities were mentioned during our interviews, but have been excluded after the preceding pretest of the online survey. This is because participants reported a high degree of cognitive exhaustion and fatigue which both can negatively affect consistency in choice and hence increase error variance (Savage and Waldman 2008). Also, other additional variables such as loyalty, perceived IT security or usefulness could be included in future research to better describe and understand the characteristics of the clusters. Finally, there is a limitation associated with the studied sample which consists mainly of German students. Although we may reason that students represent a significant portion of cloud storage user population, external validity might be limited. Students are characterised by an higher education, lower age and are notable for less disposable income as well as less crystallised attitudes (Sears 1986) which may affect the results. For example, in a meta-analysis, Peterson (2001) found that effect sizes derived from student samples frequently differ from those derived from non-student samples. On the other hand, previous research noted that when studying an innovation like cloud computing, a student sample may better represent experienced users than a random sample from a general population of users and might better foreshadow the direction in which the general population is stirring (Gallagher, Parsons, and Foster 2001). In addition, we collected data from Germany only. That is, caution should be taken when these findings are to be generalised to other countries and geographical regions, since national culture has been found to substantially affect risk and privacy perceptions as well as attitudes and preferences towards IT (Leidner and Kayworth 2006). Thus, in an effort to test the generalisability of our findings beyond the studied sample, future research could replicate the study using a non-student population and draw data from other countries. Conflict of interest disclosure statement No potential conflict of interest was reported by the authors.

Note 1. The R-script used in this study can be downloaded from http://131.173.160.123/CBC_r-script.zip

References Ackermann, Tobias, Thomas Widjaja, Alexander Benlian, and Peter Buxmann. 2013. “Perceived IT Security Risks of Cloud Computing: Conceptualization and Scale


Behaviour & Information Technology Development.” Paper presented at 33rd International Conference on Information Systems (ICIS 2012), Orlando, FL. Agarwal, Ritu, and Jayesh Prasad. 1998. “A Conceptual and Operational Definition of Personal Innovativeness in the Domain of Information Technology.” Information Systems Research 9 (2): 204–215. Aizaki, Hideo, and Kazushi Nishimura. 2008. “Design and Analysis of Choice Experiments Using R: A Brief Introduction.” Agricultural Information Research 17 (2): 86–94. Ajzen, Icek. 1991. “The Theory of Planned Behavior.” Organizational Behavior and Human Decision Processes 50 (2): 179–211. Allenby, Greg M., and James L. Ginter. 1995. “Using Extremes to Design Products and Segment Markets.” Journal of Marketing Research XXXll: 392–403. Allenby, Greg M., and Peter Rossi. 2006. “Hierarchical Bayes Models.” In The Handbook of Marketing Research: Uses, Misuses, and Future Advances, edited by Rajiv Grover and Marco Vriens, 418–440. Thousand Oaks, CA: Sage. Ambrose, Paul, and Ananth Chiravuri. 2010. “An Empirical Investigation of Cloud Computing for Personal Use.” MWAIS 2010 Proceedings. Paper 24. http://aisel.aisnet.org/ mwais2010/24. Amiri, Amin Khodabandeh, Hasan Cavusoglu, and Izak Benbasat. 2014. “When Is IT Unavailability a Strategic Risk? A Study in the Context of Cloud Computing.” Paper presented at International Conference on Information Systems (ICIS 2014), Auckland, New Zealand. Armbrust, Michael, Armando Fox, Rean Griffith, Anthony D. Joseph, Randy Katz, Andy Konwinski, Gunho Lee, David Patterson, Ariel Rabkin, and Ion Stoica. 2010. “A View of Cloud Computing.” Communications of the ACM 53 (4): 50–58. Armstrong, Just, and Terry Overton. 1977. “Estimating Nonresponse Bias in Mail Surveys.” Journal of Marketing Research 14: 396–402. Beattie, Jane, and Jonathan Baron. 1991. “Investigating the Effect of Stimulus Range on Attribute Weight.” Journal of Experimental Psychology: Human Perception and Performance 17 (2): 571–585. doi:10.1037/0096-1523.17.2.571. Bech, Mickael, Trine Kjaer, and Jørgen Lauridsen. 2011. “Does the Number of Choice Sets Matter? Results from a Web Survey Applying a Discrete Choice Experiment.” Health Economics 20 (3): 273–286. Behrend, Tara S., Eric N. Wiebe, Jennifer E. London, and Emily C. Johnson. 2011. “Cloud Computing Adoption and Usage in Community Colleges.” Behaviour & Information Technology 30 (2): 231–240. Benlian, Alexander, Thomas Hess, and Peter Buxmann. 2009. “Drivers of SaaS-Adoption – An Empirical Study of Different Application Types.” Business & Information Systems Engineering 1 (5): 357–369. doi:10.1007/s12599-0090068-x. Benlian, Alexander, Marios Koufaris, and Thomas Hess. 2011. “Service Quality in Software-as-a-Service: Developing the SaaS-Qual Measure and Examining Its Role in Usage Continuance.” Journal of Management Information Systems 28 (3): 85–126. Bhattacherjee, Anol, and Sang Cheol Park. 2013. “Why EndUsers Move to the Cloud: A Migration-Theoretic Analysis.” European Journal of Information Systems 23 (3): 357–372. Bradley, Mark, and Andrew Daly. 1994. “Use of the Logit Scaling Approach to Test for Rank-Order and Fatigue Effects in Stated Preference Data.” Transportation 21 (2): 167–184.

15

Bridges, John F. P., A. Brett Hauber, Deborah Marshall, Andrew Lloyd, Lisa A. Prosser, Dean A. Regier, F. Reed Johnson, and Josephine Mauskopf. 2011. “Conjoint Analysis Applications in Health – A Checklist: A Report of the ISPOR Good Research Practices for Conjoint Analysis Task Force.” Value in Health 14 (4): 403–413. Burda, Daniel, and Frank Teuteberg. 2013. “Sustaining Accessibility of Information Through Digital Preservation: A Literature Review.” Journal of Information Science 39 (4): 439–455. Burda, Daniel, and Frank Teuteberg. 2014. “Understanding the Benefit Structure of Cloud Storage as a Means of Personal Archiving – A Choice-based Conjoint Analysis.” Paper presented at 22nd European Conference on Information Systems (ECIS 2014), Tel Aviv, Israel. Cattin, Philippe, and Dick R. Wittink. 1982. “Commercial Use of Conjoint Analysis: A Survey.” The Journal of Marketing 46 (Summer): 44–53. Chrzan, Keith. 1994. “Three Kinds of Order Effects in ChoiceBased Conjoint Analysis.” Marketing Letters 5 (2): 165– 172. Cloud Storage Reviews. 2013. Cloud Storage Comparison. Cloud Storage Reviews 2013. http://www.cloudstoragereviews.org/ comparison/. Compeau, Deborah, Barbara Marcolin, Helen Kelley, and Chris Higgins. 2012. “Research Commentary – Generalizability of Information Systems Research Using Student Subjects – A Reflection on Our Practices and Recommendations for Future Research.” Information Systems Research 23 (4): 1093–1109. Durkee, Dave. 2010. “Why Cloud Computing Will Never Be Free.” Queue 8 (4): 20–29. doi:10.1145/1755884.1772130. Ermakova, Tatiana, Benjamin Fabian, and Rüdiger Zarnekow. 2014. “Acceptance of Health Clouds – A Privacy Calculus Perspective.” Paper presented at 22nd European Conference on Information Systems (ECIS), Tel Aviv, Israel. Farr, Christina. 2013. Why Freemium Models Are ‘Sexy’ But Overrated. VentureBeat 2013. http://venturebeat.com/2013/ 07/02/why-freemium-models-are-sexy-but-overrated/. Fishbein, Martin. 2008. “A Reasoned Action Approach to Health Promotion.” Medical Decision Making 28 (6): 834–844. Fornell, C., and D. F. Larcker. 1981. “Evaluating Structural Equation Models with Unobservable Variables and Measurement Error.” Journal of Marketing Research XVII: 39–50. Gallagher, Katherine, Jeffrey Parsons, and K. Dale Foster. 2001. “A Tale of Two Studies: Replicating ‘Advertising Effectiveness and Content Evaluation in Print and on the Web’.” Journal of Advertising Research 41 (4): 71–82. Gefen, David. 2000. “E-commerce: The Role of Familiarity and Trust.” Omega 28 (6): 725–737. Gefen, David, Elena Karahanna, and Detmar W. Straub. 2003. “Trust and TAM in Online Shopping: An Integrated Model.” MIS Quarterly 27 (1): 51–90. Gefen, D., and D. Straub. 2005. “A Practical Guide to Factorial Validity Using PLS-Graph: Tutorial and Annotated Example.” Communications of the Association for Information Systems 16 (1): 91–109. Gefen, D., D. Straub, and M. C. Boudreau. 2000. “Structural Equation Modeling and Regression: Guidelines for Research Practice.” Communications of the Association for Information Systems 4 (1): 2–77. Gensler, Sonja. 2006. “Ermittlung von Praferenzen fur Produkteigenschaften mit Hilfe der Choice-Based Conjoint


16


Analyse, Teil II.” Wirtschaftswissenschaftliches Studium 35 (6): 316–319. Giessmann, Andrea, and Katarina Stanoevska. 2012. “Platform as a Service – A Conjoint Study on Consumers’ Preferences.” Paper presented at International Conference on Information Systems (ICIS 2012), Orlando, FL. Gillett, Frank E., Christopher Mines, Ted Schadler, Michael Yamnitsky, Heidi Shey, Amelia Martland, and Reedwan Iqbal. 2011. The Personal Cloud: Transforming Personal Computing, Mobile, and Web Markets. Cambridge, MA: Forrester Research, Inc. Gracia-Tinedo, Raúl, Marc Sánchez Artigas, Adrian MorenoMartinez, Cristian Cotes, and Pedro Garcia Lopez. 2013. “Actively Measuring Personal Cloud Storage.” Paper presented at IEEE Sixth International Conference on Cloud Computing (CLOUD), Santa Clara, CA. Green, Paul E., Abba M. Krieger, and Yoram Wind. 2001. “Thirty Years of Conjoint Analysis: Reflections and Prospects.” Interfaces 31 (3): 56–73. Green, Paul E., and Venkatachary Srinivasan. 1978. “Conjoint Analysis in Consumer Research: Issues and Outlook.” Journal of Consumer Research 5 (2): 103–123. Green, Paul E., and V. Srinivasan. 1990. “Conjoint Analysis in Marketing: New Developments with Implications for Research and Practice.” Journal of Marketing 54 (4): 3–19. Haaijer, Rinus, Wagner Kamakura, and Michel Wedel. 2001. “The ‘No-Choice’ Alternative in Conjoint Choice Experiments.” International Journal of Market Research 4 (Quarter 1): 93–106. Heerwegh, Dirk. 2006. “An Investigation of the Effect of Lotteries on Web Survey Response Rates.” Field Methods 18 (2): 205–220. Hensher, David A., John M. Rose, and William H. Greene. 2005. Applied Choice Analysis: A Primer. Cambridge: Cambridge University Press. Hess, Stephane, David A. Hensher, and Andrew Daly. 2012. “Not Bored Yet – Revisiting Respondent Fatigue in Stated Choice Experiments.” Transportation Research Part A: Policy and Practice 46 (3): 626–644. doi:10.1016/j.tra.2011.11.008. Hu, Han-fen, William Moore, and Paul J. Hu. 2012. “Incorporating User Perceptions and Product Attributes in Software Product Design and Evaluation.” Paper presented at 33rd International Conference on Information Systems (ICIS 2012), Orlando, FL. Ion, Iulia, Niharika Sachdeva, Ponnurangam Kumaraguru, and Srdjan Capkun. 2011. “Home is Safer than the Cloud! Privacy Concerns for Consumer Cloud Storage.” Paper presented at 7th Symposium on Usable Privacy and Security, Pittsburgh, PA. Itani, Wassim, Ayman Kayssi, and Ali Chehab. 2009. “Privacy as a Service: Privacy-Aware Data Storage and Processing in Cloud Computing Architectures.” Paper presented at 8th IEEE International Conference on Dependable, Autonomic and Secure Computing (DASC’09), Chengdu, China. Jarvenpaa, Sirkka L., Noam Tractinsky, and Michael Vitale. 2000. “Consumer Trust in an Internet Store.” Information Technology and Management 1 (12): 45–71. Kahneman, Daniel. 1992. “Reference Points, Anchors, Norms, and Mixed Feelings.” Organizational Behavior and Human Decision Processes 51 (2): 296–312. doi:10.1016/07495978(92)90015-Y. Koehler, Philip, Arun Anandasivam, M. A. Dan, and C. Weinhardt. 2010. “Cloud Services from a Consumer Perspective.” Paper presented at 16th Americas Conference on Information Systems (AMCIS), Lima, Peru.

Krasnova, Hanna, Thomas Hildebrand, and Oliver Guenther. 2009. “Investigating the Value of Privacy in Online Social Networks: Conjoint Analysis.” Paper presented at 30th International Conference on Information Systems (ICIS), Phoenix, AZ. Lansing, Jens, Stephan Schneider, and Ali Sunyaev. 2013. “Cloud Service Certifications: Measuring Consumers’ Preferences for Assurances.” Paper presented at 21st European Conference on Information Systems, Utrecht, The Netherlands. Lee, Young-Chan, Hanh Tang, and Vijayan Sugumaran. 2012. “A Deployment Model for Cloud Computing Using the Analytic Hierarchy Process and BCOR Analysis.” Paper presented at Americas Conference on Information Systems (AMCIS), Seattle, WA. Leidner, Dorothy E., and Timothy Kayworth. 2006. “A Review of Culture in Information Systems Research: Toward a Theory of Information Technology Culture Conflict.” Management Information Systems Quarterly 30 (2): 357–399. Li, Yuan, and Kuo-chung Chang. 2012. “A Study on User Acceptance of Cloud Computing: A Multi-theoretical Perspective.” Paper presented at 18th Americas Conference on Information Systems (AMICS), Seattle, WA. Lichtenstein, Donald R., Nancy M. Ridgway, and Richard G. Netemeyer. 1993. “Price Perceptions and Consumer Shopping Behavior: A Field Study.” Journal of Marketing Research XXX (May): 234–245. Limesurvey. 2013. Limesurvey Version 2.00 + . The LimeSurvey Project 2013. http://www.limesurvey.org. Loske, André, Thomas Widjaja, Alexander Benlian, and Peter Buxmann. 2014. “Perceived IT Security Risks in Cloud Adoption: The Role of Perceptual Incongruence Between Users and Providers.” Paper presented at 22nd European Conference on Information Systems (ECIS), Tel Aviv, Israel. MacKenzie, S. B., P. M. Podsakoff, and N. P. Podsakoff. 2011. “Construct Measurement and Validation Procedures in MIS and Behavioral Research: Integrating New and Existing Techniques.” MIS Quarterly 35 (2): 293–334. Malhotra, Naresh K., Sung S. Kim, and James Agarwal. 2004. “Internet Users’ Information Privacy Concerns (IUIPC): The Construct, the Scale, and a Causal Model.” Information Systems Research 15 (4): 336–355. Manning, Stephan, Joan E. Ricart, Maria Soledad, Rosatti Rique, and Arie Y. Lewin. 2010. “From Blind Spots to Hotspots: How Knowledge Services Clusters Develop and Attract Foreign Investment.” Journal of International Management 16 (4): 369–382. Marshall, Cathy, and John C. Tang. 2012. “That Syncing Feeling: Early User Experiences with the Cloud.” Paper presented at Designing Interactive Systems Conference (DIS’12), Newcastle upon Tyne, UK. Martens, Benedikt, Jens Poeppelbuss, and Frank Teuteberg. 2011. “Understanding the Cloud Computing Ecosystem: Results from a Quantitative Content Analysis.” Wirtschaftsinformatik Proceedings 2011, Paper 65, Zurich, Switzerland. Martens, Benedikt, Frank Teuteberg, and Matthias Gräuler. 2011. “Design and Implementation of a Community Platform for the Evaluation and Selection of Cloud Computing Services: A Market Analysis.” Paper presented at 19th European Conference on Information Systems (ECIS), Helsinki, FI. Melles, Torsten. 2001. Framing-Effekte in der Conjoint-Analyse: Ein Beispiel für Probleme der Merkmalsdefinition. Aachen: Shaker. Mou, Jian, and Jason Cohen. 2014. “Trust, Risk Barriers and Health Beliefs in Consumer Acceptance of Online Health


Behaviour & Information Technology Services.” Paper presented at International Conference on Information Systems (ICIS 2014), Auckland, New Zealand. Needleman, Sarah E., and Angus Loten. 2012. When Freemium Fails. The Wall Street Journal, August 22. http://online.wsj. com/news/articles/SB10000872396390443713704577603 782317318996. Nielsen, Todd. 2012. Is Cloud Storage Becoming a Commodity? 21st Century IT 2012. http://www.21cit.com/author.asp? section_id = 2441&doc_id = 250047. Orme, Bryan K., Mark I. Alpert, and Ethan Christensen. 1997. “Assessing the Validity of Conjoint Analysis – Continued.” Accessed August 21, 2013. http://www.sawtoothsoftware. com/support/technical-papers/general-conjoint-analysis/ assessing-the-validity-of-conjoint-analysis-continued-1997. Peterson, Robert A. 2001. “On the Use of College Students in Social Science Research: Insights from a Second-order Meta-analysis.” Journal of Consumer Research 28 (3): 450– 461. doi:10.1086/323732. Punj, Girish, and David W. Stewart. 1983. “Cluster Analysis in Marketing Research: Review and Suggestions for Application.” Journal of Marketing Research (JMR) 20 (2): 134–148. R Core Team. 2013. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. Accessed September 14, 2013. http://www.r-project.org/. Reimann, Martin, Oliver Schilke, and Jacquelyn S. Thomas. 2010. “Toward an Understanding of Industry Commoditization: Its Nature and Role in Evolving Marketing Competition.” International Journal of Research in Marketing 27 (2): 188–197. Repschlaeger, Jonas, Ruediger Zarnekow, Stefan Wind, and Turowski Klaus. 2012. “Cloud Requirement Framework: Requirements and Evaluation Criteria to Adopt Cloud Solutions.” Paper presented at 20th European Conference on Information Systems (ECIS), Barcelona, Spain. Rokka, Joonas, and Liisa Uusitalo. 2008. “Preference for Green Packaging in Consumer Product Choices – Do Consumers Care?” International Journal of Consumer Studies 32 (5): 516–525. Rossi, Peter. 2012. bayesm: Bayesian Inference for Marketing/ Micro-econometrics. R Package Version 2.2-5. Accessed September 30, 2013. http://CRAN.R-project.org/package = bayesm. Rossi, Peter E., and Greg M. Allenby. 2003. “Bayesian Statistics and Marketing.” Marketing Science 22 (3): 304–328. Samuelson, William, and Richard Zeckhauser. 1988. “Status Quo Bias in Decision Making.” Journal of Risk and Uncertainty 1 (1): 7–59. Sattler, Henrik, and Adriane Hartmann. 2008. “Commercial Use of Conjoint Analysis.” In Operations Management in Theorie und Praxis, edited by Michael Höck and Kai-Ingo Voigt, 103–119. Wiesbaden: Gabler. Savage, Scott J., and Donald M. Waldman. 2008. “Learning and Fatigue During Choice Experiments: A Comparison of Online and Mail Survey Modes.” Journal of Applied Econometrics 23 (3): 351–371. Sawtooth. 2013. Conjoint Analysis Software. Sawtooth Software, Inc. Accessed November 25, 2013. http://www. sawtoothsoftware.com/products/conjoint-choice-analysis. Sears, David O. 1986. “College Sophomores in the Laboratory: Influences of a Narrow Data Base on Social Psychology’s View of Human Nature.” Journal of Personality and Social Psychology 51 (3): 515–530. Sturm, Benjamin, Jens Lansing, and Ali Sunyaev. 2014. “Moving in the Right Direction?: Mapping Literature on Cloud Service Certifications’ Outcomes with Practitioners’

17

Perceptions.” Paper presented at 22nd European Conference on Information Systems (ECIS), Tel Aviv, Israel. Teece, David J. 2010. “Business Models, Business Strategy and Innovation.” Long Range Planning 43 (2): 172–194. Teltis, Gerard J., and Gary J. Gaeth. 1990. “Best Value, PriceSeeking, and Price Aversion: The Impact of Information and Learning on Consumer Choices.” Journal of Marketing 54 (2): 34–45. Top. 2013. Why Use Online Backup? TheTop10BestOnline Backup.com, Reloaded Marketing LLC. http://www.thetop 10bestonlinebackup.com. Train, Kenneth E. 2003. Discrete Choice Methods with Simulation. Cambridge: Cambridge University Press. Trenz, Manuel, and Jan Huntgeburth. 2014. “Understanding the Viability of Cloud Services: A Consumer Perspective.” Paper presented at 22nd European Conference on Information Systems (ECIS), Tel Aviv, Israel. Trenz, Manuel, Jan C. Huntgeburth, and Daniel J. Veit. 2013. “The Role of Uncertainty in Cloud Computing Continuance: Antecedents, Mitigators, and Consequences.” Paper presented at 21st European Conference on Information Systems (ECIS), Utrecht, The Netherlands. Verma, Shalini. 2012. Forecast: Consumer Digital Storage Needs, 2010–2016. Stamford, CT: Gartner, Inc. Vriens, Marco. 1994. “Solving Marketing Problems with Conjoint Analysis.” Journal of Marketing Management 10 (1–3): 37–55. Wagner, Thomas M., Alexander Benlian, and Thomas Hess. 2013. “The Advertising Effect of Free – Do Free Basic Versions Promote Premium Versions Within the Freemium Business Model of Music Services?” Paper presented at 46th Hawaii International Conference on System Sciences (HICSS), Grand Wailea, Maui, Hawaii. Walter, Nicolai, Ayten Öksüz, Marc Waltersbusch, Frank Teuteberg, and Jörg Becker. 2014. “‘May I Help You?’ Increasing Trust in Cloud Computing Providers Through Social Presence and the Reduction of Information Overload.” Paper presented at International Conference on Information Systems (ICIS 2014), Auckland, New Zealand. Walterbusch, Marc, Benedikt Martens, and Frank Teuteberg. 2013. “Exploring Trust in Cloud Computing: A MultiMethod Approach.” Paper presented at 21st European Conference on Information Systems (ECIS), Utrecht, The Netherlands. Walterbusch, Marc, and Frank Teuteberg. 2014. “Towards an Understanding of the Formation and Retention of Trust in Cloud Computing: A Research Agenda, Proposed Research Methods and Preliminary Results.” Paper presented at TrustBus 2014: 11th International Conference on Trust, Privacy & Security in Digital Business, Munich, Germany. Wang, Cong, S. Chow, Qian Wang, Kui Ren, and Wenjing Lou. 2013. “Privacy-Preserving Public Auditing for Secure Cloud Storage.” IEEE Transactions on Computers 62 (2): 362–375. Wheeler, Bob. 2011. AlgDesign: Algorithmic Experimental Design. R Package Version 1.1-7. Accessed September 18, 2013. http://CRAN.R-project.org/package = AlgDesign. Wittink, Dick R., and Philippe Cattin. 1989. “Commercial Use of Conjoint Analysis: An Update.” The Journal of Marketing 53 (July): 91–96. Wittink, Dick R., and P. B. Seetharaman. 1999. “A Comparison of Alternative Solutions to the Number-of-Levels-Effect.” Paper presented at Sawtooth Software Conference, Sequim, WA. Zhiyun, Guo, Song Meina, and Song Junde. 2010. “A Governance Model for Cloud Computing.” Paper presented at International Conference on Management and Service Science (MASS), Wuhan, August 24–26.

Exploring consumer preferences in cloud archiving ...

Exploring consumer preferences in cloud archiving ...

Suggest Documents

ATM Consumer Preferences

Measuring consumer preferences regarding

Payments transition in India – consumer preferences

Consumer preferences towards beef cattle in Chile

International Differences in Consumer Preferences ... - AgEcon Search

International Differences in Consumer Preferences ... - AgEcon Search

Consumer Preferences and Trade in Energy Resources

Local and Global Consumer Preferences

CONSUMER PREFERENCES FOR BIOETHANOL BLENDS

Exploring passenger preferences in airline service ...

Driving consumer goods innovation in the cloud - The Consumer ...

Consumer Interest in Automation: Preliminary Observations Exploring ...

13. Consumer information, consumer preferences and ... - SSRN papers

Consumer-centric Resource Accounting in the Cloud

Consumer Protection in Cloud Computing Services ... [PDF]

Driving consumer goods innovation in the cloud

Driving consumer goods innovation in the cloud

Consumer preferences and behaviour in shopping malls in Poland ...

Transportation Oil Demand Consumer Preferences and Asymmetric ...

product category dependent consumer preferences for ... - CiteSeerX

consumer preferences for quality characteristics ... - AgEcon Search

Identifying consumer buying preferences of beef

Eliciting Consumer Preferences Using Robust ... - Semantic Scholar

Consumer Preferences for Ecolabeled Seafood - Food Marketing ...