The effect of message types on the acceptance of location-based push notifications and the mediating roles of perceived fit and skepticism

(1)

Master’s Thesis

MSc Business Administration

The effect of message types on the acceptance of location-based push

notifications and the mediating roles of perceived fit and skepticism

Student: Victoria Fandl

Thesis subject: Marketing Student-No.: 11385812

Supervisor: Dr. Alfred Zerres

(2)

Statement of originality

This document is written by Student Victoria Fandl who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creat-ing it.

The Faculty of Economics and Business is responsible solely for the supervision of comple-tion of the work, not for the contents.

(3)

Abstract

Location-based services constitute valuable opportunities for companies to link customers closely to the company, however they are still rarely used and researched. Previous scholars focused, for instance, on privacy concerns and tested different determents on people’s like-lihood of accepting location-based advertising. This thesis contributes to this field by distin-guishing between different types of location-based push notifications and examining the role of perception of fit between the app purpose and the disseminated message as well as skep-ticism regarding potential underlying intentions of the app provider. Due to the notable fit-ness hype over the past years, the main sector of interest is the fitfit-ness and health industry. In cooperation with a sports center in Amsterdam, 381 members of this gym were surveyed regarding a complimentary app offered by the sports club. The three push notification types “information-based”, “monetary incentive-based” and “community-building” were tested. It is found that information-based and monetary incentive-based messages lead to higher ac-ceptance when compared to community-building notifications. Further, higher perceived fit of app purpose and message leads to lower skepticism regarding potential motives of the app provider. Both, higher perceived fit and lower skepticism lead to higher acceptance of location-based advertising.

(4)

Table of Contents

Abstract ... 3

1 Introduction ... 6

2 The research question ... 9

3 Theoretical contributions ... 9

4 Managerial implications ... 10

5 Theoretical framework ... 11

5.1 Location-based push notifications ... 11

5.2 Acceptance of location-based advertising ... 13

5.3 Research gap: the effect of message types on acceptance ... 14

5.4 Construal Level Theory (CLT) ... 16

5.5 Community-building, monetary incentive-based and information-based messages 17 5.5.1 Community-building messages ... 17

5.5.2 Monetary incentive-based messages ... 19

5.5.3 Information-based messages ... 20

5.5.4 Formation of hypotheses of direct effects ... 23

5.6 Perceived fit ... 24

5.7 Skepticism ... 26

6 Conceptual Model ... 29

7 Research methodology ... 29

7.1 Research and survey design ... 29

7.2 Data collection and sample ... 31

7.3 Pre-test ... 32

7.4 Measurement instruments ... 33

7.5 Biases ... 33

7.6 Data analysis ... 34

7.6.1 Normality check ... 34

7.6.2 Confirmatory factor analysis ... 35

7.6.3 Reliability check ... 37 7.6.4 Randomization check ... 37 7.6.5 Manipulation check ... 37 7.6.6 Correlation matrix ... 38 7.6.7 One-way ANOVA ... 42 7.6.8 Process ... 45

8 Conclusion and hypotheses outcomes ... 48

9 Discussion ... 49

10 Limitations and future research ... 51

References ... 53

(5)

List of tables

Table 1: Scenarios ...32

Table 2: Normality Check ...35

Table 3: Pattern matrix factor analysis...36

Table 4: Manipulation check group statistics ...38

Table 5: Manipulation check independent samples test ...38

Table 6: correlation matrix ...41

Table 7: One-way descriptives message types ...43

Table 8: One-way ANOVA message types ...43

Table 9: One-way descriptives every scenario ...44

Table 10: One-way ANOVA every scenario ...45

(6)

1 Introduction

Looking at the multitude of smartphone applications available, it is the prerequisite for apps to perform appropriately and offer useful features to be downloaded. Accounting for half of the total time spent on digital media, they have become the major tool to access the internet (“The 2016 U.S. Mobile App Report”, 2016). In 2017, there will be 4.43 billion mobile phone users worldwide, representing 59.9 % of the global population (“Mobile Phone, Smartphone Usage,” 2016). Likewise Apple claimed to record cumulatively 140 bil-lion app downloads in September 2016 (“Cumulated number of apps,” n.d.). However, not only developing properly functioning apps, but recognizing the big range of possibilities to link customers closely to the company should be the focus.

Location-based services are still rarely used, but offer powerful opportunities, since they allow companies to address their customers individually with relevant content at the best time and place to increase customer loyalty and sales (Bauer & Strauss, 2016; Bruner & Kumar, 2007). Location-based messages, disseminated directly from the app, are called loca-tion-based push notifications. The users are being addressed without their explicit request

for specific content (Khosrowpour, 2013). If not done correctly, this may lead to privacy

con-cerns, which negatively affect consumers’ acceptance towards mobile advertising (Limpf & Voorveld, 2015; Okazaki, Molina & Hirose, 2012).

Consumers expect mobile (and location-based) services to be valuable, useful, in-formative, relevant, fun, entertaining, not irritating, reliable, paired with an incentive, pre-sented in the right context, to name a few properties (Pura, 2005; Xu, Oh, & Teo, 2009;

(7)

Meuli, 2013; Izquierdo-Yusta, Olarte-Pascual, & Reinares-Lara, 2015). Push notifications of mobile applications can be utilized to meet these expectations and to enhance app engage-ment by up to 88 per cent, and the retention rate by approximately two times, if done cor-rectly (“Push Messaging Drives 88% More App Launches”, 2014). Thus, it is of utmost im-portance to understand which message content needs to be dispatched in order to achieve user acceptance, especially since consumers are faced with an overload of information (Bawden & Robinson, 2009).

Previous scholars somewhat examined the effect of message content in regards to location-based advertising. They focused, for instance, on open vs. closed message content and text vs. multimedia messages. However, there is still a lack in research, especially on the effect of specific content types on the acceptance (Ketelaar et al., 2015; Xu, Oh & Teo, 2009; Merisavo et al., 2007; Pura, 2005).

In this thesis three specific message types are defined and tested, i.e. information-based messages, monetary incentive-information-based messages and community-building messages. Previous scholars suggested distinguishing between information-based and monetary mes-sages (Merisavo et al., 2007). Community-building mesmes-sages have been chosen additionally since they constitute an effective way to link customers more closely to the company, i.e. reach higher loyalty by building a brand community (Keller, 2001; Muniz & O’Guinn, 2001; McAlexander, Schouten & Koenig, 2002).

Furthermore, the mediating role of perceived fit and skepticism will be investigated thoroughly. In social marketing, “fit” is often used to evaluate the success of CSR (Corporate

(8)

Social Responsibility) activities (Varadarajan & Menon, 1988) and in the context of brand ex-tensions to determine whether or not an extension will be successful (Bottomley & Holden, 2001). This concept has already been used in regards to advergaming campaigns on social networks (Okazaki & Yagüe, 2012). However, it has not been applied in the context of ac-ceptance of location-based advertising. Transferring this knowledge to location-based mes-sage dissemination offers a new perspective. As stated above, consumers download apps following a specific purpose. If the message content is not in line with the app purpose, ac-ceptance might be influenced. Thus, it will be examined whether a higher perception of fit between the message content and the original app purpose leads to higher acceptance.

There is an increased likelihood of people examining push notifications critically, since they pop up on the phone without being requested and may interrupt users’ current activity on the phone. Previous literature suggests that skepticism enhances the consumers’ ability to realize advertisers’ underlying intentions and to recognize persuasive messages to be less honest, less reliable and more biased which leads to fewer positive responses to messages (Hardesty, Carlson, and Bearden 2002; Obermiller, Spangenberg, and Maclachlan 2005). Hence, the role of skepticism will be looked at in regards to acceptance of location-based advertising.

The main sector of interest for this thesis is the fitness and health industry as a re-markable fitness and nutrition hype, stimulated especially by the spread of internet technol-ogy and social media, could be observed over the past years (PriceWaterhouse Coopers, 2012; Rutsaert et al., 2013; Fox & Duggan, 2013; Wilson, 2007). Hence, apps that provide

(9)

could be conducted in cooperation with a high-end sports center in Amsterdam Oost that offers a complimentary mobile app for their members. The purpose of the app is offering support for a healthy and active lifestyle. People download it to keep track of their workouts and their body metrics, check class schedules and opening hours, create workout plans with their trainer, look for different workouts, watch 3D exercise demonstrations etc.

2 The research question

The research question to be answered in this thesis is:

How do different message types of mobile applications affect the acceptance of loca-tion-based services? And how is this dynamic mediated by the perceived fit of the message type with the app as well as by skepticism about possible hidden goals attributed to the app provider?

3 Theoretical contributions

This thesis conducts further research in the area of mobile advertising by using loca-tion-based information. Previous scholars focused on many different aspects of location-based advertising and services, especially underlying privacy concerns have been studied manifoldly (Krishen, Raschke, Close & Kachroo, 2017; Kim, 2016; Xu, Teo, Tan & Agarwal, 2009). However, there is a lack of research in regards to brand-related apps, as many studies focused on applications like foursquare, on location-tracking via network providers, on social media check-ins etc. (Yu, Zo, Choi & Ciganek, 2013; Kim, 2016). Some papers investigated the message content in regards to acceptance of location-based services, e.g. open vs. close message content and text vs. multimedia messages or promotional messages vs. brand

(10)

ad-vertising (Ketelaar et al., 2015, Xu, Oh, & Teo, 2009; Unni & Harmon, 2007). However, there is insufficient research emphasizing the importance of the message content and especially regarding the perceived fit between the disseminated message and the initial reason for the download of the application. No research could be found that tests the effect of community-building messages as well as informational messages and monetary incentives on ac-ceptance of location-based push notifications. Moreover, the importance of the effect of perceived fit of message content with the app purpose as well as the interplay between per-ceived fit and possible skepticism regarding the app provider’s underlying motivation have not been investigated.

4 Managerial implications

Three developments indicate the probable success of mobile advertising: the growth of mobile devices, the forecast that mobile advertising will continue to grow, and the devel-opment of increasingly sophisticated mobile devices (Izquierdo-Yusta, Olarte-Pascual, & Reinares-Lara, 2015).

Nevertheless, the decrease in app downloads as well as the changing behavior of app users, e.g. arranging apps in folders and not allowing push notifications easily, as stated above, requires app providers to find a way not only to attract consumers’ attention, but es-pecially achieve their acceptance.

Moreover, the app attention span study, published by AppDynamics in 2014, showed that users’ expectations towards app performance are increasing more and more while their

(11)

they had deleted or uninstalled at least one app due to poor performance. Since push notifi-cations are part of an app’s performance, the role of disseminating the right messages is the focus of this thesis.

Location-based services are a pivotal chance for marketers to create strong customer bonds. It is essential for them to understand which content is legitimate to receive in cus-tomers’ perception, in regards to the type of message and also if a message fits the app pur-pose. Furthermore, it is crucial to know whether customers attribute an underlying motiva-tion to the app provider when receiving specific messages, in order to purposefully apply this knowledge in practice. It will be helpful to understand the theory behind consumers’ percep-tions and behavior, but especially to have a systematic approach for making use of location-based push notifications.

The fact that the research was conducted in cooperation with a sports center allows for a “real-world” example which offers a great opportunity to obtain real-world results, that can be transferred to apps with similar features and end users.

5 Theoretical framework

5.1 Location-based push notifications

Mobile advertising is defined as “the set of actions that enable firms to communicate and relate to their audience in a relevant, interactive way through any mobile device or net-work” (Mobile Marketing Association, 2010). It allows marketers to distribute personalized information to individually target consumers. Consumers either receive advertisements in a

(12)

passive way through the push-type delivery method or proactively via the pull-type method. As extensive consumer data is available, it is possible to design personalized advertisements based on consumers’ profiles and preferences and to deliver them right to the mobile phone (Izquierdo-Yusta, Olarte-Pascual, & Reinares-Lara, 2015).

Location-Based Advertising (LBA) on mobile devices goes one step further as location data is used additionally. It enables the dissemination of advertisements and information to

consumers, based on their current location, dynamically and in real-time. Moreover, content

can be exchanged quickly by remote access (Bauer & Strauss, 2016). It is referred to as a subset of location-based marketing (Unni & Harmon, 2007). Location-based services are a broader concept that comprise besides LBA also emergency and safety-related services, bill-ing, entertainment, navigation, asset trackbill-ing, city guides, traffic updates etc. (Lim & Siau, 2003 in Xu, Oh and Teo, 2009).

In sum, location-based services can be seen as the umbrella term for location-based marketing and location-based advertising.

It is necessary to distinguish between push and pull location-based advertising. Re-search shows that privacy concerns only play a role with location-based push advertising (push LBA), not with location-based pull advertising (pull LBA) on mobile devices since pull LBA allows more feeling of control (Limpf & Voorveld, 2015). Thus, skepticism also plays a more important role in regards to push, rather than pull, messages (Krishen, Raschke, Close & Kachroo, 2017; Kim, 2016; Xu, Teo, Tan & Agarwal, 2009).

(13)

Even though, LBA shows a higher relative effect when clearly demanded rather than when unexpectedly confronted with location-based advertising messages, it is also shown that there is a risk of consumers not activating or requesting mobile location-based advertis-ing (Unni and Harmon, 2007), especially, since nowadays people are confronted with infor-mation overload, as mentioned above (Bawden & Robinson, 2009). Taking this into consid-eration as well as the fact that push notifications technology is more timely and relevant, given their popularity and prevalence (Xu, Oh & Teo, 2009), the focus in this thesis is on push notifications.

It is important to understand that push notifications offer limited space for infor-mation since they appear as teasers on the screen of a smartphone. Push notifications can be seen as announcements for further content provided via the app and animate users to open the app for further information.

5.2 Acceptance of location-based advertising

Acceptance research gains useful insights into the success or failure of new products or services (Silberer & Wohlfahrt, 2001). Overall, there is still little research on the ac-ceptance of mobile advertising compared to its expansion and importance in the economic world, whereby location-based advertising has received even less attention (Xu, Oh & Teo, 2009; Ketelaar et al., 2015). Existing studies on acceptance of location-based services are often built upon findings about mobile advertising.

(14)

found that attitude is an important precept in research on advertising and technology ac-ceptance, which leads to the suggestion that consumers’ intention to accept mobile adver-tising is positively affected by their attitude towards it (Limpf & Voorveld, 2015; Haghirian & Madlberger, 2005; Oh & Xu, 2003). Likewise, subjective norms (Bauer, Reichardt, Barnes & Neumann, 2005) and technology self efficacy (Compeau & Higgins, 1995; Neill & Richard, 2012) are significant determents on individuals’ likelihood of accepting mobile advertising. Also, ease of use, perceived usefulness and goal impediment have a major impact on the ac-ceptance, or at least on the „not avoidance“, of mobile advertising (Bruner & Kumar, 2005; Shin and Tsui-Chuan Lin, 2016). Goal impediment means being disabled to fulfill intended tasks on the mobile device. Thus, messages that randomly appear on one’s phone while ful-filling a task might be considered as disturbing, which also emphasizes why it is interesting to study push notifications in this regard.

Furthermore, it is suggested that consumer acceptance can be increased through rel-evance and monetary incentives, e.g. special offers (Patel, 2001; Rettie, Grandcolas & Dea-kins, 2005). Pura (2005) also highlights monetary value and especially emphasizes the im-portance of attractive content in order to keep valuable customers.

5.3 Research gap: the effect of message types on acceptance

Previous scholars somewhat examine the effect of message content in regards to lo-cation-based advertising. They focus, for instance, on open vs. closed message content and text vs. multimedia messages (Ketelaar et al., 2015; Xu, Oh & Teo, 2009). It is proposed that usefulness, relevance, fun, entertainment, reliability, incentives, the right context etc. have

(15)

positive effects on the acceptance (Pura, 2005; Xu, Oh, & Teo, 2009; Chang, Kaasinen, & Kai-painen, 2012; Merisavo et al., 2017; Uitz & Koitz, 2013; Richard & Meuli, 2013; Izquierdo-Yusta, Olarte-Pascual, & Reinares-Lara, 2015).

However, Pura criticizes in 2005 that, even though attractive content is key for cus-tomer retention, there is a lack of research in this area. It is crucial to study which type of value is necessary in the location-based context from consumers’ points of view (Pura, 2005). Likewise, Merisavo et al. claim in 2007 that specific attention shall be paid to the utili-ty and relevancy of mobile advertising messages. They even state that the consumers should be provided with either useful information or with a way to save time or money, based on their specific context in regards to location and personal profile through mobile advertising. As a result, they suggest future scholars to put focus on the content of mobile advertising and their effect on acceptance (Merisavo et al., 2007).

In this thesis three specific message types are defined and tested, i.e. information-based push notifications (useful information), monetary incentive-information-based push notifications (ways to save money) and community-building push notifications.

Monetary incentives are looked at since they show immediate positive effects in re-gards to purchasing behavior (Kotler, 2000; Kamakura & Russel, 1989). However, it is also essential to understand the long-term perspective, i.e. effects on attitude. Monetary incen-tive-based messages contain temporary price reductions, such as special offers, seasonal sales etc.

(16)

Information-based incentives are interesting to be examined due to the increasing importance of access to relevant information, not only due to an information overflow, but especially because of the growing spread of misleading information (Goldberg & Sliwa, 2011; Theodosiou & Green, 2003; Burke, 2011; Goudreau, 2012). Information-based push notifica-tions comprise such that announce interesting and relevant content in regards to the topic of interest (in this case fitness and health).

Community-building messages are included as they are suggested to have the poten-tial of linking customers more closely to the company (Keller, 2001; Muniz & O’Guinn, 2001; McAlexander, Schouten & Koenig, 2002) and thus represent an interesting tool for market-ers. They are linked to a social online network via the mobile application and announce that friends of the person’s online profile are nearby.

5.4 Construal Level Theory (CLT)

In this thesis the CLT will be used to explain the difference of efficacy between the three different message types. It has been used manifoldly to explain the effectiveness of mobile targeting. It links distance and abstraction, proposing that psychological distance is an essential determinant of whether primary, essential characteristics, or secondary, pe-ripheral characteristics, are mainly used to evaluate objects and events (Trope, Liberman, & Wakslak, 2007). According to this theory, people use concrete, low-level construals to repre-sent near events and abstract, high-level construals to reprerepre-sent distant events. Reprerepre-senta- Representa-tions of events in the near future (close) are very detailed, unstructured, contextual and concrete (low-level construals), whereas representation of events in the distant future (far)

(17)

are more abstract, decontextual and schematic (high-level construals). Low-level construals emphasize on how to achieve a goal, whereas high-level construals highlight why action should be taken. Cognitive processing happens more fluently if psychological distance and construal levels are matching, i.e. close and low as well as far and high, in contrast to when they are incongruent (Katz & Byrne, 2013).

Luo et al. (2014) show that the consumer construal level expounds how consumers differently evaluate mobile messages in different contexts of temporal and spatial condi-tions, which leads to variance in the effectiveness of mobile targeting. As per their study, purchase intentions of consumers are highest when they get an SMS close to time and place of the specific happening. This is explained by the shorter temporal and geographical dis-tances, which lead to consumers mentally construing the object or event more concretely and thus to enhanced involvement and purchase intention.

5.5 Community-building, monetary incentive-based and information-based mes-sages

5.5.1 Community-building messages

As stated in the introduction, one of the goals of companies using location-based services is to increase customer loyalty (Bauer & Strauss, 2016). More and more marketers have realized that profit increases hardly when focusing mainly on attracting new custom-ers, in contrast to taking care of existing customers and building loyalty enhancement pro-grams. Major progress in information and communication technologies enables brand com-munities to spread and grow, presenting lots of opportunities for companies (Armstrong and

(18)

Hagel, 1996; McAlexander, Schouten, & Koenig, 2002; Muniz and O’Guinn, 2001; Sheth and Parvatiyar, 1995 in Hur, Ahn and Kim, 2011). Thus, it can be inferred that it is interesting for managers to understand the effect of community-building messages on the acceptance of location-based advertising.

Regarding Keller’s (2001) customer-based brand equity (CBBE) model, the highest possible level is brand resonance. Brand resonance comprises loyalty, community, attach-ment and engageattach-ment. Thus, if this level is reached, customers have a relationship not only with the brand, but also with the brand community and “active loyalty” is reached, which pictures the highest form of loyalty. Likewise Muniz and O’Guinn (2001) introduced the brand community triad, which demonstrates the interaction not only between the customer and the brand, but also among the customers. McAlexander, Schouten & Koenig (2002) ex-pand this approach and define brand community as strong relationships between the cus-tomer and the brand, the cuscus-tomer and the company, the cuscus-tomer and the product in use, and among fellow customers. These relationships produce positive outcomes for both par-ties: the customer can satisfy social needs by building valuable relationships and the brand profits from the customers’ resulting loyalty (McAlexander, Schouten, & Koenig, 2002).

Individuals aspire after building and maintaining interpersonal relationships (Baumeister & Leary, 1995). The Psychological Sense of Community (Sarason, 1974) and the theory of social capital (Bourdieu, 1983; Coleman, 1988) explain the reasons for consumers to engage in (brand) communities. These are, amongst others, to achieve a feeling of similar-ity to others, belonging to a group, interdependence with others etc.

(19)

It is found that social desires, “social identity” (i.e. the cognitive self-awareness of membership in the brand community), affective commitment and perceived importance of membership impact brand behavior (Bagozzi & Dholakia, 2006). Also, brand community trust and affect lead to commitment, which leads to loyalty behaviors (Hur, Ahn and Kim, 2011).

Community-building location-based push notifications “dig deeper” since customers’ underlying feelings of belonging and, in a broader sense, desire for happiness, are ad-dressed, rather than giving merely rational incentives. People strive for happiness and most of them think about their desire for happiness daily (King & Napa, 1998; Freedman, 1978). It has been found that social affiliation produces increases in happiness (Csikszentmihalyi & Hunter, 2003). In fact, Tkach and Lyubomirsky (2006) find in their study that it is the most frequently stated strategy people choose to become happier.

5.5.2 Monetary incentive-based messages

Monetary incentive-based messages contain temporary price reductions (i.e. promo-tions), whereas information-based messages and community-building messages are not linked to concrete tangible incentives (Blattberg, Briesch & Fox, 1995; Yi & Yoo, 2011; Bar-wise & Strong, 2002, Okazaki, 2005).

Nowadays, customers are not only facing information overflow, but also price pro-motion excess. Yi and Yoo (2011) state that concepts like EDLP (every day low price) BOGOF (buy one get one free), seasonal and “special” price discounts etc. have become more com-mon, repetitive and continuous than ever before. The different kinds of promotions are

(20)

of-ten used as a tool to attract poof-tential customers’ atof-tention, to introduce new products, to create traffic, to obtain future long-term income etc. (Blattberg, Briesch & Fox, 1995). Their effect on consumer behavior has been studied extensively, showing first and foremost that customers buy the product advertised soon after exposure to the message. Whereas most research has been conducted in the area of purchase behavior or the sales volume (Kotler, 2000; Kamakura & Russel, 1989), little research focuses on less tangible factors like attitude changes due to price reductions (Yi & Yoo, 2011).

According to the CLT (Trope, Liberman, & Wakslak, 2007), people tend to cognitively process location-based messages very detailed since the nature of location-based messages is low-level construals. As stated above, information can be processed more fluently if psy-chological distance and construal levels are consistent (Katz & Byrne, 2013). Furthermore, this message type is linked to a concrete benefit, showing customers precisely what they re-ceive whereas the consequences of revealing their privacy are more abstract and intangible.

Previous research claims that sales promotions affect brand attitude and image. In fact, it is shown that repetitive price reductions can backfire at long-term brand perfor-mance (Yi & Yoo, 2011). It is therefore derived that monetary incentive-based notifications also affect attitude towards location-based advertising, which determines the likelihood of acceptance (Limpf & Voorveld, 2015; Haghirian & Madlberger, 2005; Oh & Xu, 2003).

5.5.3 Information-based messages

(21)

Non-monetary promotions (referred to as information-based messages throughout this thesis) are the advertising tool used in this research (Mela, Gubta, & Lehman, 1997; Yi & Yo, 2011). In contrast to monetary incentive-based messages, information-based messages are neither linked to concrete incentives from the customers’ points of view, nor to easily measurable results from the marketers’ points of view. Brand image building activities, like disseminating information about the company and its services/products, focus on the long-term aspect and on cognitive dimensions by creating consumers’ favorable brand attitude. They indirect-ly increase brand value – a company’s main property (Yi & Yo, 2011; Mela, Gubta, & Lehman, 1997).

Bawden & Robinson (2009) explain in their paper how – over the past decades – the always-existing challenge of finding sufficient useful information turned into the challenge of handling information overload, mostly as a result of the spread of digital information via the internet. The most important task is now filtering and selecting information, rather than finding information (Bawden & Robinson, 2009), which emphasizes the importance of choos-ing the right and relevant content for the user.

In regards to fitness and nutrition a remarkable hype, especially stimulated by the spread of internet technology and social media (Rutsaert et al., 2013), could be observed over the past years (PriceWaterhouse Coopers, 2012). People are seeking information via the web more and more. In fact, a study conducted in 2013 by the Pew Research Center re-veals that 72 per cent of American internet users looked online for information regarding health (Fox & Duggan, 2013). When doing so, they detect trustworthiness by evaluating the message content and the information source (Wilson, 2007). Due to the unregulated online

(22)

environment and the excess of information found online, people are also confronted with misleading and conflicting information, which often leaves them overchallenged and con-fused (Goldberg & Sliwa, 2011; Theodosiou & Green, 2003; Spiteri, Cornish & Moraes, 2015; Fah, Hardikar, Fox & Mackay, 2014). Thus, consumers have difficulties to distinguish be-tween correct and incorrect fitness and health information found online (Modave, Shokar, Peñaranda & Nguyen, 2014). Especially peer-to-peer networks stimulate dissemination of incorrect and inaccurate advice (Flanagin & Metzger, 2003; Neuhauser & Kreps, 2003). Since the content on social media is often generated by non-professionals, the information spread is likely to be devoid of source credibility as well as imprecise and superficial (Rutsaert et al., 2013; Vance, Howe, & Dellavalle, 2009). In fact, a study unveils that 54 per cent of mation regarding health found on social media includes imprecise and incomplete infor-mation in 2007 (Tsai, Tsai, Zeng-Treitler and Liang, 2007). It is claimed that misinforinfor-mation in regards to fitness and nutrition causes not only confusion, but also brings people changing their attitude and behavior which may result in discontent, insecurity, aggravation of dis-comfort, injuries etc. (Spiteri, Cornish & Moraes, 2015; Burke, 2011; Goudreau, 2012).

This leads to the inevitable conclusion that correct and relevant information, espe-cially in regards to health and fitness, disseminated by credible sources is highly valued and appreciated. Compared to the message type monetary incentive-based, information-based messages have a long-term focus, whereas sales promotions affect short-term buying behav-ior and might actually backfire at long-term aspects like a person’s attitude. Acceptance is determined by attitude, which is of permanent character rather than formed upon short-term actions. Thus, it is hypothesized that information-based messages have a higher

(23)

posi-5.5.4 Formation of hypotheses of direct effects

When contrasting information-based messages to community-building tools, one has to pay attention to the fact that the setting of fitness centers, sports clubs, wellness insti-tutes etc. usually provides very good opportunities for building offline relationships and communities. As it is found that personal face-to-face interaction as well as telephone calls are preferred to online and social media interaction (Cummings, Butler, & Kraut, 2002; Peris et al., 2002; Mesch & Talmud, 2007), an online community-building tool may be a valuable complement, however it is proposed that relevant information from a credible source about the topic of interest is valued more and therefore leads to higher acceptance.

In order to explain whether monetary incentive-based or community-building mes-sages are suggested to lead to higher acceptance, the CLT shall be used. People tend to cog-nitively process location-based messages very detailed since the nature of location-based messages is low-level construals. As stated above, information can be processed more flu-ently if psychological distance and construal levels are consistent (Katz & Byrne, 2013). Tak-ing into consideration that the fitness and health community is interested in content regard-ing fitness and health, promotions for focal products and/or services are detailed and con-crete (presented in low-level construals). By contrast, receiving information that somebody from one’s friend list is closeby, does not overtly contribute to the primary fitness/health goal and is thus evaluated as more abstract and inconcrete (presented in high-level constru-als). Thus, it is expected that monetary incentive-based push notifications lead to higher ac-ceptance than community-building push notifications.

(24)

H1a: Location-based push notifications that are purely information-based are more accepted than those that include information about monetary incentives.

H1b: Location-based push notifications that are purely information-based lead to a higher acceptance compared to community-building push notifications.

H1c: Location-based push notifications that include monetary incentives lead to a higher acceptance than community-building push notifications.

5.6 Perceived fit

In social marketing, the concept of “fit” is often used to evaluate the success of CSR activities and can be defined as “the perceived link between a cause and the firm’s product line, brand image, position and/or target market” (Varadarajan & Menon, 1988). And more-over, in the context of brand extensions, perceived fit between the extension and the parent brand is an important factor to determine whether or not an extension will be successful. Actually, it has been identified as the most important success driver of a brand extension (Bottomley & Holden, 2001). This concept has been used in regards to advergaming cam-paigns on social networks where higher fit has been found to lead to higher perceived brand value (Okazaki & Yagüe, 2012). Transferring this knowledge to location-based message dis-semination, it offers a new perspective to evaluate acceptance. As stated above, consumers download apps following a specific purpose. If the message content is not in line with the app purpose, acceptance might be influenced.

(25)

mes-load an app for a specific reason (a goal they have) and build expectations towards the app. Thus, depending on the specific app purpose, they have a concrete and detailed anticipation of its performance which should be in line with the push notifications they get. Based on the exhaustive inferences made previously, especially people interested in health and fitness are believed to have high expectations towards the relevance, accuracy and completeness of information. In this thesis apps that provide support for a healthy and active lifestyle are in the spotlight. This leads to the assumption that the goal of users is to increase their health and fitness. They are supposed to evaluate information that supports them in achieving this goal as a better fit than information that does not transparently facilitate this objective.

Relevant and useful information constitutes a fundamental assistance in reaching the users’ goal. Information-based messages reach from information about provided facilities (info about fitness courses, personal training, healthy food offerings etc.) to information that encourages self-initiative (nutrition tips, food recipes, training advice, meditation tips etc.) infrastructure and facilities to fitness and nutrition tips. Thus, without information, the goal of improving one’s health and fitness can hardly be reached, which is why information-based messages are suggested to match the fit-criterion best.

Monetary incentive-based messages involve information about concrete tangible price reductions of specific offerings within the field of interest (Blattberg, Briesch & Fox, 1995; Yi & Yoo, 2011). Community-building messages offer the opportunity to connect with other gym members, which may lead to a higher loyalty and closer link to the company (Hur, Ahn and Kim, 2011; McAlexander, Schouten, & Koenig, 2002; Muniz and O’Guinn, 2001). However, based on the CLT (Trope, Liberman, & Wakslak, 2007), the psychological distance

(26)

of community-building messages is suggested to be larger since the benefit of connecting with others is more abstract when related to the goal respectively the reason why the app was downloaded in the first place. Since community-building messages are not primarily and transparently linked to achieving the goal of a better health and fitness condition, while monetary incentive-based messages are, it is hypothesized that the latter lead to a higher perceived fit.

H2a: The positive effect of information-based push notifications (vs. monetary-incentive-based push notifications) on acceptance is mediated by a higher perceived fit.

H2b: The positive effect of information-based push notifications (vs. community-building push notifications) on acceptance is mediated by a higher perceived fit.

H2c: The negative effect of monetary-incentive-based push notifications (vs. infor-mation-based push notifications) on acceptance is mediated by a lower perceived fit.

H2d: The positive effect of monetary-incentive-based push notifications (vs. communi-ty-building push notifications) on acceptance is mediated by a higher perceived fit.

H2e: The negative effect of community-building push notifications (vs. information-based push notifications) on acceptance is mediated by a lower perceived fit.

H2f: The negative effect of community-building push notifications (vs. monetary-incentive-based push notifications) on acceptance is mediated by a lower perceived fit. 5.7 Skepticism

Humans naturally ascribe substance to information they receive. (Johnstone & Graf-en, 1993; Kokko, 1997 in Dunham, 2011). Based on this finding, the signaling theory

(27)

identi-fies senders of a signal as manipulators whose goal it is to change the behavior of the receiv-ers, and receivers as mind-readers who base assumptions upon the piece of information re-garding the senders’ future behavior and, as a result, adapt their behavior accordingly (Krebs & Dawkins, 1984 in Dunham, 2011).

Likewise aim the attribution theories at explaining how people deduce, from restrict-ed evidence, unobservable properties or attributes concerning objects and organisms in their environment. They describe how humans are likely to dig deeper in regards to inter-preting events and making attributions rather than simply gathering objective data (Burn-krant, 1975).

Skepticism toward advertising has been studied thoroughly since it is an essential in-dication for acceptance of advertising. It is defined as a propensity to disbelieve informa-tional assertions of advertising, or in different words, as an ability to realize companies’ un-derlying selling intentions behind messages. Skepticism varies by different personality types, meaning more skeptical people assess marketing-related offers more negatively than people who are less skeptical, however it is a consistent characteristic of consumers which influ-ences their responses to advertising (Obermiller and Spangenberg, 1998, Knowles & Linn, 2004).

In view of CSR activities, it is found that especially a lower fit with the brand causes more negative evaluations of the firm and increases skepticism toward the underlying inten-tions of the firm. Consumers re-evaluate their image of the brand and are less likely to buy the brand in the near future. It is shown that skepticism is not elicited by companies being profit-oriented, but primarily by incongruence between the fit and the activities (Becker-Olsen et al., 2004).

(28)

Transferring these findings to the focal topic of interest, it is proposed that perceived fit and skepticism are serially mediating the direct relationship of message type and ac-ceptance of location-based advertising in a way that higher fit leads to lower skepticism, which in turn leads to higher acceptance, and vice versa:

H3a: The positive effect of information-based push notifications (vs. monetary-incentive-based push notifications) on acceptance is serially mediated by a higher perceived fit and lower skepticism.

H3b: The positive effect of information-based push notifications (vs. community-building push notifications) on acceptance is serially mediated by a higher perceived fit and lower skepticism.

H3c: The negative effect of monetary-incentive-based push notifications (vs. infor-mation-based push notifications) on acceptance is serially mediated by a lower perceived fit and higher skepticism.

H3d: The positive effect of monetary-incentive-based push notifications (vs. communi-ty-building push notifications) on acceptance is serially mediated by a higher perceived fit and lower skepticism.

H3e: The negative effect of community-building push notifications (vs. information-based push notifications) on acceptance is serially mediated by a lower perceived fit and higher skepticism.

H3f: The negative effect of community-building push notifications (vs. monetary-incentive-based push notifications) on acceptance is serially mediated by a lower perceived

(29)

6 Conceptual Model

7 Research methodology

7.1 Research and survey design

To examine the relationship of the focal variables a deductive research approach was used. This research was conducted as a cross-sectional survey since conclusions about a population of interest at a specific point in time were made (Lavrakas, 2008). The

question-naire was designed via the online survey platform qualtrics, which allowed creating an

un-tracked open link to the survey that was protected against indexation by search engines. For measurement of the variables, 7-Point-Likert scales were used to make it most convenient for the respondents and to increase the response rate (Jamieson, 2004).

The survey was made available in English and Dutch because of language diversity in the sports club, most members being Dutch and a minority being foreign. Back-translation

H1 H2

H3

H2 H3

(30)

was executed to prevent grammatical and syntax errors. For that reason, the original English version was translated to Dutch and the Dutch version was translated back to English by a different person, so that the two English versions could be compared, which were semanti-cally the same, just differed in their wordings. The questionnaire comprised general instruc-tions as well as assurance of anonymity (Appendix A) and the data was collected within two weeks.

The research was conducted through a 3 (information-based, monetary

incentive-based and community-building message) x 2 (high fit vs. low/moderate fit) factorial be-tween-subjects experimental design. Thus, six different scenarios were created. In order to allow representative results, the goal was to achieve a sample of 50 persons per treatment, i.e. 300 in total. The manipulation of high fit vs. low/moderate fit was created to generate spread in the variable perceived fit. To get a sense of what is being perceived as high/moderate/low, the six scenarios were pre-tested among a sample of 30 students of the University of Amsterdam as well as friends and family of mine.

Experimental vignette studies were used since they allow for testing a large number of scenarios which would be too broad to be presented to each respondent. The respond-ents were randomly assigned to the six different short depictions of situations, referred to as vignettes (Atzmüller & Steiner, 2010), and asked to evaluate these vignettes in regards to perception of fit, skepticism, acceptance and word of mouth. Word of mouth is not part of the conceptual model, but was measured to find out more about the likelihood of people passing experiences with location-based services to friends and family since WOM is

(31)

denot-7.2 Data collection and sample

The study was conducted in cooperation with a sports center in Amsterdam Oost that offers a complimentary mobile application for its members, which many use as a sport-ing gadget. The app provides, for instance, workout schedules, information regardsport-ing group fitness classes, offers for inviting friends for a complimentary trial etc. A non-probability (convenience) sample was used since only members of the gym have access to this applica-tion. The e-mail addresses of all active members were used to introduce and explain the study and to send a link to the qualtrics survey. Furthermore, people were approached per-sonally in the gym and asked to fill in the survey with two tablets and one smartphone. Two e-mail reminders were sent to those who had not filled in the survey by that time. The par-ticipation was voluntary, however, as an incentive, participants were offered a free smoothie or a protein shake after they had filled in the survey. To assure that people did not respond multiple times, the e-mail link was linked to their IP address. Also, they had to state their name at the reception to get the beverage and this was marked in their account.

In total, the 2.135 active members received an e-mail and 117 were approached per-sonally in the gym. Ultimately, out of 395 respondents 381 completed the survey, showing completion rates of approximately 95 % from e-mail recipients and 100 % from personal in-terviewees as well as response rates of approximately 15 % from e-mail recipients and 100% from personal interviewees. However, it has to be taken into account, that some members, who got an e-mail, filled in the survey in the sports center, not via their e-mail link. Thus, the response rate of e-mail participants is slightly biased.

(32)

The respondents are pretty active (70 % going to the gym 1-3 times per week, 20 % 4-6 times per week) and show high levels of education (34 % HBO, 11 % academic Bache-lor’s, 35 % academic Master’s and 6 % PhD/Doctoral Degree, meaning that only 14 % have less than HBO degree). 97 % of the respondents are smartphone users, whereas only 44 % have the focal app installed on their phone.

7.3 Pre-test

As stated above 30 friends, fellow students and family members of mine were asked to individually evaluate every of the six scenarios in regards to fit with an app of a sports center. Table 1 shows the scenarios that were tested and the subsequent results.

Table 1: Scenarios

Intended condition Text Fit in %

Information-based message,

fit (directly to the gym related)

Imagine you get push notifications about the gym's new ser-vices when you enter, leave or pass by the gym, e.g. infos about an event (ladies night etc.), a new workshop about kettlebells held on this day or about new group classes, new trainers, new workshops, upcoming events, new services in the restaurant etc.

100 % high fit

Information-based message, no fit (not directly to the gym re-lated)

Push notifications about health-related topics, when you en-ter, leave or pass by the gym, e.g. when leaving you get a meal tip with the recipe depending on the type of workout you just did and what time of the day it is; or tips on how to improve your overall health by meditating etc.

90 % moderate fit 10 % high fit

Monetary

incentive-based message, fit (directly to the gym related)

Push notifications about special offers when you enter, leave or pass by the gym, e.g. buy one protein shake and get an-other one for free or get 20 % off the Fresh & Go meal of the day etc.

3 % moderate fit 97 % high fit

Monetary

incentive-based message, no fit (not directly to the gym related)

Push notifications about special offers when you enter, leave or pass by a Sports World partner, e.g. you pass by a G-Star shop and get a notification that you get 5 % off a specific sportswear-product.

13 % low fit 87 % moderate fit

Community-building message,

fit (directly to the gym related)

Imagine you can connect with other gym members via the app and that you receive push notifications when someone from your friend list is at the gym as well so that you can workout together.

100 % high fit

Community-building message, no fit (not directly to the gym

re-Imagine you can connect with other gym members via the app and that you receive push notifications when someone from your friend list is near you when you are at a

restau-33 % low fit 67 % moderate fit

(33)

7.4 Measurement instruments

The constructs used in this thesis were derived from validated constructs (see Ap-pendix 2) and adapted to fit the circumstances of the study. The participants were asked to mark the extent to which they agreed or disagreed to the scale items, based on a Likert-Scale from 1 “strongly agree” to 7 “strongly disagree”. The scale for the dependent variable acceptance was adapted from Merisavo et al. (2007), for word of mouth from Yu, Zo, Choi & Ciganek (2013), for the first mediator perceived fit from Okazaki & Yagüe (2012) and for the second mediator skepticism from Yoon, Giirhan-Canli & Schwarz (2006).

Control variables

Five control variables were included and held constant to account for differences among the respondents’ demographics and behavioral as well as attitudinal habits: frequen-cy of exercising, whether or not the focal app is downloaded, gender, age and level of educa-tion.

7.5 Biases

To lower the potential risk of common method bias, that might endanger the validity of the results, answer validity was cross-checked by integrating a reversed scale (genuine-ness instead of skepticism).

(34)

To diminish the risk of social desirability, the survey was sent on behalf of the sports center with the request for their members’ honest and sincere answers to improve their ser-vices.

Reasons for non-response could be non-reachability or unwillingness to participate. In this case 85 % of the e-mail recipients did not respond which may distort the results. However, the sample was fairly large, which mitigates the non-response bias.

7.6 Data analysis

After the data was imported from qualtrics to SPSS (including the viewing order of randomized participants), data cleaning and preparation for statistical analyses were the next step. Since all the scale dimensions in the survey were from 1 = strongly agree to 7 = strongly disagree, the dimensions of Acceptance, Perceived fit and WOM were reversed by adding 1 to the highest point of the scale and subtracting the respective value. The scale for skepticism was not reversed since skepticism was evaluated covertly, asking respondents to assess statements regarding sincerity/genuineness rather than skepticism. The 14 incom-plete cases were removed so that the dataset eventually consisted of 381 cases. There were three cases showing missing data that were dealt with by “listwise deletion” to make sure that only entirely filled-in cases were analyzed.

7.6.1 Normality check

(35)

positive, it means a clustering of scores at the low end, if they are negative, there is a clus-tering at the high end. Positive kurtosis values indicate that the distribution is clustered in the center and kurtosis values below 0 show that a distribution is pretty flat (Tabachnick & Fidell, 2001). The descriptives of this study show that acceptance is positively skewed whereas perceived fit and skepticism are rather normally distributed. Also, a slightly platy-kurtic distribution can be seen for acceptance and perceived fit whereas a minorly leptokur-tic distribution can be found for skepleptokur-ticism. This might be explained, for example, by to the use of a convenience sample or the cross sectional nature of the survey. There are different opinions on kurtosis and skewness acceptance levels. In this thesis indices for limits of +/- 2 were used (Trochim & Donnelly, 2006; Field, 2000 & 2009; Gravetter & Wallnau, 2014). If the levels are higher than the suggested limit, the scores can be transformed, what is also controversial due to the loss of “real data”. As can be seen in table 2, all values for kurtosis and skewness are within the proposed limit in this study, which leaves transformation irrele-vant.

Table 2: Normality Check

Variable Skewness Kurtosis

Acceptance 0,189 -1,120

Perceived Fit -0,417 -0,285

Skepticism 0,429 -0,239

Word of Mouth 0,238 -1,063

7.6.2 Confirmatory factor analysis

In order to examine the similarities between the variables, a confirmatory factor analysis was done. First, the KMO and Bartlett’s Test of Sphericity was used to find out

(36)

whether or not the data is suited for a factor analysis. This test measures sampling adequacy for each variable in the model and for the complete model. Since the KMO amounts to 0,92 and values between 0,9 and 1,0 are identified as very good, and the Bartlett’s Test results in p= 0,000, the factor analysis could be run. An initial analysis showed Eigenvalues over Kai-ser’s criterion larger 1 and in combination explained 74,26 % of the variance. It showed that the four variables can be summarized into two latent variables (Eigen values= 8,893 and 1,503). There is high similarity between skepticism and perceived fit and high similarity be-tween acceptance and word of mouth. Furthermore, skepticism was divided into two di-mensions, which simply reflects the nature of the original scale by Yoon, Giirhan-Canli & Schwarz (2006). Table 3 shows the single items that were put together.

Table 3: Pattern matrix factor analysis

Component

1 2

WOM: If I receive this kind of location-based push notifications, I will say positive things about location-based services to others.

0,943 WOM: If I receive this kind of location-based push notifications, I will highly

rec-ommend location-based services to others.

0,93 WOM: If I receive this kind of location-based push notifications, I will encourage

using location-based services to smartphone users who want to use new smartphone services.

0,906

Acc: If I receive this kind of location-based push notifications, I am willing to re-ceive further location-based push notifications in the future.

0,9 Acc: If I receive this kind of location-based push notifications, I will read all the

push-notifcations I receive in the future.

0,898 Acc: If I receive this kind of location-based push notifications, I feel positive

about location-based push notifications.

0,888 Skep Dimension 1: I believe that Sports World tries to make a good image of the

company by sending me this kind of push notifications.

0,918 Skep Dimension 1: I believe that Sports World tries to improve its existing image

by sending me this kind of push notifications.

0,914 PerFit: Sports World and this type of notification have similar images. 0,7 PerFit: It makes sense to me that the Sports World app disseminates this type of

notification.

0,662 PerFit: The ideas I associate with the Sports World app are related to this type of

notification.

0,649 PerFit: The Sports World app and this type of notification fit together well. 0,61 Skep Dimension 2:I believe that Sports World has genuine concerns about me

and interest in me when sending me this kind of push notifications.

(37)

sending me this kind of push notifications. Extraction Method: Principal Component Analysis. Rotation Method: Oblimin with Kaiser Normalization. Rotation converged in 7 iterations.

7.6.3 Reliability check

Since validated scales were used and Cronbach’s alpha values higher than 0,7 are considered as sufficient, the values 0,953 for acceptance, 0,920 for perceived fit, 0,855 for skepticism and 0,969 for WOM show that the internal validity of every scale is quite high.

7.6.4 Randomization check

To evaluate whether the diversity of groups was roughly the same for every scenario, a randomization check was done. Pearson Chi-Square and Cramer’s V calculations were used to confirm that there are no significant differences between the groups based on gender, age, education, how often people practice sports and whether or not they have the focal app on their phone. It is shown, for instance, that in scenario 1 there were 62 % female vs.

38 % maleand that in scenario 6 there were 68 % who do not have the app downloaded vs.

32 % who have it downloaded. So, the means differ, but there is no significance (p=0,502).

7.6.5 Manipulation check

An independent sample t-test was carried out to monitor the manipulation of high fit vs. low/moderate fit in contrast to the measured “perceived fit”. The results show that the variable fit (that was created due to the manipulated conditions) significantly correlates with the measured variable perceived fit. However, there is only a slight difference between the “no fit” and the “fit” condition. As can be seen in the tables 4 and 5 the mean difference is

(38)

only half a scale point. As mentioned above, the manipulation was mainly executed to create spectrum in the variable perceived fit. For further analysis the measured variable “perceived fit” will be used.

Table 4: Manipulation check group statistics

Fit N Mean SD

PercFit

no fit 187 3,8797 1,50879 fit 191 4,3940 1,33927

Table 5: Manipulation check independent samples test

Independent Samples Test

Levene's Test

(for Equality of Variances)

t-test

(for Equality of Means)

F Sig. t df Sig. (2-tailed) Mean Diff. Std. Error Diff. 95% Confidence Interval of the Difference Lower Upper Perc. Fit Equal variances assumed 3,357 0,068 -3,507 376 0,001 -0,51430 0,14666 -0,80268 -0,22592 Equal variances not assumed -3,502 368,803 0,001 -0,51430 0,14685 -0,80306 -0,22554 7.6.6 Correlation matrix

A correlation matrix (table 6) was created in order to recognize direct relationships and to anticipate the potential of the proposed mediation model due to strong or weak cor-relations between the focal variables. Since Likert scales are commonly considered as “quasi-metric”, the Pearson’s correlation matrix was used. Nevertheless, also Spearman’s correla-tion was computed and depicts very similar results.

(39)

Acceptance

Acceptance shows a significant strong positive correlation with perceived fit (r= 0,692) and a significant strong negative correlation with skepticism (r= -0,611). As already investigated in the factor analysis, there is a very strong relationship between acceptance and word of mouth (r= 0,861).

Also, a significant weak downhill relationship can be seen for acceptance and com-munity-building messages (r= 0,12), which connotes that comcom-munity-building messages lower acceptance of location-based services. There is no significant relationship between acceptance and information-based messages or monetary incentive-based messages.

Furthermore, a significant weak negative correlation between acceptance and age is noticed (r= -0,215), meaning that younger people have a slightly higher acceptance than older ones. The significant weak positive correlation between acceptance and app (r= 0,118) demon-strates that the ones who have the app installed on their phone show a slightly higher ac-ceptance.

Perceived fit

Additionally to the significant strong positive correlations with acceptance and word of mouth (r= 0,655), a significant high negative relationship with skepticism is identified (r= -0,701), which means that a lower perceived fit leads to higher skepticism. Also, the ma-nipulation check is confirmed by the significant weak positive relationship between per-ceived fit and the manipulated fit (r= 0,178). Lastly, age (r= -0,118) and education (r= -0,13) show a significant weak negative relationship with perceived fit, indicating that younger

(40)

people and lower educated people perceive a higher fit between the message type and the app purpose.

Skepticism

As stated above, skepticism correlates negatively with acceptance, perceived fit and word of mouth. Besides these relationships there are no significant correlations with other variables.

Message types

Out of the three types, only the community-building message type shows a correla-tion with the dependent variable acceptance (- 0,120) and none of the message types shows significant relationships with any of the two mediators.

Furthermore, it can be derived that people who have the app installed are younger, since app and age show a negative correlation (r= -0,193).

(41)

Table 6: correlation matrix

Accept PercFit Skep WOM Fit Info Mon Com Gender Age Edu Sports App

Accept Pearson Corr. 1 0,692** -0,611** 0,861** 0,098 0,092 0,029 -0,120* 0,062 -0,215** -0,081 -0,015 0,118*

Sig. (2-tailed) 0 0 0 0,056 0,073 0,58 0,02 0,227 0 0,115 0,775 0,022

N 378 378 378 378 378 378 378 378 378 378 378 378 378

PercFit Pearson Corr. 0,692** 1 -0,701** 0,655** 0,178** 0,082 -0,053 -0,029 0,085 -0,118* -0,130* -0,004 0,068

Sig. (2-tailed) 0 0 0 0,001 0,111 0,304 0,568 0,098 0,021 0,012 0,932 0,185

N 378 378 378 378 378 378 378 378 378 378 378 378 378

Skep Pearson Corr. -0,611** -0,701** 1 -0,614** -0,058 -0,088 0,071 0,018 -0,051 0,09 0,098 0,033 -0,01

Sig. (2-tailed) 0 0 0 0,264 0,086 0,169 0,729 0,323 0,079 0,056 0,52 0,851

N 378 378 378 378 378 378 378 378 378 378 378 378 378

WOM Pearson Corr. 0,861** 0,655** -0,614** 1 0,03 0,061 0,037 -0,097 0,053 -0,238** -0,104* -0,017 0,105*

Sig. (2-tailed) 0 0 0 0,567 0,235 0,473 0,059 0,307 0 0,044 0,736 0,042

N 378 378 378 378 378 378 378 378 378 378 378 378 378

Fit Pearson Corr. 0,098 0,178** -0,058 0,03 1 -0,015 0,019 -0,004 -0,086 -0,032 0,01 -0,01 0,057

Sig. (2-tailed) 0,056 0,001 0,264 0,567 0,773 0,712 0,938 0,095 0,535 0,852 0,839 0,267