What effect of price promotion, message frequency, location and personalisation on purchase intentions is largest when customers receive mobile marketing messages in-store?

What effect of price promotion, message frequency, location and

personalisation on purchase intentions is largest when customers

receive mobile marketing messages in-store?

Doeke Stel

Dual Award Master of Science

Advanced International Business Management and Marketing

Submission date: 04 December 2017

Newcastle University Business School: Supervisor: Dr. A. Javornik

Student number: 160747748

University of Groningen, Faculty of Economics and Business: Supervisor: Dr. B.J.W. Pennink

Master Dissertation:

What effect of price promotion, message frequency, location and

personalisation on purchase intentions is largest when customers

receive mobile marketing messages in-store?

Dual Award Master of Science

Advanced International Business Management and Marketing

Supervisors:

Newcastle University Business School: Dr. A. Javornik University of Groningen, Faculty of Economics and Business: Dr. B.J.W. Pennink

Student number:

Newcastle University Business School: 160747748 University of Groningen, Faculty of Economics and Business: 3003337

Doeke Stel Parklaan 26

9724 AP Groningen, the Netherlands

Abstract

This research studies the influence of mobile in-store marketing on consumers purchasing intentions in the retailing environment. The influence of four value drivers in mobile marketing (price promotion, message frequency, location, and personalisation) are investigated by using mobile in-store push messaging to measure the effect on customer purchase intentions. Mobile marketing is growing in importance in the marketing sector due to the ubiquity of smartphones, yet relatively few research has been conducted into the matter. In this study, a theoretical framework is developed and empirically examined in a quasi-experimental study design. A survey based on twelve hypothetical scenarios was conducted, resulting in a valid sample of 419 respondents. Each participant was randomly distributed to a treatment condition, which each contained a different aspect of a mobile marketing value driver. To identify the influence of the value drivers on purchase intentions, ANCOVA analyses are performed while controlling for influences regarding privacy concerns, perceived intrusiveness and general attitude towards mobile services.

The results indicate several important findings. Price promotion and personalisation appeared to have a statistically significant effect in driving purchase intentions, whereas message frequency and location did not. Additionally, an interaction effect between location and personalisation has been identified, in so far as that location negatively influences the positive effect of personalised mobile messaging. Meaning, being geographically close to a product promoted via a personal mobile message, lowers the positive effects of the personalised message on purchase intentions.

The main limitation of this study is the quasi experimental design based on hypothetical scenarios. Future research should therefore focus on field studies with an observational design.

Acknowledgements

This dissertation project marks the end of my Master studies at the University of Groningen and Newcastle University. I would like to express my sincerest gratitude to dr. A. Javornik from Newcastle University and dr. B.J.W. Pennink from the University of Groningen for their advice, support and feedback during the process.

Table of contents

1.1 Research background & Research Rationale 8

1.2 Research objective and questions 12

1.3 Dissertation structure 12

2. Literature review 13

2.1 Mobile marketing 13

2.2 Purchase intentions 16

2.3 Equity theory in mobile marketing 16

2.4 Hypothesis development 17

3. Methodology 22

3.1 Research paradigm 22

3.2 Research design: Quasi experimental design 23

3.3 Data collection 30 3.4 Data analysis 31 3.5 Ethical considerations 31 4. Findings 32 4.1 Preliminary analyses 32 4.2 Hypotheses testing 41 4.2.1 ANCOVA assumptions 41

4.3 Overview of hypotheses findings 45

5. Discussion 46

5.1 Price promotion 46

5.2 Message frequency 46

5.3 Location 47

5.4 Personalisation 48

5.5 Location and personalisation 48

5.6 Mobile in-store marketing messages and purchase intention 49

6. Conclusion and recommendations 51

6.1 Contributions 51

6.2 Managerial implication 51

6.3 Research limitations and future research 51

Bibliography 53

Appendices 64

Appendix A: Scenario options 64

Appendix B: Survey scenario 1 66

Appendix C: Demographic sample characteristics 74

Appendix D: Reliability analysis 76

Appendix F: Assumption testing – homogeneity of variance test 80 Appendix G: Assumption testing – matrix of scatterplots 81 Appendix H: Assumption testing – homogeneity of regression slopes test 82

Appendix I: ANCOVA Output tables 83

List of tables

Table 3.1: Overview of variables ... 25

Table 3.2: Overview constructs and measurement items of independent variables ... 27

Table 3.3: Overview constructs and measurement items of dependent & control variables .... 29

Table 4.1: Overview of respondents per scenario ... 33

Table 4.2: Sample characteristics ... 34

Table 4.3: Descriptive statistics constructs ... 35

Table 4.4: Cronbach's alpha ... 36

Table 4.5: Factor loadings per item ... 39

Table 4.6: AVE, CR, and correlation matrix and √AVE ... 40

Table 4.7: ANCOVA: Test of between subject effects ... 45

Table 4.8: Overview hypotheses findings ... 45

Table A.1: Detailed sample characteristics... 74

Table A.2: Corrected item-total correlation per construct ... 76

Table A.3: Cronbach's alpha per construct, per scenario ... 77

Table A.4: Extracted communalities per item, per scenario ... 78

Table A.5: Levene’s test of equality of error variances ... 80

Table A.6: Homogeneity of regression slopes test ... 82

Table A.7: ANCOVA test of between-subjects effects message frequency ... 83

List of figures

Figure 2.2: Conceptual model ... 21

List of abbreviations

AVE Average Variance Extracted ANOVA Analysis of Variance ANCOVA Analysis of Covariance Apps Applications

CLT Construal Level Theory

DSMM Digital Social Media and Mobile DV Dependent Variable

1. Introduction

1.1 Research background & Research Rationale

Since the year 2000, marketing has transformed severely due to the growth in importance of digital, social media and mobile (DSMM) marketing. The use of smart mobile devices by consumers has grown at a fast pace and influenced the behaviour of consumers in different market settings (Lamberton and Stephen, 2016). Over 80 per cent of shoppers in the United States of America (USA) use a smart mobile device for shopping, even within retail stores. The drastic increased importance of the role of smart mobile devices in the shopping process results from increased functionalities of smart mobile devices and awareness of mobile functions (Byrne, Holmes and Rowley, 2013). Smart mobile devices refer to handheld electronic devices such as mobile phones, tablets, phablets, smartwatches and handheld internet access devices (Shankar, Venkatesh and Naik, 2010). Smartphones have a dominant position in the mobile device market, they have become essential to consumers and many cannot seem to do without. Consumers see smartphones as more than a personal device, it is an extension of their individuality and personality. This widespread adoption of smartphones has created significant marketing opportunities for the retailing industry to reach consumers (Barutçu, 2007; Byrne, Holmes and Rowley, 2013; Grant and O’Donohoe, 2007; Grewal et al., 2016; Roach, 2009). Yet, it also poses many challenges to companies as they have to adapt to the undisputable rise of mobile behaviour of consumers. Mobile communication becomes more and more important, to such an extent that mobile marketing has the potential of becoming the primary marketing channel (Bernier, 2014; Vatanparast and Ali Hasan, 2010).

grow from $29 billion in 2015 (49 per cent of digital ad spending) to $66 billion by 2019 (72 per cent of digital ad spending), in the USA alone (eMarketer, 2015). Globally, the mobile advertising market is expected to grow similarly in value and will account for more than 50 per cent of total digital advertising spending in 2016, surpassing desktop advertising spending (Grewal et al., 2016).

The remainder of this study will focus on mobile advertising. More specifically, the focus will be on smartphone advertising as smartphones have a dominant position in the mobile device market, accounting for the majority of mobile advertising spending (Criteo, 2015; Kim et al., 2017). When referring to mobile advertising, this will focus at smartphone advertising. Many sectors are affected by the rise of mobile technologies, but it is argued the retail industry is affected most due to the relatively high price sensitivity (Ferraro, Inman and Winer, 2009; Hofacker et al., 2016; Kollat and Willet, 1967; Shankar et al., 2016). Previous research indicated that retail consumers often use their smartphone for varying in-store activities such as searching for price-comparisons, product comparisons or finding a discount voucher or promotion (Byrne, Holmes and Rowley, 2013; Shankar, Venkatesh and Naik, 2010).

Mobile advertising offers many advantages over regular digital advertising. As stated before, mobile devices are personal communication tools which are extremely individualised by the user. This allows new strategies to be developed by marketers as consumers can be approached in a customised manner more directly and frequently, creating new opportunities for retailers to target their communication (Aguirre et al., 2016; Barutçu, 2007; Byrne, Holmes and Rowley, 2013; Grant and O’Donohoe, 2007; Roach, 2009). Important features that are unique to mobile advertising are the use of GPS, Wi-Fi, beacon, Near Field Communication (NFC), and latitude and longitude coordinates for LBM. These features allow firms to send targeted promotions which can be used to increase advocacy, loyalty and incentivise purchasing, making mobile promotions highly relevant for marketers (Andrews et al., 2016). Combining these new advertising features and utilising them in the in-store environment, is named mobile in-store marketing, or mobile in-store advertising.

smartphone shoppers uses their device in-store during shopping (Google, 2013; Shankar et al., 2016). Further, one in three shoppers uses their mobile to find products in-store and in-store price comparisons is the most common in-store smartphone activity in retail stores, performed by 36 per cent of customers (Google, 2013).

In the Fast Moving Consumer Goods (FMCG) industry, 50 to 60 per cent of the purchase decision are made in-store (Ferraro, Inman and Winer, 2009; Kollat and Willet, 1967). Therefore, retailers can benefit immensely by engaging with their customer at such moments in order to persuade them into purchasing. Hence, mobile marketing is expected to have a major impact on purchase intentions.

Previous research has mainly focused on the influence of mobile marketing in non-retail environments. Until date, only three studies have investigated the impact of mobile in-store advertising in retail outlets (Bues et al., 2017; Hui, Huang, et al., 2013; Klabjan and Pei, 2011). The research of Klabjan and Pei (2011) investigated the influence of mobile in-store communication in the customer loyalty card context by tracking previous purchasing behaviour. They found that tracking technology was still in its infancy at the time and they were unable to target customers in-store during the actual shopping experience. Therefore, customers could only be targeted either before or after the shopping process. Furthermore, Klabjan and Pei (2011) propose a one-to-one marketing model which potentially leads to a 60 per cent revenue increase due to the ability of offering personalised coupons targeted during the shopping process. Overall, their research proposes personalisation is a value driver of mobile in-store marketing.

Hui et al. (2013) studied the influence of shoppers’ in-store travel distance on unplanned spending. Travel distance of customers can be increased by either relocating popular product categories or utilising mobile in-store marketing techniques. The effectiveness of relocating up to three product categories in-store increased unplanned spending by 7.2 per cent. However, the field experiment showed that mobile promotions can increase in-store travel distance of customers and subsequently increase the amount of unplanned purchasing by 16.1 per cent. Additionally, the research demonstrated that sending promotions concerning products geographically far away from the recipient, resulted in more unplanned purchasing compared to when sending promotion concerning products nearby the recipient. The study therefore proposes that coupons, or price promotions, and location are value drivers of mobile in-store marketing.

intentions. Their models, however, do each only include one factor in their analysis, not controlling for the effect of multiple value drivers in a mobile in-store marketing campaign. This in contrast to Bues et al. (2017), who was the first to conduct a study aimed at identifying how mobile in-store advertising influences purchase intentions by using multiple value drivers. This study took into account multiple factors, based on the previous studies (price promotion, location, and personalisation) and studied the effect of these value drivers on consumer purchase intentions of wine in retail stores.

This study was performed in Germany and the findings indicate that all value drivers increase the intention to purchase, but, in contrast to expectations, price promotions were the least important value driver. It appeared that the location of receiving an ad was the strongest driver and personalised messages the second most important driver (Bues et al., 2017). This study, however, has several limitations. Bues et al. (2017) did not work with an international sample containing data from multiple countries, one limitation therefore is whether the findings are specific to Germany or whether the results are more universal and can be generalised. Further, a fourth value driver of mobile in-store marketing is suggested: message frequency, based on the studies of Bues et al. (2017), Dickinger and Kleijnen (2008) and Hui et al. (2013).

1.2 Research objective and questions

In general, this research aims to study the impact of the value drivers price promotion, message frequency, location and personalisation on purchase intentions with in-store mobile marketing messages. Furthermore, it aims to identify which factor has the largest effect. Therefore, the following research objective is formulated:

• What effect of price promotion, message frequency, location and personalisation on purchase intentions is largest when customers receive mobile marketing messages in-store?

To answer the main research objective, the following research question is addressed:

• What is the effect of price promotion, message frequency, location and personalisation on purchase intentions when customers receive mobile marketing messages in-store?

The research questions will be answered by first providing an in-depth literature review of the main concepts of mobile marketing, purchase intention, the value drivers and the relationships between them. This will result in a theoretical framework on which the hypotheses are based. To test the hypotheses, primary data will be collected by setting up a survey, which will be distributed online. After collecting the data, it will be screened and cleaned in preparation for analysis. Next, the data is tested on reliability and validity followed by ANCOVA tests to analyse the data.

1.3 Dissertation structure

2. Literature review

This chapter provides an overview of relevant literature and theories for the research. The main concepts are defined and the relationships between the concepts is explained. Subsequently, the hypotheses are built on this.

2.1 Mobile marketing

Literature offers a multitude of definitions for both marketing and mobile marketing. For this study, marketing is defined as ‘the process of planning and executing the conception, pricing, promotion, and distribution of goods, services, and ideas to create exchanges that satisfy individual and organisational goals’ (Haghirian, Madlberger and Tanuskova, 2005). Performing such activities aimed at targeting mobile devices can therefore be considered as mobile marketing since it is the instantiation of marketing placed in the mobile marketing context. Although there are many perspectives on this, it is generally agreed upon that traditional marketing involves one-way communication, whereas mobile marketing allows interaction and thus involves two-way communication. As such, drawing from multiple sources in the literature, mobile marketing is defined as: using interactive wireless media (mobile technology) to communicate with customers in an interactive and relevant manner promoting goods, services and ideas, thereby generating value for all stakeholders (Balasubramanian and Shankar, 2009; Gana and Koce, 2016; Haghirian, Madlberger and Tanuskova, 2005; Mobile Marketing Association, 2009; Stanoevska-slabeva et al., 2017). Mobile marketing enables firms to interact and communicate with customers in a personalised way via the mobile web, mobile applications (apps), mobile advertising, short message services (SMS) and multimedia message services (MMS) (Gana and Koce, 2016). The importance of mobile marketing increased significantly among firms as a result of growing competition in the global market (Vivek, Beatty and Morgan, 2012). Mobile marketing has grown into an important platform for creating, and building customer engagement, for a company or brand using text messages, mobile advertising, permission-based marketing, user-generated content, mobile commerce and the delivery of mobile content (Balasubramanian and Shankar, 2009; Dickinger, Murphy and Scharl, 2005; Leppäniemi, 2008).

more technologically advanced, as a result of smart technologies, smartphones became more ubiquitous which allowed marketers to develop a wider range of services and techniques. These techniques primarily focused at pull-based practices and more advanced push-based practices, using apps. The main advantage of using apps over SMS marketing is the extensive data which can be collected via the app, allowing the marketer to continuously improve its mobile marketing messages (Almunawar, Anshari, Susanto, and Chen, 2015; Geron, 2009; Selvi, 2014).

Pull and push marketing in the mobile context differ from their meaning in traditional marketing. Traditionally, the push approach concerns using trade promotions to get a product or brand sold in the retail environment and pull regards mass advertising to address consumers generating a demand. In the mobile context, push communication comes from the marketer and is outbound, whereas pull is inbound and initiated by the customer (Andrews et al., 2016; Harmon and Unni, 2007). The terms are defined as following:

Push messages relate to any content sent by marketers to a smartphone at a time when the marketer wishes so. The push message is sent to a customer based on his or her location and can be tailored to specific product/service preferences. Push messages give marketers more control over the advertising and promotions flow (Harmon and Unni, 2007). Pull advertising in mobile promotions is sending any message to the customer upon request of the customer. It is delivered to the customer at a certain location only when it is explicitly requested for. Thus, it is the customer who initiates the request (pull) for advertisements or promotions for specific product categories near the customers’ location (Harmon and Unni, 2007). Figure 2.1 illustrates a push vs a pull strategy in mobile advertising.

Mobile advertising is the fastest growth segment in mobile marketing, with in-app advertisements taking the lead; it is expected this type of advertisement increases in threefold from USD 69 billion in 2015 to USD 196 billion in 2019 (eMarketer, 2015; Smith, 2013). Mobile advertising is defined as: ‘utilising advanced digital technologies to reach consumers on their mobile device’ (Andrews et al., 2016; Grewal et al., 2016). An important and unique feature of mobile devices is the support of location-based services (LBS). LBS can be defined as services that depend on and are enhanced by information about the position of a mobile device (Harmon and Unni, 2007). Such services use geographical position information which is send by the mobile device through GPS, Wi-Fi, Near Field Communication (NFC) or transmitters as iBeacon (Jones and Ryan, 2017; Quercia et al., 2010). LBS can be utilised by marketers to use the location context and combine it with other technologies as customer preferences resulting in (highly) targeted messages (Rowles, 2014). This type of marketing is known as location-based marketing (LBM). It is defined as ‘any application, service or campaign that incorporates the use of geographic location to deliver or enhance a marketing message’ (Johnson, Lewis and Reiley, 2016).

LBM features allows companies to send relevant and timely content to (potential) customers’ mobiles, and includes all aspects of the marketing mix in mobile marketing regarding the location setting (Gazley, Hunt and McLaren, 2015). More specifically, this is known as location-based advertising (LBA). LBA is defined as: ‘targeted advertising initiatives delivered to a mobile device from a sponsor that is specific to the location of the consumer’ (Harmon and Unni, 2007). For the in-store environment, indoor positioning systems are used by firms to target customers. These systems can target customers when near the store or entering the store. This is the moment where a final marketing action can be initiated before customers’ purchase decision is made. As stated in section 1.1, in the FMCG industry, 50 to 60 per cent of the purchase decisions are made in-store (Ferraro, Inman and Winer, 2009; Kollat and Willet, 1967). Therefore, reaching consumers at this stage can be lucrative, resulting in marketers’ willingness to make considerable investments to reach shoppers while shopping (Hui, Huang, et al., 2013). In-store advertising can trigger purchases as only a portion of consumers’ purchases are planned at the item or brand category level, the remainder are planned at a product-category level or made spontaneously (Heilman, Nakamoto and Rao, 2002).

advertisements deemed too personal or intrusive resulting in raising concerns regarding privacy. (Grewal et al., 2011; Shankar et al., 2011; Shankar, Venkatesh and Naik, 2010).

2.2 Purchase intentions

Purchase intention is defined as the probability that a buyer in a certain situation purchases a product from a seller (Crosno, Freling and Skinner, 2009). It is part of consumer behaviour towards a product, brand or firm (Howard and Sheth, 1970). Purchase intention theories aim to explain why people chose to purchase or not, and assume this choice is systematic rather than random (Chankon, Laroche and Lianxi, 1996). Purchase intention has been subject to many studies, it has been claimed that purchasing intention is rational as it is contained within bounded rationality of consumers (March and Simon, 1958). In the marketing literature, purchase intentions have commonly been used as a predictor of subsequent purchase (Grewal et al., 1998). Customers can distinguish between different brands and perceive qualities of these brands differently. It is argued these perceived quality differences increase purchase intention for brands with the higher quality perception (Amiri Aghdaie, Riasi and Seidi, 2012; Cooksey, Pappu and Quester, 2005).

Purchase intention towards a product is formed by the combination of a set of motives, reviewing alternative courses of action and decision facilitators. Motives reflect the underlying need of the buyer and focus on the product or overall product category. Reviewing alternative courses of action involves considering similar products from a different brand or substitute products which satisfies the same need. Decision facilitators are a consumers’ set of rules that matches the motives and means of satisfying the needs of the consumer. These can be influenced by internal factors, such as previous product or brand experience, or external factors such as promotions or recommendations (Howard and Sheth, 1970; Osselaer and Alba, 2002). Thus, intentions can be seen as a customer actively thinking about a certain behaviour (purchasing or not), purchasing intention can therefore be described as considering the purchase of a product or service.

2.3 Equity theory in mobile marketing

Equity theory assumes that people in a social exchange context perceive a relationship as reasonable or fair when the expected value of the outcomes is at least equal to the input (Arenson, Evans and Huppertz, 1978). Mobile in-store marketing can be seen as a special form of permission marketing as the consumer has to explicitly and consciously give permission to receive messages (Kotler and Keller, 2009). This is done by installing the app of the retailer or an affiliated firms’ app enabling to receive in-store advertisement communications. Retailers gather valuable data as the app collects personal data and tracks customer (in-store) movements. As customers consciously allow retailers to track them, they lose privacy and anonymity. Based on equity theory, customers allow this as they in return expect benefits from the retailer.

2.4 Hypothesis development

The hypotheses H1, H3 – H5, and their derivations are based on the research from Bues et al. (2017). H2 is an extension to the original study based on suggestions in the literature (Bues et al., 2017; Dickinger and Kleijnen, 2008; Hui, Inman, et al., 2013).

The literature review is extended by including studies regarding mobile advertising outside stores and SMS based communication to provide an inclusive overview of mobile in-store value drivers. This is due to a lack of research on mobile in-store advertising.

2.4.1 Price promotions

Price promotion can be defined as ‘an exchange of value, with the intent of driving a specified behaviour in the short term’ (Andrews et al., 2016). Price promotions aim at a financial value exchange which can be in the form of a price discount or a percentage discount, of which the latter is mostly used in the retail environment (Balasubramanian and Shankar, 2009).

Price promotions can take many different formats; however, this study addresses mobile marketing messages and will therefore address mobile price promotions. Mobile price promotions can be send via SMS, MMS, social media, in-app messages, emails, and pull or push messages. Mobile price promotions distinguish itself from mobile advertising by offering a value exchange based on the short term, whereas advertising aims to influence brand attitude, a longer-term strategy (Grewal et al., 2016).

Previous research using SMS coupons have demonstrated both similar and contrasting effects of price promotions on consumer purchasing intention. Most studies identified a positive effect of mobile in-store promotions on purchase intention (Choi, Hwang and McMillan, 2008; Dickinger, Murphy and Scharl, 2005; Drossos et al., 2007). However, some studies were not able to demonstrate a significant relationship between mobile price promotions and customer purchase intentions (Li, Liao and Xu, 2008).

Therefore, considering the characteristic of driving behaviour in the short term, it is predicted that receiving a mobile in-store price promotion message, incentives the customer to consider purchasing the promoted product (drive purchase intentions) more than when receiving a mobile in-store message without a price promotion.

H1: Mobile in-store price promotions messages result in a higher purchase intention than mobile in-store messages without a price promotion.

2.4.2 Message frequency

Message frequency refers to the number of mobile marketing messages sent to the customer when being in the retail store. Previous studies regarding mobile advertising show that consumers have concerns their mobile phone number can become part of advertisement distribution lists and as a result receive countless intrusive and unwanted advertisement on their mobile (Dickinger and Kleijnen, 2008). When receiving too many mobile messages, consumers feel their privacy is invaded and start perceiving the received mobile messages as intrusive.

Receiving multiple unwanted mobile advertising messages creates the feelings of a lack of control of their personal device. Consequently, customers feel invaded and cancel the message subscription or remove the app from their mobile phone. As a result, no mobile marketing messages are being received, losing the ability to influence purchase intentions by the firm (Dickinger and Kleijnen, 2008; Kleijnen, de Ruyter and Wetzels, 2007).

Several studies have confirmed the negative consequence of perceived lack of control as a result from irrelevant and inappropriate messages, wrong timing or information overload (Dickinger, Murphy and Scharl, 2005; Leppaniemi and Karjaluoto, 2005).

Following the equity theory, customers should perceive the social exchange context of receiving mobile messages in return for mobile messages as fair (Arenson, Evans and Huppertz, 1978). It is expected that sending multiple messages in exchange for a promotion for one product is deemed as intrusive rather than fair. Therefore, it is predicted that mobile in-store promotion with low message frequency results in a higher purchase intention than mobile in-store promotions with high frequency.

H2: Mobile in-store promotions communicated with low message frequency results in higher purchase intention than mobile in-store promotion with high message frequency.

2.4.3 Location

Location refers to the point where the customer is geographically present when receiving the mobile marketing message. Retailers can influence this by either sending messages to customers far away from the product, nearby the product or anywhere in between.

The influence of geographical location of the customer when receiving a message on purchase intention can be explained using the Construal Level Theory (CLT) in the consumer behaviour context. CLT in the behavioural context concerns evaluating the dimensions of psychological distance relevance to decision making (Liberman, Trope and Wakslak, 2007). Short temporal and geographical distance to the promoted product, leads to customers better grasping the mobile promotion mentally, resulting a higher involvement with the product which increases purchasing intent (Liberman and Trope, 2008; Liberman, Trope and Wakslak, 2007).

Previous research confirms that a shorter time span between receiving a mobile advertising message (SMS) with price promotions and the customer viewing the promoted product, the higher the likelihood of purchasing (Andrews et al., 2014; Li, Liao and Xu, 2008). Vice versa, the likelihood of mobile coupon redemption decreases as the geographical distance increases (Johnson, Lewis and Reiley, 2016). Additionally, multiple studies propose that the impact on purchase intention is greater with mobile advertising (SMS) if the perceived redemption effort is low (Dickinger and Kleijnen, 2008; Drossos et al., 2007). Thus, research indicates that time and distance have a negative influence on the perceived value of mobile advertising (Andrews et al., 2014; Hui, Inman, et al., 2013).

H3: Mobile in-store messages received when a customer is geographically closer to the promoted product result in a higher purchase intention than mobile in-store messages received at a greater geographical distance.

2.4.4 Personalisation

Personalisation is defined as: ‘a form of exclusivity, that is, specific customers receive exclusive offers while others will not get the same offer’ (Barone and Roy, 2010). As mentioned in section 2.1, the rise of LBA increased the possibilities for personalisation. However, his is not new to the literature as it has already been considered a vital characteristic of advertising before the rise of LBA (Ho and Tam, 2006; Li, Liao and Xu, 2008).

To personalise messages, companies gather detailed customer data of each engagement moment between the firm and customer. The personalisation process consists of learning about customer preferences, matching offers to customers and evaluate both the learning and matching process. In the retail environment, this data usually exists of previous purchases made at the retailer (Murthi and Sarkar, 2003).

Advertising that is not personal, might be not relevant and therefore provide less value. Whereas personalised advertisements in the in-store environment provide higher value, addressing individual needs (Klabjan and Pei, 2011; Li, Liao and Xu, 2008). It appears that when customers are exposed to personalised promotional offers, they are likely to spend less time on seeking information and making a choice, as they adopt the promotional offer sooner (Ho and Tam, 2006). However, there is a limit in the extend as to how far firms can go in terms of personalisation. The personalisation-privacy paradox states that personalisation can enhance as well as diminish customer purchase intentions (Aguirre et al., 2016; Gana and Koce, 2016; Ho and Tam, 2006; Xu et al., 2011).

Therefore, it is predicted that personalised mobile in-store messages have a greater effect on purchase intentions, than non-personalised mobile in-store messages.

H4: Personalised mobile in-store messages results in higher purchase intentions than non-personal mobile in-store messages.

2.4.5 Location and personalisation

the social exchange value (Aguirre et al., 2016; Arenson, Evans and Huppertz, 1978; Oh, Teo and Xu, 2009).

There are contrasting findings in the literature regarding this, as multiple studies also indicated receiving location specific mobile messages resulted in feeling of intrusiveness and manipulation (Harmon and Unni, 2007; Xu et al., 2011). Consequently, such feelings influence purchase intentions negatively (Goldfarb and Tucker, 2011).

Therefore, it is predicted that consumers perceive a personalised mobile in-store message received geographically close to the promoted product as intrusive, and a violation of their privacy. Resulting in a lower purchase intention compared to receiving a personalised mobile in-store message at a greater geographical distance. Thus, it is expected that the location of receiving a mobile marketing message negatively influences the relationship between personalisation and purchase intentions.

H5: Personalised mobile in-store messages received being geographically closer to the product result in a lower positive effect on purchase intentions than when a message is received being at a greater geographical distance.

Conceptual model

3. Methodology

The previous chapter introduced the relevant literature and theories. To answer the research questions, a suitable research design has to be developed. This chapter outlines the research paradigm, research design, data collection techniques and statistical analyses, followed by the ethical consideration for the research.

3.1 Research paradigm

A research paradigm is ‘the set of common beliefs and agreements shared between scientists about how problems should be understood and addressed’ (Kuhn, 1970). Holden and Lynch (2004) describe it as ‘a philosophical framework that guides how scientific research should be conducted’.

The two key philosophical approaches, subjective or objective approach to research, are outlined by the core assumptions of ontology (reality), epistemology (knowledge) and methodology (Holden and Lynch, 2004).

The objectivist approach is built on the assumption that social reality is objective (positivism) and cannot be affected by the act of investigation. Contrary, subjectivism (interpretivism) is built on the assumption that reality is in our minds and is subjective (Holden and Lynch, 2004).

Ontology is based on the nature of reality, whether reality has existed or is ‘the product of one’s mind’ (Holden and Lynch, 2004; Krauss and Putra, 2005). The positivism view sees reality as objective and things having existence. Ontology affects epistomology as it concerns studying the nature of knowledge and how investigation should be done.(Holden and Lynch, 2004). The objectivist view can discover the reality and transfer this to others. Methodology describes methods used to collect knowledge and is the researcher’s tool kit, representing all available means to sociel scientist to investigate phenomena (Holden and Lynch, 2004; Kraus, 2005).

This research aims to investigate theories based on emperical research and assumes social reality is objecitive and cannot be manipulated through the act of investigation. Hence, this research is based on a positivist view of the world. Quantitative data is collected by means of a survey and analysed for hypothesis testing in order to answer the research question.

3.2 Research design: Quasi experimental design

In keeping with studies of Campo, Gijsbrecht and Nisol (2000), Sloot, Verhoef and Franses (2005), Sloot and Verhoef (2008) and Bues et al. (2017), an experimental study design with hypothetical decision making in surveys will be conducted. An advantage of surveys is that a direct measure of different reactions to a mobile in-store message in different scenarios can be collected, in contrast to behavioural observations, enabling to distinguish clearly between the effects of price promotion, message frequency, location and personalisation on purchase intention (Campo, Gijsbrechts and Nisol, 2000). However, a drawback of the hypothetical decisions design might be that people can have difficulty imagining how they would act in a particular situation, or that people not always take the same action they claim they would take. The external validity of the reported effects on purchase intention can be lowered due to this limitation (Sloot, Verhoef and Franses, 2005). To limit the number of respondents having difficulty imagining certain scenarios, it is important to survey only relevant respondents. To ensure this, two screening questions will be inserted in the survey regarding smartphone possession and affinity with the chosen research object.

every participant is treated as similarly as possible. Further, a more homogeneous sample lowers the error variance. This is reached by means of the within-subject element of exposing each respondent to two scenarios, while limiting the demand effect as the scenarios are presented randomly out of a range of twelve different scenarios.

The four value drivers are each measured on two levels to measure the influence each value driver has on purchase intention. Next, the interaction effect between location and personalisation will be added to the model to see if a personalised message moderates the relationship between location and purchase intention. Additionally, the effect sizes of all value drivers are compared to asses which variable has the largest impact and can be considered the most important value driver. For measuring the four value drivers on two levels there will be a total of 12 different experimental groups. A complete overview with explanation of all scenarios can be found in appendix A.

Participants will be allocated randomly to the different scenarios. The research is based on hypothetical decision making, meaning participants will be put in a hypothetical situation and answer hypothetical questions, this approach is common in the retail context when using experimental studies (Bues et al., 2017; Campo, Gijsbrechts and Nisol, 2000; Sloot and Verhoef, 2008; Sloot, Verhoef and Franses, 2005).

The chosen research object is wine, as wine is purchased less often than other products in the FMCG industry (Kollat and Willet, 1967). Therefore, it is less likely to be an impulse purchase at the category level as customers tend to decide on purchasing wine before entering the retail store (Bell, Corsten and Knox, 2012). Despite deciding to purchase wine, customers usually do not decide on specific type of wine (brand) upfront, but make their decision based on product available and on in-store stimuli. Supermarket and liquor stores usually have a large wine assortment causing customers to be overwhelmed by the choice. Stimulus in-store can help during decision making, and is likely to be an effective manner of influencing purchase intention (Aqueveque, 2006).

3.2.1 Survey structure

Second, each participant is randomly assigned to one of the twelve hypothetical scenarios, each differing in the composition of the value drivers (independent variables) price promotion, message frequency, location and personalisation. The common factors in all scenarios are the following: all participants are asked to imagine entering a supermarket and act as if they have purposely installed an app that offers in-store advertisement messages and have agreed to receive such messages. After reading the scenario, participants were asked to answer a control question in which they had to indicate the location where a mobile message was received. This question is inserted to ensure participants have read and understand the description. To reduce the guess factor, or the chance of accidentally selecting the right answer, four options are given of which two are relevant (at the entrance or near the product) and two are irrelevant (at the parking area or at the cashier). Not answering or faulty answering this question resulted in being omitted from the research. Subsequently, several questions regarding purchase intention were measured for each scenario.

In the third and final part, several control variables were surveyed to control for (potential) influences on customer purchase intention, dependant on mobile in-store advertisements as found in the literature review. These controls measure perceived intrusiveness, the general attitude regarding mobile services, privacy concerns and customer involvement with the product category wine. Any variance caused by these variables will be controlled for in hypothesis testing. The survey with scenario 1 as example can be found in appendix B.

3.2.2 Measurement and item specification

This research aims to identify the effects of several value drivers on the purchase intention of consumers in the retail environment. To do so, five main variables are measured to answer the research question. Table 3.1 provides an overview of all variables.

Table 3.1: Overview of variables

The measurement items for the constructs used in this research are derived from previous studies to ensure the validity of the instruments. Small changes were made to adapt the measurement items to the context of this research.

The independent variables characteristics are inside the twelve different scenarios to where participants were randomly distributed using the randomiser tool in Qualtrics. The scenario

Type of variable Variable Independent Price promotion Independent Message frequency Independent Location

situations are based upon an example of Bues et al. (2017) and adapted to the research context and measured constructs. The independent variables are measured by creating dummy variables for each independent variable and assessing the responses to two different situations for each independent variable.

Price promotion is measured using two different scenario types. One type in which a 30 per cent discount is offered via a mobile message, and one type in which no discount is offered message. The measurement items are derived from studies of Drossos et al. (2007), Hui et al. (2013), and Scharl, Dickinger, and Murphy (2005) (Dickinger, Murphy and Scharl, 2005; Drossos et al., 2007; Hui, Inman, et al., 2013).

Message frequency is measured using two different scenario types. One type in which two mobile messages are received, and one type in which one mobile message is received. These measurements are derived from Dickinger and Kleijnen (2008). (Dickinger and Kleijnen, 2008)

Further, location is measured using three different scenario types. In one scenario type a mobile message is received geographically far from the product, which is conceptualised as at the store entrance. In another scenario type the mobile message is received geographically near the product, which is conceptualised as near the shelf of the promoted product. Lastly, in the scenario type of sending two mobile messages, a message is received at both locations. These measurement items are derived from multiple previous studies (Andrews et al., 2014; Li, Liao and Xu, 2008) .

Personalisation is measured using two different scenario types. One scenario type sends out personalised message(s), and one type send out general message(s). These measurement items are derived from Goldfarb and Tucker (2011) (Goldfarb and Tucker, 2011).

Table 3.2: Overview constructs and measurement items of independent variables

The dependent variable customers’ purchase intentions (PU) is measured by using multiple items. These items are derived from a study by Baker and Churchill (1977), where PU 1, 3 and 4 are based upon. PU 2 is derived from a study by Burton, Garretson and Velliquette (1999). All measurements items were adjusted to the research context, which resulted in transforming ‘the product’ into ‘the promoted wine’ (Baker and Churchill, 1977; Burton, Garretson and Velliquette, 1999).

Furthermore, several control variables are measured to control for any influences on purchase intention of customers which are not part of the study. The control variables: intrusiveness, attitude towards mobile services, privacy concerns and involvement with wine product category were identified in the literature review as potentially relevant factors when measuring purchase intention. The measurement items are derived from literature on mobile marketing, and privacy.

The scale of Ducoffe (1996) is used to measure the perceived attitude towards mobile services (AttMServ). This scale is adapted to the research context by replacing ‘advertising on the World Wide Web…’ by ‘In general, location-based advertising for retail is …to me’ (Ducoffe, 1996).

The items measuring privacy concerns (PC) 1 and 2 are based on the scales of Dinev and Hart (2006), and PC 3 – 5 are derived from Zhao, Lu and Gupta (2012). The measurement items were transformed to the research context by replacing ‘the internet’ to ‘my phone’ (Dinev and Hart, 2006; Zhao, Lu and Gupta, 2012).

Further, the items measuring the perceived intrusiveness (PINT) are based on the scale of Edwards, Li and Lee (2002). These items were adjusted to the research context by replacing the words ‘pop-up ad’ to ‘the message’ (Edwards, Li and Lee, 2002).

Construct Measurement items Indicator Source

Price Promotion Price promotion (30% discount) 1 Drossos et al. (2007), Hui et al. (2013), and Scharl, Dickinger, and Murphy (2005) No price promotion 0

Message frequency Two mobile messages 1 Dickinger and

Kleijnen (2008)

One mobile message 0

Location Mobile message at store entrance 1 Luo, Andrews, Fang, & Phang, (2013), and Xu et al. (2008) No mobile message at store entrance 0

Mobile message near product 1 No mobile message near product 0

Lastly, the involvement with wine (Winv) of the customer is measured using the scales of Chandrashekaran (2004), Laurent and Kapferer (1985) and Mittal (1985). This scale is used to measure perceived importance of products and is adapted to the research context by filling in ‘wine’ in the blank space (Chandrashekaran, 2004; Laurent and Kapferer, 1985; Mittal, 1995). The survey aims to measure participants’ intentions, perceptions, concerns and attitudes. Scale questions have proven to be a reliable and accurate measure for this (Saunders, Lewis and Thornhill, 2009). Therefore, all measurement items will consist of a seven-point Likert scale, of which the meaning may differ depending on the question (e.g. agree or likely). A seven-point scale is adopted to enable participants scoring neutral (Malhotra et al., 2007). An overview of all constructs, the measurement items, measurement indicator and source where the items are derived from can be found in table 3.3.

Table 3.3: Overview constructs and measurement items of dependent & control variables

Construct Measurement items Indicator Source

Purchase intention

(PU) ₁ How likely is it that you would look out for the promoted wine to purchase it?

1= extremely unlikely

7= extremely likely Baker and Churchill (1977) and Burton, Garretson and Vellinquette (1999) 2

Given the described situation, how probable is it that you would consider the purchase of the promoted wine?

3

Would the purchase of the promoted wine be more likely or less likely given the described situation?

4

How likely would you be to purchase the promoted wine after reading the described situation?

Perceived intrusiveness (PINT)

1 Distracting When I got the message, I

thought it was… Edwards, Li and Lee (2002) 2 Disturbing

3 Forced

4 Intrusive 1= Strongly disagree 7= Strongly agree 5 Invasive

Attitude towards mobile services (AttMserv)

1 Useful In general, location-based advertising for retail is …to me.

Ducoffe (1996) 2 Valuable

3 Important 1= Strongly disagree _{7= Strongly agree} Privacy concerns

(PC) 1 That the information I submit through my phone could be misused.

To what extend are you concerned about the following statement? I am concerned…

Dinev and Hart (2006) and Zhao, Lu and Gupta (2012) 2

That the personal information is made available to third parties due to my phone usage.

3

About submitting

information on my phone, since data might be used unrestrictedly.

4

About submitting

information on my phone, because it could be used in a way I did not foresee.

1= Not at all concerned 7= Very concerned

5 That information I submit though my phone might violate my privacy. Involvement with

product category wine (Winv)

1 I choose wine carefully. 1= Strongly disagree

7= Strongly agree Chandrashekaran (2004), Laurent and Kapferer (1985) and Mittal (1985) 2 Choosing wine is an _{important decision for me.}

3 I am particularly interested _{in wine}

3.2.3 Survey development

After establishing the measurements for the constructs and developing the survey, a pre-test was done with N=5 respondents. The respondents were asked to complete the survey the same way as if it was for the actual project, but additionally asked to think out loud and take notes of any ambiguities or questions that might arise while answering (Collins, 2003).

Respondents noted some spelling mistakes in the questions and made comments on the order of some questions. To solve this, the spelling was checked thoroughly, and the purchase intentions questions were re-ordered. Furthermore, it was noted that demographic questions could better be asked at the end of the survey, and several respondents noted the differences between scenarios was unclear. This has been addressed by moving the demographic questions to the end of the survey. To clarify the scenarios, a small introduction was added indicating the scenarios are all similar at the start but should be read carefully after the first paragraph. Additionally, some text was made bold to emphasize differences in the scenarios.

After the revision, the survey was pre-tested with N=10 respondents to check for validity and reliability of the construct measurements (Collins, 2003). All constructs proved to be valid and reliable in pre-testing.

Participants of both pre-tests were excluded from the final survey to prevent biases. This is done as the participants have developed an idea of the hypotheses being researched. The final version of the survey can be found in appendix B.

3.3 Data collection

The aim of this study is to gather a sample size large enough to represent the Dutch population well. To do so, a valid number of respondents from the Dutch population needs to be collected. Based on a population size of approximately 13,677,409 in the Netherlands (CBS, 2017), the sample size can be calculated using the Surveymonkey sample size calculator. With a confidence level of 95 per cent and a margin of error of 5 per cent, the sample size has to be at least N=385 to be valid (Surveymonkey, 2017).

3.4 Data analysis

Data collected by the questionnaire was analysed using IBM SPSS Statistics 25 (IBM SPSS, 2017). Before analysing the data, the data is prepared for analyses. This includes developing a database structure in Excel to store the data and checking the data for accuracy before analysis.

Both descriptive and inferential statistics are used. Descriptive statistics give a summary about the sample and measures, simplifying data in an organised manner. Univariate analysis examines across cases at one variable at a time, summarising the distribution, central tendency and dispersion. Inferential statistics are used to answer the research questions. As the purpose of this study is to recognise significant differences in purchasing intention, ANCOVA tests are performed for hypotheses testing.

3.5 Ethical considerations

4. Findings

Based on the research design as described in chapter three, the gathered data is analysed. At first, all data is screened, and incomplete responses were removed. Second, descriptive statistics about the sample are provided, followed by the reliability and validity tests of the constructs. Lastly, the hypotheses are tested with ANCOVA tests.

4.1 Preliminary analyses

Before entering the data into SPSS, the data is prepared for analysis in Excel. All responses are checked on meeting the research requirements regarding the screening and control questions, removing all values not meeting the requirements. Additionally, incomplete or abnormal responses are removed. Next, the sample characteristics are outlined and descriptive statistics are stated, followed by the reliability and convergent and discriminant validity analyses of the constructs by using Cronbach’s Alpha, factor loadings, Average Variance Extracted (AVE) and Composite Reliability (CR).

4.1.1 Data screening and cleaning

Data collection took place in week 45 and week 46 of 2017. The online survey was closed after collecting data for 9 days, resulting in a total number of respondents for the raw data set of N=838. To clean the data set, all respondents not owning a smartphone or never purchasing wine were removed as well as respondents who answered the control question wrong. Further, responses with a completion rate lower than 90 per cent and abnormal responses were removed. Lastly, although it is suggested by statisticians that Likert scale data does not need to be checked for outliers as the nature of the variable has a limited range (between 1 and 7), the dataset is checked for outliers by means of a boxplot. Outliers have been found in the variables of purchase intention, attitude towards mobile services and privacy concerns. Before removing the outlier values, it is checked whether the outliers affect the data by comparing the 5 per cent trimmed mean values with the mean values. This showed the mean values were only marginally affected by the outliers with -0.10, -0.06 and -0.15 respectively, indicating there is no need to remove the outliers (Field, Field and Miles, 2013).

incomplete, faulty or abnormal responses. An overview the number of respondents per scenario can be found in table 4.1.

Table 4.1: Overview of respondents per scenario

4.1.2 Sample characteristics

The aim of this study is to gather a sample size large enough to represent the Dutch population well, which required a minimum valid sample size of 385 respondents. As the collected number of valid responses is N=419, the sample represent the population well. Based on the sociodemographic criteria, however, the sample deviates from the Dutch population regarding age group and income. The age group 18 – 34 is overrepresented compared to the age group of 35 years and older. Further, low incomes are overrepresented and moderate incomes underrepresented, but high-income groups are accurately represented. The findings of the study will be compared to the findings of Bues et al. (2017) to see if the findings are similar and thus can be generalised for a wider audience. When doing so, the results should be interpreted with caution due to the sample not being fully representative for the Netherlands.

An overview of the sample characteristics compared to the Dutch population characteristics can be found in table 4.2. The Dutch population characteristics are based on the statline database of the CBS in 2015 (CBS, 2015). A more detailed overview of the demographic variables of the sample can be found in appendix C.

Scenario Respondents Scenario Respondents

Table 4.2: Sample characteristics

4.1.3 Descriptive statistics

The descriptive statistics are presented for all constructs. It provides an overview of the number of observations, mean, standard deviation, skewness and kurtosis of each item. The mean ranges from 3.90 to 5.56 with standard deviations ranging from 1.29 to 1.79.

Skewness and kurtosis are used for testing if the sample data is normally distributed. Meaning, whether it fits a “symmetrical, bell-shaped curve, which has the greatest frequency of scores in the middle with smaller frequencies towards the extremes” (Gravetter and Wallnau, 2016). A perfect normal distribution is recognised by a value of zero for both Skewness and Kurtosis. This is, however, a rare happening in social science. Skewness indicates the symmetry of a distribution and Kurtosis shows information regarding the peakedness of a distribution (Gravetter and Wallnau, 2016). Values for Skewness and Kurtosis ranging between -1.96 and 1.96 indicate normality, and can be interpreted as normally distributed (Ghasemi and Zahediasl, 2012).

Table 4.3 provides an overview of the Skewness and Kurtosis values for each measurement item. It shows all items fulfil the normality assumption, except for purchase intention question 2: ‘Given the described situation, how probable is it that you would consider the purchase of the promoted wine?’. This item has a Kurtosis value of 2.20, exceeding the threshold value of 1.96. Nevertheless, as the exceeded value is not extremely

Dutch population Sample Gender Male 49.6% 44.9% Female 50.4% 55.1% Age group 18-24 8.7% 49.4% 25-34 12.4% 32.2% 35-44 12.2% 3.4% 45+ 66.7% 8.1% Missing values - 6.9% Nationality Dutch 78.3% 72.6% Non-Dutch 21.7 % 26.4% Missing - 1%

Household income group Low (less than €20,000 per annum) 21% 46.8% Moderate (€20,000 to €50,000 per

annum) 67% 30.3%

High (More than €50,000 per annum 12% 11.2% Prefer not to answer - 11.7%

Household size 1 person 37% 33.7%

2 persons 32.7% 27.9%

3 persons 12.1% 13.1%

4 persons 12.8% 16%

high, and no other items indicate values deviating from the -1.96 and 1.96 range, normality problems are not expected to be an issue for the hypotheses testing.

Table 4.3: Descriptive statistics constructs Construct

Purchase intention N Mean SD Skewness Kurtosis

How likely is it that you would look out for the

promoted wine to purchase it? 419 5.56 1.36 -1.30 1.66

Given the described situation. how probable is it that you would consider the purchase of the

promoted wine? 418 5.43 1.30 -1.40 2.20

Would the purchase of the promoted wine be more likely or less likely given the described situation?

418 5.38 1.28 -1.12 1.27

How likely would you be to purchase the promoted wine after reading the described

situation? 416 5.24 1.31 -1.15 1.40

Perceived intrusiveness N Mean SD Skewness Kurtosis

When I got the message. I thought it was

distracting. 419 4.04 1.49 -0.18 -0.79

When I got the message. I thought it was

disturbing. 419 3.90 1.55 -0.03 -0.84

When I got the message. I thought it was forced. 417 4.29 1.63 -0.32 -0.84

When I got the message. I thought it was

intrusive. 417 4.30 1.55 -0.34 -0.48

When I got the message. I thought it was invasive. 418 4.23 1.54 -0.28 -0.56

Attitude towards mobile services N Mean SD Skewness Kurtosis

In general. location-based advertising for retail is

useful to me. 417 5.12 1.34 -1.16 1.16

In general. location-based advertising for retail is

valuable to me. 419 4.85 1.46 -0.85 0.01

In general. location-based advertising for retail is

important to me. 415 3.99 1.55 -0.16 -0.68

Privacy concerns N Mean SD Skewness Kurtosis

I am concerned that the information I submit

through my phone could be misused. 419 4.66 1.62 -0.59 -0.79 I am concerned that the personal information is

made available to third parties due to my phone

usage. 419 4.95 1.62 -0.88 -0.18

I am concerned about submitting information on

my phone. since data might be used unrestrictedly. 417 4.82 1.52 -0.72 -0.32 I am concerned about submitting information on

my phone. because it could be used in a way I did

not foresee 419 4.87 1.57 -0.86 0.04

I am concerned that information I submit though

my phone might violate my privacy. 415 4.84 1.64 -0.79 -0.32

Involvement with wine N Mean SD Skewness Kurtosis

I choose wine carefully. 419 4.94 1.39 -0.86 0.34

Choosing wine is an important decision for me. 419 4.21 1.65 -0.21 -0.89

I am particularly interested in wine. 413 4.33 1.79 -0.25 -1.10

4.1.4 Reliability analyses

The internal consistency among the constructs in each scenario is tested using Cronbach’s alpha. There is some debate among statisticians regarding the value of Cronbach’s alpha indicating reliability, some suggest a value of 0.70 or higher whereas others recommend 0.80. In general, a value for Cronbach’s alpha above 0.7 is accepted as reliable and this will therefore also be adopted in this research (Field, Field and Miles, 2013; Hinton, McMurray and Brownlow, 2014).

Table 4.4 displays the results for the reliability analysis. There are no coefficients for Cronbach’s alpha below 0.7, indicating that the internal consistency of measures is correct. There have been found two improved values for Cronbach’s alpha if an item were deleted in the intrusiveness and attitude towards mobile services construct. The difference, however, was minimal as the total improvement would be 0.01 for each construct.

Further, all corrected item-total correlations have been checked as this is suggested when there is a small number of items in the construct (Ferketich, 1991). The corrected item-total correlation coefficient indicates the way items relate to other items. Values should be 0.30 or higher as items otherwise not contribute in measuring the core factor. All values range between 0.59 and 0.91, which is well above the threshold value of 0.30 (Ferketich, 1991). The output tables displaying the corrected item-total correlations are displayed in appendix D1. According to Devellis (2016), however, there may be variances in the reliability of Cronbach’s alpha depending on the sample due to the nature of its calculation. Cronbach’s alpha is calculated by correlating the score for each scale item with the total score for each observation, and subsequently comparing that to the variance of all individual item scores (DeVellis, 2016). Therefore, an additional reliability analysis is performed per construct and per scenario. These results can be found in appendix D2.

Table 4.4: Cronbach's alpha

Additionally, a factor analysis is conducted to check for the reliability of the constructs. A distinction can be made between Exploratory Factor Analysis (EFA) and Confirmatory Factor

Construct Number of items Cronbach’s

Alpha

Purchase intention 4 0.93

Intrusiveness 5 0.90

Attitude towards mobile services 3 0.88

Privacy concerns 5 0.95

Analysis (CFA). CFA is used in trying to confirm hypotheses using path analysis diagrams to represent variables and factors. EFA is used to attempt uncovering complex patterns by exploring data and testing predictions, and can be used to discover the number of factors as well as identifying which variables fit together (Yong and Pearce, 2013). Hence, an EFA is conducted. Before conducting factor analysis, assumptions have to be met which check for the appropriateness of performing a factor analysis.

A KMO measure of sampling adequacy was conducted before the factor analysis. This index is used to determine the appropriateness of factor analysis, values vary from 0 to 1 and values above 0.5 indicate factor analysis is appropriate (Hinton, McMurray and Brownlow, 2014; Pankhania and Jani, 2012). Further, the Barthlett’s test of Sphericity is conducted to checks for a relationship between the variables. With no relationship found, there is no point in proceeding with the factor analysis. A p value <.05 indicates factor analysis is appropriate (Hinton, McMurray and Brownlow, 2014; Pankhania and Jani, 2012). Value for KMO showed 0.88 and Barthlett’s test was highly significant (p<.000), meaning a factor analysis is appropriate.

Now that it has been determined appropriate to conduct a factor analysis, there are several extraction methods available. The most commonly used are Maximum Likelihood, the Principal Axis Factor method and Principal Component analysis. A Principal Axis Factor analysis is chosen as Maximum Likelihood is more useful for confirmatory factor analysis, and Principal Component analysis is a data reduction technique for which questions have been raised if it truly is a factor analysis technique (Costello and Osborne, 2005). Therefore, Principal Axis Factor analysis is regarded a ‘true’ factor analysis technique. Next, a rotation method has to be determined for better interpretation of the factors as un-rotated factors are unambiguous (Rummel, 1988). Either orthogonal rotation or oblique rotation can be chosen, where oblique rotation is more complex. Orthogonal rotation is suggested as interpretation is easier, hence this rotation method is applied (Yong and Pearce, 2013). Two common techniques are Varimax or Quartimax. Varimax rotation is deemed more appropriate as this reduces the number of variables which have high loadings on factors and works to make small loadings smaller (Yong and Pearce, 2013).

0.92, meaning there are no reliability problems. The results of the factor analysis can be found in table 4.5.

Additionally, the extracted communalities are calculated per item and per scenario to check the extent to which each item correlates with all other items. This value should be 0.5 or higher to be reliable for further analysis (Yong and Pearce, 2013). Most factor loadings have values ranging from 0.54 to 0.93, indicating no reliability problems. The extracted communalities table can be found in appendix E.