Predictors of customer attitude toward interactive smart shelf technology and intention to use it in supermarkets

(1)

Predictors of customer attitude toward interactive smart shelf

technology and intention to use it in supermarkets

Ani Aradjanian 11397292

June, 22th Final version

MSc. in Business Administration – Marketing Track

Amsterdam Business School, Faculty of Economics and Business University of Amsterdam

(2)

Statement of originality

This document is written by Student Ani Aradjanian 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 creating it”.

The Faculty of Economics and Business is responsible solely for the

supervision of completion of the work, not for the contents.

(3)

List of Figures

Figure 1. Theory of reasoned action (Appendix) ……… 54

Figure 2. Technology acceptance model (Appendix)……….. 54

Figure 3. Technology acceptance model 2 (Appendix)………... 54

Figure 4. Unified theory of acceptance and use of technology (Appendix)……… 55

Figure 5. Conceptual model: Interactive smart shelf technology……… 22

Figure 6 to 15. Scatterplots with ZPRED versus ZRESID (Appendix)……….. 56

Figure 16 to 25. Histograms and P-P plots with ZPRED versus ZRESID (Appendix)…………. 57

List of Tables

Table 1. List of variables (Appendix)………. 60

Table 2. Exploratory factor analysis with technology beliefs items (Appendix)……… 62

Table 3. Exploratory factor analysis with attitude and intention items (Appendix)……… 64

Table 4. Item-total correlation and Cronbach’s α reliability of all factors (Appendix)………65

Table 5. Descriptive statistics and Pearsons correlations between all variables……….…. 33

Table 6. Regression analysis with intention as dependent variables……….……….. 34

Table 7. Regression analyses with attitude and intention as dependent variables………... 35

Table 8. Mediation analysis with independent variable perceived usefulness, mediator……….... 36

attitude and dependent variable intention Table 9. Direct effect, total effect of perceived usefulness on intention and indirect effect……… 36

via attitude Table 10. Regression analyses with perceived usefulness, attitude and intention as dependent…. 38 variables Table 11. Mediation analysis with independent variable perceived ease of use, mediator………. 38

perceived usefulness and attitude and dependent variable intention Table 12. Direct effect, total effect of perceived ease of use on intention and indirect………….. 39

effect via perceived usefulness and attitude Table 13. Regression analyses with attitude and intention as dependent variables………. 40

Table 14. Mediation analysis with independent variable enjoyment, mediator attitude, and…..… 41

dependent variable intention Table 15. Direct effect, total effect of enjoyment on intention and indirect effect via attitude…... 41

Table 16. Regression analyses with attitude and intention as dependent variables………. 42

Table 17. Mediation analysis with independent variable social influence, mediator attitude……. 43

and dependent variable intention Table 18. Direct effect, total effect of social influence on intention and indirect effect via ………43 attitude

(6)

Abstract

The interactive smart shelf technology has a promising future for supermarket retailers. Benefits include inventory automation, and improved customer relations. Not surprisingly, therefore, retailers strive to integrate the technology into their stores. However, retailers should not be blinded by the technology’s benefits from their own perception rather than considering the customers' motives for accepting or rejecting the technology. Yet, academic research on technology acceptance/resistance in the retail environment is limited. Previous studies in other fields have proven the technology acceptance model (TAM) (Davis, 1989) to be suitable in explaining which factors influence acceptance of and resistance to technology, e.g., acceptance of computer systems within organizations (Venkatesh & Davis, 2000); mobile shopping (Pantano & Priporas, 2016). This study applied the TAM model in the context of the supermarket industry, and investigated how customers’ cognitive (perceived usefulness; perceived ease of use), affective (enjoyment), and social (social influence) beliefs influence their attitude toward interactive smart shelf technology and intention to use it. Data from Dutch customers (N=207) were collected using a survey. Findings showed that customers’ perceptions of usefulness, ease of use, enjoyment, and social influence affected their attitude toward the technology, and their intention to use the technology positively. Furthermore, attitude mediated the relationship between usefulness, enjoyment, social influence and intention, while usefulness and attitude mediated the relationship between ease of use and intention.

(7)

1. Introduction

Imagine a trip to the supermarket, walking through a fully supplied store with shelves that tell you where the nearest item from your grocery list is, or provide you personal recommendations and special daily promotions. A supermarket where you can gain information about products without taking out your smartphone to surf to Google, since the only thing you have to do is to reach to the product you want information about, or tap on the interactive tables which are located in all aisles with screens above them. From the screens you can read information about, e.g., the food’s price, nutrition facts, potential allergens, and the carbon footprint. Does what you have just imaged sound far-fetched? It is not. Recently, the first supermarket where such shelves are established, has been designed, and many others are on their way (Accenture, 2015). Those shelves are called interactive smart shelves, part of the smart retail technology.

Smart retail technology is a retail system that can be connected to the Internet and be used interactively (Foroudi, Gupta, Sivarajah & Broderick, 2017). Retailers can provide services to customers via the network of intelligent objects and devices that can be used in real-time data collection, communication, interaction and customer feedback (Roy, Balaji, Quazi & Quaddus, 2018). Interactive smart shelves function on the basis of data from multiple sensors combined with interactive screens, motion gesture controlled devices and augmented visualization (Accenture, 2015). The study of Accenture (2015) indicated that interactive smart shelves have the potential to fulfil the gap between online and in-store experience, and improve the customer’s entire shopping experience by integrating the info-richness of the Internet into the in-store experience. However, this is from the perception of retailers, the question remains how customers will perceive the interactive smart shelf technology.

Yet, the fact that the retail environment has been subject to change is not a matter of dispute (Grewal, Roggeveen & Nordfält, 2017). With the rise of e-commerce around the world, new markets have been created and customers have been given more personalized service (Herring, Wachinger & Wigley, 2014). Herring et al. (2014) argued that due to this development, many offline retailers have added online channel to their store format, and that, therefore, the need for physical stores has lessened. Their study illustrated this by the fact that, e.g., in the United States, Gap closed more than 250 stores in 2013, and Walmart’s new stores are one third smaller than they were in 2009. In the United Kingdom, the number of vacant stores increased by 355% between 2008 and 2013 (Herring et al., 2014). However, despite the

(8)

increase in online sales and decline of physical stores, offline sales still generate 81% of retail companies’ revenue, proving that physical stores are still the scene where the vast majority of shopping takes place (Bornemann, 2015). Nevertheless, Bornemann (2015) emphasized that customers have raised their in-store shopping expectations toward offline retailers, and if these retailers fail to meet their expectations, stores will lose their customers and will become increasingly irrelevant.

To meet customer expectations, offline retailers have recognized the importance of improving in-store customer experience (Verhoef, Lemon, Parasuraman, Roggeveen, Tsiros & Schlesinger, 2009). In the literature, customer experience is described as a multi-dimensional construct, involving the customer’s cognitive (think), sensorial (sense), affective (feel), behavioural (act) and social (relate) responses to direct or indirect contact with a firm through a journey of touchpoints along search, purchase, and post purchase situations (Homburg, Jozić & Kuehnl, 2017; Schmitt, 1999). In-store customer experience includes the value that a customer perceives from the store visit (Terblanche, 2018). While Terblanche (2018) pointed out that factors that influence in-store customer experience differ between retail types, he explored factors that are important for supermarkets. Terblanche (2018) concluded that, among other things, merchandise quality and variety, internal store layout, such as easy in-store movements, and knowledgeable staff lead to positive in-store customer experiences. Thus, Terblanche (2018) recommended retailers to pay attention to these factors. In the meanwhile, technology change continues, and retailers believe that by integrating new technology into their physical stores, they can meet customers’ expectations better, and offer them unique in-store experience and stay competitive (Grewal et al., 2017).

While new technology has emerged and retailers have embraced it to improve the in-store customer experience, academic literature investigating in-in-stores technology acceptance is limited (Foroudi et al., 2018; Roy et al., 2018). Moreover, this is especially the case for the context of the supermarket industry. To give an illustration, the few existing studies on in-store technology acceptance are in the context of the electronics industry (e.g. Müller-Seitz, Dautzenberg, Creusen & Stromereder, 2009), and the fashion industry (e.g. Kim, Lee, Mun & Johnson, 2017). Müller-Seitz et al. (2009) found that customers’ cognitive beliefs, i.e., the degree to which they perceive in-store technology to be useful in completing their shopping tasks, positively affected the acceptance of technology in electronics stores. Kim et al., (2017) contributed to this with their finding that in addition to cognitive beliefs, affective beliefs, i.e., enjoyment, also play a role in technology acceptance. They found that high perceived

(9)

enjoyment of in-store technology positively affected the acceptance of technology in fashion stores.

Supermarket retailers are also increasingly integrating technologies into their store. For example, aside from the supermarket with interactive smart shelf technology, Amazon has introduced the scan and go technology. The scan and go technology offers customers the opportunity to take the desired items and leave the store without scanning or paying the item via a cashier or self-checkouts. The technology functions via an app which is connected to the customer’s bank account. By removing an item, sensors will identify the bar code and add the purchase to the virtual cart within the app. In this way, the payment has been completed (Amazon Go, 2018). Garaus et al. (2016) highlighted the lack of research on technology acceptance in the supermarket industry context, and conducted a study on customers' acceptance and perception of electronic shelf labels in supermarkets. They found that customers found electronics shelf labels easy to use, but not very useful, because they did not know the benefits of electronic shelf labels.

As technology keeps advancing at a rapid pace, electronic shelf labels have evolved into interactive smart shelves (Accenture, 2015). The displays have been improved, allowing retailers to show more product information and daily promotions (Harrison, Faigen & Brewer, 2018). Customers may better perceive the benefits of interactive smart shelves, and thus, their perception could have changed in the meantime. Furthermore, the drivers of customer technology acceptance in supermarkets have been tested incompletely. Garaus et al. (2016) only examined the influence of customers’ cognitive beliefs, i.e., usefulness and ease of use, on the acceptance of electronic shelf labels.

This thesis aims to contribute to the existing literature on the relationship between customers’ beliefs and technology acceptance in the context of the supermarket industry by investigating not only how their cognitive, but also affective and social beliefs influence their attitude toward interactive smart shelves and the intention to use it. The following question will be answered:

“How do customers’ cognitive, affective and social beliefs influence their attitude toward interactive smart shelf technology, and intention to use it in supermarkets?”.

By doing so, this thesis contributes to existing literature in several ways. First, this study contributes to the literature by empirically examining customer technology acceptance in the supermarket industry context. As Kim et al., 2017 found that the effects of belief

(10)

factors on attitude toward technology and the intention to use it vary by the type of in-store technology, this thesis proposes that it may also vary between different retail contexts. Second, previous studies in the supermarket industry have solely based technology acceptance on customers’ cognitive beliefs (e.g. Garaus et al., 2016). The purpose of this study is to provide a more complete picture of the various belief factors that influence technology acceptance in supermarkets by investigating customers’ affective and social beliefs on their attitude toward technology and intention to use it, in addition to their cognitive beliefs. It may well be that the more enjoyable and/or social influence aspects of technology explain attitude formation and use intention.

From managers’ point of view, the large investments in smart retail technology illustrate the great potential of the industry (Porter & Heppelmann, 2014). According to a recent report, the global smart retail market size was €11.93 billion in 2015 and is predicted to reach €29.77 billion by 2020 (Research and Markets, 2015). One of the main drivers is the retailer’s desire to create a seamless experience between online, and in-store shopping. However, despite the technology’s potential, its success depends on the customers’ acceptance of the technology (Roy et al., 2018). Therefore, before retailers invest time and money in integrating technologies into their stores, they should understand the factors that influence customer acceptance of in-store technology. This thesis provides retailer insight in factors that influence customers’ attitude toward interactive smart shelf technology and intention of use this in supermarkets.

This thesis is composed into six chapters. In the second chapter that follows the theoretical framework is presented. The methodology and data are presented in the third chapter. Thereafter the results are presented in the fourth chapter. In the fifth chapter the discussion is presented in which the study’s theoretical and managerial implications are discussed, the limitation of this thesis are argued and recommendations for future research are given. The sixth and last chapter contains the conclusion of this thesis.

(11)

2. Theoretical framework

This chapter starts by briefly explaining customer experience in the retail environment and the role of in-store technologies in it. Second, various behavioral intention models are discussed. Third, the hypotheses are outlined. In the fourth and final section, the conceptual model is introduced.

2.1 Customer experience and technology in the retail environment

In recent years, much attention has been paid to customer experience management and its importance for business performance (Verhoef & Lemon, 2016). Retailers have also become aware of the need to create value for their customers in the form of experiences along the search, purchase, and post purchase processes (Verhoef et al., 2009). Verhoef et al. (2009) argued that the importance of customer experience management has to do with the high rate of competition in the retail environment. Companies strive to increase the value of their products and services by creating customer experience (van den Driest & Weed, 2014). van den Driest & Weed (2014) explained the two way in which companies do this. On the one hand, companies try to deepen the customer relationship by using what they know about a particular customer to personalize their offers. On the other hand, they try to broaden the customer relationship by adding touchpoints, e.g., during the search, purchase, and post purchase processes. During these processes retailers are in contact with customers. With the rise of multi-channel retailing, including offline and online channels, direct and indirect contact points between retailer and customers have increased (Verhoef, Kannan & Inman, 2015). Verhoef et al. (2015) argued that consumers use online and offline channels interchangeable, and that therefore, it is difficult for companies to control this usage. Yet, they also highlighted the fact that due to technological developments, the distinction between offline and online retailers is disappearing. In turn, retailers can offer seamless, enjoyable, and informative experiences to their customers (Verhoef et al., 2015).

New technology provides online retailers tools to offer visual and sensory experiences to their customers (INretail, 2017). For example, on the basis of augmented reality applications, retailers can offer customers a better view of their products by placing a virtual object in the real world through a camera (Kim et al., 2017). This makes it easier for the customer to visualize the products in his or her environment. New technology also provides online retailers new delivery options which allows them to personalize delivery options to meet customers’ expectations. Examples of these delivery options are home delivery without

(12)

being home, delivery by drones, robots or self-driving cars in the future (INretail, 2017). INretail (2017) pointed out that a company called Jet.com already uses of one of these options. In New York, they deliver when customers are not at home via a keyless technology. The deliverer receives a temporary code with which he or she has access to the building. In this way, the deliverer can deliver packages if the residents are not at home (INretail, 2017).

New technology also creates opportunities for offline retailers (Grewal et al., 2017). For instance, new technology offers retailers tools to not only improve customer convenience, but also to make better informed decisions about their products and service and to respond more quickly to customer needs (Grewal et al., 2017). Examples of new technologies for offline retailers include proximity mobile payment technology, self-scanning technology, and interactive smart shelf technology (Inman & Nikolova, 2017). Proximity mobile payment technology allows consumers to pay with their smartphones for purchases in a physical store by keeping the smartphone close to the payment terminal for a short time (De Kerviler, Demoulin & Zidda, 2016). With self-scanning technology, customers can scan and pay their items themselves without the intervention of a cashier (Inman & Nikolova, 2017). Because the focus of this thesis is on customer acceptance of interactive smart shelf technology, this type of in-store technology is discussed more in detail below.

2.1.1 Interactive smart shelf technology

Interactive smart shelves are electronically connected shelves that operate on the basis of various new technologies and devices, such as cloud technology and mobile technology (Accenture, 2015). Interactive smart shelves project, including product information, cooking suggestions, information about the number of people in the store, top-selling products, and social media posts. In addition, the shelves help customers to navigate easily through the store and find the products they are looking for. Where the main objective of proximity mobile payment and self-scanning technology is to create shopping convenience, the objective of interactive smart shelf technology is to create a personalized, playful and informative shopping environment (Accenture, 2015).

The application of interactive smart shelf technology is beneficial to both retailers and customers (Inman & Nikolova, 2017). First, retailers can improve their operations and prevent lost sales (Bornemann, 2015). On the basis of weight sensors on the shelves, store personal is kept informed when the last item is removed from the shelf. In this way, it can be refilled in time (Inman & Nikolova, 2017). A second benefit Bornemann (2015) pointed out is that interactive smart shelves reduce food waste, because the shelf can automatically decline

(13)

prices as the expiration date nears. Third, where online retailers have the advantage of gathering customer data to improve their relationships with them, for offline retailers this is more difficult (Accenture, 2015). Yet, the interactive smart shelf technology provides offline retailers similar information about customer, e.g., about their demographics and preferences (Inman & Nikolova, 2017). The benefits of interactive smart shelf technology for customers include firstly that customer only receive relevant information about products and discounts, because they can request information themselves (Grewal et al., 2017). Secondly, the shelves provide customers the possibility to order a product from the store and have it delivered at home if it is too heavy or bulky to carry (Bornemann, 2015). Thirdly, since the interactive smart shelf technology prevents lost sales, it also reduces the risk of disappointment by the customers because of out-of-stocks (Inman & Nikolova, 2017).

2.2 Behavioral intention models

Technology has opened up opportunities to improve in-store customer experience (Foroudi, et al., 2018). However, retailers may face the challenge of gaining customer acceptance and usage of new technologies (Roy et al., 2018). Retailers cannot gain success from in-store technologies if they are not accepted or used by their customers. Therefore, it is important to understand why people may accept or reject technology. One of the most applied models of user’s attitude and usage of technology is the technology acceptance model designed by Davis (1989) (Venkatesh, 2000). The technology acceptance model is inspired by the theory of reasoned action of Ajzen & Fishbein (1975) (Davis, Bagozzi & Warshaw, 1989). Therefore, before discussing the technology acceptance model, the theory of reasoned action will be briefly explained below.

2.2.1 Theory of Reasoned Action

Theory of reasoned action (TRA) is a behavioral intention model from social psychology which is used for the prediction of intentions and behavior (Madden, Ellen & Ajzen, 1992). According to the theory, a specific behavior is determined by the intention of a person to perform that behavior. The stronger the intention, the higher the likelihood that the behavior will be performed (Davis et al., 1989). The intention is determined by the person’s attitude toward performing the behavior, and subjective norms about performing the behavior (Madden et al., 1992). Attitude is defined as a person’s positive or negative feeling about performing the behavior (Davis et al., 1989). Subjective norms include the perceived behavioral control within a social group (Davis et al., 1989). Madden et al. (1992) argued that

(14)

whereas a person’s attitude is influenced by the person’s behavioral beliefs, the subjective norms of a person are influenced by the person’s normative beliefs. Davis et al. (1989) explained behavioral beliefs as the person’s beliefs that performing a specific behavior will result in certain outcomes, and the evaluation (positive or negative) of those outcomes. Normative beliefs include first, a person’s belief that important persons to him/her think he/she should or should not perform certain behavior, and second, the person’s motivation to comply, i.e., the importance to the person to do what others expect of him or her (Davis et al., 1989). The theory states that other factors that influence behavior do this indirectly by influencing attitude or subjective norms (Madden et al., 1992).

Applying TRA to the interactive smart shelf technology acceptance, the likelihood of actual usage will be higher when the intention to use the technology is higher. The intention is determined by on the one hand, the consumer’s attitude, and on the other hand, the consumer’s subjective norms. In the view of that, when the customer perceives that using the technology would result in a outcome, and evaluate that outcome positively, and people who are important to the customer approve the behavior of using it, and the person is motivated to comply, the intention to use will be higher, and that will increase the likelihood of actual usage.

[Insert Figure 1; see Appendix A]

2.2.2 Technology Acceptance Model (TAM)

Above-mentioned, the technology acceptance model (TAM) is based on TRA. Similar to TRA, the TAM model proposes that a specific behavior is determined by the user’s intention (Davis, 1989). Different from TRA, initially, the TAM model did not measure the effect of subjective norms on technology acceptance (Davis & Venkatesh, 1996). Furthermore, whereas TRA states that beliefs to perform a specific behavior influence intention completely mediated through attitude toward the behavior, the technology acceptance model showed that certain beliefs not only indirectly influence intention through attitude, but also directly (Davis & Venkatesh, 1996).

Davis (1989) developed the technology acceptance model after he noticed that although computer systems have the potential to improve user’s performance, the user’s unwillingness to accept and use computer systems can implement performance gains. With the model, Davis (1989) aimed to develop a better measure for the prediction and explanation why individuals accept or reject computer systems. The TAM model assumes that the user’s

(15)

intention to use a computer system is the strongest determinant of actual system usage. Sequentially, the intention to use a computer system is determined by the person’s attitude toward using it, which in turn is determined by the person’s beliefs of perceived usefulness and perceived ease of use. Davis (1989) argued that perceived usefulness and perceived ease of use are user’s subjective assessment of performance and effort. Perceived usefulness refers to the degree to which the user believes that the use of the computer system will increase his/her job performance. Perceived ease of use describes the degree to which the user believes that the use of the computer system will be free of effort. The influence of perceived ease of use on attitude and intention is explained on the basis of two mechanisms: self-efficacy and instrumentality (Davis et al., 1989). The easier a system is to use, the greater the user’s sense of efficacy and perceived control with respect to his or her ability to perform the behavior that is necessary to use the system. Perceived ease of use is also instrumental by its influence on usefulness. Effort saved as a result of improved ease of use can be reused, allowing a person to do more work for the same effort (Davis et al., 1989). Davis (1989) emphasized that perceived usefulness is a stronger predictor of technology acceptance than perceived ease of use. Whereas difficulty of usage can discourage acceptation of a useful systems, no extent of ease of use can compensate for a system that does not increase user’s performance. Altogether, systems that are perceived to increase the overall job performance and are perceived to be easy to use are more likely to be accepted by the user (Davis, 1989).

[Insert Figure 2; see Appendix A]

2.2.3 Extension of TAM

Although the TAM model has been used in various empirical studies in which it has explained usually around 40% of the variance in intention and actual usage, the model has not been free from criticism (Venkatesh & Davis, 2000). The main criticism was that the model is incomplete. Consequently, the model is refined and extended.

An extension of the TAM model is the technology acceptance model 2 (TAM2) introduced in 2000 by Venkatesh and Davis. In this model, Venkatesh & Davis (2000) incorporated social influencing processes and cognitive instrumental processes as additional constructs, and excluded the construct of attitude toward a computer system. The scholars described social influence as a construct that includes subjective norms, voluntariness, and image. They explained subjective norms similar to Ajzen and Fishbein (1975), namely, as the perceived behavioral control within a social group. Voluntariness is defined as "the extent to

(16)

which potential adopters perceive the adoption decision to be non-mandatory" (Venkatesh &

Davis, 2000 p.188). Image is explained as the degree to which the user believes that the use of a computer system will increase his/her social status (Venkatesh & Davis, 2000). Venkatesh & Davis (2000) operationalized cognition by the relevance of the job, output quality, result demonstrability, and perceived ease of use. According to TAM2, subjective norms, perceived usefulness and perceived ease of use have a direct effect on the intention to use computer systems. Perceived ease of use has also an indirect effect on the intention to use computer systems mediated through perceived usefulness. The same applied to subjective norms, image, job relevance, outcome quality, and result demonstrability. Experience and voluntariness moderate the effects of subjective norms on perceived usefulness and intention to use computer systems. Finally, intention predicts the actual usage of them.

Another technology model is the unified theory of acceptance and use of technology (UTAUT) designed by Venkatesh, Morris, Davis & Davis (2003). In this model, four construct: facilitating conditions which is the user’s perception of the available resources and support to perform a certain behavior, performance expectancy (comparable to perceived usefulness), effort expectancy (comparable to perceived ease of use), social influence, and four moderating factors: gender, age, experience and voluntariness of use, are included. According to the UTAUT model, the four constructs, moderated through gender, age, experience and voluntariness, affect the intention to use computer systems, while the facilitating conditions and the intention predict the actual usage of them (Venkatesh et al., 2003).

[Insert Figure 3; 4; see Appendix A]

Others scholars have argued that the user’s cognitive beliefs, such as perceived usefulness and perceived ease of use, do not provide a full understanding of consumer’s attitude formation toward technology (Childers et al., 2001). Childers et al. (2001) argued that aside the cognitive beliefs about technology also the affective beliefs, i.e., enjoyment, should be examined (Childers et al., 2001). Enjoyment can be defined as the degree to which the activity of using technology is enjoyable in its own, apart from any job performance gains that may be perceived by the user (Teo & Noyes, 2011).

In their study in the context of online shopping, Childers et al. (2001) showed that aside perceived usefulness and perceived ease of use enjoyment was also a strong predictor of customers’ attitude toward online shopping. In similar vein, Van der Heijden (2004) argued

(17)

that there should be made a distinction between utilitarian versus hedonic systems. He argued that the validity of the TAM model find its limitation in hedonic systems which are used in circumstances where the objective of the user is not to improve job performance, but to have an enjoyable experience while using the system. Van der Heijden (2004) stated that perceived usefulness and perceived ease of use measure extrinsic motivation, i.e., the use of a system is based on the improved job performance. On the contrary, enjoyment measures intrinsic motivation, i.e., the use of a system is based on the joy that users derive from using the system as such. He proposed similarly to Children et al. (2001) to include the construct of enjoyment in the technology acceptance model.

This thesis is based on the original TAM model. In later versions, e.g., TAM2 and UTAUT, the attitude construct has been excluded from the models because there was lack of evidence for its direct effect on intention (Davis & Venkatesh, 1996). However, those studies were in the organizational context where individuals may use technology even if they do not have a positive attitude toward the technology (Kulviwat, Bruner II & Al-shuridah, 2009). The attitude construct plays a central role in marketing studies. Recent studies in the context of customer technology acceptance have shown direct effects of attitude on intention (e.g. Ha & Stoel, 2009; Roy et al., 2018). Therefore, in this thesis the original TAM model, which includes the construct of attitude, will be used. Furthermore, since this thesis aims to investigate customers’ cognitive affective, and social beliefs on their attitude toward technology and use intention, the TAM model will be extended by the constructs of enjoyment and social influence in addition to the constructs of perceived usefulness and perceived ease of use.

2.3 Hypotheses

2.3.1 Relationship between attitude and intention

As explained above, attitude involves a person’s positive or negative evaluation of performing a certain behavior, and is a key determinant of intention (Davis et al., 1989). Various studies have confirmed that attitude has a positive effect on intention. For instance, Ha & Stoel (2009) in the context of online shopping. Chen & Chang (2013) in the context of near-field communication (NFC) technology, including location-based service, mobile payment, peer gaming, and targeted advertising. The following hypothesis is drawn:

H1: As the attitude toward the interactive smart shelf technology becomes more positive, intention to use this technology will be higher.

(18)

2.3.2 Relationship between perceived usefulness, attitude, and intention

Perceived usefulness as a determinant of technology acceptance and resistance has been explored by several scholars in different fields of study. For instance, Weijters, Rangarajan, Falk & Schillewaert (2007) conducted a study in the field of self-service technology. They looked for predictors that influence customer’s attitude toward self-service technology, and the usage of it. Their data was collected in six stores of a grocery retail chain in Western Europe. The results indicated that perceived usefulness, perceived ease of use, and enjoyment have a positive effect on attitude toward the technology, and in turn, on the usage of it. From all the predictors, perceived usefulness appeared to be the strongest predictor of attitude and usage. Ha & Stoel (2009) investigated customer’s attitude toward online shopping and intention to shop online. Aside from perceived usefulness and perceived ease of use, they examined the influence of trust, enjoyment and the quality of website in terms of design, customer service, privacy and security, and atmospheric. They surveyed students at a university, and found that website quality had a positive influence on the consumer’s perception of ease of use, trust and enjoyment. Perceived usefulness, enjoyment and trust had a positive influence on attitude. The effect of ease of use on attitude was not significant. In line with Weijters et al. (2007), relatively to enjoyment and trust, perceived usefulness appeared to be the strongest predictor of attitude toward online shopping. Furthermore, Ha & Stoel (2009) found a positive effect of perceived usefulness on the intention to shop online. The study by Pantano & Priporas (2016) showed that perceived usefulness was a strong predictor of a person’s intention to use technology. The study of Pantano & Priporas (2016) was in the field of mobile shopping. They conducted a mix of in-depth interviews and web-based Skype interviews with 29 customers in the Italian market. The study’s major finding was that high perceived usefulness is an important driver of positive attitude toward mobile shopping. Most respondents perceived mobile shopping as convenience in terms of time and money saving. They were willing to shop via their smartphone to avoid queues in stores. Mani & Chouck (2017) studied customer resistance to smart products, such as smartwatches. Consistent with research findings on technology acceptance, they found a negative effect of perceived usefulness on users’ resistance. Customers who believe that smartwatches have a significant added value are less resistant toward it, but those who perceive smartwatches just as gadgets, perceive them to be less useful and are more resistant to it. Roy et al. (2018) examined customer acceptance of smart retail technology, such as self-service checkouts and personal shopping assistance. The respondents were asked to base their response on their most recent experience with a smart retail technology. The results indicated that high perceived

(19)

usefulness had a positive influence on attitude toward technology as well as intention to use it. The following hypotheses are dawn:

H2a: As the perceived usefulness of the interactive smart shelf technology increases, attitude toward this technology will become more positive.

H2b: As the perceived usefulness of the interactive smart shelf technology increases, intention to use these technology will become higher.

H2c: The positive effect of perceived usefulness on intention to use, is mediated through attitude.

2.3.3 Relationship between perceived ease of use, perceived usefulness, attitude, and intention

Literature has shown inconsistent findings regarding the effect of perceived ease of use on attitude toward technology and intention to use it. The study of Ha & Stoel (2009) showed that the effect of perceived ease of use on attitude toward online shopping was not significant, nor was its effect on intention to shop online. They only found an indirect effect of perceived ease of use on attitude and intention mediated through perceived usefulness. The purpose of the study of Agrebi & Jallais (2015) was to identify factors that influence intention to use smartphone for purchases. Their study differs from the study of Ha & Stoel (2009) in that it only tested the effects of belief factors on intention without the intermediation of attitude. Furthermore, Agrebi & Jallais (2015) took into account the individual’s experience with purchases with smartphones. The sample was divided into two groups: individuals who had already made purchases with their smartphone, experienced users, and individuals who had never used their smartphones for purchases, non-experienced users. For the two groups together and the two group separately, they found positive indirect effects of perceived ease of use on intention to use the smartphone for purchases mediated through perceived usefulness. They explained this by the fact that people may use their smartphones daily, e.g., for emails, or to surf the Internet. That is why the majority of the respondents may already have experiences with the use of smartphones in general. As Davis et al. (1989) suggested that the positive influence of perceived ease of use on intention to use is present when the technology is new, because then individuals are more likely to process choice options with the help of abstract, general criteria, because they have not followed the learning needed to understand and judge more concrete, specific criteria (Agrebi & Jallais, 2015). Weijters et al. (2007) found a positive effect of perceived ease of use on consumer’s attitude toward self-service technology. This also applied to Roy et al. (2018). Furthermore, Roy et al. (2018)

(20)

found a positive effect of perceived ease of use on intention to use mediated through perceived usefulness and attitude. The following hypotheses are drawn:

H3a: As the perceived ease of use of the interactive smart shelf technology increases, attitude toward this technology will become more positive.

H3b: As the perceived ease of use of the interactive smart shelf technology increases, intention to use this technology will become higher.

H3c: The positive effect of perceived ease of use on intention to use, is mediated through perceived usefulness and attitude.

2.3.4 Relationship between enjoyment, attitude, and intention

Various studies have investigated the effect of enjoyment on attitude toward technology and intention to use it. For example, Dickinger, Arami & Meyer (2008) examined the role of enjoyment in mobile technology acceptance. Their results showed that the enjoyable aspects of mobile technology are important determinants of the positive attitude toward the technology as well as intention to use it. Enjoyment appeared to be a stronger predictor of attitude and intention than perceived usefulness. This finding is not in line with later findings by, e.g., Ha & Stoel (2009) who found perceived usefulness to be the strongest determinant of attitude toward online shopping. De Kerviler, Demoulin & Zidda (2016) explored the role of enjoyment in the intention to use smartphones in stores for collecting information and contactless payments. They found that enjoyment was a strong predictor of the acceptance of contactless payment. The effect was twice as large as the effect of perceived usefulness. Enjoyment had also influence on customers’ intention to use their smartphone for information search in stores. However, the effect was not as strong as the effect on contactless payment. Kim et al. (2017) studied customers’ attitude toward virtual mirror, socially interactive dressing room, and radio frequency identification music tag, and their intention to use these technologies. The virtual mirror enables customers to try clothing and accessories without wearing any of them by overlapping the image of the customer with photos of clothing and accessories. The socially interactive dressing room consists of a mirror, camera, and a touchscreen computer with a built-in camera. In this room visitors have access to their social network sites such as Facebook, which offers customers the opportunity to ask friends for advice about their outfits. With radio frequency identification music tag retailers can play music that corresponds to the outfit style of the customer to improve the customer’s shopping experience. In line with Dickinger et al. (2008), Kim et al. (2017) found that enjoyment is a stronger predictor of positive attitude toward the technologies than perceived usefulness.

(21)

Furthermore, they found differences in the effects of perceived usefulness and enjoyment on intention to use between different technology types. That is, whereas perceived usefulness was the strongest predictor of intention to use virtual mirrors and socially interactive dressing rooms, enjoyment was the strongest predictor of radio frequency identification music tags. The following hypotheses are drawn:

H4a: As the enjoyment of the interactive smart shelf technology increases, attitude toward this technology will become more positive.

H4b: As the enjoyment of the interactive smart shelf technology increases, intention to use this technology will become higher.

H4c: The positive effect of enjoyment on intention to use, is mediated through attitude.

2.3.5 Relationship between social influence, attitude, and intention

Studies on technology acceptance have confirmed the impact of social influence on it. Kulviwat et al. (2007) examined social influence on customers’ attitude toward innovative products and their intention to use them. The innovative product used was the latest type of personal digital assistant (PDA) which was relatively new at the time of the study and was not available to the public (Kulviwat et al., 2007). The product had multimedia capabilities and allowed consumers multitasking. Kulviwat et al. (2007) found that social influence had a positive effect on attitude toward the technology, and on intention to use it. When potential users were convinced that people who are important to them have endorsed the technology, they had a more positive attitude toward it, and were more likely to have a higher intention to use it. Kulviwat et al. (2007) also found that products were more likely to be used when a person believed that the use of the product would increase his or her social status. Yang (2012) found a positive effect of social influence on the intention to use mobile shopping. Users took the opinion of people who are important to them into account in their decision making regarding mobile shopping. Gao & Bai (2014) studied the role of social influence in the acceptance of electronic toll collection. Electronic toll collection is a system that aims to eliminate delay on toll roads by collecting toll via automatic toll gates without requiring cars to stop. The results showed that social influence positively affected intention to use electronic toll collection. In their study, the respondents ranged from 24 to 30 years. Gao & Bai (2014) argued that this age range is most vulnerable to being influenced by mass media or peers in their decision to use technology. This may explain their results. The study by Koenig-Lewis, Marquet, Palmer & Zhao (2015) found similar results. Koenig-Lewis et al. (2015) investigated the impact of social influence in the field of mobile payment technology. It was

(22)

found that social influence is a predictor of the acceptance of mobile payment technology. They explained this by the fact that mobile payments are generally carried out in a public or social context where users can be observed by others. Therefore, it is more likely that mobile payment users are affected in their use of the technology by other customers. Chen & Chang (2013) explored the influence of social influence on near-field communication (NFC) technology. Chen & Chang (2013) found a positive effect of social influence on users’ attitude toward the NFC as well as a on intention to use the NFC technology. They concluded that one is more likely to use the technology when important others, e.g., family and friends, use it.

H5a: As the social influence on the acceptance of the interactive smart shelf technology increases, attitude toward this technology will become more positive.

H5b: As the social influence on the acceptance of the interactive smart shelf technology increases, the intention to use this technology will become higher.

(23)

2.4 Conceptual model: Interactive smart shelf technology

H2b + H3c H2a + H3a + H3b + H2c H3d H1 H4c H4a + H5c H5a H5b H4b

Control variables: Age, Gender, Education level Figure 5. Conceptual model of beliefs, attitudes and intentions towards Interactive Smart Shelf Technology (ISST).

Perceived usefulness

Perceived ease of use

Enjoyment Social influence Attitude toward ISST Intention to use ISST

(24)

3. Methodology

In this chapter the empirical part of this thesis is explained. First, the design is outlined. Second, the sample is defined. Third, the operationalization of the variables and the scale formation are explained. Followed by the procedure section in the fourth section. In the fifth section the data preparation is explained. This chapter ends with the description of the data analysis strategy, and assumption testing of the analysis used in the sixth section.

3.1 Research design

This thesis took a quantitative research approach utilizing cross-sectional survey methodology. The survey consisted of nineteen technology belief, seven attitude toward technology, and three intention to use technology item questions, a test question on the content of a prior shown video, and questions on gender, age and educational level. The purpose of the design was to examine how customers’ beliefs influence their attitude toward interactive smart shelf technology, and intention to use it.

3.2 Sample

The sample of this study consisted of customers who live in the Netherlands. Because there was not a list of all customers available, and the probability of each customer being selected from the total population was not known, data was collected using self-selection sampling, which is a type of non-probability sampling technique (Saunders, Lewis and Thornhill, 2009 p.213). In a self-selection sampling, a survey is published, e.g., on a forum or in a magazine, and subsequently, individuals are invited to take part in the research (Saunders et al., 2009 p.241). In this case, the survey was posted on own Facebook and LinkedIn pages as well as Facebook and LinkedIn pages of acquaintances. Furthermore, the survey was posted on Facebook discussion group pages.

Within two weeks, 223 respondents filled out the questionnaire. From the 223 responses, 207 responses appeared to be useful data for the analyses. Taking the 207 respondents together (Mage = 24.8, SDage = 5.9, age-range: 18-67 years) 66.2% were female, 0.5% low level vocational, 3.9% general secondary education, 9.2% preuniversity education, 0.5% intermediate vocational education (level 1+2), 7.2% intermediate vocational education (level 3+4), 34.8% higher vocational education 43.5% university, 0.5% post academic education.

(25)

3.3 Materials

3.3.1 Measurements

The technology belief items were grouped together into four constructs: perceived usefulness, perceived ease of use, enjoyment, and social influence. Whereas perceived usefulness and perceived ease of use measured customer’s cognitive beliefs, enjoyment measured customer’s affective beliefs, and social influence measured customer’s social beliefs. All four constructs served as independent variables. Perceived usefulness also served once as an dependent variable and once as a mediator variable. The seven attitude items formed an attitude toward technology construct, and the three intention items formed the sixth and last construct: intention to use technology. Attitude served as a dependent variable, and mediator variable. Intention only served as a dependent variable. All six constructs were based on existing scales originating from previous studies. The perceived usefulness, perceived ease of use, enjoyment, and attitude constructs were adapted from Childers et al. (2001). The social influence construct was adapted from Gai & Bai (2014), and the intention construct was adapted from Kim et al. (2017).

A description of the constructs, test question and demographic measurements are provided in Table 1, and are summarized below.

[Insert Table 1; see Appendix B]

3.3.1.1 Perceived usefulness, perceived ease of use, enjoyment, social influence, attitude, intention to use

Perceived usefulness consisted of four items and measured the degree to which the

respondents believed that the use of the interactive smart shelf technology would increase their shopping productivity. For instance, the respondents were asked to rate how strongly they agreed or disagreed with the following statement: "Interactive smart shelf technology

would enhance my effectiveness in supermarket shopping". Childers et al. (2001) reported a

Cronbach’s α of .92 and .93 for different studies.

Perceived ease of use also consisted of four items. The items measured the degree to

which the respondents believed that using the interactive smart shelf technology in completing their shopping task would be easy and free of effort. An example of an item from this construct was: “Interactive smart shelf technology would be clear and understandable”. For the four items of perceived ease of use, Childers et al. (2001) reported Cronbach’s α of .79 and .99.

(26)

Enjoyment consisted of eight items, and measured the degree to which the respondents

believed that using the interactive smart shelf technology in completing their shopping task would be enjoyable. Two items in this construct had a reverse phrasing. For instance, an example of a positively worded item: “Shopping in supermarkets with interactive smart shelf

technology would be exciting”, and a negatively worded item: “Shopping with interactive smart shelf technology would be boring”. The reverse-phrased items are important to reduce

response bias, because participants will have to read the items in case they are phrased the other way around (Field, 2009 p.675). The negative worded items were reversed during the analyses. Cronbach’s α of 0.88 and 0.93 were reported for this construct (Childers et al., 2001).

Social influence consisted of three items. The items measured the degree to which the

respondents believed that most important people to them would think they should or should not use the interactive smart shelf technology. The goal of this construct was to test whether the norms in the respondents’ social environment with respect to the use of the technology would have an influence on their own attitude toward the technology and intention to use it. An example of an item: “People who are important to me would think I should use interactive

smart shelf technology”. Gao & Bai (2014) reported a Cronbach’s α of 0.86.

Attitude was measured on the basis of seven sematic differential items. The items

measured the respondent’s overall affective evaluation of the interactive smart shelf technology in terms of seven words. For instance, “Not Wortwhile - Worthwhile” or “Bad –

Good”. The Cronbach’s α of this construct were 0.89 and 0.93 (Childers et al., 2001).

Intention consisted of three items. The items asked the respondent whether they would

use interactive smart shelf technology or visit a store that provides its customer the technology. For example by asking them the following statement: “I would do my shopping

in a supermarket that provides its customers with interactive smart shelf technology”.

All items from the perceived usefulness, perceived ease of use, enjoyment, social influence, and intention constructs were measured on a seven-point Likert-type scale ranging from strongly disagree (1) to strongly agree (7). The attitude construct was measured on a seven-point sematic differential scale. The evaluation ranged from extremely negative (1) to extremely positive (7). Higher scores indicate that interactive smart shelf technology were perceived as more useful, easy to use, enjoyable, a social environment in which the use of it is accepted, a more positive attitude toward the technology, and a higher intention to use it.

(27)

3.3.1.1.1 Items in the questionnaire

All items, sourced from different fields of study, were altered to the context of interactive smart shelf technology in supermarkets. Furthermore, because this study targeted Dutch customers, the questionnaire was administered in Dutch to ensure that everyone could participate in the study, and to avoid potential misunderstandings of English terms in the questions. For the translation of the English items into Dutch, the parallel translation procedure was used (Saunders et al., 2009 p.385). First, the items were translated into Dutch by three independent Dutch native speakers. The three versions were compared, and merged into one version. Then, the questionnaire was discussed with a fourth person. This person was asked to translate the questionnaire back into English to compare it with the original English version. Furthermore, the questions were assessed for clarity.

3.3.1.1.2 Factor structure of the constructs

Because the item questions were translated from English into Dutch, an exploratory factor analysis using a principal-axis factor (PAF) extraction with direct oblimin rotation was conducted in order to be sure that the items represent the constructs as defined. First, the nineteen technology belief items were factor analysed. The Kaiser-Meyer-Olkin measure verified the sampling adequacy for the analysis, KMO = .897, above the recommended value of .6, and Bartlett’s test of sphericity χ2 (171) = 2732.405, p < .001, indicated that correlation between items were sufficiently large for exploratory factor analysis (Field, 2009 p.659). An initial analysis was run to obtain eigenvalues for each factor in the data. The choice of factors with an eigenvalue over Kaiser’s criterion of one led to four factors which in combination explained 63.33% of the total variance. The scree plot was ambiguous and showed inflexions that would justify retaining either one or five factors. To explore these factor structures, two new factor analyses with principal-axis factor extraction and direct oblimin rotation were conducted with one fixed number of factor, and five fixed number of factors. The analysis with one fixed number of factor showed a decrease in the total explained variance from 63.33% for four factors to 41.28% for one factor. The factor analysis with five fixed factors showed that the total explained variance increased to 66.96%. However, threeitems appeared to show cross-loading on two or three factors. Because of the decrease in explained variance with one factor, high number of cross-loadings with five factors, and based on previous theoretical support, the four factor solution was preferred. This solution of the four factors, showed that item 14 had a cross-loading on two factors. These loadings remain below .5. Therefore, item 14 was removed from the factor analysis. The Kaiser-Meyer-Olkin measure

(28)

from the analysis without item 14 verified the sampling adequacy for the analysis, KMO = .885, above the recommended value of .6, and Bartlett’s test of sphericity χ2

(153) = 2474.468, p < .001, indicated that correlation between items were sufficiently large for PAF (Field, 2009 p.659). The four factors explained 62.81% of the total variance. Table 2 presents the factor loadings of the eighteen items after rotation. The items that cluster on the same factors suggest that factor one represents enjoyment, factor two social influence, factor three perceived ease of use, and factor four perceived usefulness.

Second, the attitude toward technology and intention to use technology items were factor analysed using a principal-axis factor (PAF) extraction with direct oblimin rotation. The Kaiser-Meyer-Olkin was .906 which is well above the recommended threshold of .6, and Bartlett’s test of sphericity χ2

(45) = 1713.359, p < .001, indicated that correlations between items were sufficiently large for exploratory factor analysis (Field, 2009 p.659). Two factors were extracted explaining 68.68% of the total variance, based on the Kaiser’s criterion. In agreement with the Kaiser’s criterion, examination of the scree plot revealed a levelling off after the second factor. Therefore, two factors were retained. Table 3 presents the factor loadings after rotation. The items that cluster on the same factor suggest that factor one represents attitude toward technology, and factor two intention to use technology.

[Insert Table 2; 3; see Appendix B]

3.3.1.1.3 Reliability of the constructs

The items from all constructs suggested by prior research have remained similar expect for the enjoyment construct. Following the factor analysis item 14 has been removed from this construct. First, a reliability analysis was conducted with the four items of perceived usefulness. The items provided a high reliability with Cronbach’s α = .853. None of the items would substantially increase reliability if they were deleted. Second, the four perceived ease of use items provided also high reliability, with Cronbach’s α = .854. As the results suggested, when deleting item 7, the Cronbach’s α would increase to .868. However, because the Cronbach’s α with the 4 items is above .7, which already reflects a good degree of reliability (Field, 2009 p.679), and the increase is just .014, it was decided not to delete the item. The seven enjoyment items provided a Cronbach’s α of .867. By deleting item 11, the Cronbach’s α would become .872. Because of the small increase, and the overall Cronbach’s α is excellent because it is above .8 (Field, 2009 p.679), it was decided not to delete the item. Fourth, the Cronbach’s α of the three social influence items was .949, and none of the items

(29)

would increase the reliability if they were deleted. Then, the seven attitude toward technology items showed a Cronbach’s α of .911, and could only be increase to .912 if attitude item 4 was deleted. Because of the small increase the item was not deleted from the scale. Finally, the Cronbach’s α of the three intention to use items was .945. All values were less than the overall reliability of .949, and thus are not deleted. The corrected item-total correlation from all items in the six reliability analyses indicated that all items have a good correlation with the total score of the scales (all above .3). The items were computed into scales by taking the average of the items.

[Insert Table 4; see Appendix B]

3.3.1.2 Demographic variables

All Dutch customer were free to participate in this study. To rule out potentially spurious relations, in all analyses there was controlled for age (in years), gender (1 = male, 2 = female), and educational level. The different levels of education were scaled in the following way: 1=primary education (LO), 2=low level vocational (LBO), 3=low level general (MAVO), 4=secondary education (HAVO), 5=preuniversity education (VWO), 6=intermediate vocational education, level 1+2 (MBO), 7=intermediate vocational education, level 3+4 (MBO), 8=higher vocational education (HBO), 9=university (WO), 10=post academic. These categories are adapted from the Family Survey Dutch Population, collected by Kraaykamp, Ruiter and Wolbers (2009), and will capture the most common educational levels in the Netherlands.

3.3.1.3 Video fragment and test question

In the Netherlands, supermarkets have not yet reached the point of integrating the interactive smart shelf technology into their stores. Therefore, the respondents were presented a video about the interactive smart shelf technology. The video presented the entire layout of interactive smart shelves as well as the way they could be used during shopping. The video was in English with Dutch subtitles, and lasted approximately 1.20 minutes.

To ensure that all the respondent had seen the video before answering the questions a test question was included in the questionnaire. The question stated: “What was the video

about that you just saw?”, and consisted of three answer options: 1) “a supermarket where you can use self-scanning”, 2) “a supermarket where you can request product information”,

3) “a new attraction in an amusement park”, with the second option being the correct answer. In the video the function of requesting product information was one of the main functions of

(30)

the interactive smart shelves that was discussed. Self-scanning and attractions were not discussed in the video. During the analyses, this question has been recoded into a two category variable: correct versus wrong answer, to access the usability of the data.

3.4 Procedure

The survey consisted of two sections. In the first section, the respondents were told that they would first watch a video fragment of a supermarket with interactive smart shelf technology prior to starting with the questions. In the second section, the respondents were presented the item questions, and were asked to what degree they agree or disagree with the items based on what they have had seen in the video. After these items questions, the respondents were asked to complete the survey answering a question on the content of the video (test question), and questions on gender, age, and their educational level. All questions were equipped with a forced validation. The respondent was unable to proceed to the next question before the previous question was answered, nor was the respondent able to go back to prior questions.

3.5 Data preparation

Missing values - First of all, data were screened for missing values. From a total of

223 responses, there were missing values from five respondents. Three respondents had filled out less than 30% of the questionnaire, and had not answered the fourth question which was the test question. Other two respondents had not answered the question on age.

Unengaged responses - Secondly, the data file was explored for unengaged responses.

Two respondents were unengaged as evidenced by giving the exact same response for every single item. Furthermore, they had completed the questionnaire within on average of 1 minute, which is not reasonable because the video alone lasted approximately 1.20 minutes.

Test question - Thirdly, the test question was analysed to assess the usability of the

data. The test question suggested that twelve respondents did not watch the video, since they had answered the question incorrectly.

Outliers - Lastly, the data were observed for outliers by z-transforming the variables.

Cases with z-values > -3 or > 3 are more than 3 standard deviations above or below the mean of a case, and were considered as outliers. From all the variables, the variables gender, age, and education showed outliers. The outliers were examined to ensure no data entry or instrument errors were made. Then, the Cook’s Distance measure was developed. Based on

(31)

this measure, it was determined that the outliers were not influential cases, and thus would not affect the results (Cook <1) (Field, 2009 p.217). Therefore, it was decided not to remove these respondents from the data file.

Data cleaning – The three respondents who had filled in 30% of the questionnaire, the

two unengaged respondents, and the twelve respondents who had answered the test question incorrectly, were removed from the data file. It was discovered that one of the two unengaged respondents, and the two respondents who had not answered the age question were among those who had answered the test question incorrectly. The analyses are based on data from 207 (N=207) respondents in total, without any missing values.

3.6 Data analysis strategy

The hypotheses in this study included both direct and indirect effects. The direct effects were analysed using regression analysis. The indirect effects were analysed using regression analysis in combination with the MACRO PROCESS procedure by Hayes (2012). Below the assumptions of the regression analysis are discussed.

3.6.1 Assumptions regression analysis

Variable type – The dependent and independent variables were measured on a Likert

scale, and thus, are at the interval level. Therefore, this assumption was met.

Multicollinearity – The multicollinearity was tested by conducting a correlation matrix

with all the independent variables. Correlations above .8 or .9 indicate multicollinearity (Field, 2009 p.224). The results of the correlation matrix showed that all the correlation between the variables are below .8 and thus, the assumption of multicollinearity was met.

Linearity and homoscedasticity - To test linearity and homoscedasticity, plots were

conducted of the standardized residuals (ZRESID) on the Y-axis against the standardized predicted values (ZPRED) on the X-axis for both dependent variables (perceived usefulness, attitude and intention) with the independent variables (perceived usefulness, perceived ease of use, enjoyment, social influence, and attitude). The points in the plots between perceived usefulness and perceived ease of use were completely randomly distributed, same applied for the points between perceived usefulness, perceived ease of use, enjoyment and intention. The points in the plots of the rest of the variables were also more and less randomly distributed, but the cloud point was inclined to collect more above or more below the diagram. However, none of the plots displayed a certain pattern. Therefore, it was concluded that these two assumptions were met (see Figure 6 to 15).

(32)

Independence – Durbin-Watson test was performed to test the independence between

the variables. Following the rule of < 1 or > 3 for the Durbin-Watson test (Field, 2009 p.221), it could be concluded that the residuals were independent, varying from 1.84 till 2.05. Thus, this assumption was met.

Normality –Histograms and P-P plots of the standardized residues were conducted to

test whether the data was normally distributed. Moderate to extreme deviations from the normal distribution have been observed. Therefore, the assumption of normality was not met (see Figure 16 t/m 25).

Predictors of customer attitude toward interactive smart shelf technology and intention to use it in supermarkets