Grocery retail apps and the privacy calculus: the impact of different types of personalization on the adoption intention of consumers Master Thesis MSc Marketing Intelligence

(1)

1

Grocery retail apps and the privacy calculus: the impact of different types of

personalization on the adoption intention of consumers

Master Thesis MSc Marketing Intelligence

(2)

2

Abstract

The use of smartphones and other mobile devices is growing rapidly around the world. The mobile channel has therefore become a increasingly important for retailers. Grocery retailers are

developing applications that can assist consumers prior to, during and after shopping. These applications gather valuable data and retailers are therefore devoted to increasing adoption rates. Personalization is one of the ways retailers hope to achieve a wider acceptance of these mobile shopping aides. There are several ways in which a mobile application can be personalized to the consumer, this paper argues that only service personalization and price personalization will

positively affect adoption behaviour. Furthermore, the extent to which a person’s privacy is intruded was found to negatively influence the adoption intention of consumers. Lastly, the level of privacy concerns a person holds was found to moderate the relationship between personalization and adoption intention. This paper contributes to existing literature by providing new insights into the role of different types of personalization in the privacy calculus. Consumers become increasingly aware of the way firms can harm their privacy. Managerially, this thesis recommends firms to carefully consider the impact of their privacy strategy. Especially, when it comes to the disclosure of location based data.

Problem statement

Mobile shopping

The use of mobile devices such as smartphones is growing rapidly worldwide, currently there are an estimated 2.5 billion smartphones in the world (eMarketer 2014). The increase of smartphone users worldwide has given rise to new opportunities and challenges for marketeers. For example, the increasing number of smartphone users has led to a new form of Ecommerce called commerce. M-shopping refers to using smartphones or tablets to compose, modify or place orders online (Wang, Malthouse et al. 2015). M-shopping has significant advantages over regular shopping, research has found that shoppers who order via a mobile device have higher order rates, higher order sizes and show more customer loyalty (Wang, Malthouse et al. 2015). Currently, smartphones and applications are already used during 17-21% of retail sales (Brinker, Lobaugh et al. 2012).

(4)

4 For now this digital revolution seems to be most prevalent in Asia, however it might give an idea of what the retail environment in Europe might look like. Nowadays, mobile devices are already transforming the in-store shopping experience. In 2013, 62% of US smartphone users was using their smartphone to assist with shopping at least once a month and 17% did so every week (Google Shopper Marketing Council 2013). In the Netherlands, most of the large grocery retailers have launched an app in order to serve as a supplement to traditional offline shopping. Grocery apps offer features such as grocery lists, weekly sales promotions, recipes, store locator and self-scan through the phone’s camera. Another aspect that some of these apps try to incorporate is the offering of personalized services like automatic grocery lists or personalized discounts based on previous shopping behaviour. Mobile has become an important marketing channel due to its three unique characteristics: 1) ultra portability 2) location-sensitivity, and 3) untetheredness (Shankar, Balasubramanian 2009). Mobile marketing is therefore distinctly different from mass marketing channels in that it allows firms to deliver location specific messages, track consumer responses and specifically target consumers at a much lower price. Rapid customer adoption of mobile creates a new and important touchpoint in the customer journey and retailers need to adopt a mobile mindset in order to stay competitive (Brinker, Lobaugh et al. 2012).

(5)

5 argues that app adoption is dependent on a combination of factors. It will focus on the relation between privacy issues and the level of personalization the app offers. The core of the argument is that personalized services are more relevant (feel more natural) and will therefore lead to higher chances of adoption. It can do so by offering an integration of features that is completely and automatically customized to the consumer’s characteristics and preferences. The outline of this paper will be as follows: first, models of technology adoption will be discussed and related to the use of grocery retail apps, then the role of privacy and the privacy calculus will be introduced as an important determinant of technology adoption, lastly the concept of personalization will be introduced and the trade-off between personalization benefits and information disclosure will be discussed.

Consumers use their smartphone to prior to shopping, during shopping and after shopping. During the (pre)shopping phase consumers use their phone to check the location and opening hours of the store or by making price comparisons between products. At the moment 82% of the customers who use their smartphone to assist them in with in-store shopping are using search engines to do so, while only 21% of smartphone shoppers use a store app. This is unfortunate for several reasons 1) store apps allow retailers to collect valuable customer data 2) store apps allow retailers to conveniently increase customer loyalty though loyalty programmes and 3) store apps allow retailers to be with the consumer at all times. Currently, retailers often get most of their customer data from store loyalty cards. These cards give retailers information on what customers buy, which store they buy in and when they buy. Based on this information retailers can better forecast sales, predict consumer behaviour or offer their customers more personalized discounts and contextual offers. These contextual offers have been found to trigger shopping behaviour in consumers and increase basket size (Ramanathan, Dhar 2010), frequency of order and unplanned purchases (Heilman, Nakamoto et al. 2002). Thus, this consumer data is very valuable for retailers.

Mobile data

(6)

6 a supplement to offline retail shopping, but rather a supplement to a number of connected devices that will serve as a bridge between consumers and retailers. Those retailers that will be the able to deliver the best mobile experience to their customers will see the benefits in increased order sizes, increased frequency of orders and more customer satisfaction (Wang, Malthouse et al. 2015). However, introducing an app that will be adopted by the consumer can be quite hard. Research has shown that people only use a few apps regularly, on average consumers have about 40 apps installed on their smartphone, however they only use about 15 of them actively (Gupta 2013). Only a few of the used apps belong to stores or brands, the competition on the consumer’s phone is therefore large. Furthermore, the challenge for marketers will be to design apps that help shoppers search and discover appropriate solutions and deliver a rich experience in a device-agnostic manner (Shankar, Kleijnen et al. 2016).

Personalization and privacy

The increasing technological capabilities have allowed firms to gather vast amounts of consumer data which can be processed by ever more sophisticated data analyses techniques and software. Online retailers partner up with search engines in order to gather even more data on the behaviour of their customers. By processing this data firms are able to personalize their services and marketing efforts. Personalized services offer many advantages to both consumers and firms. Consumers can enjoy products and services that better match their preferences (Vesanen 2007). It also reduces the chance of cognitive overload of consumers and increases convenience by offering relevant products and services (Ansari, Mela 2003). Furthermore, personalization can even grant firms a competitive advantage over their competitors (Murthi, Sarkar 2003). Also, personalization allows to charge higher prices (Vesanen 2007) and increase their profitability (Zhang, Wedel 2009). The rise of data driven mass marketing has also fuelled a debate concerning the privacy rights of consumers in a digital environment (Ashworth, Free 2006, LaRose, Rifon 2007, Westin 2003). Marketing organizations now capture an extensive amount of consumer data from multiple touchpoints in the customer journey. Emerging technological capabilities allow companies to easily capture consumers attitudinal responses and track online behaviours, media usage, physical location, click through rates and other detailed information used to create personalized relationships (Montgomery, Smith 2009). These developments have caused an ethical debate concerning the extent to which companies should be allowed to track and use consumer data. New laws and

regulations are being imposed in order protect consumers’ digital privacy (European Commission 2017). Because data is very valuable and many firms are trying to convince their customers to provide more data by offering them advantages like discounts. This paper aims to shed light on the degree to which

(7)

7 Furthermore, information security issues like data hacking, phishing, identity theft and compromised databases are ever more present and increased media attention for issues like these can increase the perceived vulnerability of consumers. The increased feelings of vulnerability of consumers might be even more exacerbated due to recent events concerning data breaches, for example when a US hospital lost thousands of patient records when their security was breached (O'Hara 2017). Also, recent events regarding the use of consumer data by Cambridge Analytica in the US elections may increase distrust against firms who collect data and make consumers more weary to share their data (Halpern 2018).

Research contribution

(8)

8

Conceptual model

Models of technological innovation adoption

A lot has been written about the factors that influence the adoption of technological innovations. Rogers (1983) suggests that all innovation adoption is influenced by five factors 1) relative advantage 2) complexity 3) divisibility 4) compatibility and 5) communicability. Other theories suggest that the adoption of new technologies boils down to two factors: Perceived usefulness and perceived ease of use (Davis 1989). Other theories combine several elements to come up with a framework aimed to explain adoption behaviour (Taylor, Todd 1995). Different situations require more specific theories. Other research suggest that the perceived enjoyment is an important contributing factor into the adoption of mobile shopping (Agrebi, Jallais 2015). Perceived usefulness (Yang 2010, Taylor, Todd 1995, Davis 1989), perceived enjoyment (Holmes, Byrne et al. 2013, Yang 2012, Agrebi, Jallais 2015) and convenient access (Davis 1989, Wang, Malthouse et al. 2015) are considered to be the most important factors leading to the adoption of m-shopping. This research aims to use the standard technology acceptance model of (Davis 1989) to explain the adoption intention for grocery retail apps. This model will be extended with relevant factors that explain adoption intention.

The technology acceptance model (TAM) (Davis 1989)

The TAM was developed in order to explain the reasons why people in organisations would not accept computer technology besides the presence and benefits of computer systems. According to the model there are two factors that influence people’s attitudes towards new technology 1) the perceived usefulness of this new technology and 2) the perceived ease of use.

(9)

9 to the context of this study. One of the problems with these models lies in the fact that they were both used to describe technological adoption within an organisational context. The purpose of the technology and the relation between people and the technology is significantly different within an organisational context compared to a consumer context. It is therefore important to reconsider some of the proposed hypotheses of the TAM2. For example, according to TAM2 the subjective norm has no significant direct effect on intention to use when system use is perceived to be voluntary (Venkatesh, Davis 2000).

The TAM and TAM2 are both derived from earlier behavioural psychology theories on the relationship between people’s beliefs and attitudes vs their behaviour. The Theory of Reasoned Action (TRA) and the Theory of Planned Behaviour (TPB) suggest that people base their behaviour on the attitude they hold towards that behaviour, the perceived control they have over that behaviour and a subjective norm (Fishbein, Ajzen 1975). The TRA and TPB differ from the TAM in the sense that they allow the perceived benefit and perceived usefulness directly influence a person’s behaviour rather than indirectly through attitudes. The TRA and TPB are widely used in order to explain consumer behaviour across multiple situations. However there are some aspects this theory does not take into account that can be relevant for this study. First of the TRA, TPB as well as the TAM and TAM2 all imply that consumer behaviour is derived from the conscious reasoning of the potential results of their behaviour. This conscious reasoning aspects has later been challenged by researchers such as Daniel Kahneman. Other research has shown that human behaviour is more often than not the result of unconscious rather than conscious processing (Bargh, Chartrand 1999). For this research those unconscious decisions are less relevant since we can only measure the decisions and actions that people are aware of.

(10)

10 to the privacy calculus consumers weigh the costs and the benefits against each other before they decide to provide information. Privacy concerns are an important factor for consumers when it comes to deciding to download an app and it should therefore be included into the TAM model. (Fortes, Rita 2016).

Privacy and privacy intrusion

Privacy is concept that can be defined in numerous ways, depending on which academic field of research you are undertaking. Basically, privacy boils down to ‘’the right to be left alone’’ or “ the freedom from intrusion of a person's seclusion or solitude” (Prosser 1960). Later “information privacy” was defined as the right of “individuals, groups, or institutions to determine for themselves when, how, and to what extent information about them is communicated to others” (Westin 1967). The right to be protected from intrusion can also include the right to be free from unwanted

marketing solicitations (Petty 2000). The rise of data driven mass marketing has also fuelled a debate concerning the privacy rights of consumers in a digital environment (Ashworth, Free 2006, LaRose, Rifon 2007, Westin 2003). Marketing organizations now capture an extensive amount of consumer data from multiple touchpoints in the customer journey. Emerging technological capabilities allow companies to easily capture consumers attitudinal responses and track online behaviours, media usage, physical location, click through rates and other detailed information used to create personalized relationships (Montgomery, Smith 2009). These developments have caused for an ethical debate concerning the extent to which companies should be allowed to track and use consumer data. New laws and regulations are being imposed in order protect consumer’s digital privacy (European Commission 2017).

Consumers are also increasingly concerned about the way marketeers make use of the data they provided through mobile apps (Xu, Gupta et al. 2012). In 2006, 90% of US consumers indicated that they are concerned about threats to their personal privacy (Best, Krueger et al. 2006). Many consumers are afraid that their information is being collected and used without their consent. Identity theft, phishing and hacking are also increasingly the concern of many consumers. Mobile marketing is even more affected by privacy concerns of consumers. Being interrupted by marketing efforts while you are on the move is considered to be an annoying intrusion into the personal life of the consumer. Furthermore, consumers are especially concerned when it comes to the tracking of their physical location and purchasing context (Milne, Rohm 2003). Research has shown that these privacy concerns become even more salient in public spaces (Peltier, Milne et al. 2009). There are many examples of times when firms overlook or underestimate the privacy concerns of its

(11)

11 recording the conversations their children had with the doll they were very upset (Beke 2018, Gibbs 2015). Examples like these prove that many companies are still unaware of the impact of the privacy concerns of their customers. Consumers have to give their consent for an app to be able to use tracking systems such as location tracking. Rising privacy concerns would suggest that people would be more reluctant to give permission to companies to monitor their behaviour. However, some research suggest that there is only a weak relationship between information disclosure intention and actual disclosure behaviour (Norberg, Horne et al. 2007, Keith, Thompson et al. 2013).

Another part of digital privacy is the security of consumers’ information. Consumers are becoming more conscious about the security of their personal data due to several public data breaches. The storage and security of data is important to guarantee the privacy of your customers. Also, the way a firm collects data, overtly vs covertly, influences the extent to which people feel like their privacy is violated. Research suggest that the effectiveness of personalization is moderated by whether firms collect consumer data overtly versus covertly (Aguirre, Mahr et al. 2015). According to this research overtly collected data results in higher adoption rate by consumers while personalization based on covertly collected data results in lower adoption rates. In our research design the respondent will be aware of the type of data collected. Other research found that people who desire more information transparency are also less willing to be profiled online (Awad, Krishnan 2006). A potential solution to consumer’s unwillingness to share their data would be for a retailer to focus on building trust with the consumer.

The privacy paradox and the privacy calculus

The privacy calculus is a framework that is considered to be most useful when it comes to studying the acceptance of information collection (Culnan, Bies 2003). It is defined as consumers’ internal

trade-off of the negative and positive consequences of the collection, storage, and use of personal information (Laufer, Wolfe 1977, Dinev, Hart 2006).The privacy calculus suggests that consumers

(12)

12 assessing the benefits and the costs of providing information. In reality it is often found that there is a larger gap between the information disclosure intentions of a consumer and the actual disclosure behaviour. This discrepancy between intentions and behaviours has been coined the privacy paradox (Norberg, Horne et al. 2007). Consumers face bounded rationality when it comes to information disclosure, biases and heuristics play an especially large role in low involvement situations (Petty, Cacioppo 1986). Furthermore, consumers may be unaware of the data being collected or unable to control the firm’s privacy practices (Brandimarte, Acquisti et al. 2013). Another problem of the privacy calculus lies in the fact that the potential benefits and risks are not always easy to measure or immediately clear, definitely when a choice has to be made in a short time frame. A person’s privacy preference might differ per situation depending on the medium, the type of technology used and the firm involved. It is therefore important to realize that someone’s acceptance of information disclosure is highly dependent on the context.

Hypothesis 1: The extent to which a person’s privacy is intruded negatively affects the adoption intention

Personalization

Personalization defined as “the adaptation of products and services by the producer for the

consumer using information that has been inferred from the consumer‘s behaviour or transactions” (Montgomery, Smith 2009). Personalization differs from customization in the sense that

personalization is automated by the marketer on behalf of the customer. Personalization is increasingly used by both off- and online retailers. The rise of information technology has allowed firms to gather vast amounts of consumer data which can be processed by ever more sophisticated data analyses techniques and software. Online retailers partner up with search engines in order to gather even more data on their consumer’s behaviour. By processing this data firms are able to personalize their services and marketing efforts. Personalized services offer many advantages to both consumers and firms. Consumers can enjoy products and services that better match their preferences (Vesanen 2007). It also reduces the chance of cognitive overload of consumers and increases convenience by offering relevant products and services (Ansari, Mela 2003). Furthermore, personalization can even grant firms a competitive advantage over their competitors (Murthi, Sarkar 2003). Also, personalization allows to charge higher prices (Vesanen 2007) and increase their

profitability (Zhang, Wedel 2009).

However, although there seem to be many great advantages for firms implementing a personalization approach there has also been some research suggesting that an increasing

(13)

13 consumer (Aguirre, Mahr et al. 2015, Awad, Krishnan 2006). This is also known as the

personalization paradox. This paradox suggests that a trade-off occurs between, on the one hand personalized services and on the other privacy and security concerns. Research into the

effectiveness of personalized vs non personalized advertising suggest that people in general prefer the personalized messages (Howard, Kerin 2004). However some, research also suggests that

personalization could lead to negative consumer responses when communication is personalized but the perceived service is low (White, Zahay et al. 2008). For example, when firms use a personalized salutation in an email but the email is considered to be spam, the consumer could hold negative feelings towards the personalized the firm. Furthermore, personalization might also reduce the social or brand identity of a product or service (Montgomery, Smith 2009).

Types of personalization

There are four types of personalization that can be distinguished 1) personalization of the product or service 2) personalization of price 3) personalization of promotion 4) personalization based on location. A personalization of services occurs, for example, when a website remembers your contact details from a previous check-out. Recommendation engines are another example of personalized services that benefit both firms and consumers. Personalized services have been shown to increase customer loyalty, perceived quality and customer satisfaction (Coelho, Jörg Henseler 2012).

However, consumers are not always aware of the type of data that is being used for these kind of services or how it is collected. When consumers are confronted with the type of data being used they will value the service less (Mothersbaugh, Foxx et al. 2012). Further research also suggest that consumer acceptance of personalized services depends on the type of information that is required (Xie, Knijnenburg et al. 2014). Personalized services have been found to have a positive effect on order sizes and order frequency (Vesanen 2007). Furthermore, service personalization has been found to increase customer satisfaction (Ansari, Mela 2003).

Hypothesis 2: the extent of service personalization will have a positive effect on app adoption intention

(14)

high-14 frequency purchases such as grocery shopping personalized discounts may work since they offer a relatively low absolute monetary benefit. Moreover, the personalized benefits will be a reflection of previous purchases that is often offered as an opt-in service. Furthermore, when consumers are aware of the type of data collected for the purpose of providing personalized discounts they can make a cost benefit analysis in order to figure out if they enjoy the benefits over the costs. Research suggest that services based on overtly collected data are perceived better than services who covertly collect data (Aguirre, Mahr et al. 2015). If consumers consider the costs of providing personal information high than the benefits of receiving discounts they will not be bothered if others pay less than they do.

Personalized promotions and marketing communications have become standard practice. Consumers consider personalized advertisements more relevant and more useful, making the advertisements more effective. On the other hand, consumers are not very positive about the idea of behavioural targeting. Marketing communications that are too personalized are considered to be too intrusive and will trigger privacy concerns (Van Doorn, Hoekstra 2013). Firms can counteract the negative consequences of personalized advertisements that are considered intrusive by providing more transparency (Aguirre, Mahr et al. 2015).

Hypothesis 3: Personalized price promotions will have a positive effect on app adoption intention

Recent developments in mobile communication technology have made it possible to use so-called location aware marketing (LAM) and location based services (LBS). Many apps use GPS tracking to improve their services. For example, apps like Uber use location tracking to make sure that the driver knows where to pick you up and to calculate your fee. Other apps, like weather apps can give you a notification when it is going to rain, based on your location. GPS enabled smartphones also allow firms to deliver personalized services and marketing communications based on the consumer’s geographical location. For example, customers might get a certain offer on their smartphone when they are inside or close to a firm’s store or that of a competitor. LAM allows firm to deliver

(15)

15 represent a higher level of intrusion and location based data are considered more personal.

According to the privacy calculus model the benefits have to exceed the risks in order for the application to be adopted (Norberg, Horne et al. 2007). This paper hypothesizes that location based services will not have a positive effect on app adoption intentions because location based services will be considered to be too intrusive.

Hypothesis 4: Personalized location based services will have a negative effect on app adoption intention

General privacy concerns

Consumers differ in the extent that they hold privacy concerns, some have more concerns than others for a variety of reasons. First of all, privacy concerns differ per generation, older people tend to hold more privacy concerns than younger people (Goldfarb, Tucker 2013). Furthermore, there seems to be more privacy concerns among females (Goldfarb, Tucker 2013) and lower educated people (Milne, Boza 1999). Also, consumers who are experienced with the technology used have been linked to both higher (Sheehan, Hoy 2000) as well as lower privacy concerns (Bellman, Johnson et al. 2004).

General privacy concerns have been linked to a lower preference of personalization benefits in an e-commerce context (Chellappa, Sin 2005). Furthermore, research suggest that the people who value information transparency are also the ones who are less willing to accept online personalization (Awad, Krishnan 2006). This problem makes it harder for firms to target customers that really value their privacy with personalization benefits.

Hypothesis 5a: The extent to which a person holds privacy concerns moderates the effect of price personalization on adoption intention

Hypothesis 5b: The extent to which a person holds privacy concerns moderates the effect of service personalization on adoption intention

(16)

16

(17)

17

Methodology

Three different types of personalization will be tested in this study. Personalization of price, personalization of service and personalization based on location. The respective weights that this study finds can be interpreted as the relative importance of the different types of personalization. The degree of privacy intrusion is measured by the weights on the requested access to data.

The choice based conjoint has three attributes with two levels (Price personalization, service personalization and location based personalization) and one attribute with three levels (privacy intrusion). Price personalization, service personalization, location based personalization and privacy intrusion are the independent variables. Adoption intention is the dependent variable.

Privacy intrusion is measured by three levels. Each level is contains a different degree of intrusion into the person’s personal information. The first level is that were no information is provided. In the second level people are required to give basic information like contact details. This type of information disclosure is commonly required for many services today. The last level is where people have to give their contact details as well as allow their GPS location to be tracked. Location tracking is considered to be a considerable intrusion into people’s privacy and the benefits have to significantly outweigh the costs of the intrusion for consumers’ for app to be adopted (Xu, Teo et al. 2009, Junglas, Johnson et al. 2008).

- Price personalization

o No price personalization

o Personal discounts based on previous shopping experience - Service personalization

o No service personalization

o Personalized grocery lists based on previous shopping experience - Location based personalization

o No location based personalization

o The app will send you personalized marketing messages based on your location (GPS) - Privacy intrusion

o The app does not gain access to any personal information o The app has access to your contact details

o The app has access to your contact details and GPS location - Trust (Survey) (Morgan, Hunt 1994, Beke 2018)

(18)

18 o In general you can rely on grocery retailers

o I trust that my personal information won’t end up in the hands of third parties o I trust that my personal information will not be used for identity fraud

o I trust that my personal information is stored safely - General privacy concerns (Dinev, Hart 2005)

o I am concerned that information I provide to mobile applications could be misused o I am concerned about submitting information on mobile applications, because of

what others might do with it

o I am concerned about submitting information to mobile applications , because it could be used in a way I did not foresee

Conjoint choice design

The conjoint study design will be based on a random selection of stimuli. Three alternatives are shown per choice set and a total of 18 choice sets are shown to the respondents. A number 18 choice sets was used in order to gain as much data as possible. A smaller number of choice sets was considered but due to sampling restrictions it was set as 18 in order to ensure sufficient information on variables. The sets were ordered to ensure that there is minimal overlap between sets. An alternative no-choice option was added in order to make it more realistic.

Sampling

The ideal sample for this research would be a completely random sample. Unfortunately due to the scale of this research it is impossible to do a completely random sample. Convenience sampling was therefore used. Most of the respondents will be close to the personal network of the researcher. This means that a large part of the sample population is aged between 20 and 30. The survey was distributed digitally in order to ensure maximum reach.

Model type

(19)

19 variables consists of a systematic component and an error term. The utility function can be written as Uci = Vci + eci. The error term contains all the effects that are not accounted for, for example: respondent fatigue or omitted variable bias(Louviere, Woodworth 1983). All utilities will be based on a part worth model, except for privacy intrusion, which will be checked for linearity. A balanced and orthogonal design is used. A fractional factorial design is used. All analyses will be done in R.

Attributes of a choice based conjoint study has to meet several requirements(Orme 2002, Green, Srinivasan 1990):

- Attributes should be relevant. (they should influence the choices of the consumer) - Attributes should be discriminant. (consumers must be able to discriminate between the

different choices)

- Attributes should be manageable (too many attributes makes the study too complex requires high cognitive strain from the respondent)

- Attributes should not be interrelated (they should measure different aspects of the product) Furthermore, the levels of a conjoint study also have to meet certain requirements(Orme 2002, Green, Srinivasan 1990):

- Levels have to account for situations that are larger than in reality to account for possible future scenarios

- Levels should be defined unambiguously

- The number of levels should be kept low (no more than three or four)

- The number of levels should be as balanced as possible (same number of levels across attributes). Otherwise the number of levels effect can occur (more levels would be considered more important by consumers)

- Levels should be generally acceptable - Levels are assumed to mutually exclusive

Beta’s are estimated by a maximum likelihood estimation. The maximum likelihood estimation aims to find a set of part worth utilities that best represent the observed choices using the following formula:

𝐿𝐿 = ln(𝐿) = ∑ ∑ 𝑙𝑛 𝑘 𝑛

(𝑝𝑟𝑜𝑏(𝑖𝑛𝑐|𝐽𝑛𝑐))

(20)

20 personalization, location based personalization and privacy intrusion levels. In the second part of the analysis, moderation effects will be tested by adding interaction effects to the first model. A similar procedure will be used when testing for the control variables.

Variable

Selection Dummy Binary

Price personalization Categorical (effect coded)

Service personalization Categorical (effect coded) Location based personalization Categorical (effect coded)

Privacy intrusion Categorical (effect coded)

Trust (App) Ordinal (likert 1-5)

Trust (Grocery Retailers) Ordinal (likert 1-5)

Privacy concerns Ordinal (likert 1-5)

Familiarity with app Ordinal (likert 1-5)

Downloaded the app Binary (dummy coded)

Education Ordinal (likert scale 1-5)

Gender Binary (dummy coded)

A multinomial logistic regression will be used to estimate the models. The multinomial logistic regression assumes that a linear combination of observed features together with respondent specific variables can be used to determine the outcome of the dependent variable. There are several assumptions to a multinomial logistic regression:

- Data is case specific (the independent variables have a single value for each case)

- The dependent variable can never be perfectly predicted by the independent variable for any specific case

- Multicollinearity is assumed to be relatively low - Independence of irrelative alternatives (IIA)

(21)

21

Model validation

In order to access the goodness of fit, a likelihood ration test will be performed and the pseudo R2 will be calculated. The estimated model will be tested against a NULL model, the H0 hypothesis states that there is no difference between the two models. A chi-square test statistic will be used to test this hypothesis, using the following formula:

𝐶ℎ𝑖𝑠𝑞 = −2(𝐿𝐿(0) − 𝐿𝐿𝛽∗

(22)

22

Data analysis

The data analysis and interpretation consists of two parts. Part 1 will be an estimation of the

conjoint data, the estimates will be use to analyse hypothesis 1 to 4. The second part of the analysis and interpretation will be done by combining the conjoint data with the control and moderation variables.

Data cleaning

There were a total of 102 respondents to the survey. The first three respondents were removed from the survey since they participated in order to test the survey. The data was checked for abnormalities. The elapsed time in second was checked to see if there were respondents that did the survey in an unrealistically small timeframe. Two respondents were found to have a completion time of under two minutes, however one of these respondents of also participated in the test and might therefore be faster due to familiarity with the survey. Respondent 63, 77 and 87 were removed because they only chose the no-choice option. The first three respondents, who were used in the testing of the survey were also removed. In the end the model was estimated with (n= 96) respondents.

Sample characteristics

The sample contained 45 men and 51 women. Furthermore, 44 people already downloaded a grocery retail app before and 52 have not. Most respondents in the sample have completed a university bachelor degree (43) or master’s degree (21) and only 2 respondents have a PHD. Furthermore, 13 people had only a high school degree, 2 had an MBO degree and 13 had a HBO degree. The youngest respondent is 18 years old and the oldest respondent is 58. The average age of the sample is 24.68.

Part 1: Model estimation

(23)

23 The model estimation yielded the following results:

Beta Standard error p-value

Price personalization 0.6774 0.037 < 2.2e-16 ***

Service personalization 0.0812 0.033 0.01659 * Location based personalization -0.2031 0.034 2.574e-09 *** No access to information 0.9897 0.044 < 2.2e-16 *** Access to contact details -0.2676 0.050 9.158e-08 *** Access to GPS + contact details (recovered) -0.7221 0.055 1.996125e-38*** None_option 0.2658 0.063 2.628e-05 *** Log-Likelihood -1908 Pseudo R2 _(McFadden) _0.2035

Table 1: Estimation results

Furthermore, based on these results the importance of each attribute was calculated.

Attribute Importance

Price personalization 0.3052

Service personalization 0.0366

Location based personalization 0.0915

Information disclosure 0.5665

(24)

24

Figure 2: importance of variables

Confidence intervals

2.50% 97.50%

Price personalization 0.604756 0.75013

Service personalization 0.014775 0.147674

Location based personalization -0.26999 -0.1363

No access 0.903283 1.076128

Contact details -0.3658 -0.16944

None_option 0.14191 0.389855

Table 3: confidence intervals

Based on these results we can reject the h0 for the following hypotheses:

Hypothesis 2:

The extent of service personalization will have a positive effect on app adoption intention.

From the results we can conclude that there is a positive estimate for service personalization. However the effect of service personalization is rather small (B = 0.0821) and the significance level is lower than those of the other variables. Service personalization is still only moderately significant at a (P = 0.01659), therefore we need to be careful to confirm this hypothesis. Furthermore, the results show that service personalization is the least important variables (3.66%) .Further research will be necessary in order to fully confirm this hypothesis. The results indicate that the benefits provided by personalizing the services are not considered to be very valuable by the consumer.

30%

4% 9% 57%

Importance

(25)

25

Hypothesis 3:

Personalized price promotions will have a positive effect on app adoption intention.

The results show a very clear effect of price promotion on adoption intention (B = 0.6774).

Furthermore, this is also a highly significant result (p = < 2.2e-16). Furthermore, personalized prices seem to be one of the most important variables to the respondents (30.52%). Hypothesis 2 can therefore be confirmed.

As expected, price personalization showed a significant effect on adoption intention. It has to be noted that price personalization is measured here as a discount and not as 100% personalized prices, this would also be highly unrealistic. Nevertheless, the significant effect of price personalization offer insight into the extent to which people are able to trade their personal information for discounts.

Hypothesis 4:

Personalized location based services will have a negative effect on app adoption intention.

Location based personalization shows a clear significant (P = 2.574e-09) negative effect ( B= -0.2031) on adoption intention. Location based personalization has an importance of 9.15%. Hypothesis 3 can be confirmed.

The results indicate that location based services are considered to be too intrusive, even when GPS location is not required. Research suggests that there are many advantages to location based personalization for firms, but it is important to consider that there has to be significant advantage for consumers as well.

Hypothesis 1:

Hypothesis 1: The extent to which a person’s privacy is intruded negatively affects the adoption intention

Privacy concerns are partly measured by the amount information someone is willing to disclose. Information disclosure is also found to be the most important predictor of adoption intention (56.65%). The respondents had a very clear preference (B = 0.9897) for the no information option. The contact details option was also found to have a significant negative effect (B = -0.2676).

(26)

26

Part 2: Control and moderation variables

The moderation variables that will be used are Trust and Privacy and security concerns. First a reliability analysis was performed to check whether the questions in the survey accurately measured the constructs they were supposed to measure. If the Cronbach’s Alpha was higher than 0.7, the variables could be merged together into one (Bland, Altman 1997). These merged variables were later used in the multinomial logit models.

Reliability analysis

Trust

Trust is measured in two ways. General trust towards grocery retailers and specific trust towards grocery retail apps. Both constructs are measured by a set of three questions. A reliability and correlation analysis was done in order to access whether these questions can be combined into one variable. The correlation matrix below give an indication of the degree to which the different questions correlate with each other. The variables GrT1, GrT2 and GrT3 measure the general trust in grocery retailers, while the variables AppT1, AppT2 and AppT3 measure the specific app trust. From the correlation matrix we can see that the different constructs clearly show some correlation.

Figure 3: correlation matrix trust

(27)

27

Privacy concerns

Privacy concerns was measured by three questions. A correlation and reliability analysis was performed in order to check if the three questions can be combined into one construct. The correlation matrix show high correlations (>0.7) for the three questions.

Priv1 Priv2 Priv3

Priv1 _{1 0.794763 0.70917} Priv2 _0.794763 _{1 0.78988} Priv3 _0.70917 _0.78988 ₁

Table 4: correlation matrix privacy concerns

The Cronbach’s alpha for the three constructs is 0.9063, which is higher than the threshold of 0.8. The three questions can therefore be combined to form one variable called PrivCons.

Familiarity with App

The same analyses were performed for familiarity with apps. This was measured by four questions, but question one about whether people actually downloaded the app was left out because it was used as a separate variable in the analysis . Again, a correlation matrix was made, which showed high correlations among variables.

_Fam1 _Fam2 _Fam3

Fam1 _{1 0.469484 0.556138} Fam2 _0.469484 _{1 0.814893} Fam3 _{0.556138 0.814893} ₁ Table 5: correlation matrix familiarity with app

The Cronbach’s alpha for the three variables was 0.8246, the variables were combined into one construct called familiar.

Summary table

Variable Cronbach’s Alpha Mean Standard

Deviation VIF scores GenTrust 0.8137 3.27 0.8487 1.538182 AppTrust 0.8629 2.98 1.1099 1.796112 PrivCon 0.9063 3.76 0.9941 1.612510 Familiar 0.8246 2.48 1.2690 1.284016

(28)

28

Estimating control variable effects

The effect of several control variables were checked to see if the effects of the IV’s were different for different levels of the control variables. The interaction effects found are listed in the table below.

Variable Significant Interaction effects

P value Beta

Gender No significant

interaction effects Education Price personalization

No access No choice option 1.593e-08 *** 3.169e-08 *** 0.04881 * 0.1491 0.1762 0.0913 Downloaded the app

Y/N price personalization LB personalization none option 0.0008163 *** 0.0048023 ** 669e-05 *** -0.2551 -0.1948 0.5554 Familiar with app price personalization

LB personalization no choice 8.152e-09 *** 8.569e-05 *** 0.023840 * 0.1797 0.1080 -0.1175 Trust in grocery retailers price personalization

no access to personal information

access to contact details no choice 8.500e-09 *** 2.914e-09 *** 0.0034676 ** 6.635e-13 *** 0.1377 -0.350232 0.212621 -0.529489 Trust in app service personalization

price personalization LB personalization no access to personal information

access to contact details no choice 2.146e-06 *** 6.655e-13 *** 0.0389767 * 1.055e-13 *** 0.0001380 *** 2.2e-16 *** 0.1546 0.2569 0.0684 -0.3227 0.1940 -0.5294

Table 7: Control variable summary

Significant interaction effects were found for several variables. Trust in grocery retailers and trust in grocery retailer apps was found to have an effect on most of the independent variables. For

example, respondents with more trust in both apps and retailers had a higher preference for price personalization and a lower preference towards the no-choice option. Similar results were found for those who were familiar with the app and those who had downloaded the app. Education seem to have some influence on the preference for lower privacy intrusion and price personalization. There were no significant differences in preferences established between genders.

Estimating moderation effects

Privacy intrusion

(29)

29 location based personalization (B = -0.1018, P = 0.00417 **). As expected the respondents with higher privacy concerns showed a higher preference for disclosing no information (B = 0.2592, P = 1.896e-08 ***) and a lower preference for disclosing their contact details (B = -0.1377, P = 0.00770 **). Again the no choice option was preferred more for those who had higher privacy concerns (B = 0.4911, P = 7.726e-12 ***). These results provide evidence for both hypothesis 4, 5a, 5b and 5c . The effect of personalization is being moderated by the extent to which a consumer hold privacy

(30)

30

Conclusion and discussion

At the beginning of this paper, the need for a better understanding of the trade-off between the positive and negative consequences of information disclosure and personalization was addressed. Managers should have better knowledge on the consequences of their privacy strategy and, in particular, the consequences of the type of information customers have to disclose. This paper used the context of grocery retail apps. A conjoint study was done in order to find out the degree to which different types of personalization are perceived to be of value to consumers and the extent to which privacy intrusion is tolerated for these benefits. Furthermore, general privacy concerns were introduced as a moderation variable. The result of the choice based conjoint analysis show that the most important factor when choosing a grocery retail app is the level of privacy intrusion. The second most important factor was price personalization. This suggests that the main trade-off consumers make is that between information disclosure and price personalization. Only minor positive effects were found for service personalization. Location based personalization was found to have a negative utility, regardless of the need to disclose GPS data. Furthermore, those consumers that hold higher concerns for their privacy are not only less likely to disclose information but also value personalization benefits less.

This thesis addressed the gaps found in previous research regarding the dynamics of the privacy calculus in a specific context (Dinev, Hart 2006, Beke 2018) . This research contributes to the existing literature concerning the privacy calculus three ways. First this paper makes the distinction between different types of personalization and the subsequent privacy trade-offs consumers make with regards to these different benefits. Furthermore, the conjoint study allowed for an evaluation of the relative importance’s of different personalization benefits and the level of privacy intrusion. Lastly, this study showed how consumers trade off the negative and positive consequences of information disclosure in the specific context of grocery retail apps. Managers should think carefully about the choices regarding the type of information they request from their customers, especially when it comes to the use of location based services. Furthermore, they should be aware of the trade-offs consumers make and know exactly which benefits they have to offer before they request any information. In general, consumers are becoming more aware of the potential negative consequences of information disclosure and governments are implementing new legislation in order to protect the privacy of their citizens. This development further increases the importance of a well-considered privacy strategy. This research could help managers develop a good privacy strategy by providing insights on the trade-offs consumers make.

(31)

(32)

32

References

AGREBI, S. and JALLAIS, J., 2015. Explain the intention to use smartphones for mobile shopping. AGUIRRE, E., MAHR, D., GREWAL, D., DE RUYTER, K. and WETZELS, M., 2015. Unraveling the

personalization paradox: The effect of information collection and trust-building strategies on online advertisement effectiveness. Journal of Retailing, 91(1), pp. 34-49.

ANSARI, A. and MELA, C.F., 2003. E-Customization. Journal of Marketing Research, 40(2), pp. 131-145.

ASHWORTH, L. and FREE, C., 2006. Marketing dataveillance and digital privacy: Using theories of justice to understand consumers’ online privacy concerns. Journal of Business Ethics, 67(2), pp. 107-123.

AWAD, N.F. and KRISHNAN, M.S., 2006. The personalization privacy paradox: an empirical evaluation of information transparency and the willingness to be profiled online for personalization. MIS

quarterly, , pp. 13-28.

BARGH, J.A. and CHARTRAND, T.L., 1999. The unbearable automaticity of being. American

psychologist, 54(7), pp. 462.

BART LARIVIÈRE, JOOSTEN, H., MALTHOUSE, E.C., MARCEL, V.B., AKSOY, P., KUNZ, W.H. and MING‐ HUI HUANG, 2013. Value fusion: The blending of consumer and firm value in the distinct context of mobile technologies and social media. Journal of Service Management, 24(3), pp. 268-293.

BART, Y., SHANKAR, V., SULTAN, F. and URBAN, G.L., 2005. Are the Drivers and Role of Online Trust the Same for All Web Sites and Consumers? A Large-Scale Exploratory Empirical Study. Journal of

Marketing, 69(4), pp. 133-152.

BEKE, F.T., 2018. Consumer privacy: understanding the acceptance of consumer information

collection , University of Groningen, Som Research School.

BELLMAN, S., JOHNSON, E.J., KOBRIN, S.J. and LOHSE, G.L., 2004. International differences in information privacy concerns: A global survey of consumers. The Information Society, 20(5), pp. 313-324.

BEST, S.J., KRUEGER, B.S. and LADEWIG, J., 2006. Privacy in the information age. International

Journal of Public Opinion Quarterly, 70(3), pp. 375-401.

BLAND, J.M. and ALTMAN, D.G., 1997. Cronbach's alpha. BMJ (Clinical research ed.), 314(7080), pp. 572.

BRANDIMARTE, L., ACQUISTI, A. and LOEWENSTEIN, G., 2013. Misplaced confidences: Privacy and the control paradox. Social Psychological and Personality Science, 4(3), pp. 340-347.

BRINKER, M., LOBAUGH, K. and PAUL, A., 2012. The Dawn of Mobile Influence – Discovering the

(33)

33 CHELLAPPA, R.K. and SIN, R.G., 2005. Personalization versus privacy: An empirical examination of the online consumer’s dilemma. Information technology and management, 6(2-3), pp. 181-202.

COELHO, P.S. and JÖRG HENSELER, 2012. Creating customer loyalty through service customization.

European Journal of Marketing, 46(3), pp. 331-356.

CULNAN, M.J. and BIES, R.J., 2003. Consumer Privacy: Balancing Economic and Justice Considerations. Journal of Social Issues, 59(2), pp. 323-342.

DAVIS, F.D., 1989. Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology. MIS Quarterly, 13(3), pp. 319-340.

DINEV, T. and HART, P., 2006. An extended privacy calculus model for e-commerce transactions.

Information systems research, 17(1), pp. 61-80.

DINEV, T. and HART, P., 2005. Internet Privacy Concerns and Social Awareness as Determinants of Intention to Transact. International Journal of Electronic Commerce, 10(2), pp. 7-29.

EGGERS, F. and SATTLER, H., 2011. Preference measurement with conjoint analysis. Overview of state-of-the-art approaches and recent developments. GfK Marketing Intelligence Review, 3(1), pp. 36-47.

EMARKETER, 2014-last update, 2 Billion Consumers Worldwide to Get Smart(phones) by 2016

Over half of mobile phone users globally will have smartphones in 2018. Available:

https://www.emarketer.com/Article/2-Billion-Consumers-Worldwide-Smartphones-by-2016/1011694.

EUROPEAN COMMISSION, 2017-last update, Proposal for a Regulation on Privacy and Electronic

Communications. Available: https://ec.europa.eu/digital-single-market/en/news/proposal-regulation-privacy-and-electronic-communications.

FEINBERG, F.M., KRISHNA, A. and ZHANG, Z.J., 2002. Do We Care What Others Get? A Behaviorist Approach to Targeted Promotions. Journal of Marketing Research, 39(3), pp. 277-291.

FISHBEIN, M. and AJZEN, I., 1975. Belief, attitude, intention and behavior: An introduction to theory

and research.

FORTES, N. and RITA, P., 2016. Privacy concerns and online purchasing behaviour: Towards an

integrated model.

GEFEN, D., KARAHANNA, E. and STRAUB, D.W., 2003. Trust and TAM in online shopping: An integrated model. MIS quarterly, 27(1), pp. 51-90.

GEFEN, D., 2002. Reflections on the dimensions of trust and trustworthiness among online

consumers.

GIBBS, S., 2015-last update, Privacy fears over 'smart' Barbie that can listen to your kids. Available:

(34)

34 GOLDFARB, A. and TUCKER, C., 2013. Why managing consumer privacy can be an opportunity. MIT

Sloan Management Review, 54(3), pp. 10.

GOOGLE SHOPPER MARKETING COUNCIL, 2013-last update, Mobile In-Store Research How In-store Shoppers are Using Mobile Devices. Available: https://www.thinkwithgoogle.com/advertising-channels/mobile/mobile-in-store/.

GREEN, P.E. and SRINIVASAN, V., 1990. Conjoint analysis in marketing: new developments with implications for research and practice. The journal of marketing, , pp. 3-19.

GUPTA, S., 2013. For Mobile Devices, Think Apps, Not Ads. Harvard business review, (March),. HEILMAN, C.M., NAKAMOTO, K. and RAO, A.G., 2002. Pleasant Surprises: Consumer Response to Unexpected In-Store Coupons. Journal of Marketing Research, 39(2), pp. 242-252.

HO, S.J., 2006. The Attraction of Internet Personalization to Web Users. Electronic Markets, 16(1),. HOLMES, A., BYRNE, A. and ROWLEY, J., 2013. Mobile shopping behaviour: insights into attitudes, shopping process involvement and location. Intl J of Retail & Distrib Mgt, 42(1), pp. 25-39. HOWARD, D.J. and KERIN, R.A., 2004. The Effects of Personalized Product Recommendations on

Advertisement Response Rates: The “Try This. It Works!” Technique.

JARVENPAA, S.L., TRACTINSKY, N. and VITALE, M., 2000. Consumer trust in an Internet store.

Information technology and management, 1(1-2), pp. 45-71.

JUNGLAS, I.A., JOHNSON, N.A. and SPITZMÜLLER, C., 2008. Personality traits and concern for privacy: an empirical study in the context of location-based services. European Journal of Information

Systems, 17(4), pp. 387-402.

KEITH, M.J., THOMPSON, S.C., HALE, J., LOWRY, P.B. and GREER, C., 2013. Information disclosure on mobile devices: Re-examining privacy calculus with actual user behavior. International journal of

human-computer studies, 71(12), pp. 1163-1173.

KIM, S.J., WANG, R.J. and MALTHOUSE, E.C., 2015. The Effects of Adopting and Using a Brand's Mobile Application on Customers' Subsequent Purchase Behavior. Journal of Interactive Marketing,

August, pp. 28.

KIM, C., MIRUSMONOV, M. and LEE, I., 2010. An empirical examination of factors influencing the intention to use mobile payment. Computers in Human Behavior, 26(3), pp. 310-322.

LAROSE, R. and RIFON, N.J., 2007. Promoting i‐safety: effects of privacy warnings and privacy seals on risk assessment and online privacy behavior. Journal of Consumer Affairs, 41(1), pp. 127-149. LAUFER, R.S. and WOLFE, M., 1977. Privacy as a concept and a social issue: A multidimensional developmental theory. Journal of Social Issues, 33(3), pp. 22-42.

(35)

35 MILNE, G.R. and BOZA, M., 1999. Trust and concern in consumers’ perceptions of marketing

information management practices. Journal of interactive Marketing, 13(1), pp. 5-24.

MILNE, G.R. and ROHM, A.J., 2003. The 411 on mobile privacy. Marketing Management, 12(4), pp. 40-40.

MONTGOMERY, A.L. and SMITH, M.D., 2009. Prospects for Personalization on the Internet. Journal

of Interactive Marketing, 23(2), pp. 130-137.

MORGAN, R.M. and HUNT, S.D., 1994. The commitment-trust theory of relationship marketing. The

journal of marketing, , pp. 20-38.

MOTHERSBAUGH, D.L., FOXX, W.K., BEATTY, S.E. and WANG, S., 2012. Disclosure Antecedents in an Online Service Context: The Role of Sensitivity of Information. Journal of Service Research, 15(1), pp. 76-98.

MURTHI, B.P.S. and SARKAR, S., 2003. The Role of the Management Sciences in Research on Personalization. Management Science, 49(10), pp. 1344-1362.

NORBERG, P.A., HORNE, D.R. and HORNE, D.A., 2007. The Privacy Paradox: Personal Information Disclosure Intentions versus Behaviors. Journal of Consumer Affairs, 41(1), pp. 100-126.

ORME, B., 2002. Formulating attributes and levels in conjoint analysis. Sawtooth Software research

paper, , pp. 1-4.

PELTIER, J.W., MILNE, G.R. and PHELPS, J.E., 2009. Information privacy research: framework for integrating multiple publics, information channels, and responses. Journal of Interactive Marketing,

23(2), pp. 191-205.

PETTY, R.E. and CACIOPPO, J.T., 1986. The elaboration likelihood model of persuasion.

Communication and persuasion. Springer, pp. 1-24.

PETTY, R.D., 2000. Marketing without consent: Consumer choice and costs, privacy, and public policy. Journal of Public Policy & Marketing, 19(1), pp. 42-53.

PROSSER, W.L., 1960. Privacy. California law review, 48(3), pp. 383-423.

RAMANATHAN, S. and DHAR, S., 2010. The Effect of Sales Promotions on the Size and Composition of

the Shopping Basket: Regulatory Compatibility from Framing and Temporal Restrictions.

ROUSSEAU, D., SITKIN, S., BURT, R. and CAMERER, C., 1998. Not So Different After All: A

Cross-discipline View of Trust.

SHANKAR, V., KLEIJNEN, M., RAMANATHAN, S., RIZLEY, R., HOLLAND, S. and MORRISSEY, S., 2016. Mobile shopper marketing: Key issues, current insights, and future research avenues. Journal of

Interactive Marketing, 34, pp. 37-48.

SHANKAR, V. and BALASUBRAMANIAN, S., 2009. Mobile Marketing: A Synthesis and Prognosis. SHANKAR, V., VENKATESH, A., HOFACKER, C. and NAIK, P., 2010. Mobile Marketing in the Retailing

(36)

36 SHEEHAN, K.B. and HOY, M.G., 2000. Dimensions of privacy concern among online consumers.

Journal of public policy & marketing, 19(1), pp. 62-73.

STATCOUNTER, 2018-last update, Desktop vs Mobile vs Tablet Market Share Worldwide. Available:

http://gs.statcounter.com/platform-market-share/desktop-mobile-tablet.

TAYLOR, S. and TODD, P.A., 1995. Understanding Information Technology Usage: A Test of Competing Models. Information Systems Research, 6(2), pp. 144-176.

URBAN, G.L. and HAUSER, J.R., 1980. Design and marketing of new products. Prentice hall. URBAN, J. and HOOFNAGLE, C., 2014. The privacy pragmatic as privacy vulnerable.

URBAN, G., AMYX, C. and LORENZON, A., 2009. Online Trust: State of the Art, New Frontiers, and

Research Potential.

VAN DOORN, J. and HOEKSTRA, J.C., 2013. Customization of online advertising: The role of intrusiveness. Marketing Letters, 24(4), pp. 339-351.

VENKATESH, V. and DAVIS, F.D., 2000. A theoretical extension of the technology acceptance model: Four longitudinal field studies. Management science, 46(2), pp. 186-204.

VESANEN, J., 2007. What is personalization? A conceptual framework. European Journal of

Marketing, 41(5), pp. 409-418.

WANG, R.J., MALTHOUSE, E.C. and KRISHNAMURTHI, L., 2015. On the Go: How Mobile Shopping

Affects Customer Purchase Behavior.

WESTIN, A.F., 2003. Social and political dimensions of privacy. Journal of Social Issues, 59(2), pp. 431-453.

WESTIN, A.F., 1967. Privacy and freedom. Atheneum.

WHITE, T.B., ZAHAY, D.L., THORBJØRNSEN, H. and SHAVITT, S., 2008. Getting too personal: Reactance to highly personalized email solicitations. Marketing Letters, 19(1), pp. 39-50. WINGFIELD, N., 2018-last update, Inside Amazon Go, a Store of the Future. Available:

https://www.nytimes.com/2018/01/21/technology/inside-amazon-go-a-store-of-the-future.html. WU, J. and WANG, S., 2005. What drives mobile commerce?: An empirical evaluation of the revised technology acceptance model. Information & management, 42(5), pp. 719-729.

XIE, J., KNIJNENBURG, B.P. and JIN, H., 2014. Location sharing privacy preference: analysis and personalized recommendation, In Proceedings of the 19th international conference on Intelligent

User Interfaces (IUI '14) 2014.

(37)

37 XU, H., TEO, H., TAN, B.C. and AGARWAL, R., 2009. The role of push-pull technology in privacy calculus: the case of location-based services. Journal of Management Information Systems, 26(3), pp. 135-174.

YANG, K., 2012. Consumer technology traits in determining mobile shopping adoption: An application

of the extended theory of planned behavior.

YANG, K., 2010. Determinants of US consumer mobile shopping services adoption: implications for designing mobile shopping services. Journal of Consumer Marketing, 27(3), pp. 262-270.

ZHANG, J. and WEDEL, M., 2009. The Effectiveness of Customized Promotions in Online and Offline Stores. Journal of Marketing Research, 46(2), pp. 190-206.

ZHAO, J., FANG, S. and JIN, P., 2018. Modeling and Quantifying User Acceptance of Personalized Business Modes Based on TAM, Trust and Attitude. Sustainabillity, 10(356), pp. 1-26.

(38)

38

Appendix

> summary(ml1)

Call:

mlogit(formula = Selection_Dummy ~ Price.personalization + Service.personalization +

Location.based.personalization + Information.disclosure +

None_option | 0, data = cbc, method = "nr", print.level = 0)

Frequencies of alternatives: 1 2 3 4 0.24342 0.27412 0.25219 0.23026 nr method 5 iterations, 0h:0m:0s g'(-H)^-1g = 5.04E-05

successive function values within tolerance limits

Coefficients :

Estimate Std. Error z-value Pr(>|z|)

Price.personalization -1.259666 0.069383 -18.1553 < 2.2e-16 *** Service.personalization -0.133986 0.064849 -2.0661 0.03882 * Location.based.personalization 0.369923 0.065081 5.6841 1.315e-08 *** Information.disclosure -0.908935 0.041636 -21.8304 < 2.2e-16 *** None_option -3.105392 0.192109 -16.1647 < 2.2e-16 *** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Log-Likelihood: -2060.6 > summary(ml2) Call:

mlogit(formula = Selection_Dummy ~ Price.personalization + Service.personalization + Location.based.personalization + Information.disclosure +

None_option | 0, data = cbc2, method = "nr", print.level = 0)

(39)

39 successive function values within tolerance limits

Coefficients :

Price.personalization -1.293756 0.070881 -18.2525 < 2.2e-16 *** Service.personalization -0.105710 0.065978 -1.6022 0.1091 Location.based.personalization 0.425195 0.066395 6.4040 1.514e-10 *** Information.disclosure -0.914652 0.042686 -21.4274 < 2.2e-16 *** None_option -3.027899 0.194875 -15.5376 < 2.2e-16 *** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Log-Likelihood: -1994.2 > summary(ml3) Call:

mlogit(formula = Selection_Dummy ~ Price.personalization.1..Yes. +

Service.personalization.1..Yes. + Location.based.personalization.1..Yes. +

Infdis1 + Infdis2 + None_option | 0, data = cbc3, method = "nr",

print.level = 0) Frequencies of alternatives: 1 2 3 4 0.24156 0.27648 0.25378 0.22817 nr method 5 iterations, 0h:0m:0s g'(-H)^-1g = 2.2E-05

Coefficients :

(40)

40 > summary(ml4)

Call:

mlogit(formula = Selection_Dummy ~ Price.personalization.1..Yes. + Service.personalization.1..Yes. + Location.based.personalization.1..Yes. +

Infdis1 + Infdis2 + I(X31_Gender * Service.personalization.1..Yes.) +

I(X31_Gender * Price.personalization.1..Yes.) + I(X31_Gender *

Location.based.personalization.1..Yes.) + I(X31_Gender *

Infdis1) + I(X31_Gender * Infdis2) + I(X31_Gender * None_option) +

None_option | 0, data = cbc3, method = "nr", print.level = 0)

Frequencies of alternatives: 1 2 3 4 0.24156 0.27648 0.25378 0.22817 nr method 5 iterations, 0h:0m:0s g'(-H)^-1g = 2.6E-05

Coefficients :

Price.personalization.1..Yes. 0.659284 0.053865 12.2395 < 2.2e-16 *** Service.personalization.1..Yes. 0.065289 0.049743 1.3125 0.1893414 Location.based.personalization.1..Yes. -0.146214 0.049451 -2.9567 0.0031092 ** Infdis1 0.974860 0.063718 15.2995 < 2.2e-16 *** Infdis2 -0.260150 0.073617 -3.5338 0.0004096 *** I(X31_Gender * Service.personalization.1..Yes.) 0.029344 0.068181 0.4304 0.6669129 I(X31_Gender * Price.personalization.1..Yes.) 0.035109 0.074396 0.4719 0.6369795 I(X31_Gender * Location.based.personalization.1..Yes.) -0.108087 0.068368 -1.5810 0.1138857 I(X31_Gender * Infdis1) 0.033055 0.088426 0.3738 0.7085435 I(X31_Gender * Infdis2) -0.016956 0.100548 -0.1686 0.8660855 I(X31_Gender * None_option) -0.045961 0.126708 -0.3627 0.7168075 None_option 0.291523 0.091713 3.1786 0.0014797 ** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Log-Likelihood: -1906.3 > summary(ml5) Call:

Grocery retail apps and the privacy calculus: the impact of different types of personalization on the adoption intention of consumers Master Thesis MSc Marketing Intelligence