Step into your own store : the impact of atmosphere personalization in ecommerce environments on service quality

(1)

Step into your own store

The impact of atmosphere personalization in

ecommerce environments on service quality

Marjolein Backus

Student number: 10574409 Date: 26 January 2018

MSc. in Business Administration – Marketing Track Amsterdam Business School, University of Amsterdam Supervisor: Drs. R.E.W. Pruppers

(2)

Statement of originality

This document is written by Marjolein Backus 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)

Abstract

In a time where integrated content personalization is prevailing, this research makes a first attempt in examining the implications of personalizing website atmospheres based on individual customer preferences. Using a 2x2x2 full factorial in-between subjects experimental design, 215 consumers were asked to participate in an online experiment. The findings suggest that there is no direct indication that atmosphere personalization leads to better service evaluations, yet atmosphere personalization does lead to an increase in the perception of effort a company makes for its customers. Additionally, content personalization leads to a higher perception in usefulness of the services, which is in line with previous empirical research. The insight that atmosphere personalization is different from content personalization in its consequences, suggests that personalization is not necessarily perceived as functional, but can also cause emotional responses. Further research is needed to more robustly indicate the implications.

(6)

Chapter 1: Introduction

1.1 The ecommerce environment

Over the past decade, online shopping has become almost indispensable in our lives. In 2016, roughly 42% of internet users in the United States made online purchases at least once a month, and 24% even once or twice per week (Statista, 2016). It seems as if a life without ecommerce is already unimaginable. Ecommerce can be defined as “the use of the internet to facilitate, execute and process business transactions which involve a buyer and a seller and the exchange of goods or services for money” (Delone & Mclean, 2004). In 2016, ecommerce sales had a value of $1,915 trillion worldwide, which equals to 8.7% of total retail sales, and this share is expected to increase in the upcoming years (eMarketer, 2016). The ecommerce industry is gradually growing and customers hold high expectations, which makes it hard for companies to stand out from the crowd. Therefore, companies are increasingly focussing on customer engagement, as interacting with customers on a personal level positively contributes to the overall customer experience (Sorokina, 2015). The customer experience is salient, since 64% of customers rank the experience higher than price when making brand choices and by 2020 this is expected to become the key differentiator over price and even over the product itself (Neosperience, 2014). Gartner predicted that by 2016, 89% of all retailers would plan to compete on customer experience, as opposed to 36% in 2010 (Sorofman, 2014). Hence, it is not unlikely that this prediction might have become a fact in the current market.

As a consequence, companies are hauled to the concept of customer experience design due to this shift in focus. This concept is described as the practice of designing products or services with the focus on experience, instead of merely the quality of the offerings (Longaneker, 2016). When we connect customer experience design to the online retail environment, companies can differentiate by developing a personalization strategy. Research refers to personalization in the online environment as websites tailoring products and purchase experiences to individual preferences based on personal information (Chellappa & Sin, 2005). Amazon, perhaps one of the most influential organizations today, incorporates data-driven personalization by analysing previous search and shopping behaviour of its customers (Thorp, 2015). Based on these analyses, recommendations are provided that are tailored to each individual visiting the website. Not only are customers receptive to personalization and

(7)

find it a useful aid that eases the decision making process (Tam & Ho, 2006), but personalized recommendations also have a positive impact on the online shopping experience, satisfaction and purchase intention (H. Chen, 2012).

1.2 Personalizing atmospheres

The purpose of this study is to identify how companies can bring personalization in online environments to a next level, in order to maximise the customer experience in a differentiating way. When investigating the offline retail environments in comparison to online contexts, it has been observed that there lies an opportunity for online contexts within the concept of atmospherics. Atmospherics is defined as “a tool that enables the conscious designing of space to create certain effects in buyers” (Kotler, 1973). When linking atmospherics to personalization, we immediately see that this would be simply impossible to realise in a brick-and-mortar store. One cannot change the store atmosphere according to every single individual’s preferences. However, this would actually be feasible in online contexts, as current technology allows us to personalize anything we want, based on personal preferences or previous behaviour. Therefore, this study introduces the concept of atmosphere personalization: where customers can shop in a truly personalized online store. This could be the way for companies to differentiate in an innovative manner. To illustrate, analyst Brian Solis states that the gap between web design and user preferences is where companies are invited for disruption and competition (Wallace, 2016).

It is relevant and noteworthy to indicate that up till now no scientific research has paid attention to the interplay of atmospherics and personalization, which independently are proven to be effective to enhance the customer experience and service quality. More interesting is perhaps the reason why consumers would be, or would not be affected by a personalized web atmosphere and to what extent the effects differ from existing content personalization efforts like recommendation sections. For example, previous research already pointed out that men and women prefer different web designs (Moss, Gunn & Heller, 2006), which shows that people evaluate the design of a website differently. Yet, little is known about the effect of atmosphere personalization, where every web visitor will shop within their own store design. Hence, the main research question of this thesis is formulated as follows:

(8)

RQ: How does atmosphere personalization on ecommerce websites affect consumers’ service evaluations?

As already indicated, the underlying reasons why consumers value content personalization and perhaps also atmosphere personalization are interesting to investigate. Maybe consumers like personalization because it has some sort of added value within the shopping process, or perhaps it is because they truly feel that the company makes an effort for them and value this effort. To investigate these explanations, a sub question within this thesis is formulated:

SQ: Presuming that atmosphere personalization has an impact on consumers’ service evaluations, does this occur due to the perceived usefulness of the service and/or due to the perceived company effort that is made?

1.3 Contributions

This thesis theoretically contributes to the already existing understanding that personalization is an effective strategic choice that companies can implement to increase the online customer experience (Tam & Ho, 2006; Chen, 2012). Similarly, this thesis also builds upon the knowledge that an appealing web design contributes to a better customer experience. However, where the above contributions have reproduction of previous findings as a purpose, this thesis also proposes a first attempt in closing the research gap that has been observed. This gap is one where personalization can be brought to a higher level by combining it with web atmospherics. Academic research does not show us yet whether this would be an additional and contemporary effective communication strategy that can lead to more positive evaluations of the service. Another interesting contribution is that this research attempts to point out whether or not people indeed care about a personalized atmosphere and especially the reason why. Is this because they like the phenomenon in a sense that they feel a company makes a personal effort as an additional service, or is it because they truly see added value in it? Both the combination of reproducing previous contributions with researching why people behave in a certain way due to marketing tactics is of great relevance for studying consumer behaviour. More specifically, this research contributes to the academic fields of ecommerce, customer experience design, personalization and marketing by expanding the current knowledge about online consumer behaviour.

(9)

Moreover, the findings of this study could have a substantial impact on the way marketers and web designers establish ecommerce websites. If indeed customers show different behaviour or more positive evaluations when they get the ability to shop within a personalized atmosphere that fits their preferences, companies should take this into account so that they can create the most effective website. This is valid for either positive and negative outcomes of this research. As customer experience is going to be the key differentiator by 2020 (Neosperience, 2014), companies should be ahead on creativity. In the end, a better customer experience is expected to cause more profitable returns and superior customer value. The Dutch online magazine NSMBL (www.nsmbl.nl) comes close to personalizing atmospheres. However, the layout of this website randomly changes each time you visit, probably due to rotations between a handful of different designs. Therefore, it is not in fact personalized, but merely randomized. Observing these kinds of practices in the online environment shows that organizations are trying to differentiate in an original and creative way using web atmospherics.

1.4 Delimitations of the study

As atmospherics is a broad phenomenon, this study limits itself solely to a focus on layout colours. This choice has partly been made due to time constraints, but also due to efficiency. As there is no previous data available on the combination of web design and personalization, research first needs to point out whether or not there is an actual impact. Furthermore, the experimental design used in this thesis makes it difficult to find deeper underlying explanations for the results that are found, as perhaps there are more unconscious effects that the participants of this experiment do not actively experience.

1.5 Structure

In the following chapters of this thesis, first an extensive review of the available literature on the topics will be provided in Chapter 2 and Chapter 3. Here, the concepts of personalization and atmospherics are further explained and how they influence the service evaluations. In Chapter 4, all expected effects are visually presented in a conceptual model, along with several hypotheses based on the literature. Then, Chapter 5 will give a detailed description of the methodology so that anyone has the means to reproduce this study. After that, the results of the analysis will be presented in Chapter 6, from which a conclusion for the hypotheses

(10)

can be derived. Finally, in the conclusion and discussion section in Chapter 7 several implications and limitations of this research will be provided together with some directions for future research.

(11)

Chapter 2: Personalization

2.1 Personalization in offline and online contexts

In order to understand the phenomenon on a more detailed level, it is crucial to first elaborate on the definition of personalization in an online and offline context and what types of personalization can be implemented by companies. Before the internet became a mainstream service, personalization was already a known marketing tool in the offline service industry. This type of personalization is characterized by employees that put effort into getting to know their returning customers, in order to enhance the company-customer relationship. Mittal and Lassar (1996) have defined personalization as “the social content of interaction between service or retail employees and their customers”. For this thesis, we are interested in how this concept has been translated to online markets.

Before investigating the definitions of personalization in an online retail context, it is noteworthy to mention that personalization is not only used in online retail, but is also observed on social media and more specifically on Facebook. Not only does Facebook implement personalization on the website itself by addressing members with their first name, but they also provide a service that enables customers to login with their Facebook profiles for using third party services (Park, 2014). To clarify, one can login with Facebook to use Spotify, for which Spotify receives some basic Facebook profile information in return. Having illustrated that personalization is not exclusively linked to retail, we need to look at how the concept is actually applied in online retail settings.

In such contexts, companies are able to personalize offerings, emails and any type of message by creating so-called customer profiles (Schubert & Koch, 2002). In these profiles, data is stored regarding previous interactions with the website, transaction history, indicated preferences, ratings and contextual information about time, date or place (Schubert & Koch, 2002). Another more recent study describes online personalization as “the process of individualized matching to consumer preferences through automated processes in the web environment” (Salonen & Karjaluoto, 2016). However, in the research field, disagreement about the terminology of personalization is prevailing. Some researchers address the topic by considering the term customization as similar to personalization (Parra & Brusilovsky, 2015). Others indicate that customization requires initiation of the customer, for example by actively customizing product labels for a Heineken beer bottle, while personalization is

(12)

company-initiated (Ho & Bodoff, 2014; Montgomery & Smith, 2009). In this thesis, the concept of personalization is considered as being company-initiated and hence differs from customization.

From the above sections, we can conclude that the concept shows some contrasting terminology in offline and online retail environments. Especially the social aspect of personalization in an offline environment is something to take into account when translating it to online atmospheres. From the definition by Mittal and Lassar (1996), we can conclude that some aspects of face-to-face interaction simply cannot be replaced by online technologies, such as courtesy, friendliness, helpfulness, care, commitment, flexibility and cleanliness (Cox & Dale, 2001). Additionally, it is less complicated to get emotionally engaged with individual customers with face-to-face interaction and therefore many types of online personalization strategies are in its principle rather functional. To clarify, online personalization in this sense is considered as being a useful aid for customers in their decision-making process (Tam & Ho, 2006). It effectively reduces the required effort that a customer needs to make in the decision-making process, which in turn leads to a higher customer satisfaction (Liang, Lai, & Ku, 2007). On average, people are more willing to scroll through personalized content when they are at an early stage of decision-making (Ho & Tam, 2005).

2.1.1 Types of personalization

There are multiple types of personalization that contribute to this purpose of usefulness, such as profile-based personalization, personal tools, opportunistic links and recommender systems (Wu, Im, Tremaine, Instone & Turoff, 2003). These technologies enable users to efficiently find the most relevant content on a website, but they also empower firms to increase the time spent on a website as well as advertising revenues due to re-marketing and behavioural targeting (Ho & Bodoff, 2014; Liu, Sheth, Weinsberg, Chandrashekar & Govindan, 2013). Moreover, personalization enables companies to generate more sales, increase efficiency and improve website effectiveness by collecting data (Ramnarayan, 2005). An example of an online personalization technique that is often used is the integration of recommender systems, which are characterized by the adaptation of user profiles to be able to vary the offering mix that is displayed on a website that matches a person’s preferences (Jannach, Zanker, Felfernig, & Friedrich, 2011). Chen (2012) shows that these

(13)

recommendations have a positive impact on the purchase intention of consumers. There is a large sum of studies available about the implications and applications of recommender systems, primarily discussed on a technical level in the field of IT (e.g. Ricci, Rokach & Shapira, 2011; Koren, Bell, Volinsky, 2009). Since recommender systems are prevailing in the online retail domain, this thesis will consider these systems as the overarching concept of content personalization. Content personalization will be one of the independent variables in this research, to be able to understand the possible implications of atmosphere personalization more thoroughly.

2.2 Threats of personalization

Before discussing the opportunities of personalization, we should first give special attention to a threat with regards to online personalization that cannot be ignored. In the current society, there seems to be an ongoing trend of worrying about our privacy, which is why personalization is strongly involved with these issues. Since personalization requires personal preference data which is stored by companies, people might feel that their privacy is jeopardized. For companies, it can be very lucrative to sell personal data to third parties nowadays (Riederer, Erramilli, Chaintreau, Krishnamurthy, & Rodriguez, 2011). Such third parties use personal information to create targeted advertisements, analyse consumer behaviour or to resell it again to any other company that needs a database. The perceptions and actions of consumers regarding these matters are contradicting. One the one hand, consumers are very worried about their data being sold to third parties, but on the other hand they often accept messages such as terms and conditions to make use of a service without carefully reading them through. Research describes this phenomenon as the privacy paradox, as the actions are contradicting to the thoughts of a consumer (Awad & Krishnan, 2006). Linking these insights to personalization, research has found that a high level of personalization in online advertisements leads to an increase in feelings of intrusiveness, which in turn leads to lower business performances (van Doorn & Hoekstra, 2013). However, companies can minimise this negative impact of privacy on performance, by gaining a sufficient amount of trust. When consumers feel that the company is trustworthy, it is more likely that they will intend to make use of personalized services (Chellappa & Sin, 2005; Awad & Krishnan, 2006). Ultimately, a company should operate in an honest manner by processing personal data without being intrusive.

(14)

2.3 Opportunities of personalization

The previous sections in this chapter have demonstrated that personalization can be very useful to consumers, yet it can sometimes also be perceived as intrusive regarding privacy matters. Furthermore, these sections address that the contemporary personalization strategies hold a more functional purpose for consumers than an emotional one. To clarify this, it has been found to reduce search efforts that accordingly eases the consumer decision-making process (Tam & Ho, 2006). However, emotional purposes are of great value to marketing, as they are related to the highest form of customer-based brand equity. The objective of customer-based brand equity is to invest in relationships with customers so that brand resonance can be achieved (Keller, 2001). Brand resonance is characterized as the intense involvement of customers with a brand, which is an indication of being loyal to the brand. Brand resonance consists of an emotional component and a behavioural component. The behavioural component is characterized by activity, whereas the emotional component by attachment. Building upon these insights from Keller (2001), we could say that online personalization in its current forms is lacking to appeal to the emotional component, as its benefits are mostly related to effort reduction and efficiency. Contrastingly, offline stores have the privilege to invest in emotional bonding more easily, as they are able to communicate via face-to-face interactions.

This opens a new discussion, as we need to find out in what cases personalization could also have an emotional function that is comparable to employees in a physical store providing caring and individualized service. This type of service provision is extensively discussed in the SERVQUAL model, developed by Parasuraman, Zeithaml and Berry (1988). Their model, which translates to the Service Quality Model, provides guidance for assessing customer perceptions of service quality and consists of five building blocks. These five building blocks are tangibles, reliability, assurance, responsiveness and empathy. As this thesis focusses on personalization and more specifically on the personalization of atmospheres, the most relevant building blocks to discuss are tangibles and empathy. To clarify, the tangibles are described as the physical facilities, equipment and appearance of personnel, which stresses the importance of atmosphere in a shopping environment. For example, Kotler (1973) also indicates that the store and service atmosphere can influence the customer experience amongst other factors. The other building block that is most relevant to this study, empathy, is described as the caring and individualized attention a firm provides to

(15)

its customers and is more focussing on the emotional side of providing service. The purpose of this component is to increase satisfaction that leads to better customer experiences. In the following section, both components are discussed into detail.

2.2.1 Empathy

When looking at the empathy component within the SERVQUAL model we can think about how online firms can leverage giving personal and caring attention to their web visitors. In an online context, such an interaction is harder to establish compared to an offline context (Rose, Hair & Clark, 2011). The research by Liang, Chen and Turban (2009) demonstrates the importance of empathy with regards to increasing the service quality, and introduces a theory that is called the perceived care theory. This theory suggests that customers will build a positive attitude towards a company whenever they feel that the company cares for them and pays attention to personal needs. To make an attempt in expressing this empathy component online, many companies nowadays use chat bots or personal names of employees on their webpages to make it seem more personal and human. The Dutch online electronic store Coolblue is a good example of these strategies. They have incorporated a little anthropomorphised robot figure called “Yamuda” on their customer service landing page that is happy to answer any question (www.coolblue.nl/klantenservice). Anthropomorphism, meaning that people see human in non-human forms, is widely used in marketing to draw attention and to get an increase in product evaluations (Aggarwal & McGill, 2007). Incorporating these human-like characters is an example of an attempt at changing the static perceptions about an online shopping environment, to perceptions that consider these environments similar to physical stores with employees and face-to-face interactions.

Additionally, personalization is also a tool that is used by companies to communicate to customers on an individual level in online environments. This can be clarified using the research of Kaynama and Black (2000), who demonstrate the parallelism of the SERVQUAL model and their proposed E-QUAL model. As the name already implies, the E-QUAL model is a derivative from the original SERVQUAL model in order to translate the instrument from offline to online service environments. In the E-QUAL model, the empathy component is described as the level of personalization and customization. These concepts enable ecommerce firms to provide the individualised and caring attention in an online setting that comes as close to the empathy definition in SERVQUAL.

(16)

One can argue whether personalization could be categorized under empathy, since this suggests that the purpose of integrating personalization is to build emotional bonds with customers. However, we know that personalization is often considered as being a useful aid in the shopping process, which suggests a more functional purpose (Tam & Ho, 2006). Based on these insights, the research by Liang, Chen and Turban (2009) is useful to illustrate that personalized services are evaluated by both functional and emotional factors. They suggest that personalization is used to improve the perception of customer care, rather than merely focusing on effort reduction in the shopping process. Hence, personalization has the ability fulfil both purposes.

2.2.2 Tangibles

As has been addressed before, this research suggests a new way of looking at personalization in online environments, which is why we can learn from the SERVQUAL model and more specifically from the tangibles component. The tangibles are described as all physical appearances in a shopping environment, varying from facilities to personnel. Since we know that store atmosphere positively influences the customer experience (Bonnin, 2006), we could translate this concept to the online context and this is where the great opportunity for the future of personalization enters: atmosphere personalization. In online ecommerce environments, these tangibles can only be expressed in aesthetic aspects of the website (Van Iwaarden, Van Der Wiele, Ball, & Millen, 2004). Even though online contexts are limited in their possibilities to leverage the tangibles, we know that an offline store is simply not able to personalize the store environment for every single customer. Since this is a great opportunity for online technology in order to outperform on the competition of physical stores, Chapter 3 further elaborates on how personalization and atmospherics can be a great duo and analyses the field of store and web atmospherics into detail.

(17)

Chapter 3: Atmospherics

3.1 Atmospherics in retail

In this chapter, we will follow upon the SERVQUAL model by Parasuraman, Zeithaml and Berry (1988) and more specifically discuss the tangibles. From the previous chapter, we have learned that there are opportunities for personalization to create a competitive advantage to differentiate from the offline service industry. However, since the tangibles cannot be physically integrated online, companies should be creative in finding other ways to do so. First, we should take a look at what has been investigated in the field of store design and store atmospherics. There are numerous studies that address the importance and effectiveness of a compelling store design or store atmospherics (e.g. Kotler, 1973; Bitner, 1990; Grewal et al., 2003). Kotler (1973) described the concept of atmospherics as “the conscious designing of space to create certain effects in buyers”. He defines an atmosphere by four dimensions that are described as visual, aural, olfactory and tactile. Overall, atmospherics is a relevant marketing tool in market spaces where there is a lot of competition (Kotler, 1973). By implementing this tool, firms are attempting to attract and retain customers. To demonstrate, Bitner (1990) shows that a physical setting that is aesthetically appealing has a positive impact on the customer satisfaction. Additionally, according to Grewal et al. (2003), the atmosphere in a store can interact with consumer perceptions to have an effect on behaviour and in the end, it also positively influences the purchase intentions. Both these highlighted studies demonstrate the importance of a successfully designed store atmosphere.

Dayal, Landesberg and Zeisser (2000) already indicated that the objective of online marketers should be to deliver complete and completely satisfying experiences in order to be competitive with offline stores. That is why we should investigate how an atmosphere can be personalized in online contexts. One way to integrate atmosphere personalization would be through web atmospherics, as this concept shows parallelism to store atmospherics. In order to draw conclusions about the execution of atmosphere personalization, we need to understand what defines web atmospherics. For example, one research defines web atmospherics as “the conscious designing of web environments to create positive effects (e.g. positive affect, positive cognitions) in users in order to increase favourable consumer responses (e.g. site revisiting, browsing)” (Dailey, 2004). A more recent study introduces a theory that is called the Unified Theory of Acceptance and Use of Technology (UTAUT) to

(18)

that suggests that the three components of technology, content and appearance are most important to customers when visiting a website (Al-Qeisi, Dennis, Alamanos, & Jayawardhena, 2014). Therefore, firms put a lot of effort in improving web design in order to enhance the customer experience (Vila & Kuster, 2011). To illustrate, the article by Novak, Hofmann and Yung (2000) investigates the effect of web design on the customer experience. The results suggest that a web design should arouse the customer, but not to such an extent that he or she becomes frustrated because of hard navigation through the website. On the other side, when the web design is not arousing at all, customers will quickly get bored and consequently switch to a different website. Menon and Kahn (2002) add to this that if the first few pages stimulate joy on a website, the more consumers are likely to explore the website further. However, companies need to keep in mind that when there is too much colour or too much information on a website, this has a negative consequence on further exploration intentions, as it causes an information overload (Puccinelli et al., 2009). Another supporting research has shown that design of a web layout has significant impact on the customers’ emotional arousal and attitude towards the website which in turn positively affects the intention to purchase (Wu, Lee, Fu, & Wang, 2013). Additionally, McDowell, Wilson and Kile (2016) show that certain features of web design explain the variance conversion in rates in a positive way, for example the flow of a website. Lastly, another supporting research suggests that visual appeal of a website among others positively influences the customer experience (Flavian, Gurrea & Orús, 2009).

The findings of these studies are all coherent with each other, from which we can conclude that a proper web atmosphere design has a positive effect on the customer experience. It is noteworthy to keep in mind that the judgement of a website being well-designed is formed through an interaction between the user and the designer (Moss, Gunn & Heller, 2006). In other words, the assessment is greatly determined by the perception of the customer and thus based on individual preferences.

3.2 How to personalize an atmosphere

After having discussed the importance of web atmospherics to create unique customer experiences, the next challenge is to elaborate on the way how personalization could be integrated into this phenomenon. As already has been indicated, both concepts are individually proven to be effective, which provides a foundation for why atmosphere

(19)

personalization could be effective as well. Previous research made an attempt in investigating whether men and women have different web design preferences, tested by using colours. This research found that men and women do seem to have different preferences and one of the results suggests that men mainly prefer designs with black and blue tones, whereas women have a strong preference for pink, mauve and yellow (Moss, Gunn & Heller, 2006). Interesting about this research is that it shows us that people have different preferences regarding web design colour use, which demonstrates the relevance of this thesis. Not only do men and women have different preferences, but they also make different purchase decisions based on product colours (Funk & Ndubisi, 2006). Women carefully evaluate colours in terms of attractiveness and their attitude towards the colour, whereas men have a greater eye for the product features. If we would personalize a website design not only tailored to gender specific preferences, but on an individual level, companies are able to listen to the wishes of their customers more accurately.

Building upon the definition of atmospherics by Kotler (1973), we know that the four dimensions are visual, aural, olfactory and tactile. When considering how atmosphere personalization can be realised in the online retail environment, the most appropriate factor to focus on would be visual. Visual aspects can easily be personalized with current technologies, for example by tracking colour preferences. To illustrate, colours, brightness, size and shapes, are all factors that belong to the visual dimension of atmospherics (Kotler, 1973). Not only would atmosphere personalization based on colour preferences be very convenient to integrate, but colours in marketing are also proven to induce behavioural changes. For example, from the product packaging field, we know that 62-90% of product assessments are based on the packaging colours alone (Singh, 2006). A possible explanation for why colour has such an impact on consumer behaviour can be assigned to one’s affective perception of colour, rather than to the arousal that the colour evokes (Bellizi & Hite, 1992). The deliberate choice of colour use in retail settings is demonstrated to have a significant impact on brand perceptions and evaluations (Labcreque & Milne, 2012). By introducing atmosphere personalization with colours, this deliberate choice of colour use does not need to be actively made by the company anymore, but can be automatically integrated on the online web shop.

(20)

physical store, auditory cues are often expressed through an assembled music playlist. Research demonstrates that background music in a store has a positive effect on shopping attitudes and behaviours (Sweeney & Wyber, 2002). However, for the sake of the timespan of this thesis, background music will not be further investigated, but the focus will lie on manipulating colours as part of atmosphere personalization.

Not only is manipulating colour fairly convenient to implement, but it could also bring along some interesting additional findings. From the field of Psychology, we can learn that specific colours can cause different effects. For instance, Bellizi and Hite (1992) demonstrate that more positive responses are found in retail environments that use the colour blue, opposed to red environments. Translating this to an online retail context, we see that products that are featured on a blue or purple background are evaluated more positively than when other colours are displayed (Biers & Richards, 2011). However, there is some indication that colour on websites is not the most influential factor to evoke positive responses. For example, the research by Lee and Koubek (2010) suggests that the responsiveness of a website has a greater impact on website preferences than aesthetic aspects like colour. This demonstrates that the functionality of a website outweighs the nice-to-haves, such as a fancy design. However, there is support for the moderating effect of involvement on the relationship between web design variables and customer satisfaction (Sanchez-Fanco & Rondan-Cataluña, 2010). To clarify, people that are low in involvement will pay more attention to web design cues such as aesthetics, whereas highly involved people will process the content more carefully. In sum, choosing for colour in this thesis could result in valuable consumer responses.

3.4 Threats of atmosphere personalization

Similar to what has been discussed in paragraph 2.2, the privacy concerns that consumers hold might have a negative impact on the effectiveness of our concept of atmosphere personalization. However, opposed to regular personalization, atmosphere personalization could bring along other threats on itself. This threat can be clarified by digging into the field of branding and more specifically the use of brand elements. The adequate composition of primary brand elements such as logos, slogans, jingles and packaging leads to strong brand perceptions and associations (Keller, 2005). Whenever atmosphere personalization is

(21)

integrated on a website of a brand that is characterized by its colours to a great extent, the memorability of the brand might be jeopardized.

To illustrate this problem, we analysed the Dutch online retailer bol.com. The website looks very clean and does not make use of many colours, except or their brand colour, which is blue. By personalizing the atmosphere into for example pink, the strength of the brand might fade. Due to the atmosphere personalization, the consistency in the entire integrated marketing communications of bol.com is hard to recognize, as they would still communicate the colour blue and white across their other communication channels. Keller (2001) describes the need for consistency amongst communication efforts by using the term commonality. Issues regarding commonality can relate to executional consistency problems across communication content. In our example, the person that is used to shopping in a pink atmosphere, might have less strong associations with the colour blue related to bol.com.

To stress the importance of this threat, a different study emphasises that consistency between brand image and website image influences the overall brand attitude (Muller, 2008). For consumers, their perception of the physical store and the online store might differ as a consequence of atmosphere personalization. There seems to occur a moderating effect of the perception of consistency on the relationship between the website image and the brand attitude. (Muller, 2008). Additionally, Blakeney (2016) addresses the importance of designing a consistent omnichannel brand experience, meaning that the way companies communicate ideally should be consistent across the marketing channels. In sum, the possible issue of atmosphere personalization leading to inconsistency in marketing communication efforts across channels should be carefully considered. At all times, the company using atmosphere personalization should make sure the primary brand elements are clearly visible across channels.

(22)

Chapter 4: Hypotheses and conceptual model

In the previous chapters, all relevant information regarding the topic was discussed. However, in order to be able to measure the effects of atmosphere personalization contrasted to content personalization, several hypotheses are presented in this chapter. Another variable that measures the explicit communication of personalization efforts will be added to the model so test its impact and whether it would reinforce the results. Perceived usefulness and perceived company effort function as mediators that are expected to explain the hypothesized effect on the service evaluation. In Figure 1, all expected effects are schematically displayed. In the following paragraphs, the motivation for the hypotheses as such is discussed.

Figure 1: conceptual model

4.1 Content personalization

Content personalization is widely adopted in ecommerce environments, which is why there are studies that tell us the reason why this concept is effective and in what sense. This has been discussed in the literature review chapters of this thesis. For example, recommendations – a form of content personalization – based on preferences, interests and previous search behaviour could lead to effort reduction. Previous research shows that the increase in customer satisfaction due to content personalization was mainly caused by effort reduction (Liang, Lai, & Ku, 2007). This means that consumers find personalized content useful in a sense that it eases their search for the right product. The article by Liang, Chen and Turban (2009) describes this as transaction costs, that are defined as “a cost incurred in making an

(23)

economic exchange, during the buying and selling transactions, other than the purchase cost of the product or service” (Dahlman, 1979). These costs can be tangible or intangible, but either way these costs are ought to be paid instantly by the customer. Therefore, people logically seek for the lowest transaction costs when shopping, and content personalization in an ecommerce environment opens up the opportunity to realise this. It is expected that the perceived usefulness of content personalization will be one reason why content personalization leads to better service quality evaluations. This expectation is in line with the findings that Chen (2012) proposed, where personalization has a positive impact on shopping experience, customer satisfaction and purchase intention. Additionally, customers find personalization an useful aid that eases the decision making process (Tam & Ho, 2006). Hence, the following hypothesis is formulated:

H1: The effect of content personalization on the service evaluation is positively mediated by perceived usefulness.

Next to the expectation that customers see added value in the usefulness of content personalization, they might also show positive affect to the fact that the company makes an effort for them by tailoring content based on their individual preferences. This is more of an emotional approach to determining the effectiveness of personalization. Previous research indicates that customers who feel that the company cares for them will have more positive attitudes towards the service of personalization (Liang, Chen, & Turban, 2009). Also, according to the SERVQUAL model by Parasuraman and Zeithaml (1988), the empathy dimension of service positively affects the service quality evaluation. Based on these findings, this thesis will also investigate whether positive emotional responses to content personalization occur that influence the online service evaluations.

H2: The effect of content personalization on the service evaluation is positively mediated by perceived company effort.

4.2 Atmosphere personalization

As already described in Chapter 3, atmospherics is a relevant marketing tool where there is a lot of competition, as firms use it to attract and retain customers in a differentiating way (Kotler, 1973). In this thesis, the concept of atmospherics is transferred to an online context,

(24)

where we also know from previous research that an appealing web design has positive outcomes within the customer journey (e.g. Novak, Hofmann & Yung, 2000). However, what an offline context is lacking, is that the store atmosphere cannot be personalized, as it would be simply impossible to alter the store design for each individual customer that enters the store. For this reason, there is a great opportunity for atmosphere personalization to become a differentiator for people to go shopping online. As we know that personalization and atmospherics are both effective tools, this is not an illogical line of reasoning at all. However, if this concept would indeed lead to better evaluations, we are curious to know the reason why. In order to be able to compare the outcomes with H1 and H2 regarding content personalization, the same mediators are desirable to be measured with regards to atmosphere personalization. Nevertheless, it is expected that atmosphere personalization only shows a positive effect through perceived company effort, as this new type of personalization has no direct purpose that will fulfil the functional needs and wants of the customer that is measured through perceived usefulness. It is more plausible that customers create positive evaluations because they simply like the effort that has been made by the company. Hence, it is expected that atmosphere personalization is not mediated by perceived usefulness, but is only mediated by perceived company effort that enhances the service quality evaluations.

H3: The effect of atmosphere personalization on the service evaluation is positively mediated by perceived company effort.

4.3 Explicit communication of the personalization efforts

Additionally, another independent variable is introduced, as it might influence the results of content- and atmosphere personalization significantly. It is not unthinkable that when either content personalization or atmosphere personalization is implemented, customers might not notice the efforts. Therefore, a moderator that measures the presence of explicit communication of the personalization efforts will be tested and added to the model. In other words, the extent to which a firm is mentioning that it applied personalization to increase visibility. On the one hand, it is expected that the hypothesized effects of content personalization on the service evaluation will be reinforced by explicitly mentioning the efforts. On the other hand, content personalization is already a commonly used concept that is established in the minds of the consumer. Especially product recommendations are the most used form of personalization that many people are familiar with (Ho & Bodoff, 2014).

(25)

Therefore, it is expected that customers will easily recognize content personalization, even without explicitly mentioning its presence. However, we do believe that when there is an explicit communication text on the website that stresses the content personalization efforts that have been made, people will notice the efforts sooner and might even become more likely to form stronger attitudes towards it. The reason for this is that online retailers can gain a competitive advantage by using social cues on their webpages that enhance perceptions of human interaction that create emotional bonds (Wang, Baker, Wagner, & Wakefield, 2007). The explicit communication of personalization efforts can be considered as such a social cue, as it will make the company look more engaged with its customers. For this reason, it is expected that the reinforcing effect of this message will only influence the scores on perceived company effort, as it does not have a direct purpose that would change the perception of usefulness.

H4a: The positive effect of content personalization on perceived company effort is stronger when the personalization effort is explicitly communicated to the website visitor, than when it is not.

Despite its expected impact on the hypothesized effects of content personalization, the variable of explicit communication is expected to have a larger and maybe more crucial impact on the results of atmosphere personalization. This is mainly because of the fact that people are not familiar with atmosphere personalization yet. What could happen, is that consumers do not notice the atmosphere personalization whenever firms do not mention that they implemented it. The reason for this is that people might think that the web design as such is generic and visible to anyone. Therefore, by explicitly mentioning that it is actually a personalized design might even be a critical condition under which the expected main effect could exist. Again, it is expected to have an enhancing effect on the perceived company effort that has been made. To clarify, the following hypothesis is proposed:

H4b: The positive effect of atmosphere personalization on perceived company effort is stronger when the personalization effort is explicitly communicated to the website visitor, than when it is not.

(26)

Lastly, also a direct effect of the explicit communication message is expected to occur on the service evaluation through the perceived company effort variable. The reason for this is that consumers have positive attitudes towards interaction with a brand (Chen & Dubinsky, 2003). They might consider the website as being more human, for which they subsequently create more positive attitudes (Åberg & Shahmehri, 2000). Hence, the following hypothesis is proposed:

H5: The effect of explicit communication on the service evaluation is positively mediated by perceived company effort.

(27)

Chapter 5: Methodology

5.1 Research design and stimuli development

A 2 (content personalization) x 2 (atmosphere personalization) x 2 (explicit communication) between-subjects full factorial research design is used to test the hypotheses. All the independent variables consist out of two levels, being either present or absent. First, a pre-test was conducted to determine the effectiveness of the manipulations. After that, the final experiment was conducted. For the final experiment, all 215 participants were randomly assigned to one of the eight conditions. Table 1 visually displays these conditions.

Table 1: experimental conditions

atmosphere personalization (AP) explicit comm. content personalization (CP) yes no

yes CP – AP - comm. No CP – AP – comm.

no CP – No AP – comm. No CP – No AP – comm. atmosphere personalization (AP) no comm. content personalization (CP) yes no

yes CP – AP – No comm. No CP – AP – No comm. no CP – No AP – No comm. No CP – No AP – No comm.

A website called The Wardrobe.com was developed in HTML, CSS and JavaScript to represent the stimuli. This name was developed based on observations in the ecommerce field, together with the idea that the name implies the products that are being offered in this fictional shopping environment. To test whether the brand The Wardrobe.com was easy to comprehend, we asked five people to name the first associations that this name evoked. From this small test, everyone named thoughts that were related to an online fashion store. The choice for an online fashion store can be justified by the observation that personalization is often incorporated on these sorts of websites. For example, online fashion retailers zalando.nl and asos.com have been observed. Of course, there are other ecommerce branches that offer personalized services apart from the fashion industry, such as electronics and books. However, the choice has been made to focus on fashion in this experiment.

Next, we made a first composition of products that were going to be displayed on the page, using six product cards with three male and three female jeans. Again, we asked some

(28)

relatives to give their opinions about the products that they saw. However, there seemed to be occurring some difficulties with the products, as people criticised the fact that existing fashion ecommerce websites normally do not display male and female apparel on the same page. Based on these insights, the website was altered by now integrating more unisex products. The final products were six accessories and consisted of a watch, gloves, scarf, bag, socks and a cap. By changing the offerings into unisex products, we increased the credibility that the products were hypothetically personalized, regardless of the gender of participants.

Content personalization

Content personalization is manipulated through the creation of a fictional ecommerce webpage where participants in this condition receive personalized recommendations. Due to time constraints of this study, it is not possible to conduct a field research where search behaviour and preferences can be traced to integrate accurate personalization. Therefore, we work with more hypothetical personalization where a textual indication of personalization is integrated, based on websites of renowned online websites that personalize. For example, Amazon uses “recommendations for you, @firstname”, Marktplaats uses “voor jou” (for you) and MediaMarkt “interessant voor jou” (interesting for you). Based on these examples, content personalization was first manipulated through putting a section with “personalized

offers for you” on the webpage for this specific condition. The independent variable of

content personalization can take values 1 (= present) and 0 (= absent).

Atmosphere personalization

As explained in Chapter 3, atmosphere personalization is manipulated through using colours. We have learned that colour – together with brightness, size and shapes – belongs to the visual dimension of atmospherics (Kotler, 1973). The choice for manipulating colour is motivated mainly by its convenience, as this research is the first attempt in investigating whether personalized atmospherics can be effective. If this is the case, further research can study whether other aspects of atmospherics can also be effectively personalized. The colours that are used to manipulate atmosphere personalization are based on research that depicts the most prevailing favourite colours of people, which respectively are blue, green and red (Hemphill, 1996). Hence, participants will get a question at the beginning of the experiment that askes to pick their favourite colour amongst those three and in turn the participants in atmosphere personalization conditions will receive a webpage styled in this colour accordingly. The colour itself that people choose will not influence on the atmosphere

(29)

personalization variable, as this variable can still only take the value of either 1 (= yes) or 0 (= no). However, later in the research we might want to investigate the impact of the colour itself and include it as a control variable.

Explicit communication of efforts

Lastly, another manipulation is created. The explicit communication of efforts will be manipulated by putting a text message in a popup that is embedded in the webpage, so that it will be clearly visible to the participants and they actively need to click on the message to make it disappear. Since this message should be meaningful for both content- and atmosphere personalization efforts, it should be a generic message that implies personalization in any manner. Therefore, the following sentence is developed: “Here you go! We’ve created a

unique web page just for you, based on what we think you’ll like.”. To clarify, this message

will either be present or not, so it can take the levels of 1 (= yes) and 0 (= no).

5.2 Pre-test

In a pre-test to test the quality of the stimuli material, 55 participants were asked to fill out an online survey twice through survey tool Qualtrics, so that 110 observations were collected (n = 110). The sample was congregated through a convenience sample. As it is only desirable to test whether the manipulations are effective, not all conditions were put into the pre-test survey, but only five. These conditions were: 1) content and atmosphere personalization with communication message, 2) content and atmosphere personalization without message, 3) atmosphere personalization without message, 4) content personalization without message and 5) no personalization, no message. The participants were randomly allocated to two of the conditions. The entire pre-test questionnaire is included in Appendix 1.

Content- and atmosphere personalization

To test whether the participants perceived the manipulation of both types of personalization as intended, content personalization and atmosphere personalization were measured on 7-point Likert scales with four items each and were based on the research of Srinivasan, Anderson and Ponnavolu (2002). First, a factor analysis was conducted to see whether content personalization and atmosphere personalization indeed load on different components. The principal component analysis found two components with an Eigenvalue > 1, that together explain 79.87% of total variance. Using the Varimax rotation technique, the items

(30)

that belonged to either content- or atmosphere personalization were recognized as such by the model of the factor analysis.

For both variables, the items were tested on reliability using Cronbach’s Alpha. The scale for content personalization scores high on reliability, with a = 0.891. The scale for atmosphere personalization scores high as well with a = 0.930. In case participants would not perceive the manipulation as atmosphere personalization, also a variable called colour personalization was measured and also showed a high alpha (a = 0.906). After testing the effect of the manipulation on the scale of atmosphere personalization, we will decide on whether we need to measure something else, such as mere colour personalization instead of the entire atmosphere being perceived as personalized. Since all Cronbach’s Alpha’s show high values, there is no need to delete items in order to enhance the reliability. As the factor analysis and reliability analysis both reported desired results, average scales could be created. Also, two independent variables were computed derived from the dataset, called “Content Personalization” and “Atmosphere Personalization” that measure whether the personalization type was either present (1 = yes) or not present (0 = no).

Table 2 shows the mean scores of the scales in relation to the stimuli. From this table, we can already see that the additional measure of colour personalization shows the highest mean. This could indicate that people perceive the manipulation more as colour personalization than atmosphere personalization. Moreover, we can see that there are no striking differences between the values for content personalization, which can be problematic. The tests will give us statistical evidence whether or not the content personalization manipulation has been successful.

Table 2: mean scores stimuli testing

Manipulation CPm APm COLm

Content personalization 3.32 3.31 3.77

Atmosphere personalization 3.21 3.79 4.39

Note: CPm = content personalization measure, APm = atmosphere personalization measure,

(31)

To measure the effectiveness of the manipulation, a factorial ANOVA test was conducted to compare mean scores and to test whether there occurs an interaction between the manipulations, which is not desired. The ANOVA test pointed out that the atmosphere personalization manipulation reported significant results on the atmosphere personalization measure, F(3, 106) = 10.03, p < 0.001, h2 _{= 0.09. No significant values were found for}

interaction effects, nor on content personalization. This means that the manipulation was successful as atmosphere personalization. However, as we also measured the variable “Colour Personalization”, and a similar test was analysed to compare effect sizes. The atmosphere personalization manipulation also reported significant results on the colour personalization measure, F(3, 106) = 83.02, p < 0.001, h2_{= 0.29. These results indicate that}

when using colour personalization instead of atmosphere personalization, the effect size improves from moderate to strong. These conclusions might predict that in the final experiment people will also consider the manipulation more as colour personalization than atmosphere personalization and is therefore vital to take into account. Lastly, another factorial ANOVA on content personalization was not significant. Nevertheless, the test did not report any undesired significant results either, such as an interaction or a significant effect on atmosphere personalization. Therefore, the manipulation should be altered for the main experiment in order to become successful. How this is done will be discussed in section 5.2.1.

Explicit communication

To test whether participants in condition 1 noticed the popup message, Pearson's chi-squared test was conducted on the explicit communication variable and the measure. The awareness of the message was measured by simply asking the participant whether he or she noticed a popup on the webpage and with that also what it tried to communicate as a double check. For the manipulation, a new variable was computed that either contained the explicit popup message (= 1) or did not have the message (= 0). The measure question had three answer options: “yes” (= 1), “no” (= 2) and “I don’t know” (= 3). As “I don’t know” also contains information, but is considered to be less rigorous than “no”, this option was recoded as the middle value (= 2) to become an ordinal variable. Chi-squared (n = 110) was significant, χ2(2) = 20.34, p < 0.001. However, to see whether the “I don’t know” option is of value to us, an additional chi-squared test was conducted with these cases excluded (n = 11). It could be possible that participants that did not get any popup were insecure about answering a harsh

(32)

manipulation. When excluding “I don’t know”, the chi-squared value (n = 99) improved to χ2_{(1) = 38.98, p < 0.001. In both cases, there is a positive connexion between the}

manipulation and the communication measure. Therefore, it can be concluded that the manipulation for explicit communication was successful as such.

However, in order to be sure that also the message itself was understood, the qualitative answers were examined asking people what they thought the message stated. 67% of participants that were exposed to the message could adequately remember what the message stated, the remaining 33% did not answer the question with a full answer and mainly reported “no reply”. As in the eventual research we are not looking for whether participants are able to reproduce the message that was emitted, but merely whether they recall a message, we can still conclude that this manipulation is successful as such.

5.2.1 Adaptations content personalization

Since the pre-test showed that the content personalization manipulation as such was not successful, the displayed text was therefore altered to become more robust into: “Some

personalized offers for you, based on what we think you'll like!”. Additionally, a

scenario-based method was implemented prior to the conditional exposure to bring people in the state of mind that the products that they will see are actually based on their preferences. The scenario enables us to investigate a possible and plausible state of mind (Camponovo, Debetaz & Pigneur, 2004) and are therefore very useful in this research. The scenario stated that participants should imagine that the products they would be seeing on the website are actually based on individual preferences, even though they would not be their preference in real life.

The final stimuli materials can be found in Appendix 3.

5.3 Sample

A sample of 215 participants was collected through a convenience sample combined with a simple random sample through the service of Amazon Mechanical Turk. The convenience sampling technique was very helpful taking the short time span of this research into mind. However, the convenience sample might not be very representative, as it is a non-probability sample and only included participants that were in some way connected to the researcher.

(33)

Nevertheless, by also acquiring a sample through Amazon Mechanical Turk, we have broadened the reach of the sample internationally and made use of a probability sampling method simultaneously. It is expected that both types of samples will not negatively influence the external validity of the research outcomes, as it seems strong to use both sampling techniques. Additionally, only high ranked people on Amazon Mechanical Turk were asked to participate, to minimise the chance of receiving useless responses that have not been filled in seriously.

5.4 Operationalization of output variables

Online service evaluation

The online service evaluation (DV) was measured only after the manipulation, as results from the control group tell us whether the possible differences in evaluations are caused by the manipulation. The variable was measured using an established scale developed by Lepkowska-White, Page and Youndt (2004) that measures the overall consumers’ perception of a website. This scale consists of four items and are fairly general and were altered to fit with our research purpose. For example, one of the items included the statement “My

perception of the services on this website is positive”. All four items were measured on a

7-point Likert scale, ranging from (1) Strongly disagree to (7) Strongly agree.

Additional to the measure of the online service evaluation, also attitudes towards personalization and purchase intentions were included in the experiment for follow up analyses. Attitudes were measured using a shortened version of the attitude semantic differential developed by Yoo and Donthu (2001). We picked three items that were most relevant for this thesis. Next to the attitude measure, also purchase intentions were included, based on the work of Spears and Singh (2004). Again, the number of items was reduced to three to simplify the questionnaire.

Perceived usefulness

The mediator perceived usefulness is measured through a 7-point Likert scale with values 1

(= strongly disagree) and 7 (= strongly agree), based on the research by Henderson and

Divett (2003) and Davis (1989). In this thesis, we consider the perceived usefulness as the extent to which a certain service would enhance the convenience of using that particular service for the customer, for example through effort reduction. The three principal items that

Step into your own store : the impact of atmosphere personalization in ecommerce environments on service quality