ASSESSING THE ATTRACTIVENESS OF
PRODUCT PAGE ATTRIBUTES: A
CONJOINT ANALYSIS
by
Robin Bette
University of Groningen
TABLE OF CONTENT
1 Introduction ...3
2 Literature review ...8
2.1 Attractiveness of a product page ...8
2.2 Antecedents of the attractiveness of a product page ...9
2.3 Moderating the attractiveness of a product page... 18
2.4 Conceptual model ... 23
3 Methodology ... 24
3.1 General design ... 24
3.2 Data collection ... 25
3.3 Variables ... 25
3.4 Conjoint study: attractiveness of the product page ... 26
3.5 OLS study: purchase intention ... 28
3.6 Data limitations ... 30
4 Analysis and results ... 30
4.1 Respondents’ characteristics ... 30
4.2 Conjoint study ... 33
4.3 OLS study ... 44
5 Discussion, conclusions and future research ... 53
5.1 Findings and recommendations ... 53
5.2 Limitations and future research ... 56
6 References ... 60 6.1 Articles ... 60 6.2 Books ... 67 6.3 Lecture sheets ... 67 6.4 Online sources ... 67 7 Appendices ... 68
Appendix A: Webshop product pages (pictures) ... 68
Appendix B: Questionnaire (male, version 1) ... 70
Appendix C: Sawtooth efficiency output ... 77
Appendix D: Factor analysis overview ... 77
Appendix E: Internal hit rate calculations ... 78
Appendix F: Overview relative importance and utility ... 78
Appendix G: Calculation predictive validity (hit rate) ... 79
ABSTRACT
This paper describes which product page attributes assist in (1) optimizing the attractiveness of a webshop’s product page, and (2) increasing purchase intention. Three distinct customer segments are distinguished, with specific preferences. Practical shoppers mainly prefer a size availability table. Hedonic shoppers mainly prefer a product presentation by a celebrity, and finally, value shoppers prefer a simple product page, but with a low price. Additionally, both preference for hedonic shopping characteristics, and susceptibility for social influence have a moderating effect (hypotheses 10a1 and 12a2).
1
INTRODUCTION
The growth of the internet in the past decade has led to a change in focus of both managers and researchers. Research on e-commerce, a specific subject within the internet, is increasing fast. Moe and Yang (2009) describe for example in their paper that a shift in marketing research occurred lately from offline consumer search1 to the online environment. This shift is stimulated by the availability of rich clickstream data, which is defined by Bucklin and Sismeiro (2009) as “the electronic record of Internet usage, collected by web servers or
third-party services”. Rigby (2011) adds that online sales are increasing fast and, as a consequence,
traditional brick and mortar retailers are lagging behind. These examples show that the online environment is interesting for marketing research. It strikes though, that still surprisingly little research is available about webshops, and in more detail, optimal designs, which can enhance the attractiveness of a webshop’s product page. A product page, is the online webpage where you arrive after clicking an item in the webshop. It consists of several product page elements (e.g. pictures, or customer reviews), which are referred to as (product page) attributes in the remainder of this paper.
The subject of this paper is ‘optimization of the attractiveness of a product page’. So, although large webshops probably perform research themselves, no academic literature exists yet, providing information about for example customer preferences for certain attributes, or important attributes for a product page. In short, with which attributes should a product page be composed, to reach and attract certain groups of customers, in order to increase metrics as conversion rates, purchase intention and ultimately, even revenues? Guidelines which help enhance these metrics are interesting for webshops, to become more profitable. Add that the online retail environment is increasing fast, certain knowledge might definitely be necessary.
1
In order to find an ‘optimal design’, several designs with different combinations of attributes are tested, in order to find an optimal attractive product page design. The goal of this study is to offer a persuasive product page, to ultimately increase a customer’s purchase intention. Purchase intention is defined in this paper as the extent to which a consumer is willing to buy
a product. This definition is based on information derived from papers by Jiang, Hoegg, Dahl
and Chattopadhyay (2010), and Suk, Yoon, Lichtenstein and Song (2010). For example, in the latter, purchase intention is measured by asking respondents to rate their willingness to buy a product, on a scale of one (I do not want to buy the product) to nine (I want to buy the product). In the current paper it is expected that purchase intention is influenced positively by the persuasiveness of the product page. The more attractive and persuasive the page is, the higher the purchase intention is expected to be.
The assumption that a more attractive product page leads to a higher purchase intention, is amongst others based on findings by Kotler (1973). He found that store interiors and exteriors can be designed in ways which ultimately improve purchase intention. Hence, if this is applicable to traditional brick and mortar stores, an attractive and efficient design might also be effective for online stores. A paper by Song and Zahedi (2005) underscores the importance of design. They state that understanding how online customers are persuaded by various design attributes, can help predict online customer behavior. Additionally, according to Smith and Sivakumar (2004), online retailers should determine how the content and layout of their webshops can be optimized, in order for visitors to achieve a certain type of flow. Flow is an optimal psychological state that can be activated by using the internet (Novak, Hoffman and Yiu-Fai Yung, 2000). It can for example affect a consumer’s willingness to buy, if he has an optimal experience when visiting the webshop. Hoffman and Novak (1996) add that it is important for marketers to design commercially attractive websites, in order to achieve the right state of flow.
In short, if an attractive product pages makes online shopping and browsing2 more interesting for customers, this might enhance purchase intention. The argument that website design can affect purchase intention is also supported by findings of Park, Kim, Funches, and Foxx (2012). They show for example that product related attributes at websites (e.g. price, variety of selection and sensory attributes) can influence web browsing. This can especially be important for apparel products, which are ‘feel and touch’ (or experience) products, as consumers are not able to try the products in an online shop (Park et al., 2012). A stimulation
2
of web browsing can finally result in affecting both utilitarian (e.g. easier for consumers to gather information) and hedonic shopping behavior (e.g. stimulate adventure shopping, by making the visit more ‘thrilling’). According to prior research, utilitarian shopping behavior is a rational process, focused on completing the shopping task (Batra and Ahtola, 1991). Hence, its main goal is to acquire a desired product (Babin, Darden and Griffin, 1994). Hedonic shopping behavior refers to stimulation of senses and for example having fun during the shopping process, and consumption (Hirschman and Holbrook, 1982). Concerning websites, Park et al. (2012) found that the use of price attributes affects hedonic shopping behavior positively (e.g. stimulating the hunt for bargains when the price is low, where the hunt itself is seen as a positive experience). In short, both shopping motivations can be manipulated by marketers, by offering specific attributes at the product page and increase its attractiveness.
As little research is available yet, concerning product pages, literature with subjects as
advertisements and website design is reviewed. This might provide useful information, for
The current study is divided in two parts: a conjoint analysis and an analysis with ordinary least squares (OLS). With the former technique, various product page designs with combinations of attributes can be shown to respondents. They are asked to judge which design is most attractive. Sela and Berger (2012) describe in their paper, that products consist of several attributes, which influence the product alternative that a consumer desires to buy. As certain attributes benefit some alternatives, this affects a consumer’s purchase decision. A product page consists of several attributes as well. Therefore, it is interesting to use a conjoint analysis to study which attributes are most influential in increasing the attractiveness of the page. The latter technique is used to test which attribute enhances purchase intention the most. Both techniques are thoroughly discussed in chapter 4.
The choice to use clothing as a product is based on its increasing popularity in the online environment. For example, Kim and Kim (2004) describe that online sales of clothing is growing. Hence, choosing an interesting product category makes an experiment for consumers more interesting. This might result in less respondent fatigue and boredom and consequently, better answers (Narayana, 1977).
Next to the product page attributes, several additional factors, like demographics, might influence perception of the attractiveness of a product page. These factors might be useful in identifying distinct customer segments with different preferences for certain attributes. In short, the research question of this study is:
Which attributes assist in optimizing the attractiveness of a webshop’s product page and does their effect depend on consumer demographics, hedonic and utilitarian shopping motives, or social influence?
New knowledge in the area of e-commerce is expected to be derived from this study. Its theoretical and practical relevance are now discussed briefly.
On the contrary, practical relevance concerns findings usable for marketers. They might be interested in which of the attributes of a product page influences the attractiveness of their product pages the most, distinguished by customer segments. This knowledge can be used to ultimately increase the purchase intention of visitors. An increase in customer’s purchase intention might result in an increase of sales. Sales and its relation with certain marketing actions is an important issue in marketing research. For example Verhoef and Leeflang (2009) describe in their paper that the marketing department needs to enhance its influence within the firm, which can be achieved by becoming more controllable and accountable. Rust, Lemon and Zeithaml (2004) address a similar issue and provide an example which states the uncertainty of managers, concerning which marketing initiatives to invest in. This is still often based on intuition and experience of these managers. Finally, according to Mayzlin and Shin (2011), firms can only communicate a limited amount of information about product attributes to consumers, due to budget limitations. Hence, a similar situation might exist for webshops, as for example resource constraints might prevent them from using every possible attribute at their product pages. For instance, hiring a celebrity to present their products might be expensive. Therefore it can be interesting to know which attributes are most useful for product pages, to be able to persuade different consumer segments. Here, it is also interesting to test which price consumers want to pay for an increase in an attribute level. For example, a consumer might be willing to pay more money for a product presentation by a celebrity, instead of a presentation by a random model.
In conclusion, marketing research lacks knowledge about product pages and its attractiveness. Therefore, the current study’s objective is to identify which product page attributes contribute the most to the attractiveness of the page and which are able to influence a consumer’s purchase intention. This might be related. The final goal of this study is providing recommendations to marketers, about offering a more persuasive and attractive product page.
2
LITERATURE REVIEW
This chapter concerns a review of articles, relating to the attractiveness of a product page and its five antecedents; the attributes. Potential moderators which might affect the relation between the attractiveness of the product page and the attributes are discussed as well. Several hypotheses are proposed, of which an overview can be found in subsection 2.4.
2.1 Attractiveness of a product page
A definition of attractiveness of a product page has not yet been formulated in prior literature. However, amongst others, it is closely related to aesthetics, as beauty and an attractive design might attract customers. For example, Hauser, Urban, Liberali and Braun (2009) describe in their paper that a well-designed, beautiful website can ultimately enhance profits returned from this website. Furthermore, Veryzer and Hutchinson (1998) describe in their paper that attractive designs create more favorable attitudes than unattractive ones. This suggests that attractiveness of a website might go hand in hand with positive beliefs about its products. Therefore, an attractive product page might lead to positive beliefs about the presented product, which might ultimately increase purchase intention.
Additionally, an attractive product page is assumed to be highly arousing. Arousal is defined as “the level of alertness or activation on a continuum, ranging from extreme
drowsiness to extreme wakefulness” (Sanbonmatsu and Kardes, 1988). These authors suggest
that high levels of arousal might result in a decrease of capacity to process information. This decrease might suggest that peripheral cues of a product page (e.g. pictures) are necessary to persuade a visitor, to make the page more attractive and arousing (because pictures are often more easy to process than textual messages). This issue is discussed more thorough at the product pictures subsection (2.2.1).
in this study might affect the attractiveness of a product page and purchase intention. Several other papers confirm that website design is important. For example, Kim, Kim and Kandampully (2007) state that it affects attracting, sustaining and retaining customers to the website, and Szymanski and Hise (2000) found that site design is a dominant factor in online consumer satisfaction.
Next to design, aesthetics and arousal, understandability is an important aspect of websites (and therefore assumed for webshops and product pages as well). A website should be well-organized and clear, to enhance the perceived attractiveness of the page. For example, Song and Zinkhan (2008) state in their paper that a website should be easy to navigate and to use. This makes a website clear, well-organized and probably more attractive. Voorveld, Neijens and Smit (2011) confirm in their paper that navigation and ease of use is important for websites interactivity.
Combining all these findings, the attractiveness of a product page is defined in the current paper as “the extent that a product page is visually pleasing, arousing and easy to use,
which enables its ability to persuade a visitor to continue his or her purchase process”.
Hence, a page should be attractive, which can be due to its pleasing design, but should also be clear, easy to use and understandable. So, attractiveness does not only relate to ‘beauty’.
In conclusion, the attractiveness of a product page of an e-commerce site is closely related with aesthetics, should be well-organized, and easy to use. The attractiveness of the product page represents the total utility of the five product page attributes. This utility concerns the product pages’ visitors’ preferences for the attributes. The higher the total utility of the five attributes together, the higher the perceived attractiveness of the product page.
2.2 Antecedents of the attractiveness of a product page
This subsection reviews articles which relate to the five attributes.
2.2.1 Product pictures
opportunity to experience (e.g. feel) the product (Ha and Lennon, 2010). Kim and Lennon (2008) show that this uncertainty for customers can be reduced by presenting sufficient visual information at the website (presenting product information might also be interesting, which is discussed in subsection 2.2.2).
Research in the area of advertising provides knowledge about product presentations by using pictures. Although the current study does not concern an advertisement, product pictures are expected to be an important tool for gaining a customer’s attention. Pieters and Wedel (2004) show for example that pictures are the superior element of three key advertisement elements (brand, pictures and text). These elements have important and unique effects in drawing attention. Hanssens and Weitz (1980) add that the use of photographs and illustrations in advertisements can be very effective in recalling products. Findings of Childers and Houston (1984) confirm this; recall of picture only forms of advertisements is higher than verbal only forms. Furthermore, according to Petty, Cacioppo and Schumann (1983), two routes to attitude change exist: the central and the peripheral route. The route depends on the level of involvement of the consumer. According to these authors, high product involvement has greater personal relevance than low product involvement. For example, a person who wants to buy a certain product, is highly involved with that product category. High involvement relates to the central route, and is discussed further in subsection 2.2.2. When one’s involvement with a product is low, he or she is often persuaded by simple acceptance and rejection cues. This situation refers to the peripheral route. These cues are for example product pictures or attractive endorsers. Although this study does not research one’s involvement with a product, some products can be regarded as high or low involvement products. For example, when one wants to buy a car (high involvement), extensive information search might be necessary. But when one is planning to buy a basic white t-shirt (low involvement), he or she might buy it without really considering additional information or alternatives, and is just persuaded by simple acceptance and rejectance cues. This can for example be a picture of the t-shirt itself, which might look nice enough to be bought, or the attractive endorser who seems to fit this t-shirt well.
Finally, besides academic literature, the importance of product pictures can be derived from a review of the four previously mentioned webshops, as they all use pictures to present their products.
detail by researching a specific form of product presentations. Namely, using two forms of product presenters: normal (or ‘unknown’) models and celebrities. The use of an attractive product presenter might increase attractiveness of the product page even more than showing the product only, for at least two reasons. First, as described before, attractive presenters can be useful in persuasion when customers have low involvement with a product. Second, other prior literature shows that attractive presenters can be useful, as ‘beauty’ is an important aspect nowadays. For example, Ahearne, Gruen and Jarvis (1999), describe that factors like the media, culture, or just human nature, make people focus a lot on beauty. They also state that attractiveness influences people’s actions, attitudes and judgments. Additionally, according to Kahle and Homer (1985), an attractive model might lure viewers into an advertisement, increasing their involvement with it and affecting their attitude towards it. DeShields Jr., Kara and Kaynak (1996) add that viewers of an advertisement behave according to a physical attractiveness stereotype and therefore perceive a message more favorably when an attractive endorser is used. This leads to higher purchase intentions than when the source is less attractive, which is confirmed by findings of Miller and Allen (2012) who show that the attractiveness of an endorser can influence brand outcomes.
Social comparison is an issue concerning attractiveness and beauty products. Clothing
and apparel can be seen as attractiveness-relevant products, which is identified by Bower and Landreth (2001) as a product category which enhances beauty. Such a product can be beautiful itself, like a necklace, or can enhance beauty by wearing it, like lipstick3. Clothing, used in the current study, might also be seen as an attractiveness-relevant product, as it can be beautiful itself. Clothing can be used to express oneself, or to increase self-esteem (Bloch and Richins, 1992). Next to self-esteem, people often want to be beautiful to meet the needs of social comparison standards, as product preferences are often derived from other persons’ opinions (Irmak, Vallen and Sen, 2010). Richins (1991) adds that comparison standards can be raised by showing a consumer highly attractive models in advertisements. This can even lower one’s ‘self-satisfaction’. Additionally, Smeesters, Mussweiler and Mandel (2010) underscore the importance of social comparison by investigating the influence of body weight index. They found that social comparison standards are different for persons varying in body weight who are exposed to models who also vary in body weight. For example, low body weight respondents showed a positive shift in self-esteem when exposed to thin models.
3
Finally, studies of Bower and Landreth (2001) show that highly attractive models are more persuasive than normally attractive models in presenting attractiveness-relevant product categories.
In short, all these studies underscore the usefulness of an attractive model. Because of this usefulness, it might be interesting to test if celebrities increase attractiveness even more than these ‘normal’ models. For example, they might be perceived as more attractive, as they are often well-known. In the next paragraphs, the usefulness of celebrities is discussed.
Tanner and Maeng (2012) recommend the use of celebrities, as their well-known faces are able to influence trust. Weisbuch and Mackie (2009) add that familiar people can be more persuasive in advertisements, than unfamiliar ones. Kamins (1990) adds that for an attractiveness-related product, the use of an attractive celebrity enhances credibility and attitude towards an advertisement.
Although using celebrities can be very useful, there should be a ‘fit’ between the person and the product. For example, McCracken (1989) found that the attractiveness and credibility of a celebrity are useful in advertising, but notes that celebrities are highly individualized, and because of this, they are only useful in commercials with products which match with them. Miller and Allen (2012) recommend this ‘fit’ as well. Further confirmation can be found in a paper by Fleck, Korchia and Le Roy (2012), who state that congruence exists between a product and an endorser. Because of this, a fit is necessary between the endorser and the product, which results in an association of the consumer with the endorser and his attitude or lifestyle. Finally, Praxmarer (2011) found that for attractiveness-unrelated products, attractive presenters are also more persuasive than ‘average’ presenters, due to trust and perceived competence of the presenter. Furthermore, she found that attractiveness is also effective when males consider male advertisements (or vice versa, for females). In the current study, this means that male respondents might be affected positively by a male celebrity.
Some arguments against celebrities exist as well. Sanbonmatsu and Kardes (1988) show that willingness to buy does not differ when respondents are exposed to advertisements with, or without, celebrities, as the product is not attractiveness-relevant (they used a pen in their study). Additionally, Tripp, Jensen and Carlson (1994) found that the number of products a celebrity endorses, affects his or her credibility and likeability in a negative sense. This decrease starts from a number of four products per endorser.
H1: The attractiveness of a product page is higher when a celebrity presents a clothing product, than when an unknown model presents the product.
H2: A visitor’s purchase intention is higher when a celebrity presents a clothing product, than when an unknown model presents the product.
2.2.2 Customer reviews and online product ratings
Three sources of information exist regarding guidance in a purchase decision process (Naylor, Lamberton and Norton, 2011). These sources are (1) paid agents, (2) nonpaid experts and (3) friends and family members via word of mouth communication. A new, fourth, source of information are reviews written by ‘unknown’ (and often ‘non expert’) customers, with which the focal customer has no relationship (Naylor et al., 2011). The third and fourth source are especially interesting for the current study, as webshops often use a form of word of mouth at their product pages, for example by offering a customer reviews (or online product ratings) attribute at the product page. The importance of word of mouth is extensively studied. For example, decades ago, Arndt (1967) found that positive product-related conversations are able to increase the acceptance of a new product (and vice versa for negative conversations). Additionally, Berger and Schwartz (2011) describe in their paper that word of mouth between consumers affects product adoption and sales, as consumers discuss the perceived value and quality of millions of products and services each day. Furthermore, Zhao and Xie (2011) describe that recommendations (or word of mouth) can be used by consumers to make more informed decisions (e.g. in the buying decision process). Perhaps even more important, Godes and Mayzlin (2009) describe that a company itself is capable of creating word of mouth and is therefore able to enhance sales. Opportunities arise to use it as a tool to initiate and manage interactions between consumers. Finally, West and Broniarczyk (1998) describe that word of mouth is used by consumers to decrease perceived risk when buying a specific product. This perceived risk will be higher, when the product is for example more expensive. Although risk is out of scope in the current study, it underscores the importance of word of mouth.
In short, prior literature shows that word of mouth can be very useful in increasing sales. Therefore it seems interesting to apply an attribute at the product page, based on the opinions of other consumers. These opinions can persuade (or dissuade) customers in their purchase process. In the current study, a word of mouth attribute at a product page is represented in two forms: (1) a customer reviews attribute and (2) a pictorial
The importance of customer reviews (and online product ratings) is described by Sridhar and Srinivasan (2012). They state that consumers are highly influenced by other consumers’ online product ratings. Especially high rated products increase consumers’ expectations enormously. Customers increasingly use customer reviews before an actual purchase and this source of product information gathering is gaining with respect to traditional marketing channels (Ludwig, de Ruyter, Friedman, Brüggen, Wetzels and Pfann 2013; Zhu and Zhang, 2010).
Other information that underscores the importance of customer reviews can be derived from papers concerning persuasion. Two routes to attitude change exist (Petty et al., 1983). The peripheral route, concerning product pictures is discussed in subsection 2.2.1. The
central route, relates to highly involved consumers. Highly involved consumers might be
motivated to search for (and scrutinize) product information. This information can be found (for example) in the customer reviews, where customers discuss the product’s quality. Finally, Dröge (1989), states that for high involvement products, peripheral cues like product pictures are relatively unimportant and hence, arguments are important.
In short, as clothing is an experience product (Ha and Lennon, 2010), reviews of other customers might help in reducing uncertainty regarding the quality and fit of clothing for some consumers. Hence,
H3: Presenting customer reviews and online product ratings, enhances the attractiveness of a product page.
H4: Presenting customer reviews and online product ratings enhances a visitor’s purchase intention.
2.2.3 Pictorial recommendations
According to Ying, Feinberg and Wedel (2006), product recommendation systems are important for online retailers, as these can enhance customer acquisition and retention; a customer might buy a product based on an online recommendation. Diehl (2005) states that online screening tools are able to make purchase recommendations to customers, based on some available alternatives. This can be convenient for consumers, as it can decrease their search costs, for example. Online retailers often use a recommendation system by presenting pictures which are similar to, or complement the focal product4. A consumer might become persuaded by a product which gains his or her attention at the moment of visiting the product
4
page. Hence, the shopping cart might be filled with a complementary product and the average order value of a visitor might be increased. Kamakura (2012) found that sequential market basket analysis can provide insight in how the purchase of one product can lead to the purchase of another. With this insight, a webshop is able to increase the opportunity of cross-selling (e.g. matching jeans with a t-shirt). This insight can be exploited, for example, when the webshop is capable of finding patterns of well matching products. Consequently, these patterns can be used as a form of pictorial recommendations; showing the respondent products which match well with the product they are currently viewing. According to Dellaert and Häubl (2012), it can be effective to help consumers with their purchase or search session, by showing recommendations which sort alternatives based on attractiveness of a product. Nowadays, it is even possible to provide these sets of alternatives automatically via information systems, which is also confirmed by Chung and Rao (2012). They describe that, nowadays, most online retailers are able to collect data on consumer preferences from their first visits. These preferences can become more detailed consumers actually purchase products over time, instead of just browsing. Assumed here is that a purchase shows what the consumer wants or needs and what his preferences are within a set of alternatives, whereas browsing can be random, just to obtain a lot of information.
Based on the review of the four previously mentioned webshops, two pictorial recommendation possibilities are identified. These are (1) recommended clothing pictures by other customers and shows for instance a message as, ‘other customers liked the following
items as well’, and (2) recommended trends by the webshop. Regarding trends, think of
products which are currently purchased often (or browsed for often) at the webshop, by all its visitors, or are just brand new trendy products. As recommendations might persuade visitors to purchase additional items, it might be able to enhance total sales of the webshop. Sriram, Chintagunta and Neelamegham (2006) describe in their paper that the attractiveness of a brand’s product line can differ over time. The introduction of new products is one factor that might enhance this attractiveness. Consequently, trends might be able to increase the perceived attractiveness of a product page. Reynolds (1968) adds that recognizing trends in time is crucial for success in many industries, especially the clothing industry. He states:
“There is money to be made in understanding fashion, and money can be lost in misunderstanding it. Most of the time, however, fashion helps rather than hurts marketers.”
fashion-oriented behavior occurs. In short, as trendy products are very important in the clothing industry (Reynolds, 1968; Sproles 1981), hypothesized is that consumers think that trendy products recommendations are more important than ‘regular’ recommendations by other customers. Hence,
H5: Presenting recommended trendy products (by the webshop) enhances the attractiveness of a product page more than presenting recommendations based
on other consumers’ prior purchase and browsing patterns.
H6: Presenting recommended trendy products (by the webshop) enhances purchase intention of a visitor more than presenting recommendations based on other
consumers’ prior purchase and browsing patterns.
2.2.4 Size availability table
findings are very old and distribution is heavily improved in the last decades, it shows the importance of quick and easy delivery. Hence,
H7: Presenting a size availability table enhances the attractiveness of a product page. H8: Presenting a size availability table enhances a visitors purchase intention.
2.2.5 Price
A webshop’s product page often shows a price and according to Homburg, Jensen and Hahn (2012), pricing is very important for a company, as it can strongly influence profitability. Wertenbroch and Skiera (2002) underscore the importance of willingness to pay in their paper by stating that companies and researchers need good measures for willingness to pay, as they use it for calculating product demand. Because of the importance of pricing and willingness to pay, it seems interesting to apply a price attribute at the product page.
It might be interesting to test if a minor increase in price is accepted, when the rest of the attributes of the webshop is more attractive. For example, using a model presenting the t-shirt is more expensive than only showing the t-t-shirt. Hence, a consumer’s willingness to pay might vary when using different product page attribute and different prices levels. Willingness to pay is defined as the price which makes the difference for a consumer between purchasing and not purchasing Moorthy, Ratchford and Talukdar (1997). According to Simonson and Drolet (2004), several factors determine willingness to pay. For example, the market price of a product, or the subjective value they estimate the product to be worth. Additionally, an example that underscores the effect of price, is provided by Bagchi and Cheema (2013). They studied the effect of background color on willingness to pay and found that a red background color influences willingness to pay positively in online auctions. This is an example that various factors can influence willingness to pay. Hence, it is assumed that a more attractive product page also influences willingness to pay positively. In short, the willingness to pay corresponds with price, which is an important marketing tool. Although the next hypothesis might be obvious, as people often prefer a low price, using various price levels enables the calculation of equalization prices. These correspond with willingness to pay, as they can be used to determine what an increase in an attribute level is worth. Hence,
Regarding this hypothesis, it is important to keep in mind that the attractiveness of a product page does not only relate to what the page looks like, and its ‘beauty’, but to its overall persuasiveness, as defined in subsection 2.1. Hence, a higher price is expected to decrease this overall persuasiveness, which is similar to a lower ‘attractiveness’ of the page in this study.
Finally, it is expected that price negatively influences the purchase intention as well. However, no hypothesis is formulated concerning price and purchase intention. This allows the current study to emphasize on the other product page attributes, of which little prior research is yet available (contrary to price, of which a lot has been written yet). Furthermore, respondents might focus on price too much in a purchase intention study. Because of this, the other attributes might be overlooked. An example that is not completely related, but shows the huge importance of price, is described by Cunha Jr. and Shulman (2011). According to these authors, consumers may get feelings of regret, when discovering, after a purchase, that an alternative retailer offers the same product at a lower price.
2.3 Moderating the attractiveness of a product page
This subsection discusses four groups of factors which might influence the relations of the antecedents and their relationship with the attractiveness of a product page (or purchase intention).
2.3.1 Consumer demographics
Within consumer demographics, age, gender and education level are included in the study. Based on these demographics, several segments might be distinguished, varying in consumers’ judgments of the attractiveness of the product page. However, for these three variables, no explicit research directions could be found regarding their moderating effect between the dependent variable and its antecedents. Hence, prior research is used to show that the variables are important, but unfortunately, it is not possible to propose solid hypotheses.
Gender. Concerning gender and ‘online’ shopping, Danaher, Mullarkey and Essegaier (2006)
whereas females engage in exploratory behavior and therefore prefer detailed information at the website (Richard, Chebat, Yang and Putrevu, 2010). Considering ‘offline’ shopping, Arnold and Reynolds (2003) created various hedonic shopping segments, largely based on gender and age (e.g. the minimalist segment consists mainly of middle-aged males, who prefer value shopping). Additionally, Fischer and Arnold (1990) showed that females are more involved in Christmas gift shopping. Finally, in the print advertisement context, Peterson and Kerin (1977) showed that males and females differ in perceptions of how appealing advertisements are and Bui, Krishen and Latour (2012) state that it might be wise for retailers to tailor advertising campaigns specifically for males and females.
In short, all these papers show examples that gender affects shopping behavior. For example, as females shop online longer than males and prefer detailed information, they might prefer to see a customer reviews attribute, in which they can read a lot of information. However, too little arguments can be provided to actually propose solid hypotheses regarding gender (as a moderator between de dependent variable and its antecedents).
Age. Young people nowadays grow up with using internet and might purchase products
online more often than elder people. An aged research by Reynolds (1974) concerning catalog shopping, which was considered to be ‘in home shopping’ and congruent at that time, provides findings that young people were more eager to use it, as they were more willing to take risks. Nowadays, these findings of catalog shopping might be compared with findings considering online shopping, which might still be seen by certain age groups as risky and congruent. Especially for clothing, which is an experience product (Ha and Lennon, 2010). Hence, different age groups might react different in a study regarding online purchasing. Therefore it is interesting to include in the current study. For example, perhaps elder people prefer customer reviews, in order to be able to reduce perceived risks. However, as said before, too little arguments could be found to provide solid hypotheses.
Education. In a study concerning health and food, Ma, Ailawadi and Grewal (2013) showed
In short, demographics are potential moderators in the current study. However, please keep in mind that no explicit research directions could be found regarding their moderating effect between the dependent variable and its antecedents. Therefore, no hypotheses are formulated and the findings in the current study are assumed to be the first in this area, as little research is yet available concerning product pages.
2.3.2 Shopping online
Experience with online shopping is the next potential moderator. Its two determinants in the current study are (1) comfort with using internet and (2) preference of buying clothing online.
Insight in the expectation that people with more internet experience differ from ones with less experience can be derived from research by Menon and Kahn (2002). They state that initial internet experience can influence the level of arousal that a visitor of a website experiences. For example, when they arrive for the second time at a product page, they might try to find stimulation from other aspects of the website than at their first visit. Furthermore, in a study by Hernández, Jiménez and Martín (2010), a distinction is made between customers who made their first online purchase and those who made at least one before (these are defined as experienced online customers). Their findings show that within these groups, different perceptions occur, concerning online purchase. For example, people who repurchase online, often feel more confident and experience less risk than people who make their first online purchase. Finally, Cheema and Papatla (2010) show that people who are more experienced in online shopping, think the relative importance of online information sources are less important (for example seeing the product online, or online recommendations by a friend). These authors also found that the importance of these online sources is higher for utilitarian products, than for hedonic products. This is discussed in the next subsection.
2.3.3 Hedonic and utilitarian shopping motives
A consumer always considers some utilitarian and hedonic benefits during a shopping experience (Wang, Baker, Wagner and Wakefield, 2007). People who value utilitarian aspects highly, will for example prefer fast and convenient shopping, whereas hedonism relates to pleasure aspects. An example is provided by Chitturi, Raghunathan and Mahajan (2008):
‘battery life and sound volume are utilitarian benefits of a cell phone, whereas shape and color are hedonic benefits (e.g. aesthetic appeal)’. Hedonic products are emotionally driven,
whereas utilitarian products are cognitively driven (Botti and McGill, 2011). These examples show that a product (or shopping experience) is always utilitarian, or hedonic related. Though, an experience can be both high and low in hedonic and utilitarian benefits (Moore, 2012).
Referring to the current study, hedonic and utilitarian shopping motives might influence a consumer’s perception of the attractiveness of a product page. Here, hedonic characteristics are expected to relate to product pictures and pictorial recommendations, which relates to aesthetic appeal (Raghunathan and Mahajan, 2008). On the contrary, utilitarian characteristics are expected to relate to customer reviews, a size availability table and the price level. These three attributes are more related rationality and focused on completing the shopping task (Batra and Ahtola, 1991). Based on this information, the next hypotheses are formulated.
H10a: The difference in (1) the attractiveness of a product page and (2) purchase intention, between presenting the clothing product by an unknown model and presenting it by a celebrity, is larger when preference for hedonic shopping characteristics is larger.
H10b: The difference in (1) the attractiveness of a product page and (2) purchase intention, between presenting recommended trendy products (by the webshop) and presenting recommendations based on other consumers' prior purchases, is larger when preference for hedonic shopping characteristics is larger.
H11a: The difference in (1) the attractiveness of a product page and (2) purchase intention, between presenting customer reviews and not presenting it, is larger when preference for utilitarian shopping characteristics is larger.
H11c: The effect of price on the effect of the attractiveness of a product page is stronger when the preference for utilitarian shopping characteristics is larger5.
2.3.4 Social influence
The final potential moderator is social influence. Dimensions of social influence are for example (1) normative influence and (2) social comparison. Normative influence is defined as the tendency to confirm to others’ expectations (Burnkrant and Cousineau, 1975). These authors describe that one of the most important factors that affects one’s behavior is the influence from other people. One’s susceptibility for social influence depends for example on his or her intelligence and self-esteem (Bearden, Netemeyer and Teel, 1989). Wooten and Reed II (2004) give a clear summary of a consumer’s susceptibility for normative influence. It is: “the need to identify with others, or enhance one’s image with products and brands, or the
willingness to conform to other’s expectations regarding purchase decisions”. Additionally,
Miniard and Cohen (1983) state that social influence can affect an individual’s decision making.
Social comparison concerns an individual’s behavior, based on the beliefs, thoughts
and expectations of other people (Bearden and Rose, 1990). They describe that sources of social comparison include for example the kind of clothing that an individual wears, based on a comparison with others. This is confirmed by Smeesters et al., (2010), who state that individuals might compare their own attributes and abilities with those of others and this comparison might strongly influence how they think about themselves (positively or negatively).
In conclusion, social influence is an important factor in a consumer’s decision process. It influences one’s decision what clothing to purchase (Bearden and Rose, 1990). Social influence is expected to moderate the relations of product pictures, customer reviews and pictorial recommendations with the two dependent variables. Hence,
H12a: The difference in (1) the attractiveness of a product page and (2) purchase intention, between presenting the clothing product by an unknown model and presenting it by a celebrity, is larger when the susceptibility for social influence is larger.
5
H12b: The difference in (1) the attractiveness of a product page and (2) purchase intention, between not using a customer reviews attribute, and using one, is larger when the susceptibility for social influence is larger.
H12c: The difference in (1) the attractiveness of a product page and (2) purchase intention, between presenting recommended trendy products (by the webshop) and presenting recommendations based on other consumers' prior purchases, is larger when the susceptibility for social influence is larger.
In short, people might be afraid that other people dislike their new clothing purchase. Expected is that recommendations and customer reviews (H12a and c) reduce this uncertainty,
as recommendations prove that certain items are worth buying. Customer reviews provide additional details about the product (e.g. about the fit), which can also help the customer in deciding if he or she likes the product. Finally, using celebrities presenting the clothing (H12b),
might also reduce this uncertainty. For example, a consumer might think the product is good, because a celebrity is presenting it. In this case, one might think: “David Beckham is
presenting this product, therefore it should definitely be a good product!”, as this consumer
might see him as a credible endorser.
2.4 Conceptual model
Conceptual model 1: Overview of the hypotheses concerning attractiveness of the product page
Concerning the second model, as said before, no hypothesis is formulated regarding price and its relation with purchase intention (in the second model).
Conceptual model 2: Overview of the hypotheses concerning purchase intention
3
METHODOLOGY
This chapter describes the current study’s methodology. It is divided in six parts.
3.1 General design
purchase intention questions and (4) moderator questions6. The current study is divided in two sub-studies, namely (1) a conjoint analysis concerning the attractiveness of a product page and (2) the purchase intention derived from the attractiveness of a product page. These two sub-studies are discussed later.
The used clothing product is a basic white t-shirt. The main reason for this choice is to keep the respondents’ attention at the webshop, not at the product itself. An advanced or trendy product might bias the experiment, as attention might be drawn to the product itself, instead of the product page. For example, a respondent might not like a certain type of trendy jeans and might therefore be less reliable in answering the questions in the experiment. Therefore, as a white t-shirt is a product that everybody might need, this bias is probably reduced. It probably does not matter whether one likes the product or not. Finally, the questionnaire is divided in a male and a female version, so that females rate female pictures and males rate male pictures.
3.2 Data collection
According to Hair, Black, Babin and Anderson (2009), the sample size of a conjoint analysis is preferred to be at least 200 for each group in a population. As this is not possible due to time constraints, the sample size in the current study will be approximately 200 in total. To obtain 200 respondents, snowball sampling will be used. Friends, family and acquaintances are asked to fill in the questionnaire and are asked to disseminate it further to some of their friends, family or acquaintances. Online social networks as Facebook are a common tool for spreading nowadays, as people are connected in various networks. To be able to create segments in the conjoint analysis, attempted is to obtain a representation of the real world as much as possible regarding the respondents, but time limits might make this difficult.
3.3 Variables
The product page attributes are the independent variables in both studies, though price is not used in the OLS study (for arguments, please refer to the final paragraph of subsection 2.2.5). The dependent variable in the conjoint study is the attractiveness of a product page, whereas it is purchase intention in the OLS study. In the conjoint analysis the dependent variable is derived from the total utility of the independent variables, which represents the choice of the
6
respondents per rated design. All the variables and their levels are presented in the next overview.
Level Attribute 1 Attribute 2 Attribute 3 Attribute 4 Attribute 5 Pictures Recommendations Reviews Size av. table Price 1 (base) Random model By other customers Do not use Do not use 12,75
2 Celebrity By webshop (trends) Use Use 12,95
3 13,15
Table 1: Overview product page attributes
Examples of all attributes can be found in the questionnaire in appendix B. The product pictures are presented by a random, unknown model and by a celebrity: David Beckham. The female celebrity is Angelina Jolie. Expected is that respondents recognize them, as they both participate in many movies and commercials. The three price levels in this study are based on the actual price of the shirt, which is €12,95 at Wehkamp.nl7.
Next to the attributes, some covariates and potential moderators are studied. The covariates are measured by questions regarding (1) gender, (2) age, (3) education level, (4) comfort with using internet and (5) preference of buying clothing online. These refer to the first five questions in the questionnaire. Three potential moderators are (1) hedonic shopping characteristics, (2) utilitarian shopping characteristics and (3) social influence. The first is measured by four questions (question 19 to 22 in the questionnaire), the second by three questions (question 23 to 26 in the questionnaire) and finally, social influence is measured by the last two questions in the questionnaire. An internal consistency check and factor analysis will test whether these questions actually add up to three distinct constructs.
3.4 Conjoint study: attractiveness of the product page
As described previously, this sub-study uses a conjoint technique for analysis of the data. According to Hair et al., (2009), “conjoint analysis is developed to understand consumers’
reactions to and evaluations of predetermined attribute combinations that represent potential products or services.” Hence, this technique can be used to show respondents various product
page designs and consequently, they are asked to judge which is most attractive. These designs are based on the variables and their levels mentioned in table 1. The perceived
7
attractiveness per design results in a certain level of utility, which represents the respondent’s preference for a design, or for the attributes he prefers most.
The respondents need to answer eight questions in which they have to choose one design from a set of two designs per question. The two design options are the choice sets, which means a choice-based conjoint analysis technique is used (instead of a regular conjoint analysis, where respondents need to judge for example one design per question; Hair et al., 2009). Normally, a ‘no choice option’ can also be used, next to the two design options. This provides an additional level of realism (Hair et al., 2009). However, in the current study, a scenario is created where the respondent already decided to buy the t-shirt. Only a choice which webshop (one of the two designs) is preferred is necessary. Hence a ‘no option’ is excluded from this study.
The optimal number of profiles (product page designs) is based on recommendations by Hair et al. (2009). They state that “the minimal number of profiles equals the number of
parameters to be estimated, calculated as: number of estimated parameters = total number of levels – number of attributes + 1”. Hence, the current study requires seven8 profiles, which is satisfied as it uses 15 profiles9. The compositions of the profiles are created with the program Sawtooth. This program shows the number of profiles to use, the composition of the individual profiles and the sets10, which combination of designs to use, and the number of designs. The output of the program, using two questionnaire versions, with both seven choice tasks and one hold out task, can be found in appendix C. These tasks contain two concepts and the five attributes. The design efficiency is never below 0.95 (rule of thumb), which means that this design is optimal and matches an orthogonal design (Hair et al., 2009). One important issue with creating the designs is keeping the number of questions as low as possible. A short questionnaire might reduce respondent’s fatigue and boredom and therefore enhance their answers (Narayana, 1977). Furthermore, the profiles need to make sense. For example no unacceptable profiles should be included (e.g. profiles which are extremely good or bad), as these will probably always (or never) be chosen.
Finally, as said before, some hold out designs are used, which are used for validation (Hair et al., 2009). These are excluded from the ‘estimation dataset’. This validation sample will be used to predict preference of the profiles and will be compared with the actual responses of the study. The holdout sets both contain two base levels and two second levels
8
2+2+2+2+3 levels = 11 levels – 5 attributes + 1 = 7. 9
Two questionnaire versions with both seven profiles and one holdout set (same for both versions). 10
(vice versa for both sets). This means that the price level is kept constant at €12,95, because, as said before, it might be more interesting to test the effect of the other product page attributes, as price has been researched a lot already. Hair et al. (2009) also state that price is often valued as more important than other attributes.
In short, based on the five attributes, the total utility can be calculated, which represents the attractiveness of the product page in this study and also the part-worths of all the distinct attributes, which are “estimates from the conjoint analysis of the overall preference or utility
associated with each level of each factor used to define the product or service” (Hair et al.,
2009).
3.5 OLS study: purchase intention
The second sub-study tests the purchase intention derived from various product page designs. Here it might be interesting to test which product page attributes affect purchase intention. Price is excluded, as described at subsection 2.2.5. Each respondent is asked to rate five designs with a grade from one to ten. The data is analyzed in SPSS, where purchase intention is the dependent variable. The independent variables are represented by the product page attributes, which are transformed into dummy variables. As all attributes consist of two levels, this means that all (base) levels become zero. Consequently, the second levels become one.
To test the hypotheses regarding purchase intention, the ordinary least squares (OLS) method is used. Three regression formulas are formulated. The first concerns only the product page attributes, which provides first insight and can test hypotheses 2, 4, 6 and 8:
Where,
α = the constant ε = the error term
Briefly explained, for example, attribute 1 concerns product pictures: an unknown model is coded as zero and a celebrity as one. If β1 is a positive number, purchase intention will be
When it is negative, a normal model is preferred. In this case, purchase intention will be lower when a celebrity is presented than when a normal model is presented11. The second model uses all eight covariates (of which the last three are the potential moderators). It derives basic insight of the effect of the covariates. In the final model the interaction effects are added, to see if the covariates actually have a moderating effect at the relationship between the product page attributes and purchase intention. Hence, the second model is the same as the previous one, but adds eight β’s12. These are β5 gender, β6 age, β7 education, β8 E-comfort
(comfortability of using internet), β9 E-purchase (preference for buying clothing online), β10
preference for utilitarian shopping characteristics, β11 preference for hedonic shopping
characteristics and β12 susceptibility to social influence.
The final model adds interaction effects, based on the eight covariates with their interaction effects with the four product page variables13. This leads to 4 * 8 = 32 interaction effects. Add the four product page attributes, the eight covariates and an intercept, this would lead to a model with 45 parameters. This would be possible, in theory, as a rule of thumb states that a minimum of five observations per parameter are necessary (Leeflang, Wittink, Wedel and Naert, 2000). There are 225 observations required (45 * 5) and the dataset has 960 observations available for estimation. Total observations are 1010 (202 respondents * 5 purchase intention observations), but 5% of the observations (50 cases) is used to test the model’s predictive validity. The final model with 45 parameters would be as follows.
As the parameters are closely related (all the product page attributes are multiplied with the covariates), correlations might be high. This might result in high multicollinearity. Multicollinearity is a violation in market model assumptions. Tests regarding violations are
11
e.g. -0.3 * 1, purchase intention decreases with -0.3; -0.3 * 0, purchase intention will not decrease. 12
Assumed is here that factor analysis proves that three constructs exist: hedonic, utilitarian and social influence. 13
necessary to find out if the parameters are reliable (unbiased). Violations are remedied if necessary, as they can lead to two types of problems, namely (1) wrong estimates of the parameters, which can lead to wrong conclusions about their effects, or (2) wrong estimates of the variance of the parameters, which can lead to wrong conclusions about the significance of the effects (Leeflang et al., 2000). Next to multicollinearity, the model will test for heteroskedasticity, normality and outliers. If it seems that heteroskedasticity exist, an additional model might be necessary, where weighted least squares (WLS) is performed. Finally, testing for autocorrelation is not necessary, as the dataset does not contain any time series data. This means autocorrelation is not a problem and GLS is not necessary to remedy (Leeflang et al., 2000).
Finally, the model’s validity will be tested in three ways: face, statistical and predictive validity. Validity is important to check if the model is measured in a reliable way. As described before, the final model is split in two samples, where 95% is used for estimation and the remaining part is used for validation.
3.6 Data limitations
Potential data limitations concern, first, difficulty to obtain a sample which represents a true world population, due to time constraints. This is not necessarily a problem, as the current study provides first insight in the attributes. Second, factor analysis might prove that the three potential moderators do not exist within the current study, or that perhaps (for example) only two or four constructs exist. If this is the case, the number of β’s will change in the previously mentioned equations.
4
ANALYSIS AND RESULTS
This chapter starts with an overview of the characteristics of the respondents. Then, subsequently, the conjoint study and the OLS study are discussed. Both end with an overview of the hypotheses.
4.1 Respondents’ characteristics
Briefly explained, gender consists of slightly more females than males (54.5 versus 45.5%) and the minimum age is 17. The maximum age is 82, whereas the mean age is approximately 29. The age variable is recoded into six classes14, to get more insight in age groups. The 17-27 group is mainly represented, with 67.3%. This probably means that mostly students participated. Education is mainly represented by students of the university, with 40.6%, followed by University of Applied Sciences (30.7%) and lower education (20.3%)15.
Gender Age E-comfort E-purchase
Male Female Min Max Mean Mean Mean
92 (45.5%) 110 (54.5%) 17 82 28.9 5.94 3.67
Age groups
17-27 28-38 39-49 50-60 61-71 72-82
136 (67.3%) 36 (17.8%) 11 (5.4%) 15 (7.4%) 2 (1%) 2 (1%) 100%
Education
Practical edu. VMBO HAVO/VWO MBO HBO WO
5 (2.5%) 7 (3.5%) 5 (2.5%) 41 (20.3%) 62 (30.7%) 82 (40.6%) 100%
Shopping motives (means, on a scale from one to seven)
HEDONIC SOCIAL QUALITY INFORMATIONAL
3.76 3.21 4.05 4.07
Table 2: Respondents’ characteristics
Online shopping experience was represented by two variables; comfortability with using the
internet (E-comfort) and preference of purchasing clothing online (E-purchase). Measured on a scale from one to seven (very uncomfortable / rather offline to very comfortable / rather online), the results show that the respondents are very comfortable with using internet, with an average of almost six. They seem a bit reluctant towards shopping online though, as the average is only 3.67. This is below the average grade of 4. This variable might be influenced quite highly by for example age, as young people grow up with using the internet and might be more eager to shop online. To test if there exist any effect between E-purchase and the other variables, a regression analysis is applied with E-purchase as the criterion variable and the others as predictors. The next table shows that only the variables E-comfort and quality (i.e. shopping online to find better quality) influence E-purchase significantly (p-values of 0.01 and 0.00, both < 0.05). Hence, for every grade (on the one to seven scale) of extra comfortability with using internet, preference for shopping online increases with 0.195.
14
82- 17 = 65. Divided by 6 = approximately 11. Groups are: 17-27; 28-38; 39-49; 50-60; 61-71 and 72-82. 15
Similar, for every grade of extra preference to shop online to obtain better quality, E-purchase increases with 0.452. Furthermore, this model shows an R2 of 0.30, which means that the variance in E-purchase is explained for about 30% by the independent variables.
R2 Model significance E-Comfort Quality
0.30 0.000 (yes) β = 0.195 p-value = 0.01 β = 0.452 p-value 0.00
Table 3: Regression analysis with E-purchase (DV) and the other variables as IV’s.
The last part of table 2 concerns the averages of the four potential moderators. All have a mean closely around four (on the one to seven scales), meaning that respondents, in general, score not very high or low on one of them. Please note that in the literature review three moderators were expected (hedonic, utilitarian and social influence). However, after performing a factor analysis in SPSS16, four potential moderators are found. Here, first, the Keyser-Meyer-Olkin score provided information that a factor analysis is appropriate to use with the chosen variables17 (0.745 > 0.5). Second, Bartlett’s Test Of Sphericity is significant (> 0.05), which means that correlation exists (correlation is desired with factor analysis). Satisfaction of these requirements shows that a factor analysis can be performed. Consequently, the Principal Components Analysis (PCA) is chosen for analysis, as this is the most popular method with factor analysis (Specialization Course Marketing, Marketing Research, Sheets Lecture 5, 2011). The scree plot provides three factors with an Eigenvalue above one, hence, three principal components are usable (the factors). As the normal component matrix does not distinguish the factors very well, varimax rotation is used. Although this rotation still did not provide satisfying results, it did indicate that four (instead of three) factors might be present. Repeating the process with four fixed factors to be extracted, showed that four distinct factors could indeed be derived from the output. This output can be found in appendix D. The factors refer to shopping motivations and are labeled (1) hedonic shopping motivation, (2) social influence, (3) shopping for quality and (4) shopping or browsing to obtain information. The third and fourth factor are sub parts of the utilitarian shopping motivation (for more information concerning the moderators, please refer to chapter 3.3). Finally, in an attempt to gather more evidence that these four factors can be used, they are tested for internal consistency. The next table shows the Cronbach Alphas of the four factors.
16
Factor analysis is used to determine if various variables are correlated and can be combined as factors. 17
