Influence of online store design on
purchase behaviour using click stream data
By
Malou Mulder
S1774255
Msc Business Administration
Title: Influence of online store design on purchase behaviour using click stream data. Author: M.J. (Malou) Mulder
Email: malou.mulder@hotmail.com
Study: Msc Business Administration, specialization Marketing Management. University: Rijksuniversiteit Groningen
Executive summary
Introduction – This study will discuss the influence of online store design on purchase behaviour.
More and more consumers shop online and turnover is increasing. At the same time the growth in number of consumers shopping online is slowing down and the number of online stores is increasing. This causes a challenge for online store owners to attract and keep the consumer. Therefore online store owners should establish an optimal online store design. Until now there is no connection made between online store design and purchase behaviour. Previous studies focused on either predicting purchase or the relation of online store design and purchase likelihood. Studies were not set up to examine the influence of online store design on purchase behaviour, because click stream data lacked detailed information on the visual components of the site (Bucklin & Sismeiro, 2009). The objective of this study is to explore influence of online store design on purchase behaviour and provide the online retailer a guide to designing the most effective online store for their (potential) consumers. Online store design is divided into aesthetics, value and usability (navigation and content). The study aims to answer the question: In what way do the design variables of an online store aesthetics, value and usability influence consumer’s purchase behaviour?
Method - The hypotheses are tested in the context of a commercial online store in the building
materials industry. The experiment is performed as an online A/B/n test to collect data of the consumer behaviour in the online store. A/B/n tests are commonly used in web analytics. Each online store design variable was manipulated in a main page design, which resulted in 5 different main pages. The sample size consists out of 1139 visitors of the online store. The collected data was analyzed using binary logistic regression and linear regression.
Findings - There were no significant effects found of online store design variables (aesthetics, value,
navigation and content) on purchase behaviour. It was found that repeat visitation has a positive influence on purchase behaviour, but does not moderate the effect of online store design variables and purchase behaviour. Control variables showed a significant influence of number of pages visited on purchase.
Conclusion/Recommendations - It can be concluded that online store design variables (aesthetics,
value, navigation and content) do not influence consumer's purchase behaviour. The results did show that repeat visitors increase purchase behaviour compared to new visitors. Thereby the more pages are visited the higher the likelihood that a consumer will make a purchase.
Online store owners should not base the online store design just on their opinion, but should test the different design with an A/B test to get an optimal store. Consumers are more likely to make a purchase when they visit more pages in their visit. Also visiting more pages increases transaction size. This can indicate that consumers want to be well informed before making a purchase. Also online store owners should tempt consumers to revisit the store as repeat visitors are more likely to make a purchase.
Preface
Dear reader,Thank you for taking the time to read my thesis ‘Influence of online store design on purchase behaviour using click-stream data’. It has been year since I started this study in December 2012 and with some delays along the way the thesis has come to an end.
First of all I would like to introduce myself. My name is Malou Mulder and I wrote this thesis for my Master degree in Marketing Management. I did not want to write a thesis just to graduate, I wanted to learn something along the way. During the master there was only a small part focused on online, while in practice marketing is more and more becoming online activities. I decided to write my thesis in the online context, so I would gain knowledge which I did not get in the courses I took.
For the study association MARUG (Marketing Association University of Groningen) I implemented a new design for the association website and I was wondering how it impacted the consumer. Thereby I sell bags on an auction site and was testing which content attracted more reactions on my advertisements. Taking these interests together led to the topic of online store design.
To test the influence of online store design on purchase behaviour I used A/B testing, which is a tool through which you can test different Web site pages at once to see which performs better on several metrics. Before I started this thesis I had never heard of A/B testing, but I wanted to use a method other than a questionnaire. After gaining knowledge about and working with A/B testing I belief that it is a valuable tool to use for implementing effective online activities.
That is also a goal of the thesis. How an optimal online store can be designed. Therefore a target group for this thesis are managers of online stores or online store owners who are interested to get the most out of the online store. The other target group are scientists who are interested in the influence of online store design on purchase behaviour. As previous literature did not focus on purchase behaviour, but purchase intention.
I would like to thank a few people for their contribution to the thesis. The first person is dr. Erjen van Nierop for guiding me through the process, being patient when I postponed my thesis and for being enthusiastic about my method giving me motivation to make it work. Second I would like to thank Hilgo Gabriëls and the online store for giving me the opportunity to test my assumptions in an online store and helping me with the technique to set up the A/B test. The third person is Tim Jansen for implementing the codes needed to implement the A/B test and succeeding in modifying the Google Analytics code with my deficient explanation on how to make it work and making it able to track individual behaviour. Last I would like to thank Jacob Wiebenga for taking the time to read my thesis and being my second supervisor.
Kind regards, Malou Mulder
Table of content
Chapter 1: Introduction & Research Question ... 6
1.1 Online shopping in the Netherlands... 6
1.2 Causes of growth in online shopping ... 7
1.3 Need to adapt to the growth of the market ... 9
Chapter 2: Literature Review ... 12
2.1 Effectiveness of online stores ... 12
2.2 Purchase behaviour in online stores ... 12
2.3 Variables of online store design ... 14
2.3.1 Aesthetics in online store design ... 16
2.3.2 Value in online store design ... 17
2.3.3 Usability in online store design ... 18
2.4 Moderator of the influence of online store design on purchase behaviour ... 21
2.5 Conceptual model ... 22
Chapter 3: Methodology & Research Design ... 23
3.1 Research design ... 23 3.2 Data collection ... 24 3.3 Plan of analysis ... 27 Chapter 4: Results ... 31 4.1 Descriptive statistics ... 31 4.2 Regression analysis ... 33 4.3 Control variables ... 38 Chapter 5: Discussion ... 41
5.1 Online store design variables ... 41
5.2 Repeat visitation ... 43
Chapter 6: Conclusion and Recommendations ... 44
6.1 Conclusion ... 44
6.2 Recommendations... 45
Reference list ... 47
Appendix I ... 53
5 6,3 7,4 8,2 9 9,8 0 2 4 6 8 10 12 2007 2008 2009 2010 2011 2012 x B ill io n s
Chapter 1: Introduction & Research Question
Consumers have access to the Internet since the early nineties, in the past 20 years the online environment has taken a prominent place in many consumer lives. In the Netherlands 96% of all consumers between 12 and 75 years old use the Internet , which is a total of 12,4 million users (CBS, 2012). A great deal of activities has moved from offline to online, such as keeping up to date with news and information, keeping in touch with friends, and shopping. The first possibility to shop online was already in 1994 when Pizza Hut provided the option to order your pizza online. In the past years online shopping has risen tremendously for both the consumer and retailer side. Consumers have a declining interest in traditional shopping formats and want focus on efficient use of time (Eroglu, Machleit & Davis, 2001). This growing popularity of online shopping, has made it commonplace in countries as the United States, Germany, United Kingdom and the Netherlands. According to Ed Nijpels, chairman of thuiswinkel.org ‘online shopping has become a meaningful factor in the economic landscape’ (thuiswinkel.org, 2012).
Figure 1.1: Turnover online shopping 2007-2012 (Blauw Research, 2013).
1.1 Online shopping in the Netherlands
1.1.1 Online shopping per age category
Not only turnover is growing in online shopping. Also the total number of consumers who shop online has increased from 9,6 million consumers in 2011 to 9,8 million consumers in 2012 (+2,3%) (CBS, 2013). Thereby the number of orders is growing. The number of orders per consumer has increased with 8% to a total of 8,3 order per consumer (Blauw Research, 2013). In total 88 million orders were made in 2012 (Blauw Research, 2013).
In 2012 three-quarter of all Dutch consumers (above the age of 12) has made an purchase online (StatLine, 2012). The first experience with purchasing online is usually in the segments clothing, toys or computers and accessories (Blauw Research, 2013). The highest percentage of online shoppers are in the age group 25-45 years old (see figure 1.2), although the online shoppers in this age group seem to have stabalized in the past three years. The percentage of consumers shopping online is still increasing in the other age groups, especially the age group 65-75 years old.
Figure 1.2: Percentage of consumers shopping online classified by age (StatLine, 2012). 1.1.2 Online shopping in the future
Although in 3 of the 4 age groups the percentage of consumers who shop online is still increasing, the total growth is slowing down. In 2010 the growth in number of online shoppers was 6,2%, this percentage was almost cut in half to 3,3% growth in 2011 (StatLine, 2012). In 2012 the growth slowed further down to 2,3% (CBS, 2013). In the near future the number of consumers who shop online is going to reach its limit and future growth of the online shopping market relies on number of orders and the amount spent online. According to Thuiswinkel.org (2013) the turnover of online purchases will grow to 27 billion euros in 2020.
1.2 Causes of growth in online shopping
The (expected) growth in online shopping turnover can be explained by many different reasons. First, consumers shop online for convenience in time, place and value (Gill, 2012; Eroglu, Machleit & Davis, 2001). They want to find the best deals (Gill, 2012). Second, according to Top Executive Paul Nijhof from Wehkamp.nl, the largest online retailer in the Netherlands, consumers increasingly prefer to shop online, because everything they desire is only a mouse click away and they can save on parking money (Hafkamp, 2012). Other advantages which can explain that consumers find their way to online
0 10 20 30 40 50 60 70 80 90 100
12-25 years 25-45 years 45-65 years 65-75 years
stores more often are the increasing supply, sharp prices, the confidence consumers have in online shopping, growth of the mobile channel and good online marketing. Consumers preferred to shop in brick-and-mortar stores, because the product could be taken home immediately. Nowadays many consumers see the speed of the delivery as an advantage to shop online. Some online stores guarantee that orders made before 10pm are delivered the next day. This development also contributes to further growth of online shopping.
Another factor that might contribute to the growth in online shopping is the fact that the lines between online and brick-and-mortar retailing are blurring (Gill, 2012). 43% of consumer search for information online before buying in any channel and almost a quarter of the consumers use the knowledge they gain offline for their online buying behaviour, such as trying on products in a brick-and-mortar store before purchasing them online (Gill, 2012).
1.2.1 Growth in online segments
When dividing the growth in turnover into online segments, it becomes clear that traveling is the largest online segment with a turnover of 3,8 billion euros in 2012. Telecom comes in second with a turnover of 1,25 billion, followed by clothing with a turnover of 730 million (Blauw Research, 2013). Looking to the segments in a broader perspective shows that the turnover of online products is growing faster than that of online services (Blauw Research, 2013). A possible explanation for this difference in growth might be the economic crisis affecting expenditures on traveling and visiting concerts.
Another explanation might be that large online clothing stores invest heavily in both online and offline advertising, they want to attract consumers and gain their trust. However unintentionally smaller online clothing stores profit from these high investments (Bartelds, 2012). So, the investments in advertising benefits the entire online clothing segment.
1.2.2 Growth in number of online stores
12500 15000 17500 20000 30000 37500 0 5000 10000 15000 20000 25000 30000 35000 40000 2006 2007 2008 2009 2010 2011
Figure 1.3: Number of online stores B2C (thuiswinkel.org, 2012)
1.3 Need to adapt to the growth of the market
One of the hardest challenges for an online retailer is to keep the customer. Consumers can easily switch stores, the next online store is only a mouse click away. In comparison with brick-and-mortar stores the barrier to exit the online store is much lower. The costs to visit an online store are also low, making it easy for consumers to shop around (Moe & Fader, 2004a). Due to the large growth in number of online stores, the consumer has even more choice to shop around. It is becoming more challenging for online stores to convert consumers into buyers.
Consumers form their impression of an online store In the first 50 milliseconds of viewing a page (Gofman, Moskowitz & Mets, 2009). As a result of the low exit barrier, consumer can quickly bounce back when the impression is not as expected and shop further in the next online store. Records show that almost 50% of consumers fail to go beyond the first page they see and explore the online store further (Gofman, Moskowitz & Mets, 2009).
An online retailer needs to establish an optimal online store to cope with the challenges of the low exit barriers, the growth in number of online stores and the stabilisation of the number of consumers purchasing online. Empirical evidence indicates that many online retailers do not completely understand the needs and behaviour of the online consumer (Lee, 2002). Optimizing the store based on those needs can increase sales with 50% to 200%, with examples of even 600% (King, 2008). So, to establish an optimal online store online retailers need to understand what the consumer wants. 1.3.1 Consumer behaviour in online stores
appearance of the first page. For online retailers to understand what the consumer wants and needs, they need to understand how the appearance of the first page influences consumer’s behaviour. In past research it was stated that certain atmospheric qualities in online stores are likely to affect the use and results (satisfaction, amount purchased and time spent) of consumers (Eroglu, Machleit & Davis, 2001). However this study did not prove the statement with empirical results and it is questionable whether store atmosphere is a good term to use in online store environment. Although, to a very large extent, online consumer behaviour can be studied using theory from offline consumer behaviour, store atmosphere cannot be transferred directly to the online environment. A consumer does not go physically to the online store, but stays in his/her own atmosphere (or environment) in front of the computer. The general appearance of the online store is therefore better explained by the design (use of colour, lay-out, navigation, etc.) than by store atmosphere. 1.3.2 Past research on consumer behaviour in online stores
The influence of online store design on consumer behaviour has not been researched extensively. Recent literature focused on either predicting purchasing behaviour (Bucklin & Sismeiro, 2003; Sismeiro & Bucklin, 2004) or researching the effect of online store design on purchase intentions (Resnick & Montania, 2003; Griffith, 2005; Richard, 2005; Richard & Chandra, 2005; Schlosser, White & Lloyd, 2006; Hausman & Siekpe, 2009). The connection between the online store design variables and purchase behaviour has not been made until this point. One reason why previous studies were not specifically set up to examine the influence of online store design on purchase behaviour is that clickstream data lacked detailed information on the visual components of the site (Bucklin & Sismeiro, 2009).
This study will make that connection between online store design and purchase behaviour with the use of clickstream data. Specifically the online store design which is visible on the main page, since consumer base their impression on the main page which leads to the decision to browse further and possibly make a purchase. In this study the online store design is divided into three variables: aesthetics, value and usability. Aesthetics is defined as ‘an artistically beautiful or pleasing appearance’ (The American Heritage Dictionary of the English Language). Value is defined as the extent to which the online store is unique, specific, and relevant to the consumer. Usability relates to the arrangement of the online store and the ability to facilitate consumer goals (Eroglu, Machleit & Davis, 2003). For online retailers this study will give insights into consumer behaviour and how to optimize their online store in order to attract and keep the consumer in their store. In that way they can face the challenge of today and in the future.
So, the following research question is proposed:
In what way do the design variables of an online store aesthetics, value and usability influence consumer’s purchase behaviour?
Sub questions:
1. What are the elements of aesthetics, value and usability on the main page of an online store? 2. What are the elements of purchase behaviour?
Chapter 2: Literature Review
In this chapter the online store design variables will be discussed and purchase behaviour will be defined. When the variables are discussed, hypotheses about the relation between the online store design and purchase behaviour are formed.
2.1 Effectiveness of online stores
Online stores contain characteristics of an information system as well as a marketing channel. The effective manipulation of an online store is considered to be a strategic marketing tool (Hausman & Siekpe, 2009). To use this tool in its full potential, the online store should be the result of a systematic and conscious design (Manganari, Siomkos & Vrechopoulos, 2009). A conscious design can be established by observing the consumer. Which is quite different from observing them in traditional brick-and-mortar stores, because the consumers are not physically present in online stores. They remain in their own environment. Converting a visitor into a buyer or persuade them to continue browsing is quite difficult, as many consumers bounce back from the first page they visit. The conversion rate, the ratio from visitors of the first page to purchasers, is often cited as being in the low single digits (Gofman, Moskowitz & Mets, 2009). In 2011 Google published a conversion benchmark based on the average of Google analytics. According to that benchmark the average conversion rate in the Netherlands is 1,5% (8020 ecommerce, 2012)
To effectively use the online store as a strategic marketing tool Chen, Gillenson and Sherrell (2002) came up with four guidelines:
Make users feel comfortable Create sites that are fun to use
Tempt consumers to spend more time and revisit Increase likelihood of a purchase.
The first two guidelines are ways which can lead to the last two guidelines. Also the third guideline is leading to the end goal of a purchase. Without consumers who purchase products or services an online store cannot exist. Therefore, in the latter of this chapter the focus will be on the last guideline, increase the likelihood of a purchase.
2.2 Purchase behaviour in online stores
When an effective online store is established it is likely that the likelihood of a purchase should be increased (Chen, et al., 2002). It has been shown in the past that consumers who report intentions to purchase a product possess higher actual buying rates than consumers who report that they have no intention to purchase (Berkman & Gilson, 1978). Purchase intention, however, does not equal purchase behaviour. To increase the likelihood of a purchase, an online retailer should know what influences purchase behaviour. Based on that an online store can be optimized and that effective store increases the likelihood of a purchase.
time. Consumers can directly request the needed information after the ad exposure and order the product via the Internet.
The guidelines by Chen et al. (2002) only stated what online retailers should strive for, they failed to guide online retailers how to execute those guidelines. To execute those guidelines online retailer should understand the behaviour of the consumer in an online store. Past research observed and modelled consumer’s behaviour in the decision to continue browsing [entice consumers to spend more time and revisit] (Bucklin & Sismeiro, 2003), the impact of cues present in an online store [make users feel comfortable](Richard, 2005) and the (pre-)purchase intentions [increase likelihood of a purchase](Richard & Chandra, 2005; Sismeiro & Bucklin, 2004).
However only predicting purchase behaviour lacks the specificity necessary to indicate which variables of the design influence purchase behaviour. The design is not purely an aesthetic function, it is an indicator for consumers that an online store can be trusted, which is the most significant driver of online purchase intentions (Schlosser, White & Lloyd, 2006). Therefore online store design is an important influencer of purchase behaviour. When there is no trust, there is no purchase. Many failures of online store’s goals reflect ineffective online store design (Hausman & Siekpe, 2009). Before identifying the variables of online store design, the variables of purchasing behaviour will be discussed.
2.2.1 Variables of purchase behaviour
In this study consumers purchase behaviour is captured by two variables: purchase and transaction size. Purchase can be defined as whether the consumer makes a purchase or not and transaction size as the amount spent in a purchase. It is important not to rely solely on purchase and also include transaction size to capture the entire value contribution to a company (Liu, 2007). Thereby is it appropriate to research purchase and transaction size separately. A company can have high fulfilment costs per order (e.g., due to payment processing or shipping costs) which makes the size of an order important for determining the profit margin (Liu, 2007). On the other hand, a company might consider purchase as an important indicator for future purchase behaviour (Liu, 2007). Using both purchase and transaction size as variables of purchase behaviour allows for a more accurate assessment of the influence of online store design and makes it more appropriate to generalize the conclusions of this assessment.
2.2.2 Influence of online store design on purchase behaviour
In recent online store design literature there has been a focus on two different streams of influences on purchase behaviour. One stream tries to improve purchase behaviour by focusing on the use of single indicators in the online store design (Resnick & Montania, 2003; Schlosser, White & Lloyd, 2006), such as indicators for privacy and product quality. These indicators can positively influence purchase behaviour, but consumers are more likely to shop and return to well-designed Web sites (Liang & Lai, 2002). Therefore this research focuses on the other stream, which examines the influence of the entire design of the online store.
simulation conducted they found that both human and computer factors were necessary to create positive online shopping intentions.
Richard (2005) used a different description and divided the online store design variables into high task-relevant and low task-relevant cues. Where high task-relevant cues were based on navigation, structure and content, low task-relevant cues related to entertainment (Richard, 2005). The results deviate slightly from Hausman & Siekpe (2009), they show a direct, positive link between both low and high task-relevant cues and purchase intention. The consumers who liked the low task-relevant cues were labelled as browsers, which have less interest in actually purchasing products.
2.3 Variables of online store design
The way the online store is designed, is the way consumers experience the online store. An optimal online store consists of characteristics of both the information system and the marketing channel to ensure that the online store provides the required elements (Hausman & Siekpe, 2009). The assessment of the online store relies entirely on the elements included in the online store design, because consumers do not have access to physical elements to assess the store. In order to establish an optimal design of an online store, it should include organization, presentation and interactivity (Shneiderman, 1998). Constantinides (2004) suggested that the website experience is composed of five main elements: usability, interactivity, trust, aesthetics and the marketing mix.
Soonsawad (2013) further researched and developed those elements and compared them with success cases from practice. This resulted in the Conversion rate optimization framework (see figure 2.1.), which consists out of seven elements: Catalyst, aesthetics, value, usability, persuasion, trust and interactivity. The framework shows how online retailers can apply these elements to each stage of the customer decision process (need recognition, information search, evaluation, purchase and post-purchase) to deliver an online store experience which is likely to maximize online visitors to become purchasers and reinforcing post-purchase decisions (such as future purchase decisions). Each element of the framework can be used to decide how the value of the online store can be increased for the consumer (Soonsawad, 2013). The conversion rate optimization depends on how well those elements are implemented to serve an efficient website experience to the consumers (Soonsawad, 2013). Each element is important in different phases in the consumer decision process as can be seen in figure 2.1. For example trust, which includes transaction security, customer data safety and guarantees/return policies, is of importance in the decision stages evaluation, purchase and post-purchase. When trust is shown in the online store at the point when consumers evaluate if they want to buy the products at the online store, when they actually purchase the products and when they for example want to return an item after purchasing (post-purchase), consumers are more likely to take action and become (repeat) buyers.
the consumers search products/services comprehensively (Soonsawad, 2013). Consumers base their initial perceptions of an online store on the elements on the main page (Resnick & Montania, 2003).
Figure 2.1. Conversion rate optimization framework (Soonsawad, 2013)
In the conversion rate optimization framework the early stages of the customer decision process correspond with three variables that should be present in those stages. Aesthetics, value and usability. It also includes a variable which is not visible on the main page, but is determined by the consumer: Catalyst. The three remaining elements (interactivity, trust and persuasion) will not be taken into account in this study, since they are not of importance in the early stages of the customer decision process. Thereby previous studies found that these remaining elements are not important in the early stages of the customer decision process and on the main page of the online store. First, Constantinides & Geurts (2005) found that interactivity is not important on the main page of the online store, because interactivity did not have a significant impact on the choice of vendor. It becomes more important in a later phase. Second, consumers indicated that trust was an important factor in online shopping, though their behaviour showed that it was not the case (Constantinides & Geurts, 2005). Third, persuasion is guided by clarity, which emphasizes on the online store’s value proposition with pictures, text, layout and design (Soonsawad, 2013). This is established in the early stages by value and aesthetics.
2.3.1 Aesthetics in online store design
According to Valacich et al. (2007) consumers do not only visit online stores to search for certain products or services, they also want to have a rich and entertaining experience. The term aesthetics has a broad range of definitions in marketing theory and beyond. In this study aesthetics will be defined as ‘an artistically beautiful or pleasing appearance’ (The American Heritage Dictionary of the English Language). Apart from the broad range of definitions, defining what an aesthetic object is can be very subjective. However among researchers there is often a general agreement what represents an aesthetic element (Tractinsky, 1997; Veryzer & Hutchinson, 1998). In this study aesthetics refer to the design of the online store that is visually attractive (Cai & Xu, 2011) and creates a pleasing appearance, such as proportion, colour, shape and size (Bloch, et al., 2003).
2.3.1.1 Elements of aesthetics in online stores
Apart from the recognition that aesthetics are meaningful to online stores, it should also be applied. To apply aesthetics to the structure and perception of the online stores the basic building blocks should be included colour, brightness, movement and shape (Schroeder, 2006).
Colour impacts the evaluation of consumers of a website (Gorn , et al., 2004; Resnick & Montania, 2003). Simpler webpage backgrounds are in general more effective and appealing than more complex ones (Stevenson, Bruner & Kumar, 2000). However, Martin, Sherrard and Wentzel (2005) stated that medium complexity was more favourable than low or high. On the other hand, Bruner and Kumar (2000) found no direct effect at all of website complexity on consumers’ responses. These specific elements can play an important role in increasing usage (van der Heijden, 2004). However, focusing solely on a specific design element is not enough. The separate elements should be taken together to determine the visualisation of the online store.
2.3.1.2 Influence of aesthetics
Many have argued that the online environment puts consumers in control of the information they receive (Mandel & Johnson, 2002). Thereby it is speculated that the online environment induces a state of flow that diminishes or eliminates effects of unrelated stimuli, such as aesthetics (Hoffman & Novak, 1996). Surprisingly content and aesthetics are almost equally important to the consumer. This was found in a study by Fogg, Soohoo and Danielson (2002) which stated that close to 50% of all consumers paid far more attention to ‘shallow’ aspects of an online store, than to the provided content. The aesthetics are meaningful to the success of online stores, it attracts consumer’s attention, also on a subliminal level (Mandel & Johnson, 2002; Tractinsky & Lowengart, 2007; van Schaik & Ling, 2009). Aesthetics create a first impression which results, when positive of course, in a desire to explore the online store further (Cai & Xu, 2011). When aesthetics are attractive, consumers are more likely to complete their tasks, less likely to browse other sites and it impacts their shopping behaviour (Mandel & Johnson, 2002; Richard, 2005; Rosen & Purinton, 2004), because aesthetics make sure that an online store is arousing and appealing. In addition, Mandel and Johnson (2002) demonstrate that the shopping behaviour of both the experienced online shopping consumer and the inexperienced consumer can be influenced by aesthetics.
Hypothesis
minor details in the design, like the background used can influence purchase behaviour. When consumers are primed on a particular attribute by showing this in the background of the online store, consumers are more likely to prefer the product that excelled on that attribute than any other in the product range (Mandel & Johnson, 2002). The study Zhang and von Dran (2000) confirmed that certain aesthetic elements influence purchase behaviour. According to Schlosser, White & Lloyd (2006) the entire aesthetic design influences purchase intentions, when an online store looked like it made a high investment in the visuals consumers intention to purchase was higher. The aesthetic design includes colour, brightness, movement and shape. When aesthetics are used in the right way it has an positive influence on purchase intentions (Richard, 2005; Hausman & Siekpe, 2009), therefore it is expected that,
H1: The extent to which aesthetics on the main page are attractive has a positive influence on consumer’s purchase behaviour ((a)purchase and (b) transaction size).
2.3.2 Value in online store design
The third variable which influences the early stages of the decision process is value. Value means being unique, specific and relevant (Soonsawad, 2013). Therefore the variable can be defined as the extent to which the online store is unique, specific, and relevant to the consumer. The consumer plays an important role in value, because it is based on the perception of the consumer (Soonsawad, 2013) and what the online store wants them to perceive. Value can be applied on two levels, the company and product level.
Value at the company level relates to consumer’s understanding of an online store’s uniqueness and relevance. The question to keep in mind at this level is ‘Why should the consumer buy from this online store rather than any of the other online stores?’. What separates this online store from the other competing online stores.
The product level is related to customer perception about the reasons for buying products from the online store. It is imperative that the company applies both levels of value simultaneously. At the product level the question is ‘Why should the consumer buy this product rather than any other product?’. What separates this product from other products.
Value is also referred to as the online marketing mix (Soonsawad, 2013), which includes fulfilment, product, price and promotion (Constantinides, 2004). Fulfilment refers to the way online vendors follow up orders and deliver products, which has an immediate impact on the willingness of customers to order and return to the online store (Constantinides, 2004). Elements which are included in fulfilment are alternative payment methods, speed of delivery, flexibility of deliver and order tracking (Constantinides, 2004). Where price is often referred to as being the main motivator for consumers when choosing a particular Web site, research contradicts this predominant belief (Constantinides, 2004). Price does not seem to be the main motivator for the consumer when choosing an online store. Promotion includes elements as free extra services, sales promotions or incentive programs (Constantinides, 2004).
Hypothesis
Communicating how unique, specific and relevant the store and products are can possibly compensate the lack of physical contact with the store and products (Constantinides, 2004). Thereby consumers will want to take action (Burnstein, 2012), communicating value effectively can increase the purchase behaviour of the consumer. However, the right words or phrases should be chosen, because it can make all the difference in the effect (Soonsawad, 2013). So, when the right words or phrases are chosen to communicate the value of the online store and products, the probability that a consumer will purchase from the store will increase. Therefore it is expected that,
H2: The extent to which the main page is of high value has a positive influence on consumer’s purchase behaviour ((a) purchase and (b) transaction size).
2.3.3 Usability in online store design
Usability is an important variable in online design. The variable contributes to the quality associated with the design (Bevan, 2001). When usability is high, consumers belief the company understands, respects and cares for them (Egger, 2001). Previous literature identified different definitions for usability. Bevan (2001) defined usability as a set of attributes that bear on the effort needed for use, and on the individual assessment of such use, by a stated or implied set of users. However, Eroglu, Machleit and Davis (2003) defined usability as the arrangement of the online store and the ability to facilitate consumer goals. Combining both definitions, the definition of usability in this paper is set. Usability is the arrangement of a set of attributes in the online store that has the ability to facilitate consumers goals. When the attributes of the design of the online store are arranged easy and efficient, consumers have the possibility to access the relevant information they need. This enables consumers to explore the online store further.
The Human-Computer Interaction literature focused on explaining the failure or success of online stores by the usability of the store (Petre, Minocha & Roberts, 2006). When online retailers do not pay attention to usability, it can lead to severe problems for their online store. Even minor design failures can lead to big problems for the consumer (Konradt, et al., 2003). Poor usability may not only have a negative influence on consumer convenience while shopping online, it may also misinform consumers of the products and services provided (Konradt et al., 2003). It can even discourage them to use the online store (DeLone & MacLean, 2004). Usability can benefit the online store when applied properly. The variable usability can be divided into different usability elements. These elements relate to common usability problems, such as use of ambiguous terminology, confusing navigation or slowly responding servers. The usability elements are determined by Nielsen (2000) as the following: (1) navigation, (2) response time and (3) content. Nielsen (2000) also indicated credibility as one of the usability elements, however credibility is based on consumer evaluations and is beyond the scope of this study.
2.3.3.1 Navigation
identify the appearance of the store and understand how the store is organized, categorized and has arranged its merchandise (Titus & Everett, 1995).
The consumer has the opportunity to self-direct their movement, when they recognize how products are clustered by their common characteristics or through orientation aids (Childers, et al., 2001). The self-directed movement of navigation is a challenge for marketers. On the one hand, the consumer wants an online store that permits great freedom of choice. On the other hand the marketer wants to control the way the consumer navigates through the online store. Marketers want this for two reasons. First, they want to ensure the consistency in the marketing message that is shown to the consumer (Hanson, 2000). The second reason might be that marketers want to control the navigation of the consumer in order to increase advertising revenues (Dailey, 2004). Many websites rely on different forms of advertising (banner ads, sponsorships, etc.) to generate revenue (Hanson, 2000). The latter reason is less applicable to online stores. The main goals of an online store is to keep consumers in their store and generate revenues by selling items, instead of generating revenues by letting consumers leave the online store by clicking on an ad. The two contradicting issues of freedom of choice for the consumer and level of control for the marketer might result in negative outcomes. Consumers can react in a negative way when their navigation possibilities are too controlled, it might even infer negative reactions towards the online store (Dailey, 2004).
Navigation cues are specific indicators of how the online store is organized, categorized and has arranged its merchandise. Some of those cues are text and icon links (Hoffman & Novak, 1996; Zhang & von Dran, 2000), which are intended to enhance usability (Song & Zinkhan, 2003). Text and icon links should be clear to the consumer, so they can understand easily how the categories in the online store are clustered and where certain products are located. Other navigation cues as identified by Dailey (2004) are ‘next’ and ‘previous’ links and navigation bars. ‘Next’ and ‘previous’ links are linear navigation cues and very restrictive in nature (Hoffman & Novak, 1996). A navigation bar is a set of links that are listed on each web page that the consumer can click on in order to view pages of interest to them (Dailey, 2004). The various cues offer consumers various levels of control and freedom of choice of online store navigation (Dailey, 2004). The ease of use of the interface is an element that has effects on the consumer behaviour (Venkatesh & Davis, 2000).
The navigation element is a must to meet the expectation of the consumer. According to Valacich et al. (2007) any successful website should fulfil the basic needs of the consumer, which is called the zone of tolerance and consists of hygiene factors (those features that make a Web site useful and serviceable and whose absence will make the online user dissatisfied). Navigation is one of those factors and leads to functional convenience. High levels of ease of navigation leads to an efficient online store.
Hypothesis
positive effect on purchase intentions. Therefore it is hypothesized that if the navigational cues are more easy to use, it is more likely that consumers will make a purchase.
H3: The extent to which navigational cues on the main page are clear to the consumer has a positive influence on consumer’s purchase behaviour ((a) purchase and (b) transaction size).
2.3.3.2 Response time
The second element as indicated by Nielsen (2000) is response time, which can be defined as the time between consumers input and system output (Konradt, et al., 2003). Consumers hold a minimum expectation for response time. When the time to respond takes longer than consumers’ minimum expectation level, it will lead to dissatisfaction and prohibit usage (Huang & Fu, 2009). Ten years ago response time was very important, because of low Internet speed. Today Internet speed is much faster, making response time less of an issue. Therefore, response time will not be taken into consideration in this research.
2.3.3.3 Content
Without content, the last element of usability, an online store cannot function. It is one of the main principles contributing to repeat visits (Rosen & Purinton, 2004). It can even form the core business in a business model of an online store (Molla & Licker, 2001). Content is referred to as the characteristics and presentation of information in the e-commerce system (Zhang, Keeling & Pavur, 2000). Online stores without content are simply valueless, content is a source of value (Hartman, Sifonis & Kador, 2000). Creating valued content is not as easy as it seems, it requires truly expert skills. Only transferring content from traditional media such as catalogues usually does not work because often it fails to create interaction with the consumer (Parsons, Zeisser & Waitman, 1998). It is key to get consumer engaged in the online store. Some online stores exhibit content which is uninspiring or poorly presented, while other stores are so aesthetic or graphic that it is simply too time consuming to browse through the store. The key to engaging consumers is twofold: The first is to provide content that is valuable to consumers (the substance) (Parsons, et al., 1998). For the content to be valuable it should be personalized, complete, relevant, easy to understand and secure to infer transactions via the online store and return to the store regularly (DeLone & MacLean, 2004). In order to get consumers engaged, the content must be continuously renewed to stay ‘fresh’ or provide content that is changing on an on-going basis (Parsons, et al., 1998). On top of those prerequisites, all content should be of high quality. The quality of the content can impact the success of the online store and determine if a consumer will stay in the online store or abandons it (Chen & Wells, 1999; Molla & Licker, 2001). The second half of the key to engage consumer is mastering creative programming for interactive media (the form) (Parsons, et al., 1998).
Hypothesis
is effective it should contain high quality. Highly qualitative content facilitates purchase (Rosen & Purinton, 2004), therefore the following hypothesis is formed,
H4: The extent to which the content on the main page is of high quality has a positive influence on consumer's purchase behaviour ((a) purchase and (b) transaction size).
2.3.3.4 Conclusion on usability
Usability is an important part of online store design, an easy to use online store will attract more consumers and engage them to stay longer in the online store and make repeat visits. Even inexperienced online shoppers are aware that content quality and ease of use is a part of the design (Rosen & Purinton, 2004). In the past research already showed that usability has a positive influence on purchase intentions (Richard, 2005; Hausman & Siekpe, 2009). Konradt, et al. (2003) stated that usability is a valid predictor of users decision to buy. Usability does have a wide range of implications, reaching from navigation to content. The two hypothesized variables are different aspects of usability and are therefore taken separately in the hypotheses.
2.4 Moderator of the influence of online store design on purchase behaviour
In the previous paragraphs it was hypothesized that the online store design variables have a direct influence on purchase behaviour. This paragraph shows that this relationship can be moderated by repeat visitation.
2.4.1 Repeat visitation
Based on the number of times a consumer has visited the store, two types of visitors can be identified: new visitors and repeat visitors. New visitors are consumers visiting the online store for the first time and repeat visitors are consumer which have already visited the store before the current visit. Where the online store is completely new for new visitors, repeat visitors probably understand how to use the online store and can use the knowledge they acquired in previous visit(s) for subsequent visits (Bucklin & Sismeiro, 2003). Before consumers visit the online store they already differ based on experience with the online store. Before consumers visit the online store they already differ based on experience with the online store. Research shows that new visitors browse a Web site differently from repeat visitors (Bucklin & Sismeiro, 2003; Johnson, Bellman and Lohse, 2003). 2.4.2 Influence of repeat visitation on purchase behaviour
Hypothesis
There are differences in the behaviour in online stores between new and repeat visitors. For new visitors it is the first time to visit the online store, they are trying to understand the online store by exploring it. Repeat visitors are likely to already understand the online store, have more experience and act more goal oriented. The mix of new and repeat visitors could send a false signal of change in important Web site metrics (Bucklin & Sismeiro, 2003), like purchase behaviour. Since repeat visitation, although not per se diagnostic (Sismeiro & Bucklin, 2004), has a positive influence on purchase likelihood (Moe & Fader, 2004b). According the Bucklin & Sismeiro (2003) repeat visitation (new versus repeat visitors) should be taken in to account when analysing changes in site performance. Therefore, it is hypothesized that
H5: Repeat visitation has a positive influence on the effect of aesthetics (H1), value (H2), navigation (H3) and content (H4) on consumer's purchase behaviour ((a) purchase and (b) transaction size).
2.5 Conceptual model
To summarize all hypotheses a conceptual model is developed (see figure 2.2).
Chapter 3: Methodology & Research Design
The hypotheses formed in chapter 2 are tested in the context of a commercial online store in the building materials industry. This online store was established 10 years ago and the store is continuing to grow. One category of building materials is sold and appurtenance directly related to the installation of the building material. The online store sells to both consumers and business, each groups accounts for 50% of the total customers. However, most business customers order by telephone or personal contact, so online the percentage of consumers is higher. A one category shop is chosen, so the consumers are all searching for the same product and have the same search intention. Due to the current economic crisis the entire construction and building industry is under pressure, therefore it is important for online stores which sell building materials to be as effective as possible. The selected online store can be more effective, currently its conversion rate (visitor-to-buyer ratio) is only 0.93%, where the average conversion rate is around 1,5% (8020 ecommerce, 2012).
The purpose of this research is to examine the influence of the different levels (high, low) of the variables (value, aesthetics, usability (navigation, content)) on purchase behaviour dependent variables purchase and transaction size. Unlike previous research focusing on intentions, this research focuses on actual behaviour of the consumer. This focus on actual behaviour can give insights to online store owners to increase the effectiveness of their online store. By using data from consumers’ behaviour, purchase patterns can be directly examined rather than rely on self-reported data.
3.1 Research design
This study focuses on the extent to which the different variables of online store design influence the purchase behaviour of the consumer. A conclusive research is designed to collect the information needed. The research is shaped as a true experiment, the different independent variables (value, aesthetics, usability (navigation, content)) are manipulated and their effect is measured on the dependent variables (purchase and transaction size) (Malhotra, 2009). The consumers are randomly assigned to one of the manipulations. During the test no pre-measurements are involved, which leads to a post test-only control group design. Furthermore, the experiment will take place in the field, the test will be conducted in actual market conditions. The symbolic arrangement of the experiment is as follows:
3.2 Data collection
The field experiment was conducted as an online A/B/n test to collect data on (individual) consumer behaviour in an online store. Most commonly known in web analytics is the A/B test, which is a market research method which is used to research which is the most efficient of two possibilities. Live users are randomly assigned to one of two variants: the control, which is usually the existing version and the manipulation, which is a new version being evaluated (Kohavi, et al., 2009). An A/B/n test has the possibility to include more than one manipulation to evaluate, so in this case with 4 online design variables a A/B/C/D/E test. The A/B/n test was chosen because of its capability to track actual behaviour and the possibility of testing different manipulations at the same time. For 23 days visitors behaviour on the selected online store were tracked to further research their response towards the manipulation.
3.2.1 Instrument
As mentioned in the literature review aesthetics (proportion, colour, shape and size), value (unique, specific and relevant to the consumer) and usability, which was split into navigation (text and icon links) and content (complete, relevant and easy to understand characteristics and presentation of information) are found to be variables that determine the design of the online store. The independent variables of the online store design are defined as the factors used in the online A/B/n test for each manipulation. An option to test the factors is to crosslink them in the manipulations, leading to 16 manipulations (2^4). Another option is to isolate the factors per manipulation, based on those independent variables and levels (see table 3.1), 8 different home pages can be made (4x2).
Variable Level
Aesthetics High Low
Value High Low
Navigation High Low
Content High Low
Table 3.1: Possible manipulations based on variables.
However, due to time issues to get the sufficient sample size for 16 or 8 manipulations, the choice has been made to take the control page as the low level for each independent variable and only manipulate the high levels of these variables. Each high level manipulation will be tested in an regression to determine whether change in purchase behaviour can be explained by the high level of the independent variables.
manipulation that visitors have seen to the detailed statistics from Google Analytics. A disadvantage of using Google Analytics is that the statistics do not show individual data.
This disadvantage is overcome by modifying the Google Analytics code (which is implemented in the online store) to track individuals by UserID. This does not violate the privacy of the visitor, since there is no personal identifiable information stored in the UserID. It connects each visitor to a unique number, so there is no information stored about gender, age, email address, etcetera.
The implementation of the independent variables into the manipulations was based on previous literature (Hoffman & Novak, 1996; Zhang & von Dran, 2000; Bloch et al., 2003; DeLone & Maclean, 2004; Soonsawad, 2013), best practices based on success cases of visualwebsiteoptimizer.com and the opinion of the online store owner. Best practices and the opinion of the online store owner were most important in determining the manipulations. The reason for this is that most literature did identify elements of the independent variables, but did not provide specific guidelines on how to manipulate them. Apart from the manipulation on the page, all homepages are identical.
One of the manipulations can be seen in figure 3.2 and represents the independent variable content. The manipulation includes all aspects which are important according to the literature shown in chapter 2. It contains the same content as the control homepage (figure 3.1) which should be all on the homepage according to the online store owner, so the content is complete and relevant to the consumer. In the control homepage the information is in bulk and can be easily overseen, in the manipulation the information is presented actively and in bullet points. This makes it easier for consumers to understand the characteristics of the isolation foil, because it can be red easier and faster. All other manipulations also implemented the aspects found in the literature, the other variations can be found in appendix 1.
Figure 3.2: Content manipulation homepage
After the manipulations were established and implemented in the A/B/n test of Visual Website Optimizer, the test was set live. All manipulations were live at the same moment and consumers who visit the online store were randomly assigned to one of the 5 homepages.
3.2.2 Sample size
As mentioned in chapter 1 more than 75% of the entire Dutch population shop online, which is a large population to test. To narrow the population down, a sample is taken. The sample for this experiment are all online shoppers who visit the specific online store which is studied in a certain time period. In order to avoid repeated significance testing errors, the sample size has been decided on in advance. To gain sufficient amount of data, the following hypothesis testing approach formula is used (Malhotra, 2009):
√
(
)
√
(
)
(
)
In this formula the null (H0) and the alternative (H1) hypotheses are tested in terms of the population proportions, π0 and π1, respectively. The z values are determined by the error probabilities of Type I error (α) and Type II error (β), using the Z value table in Malhotra (2009).
Every online store differs in amount of visitors and visitors that purchase (purchase behaviour), so it is difficult to establish a mean. Therefore, the sample size is determined based on proportions instead of the mean. The hypothesis testing approach is chosen, because the standard deviation for online stores is not known in advance.
To calculate the appropriate sample size the null and alternative hypotheses are based on the proportions of the conversions in online stores. Currently, in the studied online store the conversion rate is 0.93%. Ideally, this percentage should increase for the online store to become more effective. Therefore,
The allowable probabilities of Type I (α) en Type II (β) error are specified as,
The alternative hypothesis (H1) stipulates a proportion greater than that of the null hypothesis (H0), therefore it is a one sided test. For a one sided test, the z values are zα = 1.64 and zβ = 1.64. By putting all information together, the following equation is formed:
√ ( ) √ ( )
( )
This leads to n = 9,9 million. However, the current experiment consists out of 5 variations (including the control), which makes the sample size 49,5 million. With the time and money available this sample size is too large. Especially since the studied online store has an average of 150 visitors a day. So, to decrease the sample size, the formula is calculated to detect a significant difference of 5 per cent. The allowable probabilities of Type I and Type II error stay the same.
A new equation is formed based on the new hypotheses:
√ ( ) √ ( )
( )
This equation leads to n = 160. This should be multiplied by the 5 variations, which makes the sample size 800. Although niche online stores usually have a higher conversion rate than more general stores, it is not expected that the studied online store can reach such a high increase in conversion rate.
To make sure that a sufficient sample size was reached, which could detect significance in small differences, the test was run for 23 days. That was the maximum days available to test the variations, because of time and money limitations. After the 23 days a sample size of 1139 visitors was realised. While the online store to be studied was selected by the researcher, the visitors which made up the sample size were randomly assigned to one of the variations. The visitors were not aware that they participated in a study, so tracked behaviour is real. This means that the visitor cannot cheat or cause an error by giving socially preferred answers.
3.3 Plan of analysis
3.3.1 A/B/n test
Some literature has criticized that focusing on A/B/n testing implies short-term focus (Quarto-von Tivadar, 2006), because it fails to look at delayed conversion metrics where there is a gap of time from when a consumer is exposed to something and when the action is taken. Visual Website Optimizer does indeed collect the behaviour of the consumer only once, however Google Analytics tracks the consumer each time that they visit the online store. In Google Analytics the delayed conversions are tracked and has a longer-term focus. Also, the consumers see the same manipulation each time they visit the online store. Another important point is that the manipulation should be of sufficient quality to expose users to it (Kohavi, et al., 2009). With the use of Visual Website Optimizer a sufficient quality can be guaranteed, because it has an editor where you can drag and drop without changing anything else.
After the data was collected with the A/B/n test, a custom report was composed in Google Analytics and exported for further analysis.
The analyses used to further analyse the collected data are discussed in the next sections. 3.3.2 Regression analysis
The hypotheses formed in chapter two are tested with an regression analysis. A regression analysis is a statistical procedure for analysing associative relationships between a dependent variable and one or more independent variables (Malhotra, 2009). In this study the objective is to find out whether there is relationship between the independent variables and the dependent variables. A regression analysis can be well used to determine whether the independent variables explain a significant variation in the dependent variable (Malhotra, 2009).
3.3.2.1 Binary logistic regression
The data gathered on the dependent variable purchase answers the question whether a visitors purchased something or not, which means that the response is a yes (1) or no (0) answer. A linear regression produces predictions that go beyond 0 and 1, so this technique is not suitable to conduct in this study. On top of that it cannot be assumed that the data is normally distributed. Since conversion rate is in the low single digits, there are far more visitors who did not purchase than visitors who did. This makes it difficult to assume normal distribution.
Binary logistic regression is a technique that models the chance on a categorical outcome. It determines how likely it is that a subject belongs to a group. To use logistic regression it is not necessary to make assumptions about the distribution of the dependent variable. Therefore, binary logistic regression is suitable to use in this study.
In a logistic regression the dependent variable is in such a way transformed that a type linear regression is possible. However the chance on a categorical outcome reaches from 0 to 1 and at linear regression the dependent variable has to be metric, therefore the chance cannot be used directly as an outcome. To transform this the relevant chance is used in logistic regression, the odds. The odds can be seen a metric variable reaching from 0 to infinity. However the odds are still not normally distributed, a prerequisite for linear regression. The outcome is therefore not the chance nor the odds, but the natural logarithm (ln) of the odds. These are both metric and normally distributed.
visitors did make a purchase. This means that there are 27 events for the dependent variable purchase, which is valid for logistic regression. Another prerequisite is that the visitors in the sample must be independent, which means that they should not be related due to repeated observations of the same visitors, visitors that are paired (for example siblings) or clusters of visitors related by geographic proximity.
3.3.2.2 Linear regression
The other identified dependent variable of purchase behaviour transaction size is tested using a linear regression. A linear regression analysis is a statistical procedure for analysing associative relationships between a metric dependent variable and one or more independent variables (Malhotra, 2009).
Transaction size is metric data and therefore a linear regression is suitable to use. Only the transaction size is tested, which means that visitors who did not make a purchase are left out of the equation. This results in a sample of 27 visitors for the linear regression. It is assumed that the dependent variable is normally distributed.
3.3.2.2 Dummy variables
In order to conduct both regression analyses with category independent variables, the variables should be coded into dummy variables. By coding the categorical variables as dummy variables, they can be used as predictors of the dependent variables (Malhotra, 2009). A dummy variable can only take on two values, such as 0 or 1 (Malhotra, 2009). In this study there are 5 categories: aesthetics, value, navigation, content and the control. To respecify these categories K-1 dummy variables are needed. For this study this means 4 dummy variables. 4 dummy variables are sufficient, because information about the 5th category can be derived from information about the other 4 categories (Malhotra, 2009). To conduct the regression analysis, online store design is represented by 4 dummy variables X1, X2, X3 and X4 as shown in table 3.2.
Online store design
category Original variable code
Dummy variable code
X1 X2 X3 X4 Control 1 0 0 0 0 Aesthetics 2 1 0 0 0 Value 3 0 1 0 0 Navigation 4 0 0 1 0 Content 5 0 0 0 1
Note that X1 is 1 for the aesthetics category and 0 for all others. Likewise X2 is 1 for value and 0 for all others, X3 is 1 for navigation and 0 for all others and X4 is 1 for content and 0 for all others. In analysing the data, X1, X2, X3 and X4 are used to represent all categories. When all dummy variables are 0 it represents the control category.
The dummy variables X1, X2, X3 and X4 are used as predictors. For the dependent variable purchase the binary logistic regression with dummy variables is modelled as:
(
)
The linear regression with dependent variable transaction size with dummy variables is modelled as: ̂
In this study, the control has been selected as a reference category, representing the low levels of the categories, and has not been directly included both regression equations.
3.3.2.3 Moderated regression analysis
The identified moderator ‘repeat visitation’ is also a categorical variable. The variable consists out of 2 categories: new and repeat visitor. The variable is already coded into a dummy variable, where repeat visitor is 1 and new visitor is 0. Repeat visitation is 1 for the repeat category and 0 for the new category.
The moderation indicates that the relation between online store design and purchase behaviour differs by the level of repeat visitation. To test this moderation both the online store design variables and repeat visitation and interaction terms between online store design and repeat visitation are included in the regression to predict purchase behaviour. The regression with the moderation effect of repeat visitation on purchase is modelled as:
(
)
The regression with the moderation effect of repeat visitation on transaction size is modelled as:
̂
