User preference measure for browsing experience
optimization within commercial websites
SUBMITTED IN PARTIAL FULLFILLMENT FOR THE DEGREE OF MASTER
OF SCIENCE
Chanyu Sheng
11168137
M
ASTER
I
NFORMATION
S
TUDIES
H
UMAN-
C
ENTERED
M
ULTIMEDIA
F
ACULTY OF
S
CIENCE
U
NIVERSITY OF
A
MSTERDAM
25 - 07 - 2018
1st Supervisor 2nd Supervisor
Dr. F.M. (Frank) Nack Ms. Monika Matuszewska Faculty of Science Informatics Institute, Digital CRM and Brand Projects,
University of Amsterdam L’Or
é
alUser preference measure for browsing experience optimization
within commercial websites
Thesis of MSc Information Studies (Human Centred Multimedia)
Chanyu Sheng
University of Amsterdam Amsterdam, The Netherlands serena.sheng@student.uva.nl
ABSTRACT
Compelling visual and layout design have become an essential precondition to many commercial web pages. To convey more web tra�c, companies are increasingly investing more of their marketing budget into website design and content maintenance. Yet, many websites fail to help companies to reach their objectives; the performance e.g. browsing time, tra�c and conversion rates of these websites has not always met its expectation. Despite the fact that commercial web pages have adopted professional principles of web design, large amount of the content has still remained unseen by their users.
The current research was a qualitative study of how end-users were stimulated to browse and explore through commercial web pages. Taking into account that current web designs may not always meet the desire of their audience, this study aimed at reasoning the triggers that users received when they browsed through a web page and their underlining objectives for web browsing. As a result of a thorough qualitative analysis on the topic of functional and perceptional browsing preferences across 22 users using 3 versions of web prototypes, the current research gained insights with re-gards to commercial web browsing. Results suggests that features related to time e�ciency and sensation of safety were crucial to stimulate users’ browsing behavior with a precondition that the search features must be always accessible.
KEYWORDS
Layout, E-commerce, Commercial Web, User preference, Web de-sign, Web aesthetics, User experience dede-sign, Human Computer Interaction
1 INTRODUCTION
Through the rise of digital marketing and e-commerce, the compe-tition among commercial websites has become more serious than ever before. Knowing the fact that consumers who have experi-enced pleasant user journeys are more likely to associate positive emotions with the brand and more stimulated to make a purchase [17], companies tend to spend increasingly more time and money into their website design. With this, many have chosen to adopt a popular or commonly used web template for their website in order to enhance user experience and achieve high browsing, tra�c and conversion rates [7, 12, 21]. Yet, websites that deliver excellent user experience do not guarantee that users will explore the whole set of web pages provided by the company.
While the ultimate goal of a website is to achieve pleasant user experience, allowing users to engage with the content is just as
important as the general user experience. These days, companies wish to expose more content to their target group and achieve longer browsing sessions. For them, the purpose of their website is not only optimizing information query or allowing users to make a successful purchase, but also exposing product content to their target users in the hope that their users will get triggered and in-crease their intention to purchase more than they would at the �rst place. However, rapidly growing amount of commercial websites has caused users to be more selective about what they want to view and how they want to view. For many, it is challenging to �nd the balance between more content exposure and pleasant user experience [15].
Though, various previous studies published their own view on what is called a proper design, web browsing preferences within commercial environments from the users’ perspective were largely unexplored. The lack of understanding of the major stimulants or impediments to users’ longer website browsing sessions has presented a substantial problem for many designers. Especially for commercial websites that contain a wide variety of content, it is challenging to enhance user experience for reaching desirable content and at the same time exposing new content.
Previous research showed that users commonly stop or suspend their browsing session for unknown reasons [29]. To gain more insight about the problem of why users lost their interest while browsing and that certain content on websites remained unseen, the current research was devoted to the investigation of user needs when browsing commercial websites. Exploring the possibilities and challenges of web design for browsing, the current study ap-plied qualitative methods consists of user test driven observations and post-observation interviews to understand the browsing pref-erences of users. With speci�c focus on the principles of layout design and user experience design, the current research targeted on the digital marketing environment of fast-moving consumer goods (FMCG). The research questions reads as following.
Main research question: "How to stimulate the browsing behavior of web users within a commercial web environment through web design?"
In order to answer the main research question, the following sub research questions were posed. These were a set of support questions that needed to be answered in order to answer the main research question.
Sub research questions:
(2) What feature is associated with positive browsing experi-ences?
(3) What feature is associated with negative browsing experi-ences?
The current research were interested in e�ective distribution of users to web ares they have not explored. Using 3 prototypes of a commercial website www.redken.com, the research aims at dis-covering certain ways of content representation that are perceived to be the most e�ective to potential users when browsing through web pages. In the �rst and upcoming sections, literature including topics such as visual appeal, usability, user preferences and web layout apparatuses are discussed. Combining theories derived from literature research and observations of current popular web content representation methods, in the methodology section, the experi-mental setup of the current study will be elaborated upon. This is followed by a discussion of the �ndings and future implications for commercial web designers and marketers.
2 RELATED WORK
In this section, earlier work in research topics which were relevant to the current research will be elaborated upon. These topics in-cluded: role of usability and aesthetics, layout methods, features and apparatuses, metrics for the evaluation web layout and web design for browsing.
2.1 Role of aesthetics and usability
Web user experience will directly a�ect the online consumer be-havior [7, 17]. Usability, content personalization and aesthetics are critical factors for creating smooth browsing experiences [18]. While both web aesthetics and usability can contribute to an en-gaging user experience[6, 24], it has been an ongoing discussion concerning which of the two is more important.
Around the discussion whether visual appeal is beyond useful, Wang et al. thoroughly examined the role of web aesthetics and web usability [26]. According to their research, web usability has a positive e�ect on the conversion rates, but it also a�ects the willingness of users to explore through the website as they will directly be guided to their search and do not get the chance to engage with other possible content of interest. Web aesthetics on the other hand have a positive e�ect on the browsing behavior of users. The perceived online service quality and satisfaction within an e-retail environment are generally positive when users �nd the web page appealing[25, 27]. Supporting this, based on the so called Stimulus-Organism-Response (S-O-R) framework and the pleasure, arousal, and dominance (PAD) emotional model, Chang et al. veri�ed that the aesthetic aspect in web page visualization signi�cantly in�uenced the purchase behavior of users through the emotional model [3]. The appearance of a web page activates the user’s willingness to search for content and explore through pages. Only the homepage layout already a�ects user satisfaction [11] and their willingness to browse further. Moreover, visual appeal produces a greater in�uence on customers’ evaluation of trust, than ease of use [17].
For many, web usability had been recognized as an undoubtedly fundamental element within web design. In their study about the browsing behavior of online users, von der Welth and Hauswirth
proposed the DOBBS (DERI Online Behavior Study), a method to obtain insightful browsing data of web users [29]. According to their research, users tend to open multiple tabs when browsing. As users can open a tab without having any interactions with it, not all open tabs implies to a active browsing session. In order to gain insight about the online browsing behavior, one needs to analyze log data (e.g. users’ activity, session duration, quantity of active windows that contributed to the browsing activity and more). Further, simple web features can determine whether users are willing to browse further. According to Milic et al. a simple back navigation has the ability to a�ect the general user experience [13]. The duration of the browsing session is related to the users’ shift of interests. According to Jiang et al., results become less and less attractive for users the longer the browsing session proceeds [8]. Therefore, to grasp users attention and interest, a balance must be found between web aesthetics and usability.
2.2 Layout methods, features and apparatuses
In the past, many studies investigated the layout of commercial web pages. In their study about web layout for small and medium-sized enterprises (SMEs), Winkelmann et al. [30] found that the structure of the most commonly used layout for online shops consists of logo at the left upper corner, a �rst navigation layer at the top, a second navigation layer at the left side and a content display under and next to the navigation layers of the page. This layout has been considered as the foundation of many web pages and it is widely adopted by cooperations all over the world. Further, in order to understand of the impact of layout within e-commerce at a deeper level, Hausman et al. incorporated the grati�cations theory, technology acceptance model, and the concept of �ow in their study of web interface features [7]. In their experiment, web features were categorized as computer factors or human features and used to examine their e�ect on consumer online purchase intentions. According to their results the computer factors including, Indication of security/secure site, Clear displays of page contents, Presence of clear menu items on each page and more1a�ect the consequent variables more than human factors do. The so called �ow, which combines these factors with the user emotion towards the experience has a signi�cantly large impact on the purchase intention of users.
In their research where objective design components were linked with subjective aesthetics, Seckler et al. found that web users over-all prefer symmetric, clean design with a blue color in favor. [22]. Layouts that contained content or description overload were gener-ally viewed to be more complex and less attractive. Also, typicgener-ally users used more of in-text navigation features than graphic ones such as page-turning buttons and page numbers [4].
2.3 Metrics for the evaluation web layout
Studies in the past addressed key design and performance met-rics including speed, navigability, content and responsiveness, for the usability evaluation of web layouts[16]. Performance metrics such as user satisfaction, the likelihood of return, the likelihood to recommend and the frequency of use were widely adopted.
1presence of shopping cart, Up-to-date information, Un-do button, Assurance of
pri-vacy, Payment options, Purchase tracking services, Company logo, Consistent web page design, Declaration of intended use, Logical web page information, O�ers order con�rmation, Product images as thumbnails
PROTOTYPE Search Browse method CTA (Product) Display
1 Static �lter left In�nite scroll design Hover - Preview - Click Direct display
2 Relative �lter left Page based design Click on Image Categorized (preview)
3 Hover -Top �lter Load more design Static button Categorized
Table 1: Overview of currently used prototypes and their features.
Web optimization refers to the continuous process of improving website performance. It ensures that the website ful�lls customers’ expectations as well as the purpose intended by the company itself [28]. In their qualitative research, Marc Hassenzahl and Andrew Mon reported a positive correlation between perceived beauty and perceived usability [5]. While various metrics had been adopted for users’ browsing experience validation, few studies chose in-depth user data to understand and measure the browsing preferences from the users’ perspective. In their empirical research, Tan et al. examined the behavior of e-commerce platform users by asking them to think out loud [23]. The �ndings of their research suggested that various factors of website design such as reduced lostness reduce cognitive overhead facilitated users to the purchase destination.
2.4 Web design for browsing
Within the domain of online consumer behavior studies, the online browsing behavior can be roughly distinguished into two categories, goal-directed and experiential [10, 14]. A goal-directed browsing session is more task driven and can be strongly a�ected by the perceived usability of a web page. An experiential browsing session is more associated to the sensory pleasure of users and considered as a continue process of search. Along with these �ndings, in her E-business research, Koski found that factors such as anonymity, ease of access and content variety can stimulate impulsive purchases [10]. According to Koski, greater variety in products can cause com-plex information load to online consumers that might encourage impulse purchases. Similarly, in their study of impulsive purchase behavior among tourists Rezaei et al. examined the role of web site personality, utilitarian and hedonic web browsing [19]. According to the authors: "website personality was a second-order re�ective construct comprising solidity, enthusiasm, genuineness, sophistication and unpleasantness (page 60)" [20]. Both hedonistic web browsing and utilitarian web browsing positively in�uences the impulsive online shopping behavior.
Goal-directed and experiential browsing methods were insepara-ble processes and both commonly used by users. However, the web design objectives for these browsing methods did deviate. While the usability of a web page is claimed to be the most important reason for a satisfying user experience, research has found that web aesthetics is a signi�cant factor which contributes to the user’s browsing journey [26].
Along with the upswing of web 2.0, web users of the current generation are more community driven [1]. Until recent, research found that consumers who are browsing with intentions are likely to rely their browse and purchase decisions on product ratings [9].
3 METHODOLOGY
The current research aimed at discovering browsing preferences of web users within commercial environments. Based on previous literature research and pre-experimental observations, 3 interactive prototypes of a business commercial website owned by L’Oréal PPD2were made and validated by potential users.
Present in more than 130 countries and owning over 32 inter-national brands, L’Oréal is a Fast Moving Cosumer Goods (FMCG) company that aims at providing unisex beauty products to its large consumer audience. Brands of L’Oréal are divided into 4 distinct divisions i.e. L’Oréal Luxe, Active Cosmetics Division (ACD), Pro-fessional Products Division (PPD) and Consumer Products Division (CPD).
With a broad variety of professional hair products and sub-brands, www.redken.com is a international web page of Redken, one of the American brands under L’Oréal PPD. The website of Red-ken has a basic structure similar to the one that has been previously mentioned in section 2 [30]. As the website contains a su�cient amount of products and consists of a solid amount of web features that could be easily counterfeited, it was chosen as the foundation for all current implementations.
3.1 Prototypes
In total 3 prototypes were implemented based on www.redken.com (owned by L’Oréal PPD). Note that, despite the fact that the cur-rent research focused on the browsing behavior of users within commercial web pages, www.redken.com was not an e-commerce platform but a business website for consumers to �nd detailed in-formation about its products. The website did not provide users with the option to directly purchase from itself but did navigate users to online web shops which trades the products. Since the goal of the experiment was to observe the browsing behavior of users and understand the preferences of potential target users, the fact that www.redken.com was not a web shop should not be a concern to the current research. As the purchase experience was not taken into account when validating the browsing experience and the ap-pearance and structure of www.redken.com was similar to other existing e-commerce platforms, the author believed that the use of www.redken.com not a�ect the results of the experiment.
Further, the current study found it important to test up-to-date features. Therefore, to obtain a better understanding of how current FMCG websites are structured, a pre-experimental analysis was ex-ecuted before any implementations. Here, the layouts and features of 20 FMCG websites derived from the following categories of web
shops: drugstore, perfumery, supermarket and apparel were ana-lyzed3. Based on observational analysis, commonly used features
and layouts were extracted. Table 1 provides the overview of all adopted features.
To test the in�uence of web features and layout design on the browsing behavior of potential users in a natural way, the current research conducted semi-formal interviews involving 3 interactive prototypes. Each of the 3 prototypes adopted a set of commonly used features, such as a "static �lter bar", "in�nite scroll design" or "categorized information display". In this way, the participants could experience the browsing procedure, provide real-time feedback and show real-time emotions.
All prototypes were implemented with Invision Studio Version 0.92.104. For the main structure of the prototypes, all carried a
similar basic layout for online web shops that had been previously mentioned in section 2 [30]. Figures 2 to 5 depict the home screen of all prototypes. To avoid branding e�ects, logos were removed in all prototypes. Also keyboard shortcuts and the search function were not available in the current prototypes. Figure 1 shows an overview of all tested features and the prototype they belonged to. All features embedded in all prototypes were commonly used features extracted from the pre-experimental analysis. Prototype 1, 2 and 3 covered all commonly used commercial web features that were observed in the pre-experimental analysis and were used to test the potential browsing behavior of users in various ways. The general structure of all prototypes consisted of mainly 4 page navigations, as presented in Figure 1:
• The categories �lter allowed users to enter content related to its category, in this case there were 2 main categories: all haircare products or hair styling products. By selecting one of the two options, users could navigate to a category speci�c page that only showed all products which belonged to the chosen category. By clicking on any arbitrarily chosen product on the category speci�c page, the product speci�c page for that product would be shown.
• Using the By end result option, users could navigate them-selves to an end result speci�c page that was related to the �nishing results of the products. The end result speci�c page showed exclusively all products that belonged to the chosen end result (e.g. "Smooth and straight hair" or "Curly & wavy hair"). By selecting one of the given products this page, users could navigate to the product speci�c page for that product. • The Bestseller option navigated the users directly to all best
selling products.
• The Award winners option navigated the users directly to an overview of all best selling products.
• The "others" refers to web navigations that the current re-search did not cover due the limited time of implementation and to the fact that no products, but brand related visuals were displayed on theses pages.
3Appendix C shows a list of analyzed websites that was used for the current research. 4https://www.invisionapp.com/
Figure 1: General web structure of all prototypes for the current research. Here, "Others" refers to pages that do not cover product information and has been left out for the cur-rent implementation.
The �rst prototype adopted the following web features:
Figure 2: General web structure of Prototype 1 (home screen) • A static �lter bar on the left side of the interface that stayed at the same position and did not move when other interac-tions were going on. This feature aimed to test the whether a always accessible search support would stimulate the brows-ing behavior of users to browse through more pages and look for more products.
• An in�nite scroll design where users were allowed to scroll through the product feed without any interruptions. This feature would test the whether a �uently time saving user interaction would be associated with any positive browsing experiences and stimulate the user to browse through more content.
• A direct non-grouped content display where content was displayed without previous categorization. This feature ex-posed the product feed directly to the users to test whether
leaving out the on page categorization would disturb the user journey or increase the product exposure.
• A quick view hover e�ect (i.e. when the mouse hovered over the product image a preview would show). If the user wished to view the product page, he or she could click on the preview and navigate to the speci�c product page. If not, the preview would disappear once the mouse is no more positioned at the image. This feature aimed to test whether showing product content without having the user to leave the page would a�ect the way users browse and whether or not it would stimulate his or her browsing behavior. Appearance wise, the second prototype was similar as the �rst one (see Figure 3). However, the following features di�ered:
Figure 3: General web structure of prototype 2 (home screen)
Figure 4: Prototype 2 (home screen). When scrolling down the page, the left �lter bar disappeared but still took place of the interface.
• The �lter bar on the left side was not static but relative (i.e. the �lter would move away when scrolling). The �lter functions disappeared when scrolling up but did take place of the interface (Figure 4). It aimed to investigate whether users would associate a missing �lter that still took place of
the interface support with negative browsing experiences. Also it would test whether the users would be more focused on the page content without having the �lter options in the interface.
• The content was not directly displayed but �rst being catego-rized into hair care, hairstyling and hair color. This display method tried to test whether users would be triggered to browse though more pages if the products were previously being categorized.
• The product pages were page based, on each page approxi-mately 20 items were displayed. By clicking on "next page" the user could get to see the next 20 items on the next page. This feature required users to click on a button and load the "next page". It aimed at testing whether loading sectioned product display implemented in the so called "page based design" would stimulate users browsing behavior to explore more "pages".
• This prototype contained further no preview functions. By clicking on a product image, the user could be directed to the product speci�c page. In reality, the users would have opened tabs every time they clicked on a item. So, this feature aimed to test whether users could associate positive or negative browsing experiences with opening new tabs and whether or not it would a�ect the way they browse.
Finally, unlike the previous 2 prototypes, the third prototype carried a di�erent layout where the �lter bar is not visible in order to test whether the users would be more focused on the page content without having the �lter bar displayed at all in the interface:
Figure 5: General web structure of prototype 3 (home screen) • The third prototype (see Figure 5) did not carry a �lter bar at the left side of its layout and Instead, when hovered over the header, a drop-down menu appeared with the same prop-erties. This change of layout aimed to test whether hidden �lter functions would be accepted by the users and stimulate the browsing behavior as users do not get to see the �lter in screen when browsing.
• The browsing interaction design was made based the "load more design" where users have to click on load more to view to more pages. This feature aimed to investigate whether
users would explore more content when having loading breaks in between.
• Furthermore, this prototype also categorized the products according to their properties and contained no preview func-tions just like prototype 2.
• Apart from the drop-down menu, a noticeable di�erence compared to previous prototype was at the product display. As depicted in Figure 6, each product in the product display of Prototype 1 and 2 contained the product name, buttons and star ratings, whereas the product display of Prototype 3 replaced those by a "more information" button. Here, the features aimed to test the impact of descriptive content on the users browsing behavior; whether users would stop browsing earlier when no descriptive content is around.
Figure 6: Di�erence in product presentation between Proto-type 1 and 2 (left) en ProtoProto-type 3(right)
3.2 Sampling
As a commercial web page in reality could be exposed to any web users, the current research adopted the convenience sampling method, which means that the sample set was generated by virtue of accessibility [2]. In total 22 subjects participated in the experi-ment. The ages of the participants varied from 21 to 65 years. The ethnicity varied as well, but all subjects resided in the Netherlands. The participants were mainly initiated through the author’s personal social network with the preconditions that they should not be familiar with the original website to increase the internal validity of this research. Further, only participants who had basic web knowledge and online browsing experience were chosen for the experiment. Table 2 provides the demographic description of all participants.
3.3 Test procedure
To analyze the browsing behavior of online consumers, the cur-rent research adopted objective observations, semi-structured in-terviews and a questionnaire for its experiment.
ID Age Gender Nationality Status RES Group
1 19 - 25 M Dutch Student NL 1 2 19 - 25 F Dutch Student NL 2 3 26- 35 F Dutch Working NL 3 4 19 - 25 F Dutch Student NL 4 5 19 - 25 F Aruban Student NL 5 6 19 - 25 F Dutch Student NL 6 7 19 - 25 F Dutch Student NL 1 8 19 - 25 F Dutch Student NL 2 9 26- 35 M Taiwanese Working NL 1 10 26- 35 F Dutch Working NL 3 11 26- 35 M Dutch Working NL 4 12 26- 35 F Chinese Working NL 5 13 26- 35 M Dutch Working NL 6 14 26- 35 F Dutch Working NL 2 15 56 - 65 F Dutch Working NL 3 16 19 - 25 F Dutch Unemployed NL 4 17 19 - 25 F Dutch Student NL 1 18 19 - 25 F Dutch Student NL 5 19 19 - 25 M Dutch Student NL 6 20 26- 35 F Israeli Student NL 2 21 19 - 25 F Chinese Student NL 3 22 19 - 25 M Chinese Student NL 4 Table 2: Demographics of participants. Here, "ID" refers to the participant number of each subject, "Age" refers to the age group the participants belong to, "M" refers to Male par-ticipants, "F" refers to Female parpar-ticipants, "Status" refers to the current �nancial status of the participants, with an in-come around the average for those who are working, "RES" refers to the current country the participants reside and "Group" refers to the test group the participants belong to.
3.3.1 Counterbalance.As for the experimental set-up, to not confound the order in which a task was performed with the ex-perimental treatment and avoid the situation in which practice on user tasks leads to better performance and evaluation, the order of the current layout presentation was counterbalanced (i.e. subjects were provided with the prototypes in di�erent orders in such a way that each prototype was tested in each sequential position an equal number of times). In Table 2 and, the test orders of each group is shown5.
3.3.2 Procedure overview.Further, in order to study the e�ects of di�erent layouts and features on the browsing experience of the users, for all 22 experiments a within-subject design was used. Here, the independent variables were the di�erent versions of layouts with di�erent types of features. As for dependent variables, a semi-structured interview and a questionnaire with closed questions were used. The experiment took place in a lab environment where participants performed the experiment one after the other and were interviewed individually. In the exit interviews, all subjects reported having moderate knowledge of the Web, and spending an
5For details regarding the which group each participant belongs to, please refer to
Test group Test order Group 1 P1 - P2 - P3 Group 2 P2 - P1 - P3 Group 3 P3 - P1 - P2 Group 4 P1 - P3 - P2 Group 5 P2 - P3 - P1 Group 6 P3 - P2 - P1
Table 3: Test order of each group, with "Prototype" abbrevi-ated as "P".
average of about 12 hours on the Web per week. For clarity, the entire experiment was proceeded as following:
(1) The participant gets assigned to a certain test group with its test order.
(2) The participant gets a brief description of the experiment. (3) The participant �lls in the general demographic questions
in the questionnaire (Appendix B Q1 - Q5).
(4) The researcher shows the prototype, asks the participant to execute certain tasks including: search, browse and �nd most favored and least favored products.
(5) The participant executes a task while the researcher observes the participant and conducts a semi-formal interview after each task.
(6) The participant �lls in the questions about the prototype he or she tested.
(7) Step 4 till 6 will be repeated until the user has tested all 3 prototypes.
(8) General questions are asked about web browsing in the form of a semi-formal interview.
An entire experiment took approximately 60 minutes. The partic-ipants were asked to think aloud during the experiment and to use the same browsing techniques that they would use at home. Also the participant could indicate interactions that he or she would have executed in their real life browsing sessions but could not perform with the current prototype due to its limitations.
3.3.3 Data retrieval.Semi-structured interviews were conducted with all 22 participants in the observational study immediately af-ter each user test session to elicit decision making processes and criteria for certain browsing behavior6. The entire experiment was
audio recorded. During the user tests, the author took notes of the user behavior.
All user test sessions started with a general question regarding a scenario when the users intended to repurchase an item and want to �nd the item again. The intention of this question was to investigate the browsing possibilities when conducting a goal-directed search. Subsequently, users were asked to browse through the web pages and chose a most favored product and a least favored product for a certain hair type. Note that not all buttons were interactive in the current prototype, therefore speci�c scenarios were set for this question e.g. You have smooth and straight hair, what hair care product would you click and why?
6All asked questions and the questionnaire were enclosed as appendix.
After this the users were asked to provide their view on the CTA (call to action) button or image and brie�y browse through the product speci�c page. Finally, after having completed all tasks for all 3 user test sessions, general questions were asked to allow users to validate the prototypes, e.g. what feature they liked or �nd disturbing and what would stimulate them to browse.
Ensuring that the opinion were yield objectively, users were asked to complete a questionnaire after each testing session. The questionnaire was exclusively distributed to those who participated the user tests. It allowed subjects to validate the tested features and grade general user experience on each prototype.
4 RESULTS
As explained in the previous sections, participants were interviewed throughout the user test on topics such as web navigation, search, content selection and browse interest. All conducted interviews were transcribed manually and subsequently open coded using Atlas Ti Version 8.2.4 (559)7. For the entire research, the open-axial
coding method was applied; tentative labels were created for each data piece and subsequently being merged. As for the data analysis, graphs of code networks were generated to identify the relation-ships among the open codes. In this section, remarkable points yielded from the data analysis on the semi-structured interviews are explained. Also the results of the questionnaire which were used to con�rm the arguments and observation remarks of the users are discussed.
4.1 The power of search
As mentioned in section 3.3.3 The experiment initiated with a ques-tion regarding re-�nding a product. When asked about what partici-pants would do when they want to repurchase an item, the majority of 17 out of 22 (77.2 %) indicated that they would directly use the search bar to type in the name of the product and search. During the user tests, those participants would use the keyboard shortcut Ctrl-F to �nd the product name or look for search options at the right header corner. The absence of the search option could deliver instant frustration to users. Figure 7 shows the relations between various codes derived from the interviews.
Figure 7: Graphical presentation of code networks with re-spect to the search property.
Figure 8: Data visualization of participants view upon 3 types of major interaction design: in�nite scrolling, load more and page based design. Here, representative quotes that the participants used as argument for their choice are shown.
Remarkable was that, 3 Participants (13.6%) mentioned that if there was no search option incorporated or directly visible on the home screen, they would directly quit the browsing session and try to �nd the product in another way, such as using a search engine. These users were frustrated about the absence of the search function. According to participant 1: "I would honestly just directly go to the website and search by name. If it’s not there, I would exit the web page and go to Google or something, to search for the item.". Only a few participants (22.7 %) mentioned that they would initially use the �lter bars or tabs to navigate through the website and �nd the product. According to participant 2: "I would go to the category of that item. Of course if I still remember the way I got to the page I would do the same, but otherwise I would just click on the category I think the product belongs to.".
Further, 2 participants (9.1%) claimed that they were willing to go through the categories if there is no search option available. In fact the users also indicated that this is only the case when time allows. According to participant 14: "I would just search for the name, but if I �nd an item interesting at the home page, I would maybe click on that �rst and then start browsing. This is of course only the case when I have time, when I’m really looking for a product, I might just use Google.".
The search bar seemed to be a crucial feature during a goal-directed browsing session. Based on observations and yield argu-ments, a goal oriented search worked against the aim of companies to keep clients on their page. The initial mindset of participants towards this �rst question was to reach to the desired results as soon as possible. The majority (77.2 %) were only interested in the product they were looking for and not interested in browsing additional web pages. Moreover, the missing search bar a�ected
both the general user experience and their willingness to browse tremendously.
Though, results suggested the unlikelihood for goal oriented users to browse before search, a recommendation feature near the end of the search journey could be helpful for keep those users tuned. As a goal-directed search is driven by a goal, in this case �nding a desired product, when the desired product is found, the users might become willing again to receive new information and get triggered to browse again.
4.2 Browsing interaction
One of the tasks of the user tests was to �nd a favored product and a least favored product. Here, users were allowed to browse through the pages. After each the indication of their chosen products, the users were asked to elaborate upon their choices.
4.2.1 Product display.According to observations during the user tests, participants tend to browse more content and ignore the �lter bar when using the "in�nite scroll design". On average, users who used the "load more design" or "page based design" stopped their browsing session after 2 times of loading. In fact, when using the "in�nite scroll design" they scrolled through a amount products that was equal to 4 times of loading on average. These observations suggested that users who are using the "in�nite scroll design" were generally more focused on the browsing process and explored more content than when using the "load more design" or "page based design".
Further, the "In�nite scroll design" and "preview mode" were both cooperated in Prototype 1. Observations yielded that users tend to stay longer at a page that showed previews. Table 4 and
Codes A: keep browsing A: Navigate by categories A: quit A: Random click
Action: external search 0 0 1 0
Action: keep browsing 0 0 1 0
Action: quit 1 0 0 0
All in one page 1 0 0 0
Appearance 1 0 1 0
categorized display 1 0 0 0
Concern: Content recall 0 1 0 0
Content recall 0 1 0 0
Curiousity 0 0 0 1
Emotion: fearless 2 0 0 0
Feature: Bookmark 3 0 0 0
Feature: Filter inscreen 10 1 0 0
Feature: Maintain scroll position 2 0 1 0
Feature: Preview 5 0 0 0
Feature: Recommendations 15 0 0 0
images 1 0 0 0
Missing feature: account & wish list 3 1 0 0
Missing feature: brand categories 2 0 0 0
Missing feature: Product description 0 0 3 1
Missing feature: re�ned categories 2 0 0 0
Rating 3 0 0 1
Review 1 0 0 0
search direction 0 0 0 1
Table 4: Co-occurrence table of on the X-axis relevant actions, with "Action" abbreviated as "A" and on the Y-axis all relevant codes, generated with Atlas TI.
Table 5 show that the preview mode is associated to the positive browsing experience and the willingness to browse further. One of the argument for liking the preview feature was claimed by participant 8: "I’d like the preview option of one of the prototypes. The preview could prevent me from going to pages that doesn’t suit me and quitting the activity on this website.".
When asked about the preferences of browsing interaction with the product display e.g. in�nite scrolling, page turning or click-ing on the load more button, opinions seemed to vary. While the majority(54.5%) claimed to prefer the "in�nite scroll design", there were also interesting arguments mentioned why other interaction methods were preferred.
Figure 8 depicts the relationships between various elements of preferences and concern regarding the browsing interaction with the product display. To enhance the overview of the graph only few representative quotes are shown. The main arguments for in�nite scrolling was that it saved more time and its ability to scroll back at once. Here, 91.6% of the participants who had chosen "in�nite scroll design" as their desired display interaction method indicated the loading time as their argument. According to participant 18: "I like in�nite scroll, I don’t like the extra loading time. And if I want to scroll back, I can just scroll in one go.". As both the "pages based design" and "load more design" constantly required the users to click to load new content. The "in�nite scroll design" provided users a smooth user experience since the browsing session would not be interrupted by any loading issues.
However, 3 participants (13,6%) viewed this smooth browsing experience more as a concern and even claimed that the process of in�nite scrolling provided them a sense of anxiety as you never know where it ends. According to participant 5: "I would like the load more, for some reason the pages approach or the in�nite scrolling gives me anxiety. You don’t know where it ends. The load more option is just convenient. I can stop whenever I want."
For those who preferred the page turning interaction, the main advantage was that the page numbers supported them in remember-ing the item or content they might want to see again. Accordremember-ing to participant 16: "I would like the paged based design the most because then I can remember where I was.".
Finally, the "load more design" seemed to be a suitable feature for large content display. 18% Of the participants have mentioned about the customizable load more function where one gets to choose how many items he or she wants to see per scroll session. According to participant 19: "I prefer load more, especially for large amount of content display. If there is the possibility for me to change the amount of items to display I will do that and then load more. Because I don’t like to scroll through a ton of products". 4 Participants considered a customized load more function as a intermediate design that allowed them to smoothly browse but at the same time allowed them to take a break when "anxiety" struck.
4.2.2 Product property.While choosing the favored products, 11 users indicated the importance of product property e.g. name description, rating and tags. Here, ratings and descriptions were
Figure 9: Data visualization of codes related to the action of continue browsing most often referred as an indication for the product property. As
shown in Table 4, the main co-occurring reason for quitting a browsing session was when the product description was missing. According to participant 3: "I �nd it disturbing when there is no information about a product. For me it all depends on the description that determines whether or not I would click further or have the interest to browse further.". Further, as illustrated in Figure 9, the product property (i.e. what users think that the product does) had a direct relationship with the decision of whether the user still wanted to keep browsing.
4.2.3 Filter feature.During the browsing sessions, di�erent �l-ter features were used. Among the 3 �l�l-ter display designs, the static �lter option received most positive comments. Together with ar-guments such as "The �lter bar should always be visible, I do not like pages that I have to hover over. I like everything in one place. (participant 9)" or "Also, a �lter on the left side that stays in the screen allows me to go to any pages of my interest. (participant 4)", 17 users (77.3%) indicated their need of having a �lter bar always in screen. By placing the �lter function statically on the left side of the screen the participants were feeling supported to be able to navigate to their preferable content.
3 Participants (13.6%) mentioned their preferences for the drop-down �lter of the third prototype with the argument that this design would not block the usable space of the interface. As for the relative �lter in Prototype 2, all users claimed to �nd this design
inconvenient as it took space but was not usable when scrolling down the page.
Although the initial function of a �lter was search support, it seemed that it also stimulated the browsing behavior of users in-directly. As users claimed to want to have more �lter content to browse through, the �lter function was also used to explore. Here, 6 Users (27.3%) indicated their need for a more re�ned �lter func-tion, according to participant 5: "I �nd disturbing that the categories ended quickly. I would expect something like shampoo and condi-tioner, some primitive categories.". Moreover, 3 of the these users explicitly indicated their need for brand categorization. Participant 14 claimed: "First I would like to see all categories re�ned, in the end of the category, a brand option might be handy. Ideally it would be nice if you could �lter on both the brand and function category.".
4.2.4 Recommendations.According to current observations, those who were exposed to the product page and saw the recommenda-tions (91.0%), 85% tried to click on the suggestive content on the bottom. 68% Of the participants mentioned that recommendations of similar products would boost their browsing interest. Accord-ing to participant 15: "The recommendation were nice. That always stimulates me to click and browse further.". Moreover, Table 5 shows that positive browsing experience 15 times associated with the recommendations feature.
4.2.5 Bookmarks.The bookmarks i.e. the function to save or like a product was also preferred by users. This feature allowed
Codes Pos: Browse Codes Neg: Browse
Action: keep browsing 35 Missing feature: re�ned categories 4
Feature: Filter inscreen 16 Important: product description 3
Feature: Recommendations 15 Missing feature: Product description 3
Feature: Preview 7 Pos: Browse 3
Feature: Bookmark 6 Action: quit 2
Rating 6 Concern: Hover e�ect 2
Important: product description 4 Action: keep browsing 1
Appearance 3 Action: Navigate by categories 1
Feature: Dropdown �lter by hover 3 All in one page 1
Feature: Product description 3 Concern: Content recall 1
Missing feature: account & wish list 3 Concern: Filter in screen but move away 1
Neg: Browse 3 concern: no search property 1
Action: quit 2 Concern: Open a new tab 1
All in one page 2 Concern: Scroll back 1
Emotion: fearless 2 Content recall 1
images 2 Feature: Filter inscreen 1
Missing feature: brand categories 2 Feature: Maintain scroll position 1
Missing feature: Product description 2 Missing action: Back navigation 1
Missing feature: re�ned categories 2 Missing feature: brand categories 1
No content blocked 2
Review 2
categorized display 1
Concern: Hover e�ect 1
Feature: Maintain scroll position 1
No tags 1
Table 5: Co-occurrence table of the codes "Pos:Browse" and "Neg,Browse" and their co-occurring codes generated with Atlas TI. Here, "Pos:Browse" states for "positive browsing experience" and "Neg: Browse" states for "negative browsing experience".
users to �nd their content of interest easily and provides overview of preferred content. After having found their favorable products, 16 participants (68,2%) tended to click on the heart button and asked whether the product could be saved. According to participant 13: "I noticed that there is a possibility to like and store an item, that would also be extremely helpful for me to compare and �nd the products later on.".
4.2.6 On page categorization.Lastly, there seemed to be no strong preference towards whether the product display had cate-gorized its content into sections. According to observations, it did not matter whether the product display was categorized or not, for browsing purposes users generally �xated �rst at the �lter options and started their interaction with the �lters. When the �lter bar was not visible (Prototype 3), the user would either ask for the �lter or hover over the header. After �ltered at least once, the user started to scroll through the page. Here, only participant 19 mentioned her preferences towards an on page categorized display: "For me the packaging is quite important, so fancy, pretty packaging that are categorized in one section would be the best.".
4.3 Questionnaire
In order to gain insight and validate the qualitative analysis, results yielded from questionnaire are discussed below. Note that since the current research examined relative low amount of participants,
Question P1 P2 P3 Q7.1 6.18 6.18 5.59 Q7.2 5.68 5.86 4.68 Q8.1 6.59 6.33 6.32 Q8.2 6.91 7.33 6.52 Q10.1 7.18 7.57 7.00 Q10.2 7.09 7.24 6.45 Q.10.3 7.27 7.33 6.73
Table 6: Performance measure of question 7 till 10 of the questionnaire with a 10-point Likert scale. As all 3 proto-types had a similar foundation framework, the choice for a continuous 10-point Likert scale was made to achieve a higher variance and high degree of measurement preci-sion (i.e. a 10-point Likert scale o�ers more variance than a smaller Likert scale e.g. 7-point or 5-point Likert scales).
statistical data analysis was not performed as it would not be sig-ni�cant and contribute any meaningful conclusions for the entire population.
While browsing for a most and least favored product, users were asked to indicate the user experience of �nding a most favored and least favored product. Questions 7 and 8 from the questionnaire
Figure 10: Data visualization of the agreement questions of the questionnaire. Note that for "Q9: The hover ... browsing be-havior" contains only data from Prototype 1 because it is the only prototype that adopted the hover feature. Here, the "Hover e�ect" refers to the small preview.
allowed participants to validate their experience. According to the data shown in Table 6. Prototype 2, which adopted the page based design and a relative �lter, had the highest average rating followed by Prototype 1 which adopted the in�nite scroll and a static �lter. Prototype 3 which had a load more design failed to reach the par-ticipants favor. However, this could be explained to the fact that
Prototype 3 did not contain other descriptive information other than the image and the tag.
When looking at the data of Q9.1, the preview function, which was denoted as the hover e�ect, was agreed by the majority as a feature that would stimulate the browsing behavior. Noticeable as that in the same �gure, the �lter function of Prototype 1, i.e. the
static �lter bar, received most agreement associate with browsing (Q6.3).
Finally in Q10 participants were asked about the general experi-ence of using the prototype as a whole, for all 3 criteria Prototype 2 scored the highest and Prototype 3 as the lowest. Taken into account that the qualitative analysis found that none of the participants preferred the �lter feature of Prototype 2, other features such as page based approach and categorized display might have caused these results.
5 DISCUSSION
It has been a substantial problem that content remained unseen by users due to the fact that the design perspective of web devel-opers did not meet the browsing preferences of the users. Using 3 prototypes, the current research conducted a qualitative analy-sis to investigate the browsing behavior of potential users within commercial websites.
5.1 What makes people browse
To answer the posed sub-research question of what features are associated to positive browsing experiences, both co-occurrence tables (Table 4, Table 5) suggests that the �lter in screen and recom-mendation features were mostly associated to the positive browsing experience and considered as the biggest drive for keep on browsing for both goal-directed as well as for experiential browsing. Other descriptive features such as bookmarks, previews and wish lists would stimulate the browsing intention by the fact they allowed users to compare product content at a glance. For experimental browsing sessions, as 68% of the participants mentioned that recom-mendations of similar products would boost their browsing interest, one might need to think of a feature that provides users sugges-tions at an earlier stage of the browsing session to stimulate the browsing behavior. For goal-directed browsing sessions, the recom-mendations of similar products might be helpful to trigger the users’ interest to browse further. Therefore, regardless the intention of the user, a recommendation feature should be embedded at both the starter point of their browsing journey (e.g. at the home page) and near the end of their browsing journey (e.g. at the speci�c product information page).
For features that are associated to negative user experiences, re-sults yielded that the absence of the search feature and insu�cient product description would lead to negative user experiences. The presence of the search function was crucial for keeping web users on the website. Despite the fact that the search function was a key usability element and would a�ect the browsing opportunities of the users according to past studies, for goal-directed browsing ses-sions, users claimed to �nd this feature highly important and would quit the browsing session if they would not search for anything they had in mind. Further, online consumers and users were potentially seeking for safety when browsing. Features such as bookmarks and save for later buttons would provide users a sense of security and stimulate consumers to browse through the pages fearlessly. For a smooth and time saving browsing session, both observations and interview results showed that users favored the in�nite scroll option and was stimulated to browse. However, analysis on ques-tionnaire data showed that the page based approach seemed even
more favored. This can be, again, explained by the hypothesis that users seek for sense of safety when browsing. Page based design typically provides the users mental support memorize on a page number when browsing and adds a sense of security. Based on these �ndings, the current research suggested that commercial web design must contain features that allow users to save favored items without breaking the browsing interaction e.g. scrolling.
Furthermore, descriptive content such as ratings, product descrip-tion and tags, were crucial for stimulating the browsing behavior of users. According to the current analysis, product property consisted of appearance of the product, reviews, ratings, its description and tags. It could be viewed as a concern when the perceived product purpose was unclear due to lack descriptive content. Here ratings and descriptions were most often referred as an indication for the product property. Missing these information would lead to the end of the users’ browsing journey. Moreover, the �ndings showed that when browsing, participants tended to �nd an "all in one page" design pleasant. The majority mentioned that the �lter function should be visible on the interface. Also a preview function was reviewed as highly preferable by users.
To answer the sub-research question of how di�erent web fea-tures a�ect the way users browse; even when web feafea-tures contain the same content, users’ does not behave the same when the layout or interaction method di�ered. As discussed in the previous sec-tions, in�nite scrolling extended the browsing session of the users. The absence of the search feature and product description would cause users to drop o� the website. The preview mode allowed users to stay longer at a page and provided users positive emotions towards browsing.
5.1.1 Time e�iciency.When browsing, users generally preferred a short loading time. As the "in�nite scroll design" allowed users to skip the loading time, and stimulated the user to view more content, this feature could be adopted when one wants to achieve high product exposure rate. The preview function allowed users to brie�y browse to the content without having the risk of loosing the browsing position or going to a non-preferred page. Therefore, this feature is recommended when one intend to expose large amount of content but wants to have only relevant users to click through. 5.1.2 Sense of safety. It seems that the users are seeking for a sense of safety when browsing. The supportive features "static �lter", "page based design", "remain scroll position", "bookmarks" and "product property" were considered as important elements which stimulated the participants browsing behavior.
A static left �lter bar that was always visible provided the user browse support and indirectly supports users to browse. Page based design and features such as "remain scroll position" and "like but-tons" ensured that the users could easily view their favored products and compare them if necessary. Moreover, descriptive content was of great importance to keep users browsing. More than half of the participants indicated that when descriptive content such as prod-uct name, ratings or images were not there, they would immediately quit the browsing session.
To answer the main research questions of the current study of "How to stimulate the browsing behavior of web users within a com-mercial web environment through web design", both observations and data analysis were combined. Results suggests that features
related to time e�ciency and sensation of safety were crucial to stimulate users’ browsing behavior with a precondition that the search features must be always accessible.
5.2 Limitations
The current study believed to have obtained valuable �ndings re-garding the browsing preferences of potential users. However, there were limitations in the current approach due to time limitation and lacking test environments.
The current study used prototypes that lacked in completeness, interactiveness and speed. Due to these limitations, users could not perform all actions they intended to. For example, there was no search option available in the current interface or not all buttons on the header were interactive. As questions were asked to participants regarding certain features, it would have been more insightful if the prototypes could function as a website should in real life. Further, since the current research focused analysis on qualitative data, the yielded results were there to provide insight to the users’ perspec-tives and preferences of browsing. However, due to the fact that the samples were drawn from a restricted social connection of the author, the sample might be biased and not represent the opinions of the population. Moreover, the interview questions does not guar-antee that all possible information regarding browsing preference were retrieved.
6 CONCLUSION AND FUTURE WORK
A commercial website consists of various features that a�ect the way users browse. In order to solve the problem that web designs fail to meet users’ expectations and that large amount of web content remains unseen. By using a qualitative approach, the current re-search investigated web design features from the users’ perspective and answered the earlier posed main and sub-questions regarding users’ browsing behavior.
Although previous studies have claimed the negative relation-ship between search features and browsing behavior, results of the current research showed that search features (e.g. �lters, navigators and search engines) are crucial features to have in the interface. As the preference regarding interaction design methods (e.g. in�-nite scroll design, page based design and load more design) varied, observations showed that the users tend to extend their browsing session when using the "in�nite scroll design".
In order to stimulate the browsing behavior of web users within a commercial web environment, the layout must provide users a sense of security and compactness. As positive browsing experience was mostly associated with a "�lter in screen", "recommendation", "preview" feature, bookmark options and product descriptions, the current study suggest web designers to adopt all these features into the commercial web interface design. Further, negative browsing experiences could be avoided by ensuring that the product descrip-tion is representable to its funcdescrip-tion and that re�ned categories are adopted for the �lter feature.
To provide users a more realistic browsing experience, future studies could consider to execute similar tests with online websites, using other product categories. As, the current prototypes were incorporated categories only product wise, but not brand wise, many participants had mentioned the need for a more re�ned �lter
option. With this, it might be interesting for future studies to test the user preferences among re�ned �lter options. Moreover, as the commercial web environment is constantly growing, one could investigate the browsing experience of web environments that allows multiple input media such as voice and gestures.
ACKNOWLEDGMENTS
The author wish to thank various people for their contribution to the current research; all participants of the experiment for provid-ing meanprovid-ingful insights to this research. Special thanks should be given to Dr. F. Nack, the research project supervisor of the current research for his professional guidance and valuable support and to Ms. M. Matuszewska for her insightful and practical recommenda-tions on this project.
REFERENCES
[1] Russell Belk. 2014. Sharing versus pseudo-sharing in Web 2.0. The Anthropologist 18, 1 (2014), 7–23.
[2] Alan Bryman. 2008. Of methods and methodology. Qualitative Research in Organizations and Management: An International Journal 3, 2 (2008), 159–168. [3] Shu-Hao Chang, Wen-Hai Chih, Dah-Kwei Liou, and Lih-Ru Hwang. 2014. The
in�uence of web aesthetics on customersâĂŹ PAD. Computers in Human Behavior 36 (2014), 168–178.
[4] William P Eveland Jr and Sharon Dunwoody. 2016. Users and navigation patterns of a science World Wide Web site for the public. Public understanding of science (2016).
[5] Marc Hassenzahl and Andrew Monk. 2010. The inference of perceived usability from beauty. Human–Computer Interaction 25, 3 (2010), 235–260.
[6] Marc Hassenzahl and Noam Tractinsky. 2006. User experience-a research agenda. Behaviour & information technology 25, 2 (2006), 91–97.
[7] Angela V Hausman and Je�rey Sam Siekpe. 2009. The e�ect of web interface features on consumer online purchase intentions. Journal of Business Research 62, 1 (2009), 5–13.
[8] Jiepu Jiang, Daqing He, and James Allan. 2014. Searching, browsing, and clicking in a search session: changes in user behavior by task and over time. In Proceedings of the 37th international ACM SIGIR conference on Research & development in information retrieval. ACM, 607–616.
[9] Aaron C Johnson. 2017. Star-Crossed Consumers: The E�ects Of Online Rating Scale Length On Product Evaluations. (2017).
[10] Nina Koski. 2004. Impulse buying on the internet: encouraging and discouraging factors. Frontiers of E-business Research 4 (2004), 23–35.
[11] Weilin Liu, Fu Guo, Guoquan Ye, and Xiaoning Liang. 2016. How homepage aesthetic design in�uences usersâĂŹ satisfaction: Evidence from China. Displays 42 (2016), 25–35.
[12] David Lowe and John Eklund. 2002. Client needs and the design process in web projects. J. Web Eng. 1, 1 (2002), 23–36.
[13] Natasa Milic-Frayling, Rachel Jones, Kerry Rodden, Gavin Smyth, Alan Blackwell, and Ralph Sommerer. 2004. Smartback: supporting users in back navigation. In Proceedings of the 13th international conference on World Wide Web. ACM, 63–71. [14] Thomas P Novak, Donna L Ho�man, and Adam Duhachek. 2003. The in�uence of goal-directed and experiential activities on online �ow experiences. Journal of consumer psychology 13, 1-2 (2003), 3–16.
[15] Steven Pace and others. 2003. Understanding the �ow experiences of Web users. (2003).
[16] Jonathan W Palmer. 2002. Web site usability, design, and performance metrics. Information systems research 13, 2 (2002), 151–167.
[17] Supavich Fone Pengnate and Rathindra Sarathy. 2017. An experimental investiga-tion of the in�uence of website emoinvestiga-tional design features on trust in unfamiliar online vendors. Computers in Human Behavior 67 (2017), 49–60.
[18] Amit Poddar, Naveen Donthu, and Yujie Wei. 2009. Web site customer orienta-tions, Web site quality, and purchase intentions: The role of Web site personality. Journal of Business Research 62, 4 (2009), 441–450.
[19] Sajad Rezaei, Faizan Ali, Muslim Amin, and Sreenivasan Jayashree. 2016. Online impulse buying of tourism products: The role of web site personality, utilitarian and hedonic web browsing. Journal of Hospitality and Tourism Technology 7, 1 (2016), 60–83.
[20] Sajad Rezaei, Faizan Ali, Muslim Amin, and Sreenivasan Jayashree. 2016. Online impulse buying of tourism products: The role of web site personality, utilitarian and hedonic web browsing. Journal of Hospitality and Tourism Technology 7, 1 (2016), 60.
[21] Deborah E Rosen and Elizabeth Purinton. 2004. Website design: Viewing the web as a cognitive landscape. Journal of Business Research 57, 7 (2004), 787–794. [22] Mirjam Seckler, Klaus Opwis, and Alexandre N Tuch. 2015. Linking objective
design factors with subjective aesthetics: An experimental study on how structure and color of websites a�ect the facets of usersâĂŹ visual aesthetic perception. Computers in Human Behavior 49 (2015), 375–389.
[23] Gek Woo Tan and Kwok Kee Wei. 2006. An empirical study of Web browsing behaviour: Towards an e�ective Website design. Electronic Commerce Research and Applications 5, 4 (2006), 261–271.
[24] Manfred Thüring and Sascha Mahlke. 2007. Usability, aesthetics and emotions in human–technology interaction. International Journal of Psychology 42, 4 (2007), 253–264.
[25] Yong Jian Wang, Monica D Hernandez, and Michael S Minor. 2010. Web aes-thetics e�ects on perceived online service quality and satisfaction in an e-tail environment: The moderating role of purchase task. Journal of Business Research 63, 9-10 (2010), 935–942.
[26] Yong J Wang, Soonkwan Hong, and Hao Lou. 2010. Beautiful beyond useful? The role of web aesthetics. Journal of Computer Information Systems 50, 3 (2010), 121–129.
[27] Yong Jian Wang, Michael S Minor, and Jie Wei. 2011. Aesthetics and the online shopping environment: Understanding consumer responses. Journal of Retailing 87, 1 (2011), 46–58.
[28] Birgit Weischedel and Eelko KRE Huizingh. 2006. Website optimization with web metrics: a case study. In Proceedings of the 8th international conference on Electronic commerce: The new e-commerce: innovations for conquering current barriers, obstacles and limitations to conducting successful business on the internet. ACM, 463–470.
[29] Christian von der Weth and Manfred Hauswirth. 2013. Dobbs: Towards a com-prehensive dataset to study the browsing behavior of online users. In Proceedings of the 2013 IEEE/WIC/ACM International Joint Conferences on Web Intelligence (WI) and Intelligent Agent Technologies (IAT)-Volume 01. IEEE Computer Society, 51–56.
[30] Axel Winkelmann, Matthias Boehm, and Jörg Becker. 2008. Usability of" Trusted Shops" An Empirical Analysis of eCommerce Shops. SIGHCI 2008 Proceedings (2008), 8.
APPENDICES
A INTERVIEW QUESTIONS
(1) Imagine the scenario where you have purchased item x and wants to �nd item x again. What would you do?
(2) Why do you did you choose the product as most favored? (3) Why do you did you choose the product as least favored? (4) You have seen multiple ways of web interaction, for example
in�nite scrolling, load more and page based approach, which way of interaction do you prefer and why?
(5) What do you �nd disturbing and what did you �nd pleasant while browsing?
(6) What makes you to continue browsing further?
B QUESTIONNAIRE
The following questionnaire was conducted after each user test. Question 6 till 10 were repeated after each test session. The form was powers by Qualtrics and �lled in using participants’ mobile phones.
(1) Please enter your participant number: (2) Please indicate your age group:
• - - 18 years (1) • 19 - 25 years (2) • 26- 35 years (3) • 36 - 45 years (4) • 46 - 55 years (5) • 56 - 65 years (6) • 66+ years (7)
(3) Please indicate you Nationality.
(4) Please select your current �nancial status.
• Student (1)
• Working (Below average income) (2) • Working (Around average income) (3) • Working (Above average income) (4) • Unemployed (5)
• Others (6)
(5) In which country do you currently reside?
(6) Q6.1 - This question is asked with regards to the �lter feature The �lter function helped me with �nding the item. Q6.2 -This question is asked with regards to the �lter feature - I would directly use the �lter function if I entered the page for browsing purposes. Q6.3 - This �lter function would stimulates my willingness to browse through the website. (7) Q7.1 - How would you grade... - the degree to which the
�lter feature would stimulate your browsing behavior? Q7.2 - How would you grade... - the general user experience of
�nding the item?
(8) Q8.1 - How would you grade the general user experience of ... - �nding the least favored item? Q8.2 - How would you grade the general user experience of ... - �nding the most favored item?
(9) Q9.1 - The questions below are related to the CTA (Call to Action) feature. - By clicking on the button/product image I was easily navigated to the page with product information. Q9.2 - The questions below are related to the CTA (Call to Action) feature. - The hover e�ect would stimulates my browsing behavior. Q9.3 - The questions below are related to the CTA (Call to Action) feature. - I would directly click on the button/product image to get more information about the product.
(10) Q10.1 - Q. the degree to which this interface would stimulate your willingness to browse? Q10.2 - Q. How would you grade ... - the general user experience of �nding details about the product at this page? Q10.3 Q. How would you grade ... -the degree to which this interface is clear?
C LIST OF CONSIDERED WEBSITES FOR
PROTOTYPE DESIGN
(1) https://www.trekpleister.nl/ (2) https://www.etos.nl/ (3) www.kruidvat.nl/online (4) https://www.da.nl/ (5) https://www.hollandandbarrett.com/ (6) www.redken.com/products (7) https://www.kerastase.com/products (8) https://www.matrix.com/hair-care (9) https://www.iciparisxl.nl/ (10) https://www.douglas.nl/ (11) https://www.ah.nl/ (12) https://www.jumbo.com/ (13) https://www.deen.nl/ (14) https://www.dirk.nl/ (15) https://www.lidl.nl/nl/index.htm (16) www.asos.com (17) www.yoox.com (18) www.netaporter.com(19) https://www.uniqlo.com/eu/en/home/ (20) https://www.urbanout�tters.com/
