Citation
The effects of website information utility on the outcomes of user-website interactions

Material Information

Title:
The effects of website information utility on the outcomes of user-website interactions
Creator:
Hasley, Joseph Paul
Publication Date:
Language:
English
Physical Description:
xi, 116 leaves : ; 28 cm

Subjects

Subjects / Keywords:
Web sites ( lcsh )
Electronic commerce ( lcsh )
Human-computer interaction ( lcsh )
Internet surveys ( lcsh )
Electronic commerce ( fast )
Human-computer interaction ( fast )
Internet surveys ( fast )
Web sites ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 97-116).
General Note:
Department of Computer Science and Engineering
Statement of Responsibility:
by Joseph Paul Hasley.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
655269306 ( OCLC )
ocn655269306
Classification:
LD1193.E52 2010d H37 ( lcc )

Full Text
THE EFFECTS OF WEBSITE INFORMATION UTILITY ON THE
OUTCOMES OF USER-WEBSITE INTERACTIONS
by
Joseph Paul Hasley
B.A. University of Iowa, 1997
M.A. University of Iowa, 1999
A thesis submitted to the
University of Colorado Denver
in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Computer Science and Information Systems
2010


2010 by Joseph Paul Hasley
All rights reserved.


DEDICATION
This work is dedicated to my beautiful wife, Kathryn, who has been with me through every
step of this dissertation, even when I couldn't be with her.
This work is also dedicated to my parents, Paul and Geraldine Hasley, who taught me the
most important things I will ever learn: love of family, integrity, hard work, persistence.
Finally, this work is dedicated to my daughter, Bridget,"... to the moon and back."


ACKNOWLEDGEMENT
I would like to thank Dawn Gregg, my PhD adviser, for the hard work and patience she has
selflessly invested in me during the last 7 years. I would also like to thank the other members
of my committee: Peter Bryant, Madhavan Parthasarathy, Min-Hyung Choi, and A. Laurie
Shroyer (who believed I would accomplish this goal, even when I didn't) for the time and hard
work they have invested in my education.
I would like to acknowledge Robert Cooperman, David Milofsky, and David Romtvedt for their
enthusiastic and helpful participation in this project.
1 would like to acknowledge all the teachers in the Linn-Mar school system, especially Dale
Monroe, Marc McCoy, Trudy Sundermann, Suanne Huffman, Clark Weaver, Paul Knapp,
Steve Olson, Dixie Knudsen, Tom Lippert, Bob Wennekamp, Nick Spencer, Bob Blake, Jan
Bahmann, Marylou Ediburn, and Dorothy Thompson.
I would like to acknowledge the support and friendship of Alberto Segre, Thomas Gruca, Jay
Holstein, Brooks Landon, and Fatosh Kocer, who taught me at the University of Iowa and
who inspired me to pursue a PhD.
I would like to acknowledge the encouragement, friendship and love of my dear sister
Elizabeth and my brother-in-law Corey, my beloved grandparents, my Uncle George, my
cousins Brian Hasley and Jon Chase, and my close friend Cathleen Strabala. You have all,
from time to time, seen me at my worst. I hope you are proud of me today when I am at my
best.


TABLE OF CONTENTS
Figures.........................................................................x
Tables...........................................................................xi
1.0 INTRODUCTION................................................................1
1.1 Research Problem and Scope...............................................1
1.2 Topic Importance.........................................................1
1.3 Research Questions.......................................................2
1.4 Research Approach........................................................5
1.5 Contributions of this research...........................................5
1.6 Overview of this study...................................................6
2.0 LITERATURE REVIEW........................................................... 8
2.1 User-Website Interactions (UWIs).......................................... 8
2.2 UWI Antecedents..........................................................9
2.2.1 User factors........................................................9
2.2.2 Task...............................................................11
2.2.3 Context............................................................12
2.2.4 Technology factors.................................................12
2.2.5 Website design.....................................................13
2.2.6 Information Content................................................14
2.2.7 Information cues in the IS literature..............................15
2.2.7 UWI outcomes.......................................................18
2.3 Previous investigations of website information content and UWI outcomes.19
2.4 Gaps and opportunities..................................................20
3.0 THEORETICAL FRAMEWORK......................................................23
3.1 Utility.................................................................25
3.2 A new model.............................................................26
3.3 Hypotheses..............................................................28
3.3.1 Information content................................................28
3.3.2 Website Quality....................................................28
3.3.3 Engagement.........................................................30
3.3.4 Trust..............................................................32
3.3.5 Perceived Risk.....................................................34
4.0 RESEARCH METHODOLOGY........................................................37
4.1 Overview of the studies...................................................37
vi


4.2 Exploratory study: Measuring Website Information Content- The WICS.........37
4.2.1 A demonstration of the WICS...........................................38
4.2.2 Results of the exploratory study..................................... 40
4.3 Main study.................................................................42
4.3.1 Overview..............................................................42
4.4 Phase One: Identifying the cues most commonly presented within a domain...43
4.4.1 Sample and procedure..................................................43
4.5 Phase two: Determining the relative utilities of specific information cue categories
using Maximum Difference Scaling..................................................44
4.5.1 Sample................................................................44
4.5.2 Procedure.............................................................45
4.5.4 MaxDiff Results.......................................................47
4.6 Phase Three: An investigation of the relationships between information utility, the
UWI, and UWI outcomes.............................................................47
4.6.1 Sample................................................................47
4.6.2 Cue creation..........................................................49
4.6.3 Cue validation........................................................49
4.6.4 Information cue groupings.............................................49
4.6.5 Experimental website creation.........................................51
4.6.6 Procedure.............................................................51
4.6.7 Variables being studied...............................................51
4.6.7.1 Independent variable..................................................51
4.6.8 Sample Size...........................................................52
5.0 ANALYSIS AND RESULTS.........................................................53
5.1 MANOVA analysis............................................................53
5.2 ANOVA analysis.............................................................53
6.0 DISCUSSION...................................................................57
6.1 Hypothesis 1...............................................................58
6.2 Hypothesis 2...............................................................61
6.3 Hypotheses 3 and 4.........................................................61
6.4 Effects of information quantity............................................63
6.5 Overall summary of hypotheses..............................................66
6.6 Possible effects of priming and confirmation/disconfirmation...............67
6.7 Post Hoc Analysis..........................................................69
7.0 CONCLUSION...................................................................71
7.1 Limitations and Future Research............................................71
vii


7.2 Contributions...................................................73
7.3 Implications for Theory....................................... 74
7.4 Implications for Practice.......................................75
APPENDIX A..............................................................77
APPENDIX B..............................................................82
APPENDIX C............................................................. 88
APPENDIX D..............................................................91
APPENDIX E..............................................................92
APPENDIX F..............................................................94
REFERENCES..............................................................97
viii


LIST OF FIGURES
Figure
1: The Zhang and Li (2005) HCI model...............................................8
3.1: Post-UWI assessments of information quality are not able to separate the effects of
information content from the effects of how that information is framed........24
3.2: Pre-UWI methods of assessing information utility are able to isolate information content
from how that information is framed............................................25
3.3: Previous studies have focused on post-UWI measurements........................27
3.4: Relationships between Information Utility and UWI outcomes....................36
4.1: Each visitor was asked to indicate the most important and least important cues in
sixteen conjoint tasks........................................................46
x
x


LIST OF TABLES
Table
2.1: Previously identified dimensions of information quality.............................18
2.2: An example of expected financial payouts............................................21
4.1: Inter-rater reliability scores......................................................40
4.2: Paired Sample t-Tests Comparing Domains (significance)..............................41
4.3: The general cues manipulated at the experimental test websites......................44
4.4: Phase 3 participant characteristics.................................................47
4.5: Results of the MaxDiff investigation of cue utility.................................48
4.6: The cues included in each treatment.................................................50
5.1: Pairwise comparisons of the means for questions demonstrated to be significantly
influenced by information utility....................................................54
6.1: Pairwise comparisons of the means for questions demonstrated to be significantly
influenced by information utility.................................................... 59
6.2: A summary of the hypotheses investigated by this study..............................63
6.3: Pairwise comparisons of answers to question 21 for author Romtvedt..................65
A-1: Studies that have investigated the website quality construct........................ 77
B-1: The results of the WICS exploratory study and phase two of the main study...........82
C-1: Screen captures of the Romtvedt experimental websites (landing pages)...............88
C-2: Screen captures of the Milofsky experimental websites (landing pages)...............89
C-3: Screen captures of the Cooperman experimental websites (landing pages)..............90
D-1: Survey items for the main study.....................................................91
E-1: Summaries of the results of ANOVA for each of the 23 survey questions...............92
F-1: The results of ANOVA for author Cooperman.........................................94
F-2: The results of ANOVA for author Romtvedt..........................................95
F-3: The results of ANOVA for author Milofsky..........................................96
XI


1.0 INTRODUCTION
1.1 Research Problem and Scope
The world wide web has emerged as one of the primary ways businesses connect
with customers in the twenty-first century. Consequently, the user-website interaction, or
"UWI", has become a central focus of researchers and practitioners who want to know why
UWIs between e-tail businesses and their customers result in (or fail to result in) customers
making on-line purchases. In response to this question, hundreds of theoretical and
empirical studies have identified and described many of the elements that comprise UWIs,
the antecedent factors that influence UWIs, the outcomes of UWIs, and how UWI elements,
antecedents, and outcomes interact with one another (e.g. Ba & Pavlou, 2002; Fiore, Jin, &
Kim, 2005; Lim, Sai, Lee & Benbasat 2006; Pavlou & Gefen, 2004, Pavlou, 2003). The
results of these studies indicate that many factors, including user factors (experience using
computers and the internet, experience), task factors (complexity, goals), technology factors
(hardware issues, website design issues), and context, influence UWIs and their outcomes.
1.2 Topic Importance
Despite early predictions that the Web would eliminate seller-buyer knowledge
discrepancies and, thus, create a frictionless, price-based market (Anders, 1998; Kuttner,
1998), numerous studies have concluded that low prices are not the driving force behind
business-to-consumer (B2C) commerce. In fact, research has demonstrated that the Internet
does not inevitably provide lower prices than traditional mediums, nor do customers tend to
buy from the lowest priced seller (Brynjolfsson & Smith, 2000; Zo & Ramamurthy, 2009).
Rather, the main advantage of the Internet for business-to-consumer (B2C) website
customers lies in the relatively low cost of obtaining high-quality information (Alba & Lynch,
1997; Bakos, 1997). Hence, for B2C websites, success depends upon understanding how
customers use information to make decisions about what products to buy, what company to
buy them from, and whether or not to purchase the product on-line. (Devaraj, Fran, Kohli,
2002; Torkzadeh & Dhillon, 2002).
1


UWIs have received a great deal of attention from researchers, in part due to their
increasing importance to the global economy (e.g. Ahn, Ryu & Han, 2007; Karson & Fisher,
2005a; Kim & Stoel, 2004a; Pavlou 2003). In 2007, retail e-commerce was $136 billion in the
U.S. alone (US Census Bureau 2008). The results of scores of peer-reviewed empirical
studies demonstrate how understanding UWI antecedents and outcomes may be one of the
most practical and cost-effective paths toward creating sustainable competitive advantages in
the on-line environment. For example, research suggests that numerous antecedent factors
significantly influence consumer impressions of e-businesses (e.g. Barnes and Vidgen, 2001;
Kim & Stoel, 2004a, 2004b; Shchiglik & Barnes, 2004), their trust in the e-business (e.g.
McKnight, Cummings & Chervany, 1998; Pavlou, & Gefen, 2004), their willingness to transact
with the e-business (e.g. Ranganathan & Ganapathy, 2002) and ultimately, the prices they
are able to charge for their goods and services (e.g. Ba & Pavlou, 2002; Gregg & Walczak,
2008).
Although the growth of the UWI body of knowledge has greatly enhanced the abilities
of investigators to describe and predict the course of UWIs, there are still significant gaps in
the understanding of how website information content affects the outcomes of UWIs. The
purpose of this paper is to demonstrate a method of determining the information content and
information utility of website information cues, and to investigate whether website information
utility influences website visitor perceptions and actions regarding the website.
1.3 Research Questions
A significant body of research has investigated the factors that influence UWI
outcomes. Early investigations sought to identify major elements (site design, functionality,
information content, etc.) of successful websites (Huizingh, 2000; Liu, Arnett, Capella &
Beatty, 1997; Agarwal & Venkatesh, 2002). A few papers have sought to not only identify
major website components, but to assess their relative importance to website success. For
example, in their 2001 study, Mateos, Mera, Gonzalez, and L6pez, identified 4 major factors
of successful websites (Accessibility, Speed, Navigability, and Content Quality) and assigned
2


each factor a relative importance a priori, based on a 100 point scale. Ranganathan and
Ganapathy (2002) found four key dimensions of Business to Consumer websites (information
content, design, security, and privacy). Using multiple discriminant analysis, they determined
the relative importance of the factors to be security, privacy, design, and content. Zhang and
von Dran (2001) investigated the relative importance of fourteen website features
(information accuracy, information completeness/comprehensiveness, information
currency/timeliness, engaging, information reliability/reputation, information representation,
navigation, visual design, product and service concerns, readability/comprehension/clarity,
relevant information, security/privacy, site accessibility/responsiveness, and site technical
features) within six different domains (finance, e-commerce, entertainment, education,
government, medicine/health).
Hence, along with factors such as design, security, and privacy, information content
has been consistently reported to be one of the most influential factors in determining website
success. (Song & Zahedi, 2005; Evans & Wurster, 1999; Palmer 2002). However, previous
research also strongly suggests that providing "high quality" website information content is
not a function of simply providing all available information (ie: the information theoretically
required to make a perfectly informed decision) (Belonax & Mittelstaedt, 1978; Lussier &
Olshavsky, 1979) Current high-speed internet technology makes the incremental costs of
presenting information content relatively insignificant for sellers. At the same time, modern
web search-engines and shop-bot technologies have made finding large amounts of
information relatively cheap and easy for users. This does not, however, mean that users
have become more efficient or effective at finding or processing information that is relevant to
their decision-making needs. Research indicates that increasing information quantity does
not positively increase customer perceptions of information quality (Dholakia & Rego, 1998;
Kang & Kim, 2005). Because human cognition is limited, consumers cannot be expected to
find, organize, and process all information possibly available about even simple product
choices (Laroche, Kim & Matsui, 2003). In general, studies have indicated that providing a
consumer with additional information helps that individual with his decision-making task until
3


the point where the amount of information becomes so great that the individual encounters
information overload, a state where a decision-maker feels swamped by the amount of
information he is required to process in order to make a decision, which can lead to less
efficient, less effective decision making (Lee & Lee, 2004). Researchers have consistently
observed that the relationship between information quantity and information utility is that of an
inverted U-shape, indicating that additional information aids decision making to a certain
point, but then inhibits cognitive processing as information overload occurs (Jacoby, Speller
and Berning, 1974; Malhotra, 1982; Geissler, Zinkhan & Watson, 2001). Hence, although the
web has greatly reduced the costs of providing (sellers) and acquiring (buyers) information, it
does not generally afford a corresponding reduction in the cognitive costs of processing
information (Chen, Shang & Kao, 2009).
With the understanding that providing all available information does not equate to
providing high quality website information content, one of the fundamental questions being
asked by investigators is "for a given e-commerce website, what information constitutes high
quality information content, and what are the relationships between information content and
visitor perceptions and behaviors?"
Hence, this research is motivated by the following questions:
(1) What information content is presented within a given e-tail website?
(2) Are there subsets of website information cues which are common to a specific
domain?
(3) What are the relative utilities of the information cues presented at a given e-tail
website?
(4) What are the relationships between website information content utility and user-
website interaction outcomes at a given e-tail website?
To investigate these questions, this study develops and presents a survey of website
information content cue categories that are commonly included within B2C retail websites.
We then demonstrate how the survey can be used to meaningfully and reliably identify the
4


cue categories that are most commonly presented within various retail website domains. In
the second part of this study, we use concepts from the field of utility theory to explore how
customer assessments of the relative values of website information cue categories can be
calculated. Based on these assessments, we are able to compare the relative utilities of the
information cues presented within specific websites. Finally, we report the results of an
experiment that investigated possible relationships between a the calculated utility of a
specific website's information and the outcomes of user interactions with that website.
1.4 Research Approach
This article presents the results of a causal study that investigates a new model that
is rooted in prior investigations of information utility and UWIs. The independent variable
being manipulated is information utility. The independent variable is hypothesized to predict
the outcomes of the dependent variables trust, perceived risk, perceived website quality
(information quality and design quality), time spent at the website, and actual purchase
behavior.
This research uses content analysis methods, maximum differential scaling methods,
and survey methods to collect data and test the proposed research model. The unit of
analysis is at the individual level as perceptions of websites and actual purchase behavior are
considered. MANOVA, multiple-regression, and structural equation analysis are used to
examine the proposed hypotheses.
1.5 Contributions of this research
The main contribution of this research is to demonstrate that website information
cues can be systematically identified, that their relative information value can be calculated,
and whether there is a measurable, significant relationship between a website's information
value and how the website is perceived by its users. Most studies of UWIs use either a
deductive approach to test a theory or an inductive approach to formulate a theory based
upon observations. This study uses both deductive and inductive approaches. Using a
deductive approach, we investigate one way of measuring the information value of website
5


information cues. Then, using an inductive approach, we investigate the relationship
between website information value and the outcomes of UWI's.
This research builds upon several significant streams of IS research. First, we create
and demonstrate the Website Information Content Survey (WICS), a comprehensive list of
information cues. Although other information cue inventories have been reported in the
literature, these studies have focused on broad domains and have included website elements
beyond information content (Dholakia & Rego, 1998; Zhang, von Dran, Blake &
Pipithsuksunt, 2001). The WICS is unique in that it focuses specifically on e-tail websites,
and it inventories only information cues. Also, although several studies have created lists
intended to survey information content, none have limited their scope to B2C websites (as
opposed to the broader context of e-commerce sites), and most of the surveys include a
mixture of information content, functionality, and user perceptions. The WICS is specifically
designed to survey only information content within the limited scope of B2C websites.
The second major contribution of this article is to demonstrate one way that website
information cue utilities can be empirically measured on an interval scale prior to a UWI (pre-
UWI). Previous investigations have relied upon post-UWI measurements of user reactions to
websites (i.e., perceived website information quality) in order to assess the effects of
information content. This research adds to these studies by demonstrating how the influence
of information content can be predicted from pre-UWI assessments of information content
utility.
1.6 Overview of this study
Chapter two reviews previous studies that have investigated our research problem
and questions. Chapter three presents our research model and hypotheses. In chapter four,
we explain the experiments and observations conducted to test our hypotheses. In chapter
five, we analyze the results of our experiments. In chapter six we discuss and interpret the
results of our analysis. We conclude with chapter seven and a discussion of the study's
limitations, opportunities for future research, and the study's implications for theory and
practice.
6


7


2.0 LITERATURE REVIEW
2.1 User-Website Interactions (UWIs)
User-website interactions have become a major vein of information systems
research, and they are the central focus of this study. As described by Zhang and Li (2005),
Human Computer Interactions (HCIs) are the interactions between human users and the
computer technology we employ to accomplish a given task within a specific context (Figure
1). In this context, UWI's consist of three major components: the antecedents of UWIs, the
UWIs themselves, and the outcomes of UWIs. Each scenario of antecedents can be viewed
as a unique type of HCI.
Technology
Task/Job
Figure 1: The Zhang and Li (2005) HCI model.
As a general term, "User-Website Interaction" could be used to describe any
interaction between a user and a website. In the context of this research, a UWI refers to a
much more narrowly defined set of users, technologies, contexts, and tasks. Specifically,
when investigating whether and why consumer visits to websites result in, or fail to result in,
purchase behavior, UWI's refer to interactions between website users (individuals) who visit
the websites (technologies) of product or service providers (context) and purchase products
or services (task) as a result of their visit. Furthermore, in this article we focus on
8


consequential user interactions with transaction-oriented business-to-consumer websites.
Here, a consequential interaction refers to activities such as Web-surfing, browsing,
information-seeking, online shopping, or other activities that can lead to on-line transactions.
Transaction-oriented business-to-consumer (B2C) websites are defined as websites that
draw revenue directly from interactions with users, either through direct purchases of
products or services made through the website, or from other behavior (e.g., telephone calls)
that lead to purchases of products or services. In this context, web portals and search
engines are not considered transaction-oriented websites because their income is obtained
from advertisers, not directly from users purchases.
In the following sections, we briefly review the large body of research that has
focused on identifying and measuring the antecedents of UWIs, the UWIs themselves, the
outcomes of UWIs, and how these components interact and affect one another.
2.2 UWI Antecedents
According to the Zhang and Li (2005) model of HCIs, UWI antecedents are the
technologies, contexts, tasks, and user factors that influence a user's interaction with a
specific website. Although antecedent factors may change during a UWI, or as a result of a
UWI (e.g., user computer experience will increase every time a user visits a website),
antecedent factors are independent of the UWI in the sense that they can be identified and
assessed or measured independently of and prior-to the UWI. A significant body of research
has investigated the antecedent factors that influence UWIs and, in turn, the outcomes of
UWIs.
2.2.1 User factors
Research in the fields of HCI and psychology demonstrate that user attributes
significantly influence individuals' decision processes and interactions with computer
interfaces. One of the most significant attributes influencing purchase-decisions and HCIs is
an individuals' experience with the product being considered, or "product experience".
Knowledge of the product domain can help individuals identify highly relevant, informative
9


data sources (Mandel & Johnson, 2002; Sundaram & Taylor, 1998). Generally speaking,
individuals who know very little about a product will perform the least amount of search
because compared to individuals with relatively high product experience, individuals with
limited product experience have less ability to differentiate between relevant and irrelevant
data, and their lack of subject knowledge limits their ability to understand the subtleties of the
problem and relevant information (Klein & Ford 2003; Kumar, Lang & Peng, 2005; Moorthy,
Ratchford & Talukdar, 1997). Research also suggests that product experience has a strong
influence on how information is actually stored in an individuals memory. Consumers with
less experience and knowledge about a product lack a well-defined memory structure. It is
therefore harder for novices to contextualize, categorize, retain, and recall information about
a newly encountered product. Finally, consumers with limited experience in a given decision
domain often over-estimate the effectiveness of their decision making and are unable to tell
whether the information they have acquired is adequate to effectively make decisions.
(Sanbonmatsu, Kardes, Houghton, Ho & Posavac, 2003)
In addition to product experience, users' experience using the internet has also been
demonstrated to strongly influence their behaviors and beliefs (Hoffman & Novak, 1996). In
general, for less experienced users, perceptions of a website's ease of use plays a relatively
large part in determining the users' willingness to return to that site. However, for
experienced users, a website's perceived usefulness has much more influence on use
outcomes than perceived usability (Castaneda, Mufioz-Leiva & Luque 2007). In addition,
experienced Internet users are more fluent than Internet novices in regards to finding,
interpreting, using, and assessing the credibility of Internet search resources (Klein & Ford,
2003; Lazonder, Biemans & Wopereis, 2000). Through experience, individuals gain and
learn skills (Bandura, 1982; Oliver & Shapiro, 1993), and prior research has demonstrated
that tasks that initially seem difficult to complete become feasible after sufficient practice
(Allport, Antonis & Reynolds 1972; Hirst, Spelke, Reaves, Caharack & Nesser, 1980).
Lazonder, Biemans, and Wopereis (2000) found that experienced web users are more
proficient at locating useful sites, but that web usage experience did not significantly influence
10


the ability to find information within a website. Finally, studies have found that web
experience is positively related to trust on the Internet (Gefen, Karahanna & Straub 2003;
McKnight, Choudhury & Kacmar 2002), and the results of a survey by Miyazaki and
Fernandez (2001) found that internet experience is negatively related to perceived risk
toward online shopping.
2.2.2 Task
Most descriptions of task refer to how "directed" verses "exploratory" a website visit
is, and whether the visitor gains utilitarian or hedonic value from the visit. In directed (or
"goal-directed") visits, a consumer comes to a website feeling that they have enough product
information to make a purchase decision and they arrive at the site planning to make a
purchase. In this case, website visitor activity is very focused. In other cases, a visitor may
be considering a purchase, but their product choice has not been made, and the visit is
aimed at searching for information that may prove useful during future purchases. Finally, a
website visit may be purely exploratory. In this case, the website visitor has no purchase or
information acquisition goals in mind. During utilitarian tasks, a visitor seeks to fulfill a goal,
such as finding information or purchasing products or services. During a hedonic task, a
website visitor seeks enjoyment.
Two of the most popular online tasks are searching and browsing (Bodoff 2006,
Chen, Houston, Sewell & Schatz 1998; McDonald & Chen 2006). In the IS literature,
searching usually refers to information-seeking behavior with specified goals (Sutcliffe &
Ennis, 1998; Hong, Thong & Tam, 2004). Browsing is described as undirected and
unspecified navigating through hyperlinks (Bates 2002, Chung, Zhang, Huang, Wang, Ong &
Chen 2004; Hong et al. 2007). Despite these and other attempts to define and describe
tasks in the context of UWIs, only a relative few studies have investigated the influence of
task during a UWI. Hong, Thong and Tam (2007) found that the presence of animation within
a website had a larger (negative) influence during browsing tasks than during search tasks.
Moe (2003) found that directed website visits exhibited very focused shopping behavior, and
11


that web pages for targeted products were often viewed repeatedly during a shopping
session. Also, sessions that focus on knowledge-building tend to take more time and focus
on information-related website pages. Novak, Hoffman, and Duhachek (2003) found that
web users whose purpose was experiential (hedonic) were more likely to experience flow
than goal-directed users. Hazzenzahl, Kekez, and Burmester (2002) asked experimental
groups to either "have fun" at a site or to find specific information. For the task-oriented
group, appeal was strongly correlated with perceptions of the website's usability, but that
correlation was not found for the fun group.
2.2.3 Context
Zhang and Li (2005) defined context as the setting and constraints within which tasks
are carried out, and their significance for performing and completing a task. Zhang and Li
(2005) identified six contexts: organizational or workplace setting, market place (where
commerce, banking, and marketing take place), home setting (which includes issues of PC
adoption and user behavior), social environment, cultural context, and other contexts.
To date, only a few studies have investigated the effects of context in the realm of
UWIs. One of the most consistent and significant findings reported has been that context
strongly moderates the effects of aesthetic design factors. Specifically, the effects of website
aesthetics are greatly reduced when visitors act in a goal-oriented context (van Schaik and
Ling 2009, Ben-Bassat, Meyer, and Tractinsky 2006).
2.2.4 Technology factors
Most investigations of UWI antecedent factors have focused on technology factors.
Quite often, in cases where other factors (human, context, task) are studied, those factors are
treated as moderators of the effects of technology. We posit that although human, task, and
context factors are worth understanding and researching, they are largely beyond the control
of e-tail practitioners, and trying to influence visitors' computer or web use experience,
individual dispositions, etc. is not a feasible tactic. However, website hosts do have a large
12


amount of control over the technology that website visitors interact with. Consequently,
technology factors have been, and continue to be, the focus many investigations.
2.2.5 Website design
Several strains of research have investigated website design. Recent studies have
investigated optimal website organization and structure (Chen1997, Mills et al. 2002, Patel et
al. 1998, Janiszewski 1998), the use of intelligent navigation assistants (Neerincx et al.
2001), attempts to create ease-of-navigation guidance and standards (Furnas 1997, Neerincx
et al. 2001, Nielsen 2000, Lohse & Spiller 1998), comparisons of navigational layouts (Park
and Kim 2000, Pratt et al. 2004), and the effects of various design elements or dimensions on
the outcomes of UWIs (Lavie & Tractinsky 2004).
Studies consistently report that website aesthetics have a significant influence on
UWI outcomes. Schenkman and JOnsson investigated the relative effects of four website
aesthetic dimensions: beauty, mostly illustrations versus mostly text, overview and structure.
Of these dimensions, beauty was the best predictor of visitors' overall impressions of the
websites. Cyr, Head, and Ivanov (2006) investigated websites viewed from mobile (cell
phone) browsers and found that visual design aesthetics significantly influenced user
perceptions of website usefulness, ease of use, enjoyment, and loyalty intentions. Robin and
Holmes (2008) found that when identical content was presented at websites with more
aesthetically pleasing appearances versus less aesthetically pleasing appearances, the
content at the more aesthetically pleasing sites was more credible than the content of the
less aesthetically pleasing sites. There is also evidence that poor design has negative
consequences for website performance. Everard and Galletta (2005) found that website
visitors' detection of errors, poor style, and website incompleteness were inversely related to
those visitors' perceptions of website quality, trust, and intention to purchase from the
website. Finally, studies indicate that aesthetic assessments are formed relatively quickly
and tend to be relatively long-lasting. Results from Lindgaard et al. 2006 showed that users
13


make almost immediate judgments (within 500 milliseconds) of a website's attractiveness,
and that the perceptions formed during brief exposures to a website were strongly correlated
with those user's perceptions after viewing the site for extensive amounts time. Tractinsky et
al. (2006) also observed this trend when they found correlations between assessments
reported after 500 milliseconds and 10 seconds of exposure.
The high volumes of information presented at websites and the non-linear nature of
the web can make it difficult for website visitors to form cognitive models of website
information structures (Chen and Macredie, 2002, Boechler, 2001, Otter and Johnson, 2000).
For example, Danielson (2002) found that the presence of a constantly visible textual list of
website content was associated with users searching deeper within the website structure,
less use of the "back" button, greater hierarchical jumps within the website structure, and less
task abandonment. Kang and Kim (2005) found that navigation difficulty significantly
influenced website user perceptions of website entertainment, informativeness, and attitude
toward the site. Roth et al. (2010) found that website visitors have consistent expectations of
where to find various information cues within common types of websites (which they identified
as online shopping, news portals, and company web pages). The results from Roth et al.
(2010) and others (Head, Archer, and Yuan 2000, Oulasvirta 2004, Luna, Peracchio, de Juan
2002, Otter and Johnson 2000, Katsanos, Tselios, Avouris 2008) suggest that successful
website navigation and organization schemes are not so much a function of a set design
pattern (font size, colors, cue placement, etc.) as they are the results of successfully "fitting"
website navigation and organization to users' "mental models" (mental knowledge
representations) of the information provided at the website.
2.2.6 Information Content
As previously mentioned, high quality information content has consistently been
identified as a dimension of successful e-commerce websites. Knowing this, two questions
logically present themselves: What is information content, and what constitutes "high quality"
information content?
14


Resnik and Stern (1977) defined information content as "cues that enable viewers to
better achieve their own personal sets of purchase objectives." (p. 50). Information cues are
the information points that allow a consumer to differentiate between products or otherwise
make a more informed decision. Overwhelmingly, the information cues investigated in the
literature refer to discrete, explicit information cues included within a websites copy or media
content. In the context of this research, we define information cues as these discrete
elements. We do not attempt to measure or account information that may or may not be
implied through website design, organization, or visual content (Kirmani and Rao 2000). For
example, though a picture of an electronic device (such as a GPS) in use under extreme
conditions (e.g., rain) could be used to make inferences about the quality of the device, very
few IS studies have investigated these implicit cues or their effects, and we do not attempt to
account for such inferred messages here.
2.2.7 Information cues in the IS literature
Three types of studies have dominated recent investigations of website information
content. In the first type of study, various information content has been cataloged as part of a
larger attempt to describe a website or to identify the relative importance of various website
dimensions (information content, website design, service quality, etc.) For example, the
website evaluation model presented by Zhang and von Dran (2001) accounted for several
information content cues whose presence or absence could be objectively assessed. Song
and Zahedi (2005) created a list of 46 web design elements, of which 27 were specific
information cues. In an experimental analysis, the 29 most important web design elements
(of which approximately 25 were information cues) were manipulated across 32 experimental
treatments. Their results found that the information cues presented within their experimental
website treatments were significantly related to website visitors' purchase intentions and their
attitudes about the website. Zhang et al. (2001) compiled an inventory of 77 website
elements and compared their relative importance across six common web domains. Of the
77 elements they inventoried, approximately 20 were information cues, (e.g.,
15


Advertising/Lack of advertising, 'Customer reviews, responses and input, Multimedia,
Product and service description, etc.). The other items of the survey generally assessed the
presence of various functions ('No broken links, Printable/downloadable) or users reactions
to the site ('Interesting content', Informative, Intuitive interface, Appropriate Level of
content). While no specific information cues rated within the top 5 most important elements
in any domain, Completeness/Comprehensiveness of Info ranked as the first or second
most important element in four of the six domains examined. A 1998 study by Dholakia and
Rego surveyed the presence of 27 information cues and design features across 272
homepages. They used regression to examine which elements correlated with site popularity
(they used daily hit-rate as a proxy variable for site popularity). Unlike the other two studies,
their results did not find a significant relationship between website information quantity and
website popularity.
In the second type of website information content study, researchers have
investigated user perceptions of information content. As previously mentioned, the relative
ease of providing information content on the Web presents e-tailers with a paradox- they
must provide visitors with enough information to make decisions, but not so much information
as to instigate information overload (Lee and Lee 2004, Fasolo et al. 2007). Hence,
investigators have sought both objective and subjective measures of the amount of
information that websites present.
Simple subjective measurements ask users to rate the amount of information they
find at websites. Other studies have attempted to objectively measure information quantity
by counting information cues. For example, Chen et al. (2009) measured information load as
the number of brand and product options. The results of their study indicated that higher
quantities of website information content caused users to experience information overload.
Lee and Lee (2004) measured information content by counting the number of attributes and
the distributions of those attributes presented by each product option in their experimental
treatments. The results of their study indicated that increasing the amount of information
presented to decision-makers obfuscated their abilities to identify dominant product options.
16


In addition, Keller and Staelin (1987) measured information quality, defined by how useful a
point of information is to a consumer in determining an options utility, and information
quantity, defined as the number of attributes describing a choice option. They found that
increasing the quantity of information, while holding information quality fixed, increased their
subjects information overload. Conversely, increasing information quality while holding
information quantity fixed did not increase information overload.
A large body of research has investigated the factors that constitute information
quality (see Table 2.1). Several dimensions of information quality, including perceived
accuracy, completeness, relevance, and timeliness have consistently been reported in the
literature. In this context, completeness describes the quantity, breadth and depth of
information content. Accuracy refers to how well a recorded value represents the actual
value (Ballou & Pazer, 1995; Fisher, Chengalur-Smith & Ballou, 2003). Clarity describes the
consistency, interpretability, and overall understandability of information (Wang & Strong,
1996). Relevance indicates users perceptions of interestingness, informativeness, and fit-to
task. Finally, timeliness gauges the currency of data (Ballou & Pazer, 1995).
In the third type of website information content study, researchers have investigated
how website information content affects the outcomes of UWIs. For example, studies have
investigated the effects of assurance seals on website hit rates (Dholakia & Rego 1998) and
trust (Rifon et al. 2005). Others have investigated how customer feedback mechanisms
influence outcomes such as trust (Ba and Pavlou 2002), perceived risk, and price premiums
(Ba and Pavlou, 2002). The following sections discuss UWI outcomes and how they are
affected by website information content.
17


Table 2.1:
Previously identified dimensions of information quality.
Authors Focus Factors
Liu and Arnett (2000) Accuracy; Timeliness; Relevance; Flexible Information Presentation; Customized Information Presentation; Price Information; Product/Service Comparability; Product/Service Differentiation; Complete Product/Service Description; Perceived Information Quality on Product/Service; Satisfying Ethical Standard; Support Business Objectives
Wand and Wang (1996) Data Quality accuracy; completeness; consistency; timeliness
Fisher, Chengular-Smith and Ballou (2003) Meta-Data Accuracy
Wang and Strong (1996) Data Quality Believability; Accuracy; Objectivity; Reputation; Value-added; Relevancy; Timeliness; Completeness; Appropriate Amount of Data; Interpretability; Ease of Understanding; Representational Consistency; Concise Representation; Accessibility; Access Security
DeLone and McLean (1992) Information Quality Importance; Relevance; Usefulness; Informativeness; Usableness; Understandability; Readability; Clarity; Format; Appearance; Content; Accuracy; Precision; Conciseness; Sufficiency; Completeness; Reliability; Currency; Timeliness; Uniqueness; Comparability; Qantitativeness; Freedom from Bias
Barnes and Vidgen (2000) Information Quality Relevant; Appropriate Level of Detail; Content is Easy to Read; Appropriate Format; Content is Easy to Understand; Updated Regularly; Reliable
Loiacono (2000) Information Quality Accuracy and Precision; Reliability; Timeliness and Currency; Completeness; Consistency
2.2.7 UWI outcomes
In the context of this paper, the term UWI outcome refers to the thoughts, emotions,
and behaviors that a user experiences as the result of a UWI. In an attempt to identify,
define, and differentiate the most commonly reported UWI outcomes in the peer-reviewed
literature, Hasley, Gregg, and Shroyer (2010) performed a systematic literature review of the
peer-reviewed articles that reported the results of empirical studies of UWI outcomes
between 1993 and 2007. The results of the review found that although scores of user-
18


website interaction outcomes have been reported in the peer-reviewed literature, most of
those outcomes directly represent one of eight high-level user-website interaction outcomes
(confirmation/disconfirmation, trust, perceived risk, engagement, purchase intentions, actual
purchase behavior, repeat website visit intention or behavior, repeat purchase intention or
behavior) or a subdimension of those outcomes. This study investigates how information
content influences perceived website quality, trust, perceived risk, engagement, and actual
purchase behavior.
2.3 Previous investigations of website information content and UWI outcomes
The relationships between website information content and UWI outcomes have
been a major field of study in the fields of information systems (Lynch and Ariely 2000, Gregg
and Walczak 2008), marketing (Lee, Love and Han 2008, Beatty and Foxx 2004, Klein 2003),
and consumer psychology (Park, Lennon and Stoel 2005, Fiore, Jin and Kim 2005).
Typically, these investigations have fallen into one of two broad types of investigations.
In the first type of study, researchers have investigated how specific information cues
influence UWI outcomes. Song and Zahedi (2005) found that five categories of web-site
information cues (defined by them as promotion, service, external interpersonal, ease of use
and navigation, and purchase facilitation) influence visitors' beliefs about both the website
and their own self-efficacy. Positive feedback ratings were also associated with higher price
premiums for 13 of the 18 products the authors investigated. One of the most commonly
studied types of information cue has been assurance seals. Assurance seals are cues that
attempt to allay website visitor fears by offering third-party assurances of transaction security
(e.g., Verisign), privacy assurance (e.g., TRUSTe), or transaction integrity assurance (e.g.,
BBBOnLine Reliability) (Hu, Wu, Wu, and Zhang, 2009). The results of previous studies that
have investigated the how assurance seals affect UWI outcomes are equivocal. Kovar, Burk,
and Kovar (2000) found that website visitors who noticed the seal or who had been exposed
to WEBTRUST advertising had more positive expectations of their interaction with the site
and stronger purchase intention to make an on-line purchase than did their counterparts.
NOteberg, Chistiaanse, and Wallage (2003) found that the presence of a third-party
19


assurance seal at a website significantly increased purchasing likelihood and reduced
visitors' concerns about privacy and transaction integrity.
In the second type of study, researchers have used information quality (or a variant of
it) to measure how website visitors interpret information content, and the relationships
between information content quality and other UWI outcomes. For example, de Wulf,
Schillewart, Muylle, and Rangarajan (2006) conceptualized website content as credibility,
currentness, relevance, and sufficiency. Their results indicated that adequate, high-quality
information content was positively associated with higher levels of pleasure, satisfaction, and
trust. Ha (2004) found that perceived brand trust was positively associated with the quality of
information offered by the site. Agarwal and Venkatesh (2002) found a significant positive
relationship between website content quality and website visitor satisfaction. Lee, Love and
Han (2006) found a strong relationship between information quality (understandability,
readability, usefulness, clarity, and relevance of information) and website user satisfaction.
2.4 Gaps and opportunities
The large body of high-quality studies reported in the peer-reviewed literature since
the commercialization of the public Internet have demonstrated that relevant, comprehensive,
and timely information content is a critical component of successful B2C websites. Several of
these studies have also demonstrated that specific website information cues can be
identified, and that the effects of these information cues can be meaningfully measured and
assessed. As a result, a large set of heuristics for managing website information content
have evolved. However, studies also demonstrate that the effects of website information
cues are dependent upon human, task, and context factors (Jahng, Jain, Ramamurthy et al.
2006). Consequently, the external validity of many website information management
heuristics are often unknown or substantially limited.
As we mentioned in the previous section, the website information quality construct
has been used to assess how the value and effectiveness of a specific website's information
content is assessed by visitors to that site. However, for researchers and practitioners, the
usefulness of the information quality construct is subject to at least two significant limitations.
20


First, information quality has only been used to make global assessments of a
specific website's information content. The current definitions and operationalizations of
information quality were not designed to make meaningful measurements or comparisons of
the quality or value of individual information cues or cue categories. Second, assessments of
a specific website's information quality can only be made after visitors are exposed to the
website. As we will discuss in the next section, this means that perceptions of the website's
information quality may be influenced by factors other than the information cues presented at
the site (e.g., the website's design, organization, etc.)
This research posits that well-established theories and practices from the field of
utility theory can help managers make empirically-based decisions about website information
content management. In the simplest sense, utility is a subjective assessment of value or
preference. In economics, utility has been used to assess decision-makers preferences
between two or more possible investment options. For example, an investor may be asked to
choose between a "safe" investment with a guaranteed payout, or a "risky" investment option
in which one of two outcomes are possible: the relatively low probability of a very high
financial gain, or the relatively high probability of a relatively minor financial loss (see Table
2.2).
Table 2.2: An example of expected financial payouts
Possible outcomes Expected Value
"Safe" investment Guaranteed gain of $10 $10
"Risky investment 20% chance of a $100 gain 80% chance of a $5 loss (0.20 $100) (0.80 -$5) = $19
In such decisions, the mathematically derived "expected value" of each option is clear, and all
"rational" individuals should choose the same option- the one with the greatest expected gain
(or lowest expected loss). However, in reality, factors such as risk tolerance and the relative
gain or loss incurred from the investment may make the guaranteed payout from the safe
investment a more attractive option than the uncertain payout of the risky investment, even
21


though the expected value of the safe investment is lower than the expected value of the
risky investment. Alternatively, some investors may prefer the possibility of a high financial
payoff, even though the probability of the payoff is relatively low in comparison to the
relatively high probability that they will incur a financial loss. In this and other situations
where preference is not dictated solely by "rational" assessments of possible financial gain,
utility can be used to account for both objective and subjective assessments of value.
In the marketing field, utility has been used to assess preferences between product and
service options in instances where value assessments are not a function of financial gain or
loss (e.g., spending the afternoon visiting a park versus sunbathing at the beach). In this
study, we use a hybrid method of conjoint analysis called maximum differential scaling to
calculate the information utility of various categories of website information content. We also
investigate whether calculated website information cue utility can be used to make
empirically-based predictions about UWI outcomes, including website information quality,
website design quality, engagement, trust, and perceived risk. This research contributes to
UWI theory by adding a new construct (pre-UWI information utility) to previous models that
have only measured user perceptions of website information content after an actual UWI.
This study also moves managers towards empirically based (as opposed to heuristic based)
predictions of expected relationships between website information content and UWI
outcomes. Finally, the information utility measurement method demonstrated in this research
(Maximum Differential Scaling) is relatively cheap and easy to implement, and the results are
simple, concrete, and easy-to-interpret. Hence, this research is valuable not only to
researchers who are well-versed in the UWI literature, but to general practitioners as well.
22


3.0 THEORETICAL FRAMEWORK
Within the Zhang and Li (2005) framework, HCI antecedent factors are postulated to
influence HCIs and their outcomes. As our literature review demonstrates, many important
antecedent factors and their effects on UWIs and UWI outcomes have been identified.
Results also indicate that strong relationships exist between website information content and
UWI outcomes, and that understanding these relationships may provide a significant
competitive advantage to e-tail practitioners (Song and Zahedi 2005). This research
postulates that pre-UWI calculations of information utility may be an antecedent factor
capable of predicting UWIs and their outcomes.
One common method of investigating the effects of information content on UWIs and
their outcomes is to create different treatments of a website (each with differing sets of
information cues), expose users to those treatments, and then use surveys or other
techniques to gauge users' reactions to the treatments (Hassenzahl & Ulrich, 2007). In
theory, if possible confounding factors (design, task, context, etc.) are held constant, it is
reasonable to assume that differences in visitor's perceptions of the different treatments are
due to the different information cues presented by each treatment. Although this method
does provide meaningful guidance, it still requires exposing users to websites, and,
consequently, the results are often influenced by how the information is "framed".
One of the most significant factors influencing decision-making processing and
outcomes is how a choice problem is framed (Bettman et al., 1998; Payne et al., 1992) The
theory of invariance asserted that a rational decision maker would be able to consistently
define product attribute information that clearly helped them identify how to best achieve their
desired goals, regardless of how product options were presented to them (Tversky and
Kahneman, 1986). However, the more recent theory of constructive preferences
acknowledges that decision makers are not perfectly rational, and research results from the
last 30 years have indicated that decision processes and outcomes are extremely sensitive to
how choice decisions are framed (Bettman et al., 1998; Payne et al., 1992; Slovic 1995;
Tversky and Kahneman 1986; Tversky et al. 1988; Tversky and Kahneman, 1981;). For
23


example, recent experimental results from Haubl and Murray (2003) indicate that decision
outcomes can be significantly influenced by adjusting which attributes an electronic
recommendation agent presents to subjects for consideration. Payne et al.s (1992) overview
of constructive decision making illustrates that how a given problem is framed can affect
attribute and brand salience, evoke different decision making processes, and ultimately lead
to different decision outcomes. Ultimately, because of the influence of framing, post-UWI
assessments are largely unable to isolate the effects of information content from framing
(figure 3.1). For example, Gregg and Walczak (2008) found that e-bay business websites
with higher e-image (higher quality information content and website design) were able to
command a significant price premium during sales. However, as the authors noted, their
analysis did not isolate the effects of information quality from other factors.
Framing
(Design, organization^.
ra Information 1 \ V, / X UWI
un Content \ \ \ ' /V Outcome
i)
Figure 3.1: Post-UWI assessments of information quality are not able to separate the
effects of information content from the effects of how that information is
framed.
This research posits that the calculated information utilities of specific website
information cues can be empirically assessed prior to a UWI, and that there are several
advantages to making such assessments. First, because expected information utility
assessments based on statistical methods are not influenced by website design factors such
as site architecture or organization, calculated information utility allows investigators to
assess the values of information cues in a context that is largely independent of how the cues
are framed, thus isolating their true effect from possible confounding factors such as website
design or organization (figure 3.2).
24


Information
Content
Figure 3.2: Pre-UWI methods of assessing information utility are able to isolate
information content from how that information is framed.
A second major advantage is that by using modern software packages, calculated
information utility can be assessed relatively cheaply and simply before any resources are
invested into creating experimental websites. From an operational perspective, this means
that calculated information utility can provide empirical guidance from which investigators can
create their experimental websites or alter previously existing ones. Third, when information
cue utilities are empirically measured (calculated) using statistical methods, it reduces the
need to extrapolate the values of specific information cue categories from the results of other
studies. For example, as discussed in section 2.2.6, although many studies have
investigated the effects of various assurance seals at e-tail websites, the results of these
studies are indeterminate. Enabling website managers to empirically assess the relative
contributions of all the information cues included within a specific website may greatly reduce
the need for managers to extrapolate guidance from the literature.
3.1 Utility
Utility is a subjective assessment of value, desirability, or satisfaction provided or
derived from a good, service, or experience (Hair et al., 2006). As was discussed previously
in section 1.3, the subjectivity of utility is rooted in the fact that even if a seemingly objective
uwi
Outcome
25


measurement of value can be assigned (e.g., a financial value), the emotional value assigned
to an object or experience can significantly influence an individual's option preferences.
Hence, one of the major advantages of the utility construct is that utility assessments are
based upon decision makers perceived assessments of overall (emotional as well as
material) gains and losses, rather than purely "rational" assessments of gain or loss. In
addition, utility can facilitate the assessment of option attributes that have no meaningful
financial equivalent (the value placed upon a large black stereo versus a small, silver-colored
one).
In the basic additive model of utility (Huber 1974), overall utility is the sum of the
values, or "part-worths", that a user places on each attribute or factor that influences the
global assessment of utility, such that:
Utility = part-worth of factor 1 + part-worth of factor 2 + part-worth for factor n
By using statistical methods of assessing information utility, it is possible to assess
the individual part-worths of individual information cues, and, theoretically, the overall
information utility of a website that includes those cues. Additionally, if the part-worth utility
assessments are made using an interval scale, then the relative contributions of individual
information cues can be meaningfully compared.
3.2 A new model
Within our given theoretical framework, we propose a model (figure 3.3) to test the
impact of information utility on UWIs and UWI outcomes. In our proposed research model,
pre-UWI measurements of information utility are posited to be significantly related to
measurements of the actual UWI (time spent at the website, depth of search) and
consequent UWI outcomes (PWQ, trust, engagement, perceived risk, and actual purchase
behavior).
26


UWI Outcomes
UWI
Antecedent j
Factors ;
Context ;
Task ;
Human j
Technology I
-Information ............p*
Content j
-Pre-UW! i
Information ;
Utility ;
Pre-UW!
UWI
Post-UWI
Figure 3.3: Previous studies have focused on post-UWI measurements.
The UWI model investigated by this research is rooted in the Zhang and Li (2005)
model in that it studies a specific type of HCI and specific HCI antecedents and outcomes.
The new model is motivated by the results of previous studies that have reported links
between website information content and UWI outcomes (Ba & Pavlou, 2002; Wang, Beatty
& Foxx, 2004). Our method of assessing information content is based on previous studies
that have surveyed website information content (Huizingh, E.K.R.E; Liu & Arnett, 2000;
Ranganathan & Ganapathy, 2002). Our conception and measurement of information utility is
based on previous research that has explored the utilities of various website attributes (Chen,
Hsu & Lin, 2009; Van Ittersum, Pennings, Wansink, van Trijp, 2007). Our conceptualizations
and measurements of UWI outcomes are based upon previous investigations of UWI
outcomes (Gregg and Walczak 2008; Hasley et al., 2010)
27


3.3 Hypotheses
In this research, we examine the relationships between pre-UWI measurements of
website information utility and post-UWI measurements of UWI outcomes. The specific
relationships under investigation are presented in the following hypotheses.
3.3.1 Information content
Knowing that pre-UWI measurements of information utility and post-UWI
measurements of information quality are both measurements of the users' assessments of
the value-added by website information content, we hypothesize that there is a significant
relationship between information utility and website information quality. Further, because
previous studies have found significant relationships between post-UWI measurements of
information quality and other UWI outcomes, we hypothesize that there will also be significant
relationships between information utility and other UWI outcomes.
3.3.2 Website Quality
As the Internet evolved and the World Wide Web became a major channel of
commerce, researchers began to investigate consumers reactions to websites. Early studies
explored how website quality factors such as design, information content, and functionality
influenced user perceptions and behaviors (Huizingh 2000, Liu and Arnett 2000, Zhang and
von Dran 2000). In their 2001 paper, Barnes and Vidgen (2001) described and demonstrated
the WEBQUAL instrument- a version of Parasuraman et al.s (1998) SERVQUAL instrument
modified to accommodate the unique aspects of the Web environment (e.g., lack of a
physical store, lack of tactile contact with products, lack of sales personnel, lower information
search costs, etc.) In 2007, one of the most rigorous instruments for measuring website
quality (also called WebQual), was published by Loiacono, Watson, and Goodhue (2007).
Table A-1 in Appendix A summarizes several studies that have developed and validated
instruments intended to measure website quality.
28


The website quality construct borrows strongly from the Technology Acceptance
Model (Davis 1989) in that one of its major objectives is to measure the usefulness and
usability of websites. However, as Loiacono et al. (2007) pointed out, website managers
and developers require more guidance than usefulness and usability. Hence, perceived
website quality models often identify web-specific sub-dimensions of usability and usefulness
(Cao, Zhang and Seydel 2005). For example, perceived website quality models often
measure usefulness in terms of information quality, functional quality, information fit-to-task,
relative advantage; usefulness is often measured in terms of navigation quality, site
organization, etc.
In this study, we examine the possible relationships between information content and
two of the most commonly identified dimensions of website quality: information quality and
design quality. Although several PWQ instruments have prescribed that various cues should
be included within high-quality websites, the relationships between specific information cues
and PWQ have received limited attention in the UWI literature to date. Gregg and Walczak
(2008) demonstrated that various information attribute values and information presentation
schemes were positively associated with a more professional e-image, increased willingness
to transact, and higher final prices paid within the context of an on-line auction.
As previously noted, pre-UWI measurements of information utility and post-UWI
measurements of perceived website information quality are both measurements of the value
added by a website's information content. However, the possible relationships between
information utility and information quality have not been investigated in the UWI literature.
We posit that adding information to a website should make a website's information content
appear more complete, accurate, timely, and relevant as long as the additional information
significantly increases the website's information utility. On the other hand, if low or negative
utility information is added to a website, the additional information may serve to distract or
confuse visitors, or even lead to information overload. Specifically, we propose the following:
29


Hypothesis 1a: There is a positive relationship between website information utility and
perceived website information quality.
The second major dimension of website quality this study investigates is design
quality. We posit that website information content utility can significantly influence several
dimensions of website design. Results from Ben-Bassat, Meyer, and Tractinsky (2006) found
that websites perceived to be more usable by visitors were also considered more
aesthetically pleasing. In addition, Gullikson, Blades, Bragdon, McKibbon, Sparlin and Toms
(1999) demonstrated that website organization and architecture influenced visitors ability to
navigate the site and find information. Teevan, Alvarado, Ackerman, and Karger (2004)
found that even when website visitors know their information needs, they are more likely to
search for that information by navigating through a site (orienteering) than by using a within-
site search engine. Consequently, website navigation design and information organization
can greatly influence user task performance and user perceptions of the website. Also, as
previously mentioned in section 2.4, successful websites fit their organization and
architecture to "fit" visitors cognitive models of the information being presented. Therefore,
we expect that websites that provide high-utility information cues will be less likely to induce
users to engage in fruitless searching for relevant content and will, consequently, be
perceived as easier to navigate and better organized.
Hypothesis 1b: There is a significant relationship between website information utility and
perceived website design quality.
3.3.3 Engagement
Website visitors often experience strong emotional and cognitive responses during a
UWI. In this study, we use the term engagement to describe these various patters of
emotions, thoughts, and the user's state of mind following their experiences during and after
a UWI.
30


One of the principle attributes of engagement studied in the literature is flow. Flow
has been studied in traditional business channels, as well the context of athletic and cognitive
performance (Jackson, 1996; Hawkins & Hoch, 1992). In its most basic sense, flow is a
latent construct that describes immersion within a task. Empirical studies have demonstrated
that individuals experiencing flow exhibit several consistent states including arousal, focused
attention, control, telepresence, affect, elaboration, time distortion, and playfulness. Arousal
describes a state of heightened awareness and involvement with a task (Novak et al., 2000).
Focus of attention refers to the selective allocation of cognitive resources (Huang, 2006),
accompanied by a heightened differentiation of relevant from irrelevant stimuli. Perceived
control describes an individuals perception of having the knowledge, resources and
opportunities required to complete their tasks (Huang, 2006; Koufaris & Hampton-Sosa,
2002). Telepresence refers to a sense of virtually experiencing an environment (Fiore, Kim,
and Lee, 2005; Klein, 2003). Affect describes the states of pleasure and enjoyment that
occur during the state of flow (Novak et al., 2000). Time distortion refers to the fact that
individuals may lose track of time while they are experiencing flow (Novak et al. 2003).
Playfulness describes creativity and loss of self-judgment or criticalness. Elaboration
describes attempts to integrate or compare current circumstances and information to
previously encountered experiences or information (Tam & Ho, 2005), as well as attempts to
cognitively process counterarguments, source derogation, support arguments, or source
bolstering. (Yoo & Kim, 2005)
The results of previous research indicate that there is a positive link between
perceived website information value and engagement. Lin et al. (2005) found a significant
relationship between confirmation and perceived playfulness. De Wulf et al. (2006) found a
significant link between adequate, quality information content and perceived pleasure. Chen,
Clifford, and Wells (2002) found that information content, entertainment, and website
organization were strong predictors of attitude toward the site, and that information content,
entertainment, and website organization were highly correlated. In general, websites that
present high-value information content should increase both the utilitarian and hedonic value
31


of a website (Fiore, Jin, and Kim, 2005). We posit that adding high utility information cues to
a website should stimulate user attention and arousal as users are given more useful and
interesting information to consider. Users should also experience increased elaboration as
they become more aware of website and product attributes and capabilities. Alternatively,
users should experience frustration and possibly begin to withdraw from a website interaction
when low or negative utility information is added to a website.
Hypothesis 2: There is a significant positive relationship between website information utility
and engagement.
3.3.4 T rust
Trust has been widely studied across the social sciences, and it has received
considerable attention within the e-commerce domain, especially in regards to its affects on
customer behavior on the Internet, where the effects of trust (or mistrust) are heightened due
to the relative ease with which vendors can act in an opportunistic manner (Reichheld &
Schefter 2000). Several studies have found that consumers concerns about website
trustworthiness are a major obstacle to consumers willingness to share personal information
or engage in on-line transactions (Dinev & Hart 2006), and that websites that establish high
levels of trust are able to demand higher prices than websites that elicit comparatively low
levels of trust (Ba and Pavlou 2002, Reichheld and Schefter 2000).
In the context of e-commerce, trust describes a trustors willingness to be
vulnerable to a trustee in an environment of uncertainty (Gefen et al. 2003). However, in their
1995 paper, Mayer et al. pointed out that trust, per se, is difficult to measure. For example, a
survey question such as How willing are you to be vulnerable to the company hosting this
website?" is probably too existential to elicit meaningful answers from survey participants.
Consequently, various trust antecedents are often used as proxy variables for trust. Mayer
identified four central antecedents of trust: ability, benevolence, integrity, and individual trust
propensity. Individual trust propensity, or disposition to trust (McKnight et al., 2002), refers
32


to an individuals innate tendency to trust. Perceived ability refers to a trustors
determination of a trustees ability to provide the goods or services they offer in a safe and
efficient, to provide assistance if required (i.e., for product returns), and to manage
competently any personal and financial information the user provides. Perceived
benevolence describe trustor's perceptions of a trustees intentions to act in the best
interests of both parties and refrain from engaging in opportunistic behaviors. Distinct from
benevolence is integrity, which is a perception that a party will adhere to acceptable
principles and abide by the rules of an agreement. Together, ability, benevolence, and
integrity are often viewed as a trusting beliefs construct.
Results from previous studies indicate that website information content and perceived
information quality can significantly influence perceptions of trust. For example, results from
Ba and Pavlou (2002) suggested that at on-line auction sites, a greater number of positive
ratings will induce higher levels of trust, but only in the absence of negative ratings. Wang,
Beatty, and Foxx (2004) found that security disclosures and seals of approval from neutral
sources were positively related to increased trust. However, as discussed by Hu, Wu, Wu,
and Zhang (2009) in their review of the effects of third party assurance seals on trust and
other UWI outcomes, the relationships between information cues and trust are often
equivocal and difficult to extrapolate from one situation to another.
Previous studies of trust have investigated how trust between interacting parties is
established, and study results provide theoretical foundations for expecting website
information content to influence consumer's feelings of trust. According to Wang, Beatty, and
Foxx (2001) "cue-based trust" is built by credible information cues that signal the intention
(benevolence and integrity) or capability (competence) of a party. Hence, we posit that
providing high-utility information content in a well-designed user interface should increase
users' perceptions of seller ability. On the contrary, providing low or negative utility
information may have the effect of confusing the visitor, who may assume that a seller who is
unable to fulfill the users' information needs will be equally unable to fulfill their product or
service needs, resulting in lower perceived ability. A visitor could also perceive low-utility
33


information content to be a sign of a seller's carelessness or ambivalence toward customers,
causing a deterioration of perceived benevolence. Hence, we posit that there will be a
positive relationship between pre-UWI measurements of website information utility and post-
UWI measurements of trust. McKnight, Choudhury, and Kacmar (2002) concluded that
consumers form trusting beliefs based upon the "structural normality" of a web domain, and
that consumer perceptions of structural normalcy are largely based upon whether a website
environment is appropriate, properly ordered, and favorable for doing business. We posit
that high-quality information cues that represent the information cues that are commonly
presented within a website domain will increase website visitors perceptions of situational
normality and, therefore, their feelings of trust.
Hypothesis 3: There is a significant positive relationship between website information utility
and trust.
3.3.5 Perceived Risk
According to Mitchell (1999), consumer risk is a subjectively-determined expectation
of loss; the greater the probability of this loss, the greater the risk thought to exist for an
individual. (p. 168). Although its precise definition is still debated (for example, risk defined
as expectation of loss differs from risk defined as pay-off times probability), most
definitions of risk that have been applied to the e-ecommerce domain seek to measure
consumer perceptions of the potential for loss or of not realizing an expected outcome. Dinev
and Hart (2006) call these beliefs about the potential for loss risk beliefs, and we adopt this
term to describe users' beliefs about the magnitude and probability of consequences in
situations involving outcome uncertainty.
It is important to note that perceived risk is not the same as actual (or purportedly
objectively measured) risk (Mitchell, 1999). In fact, research indicates that retail customers
are very poor at assessing actual risk (Grazioli & Jarvenpaa, 2000). Consequently,
34


consumer behavior is much more influenced by perceived risk than actual risk, and most UWI
investigations that study risk as an outcome focus on consumers perceptions of risk.
At this point, we also wish to emphasize the difference between trust and risk. In the
context of e-commerce, trust is a perception about a participant or mechanism in a
transaction, while perceived risk is a perception about the consequences of the transaction
itself (Mayer et al., 1995). For example, a customer may be willing to transact with a seller in
which they do not have a high level of trust as long as the transaction presents relatively low
risk (e.g., when buying a previously-owned CD on eBay). However, this same customer may
be un-willing to buy from that same seller when perceived financial and performance risk is
high (e.g., when buying a used luxury watch) (Ba & Pavlou, 2002). Although trust and risk
can be separated theoretically and operationally, in the context of e-commerce, evidence
suggests that trust and risk have a strong interaction effect (Cho, 2006; Dinev & Hart, 2006,
Jarvenpaa et al., 2000; Pavlou & Gefen 2004). Hence, in practice it is very difficult to study
trust independently of risk without threatening the internal validity of the trust measurement
(Jarvenpaa et al., 2000; Pavlou, 2003; Pavlou & Gefen, 2004).
Many buyer-seller relationships are characterized by an information asymmetry in
which the buyer is dependent upon the seller for product information (Mishra, 1998). The
signaling theory of information economics (Spence, 1973) posits that under conditions of
information asymmetry, a signal is an action that a seller can take to convey information
about unobservable product quality (Rao, Qu, and RueKert, 1999) or skill level (Kirmani &
Rao, 2000). At retail websites, signals occur as the technical aspects of the website (website
design and information content, described in section 2.3). Information signals are most
useful in situations where product or service quality are largely unknown, and can be
assessed only after purchase (Kirmani and Rao, 2000). This research posits that information
cues perform as signals about product quality and seller capability, and that informative, high-
utility information cues will decrease information asymmetry, buyer uncertainty, and
perceived risk regarding the product or service they seek.
35


Hypothesis 4: There is a significant negative relationship between website information utility
and perceived risk.
The hypotheses being investigated in by this study are summarized in figure
Hypothesis_Model.
Figure 3.4: Relationships between Information Utility and UWI outcomes
36


4.0 RESEARCH METHODOLOGY
This research investigates how e-tail website information content can be inventoried,
a method of empirically measuring the relative utilities of various website information cue
categories, and whether relationships exist between a website's calculated information utility
and various UWI outcomes.
4.1 Overview of the studies
To address the research questions posed by this article, two studies (an exploratory
study and a three-phase main study) were conducted in sequence. In the exploratory study,
the Website Information Content Survey (WICS) was developed in order to demonstrate that
investigators can reliably and meaningfully identify which information cue categories are
present at a website, and whether subsets of cue categories are common within certain
website domains. In the first phase of the main study, the WICS was used to identify the
information cue categories commonly presented at a specific website domain (regional
authors). In the second phase, Maximum Difference Scaling was used to calculate the
relative utilities of the information cue categories commonly included within regional author
websites. In the third phase of the main study, specific information cues (pictures or text)
were created to represent each of the information cue categories examined in phase two.
Next, the information utilities calculated in phase two were used as the basis to create four
treatments for each of three regional authors' websites and a 4 x 3 full-factorial experiment
was performed to investigate the relationships between website information utility and
perceived information quality, engagement, trust, and perceived risk.
4.2 Exploratory study: Measuring Website Information Content- The WICS
The goal of this exploratory study was to develop a comprehensive inventory of the
various categories of information cues that are likely to be found within retail B2C websites. In
this context, the category will account for cue that convey a specific type of information. For
instance, any text within the website that quotes a celebrity or expert would be representative
of the category "Product endorsement (Celebrity/Expert)". Using content analysis, we
37


developed an instrument, the Website Information Content Survey (WICS), which allows for
meaningful inventory, analysis, and comparison of the categories of information cues
contained within a website. Content analysis develops a data set based on systematic
coding of documentary evidence (Hodson, 1999; Krippendorff, 1980). The intention is to
systematically create descriptions of qualitative data (in this case, the presence or absence of
information cues).
To identify the categories of information cues (or "cue categories") commonly
presented at B2C retail websites, we began by reviewing the literature and identifying
information cues categories identified in studies that investigated specific information cues,
types of information, and website quality. The literature review identified 64 different
information cue categories that were used in at least one of fourteen prior research studies.
In addition to identifying information cue categories within the literature, the authors surveyed
25 retail websites to identify additional information cue categories that had not been
evaluated in prior research. Ultimately, 88 information cue categories were compiled to
create the Website Information Content Survey. Within the WICS the information cue
categories were grouped into sections (e.g., product information, company information,
etc.) based upon which pages within a website the information cue was typically presented.
For example, the cue categories included in the product information section of the WICS are
commonly presented on web pages that describe products, cue categories included in the
company information section of the WICS are commonly presented on web pages describe
the websites host company, etc.
4.2.1 A demonstration of the WICS
In this study, we demonstrate how the Website Information Content Survey (WICS)
can be used to measure which information cue categories website visitors actually
experience, and how the survey can be used to compare the information content profiles of
various websites.
38


As part of our exploratory study, an information systems class with 21 students at a
mid-sized urban university were asked to perform a content analysis of twenty different
websites using the WICS instrument. The students were asked to determine (by indicating
yes or no) whether the specific information cue categories described in the WICS
instrument could be found on the website they were assigned to assess. The websites
represented five different e-commerce domains: insurance, consumer electronics, travel
(cruises), health care, and specialty foods and were not the ones used to develop the original
information content survey instrument. Each student was given a paper copy of the initial
WICS instrument, the URL of the website they were asked to perform the content analysis
for, and a list of products or services to find information about. Two students were given the
same site in the insurance domain- one of these students was asked to obtain a car
insurance quote, the other was asked to obtain a home insurance quote. The students used
the tools available on the individual sites to locate the information content (e.g. site maps,
menus and site specific search tools).
In order to assess the reliability of the WICS instrument, a primary subject matter
expert (SME) also used the WICS to independently assess each of the target websites.
Following the initial assessments by the students and the primary SME, a second SME re-
evaluated all information content items where the primary SME's assessment differed from
the students assessment. This resulted in a combined student/ secondary SME assessment
that was compared to the independent assessment made by the primary SME to determine
the inter-rater reliability. Inter-rater reliability is the degree of agreement among raters and is
necessary in situations where there is a degree of human judgment involved in the
measurement of variables. The inter-rater reliability of the student/secondary SME and
primary SME assessments were measured by calculating Cohens kappa which is
appropriate when comparing two sets of ratings (see Table 4.1). The inter-rater reliability for
each domain and for the websites overall exceeded the 0.70 criteria indicating the coding is
acceptable (Krippendorff, 1980; Weber, 1990).
39


Table 4.1: Inter-rater reliability scores
Cohens kappa
All Sites Electronics Medical Services Specialty Foods Insurance Cruise Lines
3-rater reliability 0.805 0.783 0.840 0.860 0.721 0.790
4.2.2 Results of the exploratory study
The prevalence of information cue categories found as a result of the content
analysis are summarized in Appendix B. The Table B-1 in Appendix B lists each information
cue category included in WICS, the source of the information cue category (if it was derived
from prior literature), and the percentage of websites in the sample that were found to contain
the information cue category, listed by domain. The frequency of occurrence across domains
offers a simple comparison of which information cue categories are common (and could
initially be inferred to be important) across various domains. Finally, Table B-1 also contains
the overall percentage of sites surveyed that were found to contain a specific information cue
category.
One simple application of the WICS survey is to determine if there are any significant
differences between the information content of sites from different domains. Table 4.2 shows
the results of paired sample t-tests which were used to determine if there were significant
differences in the information cues found for different domains. The t-tests indicate that
there were significant differences for most of the domains examined. The information cues
found on the "electronics" sites were significantly different than the cues on sites for all other
domains except the "cruise line" domain. The "medical services" sites contained information
cues that were significantly different from all domains except "specialty foods", and the
"insurance" sites also contained information cues that were significantly different from all
other sites except for the "specialty foods" sites. The fact that "insurance" and "medical
40


services" were both similar to "specialty foods but not to each other suggest that the
information cue overlap might be for different reasons. For example, most of the "specialty
food" sites and most of the "medical service" sites were for very small companies with limited
websites. Although the insurance sites were all large carriers, insurance sites have no
product to deliver and as such might appear to be missing some product related information
content cues which smaller sites that actually sell products include.
Table 4.2: Paired Sample t-Tests Comparing the Content of 5 Web Domains
(significance)
Electronics Medical Services Specialty Foods Insurance Cruise Lines
Electronics - 0.0000 0.0000 0.0006 0.8791
Medical Services 0.0000 _ 0.1458 0.0047 0.0000
Specialty Foods 0.0000 0.1458 0.1237 0.0000
Insurance 0.0006 0.0047 0.1237 - 0.0008
Cruise Lines 0.8791 0.0000 0.0000 0.0008 -
Insight can also be gained from intra-domain comparisons, which may help identify
information shortfalls, industry trends, competitive advantages, or opportunities. For
instance, understanding how frequently an information cue occurs within a domain (the intra-
domain frequency of occurrence) gives an indication of whether the presence of that cue on
a website is relatively common (or, conversely, unique) across websites in that domain. If a
cue is included at most of the websites within a domain, a website that does not provide that
cue may be perceived as less informative than the competitors who do provide the
information cue.
Our analysis of intra-domain frequencies of occurrence indicates that website
designers may not always understand what data are expected in their particular domain. For
example, our analysis found lists of product ingredients at 3 of the 4 specialty food stores we
assessed. Assuming that a significant number of specialty food customers have food
allergies, a specialty food store that does not offer a list of product ingredients may be at a
41


substantial disadvantage, especially when competing websites do prominently feature
ingredient lists. Practitioners may also need guidance identifying content that is not
appropriate for their site. For example, only one of the health services websites assessed in
our exercise mentioned price (in fact, that site lists only a range of prices: $499-$ 1500 per
eye). While price is always a consideration, prudent health care providers probably do not
want to be seen as differentiating themselves based on price alone.
Although the exact meaning of a given image or string of text (information cue) is
often subjective, the results of the exploratory study indicate that certain information cues
often occur consistently within various e-commerce domains cues, those cues are often
interpreted to have a relatively consistent meaning, and website visitors can often identify
these cues in a relatively consistent manner.
4.3 Main study
4.3.1 Overview
The websites of regional authors (defined as an author who has published at least
three books and who sells less than 1,000 books a year) were chosen as the domain for this
study for three main reasons. First, the purpose of a regional author's website is primarily
limited to marketing the author's books. Advertisements for other products or services are
rarely included at these sites. Hence, the effects of irrelevant information cues that might
distract a user or otherwise confound the effects of the targeted information cues are
negated. Second, unlike some on-line stores that offer hundreds or thousands of products or
services, the number of products (books) and the number of cues describing those products
are relatively limited. Consequently, it is relatively easy to account for all of the information
cues presented within a website, and to limit the effects of information cues that were not of
interest to the study. Finally, the consistency of the information presented at regional authors'
websites means that the same categories of information cues (with different representations
for each author) could be presented at each author's websites.
42


4.4 Phase One: Identifying the cues most commonly presented within a domain
4.4.1 Sample and procedure
To identify potential information content for the experimental websites, 3 subject
matter experts (students and professors) were asked to use the WICS to evaluate the
information content of the websites of each of five regional authors. In addition to the cue
categories listed on the WICS, the subject matter experts were asked to list 1) any
information cue categories that were not listed in the WICS or 2) any information cues that
would be specific to the niche author website domain. Based on the commonality of the cue
categories identified by the subject matter experts (SMEs) and feedback from the SMEs, the
most common information cues included within the targeted five regional author websites
were identified within. Specific instances of the most common information cue categories
(author name, picture, etc.) were then created for each author's site. The information cue
categories examined in the survey are listed in Table 4.3.
43


Table 4.3: The cue categories manipulated at the experimental test websites.
Author website information cue____________
A picture of the author___________________
Book reviews______________________________
A biography of the author (100 to 300 words)
Book synopsis_____________________________
Samples of the book (text that you can read)
Samples of the book (audio that you can listen to)
Upcoming readings/appearances by the author
(location, time)____________________________
Awards for the book or the author
The name of the publisher(s) of the author's books
A list of stores or websites where the book can be
purchased_________________________________________
The author's contact information__________________
A picture of the book cover_______________________
Price
4.5 Phase two: Determining the relative utilities of specific information cue
categories using Maximum Difference Scaling
4.5.1 Sample
The theoretical population for phase two included any individuals who purchase
books online. The sampling frame for phases two consisted of librarians at graduate
students at public libraries and universities. To recruit participants for phase two (the
information cue utility calculation task) of the study, 263 participants (university staff and
graduate students) were recruited. 86 responses were received, and 84 provided usable
responses.
44


4.5.2 Procedure
One of the main objectives of this research is to demonstrate the measurement of the
relative utilities of various website information content cue categories using an interval, rather
than ordinal, scale. One technique that facilitates this is the Maximum Differential Scaling
technique, otherwise known as "MaxDiff". MaxDiff is a measurement and scaling technique
based on the principles of best-worst conjoint analysis (Cohen, 2003). Conjoint analysis is a
decompositional approach in which respondents are presented with various product profile
options and asked to make definite choices of preference between the product options.
Traditional methods of conjoint analysis focus intra-attribute comparisons of attribute levels
(e.g., preference levels for blue, red, or silver color options), but do not allow for inter-attribute
comparisons that would allow managers to assess the relative importance of attributes (e.g.,
the relative importance of color options versus gas mileage), because the scaling of the
attributes is unique to each attribute (Hair et al., 2006). The MaxDiff method permits inter-
and intra-attribute scaling by measuring each attribute preference level on a common, interval
scale (Cohen, 2003).
To determine the relative part-worths of the twelve information cue categories
examined in this study, the Sawtooth Software suite was used to create a website that
presented fifteen different conjoint tasks. Each conjoint task presented 4 information cue
categories and asked the question, "When deciding whether to purchase a book on-line... If
you consider only these 4 features, which is the most important and which is the least
important?" An example screenshot showing one of the conjoint tasks is shown in figure 4.1.
To increase the quality of the task outcomes, we followed guidelines from Sawtooth Software
to ensure the orthogonality of the attributes under study. Orthogonality is a mathematical test
of the independence of part-worth estimates. If part-worth estimates are not independent,
then the use of an additive utility model is called into question because an additive utility
model does not account for interactions between attributes. The minimum number of
questions to ask in order to achieve orthogonality is 3(K/k) where K is the total number of
items and k is the number of items in each set (Cohen, 2003). Since this survey evaluated
45


twelve items, and presented four items in each set, a minimum of (12/4) 3 (nine) questions
were required to achieve orthogonality. Our survey asked fifteen questions to ensure the
orthogonality of the cue categories.
In this study, navigation cues were not manipulated. Although it is possible to present
all relevant information within one long, scrollable webpage, none of the author sites
examined by the subject matter experts used this organization scheme. Also, because all of
the author websites included price, and because of its overwhelming influence (Song and
Zahedi 2005), price information was not manipulated. The prices listed within the
experimental websites were the prices available from Amazon.com or, if a particular book
was not available from Amazon.com, the price available from the book publisher's website
was listed.
When deciding whether to purchase a book on-line.
If you consider only these 4 features which is the most important and which is the
least important?
Most Important Least Important
A list of stores or websites where you can purchase the book
Upcoming readings by the author (locations and times)
Book synopsis (description-summary of each book the author has written)
Samples of the book (text that you can read)
Click the Wext' button to continue
: Next :
Figure 4.1: Each visitor was asked to indicate the most important and least important
cues in sixteen conjoint tasks.
46


4.5.4 MaxDiff Results
The results of the MaxDiff exercise are shown in Table 4.5. MaxDiff assigns average scores
for each cue category from a total of 100 total points possible (the sum of the average scores
is equal to 100). The scores are assigned on an interval scale, so it is possible to assess not
only ordinal differences between the categories, but also the magnitudes of difference
between them.
Table 4.5: Results of the MaxDiff investigation of cue utility
Label Average 95% Lower 95% Upper
Samples of the book (text that you can read) 19.12 18.29 19.96
Book reviews 18.53 17.76 19.30
Book synopsis (description/summary of each book the author has written) 18.38 17.48 19.29
Samples of the book (audio that you can listen to) 9.70 8.15 11.25
A list of stores or websites where you can purchase the book 9.19 7.68 10.71
Awards for the book or the author 7.98 6.78 9.19
A picture of the book cover 7.51 6.15 8.86
A biography of the author (100 to 300 words) 6.69 5.86 7.52
The name of the publisher of the author's books 0.98 0.35 1.60
A picture of the author 0.87 0.49 1.24
Upcoming readings by the author (locations and times) 0.75 0.48 1.02
The author's contact information (e-mail and/or phone number) 0.30 0.17 0.43
4.6 Phase Three: An investigation of the relationships between information utility,
the UWI, and UWI outcomes
4.6.1 Sample
The theoretical population for phase three included any individuals who purchase
books online. The sampling frame for phase three consisted of individuals who have joined
various FaceBook groups and librarians at public libraries.
To recruit participants for phase three of the study, invitations to participate were
posted to poetry and literature FaceBook fan group pages with total group memberships of
47


4,573 as of January, 2010. Additionally, invitations were sent to 1,834 staff members of
public libraries (both local libraries and public university libraries). Characteristics of the
phase three respondents are listed in Table 4.4.
Table 4.4: Phase 3 participant characteristics
Item Value Frequency Percent
Gender (M/F) Male 166 48.1%
Female 179 51.9%
Age 18-22 81 23.4%
23-29 67 19.4%
30-39 84 24.3%
40-49 34 9.8%
50-59 65 18.8%
60-89 15 4.3%
About how long have you been using the internet? 0-2 months 3 0.9%
3-6 months 0 0.0%
7-12 months 1 0.3%
more than 1 yr. 342 98.8%
During the last 2 months, about how many hours a week have you spent online? Less than 1 hour 2 0.6%
1-3 hours 25 7.2%
4-10 hours 80 23.1%
More than 10 hours 348 69.1%
About how long have you been buying online? 0-2 months 15 4.4%
3-6 months 9 2.6%
7-12 moths 12 3.5%
More than 1 year 308 89.5%
In the past 2 years, approximately how many purchases have you made online? None- zero 4 1.1%
1 to 3 35 10.2%
4 to 10 101 29.5%
More than 10 203 59.2%
In the past 2 years, approximately how many times have you purchased books, CD's, or DVD's online? None- zero 47 13.6%
1 to 3 99 28.6%
4 to 10 116 33.5%
More than 10 84 24.3%
Please indicate the highest level of education you have completed. Some high school, no diploma 1 0.3%
High school graduate, GED or equivalent 16 4.7%
professional degree 109 32.0%
some college, no degree 71 20.8%
bachelor's degree 74 21.7%
master's degree 13 3.8%
doctorate degree 57 16.7%
48


4.6.2 Cue creation
Eleven information cues were created for each of the three experimental websites.
The information cues were appropriately customized for each of the experimental websites
(e.g., author names, book lists, book descriptions, author biographies, etc.).
4.6.3 Cue validation
To ensure that each cue used within an experimental website for the study was
representative of the cue category it was intended to represent, two subject matter experts
(different individuals from the subject matter experts who participated in other study phases)
participated in a cue sort activity that was administered through a website designed for the
task. At the website, each of the information cues was presented, and the SMEs were asked
to determine which type of information cue within the WICS the specific cue represents. The
reliability of the cue-sort was 0.90. No cue was found to be consistently un-reliable.
4.6.4 Information cue groupings
Based upon the results of the information utility scores calculated in phase two, three
clusters of information cue categories were identified (high-utility, medium-utility, and low-
utility cues). These groups were clustered because the scores for each of the three highest-
utility cue categories ("samples of the book (text)", "book reviews", and "book synopsis")
received scores of eighteen or above, and together these cue categories accounted for 56.03
of the possible 100 points allotted. The lowest score in the high-utility group (18.38 for "Book
Synopsis (description/summary of each book the author has written)) was almost twice that of
the score of the highest-scoring cue category in the middle-utility group (9.70 for "Samples of
the book (audio that you can listen to)). Together, the four cue categories in the middle-utility
group accounted for 41.07 of the possible 100 points. The lowest score for a cue category in
the middle-utility group (6.69 for "A biography of the author") was more than six times higher
49


than that of the highest-ranking cue category in the lowest category (0.98 for "The name of
the publisher of the author's books"). The lowest-scoring information set accounted for only
2.90 of the possible 100 points.
Four different versions (treatments) of each of the three participating author's
websites were created. The information cues contained within each treatment are listed in
Table 4.6.
Table 4.6: The cues included in each treatment
Treatment version Cues Included Part-worth utility Calculated utility of treatment
Maximum utility Samples of the book (text) 19.12 100
Book reviews 18.53
Book synopsis 18.38
Samples of the book (audio) 9.7
Awards for the book or the author 7.98
A picture of the book cover 7.51
A biography of the author 6.69
The name of the publisher of the author's books 0.98
A picture of the author 0.87
Upcoming readings by the author 0.75
The author's contact information 0.3
High-med utility Samples of the book (text) 19.12 97.1
Book reviews 18.53
Book synopsis 18.38
Samples of the book (audio) 9.7
Awards for the book or the author 7.98
A picture of the book cover 7.51
A biography of the author 6.69
High-low utility Samples of the book (text) 19.12 58.93
Book reviews 18.53
Book synopsis 18.38
The name of the publisher of the author's books 0.98
A picture of the author 0.87
Upcoming readings by the author 0.75
The author's contact information 0.3
High only utility Samples of the book (text) 19.12 56.03
Book reviews 18.53
Book synopsis 18.38
50


4.6.5 Experimental website creation
Based on input (requests for various fonts and design) from the participating regional
authors, three websites (one for each author) were created. Finally, four treatments of each
author's website were created. The four treatments of each author's website presented
information sets intended to represent different levels of utility (max utility, high-medium
utility, high-low utility, or high only utility). Screenshots of the landing pages for each of the
four versions of each of the three participating authors websites are shown Tables C-1, C-2,
and C-3 in appendix C.
4.6.6 Procedure
Survey participants who chose to participate in the study were asked to visit a
website and to browse the site as if they were considering purchasing a book from the
website. The version of the website an individual viewed (author Cooperman, Milofsky, or
Romtvedt and utility of high only, high-medium, high-low, or max) was assigned at random by
a computer program. Once visitors finished browsing their assigned site, they were asked to
complete an on-line survey intended to record their impressions of the website. Participants
chose the computer and the setting where they viewed the sites and completed the survey.
4.6.7 Variables being studied
4.6.7.1 Independent variable
The independent variables under investigation in phases three were website
information content and website information content utility. Information content was
operationalized as eleven website information cues. The presence of the cues were
measured on a dichotomous scale ("present" or "absent" within a website).
51


4.6.7.2 Dependent variables
The dependent variables investigated in phase three were "perceived website
quality", "trust", "perceived risk", and "engagement". A survey was administered to measure
user reactions to the website they were asked to visit. The survey questions used to
measure "perceived website quality", "trust", "perceived risk", and "engagement" were
derived from previously published studies. Several questions were modified to fit the context
of this study. The twenty-three items intended to measure outcomes were constructed as
"strongly agree strongly disagree" statements on a seven-point Likert scale. The surveys
questions are listed in Table D-1 of Appendix D.
4.6.8 Sample Size
For a MANOVA analysis, Hair (2006) recommends a sample size of at least 20
samples for each experimental treatment being investigated. Hence, the minimum number of
respondents needed for this study was 240. Ultimately, we received 349 usable responses,
and each treatment received a minimum of 20 responses.
52


5.0 ANALYSIS AND RESULTS
5.1 MANOVA analysis
MANOVA was conducted to examine the possible effects of the independent
variables "information utility" and "author" on the dependent variables measured by the
survey questions. Results of the MANOVA indicated that, in general, the effects of utility and
author on the outcomes of interest were significant (p < 0.05) for only very few sets of survey
questions.
5.2 ANOVA analysis
In addition to MANOVA, separate ANOVAs were conducted to investigate the effects
of information utility and author on each dependent variable. The results of the ANOVA
analysis are presented in Table E-1 of Appendix E.
Results of the ANOVA indicated that visitors' answers to the questions "The
information in this website is not presented clearly", "The website's design in innovative", "At
this website, one can find details about products and/or services", "The website's design is
visually pleasing", "I trust this website", "Visiting this website was fun", "The website looks
organized", and "There is a lot of information at this website" were significantly influenced by
which authors' website survey participants visited. A significant interaction effect between the
author site and the information utility of the site was also found to be significant for questions
The results indicate that the variable "author" had a strong main effect on survey
participants' reactions to the website they visited, as well as a strong moderating affect on
how information utility affected survey participants' reactions to the site. To further
investigate how the author's websites influenced survey outcomes, separate ANOVAs were
conducted to compare the four experimental websites for each of the three authors. The
results of the three ANOVAs are presented in Appendix F (Tables F-1, F-2, and F-3).
The results of the ANOVA for author Milofsky indicated that information utility
significantly influenced participant answers for questions "The information in this website is
not presented clearly", "The information in this website is sufficiently detailed", "At this
53


website, one can find details about products and/or services", "The amount of information at
this website is appropriate for the website's purpose", "This website's design is visually
pleasing", "The layout of this site is annoying", "The website looks organized", "There is a lot
of information at this website". The mean scores for these questions are shown in Table 5.1.
Table 5.1: Pairwise comparisons of the means for questions demonstrated to be significantly
influenced by information utility.
Question (i) version (mean) a) version Mean Difference (i-j) Std. Error Sig. Lower Bound Upper Bound
3 High Only High-low 1.08 0.442 0.016 0.203 1.955
3.38 High-med 0.93 0.420 0.030 0.089 1.752
Max 1.02 0.420 0.018 0.180 1.843
High-low High Only -1.08 0.442 0.016 -1.955 -0.203
2.30 High-med -0.15 0.439 0.719 -1.028 0.712
Max -0.06 0.439 0.878 -0.937 0.803
High-med High Only -0.93 0.420 0.030 -1.752 -0.089
2.45 High-low 0.15 0.439 0.719 -0.712 1.028
Max 0.09 0.417 0.828 -0.734 0.916
Max High Only -0.75 0.420 0.018 -1.843 -0.180
2.63 High-low 0.33 0.439 0.878 -0.803 0.937
High-med 0.18 0.417 0.828 -0.916 0.734
4 High Only High-low -1.34 0.389 0.001 -2.108 -0.568
4.62 High-med -0.38 0.369 0.312 -1.106 0.356
Max -0.71 0.369 0.057 -1.439 0.022
High-low High Only 1.34 0.389 0.001 0.568 2.108
5.96 High-med 0.96 0.386 0.014 0.199 1.727
Max 0.63 0.386 0.106 -0.135 1.394
High-med High Only 0.38 0.369 0.312 -0.356 1.106
5.00 High-low -0.96 0.386 0.014 -1.727 -0.199
Max -0.33 0.366 0.365 -1.058 0.392
Max High 0.71 0.369 0.057 -0.022 1.439
5.33 High-low -0.63 0.386 0.106 -1.394 0.135
High-med 0.33 0.366 0.365 -0.392 1.058
7 High Only High-low -1.41 0.479 0.004 -2.360 -0.462
4.22 High-med -1.14 0.455 0.013 -2.046 -0.244
Max -1.03 0.458 0.026 -1.939 -0.124
High-low High 1.41 0.479 0.004 0.462 2.360
54


5.63 ONIy
High-med 0.27 0.476 0.577 -0.676 1.208
Max 0.38 0.479 0.430 -0.569 1.328
High-med High Only 1.14 0.455 0.013 0.244 2.046
5.36 High-low -0.27 0.476 0.577 -1.208 0.676
Max 0.11 0.455 0.803 -0.787 1.014
5.25 High Only 1.13 0.458 0.026 0.124 1.939
High-low -0.38 0.479 0.430 -1.328 0.569
High-med -0.11 0.455 0.803 -1.014 0.787
12 High Only High-low -1.77 0.385 0.000 -2.528 -1.005
4.23 High-med -1.04 0.366 0.005 -1.764 -0.315
Max -1.22 0.366 0.001 -1.945 -0.497
High-low High Only 1.77 0.385 0.000 1.005 2.528
6.00 High-med 0.73 0.376 0.056 -0.018 1.472
Max 0.55 0.376 0.150 -0.199 1.290
High-med High Only 1.04 0.366 0.005 0.315 1.764
5.27 High-low -0.73 0.376 0.056 -1.472 0.018
Max -0.18 0.357 0.611 -0.888 0.525
Max High 1.22 0.366 0.001 0.497 1.945
5.45 High-low -0.55 0.376 0.150 -1.290 0.199
High-med 0.18 0.357 0.611 -0.525 0.888
13 High Only High-low -1.35 0.458 0.004 -2.248 -0.435
3.84 High-med -1.10 0.435 0.013 -1.957 -0.235
Max -1.28 0.435 0.004 -2.138 -0.417
High-low High Only 1.35 0.458 0.004 0.435 2.248
5.19 High-med 0.25 0.455 0.590 -0.655 1.146
Max 0.07 0.455 0.888 -0.836 0.964
High-med High Only 1.10 0.435 0.013 0.235 1.957
4.94 High-low -0.25 0.455 0.590 -1.146 0.655
Max -0.18 0.431 0.674 -1.036 0.672
Max High Only 1.28 0.435 0.004 0.417 2.138
5.12 High-low -0.07 0.455 0.888 -0.964 0.836
High-med 0.18 0.431 0.674 -0.672 1.036
17 High Only High-low 1.16 0.502 0.022 0.175 2.161
4.09 High-med 1.48 0.476 0.002 0.545 2.431
Max 1.21 0.476 0.012 0.272 2.158
High-low High Only -1.16 0.502 0.022 -2.161 -0.175
2.93 High-med 0.32 0.498 0.522 -0.666 1.306
Max 0.05 0.498 0.925 -0.939 1.033
High-med High Only -1.48 0.476 0.002 -2.431 -0.545
2.61 High-low -0.32 0.498 0.522 -1.306 0.666
55


Max -0.27 0.473 0.565 -1.208 0.663
Max High Only -1.21 0.476 0.012 -2.158 -0.272
2.88 High-low -0.05 0.498 0.925 -1.033 0.939
High-med 0.27 0.473 0.565 -0.663 1.208
20 High Only High-low -1.57 0.383 0.000 -2.326 -0.811
4.47 High-med -0.71 0.363 0.052 -1.432 0.006
Max -1.32 0.363 0.000 -2.038 -0.600
High-low High 1.57 0.383 0.000 0.811 2.326
6.04 High-med 0.86 0.380 0.026 0.103 1.607
Max 0.25 0.380 0.513 -0.503 1.001
High-med High Only 0.71 0.363 0.052 -0.006 1.432
5.18 High-low -0.86 0.380 0.026 -1.607 -0.103
Max -0.61 0.360 0.095 -1.320 0.107
Max High Only 1.32 0.363 0.000 0.600 2.038
5.79 High-low -0.25 0.380 0.513 -1.001 0.503
High-med 0.61 0.360 0.095 -0.107 1.320
21 High Only High-low -1.72 0.408 0.000 -2.519 -0.904
3.84 High-med -1.10 0.387 0.005 -1.862 -0.329
Max -1.40 0.387 0.000 -2.165 -0.632
High-low High Only 1.72 0.408 0.000 0.904 2.519
5.56 High-med 0.62 0.405 0.131 -0.186 1.418
Max 0.32 0.405 0.441 -0.489 1.115
High-med High Only 1.10 0.387 0.005 0.329 1.862
4.94 High-low -0.62 0.405 0.131 -1.418 0.186
Max -0.30 0.384 0.432 -1.064 0.458
Max High Only 1.40 0.387 0.000 0.632 2.165
5.24 High-low -0.32 0.405 0.441 -1.115 0.489
High-med 0.30 0.384 0.432 -0.458 1.064
56


6.0 DISCUSSION
The results of our exploratory study and phase one indicate that distinct information
cues commonly included within retail websites can be reliably and meaningfully identified and
inventoried. Further, within certain types of websites, specific sets of information cues are
often consistently included within websites. These results demonstrate that research
questions 1 and 2 ("What information content is presented within a given e-tail website?" and
"Are there subsets of information cues which are common to a specific domain?") can be
systematically and meaningfully answered. The results of phase two demonstrated how the
relative utilities of various information cue categories can be meaningfully measured and
interpreted. Because the utilities of the information cue categories were measured on an
interval scale, the relative utilities of the categories could be meaningfully empirically
compared. The results of our information cue category utility calculations indicated that in the
domain of regional author websites, three cue categories (text samples of the book, book
reviews, and book synopsis) were found to be extremely important, four cue categories
(audio samples of books, awards for the book or the author, a picture of the book cover, and
a biography of the author) were found to be somewhat important, and four other cue
categories (the name of the publisher of the book, a picture of the author, upcoming readings
by the author, and the author's contact information) were found to have very little utility.
These results demonstrate that research question 3 ("What are the relative utilities of the
information cues presented at a given e-tail website?") can also be systematically and
meaningfully answered.
In phase three of the main study, we conducted a 4x2 full factorial experiment to test
the hypotheses that websites with relatively high information utilities would elicit significantly
different responses from website visitors than would websites with relatively low information
utilities. Overall, our analysis of the results indicate that study participants who visited
websites with relatively high information utility did not have significantly better reactions to the
site than did participants who visited websites with relatively low information utility. Hence,
the results of this study did not find significant relationships between website information
57


content utility and UWI outcomes at a given e-tail website. Although these results were
negative, they are a significant first step towards answering research question 4, "What are
the relationships between website information content utility and user-website interaction
outcomes at a given e-tail website?"
6.1 Hypothesis 1
Hypothesis 1a posited that a significant relationship would exist between a website's
calculated information utility and visitors' perceptions of that website's information quality. In
general, the results of the survey conducted for phase three did not find a significant
relationship between calculated website information utility and visitors' perceptions of
information quality. Consequently, hypothesis 1a was not supported. Despite this, we do
consider some outcomes of the analysis to be noteworthy.
The results of the survey indicate that there were effectively no significant differences
between survey participants' responses to the experimental websites of the authors
Cooperman and Romtvedt. Specifically, for author Cooperman, ANOVA analysis of survey
responses indicated that there were no significant (p <= 0.05) differences in the information
quality of any of the treatments. For the author Romtvedt, participant answers to one survey
question about information content ("There was a lot of information at this site") exhibited
significant differences between treatments. In this case, the mean scores for the hi and
hijow treatments were very close to one another (means scores of 4.464 and 4.370,
respectively) and were significantly different (p <= 0.05) from the mean scores of the hi_med
and max treatments (mean scores of 5.708 and 5.609, respectively). The implications of this
result are discussed later in this section.
Unlike the results for authors Cooperman and Romtvedt, a weak trend did occur
among the mean scores of certain survey questions about information quality for author
Milofsky. Specifically, mean scores for the survey questions "The information in this website
was not presented clearly", "At this website, one can find details about products and/or
services", "The amount of information at this website is appropriate for the website's
purpose", and "There is a lot of information at this website") were significantly (p <= 0.05)
58


lower for the hi information utility version of the website than were the mean scores for the
other versions (hijow, hi_med, and max) of the site (Table 6.1).
Ultimately, although a few treatments do have statistically different mean question
scores for a few survey items, it cannot be implied that they demonstrate a significant
relationship between information utility and perceptions of design quality.
Table 6.1: Pairwise comparisons of the means for questions demonstrated to be significantly
influenced by information utility.
question (i) version (mean) (j) version Mean Difference (H) Std. Error Sig. Lower Bound Upper Bound
3 High Only High-low 1.08 0.442 0.016 0.203 1.955
3.38 High-med 0.93 0.420 0.030 0.089 1.752
Max 1.02 0.420 0.018 0.180 1.843
High-low High Only -1.08 0.442 0.016 -1.955 -0.203
2.30 High-med -0.15 0.439 0.719 -1.028 0.712
Max -0.06 0.439 0.878 -0.937 0.803
High-med High Only -0.93 0.420 0.030 -1.752 -0.089
2.45 High-low 0.15 0.439 0.719 -0.712 1.028
Max 0.09 0.417 0.828 -0.734 0.916
Max High Only -0.75 0.420 0.018 -1.843 -0.180
2.63 High-low 0.33 0.439 0.878 -0.803 0.937
High-med 0.18 0.417 0.828 -0.916 0.734
7 High Only High-low -1.41 0.479 0.004 -2.360 -0.462
4.22 High-med -1.14 0.455 0.013 -2.046 -0.244
Max -1.03 0.458 0.026 -1.939 -0.124
High-low High Only 1.41 0.479 0.004 0.462 2.360
5.63 High-med 0.27 0.476 0.577 -0.676 1.208
Max 0.38 0.479 0.430 -0.569 1.328
High-med High Only 1.14 0.455 0.013 0.244 2.046
5.36 High-low -0.27 0.476 0.577 -1.208 0.676
Max 0.11 0.455 0.803 -0.787 1.014
Max High 1.13 0.458 0.026 0.124 1.939
5.25 High-low -0.38 0.479 0.430 -1.328 0.569
High-med -0.11 0.455 0.803 -1.014 0.787
12 High Only High-low -1.77 0.385 0.000 -2.528 -1.005
4.23 High-med -1.04 0.366 0.005 -1.764 -0.315
59


Max -1.22 0.366 0.001 -1.945 -0.497
High-low High Only 1.77 0.385 0.000 1.005 2.528
6.00 High-med 0.73 0.376 0.056 -0.018 1.472
Max 0.55 0.376 0.150 -0.199 1.290
High-med High Only 1.04 0.366 0.005 0.315 1.764
5.27 High-low -0.73 0.376 0.056 -1.472 0.018
Max -0.18 0.357 0.611 -0.888 0.525
Max High Only 1.22 0.366 0.001 0.497 1.945
5.45 High-low -0.55 0.376 0.150 -1.290 0.199
High-med 0.18 0.357 0.611 -0.525 0.888
21 High Only High-low -1.72 0.408 0.000 -2.519 -0.904
3.84 High-med -1.10 0.387 0.005 -1.862 -0.329
Max -1.40 0.387 0.000 -2.165 -0.632
High-low High Only 1.72 0.408 0.000 0.904 2.519
5.56 High-med 0.62 0.405 0.131 -0.186 1.418
Max 0.32 0.405 0.441 -0.489 1.115
High-med High Only 1.10 0.387 0.005 0.329 1.862
4.94 High-low -0.62 0.405 0.131 -1.418 0.186
Max -0.30 0.384 0.432 -1.064 0.458
Max High Only 1.40 0.387 0.000 0.632 2.165
5.24 High-low -0.32 0.405 0.441 -1.115 0.489
High-med 0.30 0.384 0.432 -0.458 1.064
Hypothesis 1b posited that a significant relationship would exist between a websites
calculated information utility and visitors' perceptions of that website's design quality. In
general, the results of the survey conducted for phase three did not find a significant
relationship between calculated website information utility and visitors' perceptions of design
quality. Consequently, hypothesis 1b was not supported. Specifically, at two authors'
websites (Cooperman and Romtvedt), there were no significant differences in the design
quality of any of the treatments. However, at the Milofsky website, for two questions about
design quality ("The website's design is visually pleasing", "The layout of the website is
annoying") the mean scores for the high-low, high-med, and max treatments were not
significantly different from one another, but were significantly different (p <= 0.05) from the
mean score for the high only treatment. As with hypothesis 1a, although the mean scores of
60


these few questions do show significant results between treatment variations, they do not
demonstrate a significant relationship between information utility and perceptions of design
quality.
Overall, the mean scores and lower bounds of user responses to questions about
website information quality and design quality were above the mid-point score of 4 but below
5.5 (on a 7-point scale). Consequently, based upon these scores, we infer that neither the
information content nor the design of the websites were negatively received by survey
participants. On the other hand, the scores are closer to a "neutral" judgment (a score of 4)
than they are to a "high" judgment (a score of 7), indicating that the sites did not elicit
satisfaction or delight from visitors. Ultimately, these results indicate that visitors were
generally ambivalent to the content and design of the experimental websites they visited.
6.2 Hypothesis 2
Survey participant responses to the survey question "Visiting this website was fun"
were at or below the neutral score of four, indicating that survey participants did not find the
website fun to visit. As we have previously noted, although the scores for website design and
information quality were not exceptionally high, they were above the neutral threshold score
of four, and they were not so low as to imply that users were upset by any treatment's design
or content. In the absence of a clear dislike for the site, we suspect that the reasons for the
low engagement scores are due to the fact that visitors to the author's website did not find
information content or design factors that they considered to be entertaining or interesting,
resulting in limited levels of arousal, attention, and elaboration.
6.3 Hypotheses 3 and 4
The results of the statistical analysis performed for this study found no relationships
between information utility and measured levels of user trust or perceived risks.
Consequently, hypotheses 3 and 4 were not supported. Interestingly, the mean scores for
the survey question "My interaction with this website was risky" were between 2 and 3 for
each of the twelve treatments, indicating that users generally perceived very low risk.
61


Alternatively, the mean scores for the question "I trust this website" were feel between 4 and
5 for all but one treatment (the max utility treatment of the Romtvedt website, which had a
mean score of 5.348), and none of the treatments exhibited significantly different mean
scores.
These results indicate that survey participants had relatively low perceptions of risk
while visiting the site, but did not form strong opinions of trust or distrust during their visits. A
summary of the results of hypothesis testing is presented in Table 6.2.
62


Table 6.2: A summary of the hypotheses investigated by this study.
Hypothesis Question Supported? C R M
H2: Utility has positive affect on perceived information quality. 1 Hypothesis 2a was not supported for authors Cooperman and Romtvedt. Results for author Milofsky indicated that adding information beyond the minimum amount did improve scores for some questions. N N N
2 N N N
3 N N IC*
4 N N N
7 N N IC*
8 N N N
9 N N N
10 N N N
11 N N N
12 N N IC*
14 N N N
16 N N N
18 N N N
21 N IC* IC*
H2b: Utility has positive affect on perceived design quality. 5 Hypothesis 2b was not supported for authors Cooperman and Romtvedt. Results for author Milofsky indicated that adding information beyond the minimum amount did improve scores for some questions. N N N
13 N N IC*
17 N N IC*
20 N N IC*
H3: Utility has positive affect on engagement. 19 No N N N
22 No N N N
H4: Utility has positive affect on trust. 15 No N N N
H5: Utility has negative affect on risk. 23 No N N N
IC* = Inconclusive
6.4 Effects of information quantity
To examine whether the amount of information presented by the experimental
websites influenced survey responses, the mean scores for high only and high-low
treatments were compared to scores for high-medium and max treatments of each author's
websie. Theoretically, the high-low and high-medium treatments presented the same
63


quantity of information (the same number of cues), but the cues presented by these
treatments offer very different levels of information utility (calculated utility of the high-low
treatment = 58.93, calculated utility of the high-med treatment = 97.1). If the differences in
the mean scores were due to the site's information utility, the mean scores for the high only
and high-low treatments (mean scores 3.833 and 5.538, respectively) should not be
significantly different from one another because these treatments have very similar calculated
information utility scores (56.03 and 58.93, respectively). At the same time, the mean scores
for the high only and high-low treatments should be significantly different than the mean
score for the high-medium treatment. Put simply, we posit that if differences between the
mean survey scores of various websites are caused by differing amounts of information
utility, this should be evidenced by significant survey mean score differences between the
websites with the largest differences between levels of calculated website utility (ie, the high-
low and high-med versions of a website). Conversely, if differences between the mean
survey scores of various websites are caused by differing quantities of website information
content (different numbers of cues), this should be evidenced by significant survey mean
score differences between the websites with the largest differences between levels of
information quantity (ie, the high only and hi-low versions of a website).
For the Cooperman and Milofsky websites, our analysis found no instances where
the mean score of a study question was significantly lower for the high-low version of the site
than for the high-med or max treatments. As noted in the discussion of hypothesis 1a, for
author Romtvedt, participant answers to one survey question about information content
("There was a lot of information at this site") exhibited differences between treatments (Table
6.3).
64


Table 6.3: Pair-wise comparisons of answers to question 21 for author Romtvedt.
(i) Mean
version Difference Std. Lower Upper
question (mean) (j) version (i-j) Error Sig. Bound Bound
21 High Only High-low -0.03 .443 .954 -.903 .852
4.34 High-med -1.30 .452 .005 -2.191 -.399
Max -1.20 .457 .010 -2.103 -.291
High-low High Only 0.03 .443 .954 -.852 .903
4.37 High-med -1.27 .459 .007 -2.181 -.359
Max -1.17 .464 .013 -2.092 -.250
High-med High Only 1.30 .452 .005 .399 2.191
5.64 High-low 1.27 .459 .007 .359 2.181
Max 0.10 .473 .836 -.840 1.036
Max High Only 1.20 .457 .010 .291 2.103
5.54 High-low 1.17 .464 .013 .250 2.092
High-med -0.10 .473 .836 -1.036 .840
In this case, the mean scores for the treatments that had hi and high-low levels of
calculated information utility (56.03 and 58.93, respectively) exhibited very similar survey
scores (4.464 and 4.370, respectively), and the scores for the treatments with high-med and
max levels of calculated information utility (97.1 and 100, respectively) also exhibited very
similar survey scores (5.708 and 5.609, respectively). However, the scores for the group
consisting of the high only and high low utility treatments were significantly different from the
scores of the group consisting of the high-med and max information utility treatments,
indicating that survey respondents did perceive significant differences between websites that
presented different levels of information utility. The perceived differences in information
quantities apparently did not appear to influence other outcomes, since no other survey
questions exhibited a similar trend of significantly higher scores for high-utility versus low-
utility websites.
65


6.5 Overall summary of hypotheses
In summary, our analysis of the survey results implies that increasing information
quantity did not have a significant overall positive or negative impact on user perceptions of
the high-low, high-medium, or max versions of the Milofsky website. For two of the three
authors investigated, information utility was not found to significantly influence website users'
perceptions of information quality, design quality, engagement, trust, or perceived risk. This
is in agreement with Dholakia and Rego's survey which also did not find a significant
relationship between website information quantity and website popularity. However, for one
author (Milofsky), analysis indicated that for some questions about information quality and
design quality, there was a significant difference between treatments with the lowest
information utility and treatments with more than the minimum information utility, and that
adding any information beyond the minimum amount (regardless of the quantity or utility of
the information) increased the acceptability of the information content. This supports Song
and Zahedi's (2005) findings that the information cues presented within their experimental
website treatments were significantly related to website visitors' purchase intentions and their
attitudes about the website.
The lack of agreement in the findings for the different authors for this study and the
lack of agreement in studies performed by prior researchers (e.g. Dholakia & Rego, 1998,
Song & Zahedi, 2005) suggests that other factors besides information content or information
utility may be influencing users perceptions of the information content available at websites.
Thus, the results of this study do not necessarily imply that information utility cannot be useful
when creating websites and predicting user reactions to those sites, or that it should not be
considered in future research. Instead, the results of this study suggest the need for
additional research about how users interact with websites in both real and experimental
situations.
66


6.6
Possible effects of priming and confirmation/disconfirmation
We posit that the results of this study may be partially explained by theories of
priming and confirmation/disconfirmation. According to Mandel and Johnson (2002),
exposure to a prior event (the "prime") increases the availability of information that already
exists within an individual's memory. In categorical priming, a person's judgments about an
object or product is influenced by the constructs that are activated in an earlier task (Herr
1989). In the context of consumer studies, the terms confirmation and disconfirmation are
used to describe the agreement (confirmation) or disagreement (disconfirmation) between a
customers expectations of an interaction with a seller and that customers perceptions of the
seller's actual performance. McKinney et al. (2002) define the "should expectation" as the
perception of a normal standard of performance, the "ideal expectation" as the perception of
an optimal performance, and the "will expectation" as predictions of future performance. As
noted in section 2.2.1, individuals with relatively higher product experience are better able to
differentiate between product offerings than individuals with relatively low product experience.
Answers provided to question thirty ("In the past two years, approximately how many
purchased books, CD's, or DVD's online?") indicate that 24.3% of survey participants had
purchased books, CD's, or DVD's online on ten or more occasions, and that an additional
33.5% participants had purchased books, CD's, or DVD's on between four and ten occasions
during the past two years. According to the website internetretailer.com
(http://www.internetretailer.com/top500/list.asp), Amazon.com was the largest internet retailer
in America in 2009 with on-line sales of $19.2 Billion, three times that of the next largest
internet retailer, Staples.com (2009 internet sales for Staples.com were $7.7 Billion). We
posit that through visits to Amazon.com or BarnesAndNoble.com, a large number of our
survey respondents may have had their "should" and "ideal" expectations of the design and
information quality of a bookseller "primed" to be the information and design quality of major
on-line retailers, and that the designs and information content presented by our experimental
websites may have lead to disconfirmation between what users had come to expect from a
large on-line bookseller and the regional author websites they encountered as part of the
67


study. Future studies may be able to limit the effects of priming by dominant but not
comparable websites within a field by exposing experimental subjects to websites which
better represent the types of websites that are the focus of the study. As we discussed in
section 2, even extremely short exposures to websites can have the effect of forming
significant long-lasting impressions on visitors (citation), and future researcher may want to
use this fact as a basis for investigating the effects of website expectation "priming". For
instance, a researcher performing an experiment to investigate user reactions to a target
website may benefit from first exposing study participants to a series of websites that are
inferred or empirically determined to be representative of the target website's domain just
prior to exposing the participants to the target website. Such exposures could, in effect,
increase study participants' experience with the domain, and, consequently, influence
participants "should" and "ideal" expectations for websites within that domain.
Finally, we posit that the nature of a regional author's website may also cause
confusion among its visitors. For instance, for most visitors, the purpose of Amazon.com is
clearly commercial- the website exists to sell visitors books. However, although the main
purpose of a regional author's website is also to sell books, the design preferences of the
authors are often taste-specific and somewhat non-conventional. For example, the websites
of several regional authors present non-white text on a black or dark-colored background,
and one author who participated in the study specifically requested the use of red text on a
black background- a color scheme rarely seen in commercial websites. In addition, the
pictures and graphics used in the websites of regional authors often lack the polish of
professionally produced websites. Consequently, the informality of the sites may cause
some visitors to confuse the author's commercial website with a personal website, such as
would have considerably more personal information about the author. If such confusion were
to occur, visitors could leave the website disappointed at not having made a more personal
connection with the author, even though the purpose of the site was to provide inform about
and sell the author's books.
68


6.7 Post Hoc Analysis
To look for possible explanations for this study's results, we examined web server
logs to see if the UWI per se could explain why information utility failed to have an effect on
the UWI outcomes measured by this study. The server logs allowed us to "track" website
visits by observing the URLs and timestamps of website requests made by various IP
addresses. By comparing the time that a certain IP address pulled the landing page of a UR
to the timestamp that the last request for a page was made by that IP address, we were able
to gain a rough assessment of how long a user spent at the site. We were also able to count
how many pages were viewed during the visit (depth of visit). Several previous investigations
have used total time at the website and total pages visited (also called the "depth" of a visit)
as operationalizations of the UWI. For example, Danielson (2002) found that users who
visited websites that constantly displayed site maps performed fewer navigation acts but
visited more pages than users who visited websites that did not constantly display site maps,
implying that website visitors with the constantly displayed site maps visited more pages in a
more efficient manner than did visitors to website without constantly displayed site maps.
We analyzed 50 visits to the four experimental Milofsky websites and 30 visits to the
actual Milofsky website (the site posted for the author). A typical participant in the experiment
visited the site for less than two minutes and visited between 2 and 5 pages. It was not
uncommon for participants in the experiment to spend less than a minute at the website. Of
the 30 visitors to the actual Milofsky website, all but three viewed the landing page and then
exited the site (as was evidenced by the fact that their IP did not request any other pages
from the server). These results indicate that neither participants in the experiment nor visitors
to the author's actual website were interested in the site. As we have previously mentioned,
the results of the survey did not find the website design to be significantly unacceptable.
Hence, we assume that visitors performed a cursory overview of the website and almost
immediately decided that the subject (books and authors) or the authors genres (poetry and
fiction) were not of interest to them. Hence, because they were not interested in the website
69


or its topic, experiment participants viewed enough of the site to be able to answer the survey
questions, while actual visitors simply left the website.
70


7.0 CONCLUSION
7.1 Limitations and Future Research
Despite paying careful attention to this research project's design, execution, and
analysis, the study has significant limitations. Perhaps the most significant limitation of this
study is the limited external validity of its overall findings, especially since the results do not
agree with the results of previous studies of information utility that have demonstrated
relationships between information utility and consumer responses to shopping experiences
(Keller and Staelin 1987, Chen et al. 2009). By investigating the effects of information
content at three different websites within the same domain, we were able to very effectively
examine the effects of not only website information content, but also the different website
designs presented at each author's website. Overall, the fact that no website's design was
found to be overwhelmingly different (better or worse) than the other websites examined in
the study reduce the chance that the study's results were caused by the design qualities of a
single website. However, there is no evidence to suggest that the results of this study do or
should apply to other domains, especially in light of the work of Zhang and von Dran (2001),
who, as previously discussed in section 2.4, demonstrated that different website cues and
dimensions are often given very different levels of importance across different website
domains. One simple way to test the external validity of these finding would be to ask a
second group of individuals to visit and evaluate the actual websites of several regional
authors who did not participate in the study, and a third group to visit and evaluate the
websites of major on-line booksellers such as Amazon.com and BarnesAndNoble.ccom.
Simple ANOVA tests could compare the scores for these "real world" sites to those of the
experimental websites examined in the study. The differences (or lack thereof) found
between respondents' reactions to actual author websites, the websites of major booksellers,
and those examined in this study could provide significant indications of the internal and
external validity of this study's results (or lack thereof).
Another significant limitation of this study are the possible confounders that we did
not account or correct for. As with any website, we anticipate that survey participants'
71


reactions to the websites were largely dependent upon their interest in the information and
products offered at the authors' websites. In one study (Kang and Kim 2006) experts were
asked to rank the content of websites based on the number of links, graphics, words, and
banner advertisements at each site. The study indicated that when users were relatively un-
interested in the topic of the websites they evaluated, they rated sites with high information
content as more informative than sites with low information content. In contrast, subjects who
exhibited relatively high interest did not equate information content with informativeness.
Kang and Kim speculated that disinterested users used a simple heuristic (such as: a large
amount of data implies high informative value) to judge the informativeness of a site. On the
other hand, users who displayed a relatively high level of interest did not rate all information
cues equally, and were more discriminating about which information they considered
important. As previously mentioned, anecdotal evidence from our post-hoc analysis suggests
that many users had low levels of interest in the websites products (an extremely limited
selection of books pertaining to very specific genres). Consequently, we suspect that user
involvement in the task (or, more accurately, their lack of involvement), significantly
influenced the results of the study. Future studies may want to employ tactics that may
increase user involvement, such as offering a discount or other financial incentive towards
any products purchased during the exercise (e.g. Lynch and Ariely 2001) or asking survey
participants to compare treatments, so that they must pay sufficient attention to the sites to
allow comparison and preference formation. Another possible way to increase subject
involvement may be to allow experiment participants to choose from several possible
websites in the hope that they would find a site that would interest them enough to induce
significant involvement in the task. Ultimately, however, user involvement may be difficult to
induce, since even serious browsers may simply not be interested in a website or its
offerings.
Another limitation of this study is that it only considered the effects of information
cues that were expected to have positive information utilities. The results of our exploratory
study demonstrate that not all information cues may induce positive responses from website
72


visitors (e.g., including price listings on health care websites.) Future studies may wish to
examine the effects of cues with possible negative utility (advertisements, animation, etc.),
which may have significant negative effects on UWI outcomes. For example, previous work
by Hong (2004) found that including flash animation within a website did attract users'
attention, but in some cases, the inclusion of flash animation was found to decrease overall
recall.
A final limitation that we make note of are the gaps in the instrument used by this
study. In retrospect, the results of the survey may have provided more diagnostic value if
questions such as "The subject matter of this website interests me", "If I knew a fried who
was interested in the types of books sold at this website, I would recommend to them that
they visit this website", "I was disappointed by this website", or "I disliked this website". The
direct question "The subject matter of this website interests me" would be a direct measure of
the respondent's interest in the site's subject matter, and an indirect measure of involvement.
The question If I knew a friend who was interested in the types of books sold at this website,
I would recommend to them that they visit this website" asks users to judge the site through
the eyes of someone who was interested in the site's subject matter. The last two questions
may be able to discern whether low scores are simply due to disconfirmation between
expectations and actual experience, or to genuine dislike for the site.
7.2 Contributions
This research study makes significant advances towards answering three major
research questions: "What information content is presented within a given e-tail website?",
"What is the relative utility of the information cues presented at a given e-tail website?", and
'What are the relationships between information utility and user-website interaction outcomes
at a given e-tail website?"
Towards answering the first question, we developed and demonstrated the WICS,
which we have demonstrated to be capable of meaningfully indexing the information content
of many types of e-commerce websites. In both the exploratory and main studies of this
73


article, we have demonstrated how the WICS can be used by researchers and practitioners
to assess and compare website information content.
Towards answering the second question, "What are the relative utilities of the
information cues presented at a given e-tail website?" this study has identified and defined
the construct "pre-UWI information utility", and demonstrates how the construct can be
empirically measured and used to make predictions about post-UWI outcomes. This work
makes a significant methodological contribution to the field of UWI research by demonstrating
how the relative contributions of website information cues can be measured on an interval
scale. Previous studies of website content have used techniques that resulted in ordinal
rankings of information importance (Zhang et al., 2001). This study is the first UWI study to
use a technique of measuring specific information cue utility on an interval scale. This is a
major advancement for the field of UWI research because interval scales allow for meaningful
empirically-based comparisons of the relative contributions of information cues, whereas the
ordinal rankings of information importance used by previous studies do not. We anticipate
that interval measurements of information utility will be a major step towards identifying which
information cues constitute high-quality information content.
Finally, towards answering the question, "What are the relationships between
information utility and user-website interaction outcomes at a given e-tail website?" this study
investigated the relationships between website information content utility and user
perceptions of website information quality, website design quality, engagement, trust, and
perceived risk. Although the results of this study did not find measurements of pre-UWI
information utility to be predictive of UWI outcomes, we hope that future studies will continue
to investigate the relationships between pre-UWI information utility and UWI outcomes
(including and beyond those studied here).
7.3 Implications for Theory
In the context of the Zhang and Li (2001) model of HCIs, the new construct that this
research identifies and measures, "pre-UWI information utility", is an antecedent factor that is
theorized to influence UWIs and their outcomes. This study is the first UWI study to
74


specifically identify and measure pre-UWI information cue utility. Although this study focused
on a specific type of HCI (UWIs), we suspect that the concept of measuring the utility of
information content independently from the effects of how information is framed and before
an actual HCI may advance the understanding not only of B2C e-commerce oriented UWIs,
but of many other types of HCIs (e.g., HCIs with mobile web devices, GPS technologies, etc.)
In addition to the possible main effects of information utility upon UWIs and their
outcomes, we predict that future studies will find pre-UWI information utility to be a significant
moderator of the relationships between other antecedent factors, UWIs, and their outcomes.
As we have already noted in section 2.2.5, previous work by Robin and Holmes (2008) found
that website visitors' perceptions of a website's information credibility was directly affected by
the sites aesthetic appeal. Based on such results, we suspect that future research will
discover significant interaction effects between information utility and website design.
Finally, future studies may be able to apply the website information content utility
construct to answer other questions beyond those addressed by this study. For example, the
ability to measure website information cue utility may facilitate new investigations of how
information content exacerbates or alleviates website visitors' perceptions of asymmetry.
7.4 Implications for Practice
The results of this study offer a new set of informative and actionable guidance that
can be directly applied to the work of many website managers and e-commerce practitioners.
As previously discussed, the WICS instrument and the techniques for measuring information
cue utility can be applied to answer some of the most significant questions in the field of e-
commerce. The ability to systematically and successfully address the research questions
posed by this study may in fact be the difference between success and failure in the realm of
on-line retail commerce.
In addition to influencing the low-level tactics and technical aspects of website and e-
commerce management, clear and directly actionable studies like this one have the potential
to significantly impact some company's high-level business and marketing strategies.
75


Specifically, free analytical tools such as Google Website Optimizer and relatively low-cost
and easy to implement software tools such as the Sawtooth Software Suite now make
available market research techniques that were, until quite recently, only available to
companies willing to make large investments of time and capital in research and
development. The rapidly decreasing costs of powerful, easy-to-implement and interpret
website analysis tools make applying the findings of research studies such as this one to a
practitioners own e-commerce businesses much more feasible than even a few years ago.
The increasing ability to cheaply and purposefully apply the knowledge gained from e-
commerce studies may enable companies to enter markets or marketing channels that were
previously considered cost-of-entry prohibitive, too competitive, or otherwise closed. In the
near future, as managers continue to realize and embrace the potential benefits and
competitive advantages of technologies and knowledge that enable empirically-based
website management, the influences and demand for high-quality, actionable research
studies will continue to grow.
76


APPENDIX A
Table A-1: Studies that have investigated the website quality construct.

Study Definition of satisfaction Operationalization of satisfaction Type of satisfaction considered
Abdinnour-Helm et al., 2005 Satisfaction with a web site, per se, is not defined. Satisfaction with a web site (Content, Accuracy, Format, Ease of Use, Timeliness) was measured by 12
Ballantine, 2005 Satisfaction, per se, was not defined One question measured on a likert scale anchored on satisfied- unsatisfied. Level of satisfaction with the interface provide by an online retail store
Bansal et al., 2004 Website satisfaction was defined as "The mean level of overall satisfaction with a site..." p. 294 One question gathered from quarterly website customer satisfaction data provided by Nielsen/NetRatings Overall satisfaction
Cao & Gruca, 2004 Satisfaction, per se, was not defined Consumer ratings of e-tailer service provided by BizRate. Pre-purchase satisfaction (Ease of use, product selection, product information, website performance); Post- purchase satisfaction (On-time delivery, product representation, order tracking, customer support)
Cao et al., 2003- 4 The authors discuss many different aspects of buyer-seller interactions that consumers may judge as satisfactory or unsatisfactory, but they do not define satisfaction, per se. For each of the NN aspects of the purchase being assessed, customer respondents were asked "How satisfied are you with each of the following aspects of this on-line purchase?" Respondents responded using a 10-point scale rooted on "Not at all satisfied or "Highly satisfied. Price satisfaction (satisfaction with product price, satisfaction with shipping/ handling costs); Satisfaction with ordering process (ease of ordering, product selection, product information, & website performance); Satisfaction with fulfillment process (on-time delivery, order tracking, product representation, customer support)
77


Choetal., 2002 Dissatisfaction, per se, is not defined. Dissatisfaction was measured by 3 Likert scale items anchored on "Strongly dissatisfied" and "Not dissatisfied at all Dissatisfaction
Flavian et al., 2006 Satisfaction is defined as "an affective consumer condition that results from a global evaluation of all the aspects that make up the consumer relationship." (p. 4) Overall satisfaction was measured by four 7-point Likert scale items
Holland & Baker, 2001 Satisfaction with the retailer, per se, is not defined. Satisfaction with the retailer was measured by four 7-point Likert scale items Satisfaction with the retailer.
Holloway et al., 2000 Post-recovery satisfaction was defined as "the degree to which a customer is satisfied with a service firm's transaction- specific service recovery effort following a service failure" (the authors quote the definition in Boshoff, 1999, p. 237) Post recovery satisfaction was measured by 3 items, each measured by a seven-point multi- item reflective scale Post-recovery satisfaction
Jiang & Rosenbloom, 2005 At-checkout satisfaction is defined as"... customer ratings... of e-retailing services on the shopping convenience dimension." (p. 157); After-delivery satisfaction is defined as "customer ratings... of e-retailing services on the fulfillment reliability dimension." (p. 158); Overall satisfaction is defined as"... the general attitude toward the e- tailing service provider after the transaction is complete." (p. 153) At-checkout satisfaction, after delivery satisfaction, and overall satisfaction were measured by customer responses to questions at the bizrate website.
Jiang, 2002 Anticipated satisfaction was defined as"... the customer's assessment of the likely satisfaction with consuming his/her chosen product/service." (p. 172) Anticipated satisfaction was measured by three 7-point Likert scale items
78


Jin & Park, 2006 Satisfaction was defined as"... the perception of pleasurable fulfillment and occurs when retailer performance matches or is higher than consumers' expectations." (p. 203) Satisfaction was measured by three 7-point Liker scale questions.
Kim & Kim, 2006 Satisfaction, per se, was not defined Identified three dimensions of online shopping satisfaction (Safe purchasing, shopping convenience, and vendor reliability) Online shopping satisfaction, Shopping convenience satisfaction, shopping convenience satisfaction, and vendor reliability satisfaction
Kim & Lim, 2001 Satisfaction was defined as "an evaluation rendered that the consumption experience was at least as good as it was supposed to be." (the authors quote the definition in Hunt, 1977) Study participants assessed their satisfaction with 15 various aspects of a website (information, design, customer service, entertainment, etc.) Website satisfaction
Lee & Lin, 2005 Satisfaction was defined as "Customer satisfaction with an online store." (p. 167) Customer satisfaction was measured by one question measured by a seven-point Likert scale. Customer satisfaction.
McKinney et al., 2002 Overall satisfaction was defined as "an affective state representing an emotional reaction to the entire Web site search experience." (p. 298) Information quality satisfaction and system quality satisfaction "...have an evaluative nature similar to that of overall satisfaction." (p. 299) Information quality satisfaction and System quality satisfaction (both measured by 4 Likert scale questions); Overall satisfaction (measured by 5 Likert scale questions and one dichotomous question). Information quality satisfaction; System quality satisfaction; Web-customer (overall) satisfaction
Muylle et al., 2004 Web site user satisfaction is"... the attitude toward the web site by a hands-on user of the organization's web site. (p. 545, italics original) Web site user satisfaction was operationalized by 4 first-dimensional constructs (Layout, Information, Connection, Language Customization). Information was operationalized as relevancy, accuracy, comprehensibility, and comprehensiveness. Connection was operationalized as ease of use, entry guidance, structure, hyperlink connotation, and speed. Web site user satisfaction
79


Negash et al 2003 User satisfaction
Parasuraman et al 2005 e-satisfaction was defined as "...the outcome of consumer perceptions of online convenience, merchandising, site design, and site convenience. p. Liang & Lai, 2002 Two questions asking whether the individual was satisfied and pleased with their Internet shopping experience. Overall satisfaction with the Internet- shopping experience
Posselt & Gerstnpr, 2005 Satisfaction is measured in terms of confirmation/disconfirm ation. Pre-sale services (ease of finding product, product selection, clarity of product information, price, website look, shipping charges, shipping options, display of purchase amount); Post-sale services (product availability, order tracking, on-time product arrival, product representation, customer service support); Overall satisfaction. Data was gathered from customer ratings of retailers at BizRate.com that used 11-point Likert scales. Pre-sale satisfaction; Post sale satisfaction
Rodgers et al., 2005 On-line satisfaction was defined as an "... emotional reaction to an on-line service experience." (p. 314) Satisfaction was measured by three 5-point Likert scale questions.
Shamdasani et al.. 2008 satisfaction:" Four questions using a seven-point semantic differential scale Overall satisfaction
Shankar etal., 2003 Satisfaction was defined as "...the perception of pleasurable fulfillment of a service..." (p. 154); "Service encounter satisfaction is transaction specific, whereas overall customer satisfaction is relationship-specific, that is, overall satisfaction is the cumulative effect of a set of discrete service encounters or transactions with the service provider over a period of time." (p. 156) Service encounter satisfaction was operationalized by one 5-point Liker scale question; Overall satisfaction was operationalized by one 7-point Likert scale question.
Thatcher & George, 2004 Satisfaction, per se, was not defined Consumer satisfaction was measured by 3 items which were not described. Consumer satisfaction.
80


Tsai.& Huang, 2007 Overall satisfaction was defined as"... a positive affective state resulting from a global evaluation of performance based on past purchasing and consumption experience." (p. 233) Overall satisfaction was measured by four 7-point Likert scale items
Wolfinbarger & Gilly, 2003 "Global quality is designed to be a global measure across purchase experiences., while the customer satisfaction items refer specifically to the most recent purchase." (p. 195) Six questions measuring post- purchase satisfaction Overall satisfaction
Zviran et at., 2006 The authors discuss many different aspects of buyer-seller interactions that consumers may judge as satisfactory or unsatisfactory, and they discuss other authors definitions of satisfaction, but they do themselves not define satisfaction, per se.
81


APPENDIX B
Table B-1: The results of the WICS exploratory study and phase two of the main study.
Exploraotry Study- Prevalence of each cue in the 5 domains surveyed % of sites with cue Main Study, Phase 1
Information Cue Electronics (4) Medical Services (4) Specialty Foods (4) Insurance (5) Cruise Lines (4) overall Count (out of 15) Source
NAVIGATION INFORMATION
Is a navigational bar present on every screen? 4 4 4 5 4 100% 9 Aladwani & Palvia, 2002 Barnes & Vidgen 2001
Is the navigation bar consistently located? 4 4 3 3 4 86% 9 Song & Zahedi, 2005
Does the repeated Navigation structure (menus, links @ bottom of page) contain links to: New Item
a Customer service policy? 4 2 2 5 3 76% 0 New Item
a Privacy policy? 4 0 2 5 4 71% 0 Ranganathan & Ganapathy, 2002 Robbins & Stylianou, 2003 Song & Zahedi, 2005 van Iwaarden, et ai, 2003
a Site map? 3 3 1 5 4 76% 0 New Item
a Search engine? 3 0 3 3 3 57% 0 Aladwani & Palvia, 2002 Ranganathan & Ganapathy, 2002 Robbins & Stylianou, 2003 Song & Zahedi, 2005 Zhang, et al., 2001
the Home page? 4 4 4 5 4 100% 15 Dholakia & Rego, 1998
Does the site have a site map? 3 3 1 5 3 71% 0 Robbins & Stylianou, 2003
Zhang, et al., 2001


(Table B-1 Continued)
PRODUCT/SERVICE INFORMATION
Lists of products/services offered by the company. 4 4 4 5 4 100% 15 Aladwani & Palvia, 2002 van Iwaarden, et al, 2003
List of products/services that can be purchased/used at the website 4 2 4 5 4 90% 0 Zhang, et al., 2001
Prices of Product or Service 4 1 4 5 4 86% 6 Dholakia & Rego, 1998 Lynch & Ariely, 2000; Song & Zahedi, 2005
Availability of Product or Service 3 0 0 3 4 48% 0 Dholakia & Rego, 1998
Product Description
Attributes 4 0 3 4 4 71% 0 Aladwani & Paliva, 2002 Dholakia & Rego, 1998 Lynch & Ariely, 2000 Song & Zahedi, 2005 Zhang, et al., 2001
Functionality 4 2 0 3 4 62% 0 Chan & Chan, 2005
Materials 2 1 0 0 0 14% 0 Dholakia & Rego, 1998
Ingredients 0 0 3 0 0 14% 0 Dholakia & Rego, 1998
Nutritional Information 0 0 3 0 0 14% 0 Dholakia & Rego, 1998
Description of services provided. 2 4 1 5 4 76% 0 New Item
Product variations, e.g., color, size 2 3 2 5 4 76% 0 Dholakia & Rego, 1998
FAQ list of 'Frequently asked questions 4 4 1 3 4 76% 0 Zhang, et al., 2001
Product customization information 0 1 0 5 4 48% 0 Barnes & Vidgen 2001 Song & Zahedi, 2005
Claims of product superiority 4 3 1 4 3 71% 0 Zhang, et al., 2001
Comparisons to competitor's products or prices 2 1 0 4 1 38% 0 Dholakia & Rego, 1998


(Table B-1 Continued)
'Side-by-side' comparisons of products offered by company 2 0 1 0 2 24% 0 Lynch & Ariely, 2000
Ranganathan & Ganapathy, 2002
Product Benefits (or negative avoided) 4 4 4 4 4 95% 0 Dholakia & Rego, 1998
Product warnings (e.g. side effects, hazards) 0 3 2 0 0 24% 0 New Item
Product Picture
Static, 2D 4 4 4 0 4 76% 15 Dholakia & Rego, 1998 Song & Zahedi, 2005
Dynamic, 3D 2 0 0 0 3 24% 2 Dholakia & Rego, 1998
New Product Notification 4 0 2 2 3 52% 0 Dholakia & Rego, 1998 Song & Zahedi, 2005
Owner's Manual, Assembly Instructions, etc. 4 0 0 0 0 19% 0 Dholakia & Rego, 1998
Demonstration of the product in use
Image 2 4 0 0 3 43% 0 New Item
Multimedia 2 0 0 0 2 19% 0 New Item
Product preview (e.g., sample chapters for a book)
Text 0 0 0 0 3 14% 6 New Item
Multimedia 1 0 0 0 4 24% 5 New Item
Product reviews (customer, 3rd party, etc) 3 4 2 2 3 67% 10 Lynch & Ariely, 2000 Song & Zahedi, 2005 Zhang, et al., 2001
Product endorsement (Celebrity/Expert) 1 4 0 2 0 33% 0 Song & Zahedi, 2005
Product or general warranty information 4 3 1 3 1 57% 0 Barnes & Vidgen 2001 Chan and Chan, 2005 Dholakia & Rego, 1998


(Table B-1 Continued)
Staff or service provider profile s/credentials 1 3 2 4 2 57% 0 New Item
Sale information (sale prices, sale announcement, etc.) 3 1 2 2 3 52% 0 Dholakia & Rego, 1998 Song & Zahedi, 2005
Purchase/Reservation Information
Online 4 2 4 4 3 81% 8 New Item
Offline 2 4 4 5 4 90% 0 Ranganathan & Ganapathy, 2002
Product safety information, guidelines or warnings 1 1 2 1 3 38% 0 Dholakia & Rego, 1998
Contest or giveaway information 1 1 1 0 2 24% 0 Dholakia & Rego, 1998
PERSONALIZED INFORMATION
Customer name appears on website 2 0 2 3 4 52% 0 New Item
Customer preferences tracked/used on site 2 0 2 4 4 57% 0 Aladwani & Palvia, 2002 Barnes & Vidgen 2001 Loiacono et al. 2007 Song & Zahedi, 2005
Product recommendations/suggestions made 4 0 2 3 3 57% 1 Song & Zahedi, 2005
ADVERTISEMENTS
Banner Ad 1 0 0 0 0 5% 0 Dholakia & Rego, 1998 Zhang, et al., 2001
Side Ad 1 1 1 0 0 14% 0 New Item
Embedded Ad 1 0 0 0 0 5% 0 New Item


(Table B-1 Continued)
CUSTOMER SERVICE INFORMATION
Company warranty policy (blanket, for all or most products) 4 3 2 1 1 52% 0 Song & Zahedi, 2005
Return/Refund/Exchange policy 3 2 2 0 4 52% 0 Aladwani & Palvia, 2002
Order Tracking 4 0 0 1 2 33% 0 Song & Zahedi, 2005
Customer Service contact info Barnes & Vidgen 2001 Song & Zahedi, 2005 Webb & Webb 2004
Phone 4 4 4 5 4 100% 6 New Item
Email 4 4 4 5 3 95% 8 Aladwani & Palvia, 2002
Customer Service hours 2 1 0 4 1 38% 0 New Item
Indication of customer service online conversation/chat capability 0 0 0 0 0 0% 0 Robbins & Stylianou, 2003 Song & Zahedi, 2005 Zhang, et al., 2001
TRANSACTION INFORMATION
Indication of online purchase functionality 4 0 4 4 4 76% 0 Aladwani & Palvia, 2002 Loiacono et al. 2007
Taxes and other charges 4 0 3 3 4 67% 0 van Iwaarden, et al, 2003
Total price 4 0 3 4 4 71% 0 van Iwaarden, et al, 2003
List of individual items being purchased 4 0 3 3 4 67% 0 van Iwaarden, et al, 2003
Item-by-item price list of items being purchased 4 0 3 4 4 71% 0 van Iwaarden, et al, 2003
Delivery date estimation 3 0 2 1 3 43% 0 Barnes & Vidgen 2001 van Iwaarden, et al, 2003
Shipping options 4 0 4 0 1 43% 0 Song & Zahedi, 2005
Payment options 4 0 4 3 4 71% 0 Song & Zahedi, 2005 van Iwaarden, et al, 2003
Third party security assurance (seal, endorsement, etc.) 4 0 1 2 2 43% 0 Dholakia & Rego, 1998


(Table B-1 Continued)
Shopping cart status 4 0 3 2 4 62% 0 New Item
Individual accounts with login and password 3 0 3 4 4 67% 0 Ranganathan & Ganapathy, 2002 Zhang and von Dran, 2002 Zhang, et al., 2001
Information on offline modes for conducting financial transactions 2 3 4 3 4 76% 0 Ranganathan & Ganapathy, 2002
COMPANY INFORMATION
Company logo 4 4 3 5 4 95% 0 Grazioli and Jarvenpaa, 2000
Company retail sites
List 1 4 0 4 0 43% 0 Grazioli and Jarvenpaa, 2000
URL 0 1 1 2 0 19% 0 New Item
Map 2 2 1 3 0 38% 0 New Item
Partner-company retail sites
List 3 1 2 0 3 43% 0 New Item
URL 2 1 1 1 2 33% 8 New Item
Map 1 0 0 0 1 10% 0 New Item
Company contact information Ranganathan & Ganapathy, 2002
Phone 4 4 4 5 4 100% 6 Aladwani & Palvia, 2002 Robbins & Stylianou, 2003
email 4 4 4 5 4 100% 11 Aladwani & Palvia, 2002 Robbins & Stylianou, 2003
Mail address 4 4 4 5 2 90% 4 New Item
HQ Address 2 4 3 4 2 71% 0 Robbins & Stylianou, 2003
Company history 4 1 2 5 3 71% 0 Aladwani & Palvia, 2002 Robbins & Stylianou, 2003
Press Releases 4 1 1 5 4 71% 9 Robbins & Stylianou, 2003


(Table B-1 Continued)
Company Goal, Mission or Vision 4 2 2 4 2 67% 0 New Item
Celebrity endorsement of company/brand 1 4 0 0 0 24% 0 New Item
MULTIMEDIA
Does the site have Entertainment content? Loiacono et al. 2007 Ranganathan & Ganapathy, 2002
Image 1 0 0 2 4 33% 7 Dholakia & Rego, 1998
Game 0 0 0 0 0 0% 0 New Item
Multimedia 1 1 0 2 4 38% 4 Aladwani & Palvia, 2002 Barnes & Vidgen 2001 Robbins & Stylianou, 2003 Zhang, et al., 2001 Zhang and von Dran, 2002
SECURITY
Does the site require login with user name and password? 3 0 2 4 3 57% 0 Ranganathan & Ganapathy, 2002 Zhang & von Dran, 2002 Zhang, et al., 2001
Does the key/lock display on status bar for insecure pages? 4 0 3 5 4 76% 0 New Item


APPENDIX C
Screen captures of the landing page for the Max, High-med, High-low, and High-only
versions of each authors website.
Table C-1: Screen captures of the Romtvedt experimental websites (landing pages).
Romtvedt High-Low Romtvedt High-Only
88


Table C-2: Screen captures of the Milofsky experimental websites (landing pages)
<
David Milofsky
Milofsky High-Med
David Milofsky
M;:i V - . -5 r >-j V f . V V_ iy
mmmm
Map
ir
MiM
mm


Milofsky High-Only
89


Table C-3: Screen captures of the Cooperman experimental websites (landing
pages).
Cooperman Max
Robert C ooperman
Cooperman High-Low
Cooperman High-Med
Robert Cooperman
Welcome to wvwv.hobcooperman.com ?!
Cooperman High-Only
90


Full Text
Hasley, Joseph, P. (Ph.D, Computer Science and Information Systems)
The Effects of Website Information Utility on the Outcomes of User-Website Interactions
Thesis directed by Associate Professor Dawn G. Gregg
ABSTRACT
This study investigates the relationships between website information content utility
and various outcomes of user interactions with e-tail websites. Although previous research
has consistently identified high quality information content as a critical factor of successful e-
commerce websites, those studies have not reported how to identify the specific information
cues that comprise high-utility information content. In this study, we demonstrate how a new
instrument, the Website Information Content Survey, can be used to accurately and reliably
assess website information content. We also demonstrate how the MaxDiff statistical method
can be used to assess website information content utility. Finally, to investigate the
relationships between website information content utility and various outcomes of user-
website interactions (perceived information quality, perceived design quality, engagement,
trust, and risk), a 4x2 full-factorial experiment was performed.
This abstract accurately represents the content of the candidate'^^hesjs. I recommend its
publication.
Signed