16-Sep-2000&

Abstract
Web sites are being widely deployed commercially. As the widespread use and dependency on Web technology increases,
so does the need to assess factors associated with Web site success. The objective is to explore these factors in the context of
electronic commerce (EC). The research framework was derived from information systems and marketing literature.
Webmasters from Fortune 1000 companies were used as the target group for a survey. Four factors that are critical to Web site
success in EC were identi®ed: (1) information and service quality, (2) system use, (3) playfulness, and (4) system design
quality. An analysis of the data provides valuable managerial implications for Web site success in the context of electronic
commerce. # 2000 Elsevier Science B.V. All rights reserved.
Keywords: Web site success; Electronic commerce; Fortune 1000; Cybermarketing
1. Introduction
Web sites are being widely deployed throughout
industry, education, government, and other institutions.
In practice,theimportance ofthe use ofWebtechnology
for electronic commerce (EC) activities has been discussedwidely (e.g., [32,34,50,58,59,61]). ECis away of
conducting business by companies and their customers
performing electronic transactions through computer
networks [19]. EC can help business organizations
cut costs, interact directly with customers, run more
smoothly and in a more timely manner, and even better,
it can help an organization outperform its competition.
As the dependency on Web technology increases, so
does the need to assess factors associated with Web site
success. Although there has been signi®cant research
on supporting EC, existing empirical research focusing
on success factors of Web sites is mainly anecdotal and
exploratory in nature. Few studies involved more than
one or two measurement variables involved in a Web
site design. Thus, while there should be a considerable
number and variety of factors associated with Web sites
success, little knowledge exists above the combination
ofthese factors. Inaddition,the preponderanceofstudies
focuses on building security for on-line transactions
on the Web [31,43]. Customers would not pay for
products or services over the Web if ®nancial information could not be transmitted securely: secure transactions are criticaltothe success. However, securityis only
a necessary but not a suf®cient condition of designing
a successful Web site: a secure Web market does not
guarantee customers.
Information & Management 38 (2000) 23±33
*Corresponding author. Tel.: 1-662-325-1999;
fax: 1-662-325-8651.
E-mail addresses: m10cxl1@wpo.cso.niu.edu (C. Liu),
kpa1@ra.msstate.edu (K.P. Arnett). 1 Tel.: 1-815-753-1185.
0378-7206/00/$ ± see front matter # 2000 Elsevier Science B.V. All rights reserved.
PII: S 0378-7206(00)00049-5
2. Specification of Web site success
The general de®nition of IS success is: the extent to
which a system achieves the goals for which it was
designed [23]. A Web site is a new type of information technology. In the context of EC, the functions
and features provided by companies' Web sites can
be classi®ed into three phases of marketing: pre,
on-line, and after sales [39]. Any EC activity ®ts
within these three classi®cations. The pre-sales phase
includes a company's efforts to attract customers
by advertising, public relations, new product or
service announcements, and other related activities.
Customers' electronic purchasing activities occur in
the on-line sales where orders and charges are placed
electronically through Web facilities. Kotler [33]
stressed that trustworthy, dependable, and reliable
characteristics are important to trigger business
transactions. The after-sales phase includes customer
service, problem resolution etc. This phase should
generate or obtain customer satisfaction by meeting
demand and pleasing customers. Thus, a successful
Web site, in the context of EC, is one that attracts
customers, makes them feel the site is trustworthy,
dependable, and reliable and generates customer
satisfaction.
3. Theoretical framework
As EC on the Web deals with both IS and marketing
activities, literature from both areas is appropriate in
the research context. In the marketing arena, consumer
information search strategies and measuring service
quality were investigated. In the IS arena, a search was
made of IS management, measuring IS success, and
end-user computing.
3.1. Information quality
Prior research employed various measures of IS
success, including user satisfaction [2,28,35,52], business pro®tability [13,44], improved decision quality
and performance [42,49,54,62], perceived bene®ts of
information systems [20,30,51], and the level of system usage [21,22]. All of them stressed the importance
of information quality. This leads to the following
hypothesis:
H1. Information quality is directly related to Web site
success.
3.2. Learning capability
EC is an interactive function between customers
and business enterprises [9]. Many studies have
emphasized the importance of the two-way on-line
communication between customers and ®rms (e.g.,
[5,9,16,41]). Such knowledge will not only facilitate
building relational markets but also increase customers' abilities to learn how to browse and to
®nd relevant information on the Web. Business
on-line can pro®t from the interactive culture on the
Web [6].
For many potential customers, using Web technology for EC activities is a new experience. Also,
providing interactive learning tools is necessary since
consumers need to develop and apply their abilities
through exploratory behavior [60]. Thus, we propose:
H2. Learning capability is directly related to Web site
success.
3.3. Playfulness
The importance of playfulness has been emphasized
by Web site designers. A study by Rice [53] suggests
that the likelihood of a repeat visit to a Web site is
enhanced when the visitors ®nd the visit enjoyable.
In the context of marketing, hedonic value re¯ects
shopping's potential entertainment and emotional
worth [15]. A satis®ed customer not only comes from
an extrinsic reward of purchasing products or services
but also from personal and emotional reward from
purchasing-derived pleasure [29]. This suggests that
shopping on the Web produces both hedonic and
utilitarian outcomes.
There is a need for Web designers to cultivate
hedonic pleasure in site design by motivating customers to participate, promoting customer excitement
and concentration, and including charming features to
attract customers and to help them enjoy the visit. This
will lead to increased customer activities [55]. Therefore, another hypothesis is:
H3. Playfulness is directly related to Web site success.
24 C. Liu, K.P. Arnett / Information & Management 38 (2000) 23±33
3.4. System quality
According to a survey conducted by the European
Electronic Messaging Association, more than 79% of
respondents said that design quality, especially security, is the top concern of EC customers [56]. However,
security is only one aspect of designing the system
quality. Anderson and Bezuidenhoudt [3] stressed that
reliability is also needed, especially in consumer
electronic markets. A reliable system should have
quick error recovery and ensure correct operation
[10]. Thus, we propose:
H4. System quality is directly related to Web site
success.
3.5. System use
The way in which customers use a Web site for EC
is also important. Success of the IS is often employed
as a measurement of success of the entire system [27].
Also, system use can be an important determinant of
user satisfaction [12].
System use can be measured in several ways. Friedman [24] concluded that obtaining consumers' con-
®dence in EC transactions is very important. Without
it, customers will not use on-line sales and payment
functions. Customers should be able to trust the
system and use its on-line purchase capabilities [1].
They should feel that the system is both under their
control and easy to use. In addition, Web designers
should allow customers to track their on-line order
status [40]. Thus, another hypothesis is:
H5. System use is directly related to Web site success.
3.6. Service quality
Prior studies have stressed the importance of
providing high quality of service [57,63]. Business
organizations and Web designers should actively seek
ways to improve service quality at Web sites. To
make it more challenging, management and Web
designers should carefully consider how to arrange
and present customer service opportunities. This
care is necessary because of the lack of face-to-face
contact on a Web site. Thus, we propose the ®nal
hypothesis:
H6. Service quality is directly related to Web site
success.
4. Research methodology
Fig. 1 illustrates the research framework. The general methodology involved an electronic questionnaire
survey of webmasters from Fortune 1000 companies.
Webmasters are typically responsible for managing
Web sites or home pages and serve as the implementers of marketing strategy. As a result, they should
have rich information about their Web sites since these
Web sites are used as bridges to connect customers and
internal business organizations [7]. Despite early success of small business on the Web, large business
organizations have historically provided leadership in
the use of information technology [38]. Therefore, the
use of the Fortune 1000 companies as the target group
seemed most appropriate.
4.1. Sampling procedure
The mailing list of the webmasters was determined
by visiting each Fortune 1000 home page. Their URL
addresses were searched through the Netscape Search
Engine and Hoover's on-line database. At the time of
visit, the webmaster's e-mail address was recorded.
FORTUNE provides summary information on the
Web regarding the performances of the Fortune companies. A searchable database of the company,
Hoover's Online (http://www.hoovers.com), was used
to obtain URLs. Netscape's Net Search was used to
obtain URL addresses that were unavailable from
Hoovers.
The proposed questionnaire was evaluated by a
person-to-person visit to six webmasters who are
considered to be content experts. The survey questionnaire was also pre-tested for content and readability by using webmasters of the top 100 Web sites
that were earlier identi®ed by PC Magazine. The
purpose of this was to further examine the content
validity of the questionnaire and to estimate the
response rate for a large sample survey. A low
response rate of 5% from this pre-test suggested the
need for a more appealing cover letter and possibly the
use of an electronic questionnaire sent individually to
each webmaster. Both of these changes were made for
C. Liu, K.P. Arnett / Information & Management 38 (2000) 23±33 25
the ®nal questionnaire, which was delivered to available Fortune 1000 webmasters in two formats: directly
sent through e-mail and via a home page. Webmasters
were asked to select one format.
4.2. Measurement of variables
The research model was derived from the study of
IS and marketing literature. Potential measurement
variables were derived from key word searches of
electronic market, EC, electronic transaction, and
electronic marketplace in ABI/INFORM, an on-line
database marketed by University Micro®lms (UMI).
The use of on-line databases, and ABI/INFORM in
particular, as a research tool has been well established
[47]. The key word searches yielded about 1000
relevant `hits.' These were scanned by reading titles
and abstracts. All variables in the survey were measured on a seven-point Likert scale from (1) completely unimportant to (7) completely important. Table 1
shows the research constructs, their measurement
variables, and the internal reliability assessment.
4.2.1. Information quality
From our literature review, we selected the following variables for measuring information quality: accuracy, timeliness, relevance [4]; ¯exible information
presentation; customized information presentation;
price information; product/service comparability, product/service differentiation, complete product/service
description [14]; perceived information quality on
product/service; satisfying ethical standard [36]; and
support business objectives [45]. Our combined
literature from the relevant, but separate, disciplines
indicates these variables are important aspects of
IS quality. On account of the diversity in variable
identi®cation, there is no justi®cation for assigning
different weights to the variables. Thus, the average
Fig. 1. Research framework.
26 C. Liu, K.P. Arnett / Information & Management 38 (2000) 23±33
score of these variables is our measure of information
quality.
4.2.2. Learning capability
Five variables were used to measure learning capability: well organized hyperlink, help function; customized search engine [11]; interactive function
between customers and businesses, and interactive
function among customers. Again, the absence of
contrary justi®cation allows us to use the average
score of these variables our measure.
4.2.3. Playfulness
This is a ®ve-item instrument adapted from the
measurement used by Badin, Darden, and Grif®n
[8]. The variables are: enjoyment, excitement, feeling
of participation, escapism, and charming. The average
score of these variables is our measure.
4.2.4. System quality
This was measured by six variables: rapid access
(processing speed), quick error recovery, correct
operation and computation; security [17]; balanced
payment method between security and ease of use
[48]; and coordination to support all functional areas.
The average score of these variables is our measure.
4.2.5. System use
As discussed in Section 3.5, the measurement
variables of system use are: customers control of a
transaction process, ease of use, con®dence, tracking
order status, and privacy. The average score of these
variables is our measure.
4.2.6. Service quality
Quick responsiveness, assurance, reliability, empathy, and follow-up service are used to measure service
quality. These measurements are well established in
marketing literature. The average score of these variables is our measure.
4.3. Reliability of the measures
In order to ensure that the variables comprising each
proposed research construct were internally consistent, reliability assessment was carried out using
Cronbach's alpha. A low value of Cronbach's alpha
(i.e. close to 0) implies that the variables are not
internally related in the manner expected [18]. Since
the mean values of `a feature to compare product/
service with competitors' (mean3.91) and `interactive communications among customers' (mean3.96)
were lower than 4.0, indicating a relative unimportance on the scale, these variables were dropped
from further analysis. The internal consistency reliability coef®cients for the research constructs in this
study are all well above the 0.50 level. However, a
widely used rule of thumb of 0.60 has been suggested
by Nunnally [46], and therefore, the reliability coef®-
cient for learning capability (0.55) might be seen as
inadequate.
4.4. Validity of the measure
To ensure content validity, a thorough examination
was made of the relevant literature. To further reduce
the possibility of non-random errors, six webmasters
Table 1
Research factors, measurements, and reliability assessment
Hypothesis number Research construct Measure component a
H1 information quality relevant; accurate; timely information; flexible and customized information
presentation; products/services differentiation; complete description of products/
services; price information; satisfying ethical standards; perceived products/services
quality; information to support business objectives
0.78
H2 learning capability interactive function between customers and business organization;
well defined link; help function; customized search engine
0.55
H3 playfulness enjoyment; excitement; feeling of participation; charming; escapism 0.83
H4 system quality security; rapid accessing; quick error recovery; precise operation and computation;
balanced payment method between security k, ease of use; coordination
0.75
H5 system use confidence; control; ease of use; track on-line order status; privacy 0.93
H6 service quality quick responsiveness, assurance; empathy; following-up service 0.86
C. Liu, K.P. Arnett / Information & Management 38 (2000) 23±33 27
and PC Magazine's top 100 Web sites webmasters
were asked to review the questionnaire for validity
(measuring what is intended), completeness (including all relevant variable items), and readability (making it unlikely that webmasters will misinterpret a
particular question). Three questions were deleted and
®ve were reworded to improve the readability.
5. Data analysis and results
Only 762 of the Fortune 1000 companies were
found to have public home pages through Hoovers
and Infoseek search engines at the time of this study.
Of the 762 companies, a total of 689 webmaster's email addresses were collected by browsing the companies' home pages and/or completing their electronic
feedback form to request the e-mail address. It is
interesting to note that about 15 home pages of Fortune 1000 companies were created and maintained by
other companies, such as ImageSoft, FCGNet, Computer Graphics, Webvision, Internet Publishing etc.
Since these design companies are responsible for
managing their clients' home pages, their webmasters
were also included in the study.
The survey was ®rst electronically mailed to 689
webmasters of Fortune 1000 companies. The number
of undelivered and returned questionnaires was 28
so that 661 total questionnaires were mailed. This
mailing received 98 responses. A follow up noti®cation and a second copy of the questionnaire resulted in
24 additional responses, giving a total of 122
responses. Of these, three were rejected because many
items were left blank, yielding a ®nal usable response
rate of 18%. Non-response bias was examined by
comparing the industry type of the respondents to
the entire sample of Fortune 1000 companies. The
Chi-square goodness-of-®t (Chi- square12.17,
p<0.06) test showed that industry type of respondents
were not signi®cantly different from the Fortune 1000
companies as a whole.
Table 2 presents the characteristics of the respondents. The responding webmasters represent a broad
coverage of industry classes, which indicates that
the survey results can be used to explain webmasters' perceptions for design quality of electronic
marketplaces on the Web across different types of
industries.
5.1. Hypothesis testing
One purpose of the webmaster questionnaire was to
provide data in order to test the research hypotheses.
Mean values and a matrix of intercorrelations among
the research constructs were calculated. The average
response for the six items is considered by us to be the
measure of the overall web design importance value. If
the overall importance mean value rating correlated
positively and signi®cantly with the six research conTable 2
Characteristics of respondentsa
Number Percentage (%)
1. Industry
Construction 2 1.68
Finance, insurance,
and real estate
16 13.45
Manufacturing 38 31.93
Retail trade 8 6.72
Service 18 15.13
Transportation, communications,
electric, gas and sanitary services
28 23.53
Wholesale trade 2 1.68
Others 2 1.68
Missing 5 4.20
Total 119 100
2. Gender
Male 79 66.39
Female 36 30.25
Missing 4 3.36
Total 119 100
3. Age group
20±25 12 10.08
26±30 22 18.49
31±35 28 23.52
36±40 11 9.24
40±45 15 12.61
Greater than 45 25 21.01
Missing 6 5.04
Total 119 100
4. Job length as webmaster
Less than 6 months 12 10.08
6±12 months 37 31.09
13±24 months 46 38.66
Greater than 24 months 17 14.29
Missing 7 5.88
Total 119 100
a Note: The classification of the industry type is based on
Fortune Magazine.
28 C. Liu, K.P. Arnett / Information & Management 38 (2000) 23±33
structs, the six hypotheses could be supported.
The means, standard deviations, and matrix of intercorrelations among the six research constructs are
presented in Table 3. The overall web design importance rating correlated positively and signi®cantly
with all six independent constructs. The probabilities
( p values), which are shown in parentheses, are less
than 0.01. Therefore, research hypotheses H1±H6 can
be supported.
5.2. Factor analysis
In order to further determine factors associated with
Web site success, an exploratory factor analysis was
performed after hypotheses testing. Kaiser's measure
of sampling adequacy (MSA) was calculated. The
overall MSA was 0.86. In addition, all individual
variables' MSAs (except for the two that were
dropped) were greater than 0.70. This clearly suggests
that factor analysis can be used to extract research
factors [25].
Several rules are typically applied when addressing
how many factors to extract. To obtain a meaningful or
interpretable grouping of the variables, we employed
the rules of eigenvalue greater than 1, percentage of
variance extracted accounts for at least 5% of the
common variance, and the Screen test. Four factors
were extracted. To obtain a simpler and theoretically
meaningful factor pattern, an oblique rotation with
PROMAX was applied. Here, a desired level of signi®cant factor loadings should be speci®ed to explain
the factor rotation results. Various researchers have
given different cut-off values for retention based on
the value of factor loadings. Some used the cut-off
value of 0.35 [37], while others used the cut-off value
of 0.50. In order to obtain meaningful factor rotation
results, both cut-off values of 0.35 and 0.50 were
selected to evaluate the factor patterns. The cut-off
value of 0.35 obtains three additional variables for the
fourth factor. Based on Hatcher's suggestion [26] that
at least three variables with signi®cant loadings should
be included on each retained factor, a cut-off value of
0.35 was applied for this study. Table 4 presents the
factor structure with the names of the factors being
subjectively inferred from the nature of the grouped
items.
After the factor analysis, a reliability test was
performed for the extracted factors. None of the four
factors' alpha is lower than 0.6. Consequently, these
factors provide a reliable and consistent measure of
intended dimensions and no further elimination of
variables appears necessary.
The factor analysis shows that only four factors are
really justi®ed; they are: (1) information and service
quality, (2) system use, (3) playfulness, and (4) system
design quality. We note that the reliability assessment
of learning capability (a0.55) is below the normal
acceptable level (a0.60). Therefore, it is not surprising that a learning capability factor did not emerge
from the factor analysis. Also, because a service
encounter on a Web site has no face-to-face contact,
it may be so different from traditional customer
service activities that it is just a part of the overall
information quality.
6. Conclusions and managerial implications
Apparently, Web site success in the context of EC is
related to four major factors: quality of information
and service, system use, playfulness, and system
design quality. Organizations who launch Web sites
should be more aware of these factors. Based on the
results, several recommendations can be advanced.
First, business organizations and Web developers
should actively seek ways to improve information and
service quality provided through Web sites. Business
organizations and Web designers should establish a
service-oriented concept for both pre-sale and aftersale stages to provide high quality service and high
quality information. For example, a Web site may
provide a recommendation for a particular plug-in to
allow a better presentation of its products/services,
and the site might also help customers download/
upgrade their plug-in. Here, both service and information quality may be enhanced. A service-oriented
concept aims at serving customers better through all
phases of marketing activities.
Second, business organizations and Web site
designers should focus on the way in which customers
use a Web site. The results indicate the importance, in
general, of successful Web site design to system use.
Customers rather than business organizations should
control the on-line transaction process.
Third, there is a need for business organizations and
Web developers to cultivate hedonic pleasures in the
C. Liu, K.P. Arnett / Information & Management 38 (2000) 23±33 29
Table 3
Matrix of intercorrelations among study constructs (N119)a
Construct Mean S.D. 1 2 3 4 5 6 7
1. Well designed importance 5.90 0.74 1.00 (0.0)
2. Information quality 5.64 0.68 0.27 (0.0028) 1.00 (0.0)
3. Learning capability 5.33 0.83 0.26 (0.0041) 0.72 (0.0001) 1.00 (0.0)
4. Playfulness 5.10 1.04 0.35 (0.0001) 0.37 (0.0001) 0.36 (0.0001) 1.00 (0.0)
5. System quality 5.49 0.99 0.31 (0.0006) 0.64 (0.0001) 0.62 (0.0001) 0.30 (0.001) 1.00 (0.0)
6. System use 5.47 1.39 0.30 (0.0008) 0.50 (0.0001) 0.53 (0.0001) 0.21 (0.0189) 0.69 (0.0001) 1.00 (0.0)
7. Service quality 6.16 0.92 0.42 (0.0001) 0.70 (0.0001) 0.59 (0.0001) 0.39 (0.0001) 0.59 (0.0001) 0.53 (0.0001) 1.00 (0.0)
a Note: (1) p values are in parentheses; (2) the measurement scale of mean values is from 1 (completely unimportant) to 7 (completely important).
30 C. Liu, K.P. Arnett / Information & Management 38 (2000) 23±33
Web site by motivating customers to participate,
promoting customer excitement and concentration,
and including charming features to attract customers
and to help them enjoy the visit. Creativity must be
incorporated into the design process in order to obtain
customers' psychological satisfaction when engaging
in marketing on the Web.
Fourth, the results corroborated the hypothesized
direct relationship of system design quality with Web
site success. Surprisingly, security is absent in the
system design quality factor. But without security, no
customers would shop around. However, security is
only a necessary condition; alone it can not attract
customers and promote electronic marketing activities.
7. Limitations
The primary limitation of this research is that data
about Web site success was gathered from webmasters. These perceptions tell us what these important
people in the web design process believe, but they are
not necessarily grounded in fact. In addition, the
design and maintenance of an electronic marketplace
on the Web is still in relative infancy, so there is
limited knowledge for both consumers and businesses
as to how to pursue electronic marketing activities on
the Web. Although these results provide some important guidelines for the design of a Web site, continual
monitoring of the development and functionality of
Web sites will be needed. The data presented is crosssectional, and longitudinal data will likely be needed
in the future because of the dynamics of Web-enabled
commerce.
Another limitation is that the results cannot be
generalized to all businesses. It is true that large
organizations generally provide leadership in using
information technology, but differences exist between
small and large businesses, especially in using the
Web to compete. Therefore, careful use of the results
should be made, especially as to their applicability to
small businesses.