Professional Documents
Culture Documents
Wixom - Teori Kepuasan Pengguna and Technology Acceptance
Wixom - Teori Kepuasan Pengguna and Technology Acceptance
I n general, perceptions of information systems (IS) success have been investigated within two primary research
streams—the user satisfaction literature and the technology acceptance literature. These two approaches have
been developed in parallel and have not been reconciled or integrated. This paper develops an integrated
research model that distinguishes beliefs and attitudes about the system (i.e., object-based beliefs and atti-
tudes) from beliefs and attitudes about using the system (i.e., behavioral beliefs and attitudes) to build the
theoretical logic that links the user satisfaction and technology acceptance literature. The model is then tested
using a sample of 465 users from seven different organizations who completed a survey regarding their use of
data warehousing software. The proposed model was supported, providing preliminary evidence that the two
perspectives can and should be integrated. The integrated model helps build the bridge from design and imple-
mentation decisions to system characteristics (a core strength of the user satisfaction literature) to the prediction
of usage (a core strength of the technology acceptance literature).
Key words: user satisfaction; technology acceptance model; information systems success; theory of reasoned
action; system quality; information quality
History: V. Sambamurthy, Senior Editor; Ritu Agarwal, Associate Editor. This paper was received on October
9, 2002, and was with the authors 13 months for 3 revisions.
how to influence usage through design and imple- are based on a sample of 465 users of data warehous-
mentation (Taylor and Todd 1995, Venkatesh et al. ing predefined reporting software from seven different
2003). For example, designers receive feedback regard- organizations. Finally, in §5, we provide a discussion
ing ease of use and usefulness in a general sense, but of the findings and an agenda for future research.
they do not receive actionable feedback about impor-
tant aspects of the IT artifact itself (e.g., flexibility, inte- 2. Theoretical Development
gration, completeness of information, and information
2.1. Understanding Behavioral Beliefs
currency). Such guidance was a core objective in the
and Attitudes
development of TAM, but one that has received lim-
According to the expectancy-value theory developed
ited attention (Davis et al. 1989).
by Ajzen and Fishbein (1980), external variables influ-
Although user satisfaction and technology accep-
ence beliefs about the outcomes associated with per-
tance have evolved largely as parallel research
forming a behavior, which in turn shape attitudes
streams, the two approaches can and should be inte-
toward performing a behavior. Attitude, in turn,
grated (Goodhue 1988, Hartwick and Barki 1994,
influences intention to perform the behavior and,
Melone 1990, Seddon 1997). Such integration can help
ultimately, influences the behavior itself. Satisfaction
build a conceptual bridge from design and imple-
in a given situation is a person’s feelings or atti-
mentation decisions to system characteristics to the tudes toward a variety of factors affecting that situ-
prediction of usage. Ultimately, this would improve ation. As articulated in the theory of reasoned action
the predictive value of user satisfaction and augment (TRA), these relationships will be predictive of behav-
the practical utility of technology acceptance. Further- ior when the attitude and belief factors are specified in
more, by theoretically integrating the two very impor- a manner consistent with the behavior to be explained
tant IT research streams, we can answer the call to in terms of time, target, and context (Ajzen and Fish-
provide a way for perception-based IT research to bein, in press; Fazio and Olson 2003). Within the IT
more fully examine the role of the IT artifact (Benbasat literature, these ideas have taken shape in the form
and Zmud 2003, Orlikowski and Iacono 2001). of the TAM. TAM has been widely applied to under-
To accomplish this, we apply concepts from the stand the attitude one holds about the use of technol-
broader attitude literature (e.g., Ajzen 2001; Ajzen and ogy, which is used to predict the adoption and use
Fishbein, in press; Eagly and Chaiken 1993; Fazio and of information technology. The attitude construct in
Olson 2003; Haddock and Zanna 1999). Specifically, TAM represents attitude toward the behavior of using
the paper develops a model that explicitly distin- technology.
guishes the object-based beliefs and attitudes found in Over the past decade, the technology acceptance
the user satisfaction literature from behavioral beliefs literature has included a large number of empirical
and attitudes in the technology acceptance literature. tests, comparisons, model variants, and model ext-
It enumerates a set of system and information charac- ensions. As Figure 1 illustrates, researchers have ext-
teristics that influence system and information qual- ended TAM in three primary ways to provide greater
ity, describes how they in turn influence object-based understanding and explanatory power and addi-
beliefs and attitudes with the system and the infor- tional points of managerial leverage in its application.
mation it produces, and then describes how these The first approach involves introducing factors from
object-based attitudes toward the system can shape related models, such as subjective norm, perceived
the behavioral beliefs of usefulness, ease of use, and, behavioral control, and self-efficacy (e.g., Hartwick
ultimately, system usage. and Barki 1994, Taylor and Todd 1995, Mathieson
The remainder of the paper proceeds as follows. Sec- et al. 2001). A second approach involves introducing
tion 2 builds the theoretical arguments for the pro- additional or alternative belief factors to the model.
posed research model. In §3, we present the back- Most often, this includes adding key related factors
ground for a preliminary study that tested this model from the diffusion of innovation literature, such as
in the context of data warehousing. The results of an trialability, compatibility, visibility, or result demon-
empirical test of this model are presented in §4. They strability (Agarwal and Prasad 1997, Karahanna et al.
Wixom and Todd: A Theoretical Integration of User Satisfaction and Technology Acceptance
Information Systems Research 16(1), pp. 85–102, © 2005 INFORMS 87
Behavioral beliefs
Behavioral attitude
Perceived
C. External usefulness
variables
Attitude toward Intention to Usage
(e.g., demographics,
usage use
system characteristics,
personality traits) Perceived
ease of use
A. Factors from
related models
(e.g., subjective
B. Additional norm, perceived
belief factors behavioral control)
(e.g., trialability,
compatibility)
1999, Plouffe et al. 2001). A third approach has been to might influence the core beliefs in TAM, and might
examine external variables, which are antecedents to indirectly shape system usage.
or that moderate the influence of ease of use and use-
2.2. Understanding Object-Based Beliefs
fulness within the TAM, such as personality traits and
and Attitudes
demographic characteristics (e.g., Gefen and Straub
In contrast to the technology acceptance literature,
1997, Venkatesh 2000, Venkatesh and Morris 2000).
system and information characteristics have been core
Venkatesh et al. (2003) provide a comprehensive
elements in the literature on user satisfaction (DeLone
examination of eight different models and derive a
and McLean 1992). Within this literature, user satisfac-
unified theory of acceptance and use of technology.1
tion is typically viewed as the attitude that a user has
Despite this extensive research activity, only a
toward an information system; therefore, it represents
handful of TAM studies have looked explicitly at the
an object-based attitude. User satisfaction primarily
role of system characteristics as antecedents to ease
has been measured by various subsets of beliefs about
of use or usefulness (e.g., Davis 1993, Igbaria et al. specific systems, information, and other related char-
1995, Lim and Benbasat 2000). For the most part, these acteristics (e.g., IT service).
studies have treated system characteristics at a holis- This becomes clear when one examines user sat-
tic level or have looked at a limited number of fea- isfaction instruments, such as Bailey and Pearson
tures. One exception to this is the work by Hong (1983), Baroudi and Orlikowski (1988), Doll and Tork-
et al. (2001–2002) that examines how dimensions of zadeh (1988), and Ives et al. (1983) (see Table 1). These
usability (information relevance, clarity of terminol- instruments use a characteristics-based approach for
ogy, and screen design) influence ease of use and measuring user satisfaction. Although these instru-
usefulness in the context of a digital library applica- ments have been criticized for containing an arbitrary
tion. Their results show mixed effects with only rele- assortment of characteristics (Galletta and Lederer
vance influencing both usefulness and ease of use. In 1989), the items from user satisfaction instruments
their integration of the technology acceptance litera- appear to conceptually represent a relatively small
ture, Venkatesh et al. (2003) stress the need to extend number of higher order constructs. Thus, the existing
this literature by explicitly considering system and measures of user satisfaction provide a useful base for
information characteristics and the way in which they identifying and examining the underlying structure of
system and information characteristics.
1
The Venkatesh et al. (2003) study provides an excellent review of A fundamental problem with user satisfaction
TAM studies. research has been its limited ability to predict system
Wixom and Todd: A Theoretical Integration of User Satisfaction and Technology Acceptance
88 Information Systems Research 16(1), pp. 85–102, © 2005 INFORMS
External Instrument Bailey and Ives et al. Baroudi and Doll and
variables characteristics Pearson (1983) (1983) Orlikowski (1988) Torkzadeh (1988)
usage (Davis et al. 1989, DeLone and McLean 1992, correlation between behavioral attitude and the beh-
Goodhue 1988, Hartwick and Barki 1994, Melone avior itself averaged 0.54 (Kraus 1995). Thus, better
1990, Seddon 1997). However, when one considers the understanding the theoretical relationships within the
general attitude literature, the equivocal relationship user satisfaction literature can help bridge such equiv-
between user satisfaction and usage can be under- ocal findings while offering system designers a way
stood. For a belief or attitude to be directly predictive to influence usage through design based on system
of behavior, it needs to be consistent in time, target, and information characteristics.
and context with the behavior. Therefore, satisfaction
with the system and its information output is unlikely 2.3. An Integrated Model of User Satisfaction and
to be directly predictive of the use of that system. Technology Acceptance
Instead, user satisfaction needs to be recognized The investigation of relationships among object-based
as an object-based attitude (Ajzen and Fishbein 1980, beliefs, attitudes, and behaviors has been an ongoing
p. 84) whereby it serves as an external variable with challenge in the attitude-behavior literature:
influences on intention and behavior that are fully
mediated by behavioral beliefs and attitudes (Ajzen If there is one clear conclusion to be derived from the
and Fishbein 1980; Eagly and Chaiken 1993, p. 205). work on the attitude-behavior relation it is that general
attitudes will usually not provide a good basis for pre-
For example, one’s satisfaction with the reliability of
dicting and explaining single behaviors with respect
a system does not directly impact whether one will to the attitude object; correlations of single behaviors
use the system. However, beliefs about reliability cer- with general attitudes tend to be modest at best (Ajzen
tainly will affect one’s attitude toward the system, and Fishbein, in press, p. 28).
which will shape behavioral beliefs about using the
system (e.g., ease of use). It is the system behavioral For accurate prediction, beliefs and attitudes must be
belief (ease of use) that directly influences attitude specified in a manner that is consistent in time, target,
toward use and, ultimately, usage. In the user satisfac- and context with the behavior of interest (Fishbein
tion literature, the mediating behavioral beliefs and and Ajzen 1975). This is often referred to as the corre-
attitudes are absent, and inattention to this concep- spondence principle (Fishbein and Ajzen 1975) and is at
tual gap explains the equivocal relationship between the core of the power of models such as TAM where
system satisfaction and system usage (see Figure 2). beliefs and attitudes about a specific behavior (e.g.,
Empirical evidence shows that object-based attitude the use of an e-mail system), in a particular context
is generally a weak predictor of behavior (Ajzen and (e.g., work), at a particular point in time (e.g., over the
Fishbein, in press). For example, one meta-analysis next month) are found to be predictive of intention
found that the correlation between object-based atti- and behavior. Given this, we begin to construct our
tude and behavior averaged only 0.13, whereas the research model with the right half of Figure 3. Fully
Intention
Currency
Attitude
Reliability
Flexibility
System System
Integration Ease of use
quality satisfaction
Accessibility
Timeliness
consistent with TRA, TAM, and more recent deriva- be influenced by beliefs about the consequences of
tions, such as the unified theory of acceptance and purchasing the brand. Those beliefs are shaped, at
use of technology (UTAUT) (Venkatesh et al. 2003), least in part, by the attitude toward the brand itself.
the model proposes that IT usage (the target behavior In the context of IT, beliefs about using the system to
of interest) is driven by behavioral intention, inten- accomplish a particular task will be shaped, in part,
tion is determined by attitude toward use and use- by the attitude toward the system itself; indirectly
fulness, and usefulness is a function of ease of use. these beliefs will shape attitude toward use and the
Usefulness and ease of use are both assessments of eventual usage behavior.
the consequences of using a system to accomplish Given this, we introduce the left side of Figure 3,
some task. which represents the user satisfaction literature. The
More general object-based attitudes (e.g., attitudes far-left side of the model specifies key antecedents
about a system) also can be predictive of behav- to information and system quality. These specific fac-
ioral dispositions by influencing the way in which tors are derived from a decomposition and integration
information about the behavior is perceived and of factors identified in the user satisfaction literature
judged (Fazio and Olson 2003, Eagly and Chaiken (see Table 1). Although we believe these dimensions
1993). Theoretically, these serve as external variables have general applicability, it may be that the relative
that may determine satisfaction with an object, and importance of each is contingent on a specific system
that level of satisfaction subsequently may influ- and setting. For system quality, reliability refers to
ence beliefs about the consequences of using the the dependability of system operation, flexibility refers
object (Ajzen and Fishbein, in press). More specifi- to the way the system adapts to changing demands
cally, Ajzen and Fishbein (1980, p. 9) note that “exter- of the user, integration refers to the way the system
nal variables may influence the beliefs a person holds allows data to be integrated from various sources,
or the relative importance he attaches to attitudinal accessibility refers to the ease with which information
and normative considerations.” can be accessed or extracted from the system, and
Ajzen and Fishbein (1980) illustrate the use of object timeliness refers to the degree to which the system
attitudes as external variables using the consumer offers timely responses to requests for information
context. A consumer often forms an attitude toward or action.2 It is important to note that each of these
a particular brand. That attitude is shaped by beliefs
about the brand. The consumer may also develop an 2
These five antecedents to system quality were selected based on
attitude toward purchase of the brand, which will their widespread use, representativeness, and relevance to the IT
Wixom and Todd: A Theoretical Integration of User Satisfaction and Technology Acceptance
Information Systems Research 16(1), pp. 85–102, © 2005 INFORMS 91
factors reflects perceptions of the system itself and the itself, the more likely one is to find the system to be
way it delivers information. easy to use.
Information quality is shaped by four dimensions: Consistent with the notion that ease of use will
completeness represents the degree to which the sys- influence perceptions of usefulness, our model hyp-
tem provides all necessary information; accuracy rep- othesizes that system satisfaction will influence infor-
resents the user’s perception that the information is mation satisfaction. Being able to effectively interact
correct; format represents the user’s perception of how with the system is a necessary condition to obtaining
well the information is presented; and currency rep- useful information from it. Thus, an individual’s level
resents the user’s perception of the degree to which of satisfaction with the system is likely to influence
the information is up to date.3 These dimensions his or her sense of satisfaction with the information it
determine the user’s perception of the quality of the produces.
information included in the system. To summarize, our models suggest that the tech-
Next, we assert that information and system quality nology acceptance literature and the parallel user
beliefs shape attitudes about information and system satisfaction stream are not competing approaches to
satisfaction, respectively.4 This is supported by the understanding IT usage and value. Rather, they repre-
concept from the attitude behavior literature that sent complementary steps in a causal chain from key
beliefs about objects (in this case, system and informa- characteristics of system design, to beliefs and expec-
tion quality) are linked to attitude toward an object (in tations about outcomes that ultimately determine
this case, system and information satisfaction) (Ajzen usage. Next, we present a preliminary empirical test
and Fishbein 1980). of the proposed model to assess the aptness of the
At this point, information and system satisfaction proposed relationships. The test is based on a sample
represent object-based attitudes that serve as exter- of 465 users of data warehousing predefined report-
nal variables shaping behavioral beliefs. Satisfaction ing software from 7 different organizations.
with the information produced by the system will
influence perceptions of usefulness. That is, the higher 3. Method
the overall satisfaction with the information, the more
3.1. Instrument Development
likely one will find the application of that informa-
The development of the survey instrument was pat-
tion useful in enhancing work performance. A similar
terned after the process proposed by Moore and
effect is anticipated in terms of system satisfaction.
Benbasat (1991). First, groups of questions were com-
System satisfaction represents a degree of favorable-
piled from validated instruments to represent each
ness with respect to the system and the mechanics of
construct, and wording was modified to fit the data
interaction. The more satisfied one is with the system
warehousing context to be studied. Next, 10 profes-
sors and graduate students sorted the 88 initial items
context that will be explored in this study. This list is not necessarily into 17 separate categories, identifying ambiguous
exhaustive. or poorly worded items. Items were removed, and
3
These four antecedents to information quality were selected based minor wording changes were made prior to a sec-
on their widespread use, representativeness, and relevance to the IT
ond round of sorting, which did not uncover fur-
context that will be explored in this study. This list is not necessarily
exhaustive. ther problems. The three items that were categorized
4
User satisfaction instruments also refer to other categories of most accurately were selected for each construct and
object beliefs, such as service quality that could be included in this included in a random order on the survey instru-
model. However, consistent with Seddon (1997), when the focus of ment.5 Each question was measured on a 7-point,
the model is on the use of an application, we treat only the sys- Likert-type scale, ranging from 1 (strongly disagree)
tem and information characteristics, rather than the broader set of
to 7 (strongly agree).
factors that might be used to evaluate satisfaction with overall IT
services. This is not to say that such factors are not important, but
5
rather that they are focused on the broader target of the IS function Only two questions were included for information satisfaction and
rather than on the individual application. system satisfaction to reduce redundancy.
Wixom and Todd: A Theoretical Integration of User Satisfaction and Technology Acceptance
92 Information Systems Research 16(1), pp. 85–102, © 2005 INFORMS
The context of the survey instrument was the contact collected the completed surveys and returned
success of data warehousing predefined reporting them to the researchers. Response rates varied across
software. Predefined reporting software was installed organizations (see Table 3), with an overall study
and managed by the data warehousing project team response rate of 21%, yielding 465 completed surveys.
and run by users on a regular basis to provide The average age of the respondents was 42 years,
predetermined information. This context was chosen and 40% were male. The respondents had an aver-
because of its importance and widespread use in prac- age of 12 years tenure with their organization and
tice. It was hoped that widespread interest in the topic 18 years average total work experience. Their posi-
of data warehousing would encourage individual and tions in the organizations varied from clerical to
corporate participation in the study. senior management—58% were analysts; they repre-
Before implementing the survey, the instrument was sented different functional areas across the organiza-
reviewed by academics and practitioners with knowl- tion. The demographic profile of the sample is shown
edge of survey design, IS success, and data warehous- in Table 4.
ing. Minor changes were made based on their sug- The respondents were direct, voluntary users of
gestions. The resulting survey was then pilot tested data warehousing predefined reporting software.
using respondents from a large public university to On the survey, they identified their absolute usage of
identify problems with the instruments’ wording, the system and their use relative to opportunity. Both
content, format, and procedures. For this pilot test,
absolute and relative usage were measured using
surveys were distributed to 250 active users of the
a 1 to 7 Likert-type scale, with 1 representing low
university’s data warehouse; 73 responded, result-
use and 7 representing high use. The averages for
ing in a 29% response rate. Pilot participants compl-
absolute usage and relative usage were 3.6 and 4,
eted the instruments and provided written comments
respectively, suggesting that the respondents, on aver-
about length, wording, and instructions. Two of the
age, had a reasonable level of experience using the
participants were interviewed to gain a richer under-
data warehouse software. The standard deviations for
standing of the feedback. Each construct in the pilot
absolute (1.95) and relative usage (1.46) also suggest
test showed internal consistency levels exceeding
that there was reasonable variance across the sample
0.70, as measured by Cronbach’s alpha (Nunnally
1978). in usage experience. All users accessed warehouses
Based on the results of the pilot sample, minor mod- that had been in place for at least six months.
ifications were made to the survey design. The final
survey included 76 items representing the 17 con-
4. Results
structs identified in Figure 3, as well as a series of
The research model was tested using partial least
demographic and self-reported usage items. The spec-
squares (PLS), a structural modeling technique that
ified items, organized by construct, are shown in
is well suited for highly complex predictive models
Table 2.
(Barclay et al. 1995, Chin 1998, Lohmoller 1989, Wold
3.2. Sample and Joreskog 1982). PLS was most appropriate given
To obtain study participants, an e-mail announcement the large number of constructs that resulted when the
was sent to members of The Data Warehousing Insti- satisfaction and usage models were combined. PLS
tute, offering a free study to assess the success of their Graph version 2.91 (Chin and Frye 1996) was used
organization’s data warehousing data access software. for the analysis, and the bootstrap resampling method
Seven organizations from a variety of industries (e.g., (100 resamples) was used to determine the signifi-
health care, consumer goods, financial services, and cance of the paths within the structural model.
government) agreed to participate. Each organization
was asked to distribute paper-based surveys to all of 4.1. Measurement Model
the active users of its data warehouse. All surveys The test of the measurement model includes the
were confidential; no identifying personal informa- estimation of internal consistency and the conver-
tion was required. At each organization, the study gent and discriminant validity of the instrument
Wixom and Todd: A Theoretical Integration of User Satisfaction and Technology Acceptance
Information Systems Research 16(1), pp. 85–102, © 2005 INFORMS 93
Completeness
= 090
Fornell = 094
provides me with a complete set of information. 458 177
produces comprehensive information. 488 170
provides me with all the information I need. 415 187
Format
= 089
Fornell = 092
The information provided by is well formatted. 493 168
The information provided by is well laid out. 510 157
The information provided by is clearly presented on the screen. 523 155
Accuracy
= 087
Fornell = 090
produces correct information. 514 160
There are few errors in the information I obtain from . 475 178
The information provided by is accurate. 504 166
Currency
= 093
Fornell = 094
provides me with the most recent information. 505 187
produces the most current information. 496 179
The information from is always up to date. 471 177
Information quality
= 094
Fornell = 094
Overall, I would give the information from high marks. 509 168
Overall, I would give the information provided by a high 510 163
rating in terms of quality.
In general, provides me with high-quality information. 511 161
Reliability
= 090
Fornell = 093
operates reliably. 510 173
performs reliably. 515 166
The operation of is dependable. 510 156
Accessibility
= 090
Fornell = 092
allows information to be readily accessible to me. 527 170
makes information very accessible. 516 169
makes information easy to access. 514 170
Flexibility
= 086
Fornell = 090
can be adapted to meet a variety of needs. 428 199
can flexibly adjust to new demands or conditions. 373 186
is versatile in addressing needs as they arise. 400 183
Integration
= 089
Fornell = 091
effectively integrates data from different areas of the company. 478 189
Wixom and Todd: A Theoretical Integration of User Satisfaction and Technology Acceptance
94 Information Systems Research 16(1), pp. 85–102, © 2005 INFORMS
Table 2 (cont’d.)
pulls together information that used to come from different 514 177
places in the company.
effectively combines data from different areas of the company. 493 177
Timeliness
= 080
Fornell = 087
It takes too long for to respond to my requests. (RC) 426 190
provides information in a timely fashion. 507 167
returns answers to my requests quickly. 490 172
System quality
= 091
Fornell = 094
In terms of system quality, I would rate highly. 491 169
Overall, is of high quality. 512 155
Overall, I would give the quality of a high rating. 497 162
Information satisfaction
= 093
Fornell = 096
Overall, the information I get from is very satisfying. 489 180
I am very satisfied with the information I receive from . 484 178
System satisfaction
= 092
Fornell = 095
All things considered, I am very satisfied with . 461 194
Overall, my interaction with is very satisfying. 465 182
Attitude
= 089
Fornell = 091
Using is (not enjoyable/ very enjoyable). 413 186
Overall, using is a (unpleasant/pleasant) experience. 489 179
My attitude toward using is (very unfavorable/very favorable). 498 177
Intention
= 087
Fornell = 092
I intend to use as a routine part of my job over the next year. 513 194
I intend to use at every opportunity over the next year. 480 193
I plan to increase my use of over the next year. 464 188
Ease of use
= 085
Fornell = 089
is easy to use. 531 182
It is easy to get to do what I want it to do. 439 198
is easy to operate. 520 184
Usefulness
= 082
Fornell = 088
Using improves my ability to make good decisions. 504 165
allows me to get my work done more quickly. 484 180
Using enhances my effectiveness on the job. 504 176
COMP ACCU FORM CURR RELI FLEX INTE TIME ACCE INFQ SYSQ INTN ATTI EASE USEF SYSS INFS
Completeness 091
Accuracy 055∗ 087
Format 066 049 089
Currency 057 061 048 092
Reliability 062 068 057 059 090
Flexibility 068 033 047 039 041 086
Integration 069 046 050 050 050 048 088
Timeliness 055 054 052 054 074 046 047 083
Accessibility 070 054 063 055 070 054 061 070 089
Info. quality 074 076 064 064 073 054 063 060 071 091
System quality 077 071 071 056 077 057 057 067 077 085 091
Intention 053 036 040 039 041 051 042 043 058 053 057 089
Attitude 059 049 063 044 061 056 047 062 070 067 075 071 087
Ease of use 052 042 057 038 058 044 045 056 068 058 069 055 077 085
Usefulness 060 047 049 047 054 054 053 052 066 069 067 076 075 065 085
System sat. 066 041 060 039 060 058 051 057 071 067 075 067 084 081 077 095
Information sat. 067 052 055 051 061 058 054 054 063 077 072 066 073 061 080 079 096
∗
All correlations are significant at the 0.001 level.
†
Diagonal elements are the square root of AVE. These values should exceed the interconstruct correlations for adequate discriminant validity. This condition
is satisfied for each construct.
Wixom and Todd: A Theoretical Integration of User Satisfaction and Technology Acceptance
96 Information Systems Research 16(1), pp. 85–102, © 2005 INFORMS
Discriminant and convergent validity are further the amount of variance explained by the independent
confirmed when individual items load above 0.50 on variables. Together, the R2 and the path coefficients
their associated factors and when the loadings within (loadings and significance) indicate how well the data
construct are higher than those across constructs. The support the hypothesized model.
appendix contains the loadings and cross-loadings for Figure 4 shows the results of the test of the hypoth-
items used in this study; all items loaded on their esized structural model. The paths specified in TAM
constructs as expected. Furthermore, all items loaded are all significant with the direct and indirect effects
more highly on their construct than they loaded on of usefulness, ease of use, and attitude toward use
any other construct, and in all but one case among accounting for 59% of the variance in intention. As
the 784 cross loadings the differences were greater predicted, information satisfaction (0.64) had a signifi-
than 0.10. cant influence on perceived usefulness and accounted
Finally, the data were tested for multicollinearity. for 67% of the variance in perceived usefulness. Sys-
We tested for all potential collinearity problems that tem satisfaction (0.81) had a significant influence on
had more than one predictor construct. In all cases, perceived ease of use and accounted for 65% of the
the variance inflation factor was below the 5.0 level. variance in perceived ease of use.
As expected, information quality (0.43) and sys-
4.2. Structural Model tem satisfaction (0.50) had significant influences on
The test of the structural model includes estimates information satisfaction, accounting for 71% of the
of the path coefficients, which indicate the strengths variance in that measure. System quality also was a
of the relationships between the dependent and inde- significant determinant of system satisfaction (0.73),
pendent variables, and the R2 values, which represent accounting for 53% of its variance. Completeness
Completeness
0.3
3
**
*
0.45 **
Accuracy *
Information
*
0.43*** Usefulness
4* quality
0.1 0.64*** 0.47***
Format R2 = 0.75 R2 = 0.67
1 **
0.1
0.42***
Currency
Information
satisfaction Intention
R2 = 0.71 R2 = 0.59
0.25*** 0.36***
0.50***
System Attitude
Reliability satisfaction
R2 = 0.69
0.3
R2 = 0.53
8
**
*
Flexibility 0.1
4 ** 0.73***
0.50***
Timeliness
∗ ∗∗ ∗∗∗
p < 005 p < 001 p < 0001
Wixom and Todd: A Theoretical Integration of User Satisfaction and Technology Acceptance
Information Systems Research 16(1), pp. 85–102, © 2005 INFORMS 97
(0.33), accuracy (0.45), format (0.14), and currency Table 6 Alternative Models to Predict Intentions
(0.11) were all significantly related to information Plus quality Plus system
quality and collectively account for 75% of the vari- and satisfaction characteristics All factors
ance in information quality. Reliability (0.38), flexi- TAM to intention to intention to intention
bility (0.14), integration (0.10), and accessibility (0.36) Intention R2 059 060 062 063
were all significant determinants of system quality, Attitude 036∗∗∗ 035∗∗∗ 041∗∗∗ 041∗∗∗
whereas timeliness was not. The first three factors Usefulness 047∗∗∗ 044∗∗∗ 042∗∗∗ 042∗∗∗
Information 011 011
together accounted for 74% of the variance in system satisfaction
quality. System 001 003
satisfaction
4.3. Alternative Models Information 012 013
quality
The proposed research model is clearly a single con- System quality 005 014
struction and ordering of the factors that we have Completeness 010 007
presented. The core logic for the proposed model is Accuracy 001 002
Format 012∗ 013∗
based on the correspondence principle (Fishbein and Currency 007 008
Ajzen 1975), which maintains that the beliefs and atti- Reliability 020∗∗ 022∗∗
tudes that most closely correspond to the behavior of Flexibility 006 006
Integration 003 001
interest should be the most important proximal pre-
Accessibility 009 008
dictors of those behaviors. Factors that are progres- Timeliness 001 002
sively more distant (e.g., attitude toward the system, ∗
p < 005; ∗∗ p < 001; ∗∗∗ p < 0001.
system quality, assessment of specific system factors)
should provide weaker direct effects in terms of pre-
dicting intention and, ultimately, the target behavior. addition to usefulness and attitude, only two are sig-
It certainly is possible to imagine and argue for
nificant. Collectively, the nine additional direct factors
alternative models that organize the factors under
only increase explanatory power for intention from
investigation in a different fashion. Our data is a
0.59 to 0.62.
cross-sectional sample of current users; therefore, we
Next, we tested a model where all factors were
cannot claim that our results demonstrate causal rela-
included simultaneously as direct antecedents to
tionships. Thus, to more fully assess the aptness of
intention (see Table 6). Usefulness and attitude again
the proposed research model, we examined alterna-
dominate in the prediction of intention, and the
tive models to determine the degree to which each
remaining path coefficients are generally small (8 of
predicts and explains intention.
First, we examined whether system and informa- 13 are below 0.1). The explanatory power for intention
tion satisfaction and quality could serve as direct increases marginally from 0.59 to 0.63.
antecedents to intention in the same fashion as useful- Finally, we examined whether the mediating influ-
ness and attitude. The proposed model suggests that ences of information and system satisfaction are
the influence of these factors should be fully mediated important to the model. To do this, information and
through usefulness and ease of use, and the analysis system quality were analyzed as direct determinants
bears this out (see Table 6). The additional paths to of usefulness and ease of use, respectively. With-
intention are weak (0.12 or less) and provide a trivial out the satisfaction constructs, information quality
0.01 increase in explanatory power for intention. (path = 045) and systems quality (path = 066) are
Second, we examined a pragmatic model that posits indeed significant determinants of usefulness and
the system characteristics (e.g., completeness, flexibil- ease of use. Consistent with our notion that they are
ity) as additional direct antecedents to intention (see more distal predictors than satisfaction, however, the
Table 6). The theory suggests that they will be weak path coefficients are significantly lower than in the
predictors of intention. Indeed, when these nine fac- proposed research model (down from 0.64 and 0.81).
tors are added as direct antecedents to intention in Furthermore, the R2 values for ease of use and use-
Wixom and Todd: A Theoretical Integration of User Satisfaction and Technology Acceptance
98 Information Systems Research 16(1), pp. 85–102, © 2005 INFORMS
fulness decrease from 0.65 to 0.43, and from 0.67 to results support the applicability of information and
0.55, respectively. system satisfaction as external variables to the tradi-
Overall, the alternative models provide less tional TAM beliefs about usage behavior. These spe-
explanatory power than the proposed research model cific findings form the integrative mechanisms that
and, more importantly, limit our understanding of tie concepts from the user satisfaction and technol-
the role of each class of variable in shaping intention ogy acceptance literature into a single research model.
and usage. Consistent with the underlying theory, our In this study, it is clear that the mediating roles of
empirical evidence suggests that the further removed quality and satisfaction are critical. When alternatives
a factor is from the actual usage behavior, the less are examined (Table 6), each shows deficiencies rela-
predictive it will be. tive to the proposed research model. The performance
of information satisfaction and system satisfaction is
5. Discussion consistent with the original conceptualization of exter-
The findings of this study provide a preliminary test nal variables as a “bridge between the internal beliefs,
of the viability of the research model within the con- attitudes and intentions represented in TAM and the
text of data warehousing predefined reporting soft- various individual differences, situational constraints
ware. The findings are consistent with the proposed and managerially controllable interventions imping-
theoretical foundation, and they support the need ing on behavior” (Davis et al. 1989, p. 988).
to differentiate between object-based beliefs and atti- There are several limitations to this study that
tudes (system and information quality, system and should be noted. First, the study respondents were
information satisfaction), and behavior-based beliefs direct users of data warehousing predefined reporting
and attitudes (ease of use and usefulness, attitude) software within seven large organizations. Findings
when trying to predict usage behavior. cannot be generalized to other settings and technolo-
Findings show, for example, that system and infor- gies without further research. Also, the respondents
mation characteristics explain 75% of the variance for were existing users of the technology, so the results
system and information quality. Completeness, accu- do not further our understanding of preadoption
racy, format, and currency serve as antecedents to situations.
information quality, with accuracy and completeness The exclusive use of surveys for data collection sug-
serving particularly important roles. Similarly, relia- gests that common-method bias may be present in
bility, flexibility, integration, and accessibility serve as the study’s measurements. We believe that our care-
antecedent beliefs to system quality, with reliability ful attention to survey design and the application of
and accessibility having the strongest effects. reliability and validity checks instill confidence in the
Timeliness was not found to be an important an- study’s results. Additional research that employs a
tecedent belief in this context. Typically, decisions triangulation of data collection methods would best
based on data warehouse data are strategic in nature; alleviate common-method bias concerns, however.
a fast response time is not as important as other fac- One also should note the potential for nonresponse
tors. Timeliness may have a different relative effect bias in the study. A few of the organizations’ response
with other kinds of systems, especially those that sup- rates were quite low (e.g., 13%), and there is a chance
port business operations (e.g., inventory management, that individuals who chose to complete a survey have
order-entry). As warehouses evolve to serve more different perceptions from those who did not choose
operational needs, as in the cases of real-time ware- to respond. Unfortunately, the authors had no direct
housing and business activity monitoring, timeliness interaction with the study respondents. The surveys
may become more important in data warehousing. were distributed by a company contact who con-
Perhaps most important, the proposed influences ducted the publicity and follow-up efforts. The varied
of object-based attitudes on behavioral beliefs are response rates (i.e., 13% to 53%) resulted from the dif-
demonstrated by the strong significant relationships fering knowledge and control that the contact person
between information satisfaction and usefulness, and had with his or her user base. With the larger imple-
between system satisfaction and ease of use. The mentations, the contact person had much less personal
Wixom and Todd: A Theoretical Integration of User Satisfaction and Technology Acceptance
Information Systems Research 16(1), pp. 85–102, © 2005 INFORMS 99
contact and influence. With the smaller implemen- of factors across different contexts (e.g., technologies,
tations, the contact person typically had more time applications, tasks, and settings) is accumulated, it
and a greater understanding of the users, which led might provide a basis for important design guidelines
to higher response rates. At the same time, we note and standards for building effective systems.
that across the sample there was significant variance Furthermore, we believe that the research model
in user experience with the data warehouse software has diagnostic value at any stage of a system’s imple-
and varying organizational levels and positions repre- mentation or usage process. In current times, sys-
sented among the respondents across a variety of busi- tems are less likely to be built from scratch; rather,
nesses and industries. Collectively, this heterogeneity they evolve over time. Data warehousing, for exam-
should suggest that our results are relatively robust. ple, is typically described as a journey rather than
a destination—with companies adding new applica-
5.1. Implications for Practice tions, meeting unique business needs, and creating
A key benefit of this study is that managers now have architectural changes over time. Therefore, it is impor-
a way to assess system and information characteristics tant to understand design and implementation issues
and then reliably investigate their impacts on ultimate and their ultimate effects on usage at any point within
usage through the proposed causal chain. This can the system life cycle. In fact, many of the companies
help with management activities such as task priori- that participated in our study did so because they
tization and resource allocation. For example, a man- wanted to assess the effectiveness of their existing
ager who participated in this study discovered that data warehouse and make changes accordingly.
the data currency in his data warehouse was too old
to be useful to the users. The “intention to use” scores 5.2. Implications for Research
were so low that the manager decided to take imme- This study yields three implications for research. First,
diate steps to improve the system’s data currency. this study illustrates the importance that researchers
The manager had a long list of ways to improve the understand and preserve essential theoretical relation-
data warehouse, but the information from this study ships when performing empirical research in this area.
helped him focus on which tasks would be most likely We encourage researchers to consider the principle of
to improve system usage. correspondence when designing research models that
The proposed research model also provides a mech- involve object-based and behavioral beliefs and atti-
anism for understanding and assessing the relative tudes. If one is interested in understanding how sys-
influence of detailed system and information char- tem and related design and implementation attributes
acteristics; this provides important guidance to sys- might influence system and information satisfaction,
tem designers. For example, the model first can help per se, then these should be measured as a consis-
designers understand which characteristics (e.g., com- tent set of beliefs and attitudes about system char-
pleteness, accuracy, and reliability) have the most re- acteristics. Furthermore, the aptness of such models
lative importance within the context of interest (e.g., should not be based on the relationship of these factors
data warehousing, enterprise resource planning sys- directly to usage or measures of business value. Such
tems (ERP), and mobile computing). Based on this relationships can only be understood by examining
understanding, designers can look at the scores for appropriate mediating factors related to behavioral
these characteristics and determine what kinds of beliefs and attitudes, such as ease of use, usefulness,
changes can have the most meaningful impacts. For attitude toward use, and behavioral intention as spec-
example, if users perceive significant problems with ified in TAM. Comparisons across studies also need to
the accuracy of information (a very important infor- take into account the nature of the constructs as con-
mation characteristic in this study), then designers ceptualized and operationalized to avoid mixing stud-
can focus efforts accordingly. More specifically in the ies that examine factors at differing levels of abstrac-
data warehousing context, designers can address accu- tion. Attempts to integrate such studies would only
racy through such efforts as meta data, data sourcing, contribute to conflicting and contradictory results.
quality assessment tools, and business semantic lay- This study also identifies considerable opportunity
ers. Ideally, as knowledge about the relative influences for researchers to develop the portions of our model
Wixom and Todd: A Theoretical Integration of User Satisfaction and Technology Acceptance
100 Information Systems Research 16(1), pp. 85–102, © 2005 INFORMS
that feed into TAM. We acknowledge that researchers technology acceptance research streams by apply-
have worked diligently to extend TAM (see Figure 1); ing Ajzen and Fishbein’s (1980, in press) concep-
however, we encourage researchers to investigate the tualizations of object-based and behavioral beliefs
effects of the IT artifact itself as an antecedent to ease and attitudes. It developed a research model, con-
of use, usefulness, and related factors. We believe our ducted a preliminary test using survey data from 465
research model suggests ways in which this can be data warehouse users, and analyzed the data using
accomplished. structural equation modeling. Overall, the results
Additionally, it would be useful for researchers to are largely consistent with the hypothesized model
investigate whether there is a core set of system char- and demonstrate the potential to integrate concepts
acteristics that apply broadly across a wide range related to user satisfaction and technology acceptance
of systems. Studies should systematically investigate into a single unified model. This provides an oppor-
various technologies that differ on important dimen- tunity to integrate two streams of research that, until
sions. For example, systems with strong analytical now, have been treated largely as distinct, and to
capabilities can be compared with those that have build on the unique strengths of each.
stronger information richness components. We may Orlikowski and Iacono (2001) criticize past TAM
find that features that make one system useful or easy
studies because they “lack the means to account for
to use are less relevant for another kind. Whether in-
temporal and contextual variations in the patterns
formation accuracy is important overall, only relevant
discerned. To do so will require more careful the-
in a data warehousing context, or less salient when
orizing about differences in IT artifacts and their
the users create information themselves leads to fur-
role and use in different contexts and over time”
ther research questions. The research model provides
(p. 129). We believe that this study helps to leverage
a platform for more detailed thinking, theorizing, and
the rich stream of research found within the technol-
testing of such ideas. Continued research along this
ogy acceptance literature by offering a theoretical way
vein should further confirm the absolute importance
of other components of the integrated model. to integrate it with a research stream, user satisfaction,
A challenge to researchers is not to identify and test which offers a way to theorize about and account for
alternative factors in an indiscriminate fashion, but to differences in the IT artifact.
develop a way to theorize about the characteristics of The important extension that the integrated model
a system and the information it produces to arrive at provides is the conceptual understanding of the dif-
ways to predict and understand them across contexts. ference between object-based beliefs and attitudes and
In other words, researchers need to develop theories behavioral beliefs and attitudes toward use, which
about information and system characteristics to over- primarily have been investigated separately in the
come the legitimate complaint that existing sets of user satisfaction and TAM literatures, respectively. We
system and information characteristics are arbitrary believe our model provides an important step on
(Galletta and Lederer 1989). the path to providing conceptual clarity to two crit-
Finally, other research designs would strengthen ically important streams of IT research. Nonetheless,
our understanding of the integrated model. This it is simply a step. Additional research is needed to
study examined a cross-section of individuals with a develop a refined understanding of the relationships
variety of usage patterns. Other studies may want to proposed in the integrated model.
look at more-controlled subsets of users and contexts
to identify constraints and exceptions to the model. It Acknowledgments
would also be beneficial to perform longitudinal stud- The authors are grateful to the McIntire School of Com-
ies that test the proposed relationships as they unfold merce Center for the Management of Information Technol-
over time. ogy, and the Information Systems Research Center at the
University of Houston for their support of this project. The
authors also appreciate the helpful comments from Henri
6. Conclusion Barki, Izak Benbasat, Mike Morris, and anonymous review-
This study integrated the theoretical perspectives ers. Wynne Chin’s assistance with technical issues related
and empirical findings of the user satisfaction and to the data analysis is gratefully acknowledged.
Wixom and Todd: A Theoretical Integration of User Satisfaction and Technology Acceptance
Information Systems Research 16(1), pp. 85–102, © 2005 INFORMS 101
Bailey, J. E., S. W. Pearson. 1983. Development of a tool for measur- Hong, W., J. Thong, W.-M. Wong, K.-Y. Tam. 2001/2002. Deter-
ing and analyzing computer user satisfaction. Management Sci. minants of user acceptance of digital libraries: An empirical
29(5) 530–545. examination of individual differences and systems characteris-
Barclay, D., C. Higgins, R. Thompson. 1995. The partial least tics. J. MIS 18(3) 97–124.
squares approach to causal modeling, personal computing Igbaria, M., T. Guimaraes, G. B. Davis. 1995. Testing the determi-
adoption and use as an illustration. Tech. Stud. 2(2) 285–309. nants of microcomputer usage via a structural equation model.
Baroudi, J., W. Orlikowski. 1988. A short-form measure of user J. MIS 11(4) 87–115.
information satisfaction: A psychometric evaluation and notes Ives, B., M. H. Olson, J. J. Baroudi. 1983. The measurement of user
on use. J. MIS 4(4) 44–59. information satisfaction. Comm. ACM 26(10) 785–793.
Benbasat, I., R. Zmud. 2003. The identity crisis within the IS disci- Karahanna, E., D. W. Straub, N. L. Chervany. 1999. Information
pline: Defining and communicating the discipline’s Coir prop-
technology adoption across time: A cross-sectional comparison
erties. MIS Quart. 27(2) 183–194.
of pre-adoption and post-adoption beliefs. MIS Quart. 23(2)
Chin, W. W. 1998. Issues and opinions on structural equation mod- 183–213.
eling. MIS Quart. 22(1) vii–xvi.
Kraus, 1995. Attitudes and the prediction of behavior: A meta-
Chin, W. W., T. A. Frye. 1996. PLS Graph. 2.91.03.04. Department analysis of the empirical literature. Personality Soc. Psych. Bull.
of Decision and Information Systems, University of Houston, 21 58–75.
Houston, TX.
Lim, K. H., I. Benbasat. 2000. The effect of multimedia on perceived
Davis, F. 1989. Perceived usefulness, perceived ease of use, and
equivocality and perceived usefulness of information systems.
user acceptance of information technology. MIS Quart. 13(3)
MIS Quart. 24(3) 449–471.
319–339.
Davis, F. 1993. User acceptance of information technology: Sys- Lohmoller, Jan-Bernd. 1989. Latent Variable Path Modeling with Par-
tems characteristics, user perceptions and behavioral impact. tial Least Squares. Physica-Verlag, Heidelberg, Germany.
Internat. J. Man-Machine Stud. 38 475–487. Mathieson, K., E. Peacock, W. W. Chin. 2001. Extending the tech-
Davis, F. D., R. P. Bagozzi, P. R. Warshaw. 1989. User acceptance of nology acceptance model: The influence of perceived user
computer technology: A comparison of two theoretical models. resources. Database Adv. Inform. Systems 32(3) 86–113.
Management Sci. 35(8) 982–1003. Melone, N. 1990. A theoretical assessment of the user-satisfaction
DeLone, W. H., E. R. McLean. 1992. Information systems success: construct in information systems research. Management Sci.
The quest for the dependent variable. Inform. Systems Res. 3(1) 36(1) 76–91.
60–95. Moore, G., I. Benbasat. 1991. Development of an instrument to mea-
Doll, W. J., G. Torkzadeh. 1988. The measure of end-user computing sure the perceptions of adopting and information technology
satisfaction. MIS Quart. 12(2) 259–274. innovation. Inform. Systems Res. 2(3) 192–222.
Eagly, A. H., S. Chaiken. 1993. The Psychology of Attitudes. Thomson Nunnally, J. C. 1978. Psychometric Theory. McGraw-Hill, New York.
Wadsworth, Belmont, CA.
Orlikowski, W. J., C. S. Iacono. 2001. Research commentary: Des-
Eckerson, W. W. 2002. Data quality and the bottom lone. The perately seeking the “IT” in IT research—A call to theorizing
Data Warehousing Institute Report Series. 101 Communications, the IT artifact. Inform. Systems Res. 2(12) 121–134.
LLC.
Plouffe, C. R., J. Hulland, M. Vandenbosch. 2001. Research
Fazio, R. H., M. A. Olson. 2003. Attitudes: Foundation, function and report: Richness versus parsimony in modeling technology
consequences. M. A. Hogg, J. Cooper, eds. The Sage Handbook adoption decisions: Understanding merchant adoption of a
of Social Psychology. Sage, London, UK. smart card–based payment system. Inform. Systems Res. 12(2)
Fishbein, M., I. Ajzen. 1975. Belief, Attitude, Intention, and Behavior: 208–222.
An Introduction to Theory and Research. Addison-Wesley, Read-
Seddon, P. 1997. A respecification and extension of the DeLone
ing, MA.
and McLean model of IS success. Inform. Systems Res. 8(3)
Fornell, C., D. F. Larcker. 1981. Evaluating structural equation 240–253.
models with unobservable variables and measurement error.
J. Marketing Res. 18 39–50. Szajna, B. 1996. Empirical evaluation of the revised technology
acceptance model. Management Sci. 42(1) 85–92.
Galletta, D. F., A. L. Lederer. 1989. Some cautions on the mea-
surement of user information satisfaction. Decision Sci. 20(3) Taylor, S., P. A. Todd. 1995. Understanding information technology
419–439. usage: A test of competing models. Inform. Systems Res. 6(2)
144–176.
Gefen, D., D. Straub. 1997. Gender differences in the perception
and use of e-mail: An extension to the technology acceptance Venkatesh, V., M. G. Morris. 2000. Why don’t men ever stop to
model. MIS Quart. 21(4) 389–400. ask for directions? Gender, social influence, and their role
Goodhue, D. L. 1988. IS attitudes: Toward theoretical and defini- in technology acceptance and usage behavior. MIS Quart. 24
tional clarity. Database Adv. Inform. Systems 19(3/4) 6–15. 115–139.
Haddock, G., M. P. Zanna. 1999. Cognition, affect, and the predic- Venkatesh, V., M. Morris, G. Davis, F. Davis. 2003. User acceptance
tion of social attitudes. W. Stroebe, M. Hewstone, eds. Eur. Rev. of information technology: Toward a unified view. MIS Quart.
Soc. Attitudes 10 75–99. 27(3) 425–478.
Hartwick, J., H. Barki. 1994. Explaining the role of user par- Wold, H., K. Joreskog. 1982. Systems Under Indirect Observation:
ticipation in information system use. Management Sci. 40(4) Causality, Structure, Prediction. North-Holland, Amsterdam, The
440–465. Netherlands.