An experimental investigation of web-based information systems success in the context of electronic commerce

Download An experimental investigation of web-based information systems success in the context of electronic commerce

Post on 05-Sep-2016

213 views

Category:

Documents

1 download

TRANSCRIPT

An experimental investigation of WAbstractelectronic marketplace is having a profound impact onthe world economy and the way business is con-ducted. In 1996, online sales were approximatelyglassbbc@muohio.edu (B. Glassberg)8 ykim@binghamton.edu(Y.J. Kim)8 mgtsand@mgt.buffalo.edu (G.L. Sanders)8Decision Support Systems 39 (shin@uri.edu (S.K. Shin).1 Tel.: +1 716 888 2267; fax: +1 716 888 2525.In this paper, we examined Web-based information systems (WIS) success and focused on User Satisfaction in the context ofa consumer purchasing decision. The results indicate strong support for the research model consisting of three fundamental UserSatisfaction components: Task Support Satisfaction (TSS), Decision Support Satisfaction (DSS), and Interface Satisfaction. Themodel explains approximately 50% of the variance in users intention to use Web-based information systems. It is concludedthat Decision Support Satisfaction plays an important role in Web-based information systems success. In light of these findings,implications for theory and practice are discussed.D 2004 Published by Elsevier B.V.Keywords: Web information systems; Electronic commerce success metrics; Information systems success; Technology acceptance model; Usersatisfaction; Partial least squares1. IntroductionThe emergence of the World Wide Web as an* Corresponding author. Tel.: +1 716 645 2373; fax: +1 716645 6117.E-mail addresses: garrity@canisius.edu (E.J. Garrity)8success in the context of electronic commerceEdward J. Garritya,1, Bonnie Glassbergb,2, Yong Jin Kimc,3, G. Lawrence Sandersd,*,Seung Kyoon Shine,4aInformation Systems Department, Canisius College, Buffalo, New York 14208, United StatesbDepartment of Decision Sciences and MIS, Miami University, Oxford, Ohio 45056, United StatescSchool of Management, Room 240, State University of New York at Binghamton, Binghamton, New York 13902, United StatesdDepartment of Management Science and Systems, 310A Jacobs Management Center, State University of New York at Buffalo, Buffalo,New York 14260, United StateseCollege of Business Administration, University of Rhode Island, 210 Flagg Road, Kingston, RI 02881-0802, United StatesReceived 1 July 2003; accepted 1 June 2004Available onlnie 2 October 20040167-9236/$ - see front matter D 2004 Published by Elsevier B.V.doi:10.1016/j.dss.2004.06.0152 Tel.: +1 53 Tel.: +1 607 777 6638; fax: +1 607 777 4422.4 Tel.: +1 401 874 5543; fax: +1 401 874 4312.eb-based information systems2005) 485503www.elsevier.com/locate/dswexpanding rapidlyUS$500 million and have been31 529 4826; fax: +1 531 529 9689.since [64]. The total electronic commerce spendingestimates for 2004 range from US$963 billion toorganizational setting, investigating the causal pathfrom User Satisfaction to individual impacts isuppoUS$4 trillion [20]. Many firms positioned anddeployed systems to compete in the electronicmarketplace. The systems, in general, are referred toas Web-based information systems (WIS) [76]. Web-based information systems [76] represent a newfrontier for businesses trying to establish an on-linepresence where consumers are free to shop in moreefficient bfriction free markets.QIncorporating the radical changes in global market-places created by this new application of informationtechnology, Web-based systems are considerablydifferent from the traditional systems in terms of theirscope and focus [91]. WIS are usually built tofacilitate consumer oriented tasks and focus onenhancing the consumer decision making process.As the Web environment is characterized as non-linearin nature, presentation and delivery of information andfrictionless work support are very critical for attract-ing consumers and enhancing the shopping experi-ence. Lederer et al.s [46] study found that the TAMmodel did not adequately explain Web use, and thisimplies that traditional information systems havedifferent characteristics than WIS.Given the importance of electronic commerce andWIS, in terms of the magnitude of its impact onbusiness process and organization structure, there islittle research addressing fundamental issues such ashow electronic commerce systems success can bemeasured, whether existing measures of success canbe applied, and how consumers react to the onlineshopping experience. Without a clear understandingof the dynamics of Web-based system success toguide firms, proper strategies and system designs aremere speculation. A central activity for researcherswill be to define and operationalize the constructs forunderstanding the success of electronic commercesystems. This paper, while drawing on the traditionalinformation systems success literature, extends thatliterature to the development of a Web-based modelfor electronic commerce success in the context of aconsumer purchase decision.The contributions of the current study are three-fold. First, this research addresses the issue of how topredict users intent to use a particular Web siteutilizing traditional information systems successtheory. The results should be of great interest toE.J. Garrity et al. / Decision S486Web designers, IS staff and researchers. Second, byexplaining the dynamic relationship between Userfruitless [68]. Finally and most importantly, the modeldoes not explain clearly and fully the relationshipbetween User Satisfaction and individual/organiza-tional impacts.2.2. Technology Acceptance ModelSatisfaction components which influence WIS suc-cess, the current research can aid researchers in furtherrefinement of information systems success models ingeneral. Third, the current study provides a reasonablebackground for applying existing measures of infor-mation systems success to the electronic commerceenvironment.2. Review of information systems success literatureThere is a rich tradition of information systemssuccess research including user information satisfac-tion, tasktechnology fit, user involvement, andparticipation. In this paper, three important, compre-hensive studies are further examined in terms ofproviding theoretical background. They are theDeLone and McLean Model of IS Success [17], theTechnology Acceptance Model (TAM) [1315], andthe Garrity and Sanders Model of IS Success [25].2.1. DeLone and McLean Model of IS SuccessDeLone and McLean [17] suggested a model of ISsuccess that is comprised of six multi-level constructs:Information Quality, System Quality, User Satisfac-tion, System Use, Individual Impact, and Organiza-tional Impact. While the model integrates thecomprehensive dependent variables used by ISresearchers, there exist several criticisms (althoughthey do not reduce the importance of the model). First,IS Use in the DeLone and McLean model contains toomany meanings to be appropriately examined [68]. ISUse is also argued to play a problematic andcontroversial role in modeling system success (cf.Refs. [15,49,68,69]). Second, because User Satisfac-tion represents individual impacts of IS in anrt Systems 39 (2005) 485503A second stream of IS success research is focusedon the use of information technology as conceptual-ers mupport Systems 39 (2005) 485503 487ized by the Technology Acceptance Model [1315].This model of IS success asserts that Ease of Useand Perceived Usefulness are primary determinantsof System Use. The model specifically postulatesthat technology Usage is determined by BehavioralIntention to Use the technology, which is itselfdetermined by both Perceived Usefulness andAttitude.TAM has been widely studied in a number ofcontexts using different information systems. Morerecently, it has been applied to examining the use ofWeb-based information systems [27,46,54,77]. How-Fig. 1. Garrity and SandE.J. Garrity et al. / Decision Sever, this model does not capture the complexpsychological and social influences on technologyadoption [50] and long-term use [9,50]. In particular,when applied to the context of Web informationsystems, the model appears to account for only a smallportion of variance for Usage [46]. Lederer et al. [46]called for researchers to develop better instruments formeasuring antecedents of Perceived Ease of Use andPerceived Usefulness for assessing Web sites. Inaddition, they urged a reexamination of factors orlatent variables related to Web site Usage via studyreplication, and additional use of factor analysis touncover important subdimensions of the TechnologyAcceptance Model.2.3. Garrity and Sanders Model of IS SuccessGarrity and Sanders [25] extended the DeLone andMcLean model and proposed an alternative model inthe context of organizational and socio-technicalsystems (Fig. 1). The model identifies four subdi-mensions of User Satisfaction: Interface Satisfaction,Decision Support Satisfaction (DSS), Task SupportSatisfaction (TSS), and Quality of Work Life Sat-isfaction. These factors were derived from an exten-sive review of IS success research and reasoning frombasic principles of systems and general systemstheory. The four factors correspond with three view-points of information systems: the organizationalviewpoint (that views IS as a component of the largerorganization system), the humanmachine viewpointodel of IS success [25].(which focuses on the computer interface and the useras components of a work system), and the socio-technical viewpoint (that considers humans as alsohaving goals that are separate from the organizationand whereby the IT or technical artifact impacts thehuman component in this realm).5Both Task Support Satisfaction and DecisionSupport Satisfaction attempt to assess the effective-ness of an IS within the context of the organizational5 The fourth dimension of the Garrity and Sanders model isquality of work life satisfaction. This dimension addresses the fitbetween an IS and the socio-technical work world of the respectiveusers. Quality of work life satisfaction is an important dimension ofsuccess in organizational settings. However, in the context of aconsumer-oriented web information system, the users will not beimpacted in terms of work-life or job satisfaction, and thus, thisdimension will not be included in this study.viewpoint of systems. The Task Support Satisfactiondimension captures the overall set of tasks associatedwith job activities while Decision Support Satisfactionis more focused on decision support (i.e., structuring,analyzing, and implementing a decision).Interface Satisfaction assesses IS success from thehumanmachine viewpoint. Interface Satisfactionmeasures the quality of the interface in terms ofpresentation, format, and processing efficiency. Thequality of the interface is related to both Task Supportinformation systems success [17,26,35,36,52]. Fig. 2shows the proposed research model. The next sections3.1. Theoretical background of the research modelIn this research study, we are concerned withassessing User Satisfaction with Web-based informa-tion systems in the context of consumer purchasingactivities. The nature of a consumer purchasing Web-based system is closely tied to providing decisionsupport capabilities to aid consumers in product andservice purchase decisions [57].A consumer-oriented WIS can be decomposedE.J. Garrity et al. / Decision Support Systems 39 (2005) 485503488elaborate a theoretical background of the proposedresearch model and detailed hypotheses developed toexamine the relationships among the constructs ofinterest.and Decision Support Satisfaction. An improperlydesigned interface can cause users difficulty with taskcompletion or it can impair their ability to makedecisions.3. Research model and hypothesisThe model used in this study draws on the DeLoneand McLean Model [17], TAM, and the Garrity andSanders Model of IS Success [25] to examine Web-based information systems success. As outlined byLederer et al. [46], we seek to discover importantunderlying factors that will help to explain Web-basedinformation systems use and adoption. In essence, theresearch model, Web-based Information SystemsSuccess Model (hereafter, referred to as WISSM),deals with the reciprocal relationship between UserSatisfaction and systems use. This relationship hasbeen touted as one of the key constructs representingFig. 2. Web-based Information Systeinto numerous tasks, subtasks and decisions. Forexample, a product purchasing WIS may involve anumber of cognitive tasks, such as product search-ing, advice gathering, decisions on informationpresentation, rank ordering of alternatives, criteriadiscovery, learning about the domain of product use,deciding on the relevance of criteria, determiningtradeoffs among criteria, prices, and products, actualpurchase tasks such as badding to shopping carts,Qnavigating from one task context to the next,checkout, payment, confirmation, and shippingmethods. Many of these cognitive tasks can becategorized as intellectual or decision support based,while others may be more mundane tasks such asfilling out self-report information.In sum, Decision Support Satisfaction assesses thesupport provided by the WIS in the context ofdecision making. Task Support Satisfaction, on theother hand, assesses the overall support for thepurchasing task. Since this is a general assessmentof the overall task, it takes into account the overalldesign of the software and its usefulness. In additionto these variables, Trust in a Web site is regarded as animportant determinant of EC success [27,82]. Gefenand Straub [27] included a Trust item as an indicatorvariable for the behavioral intention to purchase.ms Success Model (WISSM).However, the current study does not directly incor-porate Trust as an indicator variable reflectingE.J. Garrity et al. / Decision SupporBehavioral Intention to Use an EC web site.The dependent variable, Behavioral Intention toUse an EC Web site, was taken from the TechnologyAcceptance Model [1315] and Seddons [68] respe-cification of the DeLone and McLean Model [17].Their work provides the theoretical basis for therelationship between User Satisfaction and the Behav-ioral Intention to Use an information system.System Use was not selected as the dependentvariable because of the well-documented problems inmeasuring system success (cf. Seddon [68]).6 Thispaper posits that Behavioral Intention to Use a systemrepresents an individuals perception of the benefitfrom System Use [68]. In essence, BehavioralIntention to Use a system effectively mediates theeffect of User Satisfaction on actual systems use. Thepositive relationship between User Satisfaction andBehavioral Intention has been frequently reported inthe marketing literature (e.g., Refs. [2,53,83]). Theinnovation literature also reports that dissatisfactionwith an innovation leads to the termination of theinnovation [60].It should be noted that the constructs used in thismodel are more context based than the System Qualityand Information Quality measures from the DeLoneand McLean model. Information Quality and SystemQuality were developed from a product focus asopposed to a system, tool, or goal directed focus. Ourapproach, using a more goal-oriented perspective,asserts that a system of high quality should supportusers in performing their task related responsibilities.This is the approach taken by Davis [1315] in thedevelopment of the Perceived Usefulness dimensionof IS Success. Recall that Perceived Usefulness isdefined as the users subjective assessment that usinga system will increase his or her job performance. Thedimensions of Task Support and Decision SupportSatisfaction are simply more specific measures of thetype of job support provided by the system.Finally, the Interface Satisfaction dimension maybe thought of in several ways. It provides an assess-ment of the design of the hardware and software6 According to Seddon [68], information systems use in Deloneand McLeans model has three distinct meanings: a proxy for thebenefits from use, future IS Use, and an event in a process model.interface and is closely related to Daviss Ease of Usedimension. If systems are tools to help workersaccomplish tasks, then a well-designed tool shouldbe made in such a way that it is easy and efficient touse. In addition, the interface is also the focal point ofinteraction between humans and information systems.Thus, Information Quality can also be explained byInterface Satisfaction, in the sense that the vehicle forpresenting the information (e.g., a textbox, graph,listbox, or form) cannot be separated from theinformation itself [56]. A bchunkQ of information isbeneficial to a user if it is presented in an effective andefficient manner, helping users accomplish taskobjectives. bProviding high-quality data along thedimensions of value and usefulness relative to theconsumers task contexts places a premium ondesigning flexible systems with data that can be easilyaggregated and manipulatedQ [73]. Thus, we believe aseparate assessment of the interface is especiallycritical for WIS because users must successfullynavigate the Web site to obtain informational supportfor tasks and decisions.Thus, the focus of the Web-based InformationSystems Success Model presented in Fig. 2 is on: (1)How well the system supports workers in theaccomplishment of their goals (an effectivenessmeasure via Task Support and Decision SupportSatisfaction), and (2) How well designed and imple-mented the interface is so that workers may accom-plish their goals more easily (an efficiency measurevia Interface Satisfaction).3.2. Research question and hypothesesFig. 2 illustrates hypothesized relationships betweenconstructs in the study. The major determinant ofBehavioral Intention to Use an electronic commerceWeb site is hypothesized to be Task Support Satisfac-tion which is influenced by both Decision SupportSatisfaction and Interface Satisfaction. Note that thereis no direct relationship between Interface Satisfactionand Behavioral Intention to Use an EC Web site. Thereasons are twofold. First, as discussed earlier, thecurrent study assumes that Task Support Satisfaction isa focal construct that mediates the effects of Interfaceand Decision Support Satisfaction on Behavioralt Systems 39 (2005) 485503 489Intention. The second reason relates to the way thatInterface Satisfaction affects technology acceptanceuppo[43]. Over time, the impact of Ease of Use onBehavioral Intention fades away [86].3.2.1. Interface SatisfactionLohse and Spiller [47] showed that designing onlinestores with user-friendly interfaces is critical in trafficand sales, while Szymanski and Hise [75] found sitedesign important in creating a satisfying customerexperience. The Interface Satisfaction dimension cap-tures the users overall impression of the interface interms of presentation, navigability, format, ease of useand efficiency of interaction with the Web site.Shneiderman [70] argues that an appropriate designof user interfaces includes presentation, and interactiv-ity. Zhang and von Dran [90] report that ease ofnavigation was the most important Web site qualityacross six electronic commerce domains. Poorlydesigned screens and interfaces make task supportand decision support activities more difficult. Inparticular, high interface quality in hypermedia systemslike the Web is critical in reducing cognitive overhead[11,48] and disorientation [11,78,81].According to the theory of symbolic representationin problem solving [32,40,41], there are three types ofsymbolic representation: linguistic representation,visual imagery representation, and exploratory rea-soning. These symbolic representations can bereferred to as representational information quality[73], which is, in turn, interpreted as interface quality.In most problem solving situations, various combina-tions of these three representations are needed toprovide dominant or controlling focus in problemsolving. Two critical stages in the problem solvingprocess are information search and the evaluation ofalternatives [19,34,55], and they are fundamental fordecision making and task performance.The importance of the user interface has a richresearch tradition. Assimilation Theory [3] posits thatmeaningful learning of computer systems begins withthe humancomputer interface. Davis and Bostrom[16] and Mayer [51] found that the interface can havea positive impact on task performance and decisionmaking within the context of Assimilation Theory.High levels of usability are related to improved taskperformance [80]. Ausubel [3] emphasized the impor-tance of external aids in the form of an badvanceE.J. Garrity et al. / Decision S490organizerQ that facilitates the development of newknowledge structures. The user interface as anbadvance organizerQ should enhance task performanceand decision making performance. According toNielsen [56], user-centered navigability significantlyenhances users performance with various Web tasks.Poor interface design has been cited as a key elementin a number of high profile Web site failures [58]. It ishypothesized, therefore, that Interface Satisfactionpositively influences both dimensions of satisfaction(TSS, DSS) with the system.H1a. Interface Satisfaction has a positive impact onDecision Support Satisfaction.H1b. Interface Satisfaction has a positive impact onTask Support Satisfaction.3.2.2. Decision Support SatisfactionDecision Support Satisfaction is defined in terms ofthe capability of an information system to assist indecision-making and better performance of the usersjobs [65], even when those decision-making activitiesare ill-structured, situation-specific, or involve choos-ing from a number of alternatives [44].Humans tend to adapt their decision makingstrategies to specific situations and environments[61]. Individuals can be described as bcognitivemisersQ who strive to reduce the amount of cognitiveeffort associated with decision making [71]. Thenotion that people are typically willing to settle forimperfect accuracy in their decisions, in return for areduction in effort, is well supported [8,37], and isconsistent with the idea of bounded rationality andthe concept of satisficing [72]. Due to this tradeoffbetween effort and accuracy, decision makers oftensatisfice when alternatives are numerous and/ordifficult to compare, or when the complexity of thedecision environment is high [62]. In the context ofWeb-based consumer purchasing decision systems, itis imperative not only to assess the PerceivedUsefulness of these systems (the purchasing taskitself) but also to include an assessment of the degreeto which the system supports or reduces the decisionmaking efforts of the user.OKeefe and McEachern [57] delineate five criticaltask areas that need to be modeled in a Web-basedconsumer decision support system: need recognition,rt Systems 39 (2005) 485503information search, evaluation, purchase, and after-purchase evaluation. Decision support features thatThe two consequences of dissatisfaction negativelyaffect repeat purchasing behavior. The positiveupporcan be provided by a Web-based consumer decisionsupport system include virtual catalogs, internalsearch capabilities, structured interactions, and eval-uative decision models. While a number of thesefeatures fall outside the realm of simply supportingthe decision process, they are vital to supporting theentire task process (i.e., a purchase task). Examples ofactions to support the task include placing an order,arranging a payment, or providing detailed informa-tion on the purchase task. Additionally, this wouldencompass after sale support and feedback, andresponse about the process itself.Unlike traditional organizational decision supportsystems, Web-based consumer purchasing systemsmust be designed to battractQ consumers to use thesystem [57]. The non-linear structure of the Webenvironment entails more cognitive effort than thelinear, sequential structure of the traditional systemenvironment and without sufficient support for deci-sion making and task performance, consumers tend tosimply click-away to competitors.Hence, decision support features or actions in WISshould not only support the decision process but alsosupport the entire task process [57]. This reasoningleads to the hypothesis of a positive relationshipbetween Decision Support Satisfaction and TaskSupport Satisfaction as in H2.H2. Decision Support Satisfaction is positivelyassociated with Task Support Satisfaction.3.2.3. Task Support SatisfactionAntecedents for e-commerce success examined byprevious studies highlight the importance of support-ing online consumers at each step of their activitieswith an e-commerce system. The most commononline tasks include product search, comparisonshopping, negotiation of terms, order placement,payment authorization, and receipt of product [38].The essential question surrounding the success of theonline experience relates to the net value of thebenefits and costs of finding, ordering, and receiving aproduct [42]. User satisfaction with such task supportcan be referred to as the net benefits of a system [68].The main behaviors of satisfied or dissatisfiedcustomers with product or service offerings includeE.J. Garrity et al. / Decision Scomplaining behavior, negative word of mouth, andrepeat purchasing [74]. Complaining behavior isrelationship between the level of satisfaction and therepeat purchasing behavior has been reported to bevery strong (correlation 0.882) [74]. In the informa-tion systems context, the repeat purchasing behavioris interpreted as the repeated Use of an informationsystem, which is primarily determined by BehavioralIntention to Use the information system [14,15,68,85].Hence, the relationship between Task SupportSatisfaction, as a global evaluation of systemssupport, and Behavioral Intention to Use an electroniccommerce Web site as a precursor for future Use of asystem, can be hypothesized as H3.H3. Task Support Satisfaction positively influencesBehavioral Intention to Use an e-commerce Web site.Hypotheses one through three relate to the specificpaths of WISSM. Task Support Satisfaction ishypothesized to be the primary determining factor ofBehavioral Intent to Use an EC Web site. TaskSupport Satisfaction is partially derived from Per-ceived Usefulness [25]. A separate but closely relateddimension, Decision Support Satisfaction, is hypothe-sized to positively influence Task Support Satisfaction(see H2). Decision support capabilities are importantand vital task considerations, and this is particularlytrue for systems used in electronic markets [6,7,57].4. Research methodThe research model was tested in the context of anInternet buying simulation. The subjects of this studywere given a Web-based purchasing task. The task wasto buy a digital camera under given limitation in termsof budget, time, and retail Web sites. The budget limitgiven to the subjects was US$600, which is the medianexplained as consumers tendency to complain tosuppliers and this works as a mechanism for dissat-isfied consumers to relieve cognitive dissonance,while negative word of mouth behavior to otherconsumers stands for another type of complainingbehavior of dissatisfied consumers. Repeat purchasingbehavior represents the phenomenon that satisfiedconsumers will likely buy the offerings repeatedly.t Systems 39 (2005) 485503 491price of digital cameras. This budget limit was requiredso that subjects would have a greater cognitive burdenon the decision task. It was expected that few of therespondents would have already purchased the targetproduct, while most would be familiar with the generalfeatures of a camera.E.J. Garrity et al. / Decision Suppo4924.1. Sampling methodEach subject was given two identical sets of surveyinstruments to be used in assessing the quality of twoWeb sites (CameraWorld.com and Advandig.com).Each set contained the four constructs (DecisionSupport Satisfaction, Task Support Satisfaction, Inter-face Satisfaction, and Behavioral Intent to Use) understudy.4.1.1. Data collection procedureThe data collection activity for this study consistedof three main phases. First, an expert survey, foridentifying two diverse Web sites, was conducted.Second, a pilot study, involving 97 subjects and 194observations, was performed. It was used to refine theitems and constructs used in the study. It also enabledthe researchers to clarify the wording, content, andgeneral layout of the survey instrument. Finally, amain survey was administered, involving 163 subjectsand 326 observations.In the first phase of the data collection, ten expertswho had practical and academic experiences withelectronic commerce evaluated six Web sites.7 Theexperts were asked to rate the Web sites in terms ofDecision Support, Task Support, Interface, and Over-all Satisfaction level, by responding to a 7-pointLikert scale questionnaire. The scores for DecisionSupport, Task Support, and Interface Satisfaction weresummed. Both the sum of the scores and OverallSatisfaction level consistently indicate that all tenexperts nominated CameraWorld.com as the superiorcommercial Web site. Nine of the experts votedAdvandig.com as the least desirable (inferior) of thesix Web sites.The second phase of the data collection consistedof a pilot study. The objective of this phase was to7 We selected six most popular commercial web sites special-izing digital cameras and accessories, which commonly appeared inthe lists of three Internet portal sites. The six sites include Advandig,Digital Camera, Digital Solution, Camera World, Digital CameraClub, and Digital Camera Store.focus on measurement issues and to refine theresearch instruments prior to the final data collection.Pilot survey questionnaires were distributed to 97junior and senior undergraduate business majors.The main study consisted of 163 students that werenot involved in the pilot study. Two hundred fiftysurveys were given out and 163 were returned(response rate: 65.2%). As noted earlier, the surveyinstrument consisted of two identical questionnairesthat were used by subjects to evaluate the Web sitesaccording to the research constructs. The order of theappearance of the Web site addresses on the ques-tionnaire was randomized.4.1.2. Sample frameIdentifying the sample frame for empirical researchis an important step in ensuring that the population ofinterest has been correctly identified. The idealcandidate for this study was computer literate, hadprevious exposure to Web technology and had currentaccess to a Web-enabled computer. The Web environ-ment is highly graphical, requiring knowledge of theuse of icons, frames, and scrolling. There may bemultiple browser windows open at the same time, sogeneral navigation and mouse skills were needed. Asthe purpose of this study was to predict purchasingbehavior by individuals, in-depth knowledge of Webtechnology was neither assumed nor expected in thetarget population.How much exposure to the technology is suffi-cient? Empirical studies based on specific technolo-gies, where the individual is the unit of analysis, haveemployed a 1 hour hands-on task [14,84]. Otherstudies used exposure without a required hands-ontask [39,50] and some asked subjects for purespeculation about a technology which did not exist[12]. Required courses at the institution used toconduct the study typically use the Web in a varietyof ways including distributing course syllabi, presen-tation materials and handouts, online cases, and avariety of Web-based assignments. Subjects wouldthus have satisfied the exposure constraint. Inaddition, respondents were asked to rate themselveson their computer skills and Web experience.The final requirement for the population of interestwas current availability of a computer with Webrt Systems 39 (2005) 485503access. All subjects had access to the Internet viaschool or work and many had their home computerthe average age was 21.5 years. Respondents wereasked to rate the extent of their experience with theWeband with computers in general on a scale of 1 to 7, with1 being a novice and 7 being an expert. This group ofsubjects considered themselves to be relatively expe-rienced users of the Web (4.90) and slightly aboveaverage on their use of computers (5.01). Most reportedthey logged onto the Web about 20 times a week,spending just under 16 h on the Internet (weekly). Amajority of the subjects have one to ten buyingexperiences over the Internet, while 24% of partic-ipants have never bought over the Internet. Addition-ally, subjects reported they spend about 24 h per weekon the computer itself. The majority of respondents(78%) had completed 3 years of college.4.2. Partial Least Squares (PLS)Assessment of the research model was conductedusing PLS. PLS is a structured equation modelingtechnique that can analyze structural equation models(SEMs) involving multiple-item constructs, withupporlinked to the Web through an Internet ServiceProvider.4.1.3. Subject poolHuman subjects identified for the pilot test and finaldata collection were a sample of upper level under-graduate students enrolled in management informationsystems courses at a state research institution in thenortheast. All subjects voluntarily participated in thesurvey, and were advised that they could withdrawfrom participation at any time without adverse con-sequence. The survey was conducted in an unsuper-vised environment. All the participants were givenextra credit for participating in the study. The final datacollection utilized all undergraduates enrolled inintroductory MIS classes taught by an instructor. Thiswas a required course for all business majors.Although the reliance on college students has beencriticized for many studies in applied research, there isjustification for using them under specific conditions[31]. In a meta-analysis comparing studies using bothundergraduate students and non-students, Gordon etal. [31] noted that between-groups differences are lesspronounced when no prior knowledge or familiaritywith the task is needed to make judgments, or whenthe objective is to perform simple reasoning tasks. Amajority of undergraduates taking courses are exposedto Web technology as part of the regular curriculum oftheir coursework. Any complexity or skill in perform-ing experimental tasks is reduced by the graphicalnature of the technology. Thus, only general knowl-edge of computers and English were required.The use of undergraduate students made sense inthis study because a primary focus of this research is onunderstanding Web-based decision behavior and col-lege students are the present and future consumers ofWeb technology. It should be pointed out that majorusers of the Web are individuals who have a collegedegree and are 1644 years old [33], which is similar tothe subject pool used in this study (age range 1850years old).4.1.4. DemographicsOne hundred sixty-three upper-level undergraduatesparticipated in the final survey. The demographics forthis sample are shown in Table 1. Approximately 60%E.J. Garrity et al. / Decision Sof the subjects were male and 40% were female. Theage of respondents ranged from 18 to 50 years old, andTable 1DemographicsAge Average:21.5 yearsRange:1850 yearsGender Male 60.5%Female 39.5%Education level: yearscompleted incollege2 30.6%3 54.4%4 or more 15.0%Experience with webtechnologyAverage self-rating(1=Novice, 7=Expert)4.90Experience withcomputersAverage self-rating(1=Novice, 7=Expert)5.01Online buyingexperienceNever 24%15 50%610 15%More than 10 times 11%Current web usage Average accesses perweek20Average hours perweek16Years using web Less than 1 year 6%Between 1 and1.5 years31%More than 2 years 13%More than 3 years 78%t Systems 39 (2005) 485503 493direct and indirect paths. PLS works by extractingsuccessive linear combinations of the predictors and isTable 2Operationalization of latent variablesDecision Support SatisfactionDBET Use of the web siteenables me to makebetter purchasingdecisions[65]DEFF This web site assists mein making a decisionmore effectivelyAdapted fromRef. [14]DPRI Use of the web siteenables me to set mypriorities in making thepurchase decision[65]Task Support SatisfactionTUSE This web site is moreuseful than I hadexpected[24]TEXT This web site isextremely useful[65]TQIK Using this web siteenables me toaccomplish tasks morequickly[14]TESY This web site makes iteasier to do my task[14]Interface SatisfactionICLR The informationprovided by this website is clear andunderstandable[18,2830]ILEN Learning to use this website was easy for me[14]IFRN This web site isuser-friendly[14,18,2830]IESY This web site is easy touse[14,18,2830]IWNT I found it easy to get thisweb site to do what Iwant it to do[14]IINT My interaction with thisweb site was clear andunderstandable[14]ISKL It would be easy for meto become skillful at usingthis web site[14]INAV This web site is easy tonavigate[46]Behavioral Intent to UseUINF I predict I will use thisweb site to get theinformation I needAdapted fromRef. [86]E.J. Garrity et al. / Decision Suppo494effective in explaining both response and predictorvariation [10]. PLS is a powerful approach foranalyzing models because of the minimal demandson measurement scales, sample size, and residualdistributions [89]. In addition, the component-basedPLS avoids two serious problems, inadmissiblesolutions and factor indeterminancy [22].SEM approaches, such as LISREL and AMOS, arenot able to deal with non-normal distributions [21],and they can yield non-unique or otherwise impropersolutions in some cases [22]. PLS is not as susceptibleto these limitations [88,89]. The emphasis of PLS ison predicting the responses as well as in under-standing the underlying relationship between theBehavioral Intent to UseUCRT The use of this web sitewould be critical in myfuture decisionsAdapted fromRef. [24]UCRE I would use my creditcard to purchase fromthis web site[27]USIM I would use this web siteagain for a similar taskAdapted fromRef. [1]UINC I intend to increase myuse of this web site inthe future[1]UFUT In the future, I plan touse this web site oftenAdapted fromRef. [86]All questionnaire items use a seven-point Likert scale, varying fromStrongly disagree to Strongly agree.Table 2 (continued)rt Systems 39 (2005) 485503variables [79]. For example, PLS is very useful inscreening out factors that have a negligible effect onthe dependent variables.PLS has one additional property that is useful instudying the nature of the linkages between constructsand their measures. The nature of the links is referredto as epistemic relationships, or drules of corre-spondenceT [4,21]. Two basic types of epistemicrelationships are relevant to causal modeling: reflec-tive indicators and formative indicators. In this study,we use reflective indicators for measuring theunobserved, underlying constructs.4.3. Operationalization of research variablesAs shown in Table 2, the variables of the studywere operationalized using a variety of sources. All ofthe measures were taken directly or adapted fromprevious studies on IS success such as Doll andTorkzadeh [18], Franz and Robey [24], and Sanders[65] and on technology acceptance such as Davis [14]and Lederer et al. [46].To keep the length of the instrument reasonable,four and five items were selected from the TaskSupport Satisfaction and the Decision Support Sat-isfaction categories, respectively, and eight items wereselected for Interface Satisfaction. For BehavioralIntention to Use an EC Web site, the study draws onthree previous studies including Agarwal and PrasadSatisfaction (This Web site assists me in performingmy task better).5. Analysis and resultsPLS analysis involves two stages: (1) the assess-ment of the measurement model, including thereliability and discriminant validity of the measures,and (2) the assessment of the structural model. For theassessment of the measurement model, individualIIIIIFRN 0.9014 0.1585 UCRE 0.7757 0.1879IIIE.J. Garrity et al. / Decision Support Systems 39 (2005) 485503 495[1], Gefen and Straub [27], and Venkatesh and Morris[86]. The study follows the criteria recommended byAgarwal and Prasad [1] for choosing survey items.They recommend removing items that are not relevantto the specific innovation examined in the study, andalso deleting items that are very similar to other items.By using these criteria, the items selected ensurecomplete coverage of the constructs at hand. Behav-ioral Intention to Use the EC Web site was measuredusing six items.Among them, five items involve the likelihood offuture use, while one item bUsing my credit card topurchase from this Web siteQ is used to assess levelof Trust. Trust is considered a critical aspect ofelectronic commerce success [67], and is animportant determinant of Behavioral Intention toUse the Web site [82]. This item was also used inGefen and Straub [27] to measure Intention toPurchase.Two items were subsequently excluded from thefinal analysis because of their high correlationcoefficient: one item from Decision Support Satisfac-tion (This Web site improves the quality of mypurchase decision making) and one from Task SupportTable 3Measures, loadings, and weightsDecision Support Satisfaction Task Support SatisfactionItem Loadings Weight Item Loadings WeightDBET 0.9228 0.3579 TUSE 0.8471 0.2844DEFF 0.9202 0.3862 TEXT 0.9079 0.2931DPRI 0.8952 0.3511 TQIK 0.9104 0.2743TESY 0.8848 0.2749IIESY 0.9271 0.1503 USIM 0.8831 0.2196WNT 0.8777 0.1582 UINC 0.8784 0.1761INT 0.8782 0.1439 UFUT 0.8475 0.1850item loadings and internal consistency were examinedas a test of reliability. Individual item loadings andinternal consistencies greater than 0.7 are consideredadequate [23].As shown in Table 3, loadings for all measurementitems are above 0.7, which indicates there is soundinternal reliability and all the weights are statisticallysignificant at pb0.001. The almost uniformly distrib-uted weights show each item contributes to eachconstruct equivalently. We used PLS-Graph Version2.91.03.04 to perform the analysis.5.1. Reliability and validity testsIn assessing the internal consistency for a givenblock of indicators, the composite reliability (CR),also referred to as convergent validity (see Werts etal. [87]), was calculated. All the CR values are over0.9, which suggests that the parameter estimates aresound (Table 4).The average variance extracted (AVE) was alsocalculated. AVE measures the amount of variance thata construct captures from its indicators relative to thevariance contained in measurement error. This statisticnterface Satisfaction BI to use EC Web sitetem Loadings Weight Item Loadings WeightCLR 0.8623 0.1625 UINF 0.8575 0.2112LEN 0.8695 0.1256 UCRT 0.8724 0.1923SKL 0.8368 0.1244NAV 0.7939 0.1248compared the proposed four-factor model to allpossible three-factor models (there were four suchcomparisons). In all cases the change in the v2comparing the proposed four-factor model to each ofthe three-factor models was significant at pb0.001. Thebest three-factor model involved treating DSS and TSSas one construct, which was accomplished by settingthe correlation between the two constructs to one. Thev2 difference between the four-factor model and thisthree-factor model was v2=105 (df=1, pb0.001).The analyses with both PLS and AMOS indicatethat there exists reasonable discriminant validityamong all of the constructs. It is important to notethat this study is in fact examining an informationsystems success model consisting of dependentTable 4Composite reliability (CR) and average variance extracted (AVE)Constructs CR AVE FormulaE.J. Garrity et al. / Decision Support Systems 39 (2005) 485503496can be interpreted as a measure of reliability for theconstruct and as a means of evaluating discriminantvalidity [23]. AVE values should be greater than 0.50.All AVEs for the constructs used in this study aregreater than .70. This indicates that more than 70% ofthe variance of the indicators can be accounted for bythe latent variables.The AVE can also be used to assess discriminantvalidity. The AVEs should be greater than the squareof the correlations among the constructs. That is, theamount of variance shared between a latent variableand its block of indicators should be greater thanshared variance between the latent variables. In thisstudy, the square roots of each AVE value is greaterthan the off-diagonal elements (Table 5, refer toAppendix A which contains the cross loadingsbetween the measurement items and the constructs).This indicates that there exists reasonable discriminantvalidity among all of the constructs.8The correlation between Decision Support Satis-faction and Task Support Satisfaction appears to be alittle high although valid in terms of discriminantvalidity criteria. Such a close but distinctive relation-Decision SupportSatisfaction0.9374 0.8332Task SupportSatisfaction0.9371 0.7884 CR=(Pk i)2/[(Pk i)2+Pivar(e i)]Interface Satisfaction 0.9610 0.7554 AVE=Pk i2/[Pk i2+Pivar(e i)]Behavioral Intentionto Use EC web site0.9413 0.7280k i is the component loading to an indicator and var(e i)=1ki2.ship between Decision Support Satisfaction and TaskSupport Satisfaction was expected (see Garrity andSanders [25]). Hence, Decision Support Satisfactiontogether with Task Support Satisfaction can be used toprovide more insight into the features of the Web site.To further evaluate/establish the discriminant val-idity of the constructs, we conducted discriminantanalysis tests [5] using AMOS. These analyses8 In personal correspondence with Wynne Chin, he indicatedthat the difference between the off-diagonal (correlation) and thediagonal (square root of AVE) should be greater than 0.1 in order tosecure reasonable discriminant validity.variable constructs (see Refs. [17,25] for a discussionof dependent variable models). Accordingly, thecorrelations between the research variables areexpected to be high. However, they should not be sohigh as to preclude them from being consideredseparate constructs. As noted above, the constructs inthe model do indeed satisfy discriminate validitycriteria detailed by Fornell and Larcker [29] andBagozzi and Phillips [5].5.2. Assessment of the structural modelThe path coefficients in the PLS model representstandardized regression coefficients. The suggestedlower limit of substantive significance for regressioncoefficients is 0.05 [63]. In a more conservativeTable 5Correlations of latent variablesDecisionSupportSatisfactionInterfaceSatisfactionTaskSupportSatisfactionBI to useEC WebsiteDecisionSupportSatisfaction(0.913)InterfaceSatisfaction0.501*** (0.869)Task SupportSatisfaction0.778*** 0.636*** (0.888)BehavioralIntentionto Use EC0.612*** 0.482*** 0.697*** (0.853)web siteThe number parentheses is the square root of AVE, and *** p b 0.001.ructurE.J. Garrity et al. / Decision Support Systems 39 (2005) 485503 497position, path coefficients of 0.10 and above arepreferable. As shown in Fig. 3, all path coefficientsare over 0.1 satisfying both conservative criteria andthe suggested lower limit. They are also statisticallysignificant at p=0.001, which indicates that allhypotheses (H1a, H1b, H2, and H3) are supportedby the data. Moreover, a high R2 for each endogenousvariable in the structural model shows that this modelcan be used to predict the Behavioral Intention to Usea Web site.5.3. Order effects and treatment effectsThe survey questionnaires were designed in twodifferent formats with identical measures (Format Alisted the inferior site first, and Format B listed thesuperior site first). Order effects during the experi-ments were investigated. The results of a two-wayANOVA (Table 6a) showed that there was nostatistically significant difference in Behavioral Inten-Fig. 3. WISSM sttion to Use caused by the order of the presentation ofthe Web sites on the questionnaire.The treatment effects were explored next. Table 6aand 6b present that Behavioral Intention to Use theTable 6aTests of between-subjects effectsSource Sum of squares df MSuperior/inferior sites 74.34 1 74Formats 3.603 1 3.6Superior/inferiorFormats 3.244e03 1 3.2Error 464.122 271 1.7Total 5340.028 275Corrected total 543.474 274Dependent variable: behavioral intention to use an EC web site.superior Web site (Mean=4.699, S.D.=1.192) wassignificantly higher than the inferior Web site(Mean=3.650, S.D.=1.417). As expected, most sub-jects (112 out of 156 respondents) decided to purchasea digital camera from the superior site. Additionally,there are no interaction effects between the order ofpresentation and the experimental treatment.5.4. Effects of online buying experienceExperience entails general learning by trial anderror, and an increasing awareness and familiaritywith the types of domain problems and their structure[45]. Through the accumulation of online buyingexperience, the online buyers can develop a repertoireof Internet skills, knowledge, and meta-knowledgethat leads to effective performance in a specificdomain [66]. Thus, it is clear that online buyingexperience plays a role in individuals Web siteassessment process and influences the selection of aal model results.Web site that a buyer would purchase from.The current study investigated the impacts ofsubjects online buying experience on their Intentionto Use for the four experience groups. Based on theean square F Significance Eta squared.347 43.411 0.000 0.13803 2.104 0.148 0.00844e03 0.002 0.965 0.00013the difference between group 1 and the other groups.The analysis offers evidence that a buyers onlineto the Interface Satisfaction concept, will lead tohigh bIntention to Use the Web siteQ via an indirectpath through Decision Support and Task SupportSatisfaction.Table 6bMean (S.D.) statisticsTable 7aGroup means and standard deviationsGroupno.Number ofexperienceSamplesizeMean scoredifferencea(Mn)S.D. Contrastestimate(MnM1)1 0 30 0.344 1.3622 15 64 0.914 1.676 0.5693 610 18 1.555 1.679 1.211b4 10b 14 2.119 1.878 1.774cE.J. Garrity et al. / Decision Support Systems 39 (2005) 485503498experience has significant impacts on Intention toUse. The results show that more experienced onlinebuyers tend to more clearly observe the differencebetween two sites than less experienced buyers( pb0.005). In order to investigate the divergence ofsubjects intention in different groups, mean scoredifferences9 were calculated. While subjects in group1 who have never experienced online buying do notshow much difference in their buying intention withsuperior and inferior Web sites (M1=0.344), ones ingroup 4 who have more than ten online-buyingexperiences better recognize the difference(M4=2.119). In essence, the more an individualexperiences online purchasing, the better they eval-uate the quality of Web sites, and the more likely theyintend to use a superior site.6. Discussion and implicationsThis study has found support for the Web-basedInformation Systems Success Model introduced here.Overall, the results indicate strong support for themodel. Table 8 summarizes the results of the hypothesismean score of the seven-scale items of BehavioralIntention to Use an EC Web site, the contrasts aredeveloped by first performing an ANOVA test (Table7b) and then conducting a post hoc test where theANOVA indicates a difference among treatments(Table 7a). The post hoc tests involve three t tests onFormat A Format B TotalSuperior site 4.821 (1.316) 4.585 (1.059) 4.699 (1.192)Inferior site 3.772 (1.466) 3.549 (1.377) 3.650 (1.417)Total 4.317 (1.481) 4.053 (1.334) 4.176 (1.408)testing. WISSM explains almost 50% of the variationin Behavioral Intent to Use theWeb information system(R2=0.486). All of the paths in the model are statisti-cally significant (at p=0.001 level). Compared to priorstudies, WISSM effectively predicts the intention to9 Mean score difference is the difference between the meanscore of a subjects Behavioral Intention to Use the superior site andthe mean score of Behavioral Intention to Use the inferior site.use a Web-based information system. BehavioralIntention to Use is primarily explained via TaskSupport Satisfaction. The results of this model alsoshow, as in Davis et al.s [15] findings, that TaskSupport Satisfaction (corresponding to Perceived Use-fulness) directly influences Behavioral Intention to Usean ECWeb site and Interface Satisfaction (correspond-ing to Perceived Ease of Use) has an indirect impact onBehavioral Intention to Use an EC Web site throughTask Support Satisfaction.Lederer et al. [46] have urged IS researchers toexplore and develop different factors or latentvariables that can better explain information SystemUse and lead to models that explain greater variance.In this research, important antecedents of TaskSupport Satisfaction, the Interface Satisfaction andDecision Support Satisfaction dimensions, werevalidated in the context of a Web-based purchasinginformation system. The results are consistent withresearchers who study Web usability such as Nielsen[56]. Web systems that are well developed accordinga Mean score differences (Mn) are the difference between themean scores of superior and inferior web sites.b Total mean difference=1.004, pb0.08.c Total mean difference=1.004, pb0.06.Table 7bTests of between-subjects effectsSource Sum ofsquaresdf MeansquareF SignificanceIntercept 74.34 1 138.316 51.972 0.000Experiencegroup3.603 3 12.150 4.565 0.005Error 464.122 122 2.661Total 5340.028 126Satisfaction, for this study, is the users assessment ofDecision Support Satisfaction, which involves activ-upporities such as comparing important product featuresand prices, and examining the myriad of tradeoffs inbundled attributes offered at a Web site. Given thenew demands of electronic commerce and markets,such specialized task characteristics warrant furtherconsideration and measurement by researchers.7. ConclusionThe most important determinant of Task SupportTable 8Summary of hypothesis testingHypothesis Support SignificancelevelH1a Interface Satisfaction has apositive impact on DecisionSupport Satisfaction.Yes pb0.001H1b Interface Satisfaction has apositive impact on TaskSupport Satisfaction.Yes pb0.001H2 Decision SupportSatisfaction is positivelyassociated with Task upportSatisfaction.Yes pb0.001H3 Task Support Satisfactionpositively influencesBehavioral Intention to Usean EC Web site.Yes pb0.001E.J. Garrity et al. / Decision SThis study attempts to address several of the issuesin electronic commerce research: How should Web-based information systems success be measured in thecontext of User Satisfaction, whether existing meas-ures of success can be applied, and how consumersreact to commercial Web sites. The results of thisstudy, which draw on the DeLone and McLean Model[17], the Garrity and Sanders [25] model of ISsuccess, and TAM, support the use of WISSM forexamining Web-based information systems success.The results also confirm three fundamental WIS UserSatisfaction components (Task Support Satisfaction,Decision Support Satisfaction, and Interface Satisfac-tion) that explain approximately 50% of the variancein users Intention to Use a WIS. The amount ofvariance explained in Behavioral Intention to Use,however, needs to be carefully interpreted with theunderstanding that the experiment was done within acontrived situation. The respondents were not actuallymaking a purchase.An important facet of conducting WIS research isthe availability of a dependent variable that is able toreflect the concept of system success. A set ofquestions has been presented that can be used tomeasure WIS success. This is, of course, not the onlyavailable scale nor should it be considered the finalinstrument. The brevity of the scale along with itsreasonably interpretable and clear dimensions desig-nates this measure as a useful addition to the currentliterature.This research is not without limitations. First, thisexperiment utilized a Web-based information sys-tem, focusing on a fundamental but core purchasingtask. The subjects in this study were asked todetermine a product to purchase, given a constrainedbudget. A complete assessment of consumer pur-chasing behavior in an electronic market should,however, include after sale support, order tracking,and should be extended to the areas of bneedassessment,Q advertising, and promotion (see Ref.[57]). While the current study examined thoserespondents who were considered a major usergroup, further experiments could utilize new referentgroups and control the experimental environmentmore closely. That is, future study needs to includevarious user groups (i.e., individuals in high school,those who do not go to college, and users withdisposable income) and investigate differences intheir behavior. Notwithstanding these limitations, thestudy supports the basic research question thatDecision Support Satisfaction and Task SupportSatisfaction do matter in WIS success. This followsrecommendations by Lederer et al. [46] to developmore robust constructs to explain Web-based infor-mation systems success.In the future, research on Web-based informationsystems assessment could be extended to considervariations in Web support required for different typesof products. For example, the experimental designcould manipulate differences in product informationintensity and determine the effect on purchasebehavior. Palmer and Griffith [59] call for researchinto Web-based product purchase decisions thatmanipulate the product type to account for differencest Systems 39 (2005) 485503 499in consumer binvolvementQ in the purchase. Web sitedesign for high involvement products may requiredifferent types of decision and task support. Inaddition, most research in the Web success area hasfound a strong link between Trust and purchasebehavior (e.g., Ref. [82]). For future work, it wouldbe worthwhile to consider the effect of Trust onBehavioral Intention and different User Satisfactiondimensions. Future research might also examine thepotential of the Web-based Information SystemsSuccess Model for different Web-based applications.Models of success may need to be modified astechnology changes and new models of business andcomputing evolve.AcknowledgementThe authors thank two anonymous reviewers fortheir constructive comments and suggestions onearlier drafts of this article. The authors also thankAppendix A. Cross loadings between measurementitems and constructsTUSE 0.692 0.847 0.563 0.632E.J. Garrity et al. / Decision Suppo500TEXT 0.730 0.908 0.548 0.651TQIK 0.666 0.910 0.556 0.604TESY 0.671 0.885 0.592 0.586ICLR 0.484 0.623 0.862 0.477ILEN 0.371 0.485 0.869 0.334IFRN 0.479 0.603 0.901 0.478IESY 0.444 0.579 0.927 0.428IWNT 0.495 0.588 0.878 0.462IINT 0.420 0.558 0.878 0.419ISKL 0.361 0.485 0.837 0.360INAV 0.399 0.458 0.794 0.357UINF 0.568 0.639 0.491 0.858UCRT 0.571 0.582 0.366 0.872UCRE 0.482 0.569 0.458 0.776USIM 0.534 0.665 0.521 0.883Item DecisionSupportSatisfactionTaskSupportSatisfactionInterfaceSatisfactionBI to useEC WebsiteDBET 0.923 0.691 0.455 0.527DEFF 0.920 0.732 0.511 0.580DPRI 0.895 0.706 0.402 0.568Dr. Wynne Chin for his correspondence related to thediscriminant validity issues.UINC 0.492 0.533 0.303 0.878UFUT 0.475 0.560 0.293 0.848References[1] R. Agarwal, J. Prasad, Are individual differences germane tothe acceptance of new information technologies? DecisionSciences 30 (2) (1999) 361391.[2] E.W. Anderson, M.W. Sullivan, The antecedents and con-sequences of customer satisfaction for firms, MarketingScience 12 (2) (1993) 125143.[3] D.P. Ausubel, Educational Psychology: A Cognitive View,Holt, Rinehart and Winston, New York, NY, 1968.[4] R.P. Bagozzi, A prospectus for theory construction in market-ing, Journal of Marketing 48 (1984 Winter) 1129.[5] R.P. Bagozzi, L.W. Phillips, Representing and testing organ-izational theories: a holistic construal, Administrative ScienceQuarterly 27 (3) (1982) 459489.[6] J.Y. Bakos, Reducing buyer search costs: implications forelectronic marketplaces, Management Science 43 (12) (1997)16761692.[7] Y. Bakos, The emerging role of electronic marketplaces on theInternet, Communications of the ACM 41 (8) (1998) 3542.[8] J.R. Bettman, E.J. Johnson, J.W. Payne, A componentialanalysis of cognitive effort in choice, Organizational Behaviorand Human Decision Processes 45 (1990) 111139.[9] R.O. Briggs, A technology transition model derived from fieldinvestigation of GSS use aboard the U.S.S. CORONADO,Journal of Management Information Systems 15 (3) (1998/1999) 151195.[10] W.W. Chin, The Partial Least Squares approach to structuralequation modeling, in: G.A. Marcoulides (Ed.), ModernMethods for Business Research, Lawrence Erlbaum Associ-ates, Mahwah, NJ, 1998, pp. 295336.[11] J. Conklin, Hypertext: an introduction and survey, IEEEComputing Survey 20 (9) (1987) 1741.[12] P.A. Dabholkar, Incorporating choice into an attitudinal frame-work: analyzing models of mental comparison processes,Journal of Consumer Research 21 (1) (1994) 100118.[13] F.D. Davis, A Technology Acceptance Model for EmpiricallyTesting New End-User Information Systems: Theory andResults. PhD dissertation, MIT, 1986.[14] F.D. Davis, Perceived usefulness, perceived ease of use, anduser acceptance of information technology, MIS Quarterly 13(3) (1989) 319340.[15] F.D. Davis, R.P. Bagozzi, P.R. Warshaw, User acceptance ofcomputer technology: a comparison of two theoretical models,Management Science 35 (8) (1989) 9821003.[16] S.A. Davis, R.P. Bostrom, Training end users: an experimentalinvestigation of the roles of the computer interface and trainingmethods, MIS Quarterly 17 (1) (1993) 6185.[17] W.H. DeLone, E.R. McLean, Information systems success: thequest for the dependent variable, Information SystemsResearch 3 (1) (1992) 6195.[18] W.J. Doll, G. Torkzadeh, The measurement of end-usercomputing satisfaction, MIS Quarterly 12 (2) (1988) 259274.[19] J.F. Engel, D.T. Kollat, R.D. Blackwell, Consumer Behavior,21 ed.., Holt, Rinehart and Winston, New York, 1973.rt Systems 39 (2005) 485503[20] L. Enos, Report: B2B Still Driving e-Commerce, http://www.ecommercetimes.com, Dec. 11, 2000.E.J. Garrity et al. / Decision Support Systems 39 (2005) 485503 501[21] C. Fornell, A Second Generation of Multivariate Analysis,Praeger, New York, NY, 1982.[22] C. Fornell, F.L. Bookstein, Two structural equation models:LISREL and PLS applied to consumer exit-voice theory,Journal of Marketing Research 19 (4) (1982) 440452.[23] C. Fornell, D. Larcker, Evaluating structural equation modelswith unobservable variables and measurement error, Journal ofMarketing Research 18 (1981) 3950.[24] C.R. Franz, D. Robey, Organizational context, user involve-ment, and the usefulness of information systems, DecisionSciences 17 (3) (1986) 329356.[25] E.J. Garrity, G.L. Sanders, Dimensions of information systemssuccess, in: E.J. Garrity, G.L. Sanders (Eds.), InformationSystems Success Measurement, Idea Group Publishing,Hershey, PA, 1998, pp. 1345.[26] E.J. Garrity, G.L. Sanders, Introduction to information systemssuccess measurement, in: E.J. Garrity, G.L. Sanders (Eds.),Information Systems Success Measurement, Idea GroupPublishing, Hershey, PA, 1998, pp. 112.[27] D. Gefen, D. Straub, The relative importance of perceivedease-of-use in IS adoption: a study of e-commerce adoption,Journal of the Association for Information Systems 1 (8)(2000) 121.[28] D.L. Goodhue, Developing a Theory-Based Measure of UserSatisfaction: The Task-Systems Fit Questionnaire, in WorkingPaper, 1990 Information and Decision Sciences, University ofMinnesota.[29] D.L. Goodhue, Development and measurement validity of atasktechnology fit instrument for user evaluations of infor-mation systems, Decision Sciences 29 (1) (1998) 105138.[30] D.L. Goodhue, R.L. Thompson, Tasktechnology fit andindividual performance, MISQuarterly 19 (2) (1995) 213236.[31] M.E. Gordon, L.A. Slade, N. Schmitt, The bscience of thesophomoreQ revisited: from conjecture to empiricism, Acad-emy of Management Review 11 (1) (1986) 191207.[32] T. Helstrup, One, two, or three memories? A problem-solvingapproach to memory from performed acts, Acta Psychologica66 (1987) 3768.[33] D.L. Hoffman, W.D. Kalsbeek, T.P. Novak, Internet and Webuse in the U.S., Communications of the ACM 39 (12) (1996)3646.[34] J. Howard, J. Sheth, The Theory of Buyer Behavior, FreePress, New York, NY, 1972.[35] M. Igbaria, M. Tan, The consequences of informationtechnology acceptance on subsequent individual performance,Information & Management 32 (3) (1997) 113121.[36] B. Ives, M.H. Olson, J.J. Baroudi, The measurement of userinformation satisfaction, Communications of the ACM 26 (10)(1983) 785793.[37] E.J. Johnson, J.W. Payne, Effort and accuracy in choice,Management Science 31 (4) (1985) 395414.[38] R. Kalakota, A.B. Whinston, Electronic Commerce: AManagers Guide, Addison-Wesley, Reading, MA, 1997.[39] E. Karahanna, D.W. Straub, N.L. Chervany, Informationtechnology adoption across time: a cross-sectional comparisonof pre-adoption and post-adoption beliefs, MIS Quarterly 23(2) (1999) 183213.[40] G. Kaufmann, Imagery, language and cognition: toward atheory of symbolic activity in human problem solving,Columbia University Press, NY, 1980.[41] G. Kaufmann, A theory of symbolic representation in problemsolving, Journal of Mental Imagery 9 (2) (1985) 5170.[42] R.L. Keeney, The value of Internet commerce to the customer,Management Science 45 (4) (1999) 533542.[43] M. Keil, P.M. Beranek, B.R. Konsynski, Usefulness and easeof use: field study evidence regarding task considerations,Decision Support Systems 13 (1) (1995) 7591.[44] P.J. Kirs, G.L. Sanders, R.P. Cerveny, D. Robey, Anexperimental validation of the Gorry and Scott Mortonframework, MIS Quarterly 13 (2) (1989) 183197.[45] J.L. Kolodner, Towards an understanding of the role ofexperience in the evolution from novice to expert, Journal ofManMachine Studies 19 (1983) 497518.[46] A.L. Lederer, D.J. Maupin, M.P. Sena, Y. Zhuang, TheTechnology Acceptance Model and the world wide web,Decision Support Systems 29 (3) (2000) 269282.[47] G.L. Lohse, P. Spiller, Electronic shopping, Communicationsof the ACM 41 (7) (1998) 8187.[48] D. Lucarella, A. Zanzi, A visual retrieval environment forhypermedia information systems, ACM Transactions onInformation Systems 14 (1) (1996) 329.[49] M.A. Mahmood, End user computing productivity: a newapproach to help resolve information technology productivityparadox, Journal of End User Computing 7 (2) (1995) 25.[50] K. Mathieson, Predicting user intentions: comparing theTechnology Acceptance Model with the theory of plannedbehavior, Information Systems Research 3 (3) (1991)173191.[51] R.E. Mayer, The psychology of how novices learn computerprogramming, Computing Surveys 13 (1981) 121141.[52] N.P. Melone, A theoretical assessment of the user-satisfactionconstruct, Management Science 36 (1) (1990) 7691.[53] V. Mittal, P. Kumar, M. Tsiros, Attribute-level performancesatisfaction, and behavioral intentions over time: a consump-tion-system approach, Journal of Marketing 63 (2) (1999)88101.[54] M.G. Morris, A. Dillon, How user perceptions influencesoftware use, Decision Support Systems, IEEE Software (1997JulyAugust) 5865.[55] F.N. Nicosia, Consumer decision processes: a futuristic view,Advances in Consumer Research 9 (1982) 1719.[56] J. Nielsen, Designing Web Usability: The Practice ofSimplicity, New Riders Publishing, Indianapolis, IN, 2000.[57] R.M. OKeefe, T. McEachern, Web-based customer DecisionSupport Systems, Communications of the ACM 41 (3) (1998)7178.[58] J.W. Palmer, Web site usability, design, and performancemetrics, Information Systems Research 13 (2) (2002) 151167.[59] J.W. Palmer, D.A. Griffith, An emerging model of web sitedesign for marketing, Communications of the ACM 41 (3)(1998) 4451.[60] M. Parthasarathy, A. Bhattacherjee, Understanding post-adoption behavior in the context of online services, Informa-tion Systems Research 9 (4) (1998) 362379.E.J. Garrity et al. / Decision Support Systems 39 (2005) 485503502[61] J.W. Payne, Contingent decision behavior, PsychologicalBulletin 92 (2) (1982) 382402.[62] J.W. Payne, J.R. Bettman, E.J. Johnson, The AdaptiveDecision Maker, Cambridge, New York, NY, 1993.[63] E.J. Pedhazur, Multiple regression in behavioral research:explanation and prediction, 3rd edn., Harcourt Brace CollegePublishers, Fort Worth, 1997.[64] H.R. Rao, A.F. Salem, B. DosSantos, Introduction: marketingand the internet, Communications of ACM 41 (3) (1998)3234.[65] G.L. Sanders, MIS/DSS success measure, Systems, Objec-tives, Solutions 4 (1984) 2934.[66] K.D. Schenk, N.P. Vitalari, K.S. Davis, Differences betweennovice and expert systems analysts: what do we know andwhat do we do? Journal of Management Information Systems15 (1) (1998) 950.[67] E. Schonberg, T. Cofino, R. Hoch, M. Podlaseck, S.L.Spraragen, Measuring success, Communications of the ACM43 (8) (2000) 5357.[68] P.B. Seddon, A respecification and extension of the DeLoneand McLean Model of IS Success, Information SystemsResearch 8 (3) (1997) 240253.[69] P.B. Seddon, M. Kiew, A partial test and development of theDeLone and McLean Model of IS Success, Proceedings ofFifteenth International Conference on Information Systems(ICIS 1994), 1994, pp. 99110.[70] B. Shneiderman, Designing the User Interface: Strategies forEffective HumanComputer Interaction, 3rd edn., Addison-Wesley Publishers, Reading, MA, 1998.[71] S.M. Shugan, The cost of thinking, Journal of ConsumerResearch 7 (2) (1980) 99111.[72] H.A. Simon, A behavioral model of rational choice, QuarterlyJournal of Economics 69 (1955) 99118.[73] D.M. Strong, Y.W. Lee, R.Y. Wang, Data quality in context,Communications of the ACM 40 (5) (1997) 103110.[74] D.M. Szymanski, D.H. Henard, Customer satisfaction: a meta-analysis of the empirical evidence, Journal of the Academy ofMarketing Science 29 (1) (2001) 1635.[75] D.M. Szymanski, R.T. Hise, e-Satisfaction: an initial exami-nation, Journal of Retailing 76 (3) (2000) 309322.[76] J.M. Tenenbaum, WISs and electronic commerce, Communi-cations of the ACM 41 (7) (1998) 8990.[77] T.S.H. Teo, V.K.G. Lim, R.Y.C. Lai, Intrinsic and extrinsicmotivation in internet usage, Omega 27 (1) (1999) 2537.[78] M. Thqring, J. Hannemann, J.M. Haake, Designing forcomprehension: a cognitive approach to hypermedia deve-lopment, Communications of the ACM 38 (8) (1995)5766.[79] R.D. Tobias, An Introduction to Partial Least SquaresRegression, SAS Institute, 1999, pp. 18.[80] R. Took, Putting design into practice: formal specification andthe user interface, in: H. Harrison, H. Thimbleby (Eds.),Formal Methods in HumanComputer Interaction, CambridgeUniversity Press, Cambridge, UK, 1990.[81] K. Utting, N. Yankelovich, Context and orientation in hyper-media networks, ACM Transactions on Information Systems 7(1) (1989) 5884.[82] C. Van Slyke, F. Belanger, C.L. Comunale, Factors influen-cing the adoption of web-based shopping: the impact of trust,Database for Advances in Information Systems 35 (2) (2004)3249.[83] S. Varki, M. Colgate, The role of price perceptions in anintegrated model of behavioral intentions, Journal of ServiceResearch 3 (3) (2001) 232240.[84] V. Venkatesh, F.D. Davis, A model of the antecedents ofperceived ease of use: development and test, DecisionSciences 27 (10) (1996) 451481.[85] V. Venkatesh, F.D. Davis, A theoretical extension of theTechnology Acceptance Model: four longitudinal field studies,Management Science 46 (2) (2000) 186204.[86] V. Venkatesh, M.G. Morris, Why dont men ever to stop to askfor directions? Gender, social influence, and their role intechnology acceptance and usage behavior, MIS Quarterly 24(1) (2000) 115139.[87] C.E. Werts, R.L. Linn, K.G. Joreskog, Intraclass reliabilityestimates: testing structural assumptions, Educational andPsychological Measurement 34 (1) (1974) 2533.[88] H. Wold, Causal flows with latent variables, EuropeanEconomic Review 5 (1974) 6786.[89] H. Wold, Partial Least Squares, in: S. Kotz, N.L. Johnson(Eds.), Encyclopedia of Statistical Sciences, vol. 6, Wiley,New York, 1985, pp. 581591.[90] P. Zhang, G.M. von Dran, User expectations and rankings ofquality factors in different web site domains, InternationalJournal of Electronic Commerce 6 (2) (2001/2002) 9.[91] V. Zwass, Electronic commerce: structures and issues, Interna-tional Journal of Electronic Commerce 1 (1) (1996) 323.Edward J. Garrity is Associate Professor of Information Systemsin the Richard J. Wehle School of Business at Canisius College inBuffalo, New York. He holds a PhD in MIS and an MBA from theState University of New York at Buffalo. He has published researchin Datamation, DATA BASE, Encyclopedia of Computer Science,Encyclopedia of Information Systems, Expert Systems with Appli-cations, Heuristics, Information Resources Management Journal,the Journal of Management Information Systems, the Journal ofManagement Systems, and several book chapters. He has also co-authored a book of readings entitled bInformation Systems SuccessMeasurement,Q published by Idea Group Publishing. His currentresearch interests are information systems development issues,information systems success measurement, and IT evaluationmanagement.Bonnie C. Glassberg is currently an Assistant Professor at MiamiUniversity/Ohio. She has a PhD from the University of SouthCarolina, an MBA from Indiana University, and a BSc degree fromCornell. As an Assistant Professor at the University at Buffalo(19972001), she also taught information systems at the graduateand undergraduate level. Bonnie has 10 years of industry experienceas a Senior Analyst with a Fortune 100 company. She has beeninvolved in the planning, design, implementation, and managementof major corporate systems. She has served as a reviewer for JMIS,CACM, Database, and the Mid-America Journal of Business.Yong Jin Kim is an Assistant Professor of Management InformationSystems at the School of Management at the State University of NewYork at Binghamton. He holds a PhD in MIS from the StateUniversity of New York at Buffalo. He participated in large ITprojects including a BPR implementation for Samsung LifeInsurance, an online accounting systems development project forthe Bank of Korea, and a business recovery planning project for theKorea Financial Telecommunication and Clearing Institute. Hisresearch interests are in IS success, e-commerce, and informationtechnology valuation, and knowledgemanagement. He has publishedpapers in outlets such as Communications of the ACM, DecisionSupport Systems, and Knowledge and Process Management. He hasalso published book chapters on IS Success and e-learning.Seung Kyoon Shin is an Assistant Professor of InformationSystems in the College of Business Administration at the Universityof Rhode Island. Prior to joining academia, he worked in industry asa software specialist and system integration consultant. His researchhas appeared in Communications of the ACM, and the InternationalConference on Information Systems, and Hawaii InternationalConference on System Sciences (HICSS). His current researchinterests include strategic database design for data warehousing anddata mining, Web-based information systems success, and economicand cultural issues related to intellectual property rights.G. Lawrence Sanders, PhD, is Professor and Chair of theDepartment of Management Science and Systems in the School ofManagement at the State University of New York at Buffalo. He hastaught MBA courses in the Peoples Republic of China andSingapore. His research interests are in the ethics and economicsof digital piracy, systems success measurement, cross-culturalimplementation research, and systems development. He haspublished papers in outlets such as The Journal of ManagementInformation Systems, The Journal of Business, MIS Quarterly,Information Systems Research, the Journal of Strategic InformationSystems, the Journal of Management Systems, Decision SupportSystems , Database Programming and Design , and DecisionSciences. He has also published a book on database design andco-edited two other books.E.J. Garrity et al. / Decision Support Systems 39 (2005) 485503 503An experimental investigation of web-based information systems success in the context of electronic commerceIntroductionReview of information systems success literatureDeLone and McLean Model of IS SuccessTechnology Acceptance ModelGarrity and Sanders' Model of IS SuccessResearch model and hypothesisTheoretical background of the research modelResearch question and hypothesesInterface SatisfactionDecision Support SatisfactionTask Support SatisfactionResearch methodSampling methodData collection procedureSample frameSubject poolDemographicsPartial Least Squares (PLS)Operationalization of research variablesAnalysis and resultsReliability and validity testsAssessment of the structural modelOrder effects and treatment effectsEffects of online buying experienceDiscussion and implicationsConclusionAcknowledgementCross loadings between measurement items and constructsReferences

Recommended

View more >