learning and using electronic information products and services: a field study

11
LEARNING AND USING ELECTRONIC INFORMATION PRODUCTS AND SERVICES: A FIELD STUDY VikasMittal Mohanbir S. Sawhney f ABSTRACT Because of its implications for revenues, marketers of Electronic Information Products and Services (EIPS) are interested in increasing usage of their offering. In this research we investigate how structuring the initial-learning experience can impact subsequent usage. The initial learning experience can be structured by varying the amount of process-based or content-based knowledge that consumers acquire during the initial learning phase. A field experiment where actual usage of a complex, Web-based EIPS was monitored is described. Results show that both content-based and process-based learning are important, though emphasizing either to the virtual exclusion of the other during the initial learning experience may be deleterious for usage. © 2001 John Wiley & Sons, Inc. and Direct Marketing Educational Foundation, Inc. f JOURNAL OF INTERACTIVE MARKETING VOLUME 15 / NUMBER 1 / WINTER 2001 2 VIKAS MITTAL is Assistant Professor of Marketing in the Katz Graduate School of Business, University of Pittsburgh, Pittsburgh, Pennsylvania. MOHANBIR S. SAWHNEY is McCormick Tribune Foundation Professor of Electric Commerce and Technology in the Kellogg Graduate School of Management, Northwestern University, Evanston, Illinois. Please address all correspondence to Vikas Mittal, Assistant Professor of Marketing, 360 Mervis Hall, Katz Graduate School of Business, University of Pittsburgh, Pittsburgh, PA 15260; e-mail: [email protected].

Upload: vikas-mittal

Post on 06-Jun-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

LEARNING AND USING ELECTRONIC

INFORMATION PRODUCTS AND

SERVICES: A FIELD STUDY

V i k a s M i t t a lM o h a n b i r S . S a w h n e y

f

A B S T R A C TBecause of its implications for revenues, marketers of ElectronicInformation Products and Services (EIPS) are interested inincreasing usage of their offering. In this research we investigatehow structuring the initial-learning experience can impactsubsequent usage. The initial learning experience can be structuredby varying the amount of process-based or content-based knowledgethat consumers acquire during the initial learning phase. A fieldexperiment where actual usage of a complex, Web-based EIPS wasmonitored is described. Results show that both content-based andprocess-based learning are important, though emphasizing either tothe virtual exclusion of the other during the initial learningexperience may be deleterious for usage.

© 2001 John Wiley & Sons, Inc. and

Direct Marketing Educational Foundation, Inc.

f

JOURNAL OF INTERACTIVE MARKETING

VOLUME 15 / NUMBER 1 / WINTER 2001

2

VIKAS MITTAL is AssistantProfessor of Marketing in the KatzGraduate School of Business,University of Pittsburgh,Pittsburgh, Pennsylvania.

MOHANBIR S. SAWHNEY isMcCormick Tribune FoundationProfessor of Electric Commerceand Technology in the KelloggGraduate School of Management,Northwestern University, Evanston,Illinois.

Please address all correspondenceto Vikas Mittal, Assistant Professorof Marketing, 360 Mervis Hall, KatzGraduate School of Business,University of Pittsburgh,Pittsburgh, PA 15260; e-mail:[email protected].

Today, many firms are in the business ofproviding Electronic Information Products orServices (EIPS). We define an EIPS1 as a prod-uct or service that is accessed through an electronicuser interface and contains significant informationcontent. Examples include:

● Websites such as www.gartner.com andwww.usatoday.com that aggregate informa-tion content and enable customers to ac-cess it electronically.

● Products such as the Palm Pilot where in-formation content is input and accessedusing an electronic interface.

● E-tailers such as www.amazon.com that pro-vide significant information content thatmust be consumed electronically. For in-stance, in addition to purchasing books,customers may go to amazon.com to seebook titles, compare prices, read book re-views, and take advantage of suggestionsmade through collaborative filtering.

Both the process of using an EIPS and thecontent in an EIPS can vary from being verysimple to very complex. Firms need to deter-mine the level of knowledge that consumersshould acquire about the process and contentof using an EIPS. This is important becausecustomers have limited time to acquire suchknowledge and because such knowledge criti-cally affects usage. Strategically, a firm offeringan EIPS is interested in increasing usage amongconsumers because, irrespective of the businessmodel used, increased usage, in most instances,may result in increased revenues. Typically,three classes of revenue models are being usedamong EIPS providers (cf. Rayport, 1999):

● Fee-based access: Some providers such aswww.garterninteractive.com provide accessto clients based on an annual fee. In suchcases, firms may measure the value of the

service based on the extent to which itsemployees use the website. By increasingusage, such fee-based information provid-ers can ensure that they provide value totheir client.

● Free access: Some providers such as www.usatoday.com provide free access to con-sumers but generate revenues by chargingfees to firms placing banner ads, links andso forth on their website. For such provid-ers, the larger customer base and higherusage for each customer (on average) canenable them to attract advertisers andcharge commensurate advertising fees.

● E-tailing: Providers such as www.amazon.com generate most of their revenue by sell-ing products and services. Yet, consumersusing the site for information consumptionmay access the website more often, stay onthe website longer, trust the website more,and feel relationally obligated to purchasefrom that website. This can lead to repeattraffic, longer time spent online, higherrevenues from sales, higher banner adver-tising revenues, and so forth.

However, despite its strategic importance re-search examining usage of an EIPS is sparse atbest. Though some articles have examined usageof services, their goal has been to relate usage tosatisfaction with the service provider (e.g., Bolton,1998). Our research is different. We investigatehow structuring the initial learning experience foran EIPS can affect subsequent usage. We arguethat the initial learning experience can be struc-tured based on the relative emphasis placed oncontent-oriented and process-oriented learning.In support of this contention, results from a quasi-field experiment are described where substantialdifferences in usage were obtained by manipulat-ing the amount of content and process learningduring the initial learning experience.

The Initial Learning ExperienceIn this paper we focus on the initial post-

purchase learning experience. This is distinctfrom pre-purchase learning and ongoing post-purchase learning. Pre-purchase learning (in-formation search) is aimed at helping consum-

1 The term “information product” has been defined broadly inthe literature to include information technology, advertising, me-dia, and paper-based information products (Glazer, 1991; Meyer& Zack, 1996). We adopt a more focused definition by excludingphysical information products that do not have an electronic userinterface (e.g., books, paper-catalogs, or magazines).

L E A R N I N G A N D U S I N G E L E C T R O N I C I N F O R M A T I O N P R O D U C T S A N D S E R V I C E S

JOURNAL OF INTERACTIVE MARKETING ● VOLUME 15 / NUMBER 1 / WINTER 2001

3

ers in brand evaluation and choice whereaspost-purchase learning is aimed at learningabout the “consumption” of the brand pur-chased. Within post purchase learning, initialpost-purchase learning should be distinguishedfrom ongoing, passive learning that occurs as aresult of consumption (Hoch & Deighton,1989; West, Brown, & Hoch, 1996). Likely, on-going post-purchase learning is interactive innature, evolves over time, and impacts con-sumption as usage. However, our work is limitedto examining the initial, post-purchase learningexperience.

During the initial, post-purchase learning ex-perience, consumers may engage in a variety ofactivities such as reading manuals, taking tuto-rials, or engaging in self-guided exploration ofthe EIPS. A key characteristic of the initial post-purchase learning experience is that it is limiteddue to demands on consumers’ time (Feldman& Hornik, 1981), and because this phase istypically perceived as “necessary non-work ornon-compensated time” (Schary, 1971, p. 54,Table 1). Therefore, it is critical for a firm tomanage the limited amount of time devoted tothe initial learning experience.

Process-Oriented and Content-OrientedKnowledge

The knowledge that consumers acquire dur-ing the initial learning experience can be con-ceptualized along two dimensions—process-ori-ented knowledge and content-oriented knowledge.Process-oriented knowledge results from learn-ing about how to use the EIPS. For instance, foran online grocery shopping site, consumers maywant to know how to search the database ofitems, how to use a coupon, how to set up theirpersonal shopping lists, how to use a “shoppingcart,” and so on. Content-oriented learning, onthe other hand, refers to learning the actualcontent or information residing in the EIPS.For online grocery shopping, content-orientedlearning may include knowledge about the lay-out of the virtual aisles, the nutritional informa-tion offered, and the brands and manufacturersrepresented in the store.

The distinction between these two dimen-sions of learning has been used in many con-

texts. Management theorists distinguish be-tween the acquisition of knowledge versus skillsby managers (Nass, 1994). Consumer behavior-ists (Leong, Busch, & Roedder-John, 1989) dis-tinguish between schemas and scripts. Cognitivepsychologists such as Anderson (1996) andSmith (1990) distinguish between declarativeversus procedural knowledge, or semantic ver-sus production rules.

Smith (1990) argues that both content- andprocess-oriented representations can act as cog-nitive mediators of behavior, though their rolecan be markedly different. An interesting illus-tration of this point is provided by McKendreeand Anderson (cited in Anderson, 1987; Smith,1990, p. 29). They had two groups of peoplestart with exactly the same content, but eachgroup applied different processes in using thesame content. One group merely evaluated theresult of a particular combination of LISP func-tions, while the other actually coded combina-tions of LISP functions that would achieve adesired result. The relevant content-knowledgein both cases was the knowledge of results com-puted by each primitive function and how thefunctions may be combined in LISP. Subjectswho practiced evaluation for four days in-creased greatly in their speed and accuracy atthe evaluation task—but did not improve at allin the coding test even though it involved thesame content. The process of using content in aparticular way is what becomes more efficientwith practice, as opposed to the knowledge it-self. In another study, Kolers (1976) had sub-jects practice the skill of reading text in novelorientations (right to left, backward, etc.). Notonly did these subjects improve with practice,but after one year, they were also significantlyfaster (60% to 80% based on different mea-sures) than unpracticed subjects. Moreover,subjects were slightly, though significantly fasteron pages they had practiced a year earlier. How-ever, even for new content, practice subjectswere faster than non-practice subjects. As Smithsummarizes:

Production strength is a long lasting property,decreasing only very slowly. Increases in proce-dural strength and efficiency will be distin-

J O U R N A L O F I N T E R A C T I V E M A R K E T I N G

JOURNAL OF INTERACTIVE MARKETING ● VOLUME 15 / NUMBER 1 / WINTER 2001

4

guishable from increases in accessibility ofcontent (declarative knowledge) by the meth-odological principle of specificity. Contentspecificity will not be observed, because astrengthened production is able to operatefaster on any data, not just the data with whichit was practiced. The effect will, however, beprocess specific because only tasks that involvethe strengthened production will show speed-ups from practice. Even other processes thataccess the same declarative knowledge will notbenefit from procedural strengthening.(Smith, 1990, p. 28)

These findings are equally applicable in thecontext of EIPS. For instance, if consumerslearn the content of an electronic database, butnot the process of using it, the anticipated im-provements in speed in terms of using the da-tabase are unlikely to occur. Further, once it islearnt, a process is rapidly transferable across avariety of content chunks. Similarly, once con-tent is learned, the benefits of doing so maytransfer across many processes. For instance,consumers who have learned the various cate-gories used to organize the content of an onlinenewspaper can apply all processes such assearch, retrieval, and saving information moreefficiently. This suggests that there may be asynergistic effect between process- and content-oriented knowledge in determining usage.

Process-Oriented and Content-OrientedKnowledge: How Much?

Ideally firms would want consumers to ac-quire high levels of both content- and process-oriented knowledge. However, due to the lim-ited duration of the initial learning experiencethere is a tradeoff between the amount of con-tent- or process-oriented knowledge.

We propose that the optimal amount of pro-cess-oriented learning versus content-orientedlearning for an EIPS is contingent on its relativecomplexity on the content and process dimen-sions. These are shown in Figure 1. For in-stance, using personal finance software (e.g.,Quicken) or voice recognition software is a rel-atively complex process, while using a YellowPages website or reading an online newspaper isprocedurally simple. Similarly, on the contentdimension, an EIPS can be relatively complex(deep and broad content) or simple (shallowand simple content). By understanding wherean EIPS fits into this classification scheme, itsmarketers can design an initial learning experi-ence to optimize usage. Note, however, that therelative complexity of an EIPS is likely to varybased on factors related to the customer (e.g.,knowledge and expertise), market (e.g., matu-rity of the EIPS category), and the usage goals(e.g., using the Palm Pilot to mange appoint-

F I G U R E 1Dimensions of Complexity for EIPS

L E A R N I N G A N D U S I N G E L E C T R O N I C I N F O R M A T I O N P R O D U C T S A N D S E R V I C E S

JOURNAL OF INTERACTIVE MARKETING ● VOLUME 15 / NUMBER 1 / WINTER 2001

5

ments or complex tasks like e-mailing and ad-dress book).

For an EIPS that has high level of processcomplexity but very low content complexity, theinitial learning phase may be geared towardprocess-oriented learning only. A high level ofprocess-oriented knowledge manifests itself asreduced cognitive effort and decreased time toperform tasks, without any loss in quality ofperformance (Alba & Hutchinson, 1987; Kolers,1976; Smith, 1990). As consumers gain moreprocess-oriented knowledge, they can performfamiliar tasks automatically and without con-scious control. Moreover, they can perform sev-eral tasks simultaneously without significant re-duction in efficiency. Reduced cognitive effortin executing processes may free up cognitiveresources that are now available for other tasksrelated to the usage experience. For instance,consumers who have learned how to navigate awebsite with high automaticity can spend moretime reading the content and elaborating on it.However, due to the simple and/or limitedamount of content, gains from transferring pro-cesses across various content chunks may notoccur.

Conversely, when an EIPS has very complexcontent, but a simple process, the initial learn-ing phase should emphasize content-basedlearning. Greater content-oriented learningleads to richer representations of products andconsumption experiences that are more com-plete and articulate (West, Brown, & Hoch,1996). Consumers also gain a deeper under-standing of causal relationships among the con-tent chunks, and the set of content chunks thatthey need to access to achieve specific consump-tion goals (Alba & Hutchinson, 1987). For in-stance, in an online grocery shopping experi-ence, consumers may be able to search bynutritional content, by manufacturer, or byprice to achieve different shopping objectives.

However, what if the EIPS has high levels ofboth process and content complexity? Due tothe limited initial learning period, managersface a tradeoff. We postulate that it is in suchsituations that the synergistic effect betweenprocess-oriented and content-oriented knowl-edge in affecting usage will be most pro-

nounced. First, enhanced process-orientedlearning frees up cognitive resources (Alba &Hutchinson, 1987) that can be devoted to ac-quiring and processing additional content-ori-ented knowledge. For content-poor EIPS, freedcognitive resources may not have such a positiveeffect because there is no additional content tobe acquired and processed. Second, as shown bySmith (1990), process learning is not contentspecific. After learning a particular process,consumer can apply it to new content with highlevel of effectiveness. This, in turn, may freecognitive resources devoted to more usage.Third, after learning some of the content in acomplex EIPS, consumers can apply a givenprocess more efficiently to it. Again such gainsin usage may not occur for EIPS with a single,simple process.

In summary, for a complex EIPS, both effi-ciency and elaborative gains may ensue if bothprocess- and content-oriented learning are em-phasized during the initial learning experience.If one type of learning is emphasized at theexpense of the other, such synergistic gains areunlikely to accrue. This suggests, that in a lim-ited-period training session, both types of learn-ing, rather than one type to the exclusion of theother is optimal for maximizing usage.

STUDYThe impact of managing content- and process-oriented learning on usage of a complex EIPSwas tested in a field-experiment. The EIPS usedwas the website of a commercial provider oftechnical and business information. In the ex-periment, the amount of content- and process-oriented learning during the initial learningexperience was systematically varied, and its ef-fect on subsequent usage was measured. Thetwo key strengths of the study are a behavioralmeasure of usage and longitudinal data. Basedon the conceptualization earlier, we expect par-ticipants receiving high amounts of only contentor only process learning to have lower level ofusage than those receiving moderate levels ofboth content- and process-oriented learning.

The study was conducted in conjunction withGartner Interactive’s electronic information

J O U R N A L O F I N T E R A C T I V E M A R K E T I N G

JOURNAL OF INTERACTIVE MARKETING ● VOLUME 15 / NUMBER 1 / WINTER 2001

6

web site (www.gartner.com). The GartnerGroup is the largest provider of research for ITprofessionals at Fortune 500 companies, with1997 revenues in excess of $600 million. Interms of content, the website provides an exten-sive database of research reports and docu-ments related to IT and other similar businessissues. In terms of process, the website offerssophisticated search, storage, retrieval, messag-ing, and personalizing features. Thus, using thesite entails significant complexity in terms ofprocess and content knowledge.

The Gartner Group provided free access toparticipants as part of a collaborative-learningproject for a technology course at a businessschool in the Midwest. Though not mandatory,the course assignments were designed to en-courage use of information on this website, andstudents had free access to the site for an 8-weekperiod. At the beginning of the study, studentswere offered the opportunity to attend a 1-hourtraining session. During this session we variedthe relative emphasis between content- and pro-cess-oriented learning. Next, actual usage of thesite was monitored unobtrusively over an 8-weekperiod.

ExperimentParticipants were 84 MBA students enrolled inthe high-technology course. They were told thatthey had to participate in specialized trainingsessions to derive maximum benefit from thesite. Because attendance at the sessions was rec-ommended but not required, some studentslearned to use the site on their own without

attending a training session. These studentsprovide another benchmark for assessing usage,though they self-selected themselves to receiveno training. We discuss the implications of suchself-selection later.

Those attending the training sessions wererandomly assigned to one of three training ses-sions described in Table 1. Each session lasted atotal of 60 minutes with three parts of equalduration (20 minutes). As shown in the table, a20-minute “background module” that providedan overview of the Gartner Group was commonto all three sessions. This was followed by 40minutes of content-oriented or process-ori-ented learning. The relative time spent on con-tent- versus process-oriented learning was variedin the three cells, and this constituted the ex-perimental manipulation.

In consultation with the Gartner Group, wecreated “basic” and “advanced” modules forboth content-related learning and process-re-lated learning. On the content dimension, thebasic content-oriented module (20 minutes induration) covered the content categories, doc-ument types, and areas of content coverage onthe site. The advanced content-oriented mod-ule (also 20 minutes in duration) supplementedthe basic module and exposed participants toadditional specialized content areas, finer dis-tinctions between different types of documents,and personalization features related to the con-tent offered by the site. On the process dimen-sion, the basic process-learning module (20minutes in duration) covered basic navigationand simple search procedures. The advanced

T A B L E 1Design of Training Sessions for Quasi-experiment

Session Timeline Basic Content, Basic Process High Process, No Content No Process, High Content

Part 1(20 minutes)

Background module Background module Background module

Part 2(20 minutes)

Basic content-oriented learningmodule

Basic process-orientedlearning module

Basic content-orientedlearning module

Part 3(20 minutes)

Basic process-oriented learningmodule

Advanced process-orientedlearning module

Advanced content-oriented learningmodule

L E A R N I N G A N D U S I N G E L E C T R O N I C I N F O R M A T I O N P R O D U C T S A N D S E R V I C E S

JOURNAL OF INTERACTIVE MARKETING ● VOLUME 15 / NUMBER 1 / WINTER 2001

7

process-oriented module (also 20 minutes induration) covered specialized procedures suchas advanced searches using Boolean logic, cus-tomizing document downloads to printers orhard-drive2, and specialized e-mail features forrouting documents. All the training manualswere carefully scrutinized by managers at Gart-ner, and pre-tested with a small group of stu-dents to ensure that the desired effect was beingachieved.3 Note that students who were exposedto both process- and content-oriented learningonly got 20 minutes for each type of learning.Therefore, if any effects of simultaneous learn-ing of process and content occur, then theyprovide a conservative test of our theory. Allsessions were held on the same day. Followingthis session, students were free to access the siteat will.

In addition to these three experimental cells,we also had access to usage data among studentswho chose not to participate in any training.Though not necessary for testing the hypothe-sis, data from this “fourth” cell provides an ad-ditional benchmark.

ResultsFor a total of eight weeks the actual number ofdocuments accessed by each student was re-corded. Data were made available to us for twotime intervals: Weeks 1 through 3 (early period)and Weeks 4 through 8 (later period). By com-bining data from these two periods, a cumula-tive usage index was created. This permitted usto conduct three analyses of usage: the “earlyperiod” (Weeks 1 through 3), the “later period”(Weeks 4 through 8), and cumulative usageover the 8-week period.

The usage rates are shown in Table 2. Aone-way analysis of variance was conducted forall three time periods. The overall model wassignificant for the later time period as well asthe cumulative time period (p , .01), but notfor the initial time period (p 5 .16). This sug-gests that the effects of the initial learning ex-perience do not manifest themselves immedi-ately but are subjected to a time lag.

Cumulative Usage. The cumulative usage foreach condition is plotted in Figure 2. The mostinteresting finding is that usage is highest whenpeople get basic levels of both content and pro-cess training. When participants get only highcontent training but no process training, their

2 The training manuals are available from the authors.3 Participants in the pre-test were not included in the actual study.

T A B L E 2Usage Statistics

Usage (number of documents/person)

Early period(weeks 1–3)

Later period(weeks 4–8)

Cumulative(weeks 1–8)

A. Basic process, basic content 14.00 28.43 42.43B. High process, no content 5.30 10.50 15.80C. No process, high content 4.67 6.73 11.40D. No process, no content (control group) 4.80 4.64 9.44

Mean Usage Rate (documents/week/person)

A. Basic process, basic content 4.67 5.69 5.30B. High process, no content 1.77 2.10 1.98C. No process, high content 1.56 1.35 1.43D. No process, no content (control group) 1.60 .93 1.18

J O U R N A L O F I N T E R A C T I V E M A R K E T I N G

JOURNAL OF INTERACTIVE MARKETING ● VOLUME 15 / NUMBER 1 / WINTER 2001

8

usage is just as low as those getting no trainingat all (11.40 vs. 9.44, p 5 .82). Similarly, whenthey get only high process training but no con-tent training, their usage is not significantlydifferent than those getting no training at all(15.8 vs. 9.44, p 5 .54). Only when participantsgot basic levels of process and content trainingwas their usage significantly higher than the restof the conditions (all p’s , .01). These resultsare consistent with our theorizing that, for a

complex EIPS learning only one dimension isnot enough to enhance usage. Rather, moder-ate levels of both content- and process-orientedlearning may be needed for high usage levels.

Usage Patterns over Time. Does usagechange over time? To answer this question, thedaily usage rate for early and late period wascompared within each cell. Recall that, the earlyperiod includes 3 weeks while the later period

F I G U R E 2Cumulative Usage for Experimental Cells (Weeks 1 to 8 Combined)

F I G U R E 3Usage Rate for Early and Late Period

L E A R N I N G A N D U S I N G E L E C T R O N I C I N F O R M A T I O N P R O D U C T S A N D S E R V I C E S

JOURNAL OF INTERACTIVE MARKETING ● VOLUME 15 / NUMBER 1 / WINTER 2001

9

includes 5 weeks. The daily usage rates areshown in the last part of Table 2 and in Figure3. Within each of the four experimental cells,the daily rate for the early and late period wascompared using a t test for matched sample(Welkowitz, Ewen, & Cohen, 1982). Only thedifference in the Basic Process and Basic Con-tent cell approached significance for the earlyversus late period (t 5 1.69, p 5 .11). None ofthe other differences approached significance,though the pattern of results is instructive. Forthe basic content and basic process group, re-sults show a pattern of increased usage overtime. That is, compared to other groups notonly did this group have higher usage duringthe initial period, but within this group itselfusage during the later period increased. For allthe other groups, usage was low in the initialperiod and stayed low over time. This suggeststhat a well-structured learning experience en-genders higher initial usage. Higher usage, inturn, sets up a positive feedback loop. As con-sumers use the product more, they developmore content and process knowledge, leadingto even higher usage.

DiscussionBy systematically varying the initial learning ex-perience that lasted only 1 hour, significant andpersistent differences in usage levels for a com-plex EIP were obtained. Moderate amounts ofboth process- and content-related learning resultin higher usage than high amounts of only con-tent or only process learning. Further, the ef-fects of superior learning seem to amplify overtime while ineffective learning experiences leadto consistently lower (even decreasing) usageover time.

At this point, it is also interesting to note thatusage was virtually identical among groups re-ceiving only high levels of content or processtraining and no training at all. However, what ifthose not attending any training session aresystematically different than those randomly as-signed to one of the three training sessions? Ifwe assume that those opting out of the trainingare already highly sophisticated users of thisparticular web-site, then we would have ex-pected higher usage among them. However,

their usage is not any higher than those receiv-ing only content or only process training. If weassume that those opting out of training aredisinterested in the website, then we would ex-pect significantly lower usage than those receiv-ing any type of training. Again this is not thecase: their usage is very similar to those receiv-ing only content or only process training.

Finally, it should be emphasized that, thoughconsistent with our theorizing, the observeddata are amenable to alternative explanations.For instance, it could be that increasing thetime spent learning from 20 to 40 minutes doesnot lead to a corresponding increase in learn-ing. That is, there are decreasing returns interms of actual learning as the amount of timespent learning increases linearly. Though mostdescriptive accounts of learning support in-creasing returns to time spent, this explanationneeds investigation.

RESEARCH ISSUESGiven the opportunistic nature of this field ex-periment, several limitations remain. Future re-search should address these limitations. First,though behavioral, the measure of usage onlytapped into one dimension of usage of an EIPS.Future research should use additional measuresof usage such as time spent with the EIPS, thenumber and types of information pieces usedand so forth. Second, cognitive processesthrough which process-oriented and content-oriented learning affect usage should be delin-eated using process-tracing approaches. Third,factors that moderate the relationship betweeninitial learning and usage should be investi-gated. For instance, consumer expertise andfamiliarity with a category may moderate theextent to which initial learning has an impacton usage. Naturally, these issues should be bun-dled with studies that replicate and extendthese results across different EIPS types.

Research is also needed to investigate themoderating impact of consumer learning onthe relationship between marketing mix vari-ables and usage (cf. Hoch & Deighton, 1989).For instance, high levels of consumer usage,and consequent learning of an EIPS may render

J O U R N A L O F I N T E R A C T I V E M A R K E T I N G

JOURNAL OF INTERACTIVE MARKETING ● VOLUME 15 / NUMBER 1 / WINTER 2001

10

consumers impervious to several types of com-munication attempts. Similarly, high levels oflearning-based inertia may lead to lower levelsof price elasticity. These ideas can be used toinvestigate new ways for segmenting EIPS con-sumers. Consumer differences in their ability,aptitude, and amount of learning for EIPScould become powerful segmentation variables.For instance, in predicting usage for a retailwebsite, factors such as preferences for differentproduct categories found on the website maynot be as critical as the amount of content andprocess learning for the website.

STRATEGIC IMPLICATIONSMarketers of EIPS need to manage usage be-cause of its revenue implications. In this paperwe show that the initial learning experience is acritical antecedent of usage. By optimizing theamount of process- and content-oriented learn-ing during this period, firms can maximize us-age for their brand.

However, the extent of process- and content-oriented learning is only one aspect of the ini-tial learning experience. Naturally, a firmshould also manage other aspects of the initiallearning experience. For instance, during theinitial learning experience, consumers samplefrom a variety of attributes of an EIPS. Firms canmanage the set of attributes that consumerslearn during the initial learning experience andthe sequence in which the attributes arelearned. The set of attributes a consumer learnsor is exposed to may be tailored depending onthe consumption goal whereas the sequence inwhich attributes are learned may depend on theconsumer’s prior level of knowledge and exper-tise about the category. For instance, visitors atmySimon.com can answer a series of questionsthat enables the site to identify the key con-sumer goals, segment consumers based onthem, and then expose each segment to thosesets of attributes that are central to achievingthe respective goal. Thus, process segmenta-tion, instead of preference segmentation maybe needed in the context of an EIPS.

As with most categories, customer retentionhas become a key strategic imperative for EIPS.

Part of this emphasis has been because of thehigh cost of customer acquisition and the longtime it takes before an acquired customer be-comes profitable (Mittal & Sawhney, 1998). Ourresults suggest that carefully designed initiallearning experiences can become a key sourceof competitive advantage with respect to cus-tomer retention. If the content and processknowledge acquired during the initial learningexperience are specific to the firm’s brand andcannot be transferred to others, customers arelikely to stay with the firm’s offering. For in-stance, those learning the Graffiti script for us-ing the PalmPilot, are more likely to be loyalusers of that EIPS due to the brand-specificity oflearning. As categories mature, many EIPSbrands may become similar (e.g., common userinterfaces). It is exactly in such situations thatcultivating brand-specific learning can providea competitive advantage in terms of customerretention. In other words, among the varioustypes of knowledge-based assets (Glazer, 1991),brand-specific learning among its customers is akey informational asset that a firm should care-fully cultivate and preserve.

REFERENCESAlba, J.W., & Hutchinson, J.W. (1987). Dimensions of

Consumer Expertise. Journal of Consumer Re-search, 13(March), 411–454.

Anderson, J.R. (1987). Skill Acquisition: Compilationof Weak-Method Problem Solutions. PsychologicalReview, 94, 192–210.

Anderson, J.R. (1996). ACT: A Simple Theory of Com-plex Cognition. American Psychologist, 51(4), 355–365.

Bolton, R.N. (1998). A Dynamic Model of the Durationof the Customer’s Relationship with a ContinuousService Provider: The Role of Satisfaction. Market-ing Science, 17(1), 45–65.

Feldman, L.P., & Hornik, J. (1981). The Use of Time:An Integrated Conceptual Model. Journal of Con-sumer Research, 7(March), 407–417.

Glazer, R. (1991). Marketing in an Information-IntensiveEnvironment: Strategic Implications of Knowledge asan Asset. Journal of Marketing, 55(October), 1–19.

Hoch, S.J., & Deighton, J. (1989). Managing What Con-sumers Learn from Experience. Journal of Market-ing, 52(April), 1–20.

L E A R N I N G A N D U S I N G E L E C T R O N I C I N F O R M A T I O N P R O D U C T S A N D S E R V I C E S

JOURNAL OF INTERACTIVE MARKETING ● VOLUME 15 / NUMBER 1 / WINTER 2001

11

Kolers, P. A. (1976). Reading a Year Later. Journal ofExperimental Psychology: Human Learning andMemory, 2, 554–565.

Leong, S.M., Busch, P.S., & Roedder John, D. (1989).Knowledge Bases and Salesperson Effectiveness: AScript-Theoretic Analysis. Journal of Marketing Re-search, 26(May), 164–178.

Meyer, M.H., & Zack, M.H. (1996). The Design andDevelopment of Information Products. Sloan Man-agement Review, Spring, 43–59.

Mittal, V., & Sawhney, M.S. (1998, October 19). Learn-ing, Lock-in and Loyalty: Managing Customer Learn-ing for Information Products. Financial Times.

Nass, C. (1994). Knowledge or Skills: Which Do Admin-istrators Learn from Experience. Organization Sci-ence, 5(1), February, 38–50.

Rayport, J.F. (1999). The Truth about Internet BusinessModels. Available online: www.strategy.business.com/briefs/99301/.

Schary, P.B. (1971). Consumption and the Problem ofTime. Journal of Marketing, 35(April), 50–55.

Smith, E.R. (1990). Content and Process Specificity inthe Effects of Prior Experiences. In T.K. Srull & R.S.Wyer, Jr. (Eds.), Advances in Social Cognition (Vol.3, pp. 1–59). Hillsdale, NJ: Erlbaum.

Welkowitz, J., Ewen, R.B., & Cohen, J. (1982). Introduc-tory Statistics for the Behavioral Sciences (3rd ed.).Chicago: Harcourt, Brace Jovanovich Press.

West, P., Brown, C.L., & Hoch, S.J. (1996). Consump-tion Vocabulary and Preference Formation. Journalof Consumer Research, 23(2), September, 120–135.

J O U R N A L O F I N T E R A C T I V E M A R K E T I N G

JOURNAL OF INTERACTIVE MARKETING ● VOLUME 15 / NUMBER 1 / WINTER 2001

12