Getting Something for Nothing:
The Impact of a Sample Offer and User Mode on Banner Ad Response
Maureen E. Hupfer
McMaster University
Alex Grey
J. Walter Thompson
Abstract
An Internet banner that advertised a free sample generated higher click-through than a banner ad with information only. The ad attitude and site focus of experiential users were positively affected by the sample offer, but this incentive had a negative effect on goal-directed searchers, who appeared to regard the sample-offer banner as a distraction that interfered with search goals and reduced satisfaction experienced at the host site. Beliefs about bias in the site information were unaffected and goal-directed searchers expressed more favorable return visit intentions than experiential users. Further research is needed to clarify the effects of incentive-offer banners on experiential versus goal-directed users. In particular, the Interactive Advertising Model theorizing concerning goal-directed searchers should be examined more closely.
Introduction
Concern about poor effectiveness characterizes recent commentary regarding Internet banner advertisements. With click-through estimates that vary from .25% to .70%, response rates sit at approximately one tenth of 1996 figures, and are far short of the average direct mail response rate of 1.5 to 2.0% (Drèze and Hussherr 2003; Rothenberg 2001; Sherman and Deighton 2001). Perhaps explained by poor targeting and sheer proliferation, the banner's perceived ineffectiveness prompted many mainstream marketers to drop them from media plans (Kennerdale 2001; Rothenberg 2001). Instead, advertisers gravitated toward the sophistication of rich media ads in more intrusive pop-up and pop-under formats (Edwards, Li, and Lee 2002).
Now, however, the pop-up's continued viability is being questioned. According to a 2003 Forrester Research study conducted in the United States, 64% of Internet users found pop-up ads annoying and 28% reported avoiding sites that hosted them (Murphy 2004). Because of consumer irritation, many popular sites have stopped selling pop-up ads (Musgrove 2004). More seriously, pop-up blocking software is widely available and Microsoft has announced plans to incorporate an automatic blocking feature in its Explorer browser (Murphy 2004; Musgrove 2004).
In reaction to these threats, advertisers are returning to "search advertising" or short text messages placed on search results pages (Hansell 2004). According to the Interactive Advertising Bureau, search advertising accounted for 35% of 2003 Internet advertising revenue, compared with 15% in 2002 (Hansell 2004). This trend points to the importance of understanding how user information-seeking goals might affect response to Internet advertising. Insight is offered by the Interactive Advertising Model (Rodgers and Thorson 2000), with its distinction between serious search and playful browse, but very little published evidence is available to guide advertising strategists.
Advertisers also are using the simple banner format to administer sampling campaigns ("Internet Advertising Success Stories" 1997; Sparta 2000; Whittaker 2001; "Widening the Net" 2000) and some believe that incentive-offer banners will become increasingly important (Gray 2004). Problematically, however, little is known about their effectiveness.
To address our limited understanding on these two fronts, we first review the relevant banner advertising literature from the standpoint of the Interactive Advertising Model, paying particular attention to how ad execution and the user's information acquisition mode appear to affect response. We summarize anecdotal accounts of recent banner sampling campaigns, briefly review academic research concerning sampling in more traditional environments, and identify the need for research examining banner sampling programs in conjunction with user information search goals. Next, we describe the results of exploratory experimental research in which user mode and banner sample offer were manipulated. Finally, we discuss the implications of our findings for current practice and future research.
The Interactive Advertising Model
Describing a continuum that varies from playful browsing to serious information search, the Interactive Advertising Model (IAM) proposed by Rodgers and Thorson (2000) argues that response to Internet advertising depends in part on the user's information acquisition mode. Serious goal-directed information search implies greater cognitive effort being placed on task resolution and less on processing site features such as online ads. The IAM further proposes that search mode and advertising execution interact with user motives. Users engaged in serious information search should be less affected by execution cues such as color or animation but might attend to ads that are relevant to the problem at hand. Cho (2003), for example, found that users who were asked to evaluate a movie or book review site were more likely to click on banners when they were highly involved in the product category and when the banner was congruent with the site content.
Mixed support has been found for the IAM distinction between serious search and playful browse. Li and Bukovac (1999) found no differences between surfers' and information seekers' response to animation, but did report that larger banner ads generated higher click-through rates among surfers than information seekers. Comparing differences in recall and recognition of animated banner ads among users engaged in search versus browsing behavior at a modified Center for Media Research website, Pagendarm and Schaumburg (2001) found that browsers demonstrated superior memory. Consistent with IAM thinking, they reasoned that browsers were more likely to attend to banners while those who were searching suppressed perception or overlooked typical banner placement positions.
"Banner blindness" (Pagendarm and Schaumburg 2001) may occur whenever users are absorbed in their online activities (Zhang 2000). More experienced Internet users also seem to be less susceptible to banners than novices (Dahlen 2001). Eye-tracking evidence suggests that familiarity with Internet advertising practices has contributed to viewers' ability to ignore page areas where banners typically are placed (Drèze and Hussherr 2003).
Even when users are "banner blind", however, they may process banners at a pre-attentive level (Drèze and Hussherr 2003). Banners may exert a positive impact on attitude toward the brand even when consumers do not click on them (Briggs and Hollis 1997). If so, the customary click-through, inquiry, or purchase outcome criteria used by most interactive advertising agencies (Shen 2002) may be inadequate. Consequently, Drèze and Hussherr (2003) urged advertisers to consider additional effectiveness measures, such as awareness, recall, and recognition. Rodgers and Thorson (2000) also advocated a full complement of traditional advertising effectiveness measures, including attitudes, intentions, and behavior. As with other media, it is likely that many banner ad effects are delayed; repeated exposure may be required to build awareness, preference, and conviction. A banner that fails to attract click-through on a given occasion is not necessarily ineffective.
Banners and Incentive Offers
Perhaps understandably for a new medium, far more attention has been devoted to investigating banner execution elements that are specific to the online environment (e.g., Bayles 2002; Brown 2002; Hofacker and Murphy 1998; Yoo, Kim, and Stout 2004; Zhang 2000), than to exploring how banners might work within a larger promotional campaign. We are aware of two important exceptions that have examined banners incorporating promotional information such as price reductions or discounts, gifts or offers of free product, and other incentives such as free shipping and handling.
The first of these analyzed a large database of approximately 1200 ad insertions and found that promotional information interacted with banner placement (Chtourou, Chandon, and Zollinger 2001). Among banners with general rotation or home page placement, the presence of promotional information was associated with a higher click-through rate than that found when banners lacked incentive offers. Interestingly, the inclusion of promotional information in banners at thematic and keyword search results pages, where users would have been more goal-directed, actually reduced click-through - rather than simply having less of a positive effect as suggested by the IAM.
Investigating the effects of incentive-offer banners in conjunction with emotional appeal, Xie et al. (2004) also found that incentive offers improved click-through but that their effect was moderated by the type of emotional appeal. Banners with positive emotional appeals and incentive offers generated higher click-through than those with positive appeals and no incentives. However, the reverse pattern occurred for banners with negative or no emotional appeals. Under these circumstances, ads without incentives produced a higher click-through rate than those which incorporated such offers. Incorporating analysis of over 10,000 banners from a wide variety of products and services, their results appear to generalize across involvement conditions. Because they did not examine banner placement, however, we are unable to speculate about the compatibility of their findings with the IAM.
Apart from these two studies, little published evidence pertains to the effectiveness of banners that offer gifts or free samples. Still, industry accounts indicate the willingness of consumers to complete online surveys to obtain incentives such as samples (Benitez 2001) and further suggest that electronic sampling may be effective for attracting attention, prompting product trial, obtaining personal information, and initiating a dialogue with consumers ("Internet Advertising Success Stories" 1997; Sparta 2000; Whittaker 2001).
Bristol-Meyers Squibb's sample program for Excedrin was one of the first instances of a direct response approach to banner advertising. Run during the tax season of 1997, the campaign placed sample-offer banners on the website of Turbo Tax, a popular income tax software program. They received more than double the expected response and were pleased to find that the online program delivered a younger demographic than that attracted by their television ads ("Internet Advertising Success Stories" 1997). Cosmetic firms, including Lancome, Elizabeth Arden, and Calvin Klein, also have been pleased with results obtained by using banners with sample offers ("Widening the Net" 2000). Since then, improvements to the methods by which electronic coupons can be distributed and redeemed (Blundo, Cimato, and De Bonis 2002) have enhanced firms' ability to incorporate digital vouchers for free products or services as part of an integrated communication plan.
Even though little is known about the effects of sample offers in electronic environments, these programs have been well studied in traditional retail contexts. Accounting for approximately $1.2 billion of the $249 billion spent annually on sales promotions (Fowler 2001), sampling is effective in inducing trial and evaluation. Free samples reduce the brand uncertainty inherent in trying an unfamiliar product and may be even more important for low involvement products (Ghosh, Chakraborty, and Ghosh 1995).
The success of sampling in promoting subsequent purchase behavior can be explained by principles of instrumental or operant conditioning theory. When a trial sample is favorably received, the attendant positive reinforcement can establish a favorable attitude toward the product and improve intentions to purchase as well as actual purchasing behavior (McGuinness, Brennan; and Gendall 1995a, 1995b; McGuinness, Gendall; and Mathew 1992; Motes and Woodside 2001; Nord and Peter 1982; Peter and Nord 1982; Rothschild and Gaidis 1981). Many marketers credit sampling programs with short-term boosts in sales of new products and expect to see continued growth in expenditures on sampling programs (Fowler 2001).
Recent industry surveys provide further support for the importance of sampling in trial generation and purchase. Approximately 72% of the consumers who participated in a recent survey conducted by the Promotion Marketing Association indicated that they purchased at least some of the brands for which they had received samples (Estell 2002). Respondents agreed that sampling was a good risk-free opportunity to try new products, while 89% said they would "feel better" about purchasing a product they had sampled. Most (69%) reported that product samples and demonstrations influenced their behavior more than television or radio ads, and 70% had completed surveys within the last year in order to obtain product samples. A recent National Incentive Survey conducted in Britain reported that 72.8% of respondents felt a free product sample was the promotional offer most likely to persuade them to buy ("Consumers Vote Free Samplings as Best Promotions" 2003). Sampling programs also appear to signal the company's belief in its offering. Consumers may infer that the firm is offering a free sample because they are highly confident that the trial experience will be favorable and lead to later purchase (Wellner 1998).
In addition to identifying the need for further research concerning banner ads with incentive offers such as samples, our review of the literature also indicates the importance of studying banner execution in conjunction with user information-seeking goals. To improve our understanding of how banner incentive offers and user mode might affect more than just click-through, we also must collect dependent variables pertaining to more traditional advertising effectiveness measures. Finally, we need to better determine how banners affect user attention to and perception of the host sites where they are inserted (see Becker-Olsen 2003, Zhang 2000). To address these research questions, we conducted experimental research in which user mode (experiential vs. goal-directed) and banner type (sample vs. no sample offer) were manipulated.
Experimental Research
Based on the foregoing literature review, we formulated hypotheses pertaining to the effects of user mode and a sample-offer banner on both the advertised product (click-through, attitude toward the ad and brand) and user focus on the site hosting the advertisement.
Click-through. The attractiveness of sampling programs among consumers in traditional retail contexts is well-established. Further, practitioner evidence ("Internet Advertising Success Stories" 1997; Sparta 2000; Whittaker 2001; "Widening the Net" 2000) and academic research indicate that web-based sampling can enhance click-through response in electronic environments (Chtourou, Chandon, and Zollinger 2001, Xie et al. 2004).
H1 : A sample-offer banner generates higher click-through than one without a free sample offer.
According to the Interactive Advertising Model (Rodgers and Thorson 2000), experiential users generally pay more attention to website banner ads than those who are goal-directed, and they are more affected by peripheral execution cues.
H2 : Experiential users are even more likely than goal-directed searchers to click on the sample-offer banner.
Attitude toward the Ad and Brand . In traditional retail environments, a sample offer can signal a firm's confidence in its product (Wellner 1998). Also, prior research on banner advertising indicates that banners need not be clicked to have an effect on attitudes (Briggs and Hollis 1997). According to the IAM, experiential users are more affected by banner execution elements than are those who are engaged in more directed search, indicating that user mode moderates the effectiveness of a banner sample offer.
H3a and H4a : We predict a main effect for banner type. That is, attitude toward the ad (H 3a ) and attitude toward the brand (H 4a ) are more favorable when a free sample is offered, whether users click on the ad or not.
H3b and H4b : The main effect for banner type is qualified by a two-way interaction between user mode and banner type. The ad (H 3b ) and brand attitudes (H4 b ) of experiential users are more positively affected by the free sample offer than are the attitudes of goal-directed searchers.
Focus on the Host Site. The free sample offer is expected to attract attention and divert information processing resources from the site where it is inserted. According to IAM, and consistent with "banner blindness" findings, users engaged in goal-directed searching are less affected by execution elements and may suppress their perception of typical banner placement areas. Compared with experiential users, those who are goal-directed will divert less processing capacity to a sample banner offer.
H5a : A main effect for banner type is expected; reported user focus on the host site is reduced when a free sample is offered.
H5b : The type of banner and user mode interact in such a way that the negative effect of a free sample offer on user focus is less severe when users are in a goal-directed mode.
Our 2 x 2 between-subjects design manipulated search task (goal-directed vs. experiential) and banner type (no sample vs. free sample). Undergraduate business student participants were solicited through e-mails sent to faculty distribution lists. The cover message requested feedback on an under-construction student website, provided information about the prize draw incentive (a $100 gift certificate for every 20 participants), and included a link to the study. Because the site was created specifically for the experiment and accessed through the business school server for approximately two weeks, participants could not have had pre-existing attitudes about its usefulness. With its reviews of restaurants, services, and shops in the university neighborhood, we believed the content would be of interest to the sample we studied.
The 115 students who clicked on the link to participate were first directed to an instructional web page which told them that their responses to questions concerning design issues would be applied to produce a better website for students' use. The instructions also assigned participants on a random basis to one of two search conditions. As recommended by Toms (2000), those in the experiential condition were given process instructions while participants in the search condition were assigned objective task goals. Specifically, subjects in the experiential search condition were asked to "Look around the site and just have fun!" while those in the goal-directed condition were asked to look at the reviews for two specific restaurants and decide if they agreed with the site's ratings.
An additional link found at the bottom of the instructional page assigned subjects to a banner condition by directing them to one of two versions of the experimental website. Except for the banner manipulation, the two sites were identical. Upon arrival at the home page, subjects saw a 468 by 60 pixel animated banner advertising the ingredients of the new Hershey Sidekick chocolate bar (Figure 1). The chocolate bar was selected as representative of a low-cost and low-involvement purchase for which sampling would be realistic. In addition to the product's expected appeal for student participants, the chocolate bar ad also was relatively congruent with the website's content. Although we used a real brand, this particular chocolate bar was not yet available in city stores and hence should have not been familiar to participants.
Figure 1. Home Page and Banner Ad (Phase 1)
Figure 2. Home Page and Banner Ad (Phase 2)
Figure 3. Home Page and Banner Ad (Phase 3)
The opening banner configuration was replaced by a second smiling face phase (Figure 2), which in turn gave way to a third and final phase in which the Sidekick brand was identified (Figure 3). After cycling through these three animation phases, the final banner phase appeared in a static form in the same position on each of the site's pages. Our decision to make the static banner visible throughout navigation reflected methodological concerns related to the extremely low click-through rates that characterize real world conditions. Had we chosen a format other than forced exposure, we would have required a sample size far larger than was practical for our exploratory study, in order to avoid the kinds of floor effects that could have made potential differences undetectable. Furthermore, the findings of Cho, Lee, and Tharp (2001) indicate that forced exposure to banners can have a positive rather than negative effect on attitudes. Finally, the widespread availability of blocking software points to the desirability of investigating forced exposure formats that do not interfere with or otherwise interrupt the user's primary online tasks in the manner occasioned by pop-up advertising.
The sample banner condition (Figure 3) encouraged participants to click to receive a free sample while the ad seen by participants in the no-sample condition invited them to click to receive more information. Participants were not obliged to click on the banner, but those who did received a pop-up message. In the no-sample condition, participants were thanked for clicking and informed that the banner was being used to test ad placement. No further information was currently available for the advertised product. Those who clicked in the free sample condition received the same message, but with additional instructions advising them of the campus location and schedule for obtaining their free sample Sidekick chocolate bar.
Participants in both conditions were asked to complete a series of
questionnaire items once they were satisfied that they had finished
looking at the site. To reduce the likelihood of hypothesis guessing,
general questions about Internet usage, the website's layout, and basic
demographics were included along with the dependent variables of
interest. After the experiment website was closed down, all respondents
received an e-mail debriefing about the purpose of the experiment. A
final e-mail was sent after the draw had been conducted and all winners
had been contacted.
Dependent Measures
In addition to recording the experimental condition and banner click-through, we collected dependent measures pertaining to our hypotheses for attitude toward the ad, attitude toward the brand, and perceptions of focus on site. We also included one item that assessed the effectiveness of the user mode manipulation and a multiple-choice recognition question to determine whether participants attended to and processed the ad. For purposes of results interpretation, we measured perceived bias in the site's reviews, satisfaction experienced at the site, intentions to consult the site in the future, and the amount of time per week spent on the Internet.
Attitude toward the Ad. Using five-point scales anchored by strongly disagree and strongly agree, respondents were asked to indicate their agreement with three items tapping attitude toward the banner ad. These included "The advertisement that I saw on the website is: bad/good, useful/not useful, and entertaining/dry". Because principal components analysis demonstrated that these three items formed a single factor accounting for 71.3% of the variance, the individual responses were averaged after reverse-scoring the second and third items. The resulting scale reliability (Cronbach's alpha) was .79.
Attitude toward the Brand. Using the same five-point scale format, the attitude toward the brand items included "I like this brand", "This brand can satisfy my needs", and "This brand is desirable". Forming a single factor that accounted for 81.6% of the variance, the brand attitude responses were averaged to arrive at a scale score with a Cronbach's alpha of .90.
Focus on Site. Participant perceptions of the extent to which they attended to the experimental website were assessed with four of the flow items developed by Hoffman and Novak (1996). Rather than asking participants about their Web experiences in general, they were asked to respond on the basis of their reaction to the specific experimental website. "While viewing the Westdale website, I felt.absorbed intently/not absorbed intently, not deeply engrossed/deeply engrossed, fully concentrated/not fully concentrated, focused attention/not focused attention". These four items formed a single factor that accounted for 67.0% of the variance. After reverse-scoring for absorption, concentration and attention, responses were averaged to arrive at a scale score. Reliability for focus on site was .83.
Manipulation Check. To assess the effectiveness of the user mode instructions, we included the controlling/controlled five-point Hoffman and Novak item (1996). We expected that participants in the goal-directed group, whose experience was controlled or limited by the search task, would score higher on this item than the experiential users who were simply encouraged to have fun. As with the focus on site items, participants were asked to respond with respect to the specific experimental site.
To determine the level of banner ad awareness and processing, participants were asked to identify the chocolate bar brand by choosing one of five multiple-choice responses. In addition to the correct Sidekick answer, we included four other real brands (Fastbreak, Crunch, Almond Joy, and Big Turk).
Satisfaction at Site. This five-point item (Hoffman and Novak 1996) had endpoints of satisfied and unsatisfied and was reverse scored.
Website Bias. This five-point scale, anchored with endpoints of impartial and biased, asked respondents to indicate the extent to which they believed reviews on the site were biased.
Future Site Consultation. Using
a five-point scale anchored by strongly disagree and strongly agree,
participants were asked to indicate their agreement with the following
statement: "I would consult this website before trying a shop in
Westdale"
Results
Manipulation Checks. Analysis of the item measuring respondents' perceived control confirmed that participants in the goal-directed condition felt more controlled while visiting the website than did those in the experiential condition (M goal-directed = 3.30, M experiential = 2.97). The type of banner had no effect on perceived control, nor did it interact with user mode (Table 1). Next, the ad recognition task indicated a high level of awareness and processing, with 77% of respondents correctly identifying Sidekick as the advertised brand. Consequently, we assumed that participants also had noted the sample vs. no-sample manipulation. Logistic regression analysis found that neither the experimental manipulations nor their interaction had any effect on participants' ability to recognize the brand; the chi-square omnibus test was non-significant ( p < .588) as were all individual parameters.
Table 1. ANOVA Results: User Mode Manipulation Check
Click-through. Our logistic regression results for click-through (Table 2) found that only the type of banner affected tendency to click on the ad. As expected (H 1 ), a higher click-through rate was found for the free sample banner manipulation. While 28% of respondents (17 of 60) clicked in the no-sample condition, the click-through rate improved to 51% (28 of 55) when a free sample was offered. Our H 2 predictions were not supported; experiential surfers were no more likely to click on the banner (22/59 or 37%) than were goal-directed searchers (23/56 or 41%).
Table 2. Logistic Regression Results: Click-through
Because we recognized that the pop-up message about banner placement testing might annoy respondents, we included a click-through factor in our preliminary analyses of all remaining dependent measures. By first conducting three-way ANOVAs with banner type (no sample vs. sample), user mode (experiential vs. goal-directed) and click-through (no click vs. click) factors, we could determine whether receipt of the pop-up message interacted with the experimental conditions. These analyses also included a covariate that measured the amount of time spent on the Internet. Measured with a six-point scale, the Internet use variable featured endpoints of over 40 hours a week (1) and one hour or less per week (6). With a mean of 2.95 and a modal response of 3 (between 10 and 20 hours per week), our sample was characterized by frequent Internet use. We obtained non-significant results for this covariate in all preliminary ANCOVAS. Further, no significant interactions were found among click-through and the experimental conditions. Therefore, only the two-way ANOVA results are discussed below.
Attitude toward the Ad : The ad attitude ANOVA revealed a search task by banner type interaction ( p < .037) but no significant main effects (Table 3). Because we found no main effect for ad attitude, H 3a was not supported. We found qualified support for H 3b in that the positive effect of a free sample was stronger among experiential users. However, the crossover form of the two-way interaction was more extreme than the expected fan shape (Figure 4). We had expected a strong positive effect for the sample offer among participants in the experiential condition and did find a more favorable ad attitude among those who were invited to click for a free sample (M sample = 2.96 vs. M no-sample = 2.59). Among goal-directed searchers, we had anticipated a smaller, but nevertheless positive reaction. Instead, the free sample banner had a negative effect on ad attitude (M sample = 2.67 vs. M no-sample = 3.02).
Table 3. ANOVA Results: Advertised Brand
Figure 4. Ad Attitude: Sample by User Mode Interaction
Attitude toward the Brand : Our brand attitude analysis revealed no main effect for the sample offer and a marginal interaction between the sample and user mode conditions ( p < .095). Therefore, H 4a was unsupported and only qualified support was found for H 4b . Similar to the ad attitude results, a cross-over interaction was found (Figure 5). Experiential users reacted favorably to the sample offer (M sample = 2.97 vs. M no-sample = 2.86), but goal-directed users reported poorer brand attitude (M sample = 2.83 vs. M no-sample = 3.18).
Figure 5. Brand Attitude: Sample by User Mode Interaction
Focus on Site . Because the sample manipulation had no main effect on user reports of focus on the host site, H 5a was unconfirmed (Table 4). We also failed to find support for H 5b . We did predict and obtain a two-way interaction between search task and banner type, but as with the ad and brand attitude data, the crossover form was unexpected (Figure 6). Parallel to the ad attitude results, we found that the free sample offer had a positive effect on the perceived focus of experiential participants (M sample = 2.92 vs. M no-sample = 2.55) and a negative effect on the focus of goal-directed searchers (M sample = 2.93 vs. M no-sample = 2.63).
Table 4. ANOVA Results: Host Site
Figure 6. Focus on Site: Sample by User Mode Interaction
We were surprised that the free sample had a greater rather than lesser negative impact on the site focus of goal-directed searchers. When considered in conjunction with the attitude results, this finding suggests that the sample-offer banner posed an unwelcome and ongoing distraction that interfered with their website evaluation task. Our banner did not delay or disrupt the search task, or take over the visual field as a pop-up would have done, but the forced exposure format meant that the ad could not be closed or otherwise removed from the screen. In this sense, the banner could be considered intrusive in a manner analogous to the interruptions posed by pop-up advertising. Research on pop-up advertising demonstrates that intrusiveness is associated with both ad avoidance and irritation (Edwards, Li, and Lee 2002), and that interruptions pose a more severe distraction to searchers with concrete goals, causing them to spend less time on task (Xia and Sudharshan 2002).
Our click-through results demonstrated that goal-directed searchers were just as likely as experiential users to click on the sample offer banner, and the recognition task showed that there were no differences between the two groups in their memory for the advertised brand. Therefore, goal-directed users cannot be said to have avoided the ad altogether. Under conditions that are more representative of current real world contexts, searchers may simply ignore a banner or decline to click on it even if it is noticed, as indicated by the lower promotional banner click-through rates reported by Chtourou, Chandon, and Zollinger (2001). In our experiment, exposure was forced and users were less able to ignore the banner. However, the sample offer's negative impact on the focus of goal-directed users cannot be explained by intrusiveness alone. If this were the case, all goal-directed users would have been affected negatively by the ad's continued visual presence. Instead, the incentive-offer banner seemed to constitute a distraction that an information-only banner did not pose. Perhaps the banner that offered information only was more easily ignored by searchers once it had been seen or clicked. Alternatively, its offer may have been perceived as more consistent with their site evaluation goal. Edwards, Li, and Lee (2001) found that the effect of intrusiveness was reduced when ads were relevant to the site's content, or when they provided information or entertainment value.
In contrast, experiential surfers only had entertainment as a goal, having been instructed to "just have fun" looking around the site. Neither banner manipulation should have interfered with this navigational purpose. Consequently, the free sample banner had a positive effect on ad attitude, as would normally be expected with such an offer, and as usually occurs in traditional retail contexts. The sample offer also may have prompted an "are there any other deals at this site?" frame of inquiry, thus resulting in higher reported focus on the site.
Satisfaction at Site, Website Bias and Future Site Consultation . We made no hypotheses concerning these measures but analyzed the data for further insight into the differing reactions of searchers and surfers (Table 4). Similar to results for the attitude and site focus data, our analysis of satisfaction experienced at the site revealed a cross-over interaction ( p < .036, Figure 7). Experiential users reported higher satisfaction when a sample was offered (M sample = 3.29, M no-sample = 2.81), but the opposite occurred for goal-directed users (M sample = 3.17, M no-sample = 3.48). Our experimental manipulations had no effect on perceptions about the site's impartiality whereas intentions concerning future site consultation were affected only by user mode ( p < .033). Goal-directed searchers, who were given specific directions to investigate restaurant reviews, agreed more strongly than experiential users that they would consult the site before patronizing a neighborhood shop (M goal-directed = 3.38 vs. M experiential = 2.91). It appears, therefore, that the lesser site focus reported by goal-directed users in the sample offer condition was related to a decreased sense of satisfaction with the navigation experience, and this in turn had a negative effect on their ad and brand attitudes. This negative effect did not, however, extend to their beliefs about the impartiality of the host site or their willingness to return to it.
Figure 7. Satisfaction at Site: Sample by User Mode Interaction
Discussion
As often occurs in exploratory studies, our findings require clarification and are subject to certain limitations. By placing only one banner on the website and featuring it throughout navigation, we investigated a forced exposure scenario with a banner format instead of a pop-up execution and were able to detect interaction effects with a relatively small sample. However, the uniformly high level of ad recognition suggests that the absence of competitive clutter, the banner's opening animation and its repeated exposure attracted much more attention than would be the case in current real world navigation. The sample offer did elicit a higher click-through rate than the information-only banner, but neither user mode nor the two-way interaction had an effect. A more naturalistic context may be required before differences that depend on user mode can emerge.
We believe that future research should address these compromises to external validity by incorporating within-subject banner execution manipulations as well as the between-subject search factor. If, for example, all subjects were to see an assortment of banners that were varied according to repeated vs. single exposure, and sample vs. no-sample formats, we could better determine the effects of sample offers on the response of goal-directed versus experiential searchers. The resulting increase in the number of banners viewed by individual subjects would create a navigation context that reflects current practice more closely and would reduce the attention directed toward any one banner. In addition, the use of a variety of product categories and incentive offers other than sampling would improve the generalizability of findings, while collection of consumption and product involvement data would assist interpretation.
Apart from improvements to face validity, additional effort must be directed toward understanding how user mode affects banner ad response, especially among goal-directed searchers. Previous research concerning sample offers indicates that both goal-directed and experiential users should react favorably to a free product incentive and according to the IAM (Rodgers and Thorson 2000) this response should be even more pronounced among experiential users. Basing our hypotheses on the IAM, we expected to see main effects for the sample offer, as well as fan-shaped interactions. Instead, our analyses of attitudes, focus on the site, and satisfaction with the site found a series of cross-over interactions that negated the main effects. The sample offer actually generated a negative response among goal-directed users, rather than one that was simply less positive.
Future experimentation must seek to explain why these cross-over interactions occurred. Importantly, the negative response that we observed among the goal-directed users occurred regardless of whether they clicked or not. Our preliminary three-way analyses that included a click-through factor found no interaction between click-through and either of the experimental conditions. The attitudes, focus on site, and satisfaction of goal-directed users in the free-sample condition who decided to click did not differ from the responses reported by those who refrained from clicking. Similarly, click-through made no difference to the responses of goal-directed users in the information-only condition. Put differently, experiential users who clicked in the information-only condition responded no differently from those who did click; nor did click-through have an effect on the responses of experiential users who were offered the free sample.
We have theorized that the sample-offer ad became an ongoing distraction that interfered with the search task assigned to goal-directed participants. Perhaps those who did not click felt frustrated or thwarted in having foregone this opportunity in favor of completing their search task, while participants who did click may have felt that they were inadequately rewarded for having done so. These two clicking scenarios could have produced equally negative outcomes, but for different reasons. We also have suggested that the sample-offer banner may have prompted experiential users to adopt a "deal-seeking" framework of inquiry that led them to report a higher level of focus on the site. To further examine these interpretations, we need constructs that assess perceptions of banner ad intrusiveness, beliefs about the extent to which a banner interferes with or encourages the user's site navigation goals, and a range of emotional responses that include surprise, pleasure, enjoyment, frustration, and disappointment or the feeling that one has been cheated. In addition, behavioral data such as eye-tracking and navigation logs can determine the points at which attention is captured, when click-through occurs and the amount of time spent on ad versus general site processing.
Next, we recommend a more complete battery of manipulation checks. The flow "controlling/controlled" item allowed us to confirm that participants in the experiential condition felt that they had greater control over their site experience than did the goal-directed users, who were assigned a specific search task. However, questions concerning users' perceived adherence to instructions and objective navigation data would be very useful additions. Similarly, our recognition task established that participants attended to the banner ad to the extent that 77% were able to correctly recall the brand name. With recognition at such a high level, we assumed that participants were well aware of the banner sample vs. no-sample manipulation. Nevertheless, this check could be supplemented with specific items asking whether the experimental ad had offered a free sample or information only.
Although it is acceptable to use student convenience samples for theory testing (Calder, Phillips, and Tybout 1981), we also acknowledge the desirability of conducting additional research among a broader and more representative sample of Internet users than the students we studied. Our sample's age, education, and Internet usage were similar to those of the higher-frequency users identified in a 2001 Statistics Canada report (Dryburgh 2001), but the "typical" or "average" Internet user profile is likely to shift as older consumers continue to move online.
The practical implications of our results concern both advertisers and content providers, and point to the importance of clickstream user tracking (e.g., Moe 2003) to determine whether site visitors are searching or surfing. Anyone using a site search engine, bookmarking pages or using a bookmark to arrive at a site could be classified as a goal-directed user, as would users at keyword results pages. If a user follows a logical path through a site, and spends a reasonable amount of time at each information layer, one might also assume that a more goal-directed search is being conducted. On the other hand, the absence of bookmarking or search engine use, or a less coherent navigational path with quick clicks through successive pages may indicate experiential use. If users can be classified in such a manner, then an appropriate banner can be delivered. Marketers also must consider their communication goals. If click-through is more important than subsequent ad and brand evaluations, our findings indicate that a free sample banner should be employed regardless of user search mode.
Recommendations for advertisers, however, may be at odds with the content provider's goals for focused attention on the host site. Taking advantage of the same kinds of tracking data, content providers may want to pre-test banners on a limited scale before they agree to full launch. This will be especially important for research-oriented sites where providers will be concerned with minimizing impediments to focused attention among goal-directed searchers. While it appears that beliefs about the host site's usefulness are not affected by the presence of sample offers, goal-directed searchers who find themselves distracted and dissatisfied may leave the site before their search task is completed.
Despite its limitations, we believe that this research makes a positive contribution to the existing literature on banner advertising execution effects. The relevance of our website for the student participants, coupled with its entertaining and commercial nature, meant that our experimental context more closely simulated typical online experiences than did the more specialized research-oriented sites selected by Bayles (2002), Li and Bukovac (1999), and Pagendarm and Schaumburg (2001). We are unaware of any other studies that investigate the viability of banner sampling from a user search mode perspective, and further note that very little of the existing Internet advertising research considers either the information-seeking goals of the user or the effects of ad insertions on the host site.
Finally, our findings indicate the need to examine more closely and refine the Interactive Advertising Model, especially where either forced exposure or incentive offerings are concerned. This is particularly true of goal-directed users. While we found that experiential users were positively affected by the free sample offer, a banner that encouraged goal-directed searchers to click for a free sample seemed to create an unwelcome distraction. Whether or not they clicked, the reported focus of goal-directed users was lessened and their evaluations of the ad and brand were less favorable than if only information were offered. If the upward trend in search advertising continues, an improved understanding of the effects of interactive advertising on goal-directed users will become increasingly important.
References
Bayles, Michelle E. (2002), "Designing Online Banner Advertisements: Should We Animate?" CHI Letters , 1 (April), 363-368.
Blundo, Carlo, Stelvio Cimato, and Annalisa De Bonis, A. (2002), "A Lightweight Protocol for the Generation and Distribution of Secure e-coupons," Proceedings of the Eleventh International Conference on World Wide Web, New York : ACM Press.
Briggs, Rex and Hollis, Nigel (1997), "Advertising on the Web: Is There Response Before Click-through?" Journal of Advertising Research , 37 (2), 33-45.
Brown, Mark (2002) "The Use of Banner Advertisements with Pull-down Menus: A Copy Testing Approach," Journal of Interactive Advertising ," 2 (Spring), <http://jiad.org/vol2/no2/brown> (accessed on 10/1/2003).
Cho, Chang-Hoan (2003), "Factors Influencing the Clicking of Banner Ads on the WWW," CyberPsychology and Behavior , 6 (2), 201-215.
Chtourou, Mohammed Saber, Jean Louis Chandon, and Monique Zollinger (2001), "Effect of Price Information and Promotion on Click-through Rates for Internet Banners," Journal of Euro-Marketing , 11 (2), 23-40.
"Consumers Vote Free Samplings as Best Promotions," (2003, September 4), Marketing , 4.
Dahlen, Micael (2001), "Banner Advertisements Through a New Lens," Journal of Advertising Research , 41 (July/August), 23-30.
Drèze, Xavier and Francois-Xavier Hussherr, (2003) "Internet Advertising: Is Anybody Watching?" Journal of Interactive Marketing , 17 (4), 8.
Dryburgh, Heather (2001), "Changing Our Ways: Why and How Canadians Use the Internet," Statistics Canada , Catalogue No. 56F0006XIE, 1-14, <http://statcan.ca/english/IPS/Data/56F0006XIE> (accessed 10/1/2003).
Edwards, Steven M.., Hairong Li, and Joo-Hyun Lee (2002), "Forced Exposure and Psychological Reactance: Antecedents and Consequences of the Perceived Intrusiveness of Pop-up Ads", Journal of Advertising , 31 (3), 83-95.
Estell, Libby (2002), "Sampling Sells: A Marketing Strategy that Motivates Consumers," Incentive , 176 (August), 6.
Fowler, Geoffrey A. (2001), "When Free Samples Become Saviors - Point-of-Sweat, Point-of-Thirst Marketing Offers Freebies When they are Most Needed," The Wall Street Journal Eastern Edition, (August 14), B1.
Ghosh, Amit K., Goutam Chakraborty and Debra Bunch Ghosh (1995), "Improving Brand Performance by Altering Consumers' Brand Uncertainty," Journal of Product and Brand Management , 4 (5), 14-20.
Hofacker, Charles F. and Jamie Murphy (1998), "World Wide Web Banner Advertisement Copy Testing," European Journal of Marketing , 32 (7/8), 703-712.
Hoffman, Donna L. and Thomas P. Novak (1996), "Marketing in Hypermedia Computer-mediated Environments: Conceptual Foundations," Journal of Marketing , 60 (3), 50-68.
"Internet Advertising Success Stories," (1997, May 5), Mediaweek , 52-58.
Kennerdale, Caspar (2001), "Banner Advertising: Still Alive, but Fundamentally Flawed?" Econtent , 24 (April), 56-57.
Li, Hairong and Janice L. Bukovac (1999), "Cognitive Impact of Banner Ad Characteristics: An Experimental Study," Journalism and Mass Communication Quarterly , 76 (Summer), 341-353.
McGuinness, Dalton , Mike Brennan, and Philip Gendall (1995a), "The Effect of Product Sampling and Couponing on Purchase Behaviour: Some Empirical Evidence," International Journal of Advertising , 14 (3), 219-230.
---, ---, and --- (1995b), "An Empirical Test of Product Sampling and Couponing," Journal of the Market Research Society , 37 (April), 159-170.
McGuinness, Dalton , Philip Gendall, and Stephen Mathew (1992), "The Effect of Product Sampling on Product Trial, Purchase and Conversion," International Journal of Advertising , 11 (1), 83-92.
Motes, William H. and Arch G. Woodside (2001), "Purchase Experiments of Extra-ordinary and Regular Influence Strategies Using Artificial and Real Brands," Journal of Business Research , 53 (1), 15-35.
Nord, Walter R. and J. Paul Peter (1980), "A Behavior Modification Perspective on Marketing," Journal of Marketing , 44 (2), 36-47.
Pagendarm, Magnus and Heike Schaumburg (2001), "Why Are Users Banner-Blind? The Impact of Navigation Style on the Perception of Web Banners," Journal of Digital Information , 2 (1), <http://jodi.ecs.soton.ac.uk/Articles/v02/i01/Pagendarm> (accessed on 10/1/2003).
Peter, J. Paul and Walter R. Nord (1982), "A Clarification and Extension of Operant Conditioning Principles in Marketing," Journal of Marketing , 46 (3), 102-107.
Rodgers, Shelley and Esther Thorson (2000), "The Interactive Advertising Model: How Users Perceive and Process Online Ads," Journal of Interactive Advertising , 1 (Fall), <http://jiad.org/vol1/no1/rodgers > (accessed on 10/1/2003).
Rothenberg, Randall (2001), "Famed Disasters: Hindenburg, Chicago Fire and Banner Ads," Advertising Age , 72 (March 5), 19.
Rothschild, Michael L. and William C. Gaidis (1981), "Behavioral Learning Theory: Its Relevance to Marketing and Promotions," Journal of Marketing , 45 (Spring), 70-78.
Shen, Fuyuan (2002), "Banner Advertisement Pricing, Measurement, and Pretesting Practices: Perspectives from Interactive Agencies," Journal of Advertising , 31 (Fall), 59-67.
Sherman, Lee and John Deighton (2001), "Banner Advertising: Measuring Effectiveness and Optimizing Placement," Journal of Interactive Marketing , 15 (Spring), 60-64.
Toms, Elaine (2000), "Understanding and Facilitating the Browsing of Electronic Text," International Journal of Human-Computer Studies , 52 (March): 423-452.
Wellner, Alison (1998), "Try It - You'll Like It!," American Demographics , 20 (August), 42-43.
Whittaker, Stephanie (2001), "Schick's Cutting-edge Mini-site," Digital Marketing , 2 (March), 5.
Zhang, Ping (2000), "The Effects of Animation on Information Seeking Performance on the World Wide Web: Securing Attention or Interfering with Primary Tasks?" Journal of the Association for Information Systems , 1 (March), 1-28.
About the Authors
Maureen Hupfer (Ph.D., University of Alberta ) is an assistant professor of Marketing in the Strategic Market Leadership and Health Services Management area at the DeGroote School of Business, McMaster University , Hamilton , Ontario . Her research interests include gender and information processing in both online and offline environments. Please address all correspondence regarding this manuscript to Maureen Hupfer (hupferm@mcmaster.ca).
Alex Grey ( BCom McMaster University ) is employed at J. Walter Thompson, Toronto , Ontario.
