Computers in Human Behavior 55 (2016) 867e876
Contents lists available at ScienceDirect
Computers in Human Behavior journal homepage: www.elsevier.com/locate/comphumbeh
Full length article
Tracking users' visual attention and responses to personalized advertising based on task cognitive demand Hyejin Bang a, *, Bartosz W. Wojdynski b a b
Department of Advertising and Public Relations, Grady College of Journalism and Mass Communication, The University of Georgia, GA, 30602, USA Grady College of Journalism and Mass Communication, The University of Georgia Athens, GA, 30602, USA
a r t i c l e i n f o
a b s t r a c t
Article history: Received 18 August 2015 Received in revised form 16 October 2015 Accepted 21 October 2015 Available online xxx
This study examined the effects of personalization in banner advertising on visual attention to the advertisement. A 2 (ad type: personalized vs. non-personalized) 2 (task cognitive demand: high vs. low) eye-tracking experiment (N ¼ 93) was conducted to examine how personally salient information attracts consumers' attention, and how it interacts with different levels of cognitive load for given tasks. Consistent with previous literature, participants paid relatively longer and more attention to the personalized compared to non-personalized advertisements. However, task cognitive demand was shown to moderate the effects of personalization on attention, such that the personalized advertisement was much more effective in attracting consumers' attention than the non-personalized advertisement when people were engaged in a highly cognitively demanding task. No significant interactions between personalization and cognitive demand of task were found on perceived goal impediment and attitude toward the advertisement. Implications and suggestions for future research are provided. © 2015 Elsevier Ltd. All rights reserved.
Keywords: Personalized advertising Attention Eye-tracking Cognitive load
1. Introduction Consider the following scenario: 29-year-old Sara is due to give birth in 3 months, so she and her husband have started to shop online for baby paraphernalia. In the past week, they've visited eretailers like Amazon.com or Babies “R” Us to look for car seats, strollers, and other baby gear. While browsing the Web, Sara noticed that ads for baby gear started following her across unrelated sites. Ads appear alongside her daily news articles, ads with her friends' names appear in her Facebook news feed, and she even receives email and text messages with promotional coupons. Although she'll usually at least glance at most of the ads to see if they're promoting a must-get product for her baby, she's starting to think that it's a bit creepy that all of the sites somehow seem to know she's pregnant. Experiences like Sara's are rapidly becoming the norm for a large majority of Web users. The development of Web-based behavior-tracking and database technology enables marketers to
* Corresponding author. E-mail addresses:
[email protected] (B.W. Wojdynski).
(H.
http://dx.doi.org/10.1016/j.chb.2015.10.025 0747-5632/© 2015 Elsevier Ltd. All rights reserved.
Bang),
[email protected]
tailor advertising based on consumers' interest, preference and needs (Pavlou & Stewart, 2000). Software now keeps track of a variety of users' browsing behavior, such as their past product selection, favorite celebrity, address, phone number, name, zip code and etc. This data is gathered either covertly (e.g., storing digital “cookies” on users' devices) or overtly (asking or requiring users to submit information), but in both cases the data collected can be incorporated into strategies by marketers or advertisers (Sundar & Marathe, 2010). Therefore, from the marketers' perspective, personalized advertising increases brands' ability to provide more accurate targeting, and from the consumers' perspective, it also increases message relevance or involvement to consumers (Tucker, 2011). Recent industry research (e.g., Internet Retailer, 2013) shows that the prevalence of personalized advertising continues to grow. According to the research, more than half of e-retailers provide product recommendations or web page personalization using digital “cookies” (Internet Retailer, 2013). According to the U.S. Federal Trade Commission (FTC), more than 90% of online websites store users' personal information to use it for marketing purposes (US FTC, 2000). In spite of the prevalence of personalized advertising, our theoretical understanding of the impact of personalization on
868
H. Bang, B.W. Wojdynski / Computers in Human Behavior 55 (2016) 867e876
consumer decision-making is still in its nascent stages. This may be due in part to the difficulty of tailoring advertising based on participant's personal information in experiment setting. Most research on personalized advertising has relied on survey data (e.g., Baek & Morimoto, 2012; Nyheim, Xu, Zhang, & Mattila, 2015; Xu, 2006; Yu & Cude, 2009), not experimental analysis. As a result, some of the controversy over how personalization may affect users responses to ads continues (Pavlou & Stewart, 2000; Phelps, D'Souza, & Nowak, 2001). Therefore, to answer the need for more empirical research regarding the situational impact of personalized advertising on consumers behavior and perceptions, this study investigated how personalized advertising alongside news article may work to attract readers' attention, and whether the effects of personalization on attention, recall and attitudes may vary based on the cognitive effort required by the task the user is trying to accomplish on the site. To more accurately gauge how personalized advertising affects consumers' attention on a page, participants' visual attention to on-screen stimuli was measured using eyetracking equipment. 2. Literature review 2.1. The rise and evolution of personalized advertising Personalized advertising can be defined as advertising that incorporates information about the individual, such as demographic information, personally identifying information (e.g., name, residence, and job) and shopping-related information (e.g., purchase habit or history and brand preference; Wolin & Korgaonkar, 2005; Yuan & Tsao, 2003; Yu & Cude, 2009). Unsolicited commercial email, postal direct mail, telemarketing, and text messaging can be all considered as forms of personalized advertising (Baek & Morimoto, 2012), although recent interest in the phenomenon has focused primarily on mobile and Web display advertising tailored and served to users based on their identity and behavior (e.g., Aguirre, Mahr, Grewal, de Ruyter, & Wetzels, 2015), a practice also known as online behavioral advertising (Smit, Van Noort, & Voorveld, 2014). Overall, personalization refers to tailoring of message content and delivery based on data collection or covert observation of users, to increase the personal relevance of message. This is often treated a distinct concept from “customization,” which involves users actively selecting or inputting information and receiving tailored content in response, thus enabling users to perform an active role in receiving information (Sundar & Marathe, 2010). Personalized advertising, once thought of as a promising attention-getting tactic, is no longer a new development. However, the processes by which user information is recorded and used to generate personalized ad content continue to evolve. Due to the advancement of tracking and database technology (e.g., digital cookies), advertisers are now equipped with power to tailor ad messages at individual level depending on consumers' interests and needs (Pavlou & Stewart, 2000), and their capabilities are often bolstered by cross-platform information sharing between various sites and applications (Finley, 2015). Despite the prevalence of personalized or customized messages (Kalyanaraman & Sundar, 2006; Poon & Jevons, 1997), evidence of the effects of personalization on advertising-related outcomes has been mixed (Yu & Cude, 2009). Some researchers have found that personalized messages (e.g., advertising) attract users' attention (Malheiros, Jennett, Patel, Brostoff, & Sasse, 2012; Tam & Ho, 2005) and increase message receivers' attitude toward the message or even toward the medium (Kreuter & Wray, 2003; Pavlou & Stewart,
2000) because of its perceived relevance to the self (Lang, 2006; Petty, Barden, & Wheeler, 2002). For instance, Kalyanaraman and Sundar (2006) found that the level of customization of Web site content led to more positive user attitudes toward the portal mediated by perceived relevance, novelty and involvement. Recent research has also shown that a consumer's level of perceived personalization of a message can be a far better predictor of positive attitude effects than whether the message was actually personalized (Li, 2016). Similar effects have been found for personalization in advertising (Baek & Morimoto, 2012; Howard & Kerin, 2004; Pavlou & Stewart, 2000). Specifically, Howard and Kerin (2004) discovered that consumers' ad responsiveness could be enhanced by personalization; when an ad contained a viewer's first name, the viewer was likely to have higher purchase intention for the product recommended in the ad. Tucker (2014) found that these effects can go beyond intention to behavioral responses, with personalized ads yielding a higher click-through rate if consumers have power to control the privacy setting on the webpage. While personalized advertising may have broad overall benefits to advertisers, its success with individual consumers may be moderated by other factors, such as the extent to which the mechanism of personalization raises privacy concerns among message recipients (Phelps et al., 2001; Sacirbey, 2000). White, Zahay, Thorbjørnsen and Shavitt (2008) suggested that the level of personalization, the presence of justification for personalization and the perceived utility of the message could all be important factors in determining consumers' reactance to personalization. Several studies empirically proved that consumers generally have a negative perception towards personalized advertising across media types, which may increase as privacy issues are made more salient (Sheehan & Hoy, 1999), or if the personalization is based on behavioral tracking (Turow et al., 2009). Recently, Baek and Morimoto (2012) found that individuals with high enduring levels of privacy concerns are likely to avoid personalized ads, mediated by skepticism toward the personalized advertising. These negative effects might offset the positive effects of personalized ad (Phelps et al., 2001; Sacirbey, 2000; Van Doorn & Hoekstra, 2013). 2.2. Attentional salience of personalized information In any Web use experience, consumers are bombarded with many simultaneous calls for attention, yet they are limited in their capacity for processing information. Biased competition theory (Desimone & Duncan, 1995), which argues that information in visual fields competes for cognitive processing, is highly relevant to the often-cluttered online media environment, where media context and advertisements compete for consumers' attention. According to industry studies, the amount of attention paid to online advertising is much lower than that to advertising in other media, such as TV, radio and magazine (AdNews, 2013). Therefore, what determines or controls consumers' visual processing of information in the competitive media environment is an important issue for advertisers. Ad personalization may play a role in attracting consumers' attention to advertisements that they might otherwise miss. According to previous research, it has been found that the sound or sight of a person's own name can attract one's attention, even when the name is embedded in sets of other information (Harris & Pashler, 2004; Mack & Rock, 1998; Moray, 1959; Wolford & Morrison, 1980). Moray (1959) first confirmed the attentional salience of person's own name in an experiment in which participants were asked to shadow information played in their left ear,
H. Bang, B.W. Wojdynski / Computers in Human Behavior 55 (2016) 867e876
while ignoring information played in the right ear. When the information played in the right ear only contained ordinary words, participants were not able to notice what it was. However, when the participant's own name was presented in the unattended ear, about one third of participants paid attention to and noticed it. This finding has been extended to the visual field through replications by other researchers (Harris & Pashler, 2004; Wolford & Morrison, 1980). According to their studies, when subjects' own names were presented in the middle of the digit-parity task, the response time for the task was significantly slower. This is because personally salient information (e.g., person's own name) is always meaningful to him or her and becomes emotionally salient. Similarly, research on participants' responses to media online content shows effects of personalization on attention, attitude and even behavioral outcomes. Tam and Ho (2005) found that web personalization (e.g., product recommendation based on users' preference) not only encourages consumers to pay attention to or engage with content, but also behave in desired way (e.g., purchase). They found that personalized information increases attention and the elaboration levels on the given contents. However, the dimension of attention was measured based on the order in which participants clicked on areas of the page, but not by measuring the duration or extent that participants actually paid attention to the message. In the context of advertising, Malheiros et al. (2012) similarly revealed that people are more likely to notice an advertisement featuring a high degree of personalization, such as consumers' own photographs, even though people may feel uncomfortable with the ad itself. This increased attention caused by personalization may translate to behavior, as well; a recent study found that personalized banner ads are clicked much more often than generic banner ads (Bragge, Sunikka, & Kallio, 2013). However, the majority of these studies measured attention based on participant self-reports after exposure to the content or other behavioral indicative of attention (e.g., click-rate; Bragge et al., 2013; Tam & Ho, 2005). There has been a lack of research examining whether increased attention to the ad during exposure is a mechanism by which personalization enables content producers to get through to participants. Only one published study to date (Malheiros et al., 2012) measured visual attention toward personalized advertising during exposure to the ad itself. The authors found participants looked twice as long at advertisements that included their photograph than a page containing only their name. In this study, we seek further extend the result of previous studies by hypothesizing that personally more salient information, in the form of a personalized ad, is more likely to attract consumers' attention than a non-personalized ad not only in terms of absolute visual attention, but also time to first fixation and frequency of attention. H1. For readers of an online news article, a personalized advertisement will attract a) earlier, b) greater, and c) more frequent visual attention than a non-personalized advertisement. 2.3. Cognitive demand of task According to cognitive load theory (Sweller, 1988; Sweller, Van €nboer & Paas 1998), people have finite cognitive resources Merrie at a given time to allocate to encoding, processing or retrieving information. As a result, when a primary task demands a high level of processing (e.g., performing a search-task), only a small amount of attentional capacity is available to process secondary information (e.g., advertisements). In this vein, consumers' acceptance of
869
advertising may vary on the basis of the consumer's cognitive capacity available to process advertising (Gilbert, Pelham, & Krull 1988). The amount of cognitive capacity allotted to a task may vary depending on the difficulty of task (Gwizdka, 2010). For example, a fact-finding (FF) task, which involves locating specific information (e.g., name of a specific person, or specific data) needs fewer cognitive resources than an information-gathering (IG) task, which involves the collection of multiple information for specific topics. That is, people involves in a FF task are more likely to pay attention to and process secondary information (e.g., advertising) compared to those who involves in IG task. In real-world Internet-use situations, Rodgers and Thorson (2000) argued that consumers' processing of online advertising may depend on consumers' Internet motive. This is because participants' goals for use of Web-based media may determine their level of cognitive effort devoted to the task. For example, information-seekers are more focused on the task at hand than entertainment seekers in an online environment (Hoffman & Novak, 1996; Wise, Kim, & Kim, 2009). Similarly, people who freely navigate web pages have greater cognitive capacity available for non-target information (e.g., advertisements) than those who seek specific information. Therefore, it is expected that people involved in a low-cognitivedemand task will pay attention to peripheral ads regardless of their personalization, due to the lack of demand on their cognitive resources. On the other hand, users engaged in a high-cognitivedemand task are less likely to be distracted by non-task-related or non-personally-relevant content. However, even under high levels of task cognitive demand, the attentional salience of the personalized advertising is expected to demand more attention than the non-personalized ad. Therefore, for readers of an online news article, task cognitive demand will interact with advertising personalization in influencing visual attention. H2. When people engage in a high cognitive-demand task, personalized advertising will attract a) earlier, b) greater, and c) more frequent visual attention compared to non-personalized advertising H3. When people engage in a low cognitive-demand task, there will be no significant difference between personalized advertising and non-personalized advertising regarding a) first time to fixation on the ad, b) total fixation on the ad, and c) frequency of attention.
2.4. Task interference of personalized ads The relationship between personalization and consumers' attitudinal responses to message contents has been shown to vary situationally (Yu & Cude, 2009). According to Petty et al. (2002), because customized or personalized messages may induce greater elaboration of a message, they necessarily lead consumers to invest a greater amount of effort. This has been illustrated in research that showed that customization of a Web site increases users' perceived involvement with the web page, and engenders greater scrutiny of the contents (Kalyanaraman & Sundar, 2006). Ironically, however, the high level of elaboration or attention grabbing of message led by personalization could elicit negative consequences if the message is not directly associated with users' primary goal or task (Wolford & Morrison, 1980). This is because the attention drawn by the personalized message takes some of cognitive resources necessary for users' primary task. For instance, in a study by Wolford and Morrison (1980), participants were asked to make a
870
H. Bang, B.W. Wojdynski / Computers in Human Behavior 55 (2016) 867e876
quick judgment whether the parity of two digits displayed on the screen matched, while disregarding any irrelevant object. Their experiment found that if a participant's name was presented in the middle of the two digits, it greatly reduced the speed of the participant's judgment. In other words, the personally salient information interrupted the task performance by attracting attention away from the task. In the context of online reading, advertising has shown to interrupt task performance in similar ways. Animated advertisements have been shown to successfully attract users' attention, but this attention-grabbing nature disrupts users' reading, and can lead to negative reactions to the advertisement (Simola, Kivikangas, Kuisma, & Krause, 2013). These results support the notion that advertising characteristics that increase situational salience of the ad increase the odds of that advertisement impeding users' primary task on a Web page. This task interruption can lead to perceived goal impediment (Cho & Cheon, 2004) on the part of the user, which has been shown to lead to more negative responses toward the ad (Cho & Cheon, 2004; Edwards, Li, & Lee, 2002). However, in the Simola et al. (2013) study, the presence of a salient ad only affected participants' task performance when they were engaged in a cognitively demanding task (i.e., reading task). This result indicates that any features of an advertisement that grab the user's attention – as personal information has been shown to do e could lead to perceived goal impediment, since some available cognitive resources for the primary task are allocated to attending to the advertisements. Therefore, the following hypothesis is put forth: H4. For readers of an online news article, task cognitive demand will interact with advertising personalization in influencing perceived goal impediment, so that for participants under high cognitive-demand task, personalized advertising will increase perceived goal impediment compared to non-personalized advertising, while there will be no such difference in low cognitive demand conditions. Additionally, perceived goal impediment caused by advertising has been found to lead to reactance to advertising, including more negative attitudes toward the ad or even ad avoidance (Lee, Kim, & Sundar, 2015; Li, Edwards, & Lee, 2002). Thus, it is postulated that personalized advertising will elicit negative attitude toward the ad due to the increased level of perceived goal impediment when the user encounters the ad in the process of a task that is more cognitively demanding. However, when people engage in a low cognitive-demand task, people are expected to have more favorable attitude toward the personalized ad because of its personal relevance. H5a. When people are engaged in a high-cognitive-demand task, people are less likely to have positive attitudes toward the personalized ad compared to the non-personalized ad. H5b. When people are engaged in a low cognitive-demand task, people are more likely to have positive attitudes toward the personalized ad compared to the non-personalized ad.
3. Method The major objective of this study was to examine consumers' attention toward and response to advertising personalization under two levels of task cognitive demand. To examine these relationships, the study utilized a 2 (ad type: personalized vs. nonpersonalized) 2 (task cognitive demand: high vs. low) factorial
between-subject experiment. The experiment was performed in a research lab, which contained a desktop computer with a devicemounted Tobii X2-60 eye tracker and a separate station for the researcher. 3.1. Participants A total of 101 undergraduate student research participants were recruited through a college-wide research participation pool at a public Southeastern University in exchange for extra credit. Among them, 5 participants were excluded from analysis due to unsuccessful eye-tracking calibration (missing data) and 3 extreme outliers were also excluded. Therefore, a total of 93 responses were analyzed in the study. The remaining 93 participants, participants were between 18 and 26 years in age (M ¼ 20.15), and 69.9% of participants were female. Participants all had normal or correctedto-normal vision. 3.2. Procedure Each participant signed up for a 30-min lab session. Once each participant arrived at the lab, researchers explained the overall procedure of the experiment and informed participants that a device-mounted eye tracker would unobtrusively record their eye movement. After participants provided informed consent, the calibration process for eye tracking was conducted. The researcher then loaded a link to the Web page, containing a single article and a banner ad. Then, participants read through an article at their own speed while performing the task they were assigned to. After viewing the article, participants completed a digital questionnaire containing advertising evaluations, demographics, and coffee consumption habits. 3.3. Stimulus materials For the experiment, an online news story titled “Poor Sleep and Sleep Habits in Adolescence May Raise Health Risk” was created on the basis of content from several published news articles. To engage participants in the reading task, a topic that is somewhat relevant to young adults was selected. Additionally, two versions of a fullcolor banner advertisement for a coffee shop were created to appear alongside the article. A fictitious brand (“Blenz Coffee”) was used in the advertisement to eliminate any confounding effect from participants' previous experiences or exposure to the brand. The chosen article and one of the created display ads (personalized or non-personalized ad) were placed together in a fictitious webpage titled “HealthNewsHQ.” 3.4. Manipulation The Level of personalization was manipulated by presence or absence of one's name and geographic information in the advertisement. Specifically, in the personalized advertising condition, each participant's full name, the name of college they attend, and the information about nearest location of the coffee shop were included in the advertisement (e.g., Ms. Sarah Jones, You are invited to Blenz Coffee's opening. Don't miss our exclusive offer for [University name] students. Your nearest location is…). In the nonpersonalized ad condition, viewers received a more generic message (“Always perfectly roasted, ground and brewed. Discover the Blenz coffee you'll love”). To manipulate task difficulty, subjects assigned to the high
H. Bang, B.W. Wojdynski / Computers in Human Behavior 55 (2016) 867e876
cognitive-demand task were assigned to a fact-finding task. In this high-cognitive demand task, participants were asked to find the three strong argument of the news article and to summarize the overall article. On the other hand, those assigned to the low cognitive-demand task were assigned to the free-viewing task, and only asked to provide any thoughts or feeling they have toward the news article (Gwizdka, 2010). 3.5. Measures 3.5.1. Absolute visual attention paid to ad Absolute attention to ad was measured as the total amount of time in which the user recorded fixations in the area of the story page taken up by the advertisement (plus an additional 10-pixel margin on each side of the ad to account for variations in calibration). Visual attention to the advertisement was measured using data collected via eye-tracking software. The eye tracker recorded participants' pupil fixations at a sampling rate of 60 Hz, or 60 samples per second. Time spent viewing the advertisement was recorded based on these samples, and samples were counted if they were within a region of interest containing the ads with a 10-pixel boundary on each side of the ad (1.14 inch (width) x 2.2 inch (height)) to account for peripheral attention issues (see Boerman, van Reijmersdal, & Neijens, 2015) (M ¼ 4.63, SD ¼ 6.17). 3.5.2. Time to first fixation on the ad The amount of time that elapsed between the loading of the news article and participants' first fixation within the area denoted around the ad was measured by the eye tracking software (M ¼ 55.62, SD ¼ 39.84). 3.5.3. Frequency of attention to the ad The frequency of attention on the ad was measured as the total number of fixations within the designated area around the ad (M ¼ 9.81, SD ¼ 8.65). 3.5.4. Perceived goal impediment (PGI) PGI was measured with seven 7-point likert scales including distracting, disturbing, intrusive, forced, interfering, invasive and obtrusive (e.g., The ad appeared on the webpage was distracting to the readings on the webpage; Edwards et al., 2002; Cho & Cheon, 2004). To check the internal consistency of the scale, Cronbach's alpha (a) was calculated and the PGI scale was reliable (a ¼ .92). The seven items were averaged to form a single measure of PGI (M ¼ 2.80, SD ¼ 1.29). 3.5.5. Attitude toward the ad (Aad) Aad was measured with three 7-point semantic differential items asking participants to rate their perceptions between the anchors of Bad-Good, Unfavorable- Favorable, Pleasant-Unpleasant (e.g., I consider the ad to be bad/good, MacKenzie & Lutz, 1989). The items were found to be internally consistent (a ¼ .77), and were averaged to form a single measure of attitude toward the ad (M ¼ 4.29, SD ¼ .87). 4. Results 4.1. Manipulation check To confirm that the manipulation for cognitive capacity required for tasks was successful, an independent samples t-test was conducted. The level of cognitive capacity required for the reading was
871
measured with 7-point likert scales with two items of “How difficult was it to find out the three strongest arguments and summarize the overall article (How difficult was it to read an article and tell us your opinion” in the low-cognitive demand condition) and “Please rate the level of mental effort you invested on the given task (1: Absolutely no effort; 7: Extreme effort).” Findings from a independent sample t-test revealed that participants assigned to high cognitive capacity condition devoted much more mental effort to perform the given task (Mhigh ¼ 3.97) than those of who assigned to low cognitive capacity condition (Mlow ¼ 2.98) (t ¼ 4.71, p < .001). Therefore, the cognitive capacity was successfully manipulated in the study. 4.2. Testing hypothesis 1 To test H1, eye-movement data were analyzed using independent t-tests. Results from series of t-tests revealed that personalization of ad significantly affects time to absolute visual attention paid to the ad (t ¼ 2.69, p < .001) and the number of visits on the ad (t ¼ 2.37, p < .05). Specifically, the total amount of time spent on the personalized ad (M ¼ 5.98) was much longer than the time spent on the non-personalized ad (M ¼ 2.91). Further, the frequency of visits on the personalized ad (M ¼ 11.56) was greater than the number of visits on the non-personalized ad (M ¼ 7.59). Although the effect of personalization did not have significant effect on the time to first fixation on the ad (Mpersonalized ad ¼ 50.64 vs. Mnon-personalized ad ¼ 61.95, t ¼ 1.37, p ¼ .18), the direction of the difference was consistent as we expected. Therefore, H1b and H1c were supported while H1a was rejected. 4.3. Testing hypothesis 2 and 3 H2 and H3 predicted that the level of attention to personalized ad varies across the level of task difficulty. To test H2 and H3 the three eye-tracking variables were analyzed using two-way ANOVAs. The result of series of two-way ANOVAs found significant interactions effect between personalization and cognitive demand on absolute visual attention, F (1,89) ¼ 5.23, p < .05, and on the frequency of fixation on the ad, F (1,89) ¼ 3.94, p ¼ .05. However, there was no significant interaction effect on time to first fixation on the ad, F (1,89) ¼ 1.99, p > .05 (See Table 1). To determine the nature of the significant interaction effects, planned contrasts were conducted on absolute visual attention and time to first fixation. The results indicated that when people are engaged in the high cognitive demanding task, people were likely to spend much longer time on the personalized ad (M ¼ 8.01) compared to the non-personalized ad (M ¼ 2.47) (F (1, 89) ¼ 11.64, p < .01) (See Appendix for visual heat maps). Participants also showed more frequent allocation of attention on personalized ad compared to that on non-personalized ad (Mpersonalized ad ¼ 13.04 vs. Mnon-personalized ad ¼ 6.09, F (1, 89) ¼ 9.09, p < .01). Although there was no significant difference between personalized ad and non-personalized ad regarding time to first fixation (Mpersonalized ad ¼ 47.25 vs. Mnon-personalized ad ¼ 68.92, F (1, 89) ¼ 3.83, p ¼ .06), the direction was consistent with our prediction. Thus, H2a was rejected while H2b and H2c were supported (See Fig. 1). However, when people are exposed to the low cognitive demanding task, there was no significant difference between personalized and non-personalized ad regarding time to first fixation on the ad (Mpersonalized ad ¼ 54.91 vs. Mnon-personalized ad ¼ 53.04 F (1, 89) ¼ .02 p > .05), absolute visual attention to the ad (Mpersonalized ad ¼ 3.43 vs. Mnon-personalized ad ¼ 3.49 F (1, 89) ¼ .01
872
H. Bang, B.W. Wojdynski / Computers in Human Behavior 55 (2016) 867e876
Table 1 Two-way ANOVA. DV
Factors
df
F Value
p
Absolute visual attention
Personalization Cognitive difficulty Personalization x cognitive Personalization Cognitive difficulty Personalization x cognitive Personalization Cognitive difficulty Personalization x cognitive Personalization Cognitive difficulty Personalization x cognitive Personalization Cognitive difficulty Personalization x cognitive
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
5.03 2.11 5.23 1.41 .24 1.99 4.06 .01 3.94 .20 1.26 .58 .57 .18 1.04
.05 .05 >.05 >.05 .05 ¼.05* >.05 >.05 >.05 >.05 >.05 >.05
Time to first fixation
Frequency of fixation
Perceived goal impediment
Aad
difficulty
difficulty
difficulty
difficulty
difficulty
supported.
10 9 8
8.01
5. Discussion and conclusion
7 6 5 4
Personalization 3.49
Non-personalization 3.43
3
2.47
2 1 0 Cognitive Difficulty_low
Cognitive Difficulty_high
Fig. 1. Personalization x cognitive difficulty on absolute visual attention.
p > .05), and frequency of fixation (Mpersonalized ad ¼ 9.57 vs. Mnon¼ 9.51 F (1, 89) ¼ .00, p > .05). Therefore, H3a, H3b and H3c were all supported. Even though H3 was supported, contrary to our prediction, both personalized and non-personalized ads received relatively low level of attention. This issue is addressed in the Discussion section.
personalized ad
4.4. Testing hypothesis 4 H4 predicted an interaction between personalization and cognitive demand of task on perceived goal impediment. Results of a 2 2 ANOVA showed no significant interaction effect of personalization and cognitive demand of task on PGI, F (1, 89) ¼ .58, p > .05. Therefore, H4 was not supported. 4.5. Testing hypothesis 5 H5 predicted that the participants' attitudes toward the personalized ad would be moderated by task cognitive demand. To test the relationship between ad personalization and task cognitive demand on attitudes toward the ad, a 2 2 ANOVA was conducted with personalization and task cognitive demand as fixed factors and Aad as the dependent measure respectively. No significant interaction effects of personalization and cognitive capacity on Aad were found, F (1, 89) ¼ 1.04, p > .05. Therefore, H5 was not
Based on the notion that personally salient information attracts people's attention (Harris & Pashler, 2004; Mack & Rock, 1998), this study investigated the impacts of personalized ad on attention using eye-tracking data. Consistent with the implications of previous literature, it was found that personalized advertisements attract significantly longer and more attention than nonpersonalized ads, indicating the strong attention-grabbing effect of personalization. Even though it was found that participants' gaze stayed longer on the personalized ad than on the non-personalized ad, there were no differences between two conditions regarding the time to first fixation. There is also reason to believe that the task context in which a Web user encounters a personalized ad may shape attention. That is, the level of cognitive effort required for the task moderated the effect of personalized ad on attention. Specifically, when people engaged in the high-cognitive demand task, people tended to pay much deeper and longer attention to personalized ad compared to non-personalized ad. As implied in limited capacity theories, when people are involved in goal-directed or high-cognitivedemand tasks, they have very limited amount of cognitive resources left for the processing task-irrelevant information such as advertising. Based on this theory, a number of scholars and practitioners have believed that placing ads alongside content that requires a large amount of cognitive effort to process might doom them to fail to attract consumers' attention. However, the result of this study indicates that personally salient advertising could be the successful way to get through the cognitive load and be noticed by consumers while generic ad messages are easily ignored. On the other hand, when people are involved in a lowcognitive-demand task (i.e., free-viewing task), people paid relatively low amount of attention to both personalized and nonpersonalized ad contrary to our prediction. Because people with low cognitive demand task have enough cognitive resources to allocate on irrelevant information, participants were expected to pay enough attention to both ads regardless of the personalization. However, the result of the current study may indicate that when people are not strongly engaged and involved in the media consumption process on the web, they tend to show habitual and
H. Bang, B.W. Wojdynski / Computers in Human Behavior 55 (2016) 867e876
automatic ad avoidance phenomenon for both personalized and non-personalized ads, a phenomenon called “banner blindness” (Benway, 1998). In our experiment, the positioning of the advertisement in the right rail of the page, a common advertising area, may have triggered some avoidance on the part of participants that would not be present if the ad were embedded more within the content itself. Additionally, people may spend less time on the web pages when they do not feel interesting or engaging, and this could possibly result in the overall low level of attention to contents on the webpage. Indeed, in the study, people assigned to low cognitive demand task spent significantly less time (M ¼ 65.96 s) compared to people with high cognitive demand task (M ¼ 96.55 s) (t ¼ 3.93, p < .01). Therefore, further research is needed to see if the personalized information attracts more attention than non-personalized advertising under low cognitive demand, but interesting and engaging task situation (e.g., entertainment or shopping). Another noteworthy finding of the current study is the nonsignificant interaction of personalization and cognitive demand of the task on perceived goal impediment and Aad. Theoretically, the attention-grabbing feature of advertising could have detrimental effect on attitudinal responses to the ad especially under goaldirected condition because of its perceived goal impediment (distractor devaluation theory; Raymond, Fenske, & Tavassoli, 2003; Simola et al., 2013). However, the finding of our study indicated that even though personalized ad attract more and longer attention from consumers under highly goal-directed situation, contrary to the theoretical expectation, it was found not to elicit high level of perceived goal impediment and negative attitude toward the ad. This might be because personalized messages increase the perceived value and perceived utility of the message (Naughton, Jamison, & Sutton, 2013; Xu, 2006) led by personal relevance and involvement (Kalyanaraman & Sundar, 2006), which may possibly offset the feeling of goal impediment and negative responses, such as ad avoidance (Baek & Morimoto, 2012; Pasadeos, 1990). In fact, Xu (2006) suggested that personalization tactic is the effective way to elicit positive attitude toward the mobile advertising by increasing ad utility and reliving the ad intrusiveness. Therefore, even though the salient feature of ad is known to have detrimental effect on task performance and in turn attitudinal response (Simola et al., 2013; Wolford & Morrison, 1980), the current study suggests that the personalization feature of ad could be a strategy to avoid the subsequent detrimental effect while grabbing attentions from consumers. Therefore, the results of this research offer a significant contribution to the advertising and consumer psychology literature. In spite of the continuous growth of personalized advertising, the theoretical understanding of it was in exploratory stages. The current study further supports the findings of psychology literature that emphasizes the role of personally salient information on attention. Furthermore, as the present work found the moderating effects of cognitive demand of tasks, our findings deepen understanding of how personalized advertising works in different web contexts or media motivation. By tailoring advertising based on personal information in experiment and incorporating physiological measures with self-report measures, this study provides more enhanced understanding of personalized advertising. Even though the salience feature of ad, that greatly attract users' attention, has been considered to generate negative responses to the ad, the finding of the current research suggests the possibility that negative effect of attention-grabbing of ad on ad intrusiveness and attitudinal responses could be moderated by the feature of the ad
873
salience. This empirical investigation also provides managerial implications for advertising practitioners of where to place personalized advertising. For example, if marketers advertise products that set their goal to attract consumers' attention to increase awareness or to place the products in consumers' consideration set, they could have benefit from placing the personalized ad in the webpage, where consumers are likely to cognitively engaged rather than the web pages where consumers freely navigate. As advertisers and content producers continue to find new ways to target and craft their appeals to users, understanding the conditional nature of the effects of this targeting will play a key role in shaping their effectiveness. As with all empirical investigations, there are several limitations that should be acknowledged. First, we only incorporate participants' personal information such as name, school name, and geographic information in the ad. However, as the definition of personalized advertising and research by industry implies (Internet Retailer, 2013), users' past product choice or search is one of the predominant data that is used for marketing purpose. Therefore, we recommend that other scholars also investigate how advertising that incorporate those behavioral data affects consumers' attention and attitude toward the ad. Also, even though the manipulation check for the cognitive demand of tasks was successful, the mean score of cognitive effort scale for the high cognitively demanding task was below the neutral score of 4. That is, although the highly cognitive demanding task was indeed required significantly more cognitive effort compared to the lowly cognitive demanding task, the task itself was not difficult enough for participants to perceive that they were using substantial cognitive resources. Future research with stronger task manipulations is needed. Furthermore, the study only had about 93 participants due to the time-intensive nature of data collection. Finally, this study employed limited types of product category (i.e., coffee shop) and advertising format (i.e. banner ad). Therefore, the generalizability of the findings is limited. Therefore, future research is needed to see whether the similar effects can be replicated across different categories, media and advertising formats (e.g., embedded ad, editorial ad). Thinking back to online shoppers like future-mom Sara mentioned at the beginning of the study, the current study suggests that the attention-grabbing effects of the personalized ads Sara receives in the course of her Web browsing is likely to differ depending on the cognitive demand of the tasks she engages in. An advertisement for baby gear is likely to attract greater attention from Sara when it appears along with the daily news articles, and more likely to lead banner blindness when it appears on her Facebook news feed. Although previous studies examining effects of ad personalization have found mixed results, this study suggests that task characteristics may be a key moderator which determines the attention-grabbing effects of personalized ad, and the study provides a general picture of how personalized advertising works differently under different contexts. While additional research on situational effectiveness of personalized advertising is needed, we hope that the present study and its results influence future scholarship in this area. Appendix A. Absolute Fixation Duration (Personalized & High cognitive demand task).
874
H. Bang, B.W. Wojdynski / Computers in Human Behavior 55 (2016) 867e876
H. Bang, B.W. Wojdynski / Computers in Human Behavior 55 (2016) 867e876
B. Absolute Fixation Duration (Non-personalized & High cognitive demand task).
875
876
H. Bang, B.W. Wojdynski / Computers in Human Behavior 55 (2016) 867e876
References AdNews. (2013, June 13). No one's paying attention to digital ads…well, 8% are. Retrived from http://www.adnews.com.au/adnews/no-one-s-paying-attention-todigital-ads-well-8-are. Aguirre, E., Mahr, D., Grewal, D., de Ruyter, K., & Wetzels, M. (2015). Unraveling the personalization paradox: the effect of information collection and trust-building strategies on online advertisement effectiveness. Journal of Retailing, 91(1), 34e49. Baek, T. H., & Morimoto, M. (2012). Stay away from me. Journal of Advertising, 41(1), 59e76. Benway, J. P. (1998, October). Banner blindness: The irony of attention grabbing on the World Wide Web. In Proceedings of the Human Factors and Ergonomics Society Annual Meeting. Vol. 42(5), 463e467. Boerman, S. C., van Reijmersdal, E. A., & Neijens, P. C. (2015). Using eye tracking to understand the effects of brand placement disclosure types in television programs. Journal of Advertising, 44(3), 196e207. Bragge, J., Sunikka, A., & Kallio, H. (2013). An exploratory study on customer responses to personalized banner messages in the online banking context. Journal of Information Technology Theory and Application (JITTA), 13(3), 2. Cho, C., & Cheon, H. J. (2004). Why do people avoid advertising on the internet? Journal of Advertising, 33(4), 89e99. Desimone, R., & Duncan, J. (1995). Neural mechanisms of selective visual attention. Annual review of neuroscience, 18(1), 193e222. Edwards, S. M., Li, H., & Lee, J. H. (2002). Forced exposure and psychological reactance: antecedents and consequences of the perceived intrusiveness of pop-up ads. Journal of Advertising, 31(3), 83e95. Finley, K. (2015, May 6). Facebook and IBM team up to supercharge personalized ads. Wired. Retrieved from: http://www.wired.com/2015/05/facebook-ibm-teamsupercharge-personalized-ads/. Gilbert, D. T., Pelham, B. W., & Krull, D. S. (1988). On cognitive busyness: when person perceivers meet persons perceived. Journal of personality and social psychology, 54(5), 733e740. Gwizdka, J. (2010). Distribution of cognitive load in web search. Journal of the American Society for Information Science and Technology, 61(11), 2167e2187. Harris, C. R., & Pashler, H. (2004). Attention and the processing of emotional words and names not so special after all. Psychological Science, 15(3), 171e178. Hoffman, D. L., & Novak, T. P. (1996). Marketing in hypermedia computer-mediated environments: conceptual foundations. The Journal of Marketing, 50e68. Howard, D. J., & Kerin, R. A. (2004). The effects of personalized product recommendations on advertisement response rates: the “try this. It works!” technique. Journal of consumer psychology, 14(3), 271e279. Internet Retailer. (2013). Site personalization has room to grow among top 500. retailers at https://www.internetretailer.com/2013/05/15/site-personalizationhas-room-grow-among-top-500-retailers. Kalyanaraman, S., & Sundar, S. S. (2006). The psychological appeal of personalized content in web portals: does customization affect attitudes and behavior? Journal of Communication, 56(1), 110e132. Kreuter, M. W., & Wray, R. J. (2003). Tailored and targeted health communication: strategies for enhancing information relevance. American Journal of Health Behavior, 27(Suppl. 3), 227e232. Lang, A. (2006). Using the limited capacity model of motivated mediated message processing to design effective cancer communication messages. Journal of Communication, 56(1), 57eS80. Lee, S., Kim, K. J., & Sundar, S. S. (2015). Customization in location-based advertising: effects of tailoring source, locational congruity, and product involvement on ad attitudes. Computers in Human Behavior, 51, 336e343. Li, C. (2016). When does web-based personalization really work? the distinction between actual personalization and perceived personalization. Computers in Human Behavior, 54, 25e33. Li, H., Edwards, S. M., & Lee, J. H. (2002). Measuring the intrusiveness of advertisements: Scale development and validation. Journal of advertising, 31(2), 37e47. Mack, A., & Rock, I. (1998). Inattentional blindness. The MIT Press. MacKenzie, S. B., & Lutz, R. J. (1989). An empirical examination of the structural antecedents of attitude toward the ad in an advertising pretesting context. The Journal of Marketing, 48e65. Malheiros, M., Jennett, C., Patel, S., Brostoff, S., & Sasse, M. A. (2012, May). Too close for comfort: a study of the effectiveness and acceptability of rich-media personalized advertising. In Proceedings of the SIGCHI conference on human factors in computing systems (pp. 579e588). ACM. Moray, N. (1959). Attention in dichotic listening: affective cues and the influence of instructions. Quarterly journal of experimental psychology, 11(1), 56e60. Naughton, F., Jamison, J., & Sutton, S. (2013). Attitudes towards SMS text message smoking cessation support: a qualitative study of pregnant smokers. Health
education research, 28(5), 911e922. Nyheim, P., Xu, S., Zhang, L., & Mattila, A. S. (2015). Predictors of avoidance towards personalization of restaurant smartphone advertising: a study from the millennials' perspective. Journal of Hospitality and Tourism Technology, 6(2), 145e159. Pasadeos, Y. (1990). Perceived informativeness of and irritation with local advertising. Journalism & Mass Communication Quarterly, 67(1), 35e39. Pavlou, P. A., & Stewart, D. W. (2000). Measuring the effects and effectiveness of interactive advertising: a research agenda. Journal of Interactive Advertising, 1(1), 61e77. Petty, R. E., Barden, J., & Wheeler, S. C. (2002). The elaboration likelihood model of persuasion: health promotions that yield sustained behavioral change. In R. J. DiClemente, R. A. Crosby, & M. Kegler (Eds.), Emerging theories in health promotion practice and research (pp. 71e99). San Francisco: Jossey-Bass. Phelps, J. E., D'Souza, G., & Nowak, G. J. (2001). Antecedents and consequences of consumer privacy concerns: an empirical investigation. Journal of Interactive Marketing, 15(4), 2e17. Poon, S., & Jevons, C. (1997). Internet-enabled international marketing: a small business network perspective. Journal of Marketing Management, 13, 29e41. Raymond, J. E., Fenske, M. J., & Tavassoli, N. T. (2003). Selective attention determines emotional responses to novel visual stimuli. Psychological Science, 14(6), 537e542. Rodgers, S., & Thorson, E. (2000). The interactive advertising model: how users perceive and process online ads. Journal of Interactive Advertising, 1(1), 41e60. Sacirbey, O. (2000). Privacy concerns, advertising hamper E-commerce. IPO Reporter, 24, 11. Sheehan, K. B., & Hoy, M. G. (1999). Flaming, complaining, abstaining: how online users respond to privacy concerns. Journal of advertising, 28(3), 37e51. Simola, J., Kivikangas, M., Kuisma, J., & Krause, C. M. (2013). Attention and memory for newspaper advertisements: effects of Adeeditorial congruency and location. Applied Cognitive Psychology, 27(4), 429e442. Smit, E. G., Van Noort, G., & Voorveld, H. A. (2014). Understanding online behavioural advertising: user knowledge, privacy concerns and online coping behaviour in Europe. Computers in Human Behavior, 32, 15e22. Sundar, S. S., & Marathe, S. S. (2010). Personalization versus customization: the importance of agency, privacy, and power usage. Human Communication Research, 36(3), 298e322. Sweller, J. (1988). Cognitive load during problem solving: effects on learning. Cognitive Science, 12(2), 257e285. €nboer, J. J. G., & Paas, F. G. W. C. (1998). Cognitive architecture Sweller, J., Van Merrie and instructional design. Educational Psychology Review, 10(3), 251e296. Tam, K. Y., & Ho, S. Y. (2005). Web personalization as a persuasion strategy: an elaboration likelihood model perspective. Information Systems Research, 16(3), 271e291. Tucker, C. (2011, May). Social networks, personalized advertising, and perceptions of privacy control. In Proceedings of the tenth workshop on the economics of information security (WEIS). Turow, Joseph, King, Jennifer, Chris Jay, Hoofnagle,, Amy, Bleakley, & Michael, Hennessy (September 29, 2009). Americans reject tailored advertising and three activities that enable it. Available at SSRN http://ssrn.com/ abstract¼1478214. U.S. Federal Trade Commission. (2000). Privacy online: Fair information practices in the electronic marketplace (A Report to Congress, May). Van Doorn, J., & Hoekstra, J. C. (2013). Customization of online advertising: the role of intrusiveness. Marketing Letters, 24(4), 339e351. White, T., Zahay, D., Thorbjørnsen, H., & Shavitt, S. (2008). Getting too personal: reactance to highly personalized email solicitations. Marketing Letters, 19(1), 39e50. Wise, K., Kim, H. J., & Kim, J. (2009). The effect of searching versus surfing on cognitive and emotional responses to online news. Journal of Media Psychology, 21(2), 49e59. Wolford, G., & Morrison, F. (1980). Processing of unattended visual information. Memory & Cognition, 8(6), 521e527. Wolin, L., & Korgaonkar, P. (2005). Web advertising: gender differences in beliefs, attitudes, and behavior. Journal of Interactive Advertising, 6, 125e136. Xu, D. J. (2006). The influence of personalization in affecting consumer attitudes toward mobile advertising in China. Journal of Computer Information Systems, 47(2), 9e19. Yuan, S. T., & Tsao, Y. W. (2003). A recommendation mechanism for contextualized mobile advertising. Expert Systems with Applications, 24(4), 399e414. Yu, J. H., & Cude, B. (2009). ‘Hello, Mrs. Sarah Jones! We recommend this product!’Consumers' perceptions about personalized advertising: comparisons across advertisements delivered via three different types of media. International journal of consumer studies, 33(4), 503e514.
