Master Thesis
What is Really Too Much? The Influence of Differing Degrees of Online Persuasive Message Personalization on Implicit and Explicit Website Attitude
By Robin C. Voboril Hammersley Student Number: 11108193 Supervisor: Dr. Guda van Noort
University of Amsterdam Graduate School of Communication Master´s programme Communication Science
Persuasive Communication Date of Completion: 02.02.2017
2 Abstract
In the Netherlands internet advertising represents one of the fastest growing advertising branches in terms of expenditure. The present study aims to identify the mediating influences of intrusiveness between differing degrees of advertising personalization and attitude towards the website. User’s perceived intrusiveness of the advertisement is expected to mediate the relationship between personalization and website attitude. Reactance and information processing theory formed the basis of a theoretical model which was tested with a scenario-based experiment at the University of Amsterdam. Two of the three hypotheses were supported. Intrusiveness did indeed mediate the relationship between both implicit and explicit attitude towards the website. Higher degrees of personalization, specifically utilizing name personalization causes increased feelings of intrusiveness which in turn negatively affects participant’s attitude towards the website both implicitly and explicitly. A specific difference in the total implicit and explicit attitude dimension was not found to be present, however.
3
Introduction
Internet advertising spend in the Netherlands has seen a significant growth over the past couple of years. In 2015 the online portion of all advertising spend saw an increase of 8.3% compared to the previous year and the estimated forecast for 2016 is 7.8% as of March 2016. Although lower than the previous year internet advertising still eclipses all other media channels (IAB & Deloitte, 2016). Expectations are that this trend will not slow down anytime soon, as it stems from the general progression of people spending more and more time on the internet, leaving less for alternatives (eMarketer). Furthermore, due to the relatively recent emergence of internet advertising it is still very much underdeveloped compared to
advertising on competing media channels.
The internet is developing rapidly however, and database technology, behaviour-tracking and personalized internet advertising as a whole continues to grow with it. Industry research shows that already in 2013, using browser cookies, more than half of retailers on the web provide product recommendations or web page personalization (Internet Retailer, 2013). Marketers are thus enabled to target more accurately and tailor advertising to their consumers' preferences, needs and interests (Pavlou & Stewart, 2002) by increasing message relevance or and involvement (Goldfarb & Tucker, 2011).
Yet, personalized online advertising can be considered a double-edged sword, increasing purchase intentions but also feelings of intrusiveness which in turn negatively affect outcomes such as purchase intentions (Van Doorn & Hoekstra, 2013). Intrusiveness in itself has been recognized as a common complaint of distractive advertising as early as 1968. It is a description of the degree to which an advertisement poses an unwelcome distraction from the task and Bauer and Greyser (1968) recognized that it was a major cause of offline advertising annoyance. Similarly, current online ad intrusiveness research has established that
4 it can negatively affect various perceptions and attitudes (McCoy, Everard, Polak, & Galletta, 2008; McCoy, Everard, Galletta & Moody, 2015; Van Doorn & Hoekstra, 2013; Li, Edwards & Lee, 2002; Diao & Sundar, 2004; Morimoto & Chang, 2006).
Findings from numerous studies have also suggested that there is a disconnect between the implicit and explicit dimensions of attitude (Gawronski & Strack, 2004; Gregg, Seibt, & Banaji, 2006; Häfner & Trampe, 2009). In current cognitive and affectual research a
distinction is made between decisions made on a conscious and a subconscious level. Current models which investigate a change in attitudes distinguish between the more deliberate and propositional self-reported explicit attitudes and the automatically activated associations in our memory from which implicit attitudes are derived (Gawronski & Bodenhausen, 2006; Wilson, Lindsey, & Schooler, 2000). Strick et al. (2011) thus further suggest that conclusions which are based solemnly on explicit attitudes might be premature, and perhaps need to be changed if implicit attitudes are considered. This insight can easily be transferred to the field of intrusiveness research, as advertisements usually have some degree of subliminal influence on the consumer as suggested by the mere exposure effect (Matthes, Schemer &Wirth, 2007). Online adverts´ subconscious (implicit) influence on the consumer due to their intrusiveness can thus be significantly different from their conscious (explicit) influence. This has some important implications for many fields of research, including, the field of online personalized advertisements, as there are, in a sense, two sides of a single outcome to consider.
The outcome variable of the present study is also a form of attitude. Attitude has been either acknowledged as or been a major focus in studies that investigate prominent outcome effects such as revisit intention, use frequency, site recommendation (McCoy et al. 2015), purchase intention (Lin & Kim, 2016) or ad avoidance (Li, Edwards & Lee, 2002), as a result of negative advertising effects. Explanations are offered by the theory of planned behavior by Ajzen (1991), as well as the theoretical model of the persuasion process as proposed by
5 McGuire (1968). All of which posit attitude as being one of the major antecedents to the intention of performing a behavior and thus also the adoption or execution of a behavior. The focus of this study will therefore specifically be the attitude which the consumer has towards the website on which the advertisement is located, or website attitude.
The current thesis will explore the influence of differing degrees of e-commerce advertisement personalization on user’s implicit and explicit website attitudes through their perceived intrusiveness. Advertisement personalization degrees will be expressed through the use of online ads containing increasing levels of information with higher distinctiveness (i.e., browsing data vs. browsing data and name vs. browsing data and transaction history), making the individual more uniquely identifiable (McCoy et al., 2008; Van Doorn & Hoekstra, 2013). A higher distinctiveness of information increases the user´s feelings of intrusiveness (Van Doorn & Hoekstra, 2013), which in turn should negatively affect website attitude due to the ads increasing ability to distract from the websites content and interfere with the users goal pursuit (Diao & Sundar, 2004; Morimoto & Chang, 2006). The research question which follows from this is:
What is the mediating influence of perceived advertising intrusiveness between differing degrees of advertising personalization, and implicit and explicit website attitude?
The present study also focusses on addressing a gap in current research which stems from the prevalence of explicit outcome measurements and the aforementioned disparity between implicit and explicit results. For instance while most studies agree that a higher degree of personalization leads to more perceived intrusiveness, resulting in negative advertising effects (McCoy et al., 2008, McCoy et. al., 2015, Van Doorn & Hoekstra, 2013;
6 Li, Edwards & Lee, 2002), Okazaki et al. (2009) for example also acknowledge that a higher degree of relevancy of the ad content for the individual can result in a positive disposition within the consumer due to feeling of gratitude for receiving the right information at the right time and relieving them of the need to search further. This study will therefore also examine the total disparity between the implicit and explicit attitudinal outcome of the experiment.
Investigating both the implicit and explicit effects of online ad personalization is an important task that could enable advertisers to more accurately predict the outcome of their efforts. An advertiser utilizing online ad personalization has a distinct expectation, based on explicit reported outcomes, concerning the influence his specific persuasive message will have. Yet the implicit influence of his ad might actually be counterproductive to what he is trying to achieve. This is compounded by the fact that the personalization of advertising efforts on the internet, which use more personal distinct information about the consumer and go further than the information gathered through his or her browsing history, is becoming ever more prevalent (Van Doorn & Hoekstra, 2013).
Theoretical Background
The current thesis will explore the influence of differing degrees of online
advertisement personalization on user’s perceived intrusiveness and the resulting influence of perceived intrusiveness on implicit and explicit website attitudes. The following chapter will investigate the relevant literature and theoretical background to establish the hypotheses and ultimately arrive at the hypothetical model which is the foundation of this thesis.
Online personalized advertising refers to the customization or tailoring of internet-based advertising efforts to an individual’s own preferences using some amount of his or her distinct personal information (Chellappa & Sin, 2005). This can include personally
7 identifying information (such as name or occupation), shopping-related information (such as browsing history or brand preferences) and demographic information (Yu & Cude, 2009). Personalized advertising on the internet, and on mobile for that matter, although a broad category, has in the recent years focused heavily on advertising efforts based on user’s identity and behavior (Aguirre et al., 2015), also known as online behavioral advertising (Smit, Van Noort, & Voorveld, 2014). Facilitated through ever evolving tracking and data storage methods (Bang & Wojdynski, 2016), the collected information can be displayed in a variety of ways, one of the most prevalent and relevant still being traditional banner ads (eMarketer, 2017). They will make up the stimulus object for this study’s experiment as a result.
Online personalized advertising has been investigated by numerous studies and higher degrees of personalization, expressed in more personal and distinct consumer information used, were found to result in a higher relevance and a more fitting offer for the consumer (Okazaki et al. 2009). Personalization has been linked to positive advertising effects, such as increased purchase intention (Goldfarb & Tucker, 2011), brand attitude, click intention (De Keyzer, Dens & De Pelsmacker, 2015) or general decision outcomes (Tam & Ho, 2006). Many studies have also investigated the negative side of personalization; that a higher degree of personalization (more distinct personal information used) leads to more perceived
intrusiveness, resulting in negative advertising effects (McCoy et al., 2008, Van Doorn & Hoekstra, 2013; Li, Edwards & Lee, 2002) and thus possibly also an undermining of the intentions of the advertiser. Particularly Van Doorn and Hoekstra (2013) have shown how increased degrees of personalization, in particular adding personal identification or transaction information, instead of utilizing only browsing data, can impact purchase intention through the participants perceived intrusiveness. It was found that higher degrees of personalization indeed increase feelings of intrusiveness which negatively affects purchase intention. This
8 study builds on their research to the extent that personalization also describes to what degree the stimulus object: banner ads, will be tailored to the personal information of the character within a hypothetical scenario (presented to the participants using name identification, transaction information and browsing data). Browsing data alone is expected to be the condition with the lowest distinctiveness and adding either transaction data or a name will increase distinctiveness and thus potentially increase intrusiveness.
Intrusiveness is a common variable in studies investigating the effects of online advertising customization (McCoy et al., 2008, McCoy et. al., 2015, Van Doorn & Hoekstra, 2013; Li, Edwards & Lee, 2002) and is most often used as an intermediating variable between the independent variable and various advertising effects. Intrusiveness as a construct generally describes the degree to which an advertisement poses unwelcome distraction from the task with which an individual is currently occupied (Li, Edwards & Lee, 2002; McCoy et. al., 2015). In the context of online personalized advertising it thus describes negative feelings in the consumer as a result of being exposed to an ad which uses unwelcome personalized content that the consumer might perceive as an invasion of her or his personal space and thus distract him from his/her online task. Li, Edwards and Lee (2002) were some of the first researchers to investigate online ad intrusiveness and its effect on irritation and avoidance, while also developing a seven-item scale that can be used to measure the perceived
intrusiveness of advertisements across media. The study revealed that a larger degree of perceived intrusiveness has a positive influence on annoyance and consumers are thus more likely to engage in avoidance. This effect was compounded by the fact that the goal directed nature of internet advertising also causes consumers to experience more ad intrusiveness than comparable advertisements in traditional, more passive media (Li, Edwards & Lee, 2002). Feelings of perceived intrusiveness as a result of personalization are thus expected to
9 Attitude toward the website is defined as the consumer's predisposition to evaluate a website either favorably or unfavorably as result of web-based content (Cases et al., 2010). The theory of planned behavior by Ajzen (1991), the theory of reasoned action as well as the theoretical model of the persuasion process as proposed by McGuire (1968) posit attitude as being one of the major antecedents to the intention of performing a behavior and thus also the adoption or execution of a behavior. Fortin and Dholakia (2005) also found it to be an
important antecedent of internet behavior and it was deemed to positively influence attitude toward the product yielding higher product evaluations (Tomaseti, Ruiz & Reynolds, 2009). The main antecedents of website attitude were found to be predominantly emotion-based evaluations of the website; including how entertaining or effective it is (Chen & Wells, 1999). Degrees of personalization and the resulting intrusiveness can be considered to have a direct impact on how entertaining and especially how effective it is in the users mind. Website attitude will thus accurately reflect impact of intrusiveness and personalization on the participant’s affective response as a result of the manipulation. Additionally, McCoy et. Al (2015) investigated website attitude as a direct result of advertising intrusiveness, which expands insights into personalized online advertising, gained by Van Doorn and Hoekstra (2013) toward purchase intention and attitude towards the ad, to include website attitude.
The main theoretical explanations for the effects of intrusiveness and personalized ads on attitudinal effects can be found in psychological models of information processing; such as the reactance theory (McCoy et al., 2008, McCoy et. al. 2015, Van Doorn & Hoekstra, 2013). Brehm (1966) developed the reactance theory as an explanation for the reactions which people experience as a result of having their freedom of choosing and acting upon a chosen behavior restricted. It assumes that that someone who experiences the threat of or actual loss of freedom in choosing a behavior will value that specific behavior more than its alternatives (Brehm, 1966). McCoy et al. (2016) stated that when applied to online advertising, the
10 expectation of free choice stems from accessing the websites content to perform a task
without interruption, while the online advertisement is then perceived as a barrier restricting the ability to access the web content or to perform the task in the most efficient manner possible (McCoy et al. 2016). The user will form attitudes or beliefs as a result and based on the kind of persuasive attempt, possibly including reactance if the restriction of freedom is strong or the individual is susceptible enough (Bennett & Robinson, 2000). Moreover, there is additional cognitive effort required on the side of the user when interrupted while performing a task (Bailey et al., 2001). Varying subjective reactions, based on the type and nature of the interruption are a result of this additional effort that is required (Reid & Nygren, 1988). These differing reactions could however, be an indication of the disconnect between implicit and explicit attitudes.
A major contribution of this study is the investigation of the incongruities in outcome measurements of the implicit and explicit dimensions when investigating the same construct. As touched upon the present study´s outcome variable website attitude distinguishes between implicit and explicit website attitude. Studies have shown that implicit attitude of the same element can be disconnected from its explicit counterpart (Gawronski & Strack, 2004; Gregg, Seibt, & Banaji, 2006; Häfner & Trampe, 2009). There are several theoretical explanations for why implicit and explicit attitude towards the site may not be strongly related. These are based on known influences of responses to self-report measures (the most common form of outcome measurement in advertising research) which affect implicit measures in a variety of different ways. One of these influences is impression management, the conscious or
subconscious regulation of one’s perception in the eyes of others would justify answering the self-report measures in accordance to social norms, especially with a researcher present (Tedeschi, Schlenker, & Bonoma, 1971). Another is demand characteristics, where supposed insights into the experiments intent influence the participants to adhere to that intent (Orne,
11 1962). Evaluation apprehension, posits that a reduction of performance occurs for participants of an experiment due to the possibility of disapproval by the observers (Rosenberg, 1969). Lastly, enhancement could describe how participants choose an outcome to their self-reports, which enhances other´s or their own perception of themselves (Taylor & Brown, 1984). All these influences are based on the perceptions of the public opinion; the participant will be influenced according to what he perceives the social norm should be.
According to the above mentioned relationships it can be concluded that a higher amount of personal information in a banner ad may pose more of a distraction, which
interferes with the consumer’s cognitive processing and interrupts his goal pursuit, leading to more perceived intrusiveness. Intrusiveness, as this study´s mediator and an indicator to what extent the advertisement poses an unwelcome distraction, necessitating additional cognitive effort will result in a negative impact on attitude towards the site. This is true for both implicit and explicit website attitude, as the general direction of the effect should not be affected by self-report influences. The following hypotheses can be derived from this:
H1a: A higher degree of personalization will result in a more negative explicit website
attitude, mediated by perceived intrusiveness.
H1b: A higher degree of personalization will result in a more negative implicit
website attitude, mediated by perceived intrusiveness.
As stated and unlike most studies investigating intrusiveness of advertising
personalization; a major aim of this thesis is to investigate the disparity between implicit and explicit website attitude. But how is the difference between the two dimensions of attitude toward the website expected to deviate? In what way do report influences affect
self-12 report measures that exempt implicit measures of attitude? O'Donnell and Cramer (2015) investigated general attitude toward data collection for the use of personalized ads and found that attitudes were fairly neutral towards personalization with a slight majority (51%) being strongly concerned with the issue. These insights into consumer attitude can be combined with findings from for example Okazaki et al (2009) on the positive influences of
personalization, such as feelings of gratitude for receiving the right information at the right time and relieving them of the need to search further. From these insights a hypothesis on the relation between implicit and explicit website attitude can be developed. It states that with general public attitude towards personalized ads being slightly negative, the influences on self-report measures (such as demand characteristics) will cause a compounding effect with participants of the experiment of the current study. They should be driven to answer the self-report measures more negatively. Their implicit attitude will not be affected however and should accurately portray their attitude as a result of the advertisements alone. The hypothesis is thus:
H2: All participants’ implicit website attitude will be significantly more positive than
their explicit attitude.
From the conclusions drawn up previously and the development of the hypotheses for this study the following hypothetical model, illustrated in figure 1, can be constructed. It depicts the major constructs of this study and the hypothesized relationships that link them. In continuation the methodology of the experiment to prove these relationships will be laid out.
13
Figure 1 Hypothesized Model
Methodology
Sample
The sample consisted of young adult students (N = 120), between the ages of 18 and 35 (M = 22.83, SD = 2.44) from the University of Amsterdam. 66.9% (N = 81) of the participants were female (M = 1.68, SD = 0.47) and, 83% (N = 75) of all participants were living in the Netherlands at the time. The most common highest completed level of education was a bachelor´s degree with 57% (N = 69), followed by high school degrees with 34.7% (N = 42). There were no participants which had not reached the high school level of completed education. An initial observation of the data revealed that none of the 120 current participants had to have been eliminated, no outliers were detected, the completion rate of study was 100% and the IAT error rates for all participants were well within the acceptable margin of 20% (Greenwald, Nosek & Banaji, 2003; Karpinski & Steinman, 2006).
Explicit Attitude Implicit Attitude Degrees of Personalization + H1a H2 Perceived Intrusiveness Website attitude H1b + H1a H1b - -
14
Design
The present experiment was a 3 (degrees of personalization) x 2 (website attitude) mixed-subjects factorial design, with personalization being a between- and website attitude a within-subjects factor. Personalization was a true experimental factor and made up of the conditions, (1) browsing history only, (2) browsing history and name and (3) browsing and transaction history, referring to the amount and type of personal information included in the persuasive message. Website attitude was a repeated measures factor and its conditions reflect the two cognitive dimensions which are central to this thesis, implicit and explicit website attitude.
Operationalization
‘Degrees of personalization’ the independent variable of this study was
operationalized with the use of more distinct personal information in the second and third conditions. In overall three conditions the advertisements were tailored to the protagonist by including (1) browsing history only, (2) browsing history plus the protagonists name (Henny de Vries, a typical Dutch unisex name) and (3) browsing history in addition to a reference to previous online purchases or transactions. The following sentences have been chosen to represent the three conditions of ‘degrees of personalization’:
‘Browsing history only’: “Opt in for renewable energy and purchase a solar installation from S&S Solar now!” ‘Browsing history and name: “Hi Henny, opt in for renewable anergy and
purchase a solar installation from S&S Solar now!”
15 Browsing and transaction history: “You have purchased from us before, now extend
your renewable energy potential and purchase a solar installation from S&S Solar now!”
The examined banner advertisements were congruent in their advertised content with the content displayed on the website. This specifically refers to the fit of the advertisements to the overall context of the situation and has been found to have a positive influence on
consumers purchase intentions, but also to possibly increase perceived intrusiveness (Van Doorn & Hoekstra, 2013). By fitting the advertisements to the situation per default it is expected that the difference between implicit and explicit website attitude will automatically become more distinct, as fitting the ad to the situation is likely to increase intrusiveness but also gratitude for receiving relevant information at the right time (Okazaki et al. 2009).
Procedure
The sample of participants was randomly split into three different groups. Each of these groups was assigned to one of the experimental conditions set by the personalization variable. As website attitude was a repeated measures factor each of the three personalization conditions had have both their implicit attitude as well as their explicit attitude measured as part of the website attitude outcome condition. After the assignment into the different groups all of the participants were introduced to the pre-experimental material (T0), presented on a
computer screen, which consisted of being introduced to the hypothetical scenario.
The participants were to imagine they are Henny de Vries, a gender neutral name in The Netherlands also used in the study by Van Doorn and Hoekstra (2013), to enable each participant to identify sufficiently with the scenario, regardless of their gender. The
16 current occupation. Being an environmentally conscious citizen (according to the scenario) and just having been offed a raise; the participant has decided to reduce their carbon footprint on the world by purchasing a solar panel or photovoltaic installation for their home. They were shown the homepage of a mock website which they were told to have purchased from in the past and have been browsing the previous day for their research into solar panel
installations. This website page did not contain any manipulation material yet. The
participants were then asked to imagine that they are returning to the website the next day to continue their research after a night of consideration. They were then presented with the same mock website on which, depending on the experimental condition, one of the three
personalization stimuli was included using a banner ad (T1, see appendix for examples of
manipulation material).
In continuation, after the exposure to the website with the advertisements the
participants´ explicit and implicit website attitudes were measured using a questionnaire and single category implicit associations test (SC-IAT), respectively. Nosek et al. (2005) have argued that participant’s psychometric properties or the correlation between the measures are not influenced by the order of implicit and explicit measures. Bosson et al. (2000) have suggested, however, that assessing explicit attitudes first influences subsequent implicit attitudes and can artificially inflate correlations. Therefore implicit attitudes using the SC-IAT were measured first, followed by explicit attitude.
In conclusion, the participants were thanked and debriefed by conveying the
experiment´s intentions and the aim of the current research project, as well as by disclosing that the experimental material was entirely fictitious. On average it took about 20 minutes to complete an experiment for one participant.
17
Scales and Measurement
The measurement of perceived ad intrusiveness was achieved through the application of the 10 item (e.g., ‘I think this offer is disturbing’ or ‘I think this offer is uncomfortable’) scale, developed to measure perceived intrusiveness of advertisements by Mooradian (1996) and further adapted by Edwards et al. (2002), along a 7-point Likert-scale (see table 1). The scale was found to be a reliable measure of perceived intrusiveness by various studies investigation perceived intrusiveness (McCoy et al., 2008 & 2015; Van Doorn & Hoekstra, 2013).
Explicit attitude was assessed using a questionnaire constructed from 3 items from the 6 item scale to assess website attitude developed by Chen and Wells (1999) and similarly applied by Galletta et al. (2006) and Mcoy et al. (2016, see table 1). Only three of the items were chosen as only they can apply to the evaluation of non-interactive websites. The items chosen were (1) ‘Compared to other websites I would rate this as’ (from ‘one of the worst’ to ‘one of the best’), (2) ‘I would like to visit this Web site again in the future’ and (3) ‘I feel comfortable surfing this website’ (both from ‘entirely agree’ to ‘entirely disagree’) all of which were expanded from a 5-point to a 7-point bipolar Likert-scale by Huang and McMillan (2002). All items were found to reliable measures of the overarching concept of website attitude (Chen & Wells, 1999; Huang & McMillan, 2002), however another item measuring general attitude towards the website was included to add to the reliability of the scale.
Since the scale was modified quite heavily, a factor analysis to test its reliability was conducted. As the factors were expected to be independent of one another and only one factor was expected to be returned (due to the low amount of items) a maximum likelihood factor analysis with no rotation is chosen to assess the applicability and reliability of the construct. The KMO measure returned a ´middling’ value well above the necessary minimum value of
18 .5, KMO = .729 (Hutcheson & Sofroniou, 1999). Multicollinearity is certainly not present as the value was well above the minimum of 0.00001 (determinant = .146). The individual KMO values for the items are also all above .71, well above the accepted minimum of .5 (Field, 2013). The factor analysis was run and returned only one factor above the Kaiser criterion, which explained 68.64% of the total variance. This factor thus expresses the participant’s explicit attitude towards the website. The reliability of the single factor explicit website attitude, after items ‘I feel comfortable surfing this website’ and ‘I would like to visit this web site again in the future’ were reversed, was high (Cronbach´s α = .85). The final scale
employed in this study was implicit website attitude.
A single-category implicit association test (SC-IAT, N = 120, M = 93.54, SD = 3.8), based on the original IAT by Greenwald et al. (1998), but further adapted by Karpinski & Steinman (2006), was used to measure participant´s implicit attitude towards the website (see table 1). The error rates of the SC-IAT were, with an average error rate of 6.46%, well within the acceptable boundary of 20% also applied by Karpinski, and Steinman (2006). The SC-IAT is like any other SC-IAT, a dual categorization task. Participants are asked to quickly
categorize a word or an image stimulus into the positive or negative categories by pressing the “e” or “i” letter on the computer keyboard over the course of two stages with altering
combinations of the attitude objects and target words. Participants will see three types of stimuli, divided into two sets. (1) Pictures of the website used in the study (see image 6), as well as (2) negative and positive words associated with advertising, derived from the intrusiveness measurement scale (table 1) by Edwards et al. (2002) and Mooradian (1996), specifically developed for this study. The scale from which the target words are derived was found to be a reliable measure of perceived intrusiveness by various other studies (McCoy et al., 2008; Van Doorn & Hoekstra, 2013). It is therefore argued that the vocabulary used in the scale accurately identifies ones disposition to advertising and because the participant’s
19 attitude towards the website is a direct result of the advertisement and the intrusiveness of it, the vocabulary used is also appropriate in describing attitude towards the website within the SC-IAT. This is reflected in the good error rates of the test. The final, most relevant output of the SC-IAT for this study is the D-Score. It describes the participant´s average latency in performing the tasks during the SC-IAT, it ranges from -2 to 2 and constitutes the second outcome variable of this study, implicit website attitude.
The reasoning behind this choice of this implicit measurement tool lies in the nature of the dependent variable which will be examined in this study. A traditional IAT requires two contrast categories or attitude objects (in this case, images of the website are one attitude object), which are compared and set against the target words associated with the attitude object (see ‘implicit attitude’ in table 1). A complimentary category to that of the website images is needed. An identifiable duality is thus required between the one attitude object (website images) and its complimentary category as they, together with the target words must be definitively assigned to either the positive or negative category (Karpinski & Steinman, 2006). What is a natural complement to a website and how does one define whether one website should be assigned to the ‘good’ or ‘bad’ category in general, and especially within the 1.5 second timeframe, the small picture size and low resolution of the IAT? Using the SC-IAT this problem can be addressed. It only requires one attitude object which must be
assigned to either the ‘good’ or ‘bad’ categories. Thus, pictures of websites can be shown without the need to identify small intricacies within them. All that is necessary of the participants is to see that the image depicts a website.
20
Table 1 Scale Items
Construct Items Mean SD Source
Intrusiveness I think this offer is disturbing 3,9 1.672
Measured on 7 point Likert-scales by agreeing to the following statements from “not at all” to “completely”
I think this offer is obtrusive 3,97 1.523 Edwards et al. (2002) and Mooradian (1996)
I think this offer is alarming 3,45 1.665 I think this offer is irritating 3.98 1.647 I think this offer is annoying 4.08 1.757 I think this offer is uncomfortable 3.88 1.524 I think it is uncomfortable that personal
information is used in this offer
4.7 1.622
The supplier knows a lot about me 4.48 1.353 This offer gives me an uneasy feeling 4.07 1.494 This offer gives me an unsafe feeling 3.78 1.625
Explicit Attitude Compared to other websites I would
rate this as:
4.25 1.102
Measured on 7-point Likert-scales. Subjects respond to the statement with from “one of the worst” to “one of the best” or from “entirely agree” to “entirely disagree” with the statement
My overall evaluation of the website was:
4.09 1.366 Chen & Wells (1999), Galletta et al. (2006) and Mcoy et a. (2016)
I would like to visit this Web site again in the future
3.76 1.36
I feel comfortable surfing this website 3.89 1.389
Implicit Attitude Disturbing - Delightful
Single category implicit association test (SC-IAT), two types of stimuli with two sets of items (good and bad target words) for one and one set for the other. Second type of Items are images, see image 5. Implicit Attitude
Alarming - Comforting
Adapted from Van Doorn & Hoekstra, 2013 Obtrusive –Unobtrusive Irritating – Calming Annoying – Helpful Uncomfortable – Pleasant Vulnerable - Protected
21
Results
In order to verify whether the stimuli presented as the independent variable in the current experiment were indeed effective and worked as intended a manipulation check using a one-way ANOVA, based on the mean score of intrusiveness and various planned contrasts was conducted. The test examined the differences between the three conditions of the independent variable. Condition 1 represented browsing history only (M = 3.59, SD = 1.11), condition 2 browsing history and name (M = 4.72, SD = 1.14) and condition 3 browsing and transaction history (M = 4, SD = 1.34). The analysis revealed that there was a significant difference between the groups, F(2, 117) = 9, p < .001, ω = .34. Planned contrasts revealed that including any type of personalization in addition to browsing history significantly increased intrusiveness, t(117) = 3.31, p = .001, r = .29, compared to only using browsing history. Furthermore, including name personalization significantly increased intrusiveness compared to using transaction information, t(117) = -2.67, p = .009, r = .24.
The first hypotheses of this study explored the influence of intrusiveness as a mediator between the independent variable degree of personalization and the dependent variables explicit and implicit website attitude respectively. To test the first set of hypotheses as well as assess possible covariate influences among the control variables, age, gender, and education level on the dependent variables and the mediator perceived intrusiveness, a mediation analysis using the PROCESS mediation, moderation and conditional process analysis add-on for SPSS, developed by Preacher and Hayes (2008; or Hayes & Matthes, 2009) was
conducted. The model used was the standard simple mediation model (model 4). It should be noted that for this analysis the certain relationship directions are reversed due to utilizing a recoded version of the independent variable in which condition 1 (browsing history only) is
22 the reference category by assigning it value 3. Conditions 2 (browsing history and name) and 3 (browsing and transaction history) were recoded to conditions 1 and 2 respectively.
Regarding the covariates; degrees of personalization were the independent variable, perceived intrusiveness the mediator, age, gender and education the covariates and explicit website attitude the dependent variable. There was no significant relationship of any of the covariate variables on perceived intrusiveness. Neither for age, F(1, 114) = .19, p = .66, nor gender, F(1, 114) = .003, p = .96, nor level of completed education, F(1, 114) = .09, p = .77. There also was no significant relationship of any of the covariate variables on explicit website attitude, with the exception of age F(1, 114) = 4.95, p = .03. But neither gender, F(1, 114) = .13, p = .72, nor level of completed education, F(1, 114) = .13, p = .72 had a significant influence.
The first main analysis (for H1a), as stated, had the dependent variable degree of personalization, the mediator perceived intrusiveness and explicit website attitude as the outcome, in addition to the control variables. The analysis showed that there was an indirect effect of degree of personalization on explicit website attitude through perceived
intrusiveness, b = .27, BCa CI [-0.430, -0.142]. This is supported by the completely
standardized effect PM = .21, 95% BCa CI [0.111, 0.312]. There was no direct effect between degree of personalization and explicit website attitude, b = .15, p = .14, when intrusiveness was present in the model. However, there was a significant effect of degree of personalization on perceived intrusiveness, b = .-.55, p < .001, and there was a significant effect of perceived intrusiveness on explicit website attitude, b = -.49, p < .001. These results indicate that Intrusiveness indeed mediates the relationship between personalization and explicit website attitude (figure 2 shows the relevant mediation model). More specifically, as the degree of personalization is heightened, intrusiveness increases (direction of effect is reversed due to a
23
Figure 2 Model of relationships for H1a
change in reference category) and as a result explicit website attitude decreases, which supports H1a.
For the second main analysis the dependent variable degree of personalization, the mediator perceived intrusiveness, the outcome implicit website attitude (D-score), as well as the control variables age, gender, and education level were included. There was no significant relationship of any of the covariate variables on the outcome implicit website attitude. Neither for age, F(1, 114) = .17, p = .66, nor gender, F(1, 114) = .004, p = .96, nor level of completed education, F(1, 114) = .05, p = .77.
Similarly as with H1a, the results of the mediation analysis for the second part of the first hypothesis (H1b) indicated that there was an indirect effect of degree of personalization on implicit website attitude through perceived intrusiveness, b = .026, BCa CI [0.001, 0.064]. Compared to the previous analysis the effects size can be interpreted as quite small however,
PM = .06, 95% BCa CI [0.002, 0.153]. There was also a direct effect between degree of personalization and implicit website attitude, b = .08, p = .049, as well as no significant effect of perceived intrusiveness on implicit website attitude, b = -.05, p = .06. These results show that there was also a mediating role of perceived intrusiveness between higher degrees of personalization and the outcome variable implicit website attitude (figure 3 shows the relevant mediation model). H1b has thus been supported.
Degrees of Personalization
Perceived Intrusiveness
Explicit Website Attitude
Direct effect: b = .15, p = .14
Indirect effect: b = .27, BCa CI [-0.430, -0.142]
24
Figure 3 Model of relationships for H1b
The final hypothesis investigates the relationship between the two outcome variables implicit and explicit website attitude. H2 stated that the expected relationship would reveal an overall lower implicit website attitude than explicit website attitude. To investigate the
differences between the participant’s implicit and explicit attitudes a correlations analysis was run to see how the implicit d-score measure of website attitude and the overall mean explicit website attitude were related. The analysis revealed that, as expected, there was no significant correlation between the two dimensions, r = .16, p (2-tailed) = .08. Indicating that they do not behave in conjunction with one another, as is necessary for H2 to be the case. The two
dependent variables were simply analyzed according to their means. Implicit website attitude, (M = 1.96, SD = .33) represents a .02% negative implicit evaluation of the website and
explicit website attitude (M = 3.83, SD = 1.08), represents a .05% positive explicit attitude towards the website as a result of the advertisement, which together represent an insignificant effect F(1, 117) = 0.02, p = .9. The participant´s implicit website attitude was not significantly higher than their explicit website attitude, thus there is no support to be found for H2.
Degrees of Personalization
Perceived Intrusiveness
Implicit Website Attitude
Direct effect: b = .08, p = .049
Indirect effect: b = .026, BCa CI [0.001, 0.062]
25
Discussion
The aim of this study was to investigate the mediating influences of perceived advertising intrusiveness between differing degrees of personalization of the advertisement banner and both implicit and explicit website attitude, as well as investigate the overarching differences of implicit and explicit website attitude. To accomplish this, a scenario-based experiment within the photovoltaic industry was conducted in which the relationships between three degrees of personalization, perceived intrusiveness and both implicit and explicit website attitude were investigated. Although this kind of advertising personalization and the role of intrusiveness has been a topic of previous literature (Van Doorn & Hoekstra, 2013), this is the first study to (1) examine whether this effect holds true for attitude towards the website and to (2) also include another type of outcome measurement (an implicit
association test), past the traditional use of self-report measurements. This study (3) also sought to fill the gap in research on the disparity between the implicit and explicit attitudinal outcomes in the field of personalized advertising and examine them in relation to
intrusiveness.
The analyses revealed that there were indeed mediating influences of intrusiveness to be found between degrees of personalization and both explicit and implicit website attitude. Thus, in support to the findings of Van Doorn and Hoekstra (2013), utilizing additional distinguishing information in advertising increases the participant’s perceived intrusiveness which has a negative influence on both their explicit and implicit attitude towards the website. This effect was far smaller in the case of implicit website attitude, indicating that implicit website attitude is much less influenced by the construct of intrusiveness. This could be the result of external influencing experimental artifacts such as impression management or demand characteristics, which influence the participants on a subconscious level, mitigating
26 the reduction in attitude as a result of intrusiveness. The results show however, that the
implicit attitude dimension is still influenced by intrusiveness and implicit advertising effects as a result of it should not be discounted.
Another major contribution of the present study was to examine whether, as a result of several factors such as aid in goal pursuit, evaluation apprehension or impression
management, implicit attitude towards the website would be more positive than explicit attitude across all conditions. This relationship was not found to be present. Implicit attitude towards the website was not significantly more positive than implicit attitude.
There are three main limitations of this research project which, compounding on one another could have also contributed to the inconclusiveness of the results for the third hypothesis and should be addressed when other studies, similar in scope and aim, consider further research in this area. A major limitation which is commonly named in studies using implicit measures is that (1) conclusions drawn from the IAT could, at least partly, stem from factors other than the expected affective influences (Karpinski & Steinman, 2006). This could certainly be the case for this study as (2) demand characteristics could have occurred, when conducting the SC-IAT together with the self-reported measurements. This is because priming of the research aim can occur towards either the IAT task or the self-reported
measurements, depending on which of those is conducted first (Karpinski & Steinman, 2006). The final limitation of this study was (3) an experiment environment which was not quite ideal. It was necessary to utilize an openly accessible area of the University of Amsterdam, where noise and distractions were possible. These could have influenced the sensitive response time measurements of an IAT. An environment more suited for concentration intensive tasks could be beneficial for future studies in this field.
In conclusion, there are several important findings from this study which are beneficial for online advertising practitioners to acknowledge. Firstly, using name personalization in
27 addition to browsing history caused the highest positive influence on the participant’s
perceived intrusiveness, compared to browsing history only on the one hand, but also browsing and transaction history on the other. This indicates that name personalization’s, being a very personal and starkly distinguishing feature, could quickly backfire and cause negative reactions towards the website. Secondly, user’s attitude towards the website is a relevant outcome to consider as it is significantly influenced by feelings of intrusiveness. If practitioners engage in advertising personalization on their website, care should be taken as too high of a degree of personalization can have a negative effect on their host site. Finally, in line with Van Doorn and Hoekstra (2013) caution needs to be taken, specifically when
including name personalization in the persuasive message as it will significantly increase intrusiveness and as a result also negatively affect both the user’s implicit and explicit attitude towards the website.
28 Appendix
Examples of the Manipulation Material
The following examples (images 1-4) represent the advertising banners created to be the manipulation material of the experiment. They were integrated into the websites presented to the participants, depending on the condition which the participant has been assigned to. All brand names, logos and slogans present in the original versions of the images have been changed or removed in order to not influence the participants during the experiment.
Image 1 Ad example for the ‘browsing history only’ condition
29 Image 5 represents examples of those images that were used as second type stimulus material in the SC-IAT. They had to be sorted into either the good or bad category together with the target words, depending on the current test stage.
Image 3 -5 Second set of the SC-IAT stimulus material
Image 6 represents the website made for the experiment together with the ‘browsing history only’ banner ad condition (framed). The participant’s first exposure to the website, occurring after they have been asked to immerse themselves in the hypothetical situation, will not yet contain an ad in the indicated position. For the second exposure, when they return to the website the following image or one with another personalization condition will be chosen. The positions of all the ads have been chosen to be in place where there were either ads or news already present on the original version.
30
References
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human
Decision Processes, 50(2), 179–211.
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), 34–49. Bailey, B. P., Konstan, J. A., & Carlis, J. V. (2001). The effects of interruptions on task
performance, annoyance, and anxiety in the user interface. In M. Hirose (Ed.),
Proceedings of 2001 Human-Computer Interaction, Interact 2001 Conference, 593–
601.
Bang, H. & Wojdynski, B. W. (2016). Tracking users' visual attention and responses to personalized advertising based on task cognitive demand. Computers in Human
Behavior, 55, 867–876.
Bauer, R. A. & Greyser, S. A. (1968), Advertising in America: The Consumer View, Boston: Harvard University.
Bennett, R. J. & Robinson, S. L. (2000). Development of a Measure of Workplace Deviance.
Journal of Applied Psychology, 85(3), 349.
Bosson, J. K., Swann, W. B. & Pennebaker, J. W. (2000). Stalking the perfect measure of implicit self-esteem: The blind men and the elephant revisited? Journal of Personality
and Social Psychology, 79, 631–643
Cases, A.-S., Fournier, C., Dubois, P.-L. & Tanner, J. F. (2010). Web site spill over to email campaigns: The role of privacy, trust and shoppers' attitudes. Journal of Business
31 Chellappa, R. K. & Sin, R. G. (2005). Personalization versus Privacy: An Empirical
Examination of the Online Consumer's Dilemma. Information Technology and
Management, 6(2-3), 181-203.
Chen, Q., & Wells, W. D. (1999). Attitude toward the site. Journal of advertising research, 39(5), 27–38.
De Keyzer, F., Dens, N. & De Pelsmacker, P. (2015). Is this for me? How Consumers Respond to Personalized Advertising on Social Network Sites. Journal of Interactive
Advertising, 15(2), 124–134.
Diao, F. & Sundar, S. S. (2004). Orienting response and memory for web advertisements: exploring effects of pop-up window and animation. Communication Research, 31(5), 537–567.
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), 83–95.
eMarketer (2017). For Native Advertising, 2017 Looks Like a Year of Clarity, Two categories will emerge. Advertising & Marketing. Retrieved from https://www.emarketer.com/ Article/Native-Advertising-2017-Looks-Like-Year-of-Clarity/1015122.
Fortin, D. R. & Dholakia, R. R. (2005). Interactivity and vividness effects on social presence and involvement with a web-based advertisement. Journal of Business Research, 58(3), 387–396.
Galletta, D. F., Henry, R. M., McCoy, S. & Polak P. (2006). When the wait isn’t so bad: the interacting effects of web site delay, familiarity, and Breadth. Information Systems Research. 17(1), 20–37.
32 evaluation: An integrative review of implicit and explicit attitude change.
Psychological Bulletin, 132, 692–731.
Gawronski, B., & Strack, F. (2004). On the propositional nature of cognitive consistency: Dissonance changes explicit, but not implicit attitudes. Journal of Experimental Social
Psychology, 40, 535–542.
Goldfarb, A., & Tucker, C. (2011). Online display advertising: targeting and obtrusiveness.
Marketing Science, 30(3), 389–404.
Gregg, A. P., Seibt, B., & Banaji, M. R. (2006). Easier done than undone: Asymmetry in the malleability of implicit preferences. Journal of Personality and Social Psychology, 90, 1–20.
Greenwald, A. G., Mcghee, D. E., Schwartz, J. L. K. & Kruglanski, A. W. (1998). Measuring Individual Differences in Implicit Cognition: The Implicit Association Test. Journal of
Personality and Social Psychology, 74(6), 1464–1480.
Greenwald, A. G., Nosek, B. A., Banaji, M. R. & Devine, P. (2003). Understanding and Using the Implicit Association Test: I. An Improved Scoring Algorithm. Journal of
Personality and Social Psychology, 85(2), 197–216.
Häfner, M. & Trampe, D. (2009). When thinking is beneficial and when it is not: The effects of thin and round advertising models. Journal of Consumer Psychology, 19(4), 619– 628.
Hayes, A. F. & Matthes, J. (2009). Computational procedures for probing interactions in OLS and logistic regression: SPSS and SAS implementations. Behavior Research Methods, 41, 924–936.
Hwang, J. S., & McMillan, S. J. (2002). The role of interactivity and involvement in attitude toward the Web site. Paper presented at the American Academy of
33 IAB & Deloitte (2016). IAB report on Online Advertising Spend, The Netherlands 2015. IAB
report.
Internet Retailer (2013). Site personalization has room to grow among top 500 retailers. Retrieved from https://www.internetretailer.com/2013/05/15/site-personalization-has-room-grow-among-top-500-retailers.
Jack W. Brehm, A Theory of Psychological Reactance, Academic Press, New York, 1966.
Karpinski, A., & Steinman, R. B. (2006). The single category implicit association test as a measure of implicit social cognition. Journal of Personality and Social Psychology, 91, 16–32.
Li, H., Edwards, S. M., & Lee, J. (2002). Measuring the Intrusiveness of Advertisements: Scale Development and Validation. Journal of Advertising, 31(2), 37–47.
Lin C. A. & Kim, T. (2016). Predicting user response to sponsored advertising on social media via the technology acceptance model. Computers in Human Behavior, 64, 710– 718.
Matthes, J. Schemer, C., & Wirth, W. (2007). More than meets the eye: Investigating the hidden impact of brand placements in television magazines. International Journal of
Advertising, 26(4), 477–503.
McCoy, S., Everard, A., Polak, P. & Galletta D. F. (2008). An Experimental Study of Antecedents and Consequences of Online Ad Intrusiveness. International Journal of
Human–Computer Interaction, 24(7), 672–699.
McCoy, S., Everard, A., Galletta D. F. & Moody, G. D. (2015). Here we go again! The impact of website ad repetition on recall, intrusiveness, attitudes, and site revisit intentions. Information & Management, 54(1), 14–24.
34 Borgatta & W.W. Lambert (Eds.), Handbook of personality theory and research (1130–1187). Chicago: Rand McNally.
Morimoto, M., & Chang, S. (2006). Consumers' attitudes toward unsolicited commercial e- mail and postal direct mail marketing methods: intrusiveness, perceived loss of control, and irritation. Journal of Interactive Advertising, 7(1), 8–20.
Mooradian, T. A. (1996). Personality and ad-evoked feelings: the case for extraversion and neuroticism. Journal of the Academy of Marketing Science, 24(2), 99–109.
Nosek, B. A., Greenwald, A. G., & Banaji, M. R. (2007). The Implicit Association Test at age 7: A methodological and conceptual review (265–292). In J. A. Bargh (Ed.),
Automatic processes in social thinking and behavior. Psychology Press.
O'Donnell, K. & Cramer, H. (2015). People's Perceptions of Personalized Ads. Proceedings
of the 24th International Conference on world wide web, 1293–1298.
Okazaki, S., Li, H., & Hirose, M. (2009). Consumer privacy concerns and preference for degree of regulatory control. Journal of Advertising, 38(4), 63–77.
Orne, M. T. (1962). On the social psychology of the psychological experiment: With particular reference to demand characteristics and their implications. American
Psychologist, 17(11), 776–783.
Pavlou, P. & Stewart, D. (2002). From consumer response to active consumer: Measuring the effectiveness of interactive media. Journal of the Academy of Marketing Science, 2002, 30(4), 376–396.
Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research
35 PwC & IAB (2016). IAB internet advertising revenue report. 2015 full year results. An
industry survey conducted by PwC and sponsored by the Interactive Advertising Bureau.
Reid, G. B., & Nygren, T. E. (1988). The subjective workload assessment technique: A scaling procedure for measuring mental workload. In Hancock P. A. & Meshkati N. (Eds.), Human mental workload, Amsterdam: North-Holland, 185–215.
Rosenberg, M. J. (1969). The conditions and consequences of evaluation apprehension. In R. Rosenthal & R. L. Rosnow (Eds.), Artifact in Behavioral Research, (279 -349). New York: Academic Press.
Strick, M. A., Baaren, R. B., Holland, R. W. & Knippenberg, A. F. M. (2011). Humor in Advertisements Enhances Product Liking by Mere Association. Psychology of
Popular Media Culture, 1, 16–31.
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, 15–22.
Tam, K. Y., & Ho, S. Y. (2006). Understanding the impact of web personalization on user information processing and decision outcomes. MIS Quarterly, 30(4), 865–890. Taylor, S. E., & Brown, J. D. (1984). Illusion and well-being: A social psychological
perspective on mental health. Psychological Bulletin, 103, 193–210.
Tedeschi, J . T., Schlenker, B. R. & Bonoma, T. V. (1971). Cognitive dissonance: Private ratiocination or public spectacle? American Psychologist, 26(8), 685- 695.
Tomaseti E., Ruiz, S. & Reynolds, N. (2009). Flow and Attitude Toward the Website on the Evaluation of Products Present By Means of Virtual Reality: a Conceptual Model. In A. L. McGill and S. S. Duluth (Eds.), NA - Advances in Consumer Research (916-918). MN: Association for Consumer Research.
36 Van Doorn, J., & Hoekstra, J. C. (2013). Customization of online advertising: The role of
intrusiveness. Marketing Letters, 24(4), 339—351.
Wilson, T. D., Lindsey, S., & Schooler, T. Y. (2000). A model of dual attitudes.
Psychological Review, 107, 101–126.
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), 503–514.
Robin C. Voboril Hammersley, Amsterdam the 03.02.2017
