The effect of tailoring online narrative advertising and the role of ad skepticism : does tailoring narrative advertising positively affects persuasive outcomes and what is the moderating effect of ad skepticism?

THE EFFECT OF TAILORING ONLINE NARRATIVE

ADVERTISING AND THE ROLE OF AD SKEPTICISM

Does tailoring narrative advertising positively affects persuasive outcomes and what is the moderating effect of ad skepticism?

Name: Selman Aqiqi
 Student number: 10475199

Email: selman.aqiqi@student.uva.nl Graduate School of Communication

Track: Persuasive Communication Master’s Thesis

Supervisor: dr. Stephanie Welten Date of completion: 23/06/2015

Abstract

This study aims at investigating the effect of tailoring narrative advertising on ad attitude, brand attitude, behavioral intention and behavior. Tailored narrative advertising is compared to non-tailored narrative advertising and argumentative advertising (control). More precisely, tailoring entails matching six characteristics of readers to the protagonist in the narrative advertising. By increasing the reader-protagonist similarity, self-referencing (mediator) is encouraged which should result in more favorable persuasion outcomes. Furthermore, the moderating role of ad skepticism on tailoring is investigated. Results of an online experiment among 150 participants showed that, compared to a non-tailored ad and an argumentative ad, a tailored ad leads to significantly more favorable brand attitude and behavior, but not ad attitudes and behavioral intentions. Implications for practitioners are provided. In contrast to prior research, there is no clear evidence for the mediating role of self-reference. Results do yield insights into the role of ad skepticism. The study provides a number of promising leads for further research.

Keywords: narrative advertising, tailoring, reader-protagonist similarity, self-reference, ad skepticism, online advertising, ad attitude, brand attitude, behavioral intention, behavior

Introduction

Humans have used the power of stories to convey wisdom and knowledge from one generation to the other. A great part of the information that is acquired in a lifetime comes in the form of stories, otherwise called narratives (Boyd, 2009). Narratives not only lend themselves to transmit information, but are also often used to change beliefs and attitude. Think of the example of parents that use scary stories to teach their children why not to play with fire. Beyond

childhood, individuals use stories to make sense of the world around them, including experiences with products and brands (Adaval & Wyer, 1998). The potential of narratives to market these products and brands has not gone unnoticed to the advertising world. Advertisers have been keen on employing the persuasive effect of narratives to tell stories about their brand (Escalas, 1998). This is not miraculous since "people think narratively rather than argumentatively or

paradigmatically" (Weick, 1995, p.127). Indeed, much of the information that humans receive is stored and retrieved in memory in the form of episodes or stories (Schank, 1990).

Narratives are thus important in human persuasion. A thorough scientific understanding of how they impact individuals is however still limited. Green (2008) points towards the importance of uncovering the active ingredients of narrative persuasion. In other words, it is crucial to understand what does and what doesn't work. One of the most promising venues in narrative persuasion is the tailoring of a narrative to the individual. This is the adaptation of a persuasive message to match certain characteristics of its receiver such as gender, age, ethnicity or personal traits (Kreuter & Wray, 2003).Tailoring narrative persuasive messages has been the subject of research, but only in the context of health education and with only a few adaptation points. Adaptation points are the parts of the story that are tailored to the reader. Nonetheless,

this research has been promising because it shows that tailored health messages are more effective than one-size-fits-all messages because they allow readers to relate more to the story (Noar, Benac & Harris, 2007; De Graaf, 2014). Replicating these findings to an advertising environment poses a challenge because consumers are then faced by a message sender that is driven by commercial rather than public health motives. If aware of these motives, consumers might raises their defenses, resulting in tension between persuasion through tailoring and resistance to manipulation. Communication research should therefore take up the challenge and investigate to what extent the potential of tailoring narratives can be unlocked for consumer advertising as well. One of the areas where this challenge can take place is online advertising. Online advertising has radically changed since the rise of social media over the course of the last decade. With the explosive growth of social media use, more and more personal data is collected from users. This data is stored and analyzed by companies which allows them to aggregate highly personalized profiles of each individual (Torres, 2012). Consequently, each user can be targeted with online advertising according to the consumer segment to which they belong (Ahmad, 2003). The increasing level of detail of collected personal information nowadays allows going beyond segments; online advertising can be tailored to the individual level. This type of tailoring may hold great practical benefits for marketers, but also valuable insights for the field of communication in general as tailoring holds great persuasive potential. Studies on computer-tailored persuasion are however still divergent in their findings with regard to the effectiveness of this type of tailoring (Dijkstra, 2008). This study hopes to contribute by filling this knowledge gap on tailoring narratives and providing practical insights for

An important issue that arises from the use of personal data to target consumers with advertising is privacy. As companies collect more and more data, concern over privacy among social media users is growing (Torres, 2012). Consumers are made aware through terms and conditions that their data is sometimes collected for advertising purposes. However, this knowledge does not prevent feelings of intrusion into the private sphere. Consumers are

becoming more aware of how they are being targeted with ads. They have developed knowledge on these persuasion techniques and they have become wary of advertising (Friestad & Wright, 1994). Obermiller and Spangenberg (1998) have the coined the term ad skepticism to describe the extent to which consumers engage in critical analysis of persuasive messages and

consequently accept or refuse the content of this message. Consumers that are highly skeptical are also more difficult to persuade (Obermiller and Spangenberg, 1998). Therefore, it is important to investigate if and how the potential advantage of tailoring over non-tailoring of messages can be sustained when targeting highly skeptical consumers with online narrative ads.

In sum, the aim of this study is to examine the effectiveness of tailored advertising narratives that employ multiple adaptation points in comparison to one-size-fits-all narratives and non-narrative advertising. The effect of ad skepticism on tailoring narrative advertising is also investigated. This leads to the following research question:

RQ: To what extent is tailored online narrative advertising more persuasive than one-size-fits-all narrative advertising and non-narrative advertising and to what extent is this effect moderated by ad skepticism?

Theoretical Background

A narrative generally consists of "an identifiable beginning, middle and end that provide information about scene, character and conflict; raises unanswered questions or unresolved conflict; and provides resolution" (Hinyard & Kreuter, 2006, p.778). When readers are presented with a narrative they sometimes enter a mental state called transportation in which they become unaware of the world around them. They are transported into the world of the story and

experience the events taking place there (Green & Brock, 2000). While being transported, readers who are immersed in the story will be less likely to question the facts presented there. In other words they will refrain from counter-arguing (Green & Brock, 2000). The lessons learned from those narratives can therefore impact the reader and his or her beliefs (Prentice, Gerrig & Bailis, 1997). For this reason transportation can be a powerful tool for persuasion by reducing counter-arguing and this potential has been embraced in advertising.

Narrative advertising uses the power of transportation by presenting to consumers a story involving a character that will go through events in which the benefits of a product will help him achieve his goals (Ching, Tong, Chen & Chen, 2013). Think of a television commercial for cars; it usually involves a male character who will get his hands on the newest model from car brand X and experience ultimate pleasure when driving it. Narrative advertising lets consumers imagine themselves using a product and the positive consequences of that use (Pagdet & Allen, 1997). When the consequence of this consumption is positive (which it often is), consumers will evaluate the product and the advertisement more positively as well (Wentzel, Tomczak &

Hermann, 2010; Chang, 2009). Research further shows that narrative advertising that allows consumers to get more transported leads to more positive attitudes towards the brand and the advertisement (Chang, 2009).

In contrast, non-narrative advertising does not employ story-form; there are no characters, no plots and no chronology (Deighton, Romer & McQueen, 1989). It presents products in a factual manner and is organized in a logical and rational way. Think of an

advertisement for men's disposable razors, these often praise the sharpness of the blades and the smooth shave, for example. Arguments are at the base of this type of advertisement, which is why it is also referred to as argumentative advertising (Pagdet & Allen, 1997). Narrative and argumentative advertising are different mainly in the way its information is processed.

Consumers try to process that information in a way that is natural and familiar to them. When confronted with a narrative, they relate elements of the story in a meaningful way (Wentzel et al., 2010). They try to make sense of the events involving characters and their goals, and in the case of narrative advertising, also products, brands and their benefits. Since consumers already think narratively it is easier for them to process product information and relate to brand experiences if both come in story form. This in turn, will lead to more favorable brand attitudes than in the case of argumentative advertising (Escalas, 2004; Adaval & Wyer, 1998).

Consumers often try to relate the information contained in advertising to their own reality and frequently judge it by the extent to which it is relevant to themselves, a process called

self-referencing (Debevec & Iyer, 1988). When individuals engage in self-self-referencing they are

knowledge (Burnkrant & Unnava, 1989). Research on self-referencing shows that persuasive narratives that allow individuals to relate to themselves enjoy the advantage of higher

transportability and indeed more favorable attitudes towards a product and story-consistent beliefs (Ching et al., 2013; Escalas, 2007; De Graaf, 2014). In addition, De Graaf (2014) showed that self-referencing acted as a mediator for story-consistent belief. These studies did not

specifically investigate tailoring narrative advertising in an online environment which - as mentioned earlier - comes with its own set of challenges. Self-referencing in narratives can be achieved in many ways, one of them being similarity between the reader and the main character of the story. For example, health messages in story format that included reader-character

similarity enjoyed more effective persuasion (De Graaf, 2014). Reader-character similarity can refer to many things; demographic variables, outer appearance, personal and psychological traits (Hoffner & Cantor, 1991). It can be achieved by adapting the story to the reader. This can include the tailoring of the main character or protagonist of the story to the reader. Research has showed that narrative advertising that features some similarity between reader and protagonist has proven to be more effective than advertising that is not tailored to readers. (Andsager, Bemker, Choi & Torwel, 2006; Hende, Dahl & Schoormans, 2012). Theoretically, in the field of online

advertising, tailoring can be achieved by matching a large part of the stored personal information to the narrative advertising that is presented to internet users. The result would be that each user reads a narrative advertising featuring a protagonist that resembles them a great deal.

In summary, tailored narrative advertising is expected to lead to more positive persuasion outcomes than one-size-fits-all and argumentative advertising. As most previous studies looked at attitudes towards the brand and advertisement only, it is the goal of this study to extend that

research into more concrete persuasion outcomes. It will also investigate the effect of tailoring on the intentions of performing ad consistent behavior and actual behavior by consumers.

Additionally, the mediating effect of self-referencing will be investigated. Based on this information the following predictions are formulated:

H1: Online narrative advertising that is tailored on multiple points will lead to more positive ad

attitude, brand attitude, behavioral intention and behavior than one-size-fits-all narrative advertising and argumentative advertising.

H2: The effect of tailoring on ad attitude, brand attitude, behavioral intention and behavior will

be mediated by self-referencing.

Ad Skepticism

Ad skepticism is the extent to which consumers show general disbelief of claims that are made in advertising (Obermiller & Spangenberg, 1998). Consumers are confronted with advertising every day; some of them develop during their lifetime the ability to recognize these persuasion

attempts and also how to cope with them. This phenomenon is captured in the framework of the Persuasion Knowledge Model by Friestad and Wright (1994). The model states that consumers acquire knowledge on persuasion techniques which is then sometimes activated when they aware of a manipulative purpose such as in advertising. This knowledge does not automatically lead to resistance to the persuasion, however. Consumers will typically evaluate a number of factors such as trustworthiness of the source or the strength of the argument. After this evaluation, they may or may not engage in resistance towards advertising (Campbell & Kirmani, 2000). Ad skepticism is one strategy of resistance - it moderates the response towards the advertising by

lowering brand attitude and consequently purchase intention (Obermiller & Spangenberg, 2000). Ad skepticism has also been shown to moderate persuasion through narratives by lowering transportation and lowering brand evaluations (Escalas, 2007). However, that study did not include tailoring and self-referencing; but its findings should have implications on tailoring narrative advertising in an online context where privacy concerns are prevalent. On the one hand, consumers who are aware that advertisers are purposely presenting them a personally tailored narrative are expected to show less positive persuasion outcomes than if they were presented with a one-size-fits-all narrative. On the other hand, consumers who are less aware of the manipulation should report more positive outcomes when seeing a tailored narrative than a one-size-fits-all message because they engage in self-referencing. This leads to the following prediction on the interaction effect between the type of ad and ad skepticism:

H3: Ad skepticism will moderate the effect of tailoring on ad attitude, brand attitude, behavioral

intention and behavior. Consumers who are highly skeptical towards ads will report less

positively on these outcomes if they see tailored advertising than one-size-fits-all advertising and argumentative advertising. Conversely, consumers who are less skeptical will report more

positively on the outcomes if they see a tailored advertising than a one-size-fits-all advertising and an argumentative advertising.

Conceptual Model

Figure 1 shows the conceptual model of this study in clarification of the relation between the independent, dependent, moderating and mediating variables.

Figure 1. Conceptual model.

Method

Design

In order to investigate the effect of different type of narrative advertising as described in the hypothesis, quantitative research methods were used. This study employed a three condition (type of ad: tailored, one-size-fits-all, argumentative) between-subject experimental design with one continuous observed variable (ad skepticism). Ad skepticism was then split at the median and used as dichotomous level variable (‘low’ vs. ‘high’). The experiment and data collection was done in an online environment with a questionnaire. This was done in order to attract a large number of participants. Furthermore, this online method facilitates the presentation of stimuli in a way that is similar to what Internet users are accustomed to see when they go online. This makes participating in the experiment easier, more true to life, and enhances external and ecological validity of the study.

Stimuli

In order to test the formulated hypotheses, three promotional emails were designed. These emails were presented as originating from a monthly newsletter of the fictional Travellers Magazine publication (see Appendix 1-3 for used stimuli). All three types of emails contained a

promotional message from a fictional airline named InterStar Airlines. This airline was devised for the purpose of this study. The choice for an unknown airline was done in order to prevent existing attitudes towards established companies to influence the results. In order to increase the credibility of the stimuli, InterStar Airlines was presented as a new company within the Sky Team. Sky Team is an established alliance of existing international airlines.

The promotional message in the tailored and one-size-fits-all conditions was in the form of an advertorial. The advertorial is presented as a story by a consumer about his experience with InterStar Airlines. The argumentative condition, however, featured a regular advertisement for InterStar Airlines. The difference in stimuli was necessary to increase external and ecological validity. Presenting a narrative within a traditional advertisement is unconventional and would possibly entail an unwanted negative reaction. This would give the argumentative condition an unfair advantage. However, for the sake of internal validity both the advertorial and the

advertisement were executed in a most similar possible way. Both used one image and text. The tailored narrative counted 158 words, the one-size-fits-all narrative counted 144 words and the argumentative condition counted 124 words. Finally, all stimuli contained the same amount and type of information albeit in different wording.

Both narrative conditions contained a consumer experience story. Based on existing consumer research on narrative persuasion a number of fixed story-elements were incorporated.

Firstly, a narrative must have a beginning, middle and ending that consists of episodes in which events takes place that allows causality (Escalas, 1998). Secondly, a narrative should feature one or more protagonists with one or more goals to achieve. These goals are achieved by

experiencing the benefits of a product or service (Van den Hende & Schoormans, 2012).

The tailored narrative advertising was presented as a story written in diary-format by one of the readers of the fictional Travellers Magazine. The story tells the events experienced by the protagonist and starts with planning a summer trip, via booking plane tickets, to the actual flights experience and ends with a final judgment of the total experience. Throughout the story, the positive aspects of the experience are being underlined to achieve persuasion effects. The final evaluation of the company and its service is positive and satisfying. The story includes six tailoring points where the protagonist shows similarities with participants. These similarities are gender, age, occupation and country of residence of the participant. Furthermore the story is adapted to feature the favorite continent and preferred type of vacation of each participant. This information is collected in the survey at the beginning of the experiment. The advertorial shows a picture of a traveler in an airport setting which is also adapted to gender. The logo of InterStar Airlines is also depicted.

The one-size-fits-all narrative condition also features a story in diary format. The content is identical to the tailored narrative except that there are no tailoring points. Just as with normal narrative advertising, the characteristics of the protagonist and travel wishes are fixed. Every participant in that condition was exposed to a story about Jack (aged 35) from Canada, a working man who wishes to have an adventure trip in Africa. These characteristics were randomly chosen to imitate a real life narrative ad.

The argumentative condition features a typical advertisement which sums up all the positive aspects of the brand and product that are mentioned in the narrative conditions. There is however no protagonist and no storyline to be detected.

Participants

In order to study the effect of tailoring across all ages, genders and other personal characteristics it was necessary to accommodate a large group of participants. The target population of this study was therefore everyone aged 18 years and older. The minimum age of 18 was chosen due to practical reasons as enrolling minors requires complex procedures. Participants were recruited through the social networking site Facebook and via email.

A total of N = 150 participants completed the survey. There were 90 females (60%), 60 males (40%) and the average age was 27.10 (SD = 8.25). One percent of participants finished some kind of elementary education, 3% finished high school, 33% attained a Bachelor’s level diploma, 61% attained a Master degree diploma and 3% finished some kind of post-Master program.

Procedure

Participants who followed the invitation link were presented with a short explanation of the study and contact information. After giving consent to the study, participants were randomly assigned to one of the three conditions (tailored narrative, one-size-fits-all narrative and argumentative). Each condition contains demographic questions, questions on personal travel preferences and a filler task. Participants were then presented with a short explanation in which they were asked to imagine going through a series of events involving surfing the web. They were then presented

with a stimulus according to the condition. Participants were then guided through a number of measures; these are described in more details in the ‘Measures’ section. Ad skepticism was measured towards the end of the survey. At the very end, participants were debriefed with an expression of gratitude and the note that InterStar Airlines is a fictional brand designed for the study. A field was left open for suggestions and comments.

Measures

Ad attitude

Ad attitude measures how participants evaluate the advertisement itself. It was measured by adopting a scale from Smith et al. (2007). The semantic differential scale consisted of four items. The items were ‘bad/good’; ‘unfavorable/favorable’; ‘not likable/likable’; ‘unpleasant/pleasant’ and they were measured by using a 7-point scale (‘1’ being most negative and ‘7’ most positive).

Items on this scale were tested for reliability. They had a Kaiser-Meyer-Olkin value of . 87, which proves the sampling adequacy. A principal component analysis (PCA) with Varimax rotation showed that all items formed a single unidimensional scale. Only one component was found with eigenvalues higher than Kaiser’s criterion of 1 (EV = 3.51, R2 _{= .87). Scale reliability}

was good (Cronbach’s α = .88) and the items were computed into a single scale called ad attitude.

Brand attitude

The other important persuasion outcome was attitude towards the brand. This is the extent to which receivers value the brand positively. In order to measure this, a 7-item semantic

differential scale was used (‘1’ being most negative and ‘7’ most positive). The scale included: ‘dislike very much/like very much’; ‘useless/useful’; ‘worthless/valuable’; ‘not important/ important’; ‘not beneficial/beneficial’; ‘not fond of/fond of’; ‘not enjoyable/enjoyable’. This is an existing scale adopted from Machleit and Wilson (1988).

This scale was also tested for reliability. The items in this scale had Kaiser-Meyer-Olkin value was .90 which verified the sampling adequacy. Principal component analysis with Varimax rotation showed that all 7 items formed one component with an Eigenvalue above 1 (EV = 4.98,

R2 _{= .71). Scale reliability was also good (Cronbach’s α = .93). The items were therefore}

combined to form a single scale called brand attitude.

Behavioral intention

Another persuasion outcome that was measured is the extent to which participant intend to show ad-consistent consumer behavior. This was measured with two statements that could be answered using a 7-point Likert scale with ‘1’ representing ‘very unlikely’ and ‘7’ representing ‘very likely’. The two items were: ‘Indicate how likely it is you would book a ticket from Interstar Airlines.’ and ‘Indicate how likely it is you would recommend InterStar Airlines to a friend.’ This scale was adopted from Lee, Bun and Browne (1995).

The two items measuring intention to purchase tickets from InterStar Airlines had a Kaiser-Meyer-Olkin value of .5 which is barely enough to verify sampling adequacy. Principal component analysis with Varimax rotation showed that the 2 items formed a single

unidimensional scale with one component with an Eigenvalue above 1 (EV = 1.75, R2 _{= .87). The}

scale had a Cronbach’s alpha value of .85 which is good, the two items were combined into a single scale named behavioral intention.

Behavior

Another measure probed the direct behavior of participants after seeing the stimuli. This

consisted of two questions. First, participants were asked to indicate on a 7-point Likert scale (‘1’ being ‘disagree’, ‘7’ being ‘agree’) to what extent they agreed with the following statement: ‘I am interested in receiving one email with interesting travel offers from InterStar Airlines.’ Second, participants were asked to choose whether they allowed InterStar Airlines to send them one promotional email, measured in a binary level variable (‘0’ being ‘no’, ‘1’ being ‘yes’).

Mediator

This study also aims to investigate the mediating role of self-referencing in tailored narratives. Therefore, an adaptation of the scale by Dunlop et al. (2002) was used. It consists of four-item questionnaire which were adapted to the subject of traveling. The items are: ‘How much did this ad make you think about you and your travel experience?’; ‘How much did you think about what it would be like if the events shown in the ad happened to you?’; ‘To what extent did you think the ad related to you personally?’; ‘To what extent were you reminded of your own experiences while reading the ad?’. The questions can be answered using a 7-point Likert scale, ‘1’ being ‘not at all’ and ‘7’ being ‘a great deal’.

The items measuring self-referencing were also tested for reliability. Kaiser-Meyer-Olkin value was .76 which is sufficient to verify sampling adequacy. Principal component analysis with Varimax rotation showed that all 4 items formed one single unidimensional scale with one

component that had an Eigenvalue above 1 (EV = 2.96, R2 _{= .74). Scale reliability was also}

satisfactory (Cronbach’s α = .88). Consequently the items on this scale were combined into a single scale named self-reference.

Moderator

As proposed in the hypothesis, ad skepticism affects the advantage of tailoring narrative advertising negatively. It functions as a moderator for the main effect. Ad skepticism was measured using the SKEP scale developed by Obermiller and Spangenberg (1998). SKEP is a nine-item scale consisting of statements and opinions that revolve around consumer tendencies of disbelief towards advertising. Participants indicated on a 7-point Likert-scale how much they (dis)agreed with each statement (‘1’ being ‘strongly disagree’ and ‘7’ being ‘strongly agree’). The statements were: ‘We can depend on getting the truth in most advertising’; ‘Advertising's aim is to inform the consumer.’; ‘I believe advertising is informative.’; ‘Advertising is generally truthful.’; ‘Advertising is a reliable source of information about the quality and performance of products.’ ; ‘Advertising is truth well told.’; ‘In general, advertising presents a true picture of the product being advertised.’; ‘I feel I've been accurately informed after viewing most

advertisements.’; ‘Most advertising provides consumers with essential information.’

Reliability of the items measuring ad skepticism was also tested. The Kaiser-Meyer-Olkin value for those items was .87, which is enough to verify sampling adequacy. Principal

component analysis with Varimax rotation showed that the 8 items formed a single unidimensional scale with one component which had an Eigenvalue above 1 (EV = 4.56,

R2 _{= .57). Reliability was good with a Cronbach’s alpha value of .89. All items on this scale were}

combined to form a single scale named ad skepticism. A median split was then used to compute a second dichotomous scale for ad skepticism, resulting in a ‘low’ and ‘high’ ad skepticism group.

Control Variable

Participants were asked for a number of demographic information. This included gender, age, occupation, education level and income. Furthermore, recent events and incidents in the aviation industry may have had an influence on how participants feel about airlines. This could have been a confounding factor in the experiment; therefore a control variable was included to measure this. At the end of the survey participants were presented with the following statement: ‘Recent incidents in the airline industry are a source of worry to me’. They were then asked to indicate on a 7-point bipolar scale to what extent they agreed with that statement (‘1’ being ‘disagree’ and ‘7’ being ‘agree’).

Manipulation Check

In order to test whether the designed narrative stimuli were indeed perceived as narrative and in order to replicate methods of existing research on narrative persuasion, a manipulation check was incorporated. Participants were asked to indicate on a 7-point bipolar scale (‘1’ being ‘disagree’ and ‘7’ being ‘agree’) to what extent they (dis)agreed with the following statement: “Think about the advertisement/advertorial on InterStar Airlines you just read. Would you say it resembles a story?”

Results

Preliminary results

Data preparation

Before proceeding to statistical analysis, the collected data was controlled for mistakes and faulty responses. A total of 154 responses were registered, four responses were removed because the duration of their participation was longer than an hour. All remaining 150 responses were complete and lasted under an hour, therefore, no further omissions were made.

Randomization Check

A randomization check was done to control whether participants were equally divided over all three conditions. A total of n = 54 were randomly assigned in the tailored narrative condition, n = 41 in the one-size-fits-all condition and n = 55 in the argumentative condition. A one-way

ANOVA showed that were no significant differences between the conditions when looking at age, F (2, 147) = .14, p = .869, gender, χ2 _{(2) = 1.56, p = .458 and education level, χ}2 _{(8) = 11.10,} p = .196. In conclusion, randomization has been successful.

Manipulation Check

In order to control whether the experimental conditions with a tailored narrative or a one-size-fits all narrative differed with regard to narrativity from the argumentative condition a manipulation check was done. A one-way ANOVA showed that the experimental conditions did differ

significantly with regard to narrativity (F (2, 147) = 14 .12, p < .001, partial η2 _{= .16).}

Participants in the tailored condition were exposed to the stimuli with highest narrativity (M = 4.80, SD = 1.62), participants in the one-size-fits-all conditions scored slightly less (M = 4.63,

SD = 1.34) while participants in the argumentative condition scored lowest (M = 3.35, SD =

1.60). A post hoc Tukey test showed that the only significant difference was to be found between both narrative conditions (tailored narrative Mdifference = -1.45, p <.001; one-size-fits-all narrative Mdifference = -1.30, p <.001) and the argumentative condition. The difference between the two narrative conditions was not significant at p =.867. It is therefore safe to conclude that the designed stimulus was indeed narrative of nature.

Alternative Explanations

A number of analyses were executed to control for possible confounding factors that may affect statistical outcomes. A bivariate correlation analysis between gender, age, worry about airline industry and all dependent variables (except for the binary-level behavior variable) was

executed. Gender correlated weakly but significantly with brand attitude (r = .20, p = .014), ad attitude (r = .17, p = .038) and behavioral intention (r = .19, p = .023). Worry about the airline industry correlated weakly but significantly with brand attitude (r = .23, p = .006), ad attitude (r = .22, p = .007), behavioral intention (r = .18, p = .025) and interest in email (r = .24, p = .003). Age did not show any correlation with the independent variables. Furthermore, a one-way ANOVA test showed that different education levels did not differ with regards to the dependent variables (see Appendix 4). Additionally, a one-way ANOVA test showed that gender, age and worry about airline industry did not differ across the two groups of participants with regard to their decision to receive a promotional email from InterStar Airlines or not (see Appendix 4). The same groups did not differ when looking at education levels (see Appendix 4). Control variables that showed correlations with respective dependent variables were included as covariate in statistical analyses.

Hypothesis Testing

The first hypothesis proposes that tailored narrative advertising leads to more positive ad attitudes, brand attitudes, more ad consistent behavioral intentions and behavior as compared to one-size-fits-all narrative advertising and argumentative advertising. Furthermore, the third hypothesis proposes that the advantage of tailoring narratives as compared to one-size-fits-all narrative and argumentative advertising is moderated by ad skepticism. Both hypotheses will be tested and discussed for each dependent variable separately.

Ad attitude

A two-way analysis of covariance (ANCOVA) was performed to assess the effect of the type of ad in conjunction with ad skepticism (dichotomous) on ad attitude, while controlling for gender and worry about airline industry. When looking at the effect of type of ad, no significant

difference was found F (2, 142) = 1.51, p = .226. A significant but small effect of ad skepticism was found F (1, 142) = 5.20, p = .024, partial η2 _{= .04. Finally, there was a marginally significant}

small interaction effect between type of ad and ad skepticism F (2, 142) = 2.45, p = .090, partial

η2 _{= .03.}

The interaction effect was not significant and therefore further investigation was not technically warranted. However, the reported ad attitude was split between the two groups to visualize the means. This was done to look for differences between the ‘low’ and ‘high’ ad skepticism groups and see if findings at least follow the direction of the hypotheses. Although not significant, the difference in means of ad attitude in the ‘low’ ad skepticism group points towards the hypothesized advantage of tailoring narratives (see Figure 2). In the ‘high’ ad

skepticism group this advantage seems to disappear and the means of ad attitude are generally higher than in the ‘low’ ad skepticism group.

!

Figure 2. Mean ad attitude for tailored, one-size-fits-all and argumentative conditions in ‘low’ and ‘high’ ad skepticism groups, including standard deviation between brackets.

Despite the lack of significant main and interaction effects, further analysis was done to investigate better the effect of type of ad with the two different levels of ad skepticism. An analysis of variance was performed to probe for differences between the type of ad within the ‘low’ and ‘high’ ad skepticism groups separately. Gender and worry about airline industry where not controlled for in order to obtain post hoc results. The ANOVA-test yielded significant

variation among type of ad within the ‘low’ ad skepticism group, F (2, 83) = 5.08, p = .008. A post hoc Tukey test showed that the tailored condition and argumentative condition groups differed significantly (Mdifference = 1.11, p = .006). The one-size-fits-all condition group did not

differ significantly from the two other groups, (both p > .05). As hypothesized, there was no difference between type of ad in the ‘high’ ad skepticism group, F (2, 61) = 0.08, p = .924. These

findings need to be considered while keeping in mind the lack of initial significant main and interaction effects.

Brand attitude

The same set of statistical analyses used for ad attitude were repeated for brand attitude as a dependent variable while controlling for gender and worry about the airline industry. A two-way analysis of covariance showed that a significant but small effect of type of ad was found, F (2, 142) = 3.48, p = .033, partial η2 _{= .05. Participants in the tailored condition reported the highest}

brand attitude (M = 4.50, SD = 0.95), while those in the argumentative condition reported the lowest brand attitude (M = 3.92, SD = 1.24). Reported brand attitude for the one-size-fits-all condition was in the middle (M = 4.22, SD = 0.96). A post hoc Bonferroni test showed that the tailored condition and the argumentative condition differed significantly (Mdifference = 0.51, p = .

028). The one-size-fits-all group did not differ from the two other conditions groups (both

p > .05). Ad skepticism also had a significant but small effect, F (1, 142) = 11.20, p = .001,

partial η2 _{= .07. There was however no significant interaction effect to be found F (2, 142) =}

1.42, p = .244.

As with ad attitude, reported means were visualized across the two different ad skepticism groups in spite of the insignificant interaction results. This was done to see if the effects followed the direction of the hypothesis. A plotted chart of the means (see Figure 3) divided across ‘low’ and ‘high’ ad skepticism shows comparable results as with ad attitude with regard to hypothesized direction of effects.

!

Figure 3. Mean brand attitude for tailored, one-size-fits-all and argumentative conditions in ‘low’ and ‘high’ ad skepticism groups, including standard deviation between brackets.

Again, despite the lack of a significant interaction effect, further analysis was done to get more insights into the effect of ad skepticism. An analysis of variance was done on both ad skepticism group separately, without controlling for gender and worry about airline industry. Within the ‘low’ ad skepticism group a significant difference was found between the type of ad, F (2, 83) = 5.79, p = .004. A post hoc Tukey test showed that, again, the tailored condition and the

argumentative condition differed significantly (Mdifference = 0.99, p = .003). The one-size-fits-all

group did not differ from the two other conditions groups (both p > .05). Also, the high ad skepticism group did not show any significant difference between type of ad, F (2, 83) = 2.07,

p = .813.

Behavioral intention

An analysis of covariance was done with type of ad and ad skepticism as fixed factors, gender and worry about airline industry as covariates and behavioral intention as the dependent variable. There was no significant difference found between the type of ad, F (2, 142) = 1.61, p = .204.

There was however a significant but small difference found between the two ad skepticism groups, F (1, 142) = 5.12, p = .024, partial η2 _{= .04. A marginally significant and small}

interaction effect was detected, F (2, 142) = 2.50, p = .086, partial η2 _{= .03.}

As with the previous dependent measure, further analysis was performed on the two ad skepticism groups despite the lack of interaction effect. The purpose is to see in which direction the moderator affects behavioral intention and findings should therefore be considered in that context. A plotted chart of the means (see Figure 4) shows that in the ‘low’ ad skepticism group, the direction of main effect follows the hypothesis. This is not the case in the ‘high’ ad

skepticism group.

!

Figure 4. Mean behavioral intention for tailored, one-size-fits-all and argumentative conditions in ‘low’ and ‘high’ ad skepticism groups, including standard deviation between brackets.

A one-way analysis of variance in both ad skepticism groups separately, shows that in the ‘low’ ad skepticism group a significant difference was found between the type of ad, F (2, 83) = 4.64,

p = .012. A post hoc Tukey test showed that again the tailored condition and the argumentative

condition did not differ significantly from the other conditions (p > .05). Finally, the high ad skepticism group did not show any significant difference between the type of ad, F (2, 61) = 0.92, p = .402.

Interest in receiving an email

An analysis of covariance was done with type of ad and ad skepticism as fixed factors, worry about airline industry as covariate and interest in email as dependent variable. Gender was not included in this analysis because it did not correlate with the dependent variable. For the sake of completeness, a separate analysis of covariance showed that including gender as covariate would lower the R2_{-value of the model from .20 to .18. Gender was therefore left out in subsequent}

analysis.

The analysis of covariance showed that there was a significant moderate difference between the type of ad, F (2, 142) = 11.34, p < .001, partial η2 _{= .14. Participants in the tailored}

condition showed most interest in receiving an email (M = 4.11, SD = 0.95). Participants in the argumentative condition showed less interest (M = 2.71, SD = 1.84). Those in the one-size-fits-all showed the least interest (M = 2.51, SD = 1.40). A post hoc Bonferroni test showed that the tailored condition differed significantly from the one-size-fits-all condition (Mdifference = 1.56, p

< .000) and the argumentative condition (Mdifference = 1.28, p < .000). The one-size-fits-all group

did not differ from the argumentative condition (p > .05). There was a marginally significant difference found between the two ad skepticism groups, F (1, 142) = 3.10, p = .081. There was however no significant interaction effect found between type of ad and ad skepticism, F (2, 83) = 1.73, p = .182.

A plotted chart of the means (see Figure 5) shows that in the ‘low’ ad skepticism group, the direction of effects follows the hypothesis, although this is not the case in the high ad skepticism group.

!

Figure 5. Mean interest in email for tailored, one-size-fits-all and argumentative conditions in 'low' and high ad skepticism groups, including standard deviation between brackets.

In order to findings whether the same results can be replicated as with the previous dependent variable with regard to the direction of effect, a split between ‘low’ and ‘high’ ad skepticism was performed. Findings should therefore be considered while keeping in mind that this is done despite the lack of interaction effect. A one-way analysis of variance in both ad skepticism groups separately, shows that in the ‘low’ ad skepticism group a significant difference was found between the type of ad, F (2, 83) = 14.85, p < .000. A post hoc Tukey test showed that the

tailored condition significantly differed from the one-size-fits-all (Mdifference = 1.83, p < .000) and

the argumentative condition (Mdifference = 2.08, p < .000). The one-size-fits-all condition group did

not differ significantly from the argumentative condition group (p > .05). The high ad skepticism group did not show any significant difference between the type of ad.

Opting to receive an email

A logistic regression with the type of ad and ad skepticism (continuous) as independent variable and opted for email as dependent variable was performed. The tailored condition was used as a reference variable when coding the type of ad as dummy variables. It showed a significant effect of type of ad (p = .010). Participants in the one-size-fits-all condition were, compared to the participants in the tailored condition, less inclined to opt-in for the email, B = -3.02, p = .031. Participants in the argumentative condition were even more less inclined than their counterparts in the tailored condition to opt-in for an email, B = -3.84, p = .004. There was no significant effect of ad skepticism (p = .408). There was a marginally significant interaction effect between type of ad and ad skepticism (p = .064). A logistic regression with the one-size-fits-all condition as reference showed no difference with the argumentative condition (p > .05). The only

significant difference in type of ad was thus between the tailored and the argumentative condition.

As with the other dependent measures, lack of interaction effects did not warranty further statistical analysis. Nevertheless, further analysis was performed to explore the direction of effects of type of ad across different ad skepticism groups. Two logistic regression analyses were performed for both ad skepticism groups separately with type of ad as independent variable and Opting for email as dependent variable. The tailored condition was kept as the reference type of ad condition. The analysis for the ‘low’ ad skepticism group resulted in a significant model, χ2 ₍₂₎

= 16.27, p < .000. It showed a significant difference between the type of ad. Participants in the one-size-fits-all condition were, compared to those in the tailored condition less inclined to opt-in for an email, B = -1.54, p = .007. Participants opt-in the argumentative condition were, compared

to those in the tailored condition the least inclined to opt-in for an email, B = -2.13, p < .000. There was however no significant difference between the one-size-fits-all and the argumentative condition (p > .05). The same logistic regression for the ‘high’ ad skepticism group did not result in a significant model, χ2 _{(2) = 1.10, p = .578. Consequently no significant difference was found}

between the type of ad.

Preliminary summary

A summary of findings in Table 1 shows a mixed picture of the hypotheses. The hypothesized effect of tailoring narrative ads is not supported in all dependent measures, though marginal significant effects exist and further inspection reveals patterns that are somewhat supportive. Hypothesis 1 is therefore partially supported. There was however no support for the expected role of ad skepticism, Hypothesis 3 was therefore not supported.

Table 1. Support for Hypothesis 1 and 3 and across all dependent variables and effect of type of ad within the ‘low’ ad skepticism group. * represents a significance level of p < .05.

The role of self-reference as a mediator.

Further analysis was necessary to investigate the mediating role of self-reference. To get a general first impression, a two-way ANOVA test with type of ad and ad skepticism as independent variables and self-reference as dependent variable was executed. There was a

significant but small difference between the type of ad, F (2, 144) = 4.56, p = .012, partial η2 _{= .}

06. Participants in the tailored condition reported the highest level of self-referencing, M = 4.30,

SD = 1.55. Those in the argumentative condition reported the lowest levels of self-referencing, M

= 3.41, SD = 1.62. Participants in the one-size-fits-all condition were in the middle, M = 3.97, SD = 1.32. A post hoc Turkey test showed that the only significant difference was between the tailored and the argumentative condition, (Mdifference = 0.90, p = .006). There was no significant

difference between the one-size-fits-all and the other two conditions (both p > .05). There was also a significant but small difference found between the two ad skepticism groups, F (1, 144) = 2.52, p = .049, partial η2 _{= .03. There was however only a marginally significant interaction}

effect found between type of ad and ad skepticism, F (2, 144) = 2.52, p = .084. A plotted chart of the means (see Figure 6) shows that in the ‘low’ ad skepticism group, the direction of effects follows the hypothesis; this is not the case in the ‘high’ ad skepticism group.

!

Figure 6. Mean self-reference for tailored, one-size-fits-all and argumentative conditions in ‘low’ and ‘high’ ad skepticism groups, including standard deviation between brackets.

Although the interaction effect was not significant, further analysis was done to explore the effects of the moderator on self-reference. A one-way analysis of variance in both ad

skepticism groups separately, shows that in the ‘low’ ad skepticism group a significant difference was found between the type of ad, F (2, 83) = 6.64, p = .002. A post hoc Tukey test showed that the tailored condition differed significantly from the argumentative condition (Mdifference = 1.50, p

= .001). The one-size-fits-all condition did not differ significantly from the other two conditions (both p > .05). The ‘high’ ad skepticism group did not show any significant difference between the type of ad, F (2, 61) = 0.22, p = .805.

In order to further investigate the mediating role of self-reference, which is essentially moderated mediation, a more complex analysis was necessary. While traditional mediation analysis prescribes a causal-step approach (Baron & Kenny, 1986), more recent insights point towards bootstrapping analysis as a valid approach to investigate mediation (Preacher & Hayes, 2008). This approach also allows researching moderated mediation, which is the hypothesized path of effects in the conceptual model of this study. Moderated mediation can be detected when the level of a mediator depends on the level of a moderator (Preacher, Rucker & Hayes, 2007). The effect of a mediator is significant when its confidence interval (CI) does not include the zero-point.

The PROCESS software for the Statistical Package for Social Sciences (SPSS) developed by Hayes provided a practical solution to perform bootstrapping analyses. The software allowed to include the covariates in this study, which increased the quality of the statistical analysis. A PROCESS regression analysis was therefore performed for each dependent variable with the correlating demographic variables as covariate. For these analyses, the choice was made to

investigate the effect of tailoring narratives more deeply by selecting two levels of the independent variable type of ad, namely the tailored and the one-size-fits-all condition. This meant leaving the argumentative condition out of the equation. This was done to detect possible differences between the two narrative conditions as a result of the tailoring manipulation. Consequently, dummy variables of the narrative conditions were used in PROCESS.

Figure 7 provides an overview of the mediation analysis with all its components. The initial effect of tailoring narratives is represented with ‘c’. The direct effect of tailoring within the mediation model is represented by ‘c’’. The effect of tailoring with the moderator ad skepticism is represented by ‘a’. The relation between self-reference and the dependent measure is

represented by ‘b’. Finally the focus of the moderated mediation analysis is the indirect effect.

The results (see Table 2) showed that there was only a significant initial effect of type of ad for behavioral intention, interest in email and opted for email. Furthermore, direct effect was only significant for interest in email and opted for email. There was however no significant relation to be found between the independent variable (type of ad) and the mediator (self-reference). A significant relation is a requirement for mediation. Since this requirement was not met, Hypothesis 2 could not be supported.

Table 2. B-coefficients of the components in the moderated mediation model and corresponding Standard Error across all dependent variables. Significance levels are represented as follow: * p < 0.1; **: p < 0.05; ***: p < 0.01, n.s. = not significant.

Discussion & Conclusion


The purpose of this study was to investigate the effect of tailoring advertising narratives to individuals on a number of persuasion outcomes (H1). The role of self-reference as a mediator (H2) was investigated alongside the moderating role of ad skepticism (H3). Tailoring was achieved by presenting to participants a story that featured a protagonist which resembles themselves on six points. It was expected that tailored narrative advertising would lead to more

positive persuasion outcomes as compared to one-size-fits-all narrative advertising and argumentative advertising. Results showed that this hypothesis was supported for three out of five dependent variables; these were brand attitude, interest in email and opting for email. The effect sizes of these variables were, however, modest. Hypothesis 1 was, therefore, partially supported. Furthermore, it was expected that self-reference would mediate the effect of tailoring. Results did not provide enough evidence for this. Hypothesis 2 was therefore not supported. Finally, ad skepticism was expected to moderate the effect of tailoring; there was not enough evidence for this. Consequently, Hypothesis 3 was not supported either.

Based on these findings, it can be said that tailoring narrative advertising provides an advantage on some but not all aspects of marketing of a consumer product. More specifically, participants evaluated the brand more positively when they were shown a tailored narrative advertising. Additionally, participants showed more interest in receiving a promotional email after having seen this type of advertising. Also, more participants opted to receive that

promotional email in the tailored condition than in the other conditions. In contrast, tailoring did not lead to significantly better evaluation of the ad or to significantly more favorable behavioral intentions. When looking at the means of the dependent variables, however, results showed that participants in the tailored condition reported having more positive persuasion outcomes than participants in the one-size-fits-all and the argumentative condition. When considering only the direction of effects and not their significance, one could argue that tailoring indeed worked in the predicted way.

One point of interest is the lack of significant main effect on behavioral intention and the presence of main effect on interest in email and opted for email. The two latter measured

ad-consistent behavior which is generally harder to influence than mere intentions (Azjen &

Fishbein, 1977). A possible explanation could be that behavioral intention included the likelihood of purchasing a ticket from InterStar Airlines in the future and the likelihood of recommending the airline to a friend. These behaviors entail much more risk than just opting to receive one promotional email from InterStar Airlines. Although they were just behavioral intentions, the outcomes were much important than the actual behavior participants were probed for, which possibly explain the skewed results.

Hypothesis 2 predicted a mediation effect of self-reference between type of ad and the dependent variables. Results of the study showed that there was not enough evidence to support this hypothesis. Contrary to existing theories on self-reference in narrative persuasion (De Graaf, 2014), it was not possible to recreate the mediation effect of increasing similarity between receivers and protagonists. The reason behind this could be that since this study looked at a moderated mediation, the sample size was too small to detect any significant effects for the whole model. The tailored narrative was, however, effective in invoking more self-referencing than the argumentative narrative. Also, self-reference did correlate significantly with all

dependent measures. That aspect does resonate with existing literature as De Graaf (2014) also showed that higher levels of self-reference lead to better persuasion outcomes. One of the requirements of causal-step approach for mediation analysis by Baron & Kenny (1986) is the difference between the direct effect of the independent variable on the dependent variable outside (c) and inside (c’) the mediation model. The direct effect inside the model should disappear in favor of the mediator variable. The model in this study showed that for ad attitude, brand

the model (c’) diminished from the effect outside the model (c). Although not significant, these findings point towards the hypothesized mediation effect. Speculating on this, further research with a larger sample could validate these findings.

The results of the experiment did not yield any significant findings with regard to the expected moderating effect of ad skepticism (Hypothesis 3). This is, probably, also the result of the modest sample size. It is, however, interesting to note that although not significant, the moderation results on the dependent variables showed a trend. After splitting the sample in a ‘low’ and ‘high’ ad skepticism groups, a clear pattern arose in which all the means within the ‘low’ group followed predictions of the main effect. Indeed, participants in the tailored condition reported higher ad attitude, brand attitude, behavioral intention, showed more interest in a promotional email and opted more often to receive that email than participants in the one-size-fits-all and the argumentative condition respectively. This effect was not found in the ‘high’ ad skepticism group where participants did not differ much, tailoring did not seem to increase persuasion there. It should be noted that participants in that group were generally more persuaded than the ‘low’ ad skepticism group. Therefore, it could be argued that tailoring narrative advertising only offers an advantage with individuals that are less skeptical of advertising. When looking at participants that are more skeptical, tailoring does not provide a persuasion advantage but it does not hurt either. These claims would have to be substantiated with significant results.

Another point of interest is found in the dependent variable interest in email. When looking at the ‘high’ ad skepticism group, it is worth noting that participants in the tailored condition showed more interest in receiving an email than participants in the other two

conditions. Although the difference is not significant, this is interesting because it possibly means that tailoring helped participants overcome their skepticism towards advertising by raising curiosity about the offer by InterStar Airlines. In other words, even though these participants are wary of advertising, they probably did not want to miss out on information that could have been relevant for them. Drawing upon that speculatively, one could say that as a mechanism of drawing attention and providing information, tailoring can work even with highly skeptical consumers. Consequently, tailoring can prove to be useful in situations where changing awareness of a brand or product is more challenging that changing attitudes or behavioral intentions.

Finally, results showed that across all dependent variables, no significant difference was found between the one-size-fits-all narrative advertising and the argumentative advertising. This is in stark contrast with existing studies that show that narrative advertising enjoys better

persuasion outcomes than argumentative advertising (Ching et al., 2013; Escalas, 2007). One reason for the contradicting findings may lie in the execution of the one-size-fits-all narrative advertising. It showed an ad that featured the same number of tailored information as the tailored condition. The result was an advertising that is full of irrelevant personal details which might feel unnatural and therefore off-putting. A real narrative advertising would probably focus more on the consumption experience of the brand rather than on personal details. This may have been the reason why the one-size-fits-all advertising performed better, but not significantly better than the argumentative advertising. The execution style was nonetheless important for the sake of internal validity of the experiment. It allowed for a fair comparison between the conditions.

Practical & theoretical implications


Although not significant across the board, findings of this study show a clear trend which is that tailoring a narrative advertising provides a persuasive advantage. Narrative advertising comes in various forms, from ads in television to print magazine and from spots on the radio to

commercials on YouTube. Tailoring can therefore take a multiplicity of forms and execution styles. The most obvious venue for tailoring is online advertising because of the sheer scale of personal information available and the ease of implementation. This study employed email marketing which limits the scope of practical implications somewhat. However, practitioners in the field that are interested in enhancing online advertising campaigns that use the narrative approach should thus look into the possibilities of tailoring their message, not only in email but in other channels as well. Although tailoring online advertising may bring with it resistance due to privacy concerns, findings of this study suggest otherwise. Persuasion did not decrease with tailoring in highly skeptical participants. Persuasion did increase in less skeptical participants. This provides an interesting segment of consumers in which narrative advertising can benefit from tailoring. When practitioners identify this type of consumers while setting up a marketing campaign, it may be worth the expense to put some effort in tailoring the message. They can do so without fearing the risk of alienating highly skeptical consumers that happen be exposed to that same type of message.

This study was able to reproduce existing knowledge on tailoring persuasive messages for some but not all persuasion outcomes (Andsager, Bemker, Choi & Torwel, 2006; Hende, Dahl & Schoormans, 2012). As with the existing literature, reader-protagonist similarity was at the center of this study but with a focus on online advertising. This venue still forms an

undiscovered territory when it comes to tailored narrative advertising and that is where the scientific contribution of this study lies. Tailoring persuasive consumer messages has not received as much attention as health messages in academic research, which is where this study also hopes to contribute. The findings are promising as they point towards the same direction of the positive effect of tailoring. This study also yielded some interesting findings on ad skepticism and tailoring. It was hypothesized that ad skepticism would lead to resistance within participants due to concerns over privacy when faced with a highly personalized ad. However, results suggest that ad skepticism does not mitigate the effect of tailoring and that in absence of ad skepticism, tailoring offers a clear persuasive advantage. This interplay is not in line previous research (Escalas, 2007) and deserves further investigation. The reason behind this could be that previous research was not in an online context. Online advertising in social media and email is a relatively new field which is continuously developing and comes with its own ever-changing rules and customs. Internet users have possibly developed reaction mechanisms that are different from with traditional media and traditional advertising.

Limitations & Further research


Inevitably, this study knows some limitations. For example, it used an airline as a focus brand. While this type of product was well suited for tailoring a narrative advertising, it may not be suited to measure persuasion within the small scale of this experiment. Airplane tickets are a ‘high’ involvement product and these requires a considerable amount of thought before purchase due to the risk this action entails (Rossiter, Percy & Donovan, 1991). Participants were forced to form an opinion about the brand and decide how they will deal with in a short amount of time.

This could have had a negative impact on the validity of the findings. Further research should look at tailoring narrative advertising for low involvement products with less risk. This study used email as a marketing channel, it would be beneficial when further research investigates more modern ways of online marketing such as advertising on social media and interactive video advertising. Furthermore, this study used a number of personal characteristics as points of

tailoring. Further research could replicate this in combination with other concepts that lend themselves to tailoring such as personal traits and psychological tendencies (promotion- vs. prevention-focused persons, extravert vs. introvert persons). This type of tailoring can open new venues for narrative advertising and unlock its potential even further.

Conclusion


It has been demonstrated that advertising that uses the power of narratives enjoys better persuasion outcomes than advertising that does not. Making these advertising narrative more relevant to consumers can increase its effectivity further and this can be done in several ways. Although not significant across the board, the outcomes of this study show that tailoring is one way to go for practitioners who want to increase relevance for consumer. Narrative advertising that is tailored leads to more favorable brand attitudes and consumer behavior. Other aspects of persuasion may also benefit from tailoring as well and there lies a scientific opportunity. Further research on tailoring can show how much more persuasion potential lies in narrative advertising.

Reference

Adaval, R., & Wyer, R. S. (1998). The role of narratives in consumer information processing.

Journal of Consumer Psychology, 7(3), 207-245.

Ahmad, R. (2003). Benefit segmentation: a potentially useful technique of segmenting and targeting older consumers. International Journal of Market Research, 45(3), 373-388. Ajzen, I., & Fishbein, M. (1977). Attitude-behavior relations: A theoretical analysis and review

of empirical research. Psychological Bulletin, 84(5), 888.

Bandura, A. (1986). Social foundations of thought and action. Englewood Cliffs, NJ, 1986. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social

psychological research: Conceptual, strategic, and statistical considerations. Journal of

Personality and Social Psychology, 51(6), 1173.

Boyd, B. (2009, April 15). On the origin of stories: Evolution, cognition, and fiction. Harvard University Press.

Burnkrant, R. E., & Unnava, H. R. (1989). Self-Referencing A Strategy for Increasing Processing of Message Content. Personality and Social Psychology Bulletin, 15(4), 628-638.

Campbell, M. C., & Kirmani, A. (2000). Consumers' use of persuasion knowledge: The effects of accessibility and cognitive capacity on perceptions of an influence agent. Journal of

Consumer Research, 27(1), 69-83.

Chang, C. (2009). " Being Hooked" By Editorial Content: The Implications for Processing Narrative Advertising. Journal of Advertising, 38(1), 21-34.

Ching, R. K., Tong, P., Chen, J., & Chen, H. (2013). Narrative online advertising: identification and its effects on attitude toward a product. Internet Research, 23(4), 414-438.

Escalas, J. E. (1998). Advertising narratives: What are they and how do they work? In B. Stern (Ed.), Representing Sonsumers: Voices, Views, and Visions (pp. 1–23). New York: Routledge & Kegan Paul.

Escalas, J. E. (2007). Self-referencing and persuasion: narrative transportation versus analytical elaboration. Journal of Consumer Research, 33(4), 421-429.

Debevec, K., & Iyer, E. (1988). Self-referencing as a mediator of the effectiveness of sex-role portrayals in advertising. Psychology & Marketing, 5(1), 71-84.

Deighton, J., Romer, D., & McQueen, J. (1989). Using drama to persuade. Journal of Consumer

Research, 335-343.

Dunlop, S., Wakefield, M., & Kashima, Y. (2010). Pathways to Persuasion: Cognitive and Experiential Responses to Health-Promoting Mass Media Messages. Communication

Research, 37(1), 133-164.

De Graaf, A. (2014). The Effectiveness of Adaptation of the Protagonist in Narrative Impact: Similarity Influences Health Beliefs Through Self-Referencing. Human Communication

Research, 40(1), 73-90.

De Pelsmacker, P., Geuens, M., & Anckaert, P. (2002). Media context and advertising effectiveness: The role of context appreciation and context/ad similarity. Journal of

Advertising, 31(2), 49-61.