Does targeting moderate the need for credibility?

 

Master Thesis 

 

  

Does targeting moderate the need for credibility? 

 

 

  

Author: Andrei ­ Victor Stan  

Student No: 10986839 

Submission Date: 29 January 2016 

Qualification: MSc Business Administration ­ Marketing Track  

Institution: University of Amsterdam  

Prof. Dr. J. Y. Guyt 

  

 

 Statement of originality 

  

  This document is written by Andrei ­ Victor Stan who declares full responsibility for       

the contents of this document.   

 

I declare that the text and the work presented in this document is original and that no        sources other than those mentioned in the text and its references have been used in        creating it. 

 

The Faculty of Economics and Business is responsible solely for the supervision of        completion of the work, not for the contents.         Andrei­Victor Stan,   29 January 2016.                        

 ​

Table of contents:  

1. Introduction………5  2. Literature Review……….8  2.1. Credibility………..8  2.2. Targeting………10  2.2.1. Interaction between Targeting and Credibility………12  2.3. Purchase Intention………13  2.3.1. Theory of planned Behavior & ELM………....……….13  2.3.2. Interaction between Credibility and Purchase Intention………15  2.3.3. Interaction between Targeting and Purchase Intention...……….16  2.4. Research  model & Hypotheses……….18  3. Methodology………...20  3.1. Purpose of the Research……….20  3.2. Research Design ……….21  3.3. Choice of Product ………22  3.4. Survey Setup & Data Collection……….23  3.4.1. Procedure……….23  3.4.2. Introduction Part………..25  3.4.3. Conditions……….26  3.4.4. Sample & Data Collection………..27  3.4.5. Measurements……….27  4. Data analysis ……….30  4.1. Analysis Steps ……….31  4.2. One­Way ANOVA ...………35  4.3. Hierarchical Regression………..38  4.4. Hypotheses Testing……….43  5. Discussion and Conclusion ……….45  5.1. Theoretical Contributions ………...47  5.2. Managerial Contributions ………...48  5.3. Limitations and Further Research ……….48  6. References ………50  7. Appendices……….54 

Abstract: 

In these days Social Media is part of the daily life of more than 1 billion people.        People argue whether social platforms are credible or not as they attract much of the brands’        attention and information becomes suspicious.  

People use social media for sharing content and participating in social networking as        the main purpose, therefore when a brand is being followed the main reason is for getting        information about it or its products. The previous research on the credibility that brands have        in the online setting found that peers and friends are the most effective endorsers of brands        and products.  

This paper accepts the more positive effects of people against brands in inducing a        purchase behavior to their followers. What the paper doubts is whether the effect remains        the same in case of the same sources when proper targeting methods are moderating the        interaction with consumers. The concept of the study asks whether targeting moderates the        need for credibility in the online setting.  

The idea was tested by designing an experimental study on bikes followed by a        questionnaire. The survey tested four conditions based on high and low levels of credibility in        poor and proper targeted settings. When applying behavioral techniques for targeting as        proper targeting methods the credibility influence of brands resulted in close purchase        behavior means as friends did, which is against the previous findings. The relevance of the        paper is proving that brands have are mastering targeting tools in social media that inspire        the consumer a credibility and trust similar to friends recommendations.   

       

Preface 

This study was the research project conducted for my master thesis at University of        Amsterdam. The paper is inspired from the knowledge gathered in the courses of the MSc in        Business Administration ­ Marketing Track. 

Much of the work was not possible without the help of Dr. J. Y. Guyt. The        collaboration with him as the supervisor of this thesis was inspiring. Every aspect was        discussed in a friendly and open manner which gave me the motivation to finish the paper.  

Furthermore my family deserves special thanks for the support offered during the        entire master course and the thesis writing process. Without their support this study could        not be possible.  

 

Andrei­Victor Stan,   29 January 2016. 

 

 

 

 

 

1. Introduction 

The last 20­40 years have proved that the world of advertising and targeting is        dynamic and volatile, especially due to the powerful tools technology provides. The        traditional old channels to reach the consumers as the TV or the radio are losing face        against the Web. One of the main reasons of the switch towards the online environment is        that the consumers are empowered to skip advertising, and focus only, or at least mostly, on        the content they wish. Time has become a valuable asset for people and the ability to skip or        forward advertising forced marketers to create new strategies to reach audiences: banner        ads, pop­up ads, behavioral advertising etc. In 2010 already 20% of the ads were viewed on        Social Networks (Dutta, 2010), and consequently the attention of marketers switched        towards this environment. Nevertheless the Web users have more power in deciding        whether something is trustworthy or credible, therefore the marketers needed to learn how        to deal with the impact of credibility of advertising against friends’ opinions. Therefore        platforms as Facebook are not only social communities but true marketing platforms for        companies (Curran et al., 2011)​        . Besides allowing companies to share and promote their        products and services, Social Network Sites (SNS) offer cheap solutions for sponsored ads        that can reach a particular audience at a deeper level of targeting. Staying in touch with the        audience is useful for companies and on the their side, users can give transparent negative        or positive feedback, or even counteract with brand­related User­Generated Content. The        Web is a free market where both brands and consumers can freely exercise the right to        express any opinion, positive or negative. This topic has been covered by Cheong &        Morrison (2008) according to whom consumers’ reactions to advertising are influenced by        the content creator source, either User­Generated or Brand­Generated content. 

This study will extend the knowledge on credibility of advertising sources on SNS by        analyzing the moderation effect that targeting has in reducing the need for credibility. This       

paper argues that under certain targeting conditions, which SNS’ targeting tools do provide        at this moment, BGC can resist against the strong credibility that UGC and peer pressure        provide to fellow users of Social Media Platforms. In this sense, the following research        question was formulated​    : Does targeting moderate the need for credibility? ​               The paper is at        first of scientific relevance. Previous research focused on the effects of UGC or BGC as        separate objects of study, while the comparison between them has been little explained in        the same study. Even more the effect of brand related content has not been measured in        terms of credibility of the source who created the content. Finally the moderation effects of        targeting on the relationship between credibility and purchase behavior will contribute to        existing scientific knowledge as it focuses on new targeting strategies on SNS, specifically        behavioral targeting. The reason why this study is analyzing the effects of different content        sources is because content is what brings value to the online environment. Content is what        the Web is made of and its relevance is analyzed through the purchase intention measure.  

The research question was translated into a theoretical framework based on the        independent variable, the credibility of the creator of online brand related content. The effect        of credibility on the consumer was tested using purchase intention as the measure of the        effect of different credible sources on SNS users. In order to explain this relationship the        level of targeting (poor or proper) was added as a moderating variable of the relationship,        that would influence the relationship in a negative sense.  

To measure and test the relationships between Credibility, Purchase Intention and        the level of Targeting a research was conducted through an experimental study. Using        Facebook tools to build four experimental conditions, fictitious posts were created to express        four conditions derived from the independent variable and the moderator. The experiment        was placed after the respondent was exposed to a contextual setting defining the level of        targeting and the moderating effects on purchase intention. The post was then followed by a        set of questions measuring the construct of credibility and purchase intention. The setting for       

the research was Amsterdam, The Netherlands, where the network of the researcher is        based. Nevertheless the experimental conditions can be applied on scale based on the        positive results of the statistic analysis, so the findings were generally applicable at an        abstract level.  

The main contribution of this paper is at first academical, as it offers scholars statistic        data on the relationship between credibility and purchase intention. The survey conducted        proved the benefits and the strong influential power of SNS on the purchasing likelihood.        The paper is based on the rapid increase in online behavioral targeting campaigns that use        the tools provided by social media platforms as Facebook. The perception that people hold        on the source of a post is formed accordingly to the credibility the source holds, and low        credibility is typically assigned to brands while high credibility is held by consumers, peers or        friends. This study considers peers as the most credible online endorsers of brands and        products, and takes this as granted according to the vast previous research in practice        (Kaplan & Haenlein, 2010; Goldfarb & Tucker, 2011; Smith et al., 2012). The reason why this        study was conducted came from the debate that if targeting tools in the online setting can        capture enough information about the consumer, than brands and companies can build        proper targeted advertising campaigns that could overcome the strong credibility that friends        have acquired so far on SNS.  

The paper is more than that relevant for marketers in practice, as it describes the        effects of behavioral targeting and gives statistical insights on consumer purchase behavior        when being subject to advertising. Using the results of the experiment marketers can apply        the characteristics of the fictitious brand­related piece of content that were used in this        survey. The results of the ANOVA tests proved that there is a statistical difference in the        purchase  intention between the four conditions created. The results translate into a        significant effect on purchase intention in the conditions where proper targeting was applied.        In this sense the implications of targeting can negatively moderate the relationship between       

credibility and purchase behavior. Therefore using the correct advertising tools marketers        can overcome the power of friends or peers.  

 

2. Literature Review 

This chapter will start by introducing the existing knowledge on the topic of the paper,        examining the previous research on credibility and targeting. The first subchapter explains        the role of credibility in the marketing field. The second subchapter reviews the evolution of        targeting as a marketing tool and its effects on the credibility of an ad creator. The third        subchapter is dedicated to the measurement scale of the experiment in this research paper,        the purchase intention of the customer. The purchase intention has been widely examined        but it remains the most used measurement scale of this type of experiments. In the last        subchapter all the variables form the theoretical framework and the research model with two        hypotheses is created.  

 

2.1. Credibility 

Growing Internet usage for research and information has long ago raised concern        about the credibility of the creator­source (France, 1999; Tucher, 1997). The credibility of        online information rises in importance as the effects of traditional advertising are declining.        Respondents of a poll on CNN believability reviewed CNN.com as 14% more trustful than        CNN TV channel (Greer, 2003). A problem that can arise with online content though, is that        the consumer might be unable to distinguish the content source: between regular users        (User­Generated Content) or professional brand generated information (Brand­Generated        Content). Internet is a vast library of information, and data needs to be filtered by the users        in order to find the right answer to their search. One of the first and principal methods used       

by the human brain in filtering the online information is by the credibility of the source        (Wathen & Burkell, 2002). 

Advertising scholars define a credible source as a source of “correct knowledge”        (Hass, 1981, p. 143). Credibility (or trust in this context) has been considered by researchers        to have a central role in the online interaction with brands (Lee & Tan, 2003; Goode & Harris,        2007). Information from sources rated as highly credible leads to greater attitude change        among those receiving a message; low­expertise sources typically produce no changes in        attitude (Milburn, 1991). Scholars’ surveys results showed that amongst online influencers,        the most trusted source of content would be “a person like myself”, which mostly translates        into a “friend”. 

Texts, pictures and videos posted online on a public SNS by users outside their        professional practices are defined as UGC (Kaplan & Haenlein, 2010). If this content,        despite the fact that is not created by a company, is related to a brand or company, than it        will be defined as brand­related UGC (Smith, Fischer & Yongjian, 2012). On the other side,        when content is created by brands or companies than it will be defined as BGC.        Nevertheless after years of TV and radio advertising consumers are not fooled by the        presence of brands in Social Media. In this sense BGC is of less credibility when compared        to UGC, when considering a peripheral route of analysis. At a closer look on the existing        literature and experiments in practice it is still debatable whether in the case of a more        interested and informed audience UGC remains the better predictor of consumer attitude or        behavior change.  

Cheong & Morrison (2008) explain that the majority of people are not influenced by        Brand Generated Content across mass media, and that the effect of BGC is influencing        Opinion leaders. The next step in their theory is that Opinion Leaders spread their attitudes        and beliefs and their supporters believe in their sayings. This closes the gap between BGC       

and consumer in mass media through an indirect relation between BGC and consumers        mediated by Opinion Leaders (Cheong & Morrison, 2008). Based on scholars’ rationale that        UGC is more credible in online media and considering that Opinion Leaders are part of the        UGC category, than the effect of BGC in Social Media is validated by the credibility of        Opinion Leaders, as non­commercial agents. 

 

2.2. Targeting 

Since consumers spend more time on Internet than ever before, companies’ attention        switches to the online environment in what concerns advertising procedures. Technological        advancements enable both people and companies to easily access social media platforms.        There is vast amount of funds directed to online marketing campaigns (Mediafeitenboekje,        2008) and the highest ROI comes from Behavioral Targeting (BT). The initial targeting        technologies in online social were Contextual and Content Targeting. Contextual targeting        puts emphasis on the link between the host website and the advertising content. It is an        indirect method of proper targeting by placing relevant advertised products on same context        host websites in order to reach a more involved audience.   

By combining data using cookies a customer’s profile is created and better targeting        can be applied based on socio­demographical and historical data. This particular mix of data        results in what is called Behavioral Targeting, which companies use as a more efficient        targeting procedure in the online marketing space. Through this technique ads will only be        displayed to the individuals with the highest potential to be influenced (Liberali, 2014).        Consequently this particular customers are selected according to the data describing their        online behavior. 

We live in a dynamic advertising age where strategies change at a fast phase and        mainly based on how affective data can be used to create a more exact profile of the        consumer. Spendings on consumer data is constantly increasing (Hallerman, 2007).        Companies like eBay, Amazon or Financial Institutions hold valuable data about their        consumers and privacy concerns are increasing constantly in internet users’ view. And while        the scholars knowledge on privacy concerns is increasing, the targeting mechanisms is        moderating the effect of privacy on users on web. And in this sense privacy can be        translated into User Generated Content or Brand Generated Content. User Generated        content is perceived as respecting consumers’ privacy more as it comes from peers (Lafferty        & Goldsmith, 1999). Peers are more credible and trustworthy therefore they make followers        feel more secure. On the other side Brand Generated Content is perceived as more official,        less trustworthy and so the privacy concerns are increased, as brands are financial        organizations looking for profit.  

Starting with the 1980s, studies show significant decrease in advertising credibility,        under the reason that ads did not present “an accurate picture of advertised products”        (Shavitt, Lowrey, & Haefner, 1998, p. 8). More specific, the attitude towards ads is influenced        by demographic factors and one’s particular historic with the advertising brand or the general        attitude towards the product or brand in the ad. Furthermore, attitudes towards advertising        differ by gender, age, education, income or religion. The credibility of the advertising­source        is also important in this context; Lafferty and Goldsmith (1999) concluded that credible        endorsers counteract advertising mistrust and that credible advertising sources can influence        attitude and behavior (p. 110). 

  Online targeting has been important for researchers, which paid attention to search        and content based targeting tools as Google AdWords and AdSense. Search­based targeted        advertising uses ads to respond to user’s search demand on searching engines while       

content based targeting is applied on websites where the central focus of the website        matches the ad content. The rise of Social Media forced practitioners to use SNS (Social        Networks Systems) in a more sophisticated targeting manner: while search or content based        targeting is focusing one little user­generated data (vocabulary, words), social networks offer        more sources of user related data that in one way define Behavioral Targeting. The data        provided by SNS includes historical search, user­generated tags (Hashtags) or browsing        behavior. 

Behavioral Targeting increases the effectiveness of advertising campaigns (Yan, Jun        et al., 2009) and plays an important role in the targeting world at the moment. One        possibility, that has as of yet not been researched is the possibility that Behavioral Targeting        can overcome the credibility of friends or peers. The advantage of Behavioral targeting (BT)        is that more data input translates into clearer targeting and less ads must be used to reach        the desired audience. BT is about delivering ads to users according to the behavioral data        coming from SNS. Therefore resources can be more effectively used to satisfy consumers’        preferences and be a better source of information for the subject. 

 

2.2.1. Interaction between Targeting and Credibility 

Advertisers use endorsers as credible sources to influence consumers’ attitudes and        purchase intentions (Goldsmith et al., 2000). Endorsers are seen as indicators of proper        targeting conditions, enabling better response to online content. Proper targeting is        associated by the same authors with positive evaluations of the ad in their study, which can        further be associated with credibility of the source. This relationship has therefore been        tested before, but in another context. Even more Goldfarb & Tucker (2011) experiment if        matching an ad to the website content has positive effects on purchase behavior and       

conclude that matching these two variables can have positive impact on purchase intention;        in this scenario we consider that matching stands for proper targeting. Practitioners in the        industry believe that Behavioral Targeting, seen as a proper targeting method in the online        environment, helps to delivering information (ads) to the most likely users to associate with        that particular ad, and therefore the most likely consumer to be influenced in changing        behavior or attitudes. 

Behavioral Targeting as a new marketing strategy to promote products or services is        yet to be considered understood completely. Only few academic papers (Yan, Jun et al.,        2009) consider behavioral targeting a proper targeting method. Nonetheless it has been paid        increasingly more attention in the last few years than to other targeting methods (Soza et. al,        2008) and therefore this study will try to prove its effectiveness in influencing consumer’s        attitude and behavior. According to the definitions of Yan, Jun et al. (2009) and other        academic papers on BT, and generalizing it as a proper targeting method, we can argue that        proper targeting is used to influence the most relevant users based on the credentials of a        credible relationship with the consumer, and on the combination of user­data provided by        Social Media Networks.  

 

2.3. Purchase Intention 

2.3.1. Theory of Planned Behavior & ELM 

  Purchase intention can be defined as an individual’s conscious plan to make an effort        to purchase a brand (Spears & Singh, 2004). According to Goldsmith and Horowitz (2006)        prior to purchasing from a brand, consumers will look for information online, where content        and information provided by other consumers appeared to be more important than brands        activity. The effectiveness of advertising is most relevant when measuring the        click­through­rate (Dreze & Hussherr, 2003). And the average click­through­rate is the       

response of the consumer induced by the ad. Since most of the ads purposes is to influence        purchase behavior, clicking an ad is seen as an intention to purchase, following Armitage &        Christian’s (2003) theory of planned behavior. The authors of the Theory of Planned        Behavior explain that the relationship between attitude and behavior is mediated by        behavioral intentions, which are seen as the necessary motivation to perform a certain        behavior. Behavioral Intentions are the most appropriate to predict behavior as they reflect        the consumer’s desire to take action. From the online advertising perspective clicking and ad        is a step closer towards the behavior the ad is supposed to stimulate, and therefore        advertising click­through­rate is considered a strong predictor of Purchase Behavior.  

The effectiveness of online advertising campaigns is finally expressed by sales        increase, but in order to measure and conduct experiments Armitage & Christian’s (2003)        Theory of Planned Behavior suggests that Purchase Intention is a strong predictor of        Purchase Behavior. 

There are only three elements of an advertising that are responsible to explain the        outcomes on the viewer: the message of the ad, the ability (motivation) to process the ad        and the environment where the ad is placed. The motivation to process and evaluate        advertising means that the level of involvement of the exposed viewers must be taken into        consideration to understand how they perceive ads. Viewers’ interest and desire to translate        the message from an ad is variable based on feelings and prior experience.  

The role of online advertising campaigns is to modify consumer behavior. And since        every consumer is different in its way, they interpret ads from different points of view. To        understand what the role of targeting is in advertising Petty and Cacioppo’s (1979) explain        through the Elaboration Likelihood Model two routes of persuasion that can be applied when        interacting with an ad, central and peripheral. To follow a central route of persuasion one        must prove a high level of motivation to process the message and good cognitive ability to        understand an ad. In a more simplistic way to define a central route to persuasion the       

consumer must focus on the message and quality arguments of the ad, which involves a        cognitive, rational analysis approach. The peripheral route to persuasion involves an        audience concentrated on secondary cues and non­contextual arguments. This category of        consumers base their attitudes change on quantitative factors, number of positive or        negative cues, colors, musical theme, subjects (famous people), words used etc. Their        decision­making process is not based on product or service related technical characteristics        and therefore they do not make pure rational analyses.  

On account of the ELM, to successfully influence behavior of the viewer marketers        should take into consideration the level of involvement of the consumer. Applied to what        Social Media is today the Theory of Planned Behavior and the ELM are still applicable to the        extent that by proper online targeting conditions a high level of involvement is reached        therefore the central message and argument is the focus of the viewer. On the other side in        poor targeting conditions the consumer will be more focused on peripheral cues, secondary        arguments and information. The purpose of advertising is to generate consumer behavior.        The reason this paper is focused on the intention to purchase is because consumer attitude        is the most effective predictor of behavior. A proper targeted advertising campaign will be        more effective due to a better understanding of the consumer. The application of the ELM in        online advertising offers information on the cues that should be applied and so better        outcomes can be expected, especially through behavioral targeting.  

 

2.3.2. Interaction between Credibility and Purchase

 

 

 

 

 

 

Intention 

  Goh, Heng and Lin (2013) support the stronger influence of UGC compared to BGC        on the consumer intention to buy. They found that engagement in social media brand        communities leads to a positive increase in purchase intentions, and that UGC exhibits a        stronger impact than BGC on consumer purchase behavior. Brand communities are related       

to increased credibility, therefore credibility explains why Purchase intention increases or        decreases as a result of consumption of UGC or BGC. 

  Privacy concerns also explain the relationship between Credibility and Purchase        intention. Increased privacy concerns appear in case of contact with BGC and low privacy        concerns in contact with UGC (Van Noort, Kerkhof and Fennis, 2008). And when privacy        concerns are high consumers are more preventive and sensitive to changing behavior.        Increased privacy concerns stand for low credibility sources of information as brands or        official organizations. By high credibility we make reference to peers, or friends which        generate low privacy concerns, associated with high credibility.  

The reason why ads may be considered intrusive, bulging and interfering is on one        hand the privacy concerns. The effect of privacy concerns on purchase behavior has been        researched and is considered that the method of approval, of acknowledgment of        information is regulated by privacy matters (Wathieu & Friedman, 2009). In cases where        privacy is considered violated, the consumer will act in a more delicate manner, which        translates into a diminished acceptance to the change in behavior or attitude that ads try to        induce.  

2.3.3. Interaction between Targeting and Purchase

 

 

 

 

 

 

Intention 

  The literature on targeting moderating effects on the relationship between credibility        of the source and the purchase intention suggests that targeted ads, by increasing the        possibility of getting the attention of the consumer may result in the manipulation of        consumer’s behavior. Consumers are tolerant to targeted advertising as they suppose the        information received may be useful (Cho & Cheon, 2003), but they are skeptic on the basis        of being manipulated. In this sense, it depends on the quality of the targeting campaign       

whether there is significant influence on the purchasing behavior. Goldfarb and Tucker        (2010) explain how targeting and obtrusiveness of online ads result in increased purchase        intention when used separately. If an ad is well targeted and obtrusive in the same time the        positive effects held by each conditions separately disappear under the increased privacy        concerns of people, who feel their personal space is being invaded by brands and therefore        counteract in a negative manner. 

The scholars in the field of consumer response to advertising explain how sometimes        a well targeted campaign may result in negative beliefs on the behalf of the viewer, who        might feel manipulated (Campbell, 1955). In this situation the effect of targeting on purchase        behavior will be negative, being perceived as circumspect. Specifically in the online        environment the chances of this effect to appear are even higher since behavioral targeting        is possible; so despite the fact that people accept targeted ads on the grounds that what ads        promote is useful information about products, brands, services etc (Wang, Chen and Chang,        2008), finding the right balance between how intrusive an ad is and to what extent that ad is        needed by the customer is still a debatable topic for each particular case. The reason why        ads may be considered intrusive, bulging and interfering is the privacy concerns. 

According to the definitions of Yan, Jun et al. (2009) and other academic papers on        BT, and generalizing it as a proper targeting method, we can argue that proper targeting is        used to influence consumers. This is based on the credentials of a credible interaction with        the consumer and based on the combination of user­data provided by Social Media        Networks.  

       

2.4. Conceptual Model & Hypotheses 

The goal of this study is to extend the knowledge of scholars and practitioners on the        relationship between Credibility and the Purchase Intention of online content consumers.        The paper is analyzing this relationship in the context of Proper Targeting as a moderator of        the need of credibility in influencing the purchase behavior. The research question is as        follows: ​ Does targeting remove the need for the credibility in an online setting? In this                              context friends (or peers) stand for UGC (User­Generated Content) and brands for BGC        (Brand­Generated Content). The relationship is first tested under Poor Targeting conditions,        and then under Proper Targeting conditions.             

The framework proposed will be tested using a 2x2 factorial design: two sources of        high and low credibility (UGC and BGC) tested under two moderating conditions, Poor and        Proper Targeting. The model will be better explained in the following figure (Figure 1):  

      Figure 1. 

 

The horizontal axis stands for the level of Credibility of the advertising source and the        vertical axis for the level of Targeting of the ad; the tables inside the system of axis form the        conditions based on which the Purchase Behavior of an online content consumer is being        influenced. 

 

The following two hypotheses will be tested:   

● H1: A higher level of credibility will positively influence the purchase likelihood. 

● H2: A better targeting strategy will negatively moderate the influence of credibility on        the purchase likelihood. 

      

3. Methodology 

This chapter firstly explains the purpose of this research and then describes the        design that was used according to the other relevant academic paper on this topic. The third        subchapter explains the motives behind choosing the central product of the experiment and        the SNS platform to support the study. The last subchapter is about the survey creation and        the data collection procedures. This subchapter is divided into five sections that describe in        detail the procedures followed in the survey, the introductive part of the questionnaire and        the conditions following the introductory sections. Then the data collection procedure is        explained alongside the two measures of the questionnaire with their constructs.  

 

3.1. Purpose of the research  

To test the hypotheses described in the previous chapter an empirical study was        designed and conducted over the internet. The experiment follows a real life situation        expressed by brands and people on social media. The experimental conditions try to        recreate common situations consumers encounter at a daily basis when surfing social media        platforms.  

This research was conducted to fill in a gap in the existing literature on the        moderating effects of targeting on the purchasing likelihood that people enhance when using        social media. Nevertheless, the paper is also of important relevance from a managerial        perspective in the online advertising field. The results of the experiment should improve the        knowledge of marketers on how UGC and BGC perform in social media in a century driven        by trust between people and privacy concerns regarding official organizations and        institutions. The experiment explains how brands and regular web users influence their        audience with the content they create, following the same theories that fathers of marketing        set up long ago, as Petty & Cacioppo (1979, 1983) , Campbell (1955). The experiment seeks       

to recreate most optimal targeting conditions and under this circumstances to switch the        consumer perception on their credibility towards branded content. If well­targeted content        reaches the desired audience and enables a positive purchase intention at a close rate as        the UGC enables, than this research will prove that targeting is the necessary moderator in        the relationship between brands and consumers’ purchase intention. The theory could also        be applied on other outcomes that brands are looking for in their marketing campaigns, but        this particular experiment uses purchase intention to measure the effects of targeting based        on Petty & Cacioppo (1983) work. To sum up this experimental research provides a better        understanding of the variables in advertising that drive purchase behavior.  

 

3.2. Research design  

Following the academic structure of such a research paper specific steps were        followed to conduct the survey. The process started by covering the existing literature in        practice and working on the question of what affects purchase intention in advertising        campaigns. In order to answer the hypotheses the academic work of Kaplan & Haenlein        (2010) and Petty & Cacioppo (1979) was of high relevance. Based on the review of their        work the following variables were found to influence purchase likelihood in advertising:        credibility of the content creator source and the nature of the source (Brands vs Peers).        Following this variables the research question was defined: ​       Does targeting remove the need          for credibility in an online setting? 

Following the formulation of the research question a comprehensive relevant        literature was covered in order to explain the variables of the framework and the        relationships between them. All connections between Content Source, Targeting and        Purchase Intention were analyzed and explained and four experimental conditions resulted.        The four conditions were further explained by only two hypotheses to ease the measurement       

of the results. Testing the four conditions was possible through an online experimental        survey. The main arguments for choosing this empirical method of data collection was that it        could generate the necessary amount of responses in a short period of time and because        setting up the experiment was affordable and many options were available.  

The deductive approach of this study best provided answers using a cross­sectional        survey design. This method of research provided data to which applied statistics tests were        used as the next step after the hypotheses formulation. Finally the conclusions were drawn        and applied at an abstract level, further to be used as general findings. This method of        research has been chosen as it can best measure attitudes and behavior and then        generalize the sample to the population of online world participants. 

 

3.3. Choice of product  

The central cue of this experiment is a daily used product, that recreates real life        social media posts. Facebook was used as the experimental platform, and the post was        created using Facebook posting service. The platform was chosen as it is the most common        used social media network at this moment, with 1.5 billion users across the world        (www.facebook.com). Facebook provides companies strategic tools to promote their        products and brands by targeting their followers and other users of the platform through        behavioral techniques.  

The choice of the central product of the experiment was based on the personal        network of possible respondents of the researcher. The fictitious product, a (Lordman) Bike        eliminated the effect of cognition or familiarity as the brand (Lordman) does not exist. In this        sense previous attitudes and affect did not influence participants’ choice of answers in the        questionnaire.  

The manipulation of the two independent variables was as follows: two levels were        considered for the credibility of the source, specifically high or low. For high credibility a        regular user of Facebook was assigned while for the low credibility condition brands were        considered. The two levels of credibility were each tested in poor and proper targeting        contexts, which were used to moderate the purchase intention. The moderator variable        manipulates the relationship between credibility and the purchase intention by reducing the        need of credibility when proper targeting is applied and vice versa. The moderating        conditions were based on the personal network of the researcher and behavioral practices        were used to describe the context of the ad placement, close to the real environment of the        majority of the respondents:  

1. Poor Targeting Condition: “Assume you will be studying in Amsterdam and        you receive housing located ​       20 km away from the University ​           and the city      center. The most accessible and appropriate method of transportation around        the city is by​ public transportation. “ 

2. Proper Targeting Condition: “Assume you will be studying in Amsterdam and        you receive housing located ​         2 km away from the University ​         and the city      center. The most accessible and appropriate method of transportation around        the city is by​ bike. “ 

  

3.4. Survey setup & Data Collection 

3.4.1. Procedure 

The questionnaire was created using Qualtrics.com survey software to capture and        collect responses from SNS users. The first steps in creating the questionnaire were        introducing the respondents to the purpose of the research. Secondly they were presented        one of the four experimental conditions, according to the randomization assigned by       

Qualtrics. The introductory section created the targeted context before exposing the        participant to the experimental image, a Facebook post. After enough data collected, all        responses were translated into SPSS Statistic Software to measure the variables. 

Firstly the questionnaire presents the purpose of the research and the time        necessary to fill in one survey, of about 3 minutes. Respondents were asked to express        personal opinions and attitudes about a Facebook post that was similar to the ones on a        regular News Feed. If the respondents found the survey had the possibility they were asked        to fill it in more than once as every time opened the survey would be different (based on the        four conditions). (Appendix 1) 

The second page started with the manipulation material: first the contextual set up        was presented in order to create a Proper Targeted environment or a Poor Targeted        environment. The high (Carla Thompson) or low (Lordman Bike) credibility sources were        added with a picture created via Facebook posting service to induce the respondent the        feeling of a real social media post. Every time the survey was opened one of the four        conditions were presented to the respondent.  (Appendix 2.1 ­ 2.4) 

The questions following the experiment were on the same page so that the        respondent could check the experimental condition anytime until the end of the questions.        The survey created consisted of three types of questions (Appendix 3 ­ 5): two single answer        questions, two scale question (Strongly Disagree ­ Strongly Agree) and one open question.        The two single answer and the open question were demographic questions about education,        gender and age (Appendix 5). The two main questions mostly relevant for the model were        placed after the manipulation material and each of them are relevant for one hypothesis.  

     

3.4.2. Introduction Part

  

To test the two hypotheses a 2x2 (UGC and BGC x Proper and Poor Targeting)        design was used when setting up the experimental conditions. The four experimental        conditions are all based on the same image, but under a different context (Appendix 2.1 ­        2.4). Two fictitious content sources were created, LORDMAN Bike Official for BGC (low        credibility) and Carla Thompson for UGC (high credibility), using the two situational contexts        that were created for Proper (LTG) and Proper (PTG) Targeting.   C1. UGC + LTG ­ post offered by a High Credibility source in a Poor Targeted context   (Appendix 2.1)  C2. UGC + PTG ­ post offered by a High Credibility source in a Proper Targeted context  (Appendix 2.2)  C3. BGC + LTG ­ post offered by a Low Credibility Source in a Poor Targeted context  (Appendix 2.3)  C4. BGC + PTG ­ post offered by a Low Credibility Source in a Poor Targeted context   (Appendix 2.4)   

The introductive part initiates the respondent into an environment that recreates a        real­life context, and is based on behavioral targeting techniques used in advertising. The        contextual setting is based on students in Amsterdam. The reason students were chosen is        because most of them do not own cars therefore bicycles and public transportation have the        most potential to be chosen by a student as means of transportation. Furthermore        Amsterdam was chosen as the location of the experiment as most respondents are studying        or living in Amsterdam and the city is famous for using bikes and public services as the most        common methods of moving around the city. The need of using a bike or public        transportation was highlighted in the introductive part so that the respondent is aware of       

what is the most appropriate mean of transportation in their condition. The context created in        the experiment is based on Behavioral Targeting practices from the online environment        where marketers use data similar to the one described here to launch their marketing        campaigns.  

 

3.4.3. Conditions 

The four conditions that stand for the manipulation material of the experiment are        designed to recreate content that can be found in Social Media.  

C1. UGC + LTG ­ post offered by a High Credibility source in a Poor Targeted context  The first condition is a post generated by the fictitious person Carla Thompson (high        credibility source) recommending a bike for respondents that do not need a bike (Poor        Targeted Context). 

C2. UGC + PTG ­ post offered by a High Credibility source in a Proper Targeted                                context 

The second condition is a post generated by the fictitious person Carla Thompson        (High Credibility source) recommending a bike for respondents that need a bike (Proper        Targeted Context). 

C3. BGC + LTG ­ post offered by a Low Credibility Source in a Poor Targeted context  The third condition is a post generated by the fictitious brand Lordman Bike (low        credibility source) recommending their bike to respondents that do not need a bike (Poor        Targeted Context). 

C4. BGC + PTG ­ post offered by a Low Credibility Source in a Poor Targeted context   The fourth condition is a post generated by the fictitious brand Lordman Bike (low        credibility source) recommending their bike to respondents that need a bike (Proper        Targeted Context). 

Each participant in the survey will be randomly assigned to one of the four conditions        using the research software Qualtrics.com. The study did not use a control group as it        focused on the differences between the four conditions. Since behavioral targeting is about        using the most appropriate strategies for each platform it applies to, the posts of Carla        Thompson and Lordman Bike were accompanied by a picture of a pleasant girl on the new        bike launched by Lordman. The same picture was used for all four conditions, as pictures        increase the efficiency of an ad being more attractive than plain text.  

 

3.4.4. Sample & Data collection 

Simple random sampling will be used to select participants in the questionnaire.        Master students in Netherlands will be contacted through Social Media channels and then        asked to forward the questionnaire to other Web users. We consider that any Web user is        part of the relevant sample of the population. Nevertheless, the majority of respondents were        from the personal network of the researcher. The research connected all the resources        towards a sample as large as possible. More than 330 responses were collected, a sample        that ensured the level of relevance to generalize the findings after the analysis of the data.        Out of the total number of responses only 50 responses for each condition were fully        completed, therefore a total of 200 responses were used in measuring the data.  

 

3.4.5. Measurements 

The second page of the survey started with the manipulation material: first the        contextual set up was presented and then the high (Carla Thompson) or low (Lordman Bike)        credibility sources were added to the Proper and Poor contexts. The picture created via        Facebook posting service induced the respondent the feeling of a real social media post.  

   

The four conditions: 

1. Next you will see a Facebook message posted by an ​       unknown person Carla      Thompson who is recommending buying a LORDMAN bike. ​       You are not      friends ​on Facebook with this person. (Appendix 2.1) 

2. Next you will see a Facebook message posted by your very close friend              Carla Thompson ​   who is recommending buying a LORDMAN bike. You are        already friends ​on Facebook with this her. (Appendix 2.2) 

3. Next you will see a Facebook message posted by LORDMAN Bike        Amsterdam Official recommending a bike. You are not a follower of this                  brand on Facebook and ​don't know this brand. ​(Appendix 2.3) 

4. Next you will see a Facebook message posted by LORDMAN Bike        Amsterdam recommending a bike. You are ​       already a follower ​       of this brand    on Facebook and you ​know ​this brand. (Appendix 2.4) 

 

In order to measure the constructs the following two questions were formulated:    

1. “What do you feel regarding the source who posted?​” (Appendix 3) 

Credibility (independent variable) 

The first question is designed to test the respondent’s view on the perceived        credibility of the source of the post, either is the fictitious friend Carla Thompson or the        Lordman Bike brand. The trust and credibility that people assign to a brand related post is        expected to vary according to the quality of the targeting. Naturally targeting should be        considered only in the case of brands, but in social media brands are being promoted by        regular users as well. And regular users have their personal audience, followers or friends,        that have more or less similar preferences, which are the basic concepts of a targeting        strategy. Furthermore social media gives access to anyone to follow or be friends therefore       

the concept of being friend does not apply as in the real life. In this sense poor or proper        targeting can be applied to the audience one has on his social media profile, and in this        particular case on the Facebook account.  

Source credibility is the combined effect between trust and expert knowledge. A        study of Willemsen et al. (2012) provides an 8­item measurement scale (four of the items        measure trust – honest, sincere, reliable, trustworthy ­ and four measure the level of expert        knowledge – skilled, experienced, knowledgeable, qualified). According to the authors a        seven­point semantic differential scale from 1 (very low) to 7 (very high) was the most        appropriate. For this study the four measures of trust by Willemsen et al. (2012) were used        alongside a fifth clear measure of Credibility that was added to assure on the reliability of the        first four answers (Appendix 3). 

 

2. “After seeing this post on my News Feed:” ​(Appendix 4) 

Purchase Intention (dependent variable)  

The second question is designed to measure the effect of the two independent        variables, credibility and targeting as moderator, answering both hypotheses. The purchase        intention that builds up in the respondent’s cognition was being measured using the same        7­points scale as in the first question.  

The purchase intention only influenced by the Credibility of the source was measured        by comparing results between UGC conditions and BGC conditions, in order to answer the        first hypothesis. To conclude whether the second hypothesis applies the targeting moderator        was taken into consideration as reducing the effect of credibility. The differences between        Poor and Proper Targeting conditions were measured to conclude if targeting can reduce the        positive impact of friends credible recommendations against brand advertising. 

The only dependent variable, purchase intention, is one’s deliberate plan to make an        effort to buy a product from a brand (Spears & Singh, 2004). The scale to measure purchase        intention comes from the study of Dodds et al. (1991). Their research provides a 3­item        scale with the same scale (1­7) used in the first question: The likelihood to purchase a        product, the probability of purchasing a product and the willingness to purchase a product.        The scale had answers from 1 (Strongly Disagree) to 7 (Strongly Agree). The items were        modified to match the conditions of this experiment, the intention to purchase a LORDMAN        Bike (Appendix 4).  

 

4. Data Analysis

 

The data analysis chapter is divided into sections that describe the entire data        analysis process. The first subchapter explains the first steps into verifying the errors and the        reliability of the data. Finally the scale means, standard deviations and the correlation tables        were computed for each of the four conditions. The second subchapter explains the        performance of the ANOVA tests, alongside Levene’s test and the Multiple Comparisons        table using Scheffe. This subchapter explains the effects of the moderator in the second        hypothesis, and the results are summarized in the graph at the end of the subchapter. The        third section of the data analysis process included the hierarchical regressions performed to        examine the relation between the independent variable and the dependent variable, hence it        explains the first hypothesis. Finally the last part of the chapter is testing the two hypotheses        formulated in the previous pages. A summary of the data analysis results is being provided        in a representative table.  

     

4.1. Analysis Steps 

The first step in the data analysis process was conducting a frequency check to verify        if there were any errors in the data. More errors was found, cases where participants wrote        the gender in the age section or did not finish the survey. These cases were excluded. The        cases where any piece of data was missing were excluded so that only the cases with        complete responses within the entire variable set were used. In total 200 responses were        kept for the data analyses, out of 331. The 200 were split equally amongst the four        conditions.  

The second step was the reliability analyses of all items of Source Credibility and        Purchase Intention. Three out of four Cronbach’s Alpha of the sources were higher than 0.7.        High Credibility (UGC) Sources had Cronbach’s Alpha of .894 (Poor Targeting) and .362        (Proper Targeting) in the first 2 conditions, and Low Credibility Sources (BGC) had        Cronbach’s Alpha of .919 (Poor Targeting) and .835 (Proper Targeting). The Cronbach’s        Alpha of only .362 of the UGC under Poor Targeting conditions could be increased to .896 to        create a reliable dataset after the third item was deleted (“The sender of the message was        sincere”). Therefore the third item was excluded considering that four items are enough to        generate useful data. In this case the results confirm that the five scales are good for three        conditions: honesty, sincerity, reliability, trustworthiness and credibility, and four scales for        one condition. The Correlated Item­Total were all higher than 0.3 in all the four conditions,        and the Cronbach’s Alpha if Item Deleted lower than 0.1 in all four contexts.  

Nevertheless, Purchase Intention Cronbach’s Alpha is 0.629 in the UGC and Poor        Targeting Conditions. If any of the three construct of this question was removed the        Cronbach’s Alpha would decrease even more, so all items were kept even the indicator        remained below the recommended level of 0.7. Still all Cronbach’s Alpha if Item Deleted        were lower than 0.1 and all Item­Total Correlations were higher than 0.3 so the scale was       

considered reliable at .629. In some cases a Cronbach’s Alpha higher than 0.5 is still usable.        In the UGC and Proper Targeting condition the Cronbach’s Alpha was 0.625 with the        Item­Total Correlations all higher than 0.3 and The Cronbach’s Alphas all lower than 0.1,        based on which the scale was considered reliable following the same standards as in the        first condition. In the BGC and Poor Targeting Condition the Cronbach’s Alpha was above        0.7, with a value of 0.828. This scale was reliable and the exclusion of some constructs was        not necessary. Furthermore all Cronbach’s Alpha if Item Deleted were lower than 0.1 and        the Item­Total Correlation were all higher than 0.3. Finally the BGC and Proper Targeting        condition had Cronbach’s Alpha of 0.806 so no item needed to be excluded. Also all the        Item­Correlation were higher than 0.3 and all Cronbach’s Alpha if Item Deleted were lower        than 0.1. Since all scales were reliable, the following step was to compute the scale means        for the Source and Purchase Intention, which were coded into different variables. The mean        function was applied for all five items of Source Credibility and for the three items from the        Purchase Intention scale.  

The next step was computing the means, standard deviations and correlation table        for each of the four conditions, which are presented in the following four tables:   1. UGC +   Poor Targeting  Mean  Std.  Deviation  1  2  3  4  5  1.Source TOT  3.65  1.20  1          2.Purchase TOT  2.54  0.77  0.449** /   0.001  1        3.Gender  1.56  0.50  0.107 /   0.460  ­0.024 /   0.870  1      4.Age  24.32  3.58  ­0.070 /   0.630   0.066 /   0.649  0.057 /   0.693  1    5.Education   2.16  0.68  0.175 /   0.225  0.206 /   0.152  0.210 /   0.142  0.389** /   0.005  1      Table 1. 

 

In the first condition (Table 1) the Pearson Correlation between Source Credibility        and Purchase Intention is significant and explains .449 of strength of the linear relation        between the two variables. The correlation is positive and there is enough evidence ( Sig =        .001) to claim that the correlation exists.     2. UGC +   Proper Targeting  Mean  Std.  Deviation  1  2  3  4  5  1.Source TOT  5.29  2.16  1          2.Purchase TOT  4.98  0.90  ­0.78 /   0.589  1        3.Gender  1.70  0.46  0.098 /  0.496  0.006 /   0.964  1      4.Age  24.96  4.85  0.168 /   0.242  ­0.060 /   0.677  ­0.169 /  0.241  1    5.Education   2.34  0.68  0.218 /   0.128  ­0.189 /   0.189  ­0.058 /   0.691  0.187 /   0.192  1    Table 2.   

In the second condition (Table 2) the Pearson Correlation between Source Credibility        and Purchase Intention does not have enough strength to explain the linear relation between        the two variables. The correlation is negative and there is not enough evidence ( Sig = .589)        to claim that the correlation exists.           

3. BGC +   Poor Targeting  Mean  Std.  Deviation  1  2  3  4  5  1.Source TOT  3.78  01.23  1          2.Purchase TOT  2.51  0.95  0.515** /   0.000  1        3.Gender  1.62  0.49  0.278 /   0.051  0.339* /   0.016  1      4.Age  24.34  2.71  ­0.355* /   0.011  ­0.221 /   0.122  0.007 /   0.961  1    5.Educ   2.26  0.63  0.175 /   0.225  0.067 /   0.642  0.391** /   0.005  0.114 /   0.431  1    Table 3.   

In the third condition the Pearson Correlation between Source Credibility and        Purchase Intention is significant and explains .515 of strength of the linear relation between        the two variables. The correlation is positive and there is enough evidence ( Sig = .000) to        claim that the correlation exists.     4. BGC +   Proper Targeting  Mean  Std.  Deviation  1  2  3  4  5  1.Source TOT  5.03  0.86  1          2.Purchase TOT  4.85  0.97  0.499** /  0.000  1        3.Gender  1.52  0.50  ­0.216 /   0.132  ­0.160 /   0.267  1      4.Age  24.50  3.05  0.002 /   0.992  ­0.151 /   0.295  0.146 /   0.313  1    5.Educ   2.18  0.66  ­0.053 /   0.715  0.010 /   0.944  0.020 /  0.893  0.238 /   0.096  1    Table 4. 

In last condition the Pearson Correlation between Source Credibility and Purchase        Intention is significant and explains .449 of strength of the linear relation between the two        variables. The correlation is positive and there is enough evidence ( Sig = .000) to claim that        the correlation exists.  

 

4.2. One­way ANOVA 

The next step in the analysis process was performing a One­Way ANOVA analysis to        explain the effect of the independent variable “credibility of the source” on the dependent        variable the “purchase likelihood”. The reason why One­Way ANOVA was used is because it        compares the differences between and within the groups, which in this case are the four        experimental conditions.   

The first table when running ANOVA is the Descriptive Statistics table of the        dependent variable, the purchase likelihood (Table 5). The table shows the means of the        dependent variable in all four conditions, together with the standard deviations. The total        number of 200 respondents that was analyzed was split equally between the four conditions. 

  Table 5. 

The second table analyzed was Levene’s Test of Equality of Error Variances (Table        6). This test looks on the quality of the variances among the four condition to assure that the        purchase likelihood as the dependent variable has equal variance among the four        conditions. With a Sig=.819 greater than 0.05 Levene’s Test proves that the variances        among the four conditions are equal. 

  Table 6. 

The next table represents the results of the ANOVA test, called the Tests of        Between­Subjects Effects (Table 7). The independent variable in this table (CONDITION)        has the ANOVA test result F = 116.4545 and the P­value, also named the significance of the        test, of Sig. = .000. The significance is lower than 0.05 so there is a significant difference in        the purchase likelihood between the four conditions.   

The clinical significance of this result is explained by the Partial Eta Squared, or the        effect size of the experiment. Besides the fact that the relationship between the purchase        likelihood is statistically significant, by having a Partial Eta Squared value of .641 that is        closer to 1 than to 0, the relationship will also have a strong clinical effect. Furthermore,        having the Observed Power of 1.000, which is higher than .80, the analysis is considered        accurate, and the effect of the treatment (the four condition) is strong enough to generalize        the findings.  

Table 7. 

The next table was part of the Post Hoc Tests and included the Scheffe test to        compare the multiple correlations between the four conditions (Table 8). The table is making       

comparisons between the four conditions and compares them to every other elements of the        group. Every condition is compared to all the others individually, which is useful only        because the ANOVA test was significant. And since ANOVA was significant the next step        was to find which of the four conditions is significantly different.  

In order to find which condition is significantly different a column with the P­values        was created. Whenever the P­value was lower than .05 there is a significant difference in        that case. Out of the twelve comparisons, eight were statistically different confirming the        ANOVA results, but four show no statistical difference when compared. There is no        significant difference between the purchase likelihood with respect to the credibility of the        source when the first condition (UGC + Poor Targeting) is compared to the third condition        (BGC + Poor Targeting) as P=.999. In the same manner there is no significant difference        between the second condition (UGC + Proper Targeting) and the fourth condition (P=.910,        BGC + Proper Targeting), when comparing the third condition (BGC + Poor Targeting) with        the first condition (P=.999, UGC + Poor Targeting) or the fourth (BGC + Proper Targeting)        condition with the second condition (UGC + Poor Targeting). What the Multiple Comparisons        Scheffe test shows is that there is no difference between Low or High credibility sources        when the same targeting level is applied, either Poor or Proper or targeted.  

Finally the SPSS software provides the graph with the four experimental conditions        on the horizontal axis and the mean number of the purchase likelihood of the respondents        (Table 9). It is a quick method to visualize how high the purchase likelihood means are in the        Proper Targeted conditions, both with high and low sources of credibility, and the low values        that the purchase likelihood takes under Poor Targeted conditions, both in the case of high        and low sources of credibility. The significant difference (based on F = 116.4545 and Sig. =        .000) between the conditions and the means of both credibility sources in the poor vs proper        targeted conditions support the second hypothesis: A better targeting strategy will negatively        moderate the influence of credibility on the purchase likelihood. 

  Table 9. 

4.3 Hierarchical Regression 

Next a hierarchical regression analysis was performed to examine the relation        between the independent variable Source Credibility and the dependent variable the        Purchase Intention of the respondents. The hierarchical regression was performed after        controlling for gender, education and age. The three control variables helped to keep the        effect of Source Credibility constant in the Purchase Intention.