Search engine advertising : the impact of emotions and gender differences on search engine advertising effectiveness : a field experiment

(1)

SEARCH ENGINE ADVERTISING

THE IMPACT OF EMOTIONS AND GENDER DIFFERENCES ON

SEARCH ENGINE ADVERTISING EFFECTIVENESS:

A FIELD EXPERIMENT

Erna Omerčić

11872012

22.06.2018

Final Master Thesis

MSc. in Business Administration – Digital Business

Amsterdam Business School

Universiteit van Amsterdam

Supervisor: Sara Valentini

(2)

Statement of originality

This document is written by Student Erna Omerčić (11872012) who declares to take 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.

(3)

ABSTRACT

THE IMPACT OF EMOTIONS AND GENDER DIFFERENCES ON SEARCH ENGINE ADVERTISING EFFECTIVENESS: A FIELD EXPERIMENT

In light of the increasing importance of search engine advertising, it is essential to outline what ad content and targeting strategy can improve the effectiveness of search engine advertising. This study will focus the attention on the use of emotions in search engine advertising and the role of gender differences. A conceptual model has been developed based on hypothesis formed by previous theories in order to help answer the research questions: What is the impact of a gender targeting strategy on the effectiveness of search

engine advertising in terms on click through rates and conversion rates? How can search engine advertising be designed by using emotions to increase effectiveness? Does the interaction between gender differences and the use of emotions produce a differential impact on search engine advertising effectiveness? Data for this study were collected

through a field experiment. Cooperation with a Dutch tour operator allowed this study to be executed through use of Google Analytics. Eleven different groups within three comparable AdWords ad-groups have been involved in this experiment (N= 11.182). Results show that the use of emotions has an impact on click through rates and conversion rates. By contrast, targeting by gender differences is moderately effective. More specifically, two emotions have been examined: joy and surprise. These emotions have slightly different impact on click through rate and conversion rate, but both have a positive impact on the outcome results. This study contributes to the existing literature as this study shades lights on the role of emotions in search engine advertising and the role of gender targeting and their interaction effect. Additionally, to the sake of my knowledge, this is the first study investigating the role of emotions and gender through a field test done in collaboration with a company.

KEYWORDS: search engine advertising, search ad design, search engine advertising

(5)

1. INTRODUCTION

This dissertation will focus on the further identification of search engine advertising effectiveness in regard to ad design. This is relevant as search engine advertising is becoming the most essential marketing channel used by businesses to extend market and to win market rivalry (Indartoyo, Rahayu, Budiwan, Bismo, & Sadeghifam, 2016). Today, search engine advertising is a primary marketing channel in online advertising (Kamis & Stohr, 2006; Jansen, Z. Liu, & Simon, 2013). According to eMarketer (2015), Ratcliff (2016) and Boswell (2017) Google is by far the most used search engine navigation tool. Therefore, in this study only search engine Google (data) will be taken into consideration.

Due to the novelty of search engine advertising, a bouquet of related topics requires further investigation. There seems to be a discrepancy between traditional offline marketing effectiveness and online marketing effectiveness (Danaher, Wilson, & Davis, 2003). Targeting is often used in traditional marketing, according to Bergemann & Bonatti (2011) targeting in advertising might lead to higher effectiveness as it improves the quality of the connection between the consumer and the advertising message. While a study by Jansen, Moore & Carman (2012) found the opposite effect in search engine advertising. The authors examined the relationship of gender targeting on impressions, clicks, costs, sales, orders and items in search engine advertising (Jansen, Moore & Carman, 2012). Their results contradict the general literature about targeting, that states that targeting allows marketers to show personalized content, leading to more effectiveness for marketing (Aslam & Karjaluoto, 2017; Chaffey, 2011). This discussion stimulates further research on the effect of gender targeting on search engine advertising effectiveness.

(6)

consumer’s decision-making process (Bigné, Andreu & Gnoth, 2005; Mehta & Purvis, 2006). Males and females do have significant variances, especially with regard to emotions. Huddy (2002) stresses the statement that females are more attached by emotions. The role of gender differences and the use of emotions may have intriguing interaction effects on search engine advertising effectiveness (Allen & Haccoun, 1976; Grossman & Wood, 1993; Fisher & Dubé, 2005). Specific emotions may have a differential impact on consumers behaviour (Li, Walters, Packer & Scott, 2017). As mentioned by Teixeira & Wedel (2012), positive emotions as joy and surprise are an important approach to engage customers in advertisements. This study contributes to the literature as it exclusively investigates the interaction effect of the emotion joy and surprise with gender on search engine effectiveness.

Although some researchers studied the impact of emotions in online advertising (Lohtia, Donthu, & Hershberger, 2003; Goldfarb & Tucker, 2011) or the role of gender targeting in search advertising (Banerjee & Dholakia, 2012; Kemp, Bui & Chapa, 2012; Teixeira & Wedel, 2012; Jansen, Moore & Carman, 2012), to the best of knowledge, no previous researcher investigated gender targeting and the interaction with the use of emotions in search engine marketing. In addition, this research overcomes some of the limitations of previous studies, as it uses a large sample of actual search engine advertisements experiment to examine the (interaction) effect of gender and the use of emotions on the effectiveness of search engine advertising. This was not achievable before, because Google only introduced Demographics For Search Ads (DFSA) near the end of 2016 (Google, 2016; Irvine, 2017). Because of the relatively short existence of DFSA, this practically means no previous academic studies based on demographic targeting in search engine advertising would have been able to rely on field data.

(7)

Company:

The execution of the experiment of this study is done in cooperation with a company. The company is a leisure specialist with more than 15 years of experience in the leisure industry. Offering hotel-room deals in 8 different countries, moreover over million travellers are served yearly.

Research objective:

Evaluating the contribution of (a) emotional ad content, (b) targeting by gender and (c) their interaction effect, in order to identify how it (a, b and c) can contribute to click through rates and conversion rates in search engine advertising.

Research Questions:

Q1) What is the impact of a gender targeting strategy on the effectiveness of search engine advertising in terms on click through rates and conversion rates?

Q2) How can search engine advertising be designed by using emotions to increase effectiveness?

Q2a) is the use of emotions able to increase click through rates and conversion rates? Q2b) if yes, are there differences taking into consideration different emotions? Which specific emotion is more effective?

Q3) Does the interaction between gender differences and the use of emotions produce a differential impact on search engine advertising effectiveness?

(8)

2. THEORETICAL FRAMEWORK

This chapter will provide an overview of earlier scientific research concerning (gender) demographic targeting and search engine advertising. The subparagraphs will be subdivided into the variables of the research model. After, the dependent variables and the independent variables will be discussed, and hypotheses will be developed. The last subparagraph will show the conceptual framework with the assumed relationships.

2.1 SEARCH ENGINE ADVERTISING

In this subparagraph search engine advertising effectiveness will be discussed by first looking at online advertising, then at search engine advertising in Google and finally, at the most often used key performance indicators.

For many marketers, advertising is their daily occupation. Traditionally, advertising is identified as a form of communication that is structured, paid, and non-personal (Belch & Belch, 2004; Duncan, 2002). Advertising is designed to broadcast information about services, ideas and goods in a convincing and creative style (Belch & Belch, 2004; Duncan, 2002). Belch & Belch (2004) and Duncan (2002) also state that the main goal for any marketing campaign that consumers are exposed to is, is to influence the brand awareness, through feelings, or behaviour in a desired way about a brand.

Advertising is not a new subject in literature as advertising exist for quite some time already. However, what is relatively new is online advertising. Online Advertising occurred due to digitalization (McMillan & Childers, 2017). Already in the 1990s internet made new forms of digital communication possible (Negroponte, 1995). Then in 2005, the term Web 2.0 was introduced. The Web 2.0- period could be identified with online occasions that have

(9)

characteristics of user participation (Bulik, 2006; Chaffey, 2011). Nevertheless, in academic advertising literature, new media terms such as viral videos, blogging, advergaming, and keyword search, were largely presented by 2014 (An, Jin, and Park 2014; Cho, Huh, and Faber 2014; Fernando, Suganthi, and Sivakumaran 2014; Yoo 2014). Additionally, in 2014, researchers discovered digital media approaches that could increase advertising outcomes by conducting academic studies (Duff, G. Yoon, & Anghelcev, 2014).

When discussing about online advertising, several ways exist to execute online communications. According to Chaffey (2011) marketers can make use of online PR, online partnerships, interactive (banner) ads, opt-in e-mails, social media marketing and search marketing in order to lead online users to the company’s website. Search marketing is becoming most the essential marketing channel that is used by businesses in order to extend market and to win market rivalry (Indartoyo, Rahayu, Budiwan, Bismo, & Sadeghifam, 2016). As stated by Indartoyo et al. (2016) search marketing can be separated in two practises search engine optimization (also known as: organic search) and search engine advertising (also known as: paid search).

With search engine advertising, marketers can bid on the chance to show ads next to organic search results in a search engine like Google (Google, 2018).

Zhang et al. (2014) state that in advance of the auction, advertisers have to create ad groups and bids on some keywords for the ad group with their match types. These so-called match types can either be broad matched (advanced match) or exact matched. The Google-constructed algorithm behind search engine advertising depends on two main elements: quality score and bid (Zhang, Zhang, Gao, Yuan, & Liu, 2014).

(10)

Quality scores in Google AdWords are estimations of the quality of the search ad. A high quality score is based on historical data about the click through rate, ad relevance and the landing page of the ad (Chen et al, 2009; Google, Over de kwaliteitsscore, sd). A high quality score can decrease the cost per click for a business (Zhang, Zhang, Gao, Yuan, & Liu, 2014). In Google’s AdWords bidding is conducted per keyword or search term. After identifying what keywords a marketer wants to bid on, a match type should be set. Then comes the decision of how much will be maximally paid for each click. This bidding system is called: Pay Per Click (PPC) (Google AdWords, sd).

Marketers only have to pay per click of an online user in case of choosing for a PPC pricing model. Online advertising through Google AdWords allows marketers to target ads anytime in order to reach a specific group of online traffic (Google, AdWords: SEO vs. PPC?, 2018). Today, Search engine advertising is a primary marketing channel in online advertising (Kamis & Stohr, 2006; Jansen, Z. Liu, & Simon, 2013). According to eMarketer (2015), Ratcliff (2016) and Boswell (2017): Google is by far the most used search engine navigation tool. Therefore, only data from Google’s search engine will be taken into consideration in this study. Also, because it is Google’s only advertising tool within search engines. In addition, worldwide a lot of money is being invested in the world of Search Engine Advertising. It is forecasted by eMarketer (2015), based on a multipronged approach, that search ad spending will rise to 118.81 billion US Dollar by 2018. Next to that, Search engine advertising seems to be considered as the best way to draw visitors to the company’s website (Rangaswamy, Giles, & Seres, 2009; Karjaluoto & Leinonen, 2009).

(11)

Several key performance indicators (KPI) exist to measure the effectiveness of a search engine advertising campaign. When analysing the studies of Rutz & Trusov (2011), Rutz & Bucklin (2011) and Ghose & Yang (2009), click through rates (CTR) and conversion rates (CONV) are common Key Performance Indicators that can be used to measure the success of search engine advertising practices.

Click trough rate refers to the ratio between the number of clicks of an ad and the number of times an advertisement has been shown to the online customer. So, the formula that belongs to CTR is: CTR= clicks/impressions.

Conversion rates indicates the ratio of the number of purchases that an online user makes from the advertisement as a percentage of the number of clicks on the advertisement. So, the formula that belongs to CONV is: CONV = conversions/clicks. According to Yang, Lin, Carlson & Ross (2016), the below KPIs can be used to judge search engine advertising effectiveness. Basically, this confirms the previous found literature about search engine advertising measurements.

It seems in literature that several aspects can influence the effectiveness of online advertising. Applying targeting is known for having a positive effect on online advertising (Bergemann & Bonatti, 2011). This will be further discussed in the next subchapter.

(12)

2.2 DEMOGRAPHIC TARGETING

In this subparagraph, the main focus will be on demographic (gender) targeting. First, targeting in advertising will be discussed. Second, literature about demographic targeting will be touched upon. And finally, the moderator emotion in communication will be highlighted.

Bergemann & Bonatti (2011) state that due to internet many opportunities occurred for targeting in online advertising. As mentioned in the previous subchapter, targeting in advertising might lead to higher effectiveness as it improves the quality of the connection between the consumer and the advertising message (Bergemann & Bonatti, 2011). In addition, Edwards et al. (2003) and Wang et al. (2008) argue that targeting advertisements is perceived as useful because consumers observe the information as valuable.

The definition of targeting comes from segmentation. Market segmentation separates a homogeneous market in smaller divisions based on several characteristics (like: resources, attitudes, buying practice, wants and needs, locations etc.) in order to reach the target audience more effectively and efficiently with goods and services that match their unique desires (Armstrong, Adam, Denize, & Kotler, 2014; G. Budeva & R. Mullen, 2014). Market targeting includes breaking a market into segments and then focusing marketing efforts on one or more key segments (Kotler & Armstrong, 2010).

According to literature, several techniques of targeting can be applied in marketing. Armstrong et al. (2014) state that demographic targeting is the most used method for targeting due to its easiness of measuring compared to other ways of targeting, due to the type of variables. Next to the fact that demographic targeting is by far most used, it is also

(13)

significantly important, as demographic characteristics always need to be identified when targeting no matter what type of method is applied (Armstrong, Adam, Denize, & Kotler, 2014). Demographic targeting can be defined as a manner of separating the market based on demographic characteristics like gender, age, generation, educational level, income, family status, occupation, religion, race, ethnicity and nationality (Martin, 2011; Armstrong et al., 2014). In terms of search engine advertising, this form of targeting is applied in order to narrow down to whom search engine advertisements will be shown (Chaffey, 2011).

GENDER

One popular way of demographic targeting is gender targeting. Gender has been an interesting topic of research across research areas and across years. Males and females do have significant variances and it is therefore obvious to approach them both differently (Banerjee & Dholakia, 2012). A study by Klink (2008) found that there is a difference between how males and females response towards different content characteristics. Kempf et al. (2006) state that females evaluate content in a more detailed process than males. Different genders have different response to both marketing and advertising stimuli (Banerjee & Dholakia, 2012). An example is that males do not experience online shopping differently than conventional shopping (Dittmar, Long, & Meek, 2004). Whereas females prefer conventional shopping rather than online shopping (Dittmar, Long, & Meek, 2004). McMahan et al. (2009) state that males like interactivity (e.g. videos and web-games), while females are more likely to respond to the company’s communication links. In addition, males have a strong preference for efficiency and convenience (Rodgers & Harris, 2003), whereas females prefer emotional and social interactions (Rodgers & Harris, 2003). In addition, Huddy (2002) stresses the statement that females are more attached by emotions. To add on, McGuinness (1976) states that females are more sensitive than males in most

(14)

emotional ways. The application of gender targeting seems to have an influence on click trough rates, also females seem to be less affected than males (Banerjee & Dholakia, 2012). Besides that, Melnyk et al. (2009) found that males and females have different behaviour in responses towards company’s offers. Due to all the differences between males and females, it seems straightforward to approach each gender according to their preferences in order to increase marketing performance (Chaffey, 2011). It leads to hypothesize the following:

H1a: Applying gender targeting vs. no gender targeting leads to a higher click through rate. H1b: Applying gender targeting vs. no gender targeting leads to a higher conversion rate.

2.3 USING EMOTIONS IN ADVERTISING

Emotions in advertising seem to be a key factor in the strategy to customer engagement in advertisements (Kemp, Bui & Chapa 2012; Teixeira & Wedel 2012; Kover, Goldberg, & James 1995). In traditional marketing, using emotions in advertising has not gone unnoticed. Emotions have an impact on consumers’ decisions making process (Mehta and Purvis, 2006). Emotional appeal in advertising is a popular technique of increasing attention and generation action (Holbrook and Batra, 1987). Particularly in the tourism industry, as it is a service, which is often valued based on emotional experiences. As emphasized by Friestad & Thorson (1986) the emotional meaning of an ad can be better understood by focusing on the process and experience of a person towards the ad (Goldfarb & Tucker, 2011). In order to understand the process of different types of messages, the Elaboration Likelihood Model (ELM) can be used. The ELM entails two directions of persuasion, through which the message affects the attitude (Petty & Cacioppo, 1986). One of the two is the central route, the other one is the peripheral route. According to Ho & Bodoff (2014) the central route is when there is a high depth of processing. This occurs when a person

(15)

processes the message carefully. Whereas the person who follows the peripheral route, processes the message with low depth and less careful, because the process is based on associations. In the context of advertising, this means that people who find an advertisement uninteresting, the information will be processed via the peripheral route which has a negative effect on purchase decision, attention and recall. Alternatively, in order to create a positive influence on the purchase decision, the attention and recall an advertisement should be interesting, because then it is processed via the central route (Lancée, 2014).

It is proven by Bigné, Andreu & Gnoth (2005) that positive emotional conditions have an influence on the customer’s behavioral intentions. What also influences the behavior of customers, and mainly their purchase intent, is the content of an ad. From historical literature, it is known that warm emotions (e.g. love, joy, surprise) have a positive influence towards advertisements as it ultimately increases purchase intentions (Biel & Bridgwater, 1990; Stayman & Aaker, 1993). In addition, Teixeira & Wedel (2012) state that positive loaded emotions, like surprise and joy, seem to have a key influence in the strategy to customer engagement in advertisements.

When analysing present literature, two theories of emotions are leading in advertising research. These are Basic Emotion Theory and Dimensional Theory. The Basic Emotion Theory refers to the emotions which are present from birth, which are happiness, sadness and anger (Chamberlain & Broderick, 2007). While the Dimensional Theory identifies a set of shared dimensions of affect that can be used to distinguish individual emotions from one another. Within this Dimensional Theory, the Pleasure-arousal-dominance framework is most common (Russell & Mehrabian, 1974). This framework covers the entire set of human emotions in three independent dimensions: pleasure, arousal and valence (Russell & Mehrabian, 1974). This study will focus on the Dimensional Theory. Primarily in

(16)

advertising research the Dimensional Theory is used (Mauss & Robinson, 2009). Because according to Huang (2001) emotions in marketing are short-lived and hardly present in the pure form. It is concluded by Li, Walters, Packer & Scott (2017) that pleasure is the best predictor of effectiveness in tourism advertising according to the Dimensional Theory. When focusing on the dimension pleasure, several emotions represent pleasure. Two emotions which have been mentioned many times before as being influential are joy and surprise. It must be mentioned that with surprise a pleasant surprise is meant. Both emotions will be included in this dissertation. Underneath an overview is provided of both emotions:

(Desmet, 2012) Figure 2; Emotional words

So, it can be expected that advertisements which contain emotional (joy and surprise) loaded content, result in a higher click through rate (behavioural intentions) and conversion rate (purchase intent). The following hypotheses belong to this expectation.

H2a: Using emotion (joy) in search engine ads vs. no emotion in search engine ads increases

the likelihood of a higher click through rate.

H2b: Using emotion (joy) in search engine ads vs. no emotion in search engine ads

increases the likelihood of a higher conversion rate.

H2c: Using emotion (surprise) in search engine ads vs. no emotion in search engine ads

increases the likelihood of a higher click through rate.

H2d: Using emotion (surprise) in search engine ads vs. no emotion in search engine ads

EMOTION DEFINITION EMOTION WORDS

Joy Joy refers to the experience of being pleased about something. This emotion is also experienced when someone is reminded of a (past) joyful activity.

Joy, pleasure, happy, cheerful, good, delighted, bliss, overjoyed, wonderful, rejoice, smile, enjoyment

Surprise Surprise is experienced when an event was unexpected. The unexpected event is pleasurable/ desirable. This emotion comes with the feeling of amazement.

Surprise, startled, amazement, astonished, dazzled

(17)

2.4 INTERACTION BETWEEN GENDER TARGETING AND USE OF EMOTIONS IN SEARCH ENGINE ADVERTSING

As concluded by previous literature, the emotional perception between males and females seem to differ. A study by Klink (2008) found that there is a difference between how males and females response towards different content characteristics. Already for a long time, females are stereotyped as more emotional human beings than males (Allen & Haccoun, 1976; Grossman & Wood, 1993; Fisher & Dubé, 2005). Huddy (2002) states that females are more attached by emotions. To add on, McGuinness (1976) states that females are more sensitive than males in most emotional ways. Females are not only more emotional than males, females also are better at decoding emotions (Brody & Hall, 2008). In addition, females prefer interaction and emotions (Rodgers & Harris, 2003), while males have a strong preference for efficiency and convenience (Rodgers & Harris, 2003). According to Broverman et al. (1972) males express emotions less as they do not want to show weakness, vulnerability or dependency. However, this does not mean that males do not experience emotions (Fisher & Dubé, 2005). It is therefore expected that males will also be affected by the use of emotions in advertising, but this effect might be lower than in case of females (see H3).

H3a: Search engine ads with emotionally loaded content will lead to higher click through

rates among female users vs male users.

H3b: Search engine ads with emotionally loaded content will lead to higher conversion

(18)

2.5 CONCEPTUAL FRAMEWORK

Based on a conducted literature review, the subsequent conceptual framework in figure 3 has been proposed. The framework underneath, consists of two dependent variables that can be grouped together into one main variable, called Search engine ad effectiveness. Next to that, there are two main factors that we believe have an impact on search engine advertising effectiveness, which are Demographic targeting and Emotional Ad Content. Female and male, belong to the Gender, a characteristic of Demographic Targeting. Joy and Surprise have been used to identify Emotional Ad Content. The two main independent variables will both separate (H1, H2) and together (H3) be tested. Arrows are used to display relations.

(19)

3. RESEARCH METHODOLOGY

3.1 RESEARCH DESIGN

3.1.1 RESEARCH PHILOSOPHY

The research philosophy answers the question of what is considered as tolerable knowledge. This research is based on social science, as within this dissertation a phenomenon is studied that exists in reality. Bryman & Bell (2011) state that ontology is concerned with the nature of science, therefore it fits this research. What is even a better fit, is objectivism as all social actors will be approached independently. “Objectivism is an ontological position that

implies that social phenomena confront us as external facts that are beyond our reach or influence” (Bryman & Bell, 2011, p.21).

In this research, hypotheses are made based on theories (previous literature) and will be tested. This refers to a deductive research approach. A deductive approach can be defined as a method which contains theories and hypotheses (Bryman & Bell, 2011). Literature will be quantitatively tested during field research in order to either confirm or reject hypotheses.

3.1.2 RESEARCH STRATEGY

This research will investigate the relation between gender targeting in search engine ads with emotional ad content based on a quantitative approach. A field experiment in collaboration with a company will be conducted to be able to support or reject the hypotheses. Thus, it can be concluded that an experimental design fits this research strategy best. Two main groups are needed, the experimental group (receives the treatment) and the control group (receives no treatment) in order to do a before-and-after analysis (Bryman & Bell, 2011). Chandon (2003) confirms this by saying that in order to measure effectiveness of different marketing actions, one groups should be exposed to one version of the stimulus

(20)

while the other group to the other version of the stimulus. In case of this research, eight experimental groups will receive a treatment. Three groups will be the ultimate control group, receiving no treatment. The use of three control groups will be further explained in the next paragraph.

Table 1

Overview experimental and control groups

Group Type of Group Emotion Gender

G1 experimental group no male

G2 experimental group no female

G3 experimental group joy male

G4 experimental group joy female

G5 experimental group surprise male

G6 experimental group surprise female

G7 experimental group surprise no

G8 experimental group joy no

G9a control group – Weekendje weg no no

G9b control group – Weekend weg no no

(21)

3.1.3 EXPERIMENT DESIGN

The experiment will be designed as follows. First of all, it must be mentioned that this experiment will take place within the field. Meaning that the circumstances will not be manipulated. Recently, Google introduced Demographics for Search ADS (DFSA), this new tool makes it possible to target based on gender. (Google, 2016; Irvine M. , 2017) For this experiment, DFSA will be used to show specific ads only to males or females.

As the experiment is performed in collaboration with a company, that takes part of the leisure industry, their rules had to be respected. The only possibility for executing the experiment, with respect to the guidelines of the company, was to select three highly similar ad groups in Google AdWords. This means that individuals exposed to different experimental groups are not allocated randomly into groups. However, everything possible has been done to mimic randomization of the sample. The three used ad groups were similar in terms of exposed individuals in order to be able to compare each experimental ad with its original control ad.

On the next page, the ads that are have been used will be visible. All ad groups and group numbers are shown including the control ad (with no emotion) and the experimental ad. All ads have four sitelink-extensions, these are consistent over the whole experiment. In Dutch these are: Beoordelingen, Prijzen, Foto’s and Faciliteiten. In English: Reviews, Prices, Photos and Facilities. All ads also contain four highlight-extension: Beoordeeld met een 8.3, Eigen theater, Uitstekende restaurants and Heerlijk ontbijt. In English they mean: Rated with an 8.3, Own theatre, Excellent restaurants and Delicious breakfast.

(22)

v Weekend Weg (weekend trip)

This ad group is used for the experiment with the emotion Surprise.

v Weekendje Weg (weekend trip- diminutive)

This ad group is used for the experiment with the emotion Joy.

Figure 4; Original version (without Surprise). Used for G9b

Figure 5; Experimental version (with Surprise). Used for G5, G6, G7

Figure 6; Original version (without Joy). Used for G9a

Figure 7; Experimental version (with Joy). Used for G3, G4, G8

Weekend trip| take advantage of our specials

Free extra night, VIP upgrade or other extras. Book now & take advantage!

Surprisingly Weekend trip| Surprise yourself with our specials

Be amazed with our VIP upgrade or free extras.

Book now & take advantage!

Weekend trip| take advantage of our specials

Wonderful Weekend trip| Enjoy our specials

Have pleasure with a blissful VIP upgrade or free extras.

(23)

v Weekendweg (weekend trip- without space)

Figure 8; Original version (without gender wording). Used for G9c

Figure 9; Experimental version (with Male wording). Used for G1

Figure 10; Experimental version (with Female wording.) Used for G2

Fancy a Weekend trip?| take advantage of our specials

Sir, you do fancy a Weekend trip?| take advantage of our specials

Madam, do you fancy a Weekend trip?| take advantage of our specials

(24)

The original ad groups (weekendje weg, weekend weg & weekendweg) have been duplicated, in order not to affect the quality score of the original ad group. Because of this decision, one week (24-04-2018 till 30-04-2018) was needed in order to get the duplicated ad groups at the same level as the originals. It must be mentioned, that during the experiment the original ad groups have been paused in order to exclude keyword cannibalism between the original and duplicated ad groups. After that week (24-04-2018 till 30-04-2018), three more weeks were needed to complete the entire experiment. Hereunder an overview is provided of the operation of the experiment:

In the first week of the experiment (01-05-2018 till 07-05-2018) the ads of the groups G7 and G8 have been exposed online within Google’s search engine. In the second week (08-05-2018 till 14-(08-05-2018) groups number G1, G3 and G5 were live. This was done by using DFSA. This tool is used to determine that ads should be shown to male users for 100% and 0% to female users. In week three (15-05-2018 till 21-05-2018), this was the other way around, ads were a 100% shown to female users and 0% to male users for the groups G2, G4 and G6.Additionally, data according to the ultimate control groups (G9) was needed. This data is taken from the original ad group, data from seven days (24-04-2018 till 30-04-2018) prior to the experimental period has been used as the data for the control variable.

(25)

3.2 SAMPLE

The experiment will be in form of A/B testing. Eight treatment groups + three control groups will be created, and the experiment will be executed in Google AdWords.

The sample selected for this experiment should represent the whole population. It can be stated that the population is defined as all users of Google’s search engine, as long as the user of the search engine shows interest in the company’s offers. It is chosen not to investigate online users with an age below 18, as this group is not likely to book a trip offered by this company (tour operator). According to CBS (2017) from the Dutch society above with an age above 12 years (=14901937), on average 92.9% (=13.843.899) aged above 18 and uses internet on a daily base (CBS, Internet; toegang, gebruik en faciliteiten, 2017). On average, 95,66% of Dutch internet users use Google as a search engine (Borgers, 2017). It can therefore be calculated that 13.843.899 x 95,66% = 13.243.996 is the total population for this experiment. According to the sample size calculator: With a confidence level of 95%, a confidence interval of 5 and a population of 13.243.996 the number of users needed for one experiment is 384 (SurveySystem, 2018). In total, 384 x 9= 3.456 online users are required in order to be able to generalize the results of this experiment. The experiments that will be executed in Google AdWords, should have at least 384 impressions. This means that the ad has to be shown in Google’s engine a minimum of 384 times.

(26)

3.3 OPERATIONALIZATION OF TERMS Table 2

Variable Description Measurement Level

CONVERSION_RATE (dependent variable)

Indicates the ratio of the number of purchases that an online user makes from the advertisement as a percentage of the number of clicks on the advertisement Measured in percentages Ratio CONVERSION (dependent variable)

Indicates the number of purchases that an online user makes from the advertisement Measured in numbers Ratio CLICK_THROUGH_RATE (dependent variable)

Indicates the ratio between the number of clicks

of an ad and the number of times an advertisement has been shown

Measured in percentages Ratio CLICK (dependent variable)

Indicates the number of clicks Measured in numbers Ratio GENDER (independent Variable)

Indicates whether the Online user is a male or female Dummy variable (0 = Male, 1 = Female) Nominal EMO_JOY (moderator variable)

Indicates whether the search ad has emotive (JOY) content Dummy variable (0 = No, 1 = Yes) Nominal EMO_SURPRISE (moderator variable)

Indicates whether the search ad has emotive (SURPRISE) content Dummy variable (0 = No, 1 = Yes) Nominal EMO_NONE (independent + moderator variable)

Indicates whether the search ad has no emotive content

Dummy variable (0 = No, 1 = Yes)

(27)

3.4 MEASUREMENT

Often in business, search engine ad effectiveness is measured by using quantified data. The performance of search engine advertising can be measured with use of Google Analytics. Google Analytics reports all undertaken actions in Google AdWords. As determined based on previous literature, the measurement variables are click through rate and conversion rate in order to detect search engine ad effectiveness. Both variables can be found in Google Analytics. In addition, Google Analytics shows how ad campaigns performed. So, it is convenient to conclude which click through rate belongs to which Google AdWords Ad Group.

This experiment will be set up in a field setting. Google AdWords is going to be used to host this experiment. After all involved groups have been tested for one week the results will be viewable in Google Analytics. Google Analytics shows per ad group the results of clicks in numbers, click through rate in percentages, conversions in numbers and conversion rates in percentages. After all data is collected in Google Analytics, data can be exported to Microsoft Excel. This software will be used to statistically test the relation between variables and to identify if the differences between experimental and control groups are significant. In order to be able to reject or confirm the hypothesis, statistical test should be executed. Based on the quality of the data, it is determined to execute a t-test for differences between proportions. In coordination with the supervisor, it has been determined to execute the statistical test manually in Microsoft Excel. Several calculations had to be completed in order to compute the probability of the differences in proportion: Standard Error (𝑝̂), Variance of Proportion (S𝑝̂) and the Standard Deviation (Z). The used equations and their connections are displayed on the next page.

(28)

Afterwards, the Z-value is converted into a P-value by using a predetermined equation in Microsoft Excel: TDIST(x,degrees_of_freedom,tails). “x” refers to the numeric value at which the distribution is evaluated. After, the number of degrees of freedom are indicated. Finally, the number of distribution tails are specified. In this data analysis, a two-tailed distribution is chosen. The outcome of this equation results in a P-value. A higher Z-value (either: < -1.96 or > +1.96), decreases the P-value. A pre-chosen probability (p) has been established in order to discover significant (significance level = alpha) differences. Regularly, a level of .05 for probability is applied. For a p=.05, a confidence level of 95% can be guaranteed.

(29)

3.5 PRE-TEST

In order to be confident about the ability of users to discover the emotion within the treatment advertisements, a pre-test is executed. In this pre-test, four groups have been created. Respondents were collected based on convenient sampling (4 x 25 respondents). It has been chosen to test the recognition of the emotions. Recognition refers to the user’s capability to confirm previous exposure to the emotion. This is often measured in marketing by asking: Do you remember having seen….? Did you recognise …? (Keller, 1993). First of all, in the introduction of the pre-test

an outline of the situation was given. This outline contained the following text:

At the first page of the pre-test, the respondent

was exposed to one version of the search engine ad (this could be emotion joy, emotion surprise or no emotion). The respondent was able to look at the search ad, as long as he or she wanted. After, on the second page, the respondent was asked to answer to following question:

Did you recognise the emotion joy * in the search ad you just saw?

1) Yes, I did recognise “joy*” No, I did not recognise “joy*”

* In case of the emotion surprise. Joy will be preplaced by surprise from the example above.

At the third page of the pre-test, the respondent was asked to answer the next question: Would you click on the ad you just saw?

1) Yes 2) No

Finally, at the last page of the pre-test, the respondent had to fill in his or her gender, by answering:

Please identify your gender

1) Male 2) Female

Dear respondent,

Please imagine that you are searching for a city trip deal. The following page shows a possible ad which might pop up in Google’s search results.

Thank you in advance, Erna

(30)

The pre-test ran for one week (24-04-2018 till 30-04-2018). Students at the University of Amsterdam have been asked to fill in the pre-test. Below an overview is given about the results of the pre-test.

Table 3

Results pre-test

Recognised an emotion Not recognised an emotion JOY Experiment Male: 10

Female: 9 Overall: 19 (79.2%) Male: 1 Female: 4 Overall: 5 (20.8%) Control Male: 5 Female: 2 Overall: 7 (28.0%) Male: 11 Female: 7 Overall: 18 (72.0%)

SURPRISE Experiment Male: 9 Female: 11 Overall: 20 (74.1%) Male: 5 Female: 2 Overall: 13 (25.9%) Control Male: 1 Female: 3 Overall: 4 (16.0%) Male: 9 Female: 12 Overall: 5 (84.0%)

It can be concluded that overall, the largest part of the sample recognised the applied emotional wordings. There was no significant difference between males and females between the recognition of the emotions and between.

(31)

3.6 CREDIBILITY OF THE RESEARCH

Measurement of a study should be reliable and valid to be able to execute a valuable research. Reliability identifies the overall consistency of a measure (Bryman & Bell, 2011). As this study follows a field experiment design, which means that the data already contains real observed behaviour. This makes it unnecessary to test for reliability.

When inspecting validity, which secures if the indicator actually measured the concepts (Bryman & Bell, 2011), two types of validity may be checked. Internal validity and external validity. Field experiments are conducted in a natural environment of the participants, this causes high external validity and experimental realism. Which makes the results generalizable (Roe & Just, 2009).

However, at the same time, this makes it hard to control extraneous variables, these variables may influence the results of the experiment. Because of that experimental designs score low on internal validity. As it is simply not able to exclude all possible factors that may explain the results (Roe & Just, 2009). Although it was not able to control variables in this experiment, it must be mentioned that everything imaginable has been done in order to mimic randomness. Like selecting three highly similar ad groups in Google AdWords for the experiment.

(32)

4. RESULTS

In the fourth chapter, outcomes of the field experiment will be presented. In chapter two hypotheses have been developed. First descriptive statistics of the collected data will be discussed. After, results of the executed statistical tests will be given per hypothesis in order to determine whether the hypothesis have to be confirmed or rejected.

4.1 DESCRIPTIVE STATISTICS

In this sub-chapter the data will be described, and frequencies of the examined variables will be discussed. The variables used in this study are:

Gender (targeting): Male, Female, None Emotional Ad Content: Joy, Surprise, None

Search Engine Metrics: Click, Click Trough Rate, Conversion, Conversion Rate

In total, 11.182 users participated in this experiment. This is calculated based on the total amount of impressions of all groups. As that is the exact number of how often the advertisement has been shown to a user of Google’s search engine. Of all users, 3811 (34.08%) were male while the remaining 7371 (65.92%) were female. 3306 (29.57%) users belonged to the experimental groups and 7876 (70.43%) users received a control version. The fact that the number of users of the control groups is higher, is because of more control groups than experimental groups. Next to that, the use of emotions in advertising was applied in 5884 (52.62%) cases. Of which 3702 (62.92%) was with the emotion Joy, while 2182 (41.19%) was with the emotion Surprise. 5298 (47.38%) ads had no emotional wording.

(33)

The total number of clicks was 869, which results in a click through rate of 7.77% of all ad groups. The total numbers for conversions were lower. The total number of conversions was 21.9, this leads to a conversion rate of 1.52%. Beneath, a table is included which provides insights in the frequencies and percentages of impressions, clicks, click through rates, conversions and conversion rates among all included variables.

Table 4

Descriptive statistics: quantities and percentages

Impressions Click CTR Conversion Conv. Rate

Total 11182 869 7.77% 13.2 1.52% Male 3811 301 7.90% 3 1.00% Female Male % Female % Emotion No Emotion Emotion % No Emotion % Joy Surprise Joy % Surprise % 7371 34.08% 65.92% 5884 5298 52.62% 47.38% 4702 2182 62.92% 41.19% 568 34.64% 65.36% 498 371 57.31% 42.69% 316 182 63.45% 36.55% 7.71% 8.46% 7.00% 8.54% 8.34% 10.2 22.73% 77.27% 8.5 4.7 64.39% 35.61% 5 3.5 58.82% 41.18% 1.80% 1.71% 1.27% 1.58% 1.92%

(34)

4.2 HYPOTHESIS TESTING

This subchapter discusses the statistical test executed as well as all the results. All hypothesis will be tested with the corresponding analysis as elaborated in chapter 3.4.

4.2.1 HYPOTHESIS H1

Hypothesis H1 is concerned about the effect of applying gender targeting on either click through rate (H1a) and conversion rate (H1b). The results of the t-test for differences between proportions show that the assumed relationship between targeting and click through rate (H1a) is significant (p=.002). But the Z-value of the t-test shows a negative number, this implies that the effect is actually negative. Meaning that applying gender targeting does not lead to higher click through rates. Therefore, H1a is not supported.

H1a: Applying gender targeting vs. no gender targeting leads to a higher click through rate. Table 5

H1a: Sample size, Proportion (CTR) and t-test of difference between proportions

N CTR T-test

P-value 2-tailed Gender (G1+G2) 489 5.73%

Control (G9c) 1230 7.72%

Difference (Gender – Control) -3.302 .002

Hypothesis H1b looks at the same independent variable as H1a, but the dependent variable is different. The data showed that the ad in the control group (G9c) did not generate any conversion. This makes it obvious that there exists a significantly positive difference between applying gender targeting and not applying gender targeting on conversion rates. Therefore, Hypothesis H1b can be supported.

H1b: Applying gender targeting vs. no gender targeting leads to a higher conversion rate. Table 6

H1b: Sample size, Proportion (ConvRate) and t-test of difference between proportions N Conv. Rate T-test

P-value 2-tailed Gender (G1+G2) 489 3.57%

Control (G9c) 1230 0.00%

(35)

To analyse whether there is a difference between the application of the use of emotions in advertising (joy or surprise) on search engine results, a t-test for differences between proportions is executed. The outcome demonstrates that applying the emotion joy in search engine advertisements has a positive and significant effect on the growth of click through rates (Z=3.610, P=.001) than the control group. Therefore, hypothesis H2a can be accepted.

H2a: Using emotion (joy) in search engine ads vs. no emotion in search engine ads increases

the likelihood of a higher click through rate.

Table 7

H2a: Sample size, Proportion (CTR) and t-test of difference between proportions

N CTR T-test

P-value 2-tailed Joy (G8) 2234 8.28%

Control (G9a) 2658 7.37%

Difference (Joy – Control) 3.610 .001

For hypothesis H2b the same t-test is used to derive the results. Based on the results of the test it seems like there is a positive difference between applying emotions in advertising (joy) and not applying emotions in advertising (joy). Although the positive relationship, the effect is not big enough (P= .052) to call it statistically significant. Because of that, hypothesis H2b will be rejected.

H2b: Using emotion (joy) in search engine ads vs. no emotion in search engine ads

Table 8

H2b: Sample size, Proportion (Conv. Rate) and t-test of difference between proportions

N Conv. Rate T-test

P-value 2-tailed Joy (G8) 2234 1.62%

Control (G9a) 2658 1.28%

(36)

Both hypothesis H2c and H2d are developed in order to test the effect of applying the emotion surprise within the search ads in Google’s search engine. For hypothesis H2c a statistical significance is found (P= .015). And the Z-value of the t-test (for difference between proportions) Z=2.554, meaning that the difference between both groups is positive and thus H2c is supported and can be confirmed.

H2c: Using emotion (surprise) in search engine ads vs. no emotion in search engine ads

increases the likelihood of a higher click through rate.

Table 9

H2c: Sample size, Proportion (CTR) and t-test of difference between proportions

N CTR T-test

P-value 2-tailed Surprise (G7) 833 6.96%

Control (G9b) 921 5.65%

Difference (Surprise – Control) 2.554 .015

Hypothesis H2d presumes that the use of emotions in advertising with the emotion surprise will lead to higher conversion rates. The analysis of the executed t-test shows that indeed, when applying emotions in advertising (surprise) the ratio of conversion is higher than when there is no emotion in advertising (Z=2.671, P=.011).

H2d: Using emotion (surprise) in search engine ads vs. no emotion in search engine ads

Table 10

H2d: Sample size, Proportion (Conv. Rate) and t-test of difference between proportions

N Conv. Rate T-test

P-value 2-tailed Surprise (G7) 833 3.45%

Control (G9b) 921 2.31%

(37)

In order to test the hypothesis H3a t-test for difference between proportions was conducted. For H3, the sample proportions were: males that were exposed to emotionally loaded ad content (N=951) and females that were exposed to emotionally loaded ad content (N=1866).

H3a: Search engine ads with emotionally loaded content will lead to higher click through

rates among female users vs male users.

Table 11

H3a: Sample size, Proportion (CTR) and t-test of difference between proportions N CTR T-test P-value 2-tailed Female (G4+G6) 1866 9.27%

Male (G3+G5) 951 8.62%

Difference (Female – Male) 1.540 .122

Based on the results it can be concluded that males that were exposed to emotionally loaded ad content had no significant higher click through rate based on statistics than females that were exposed to emotionally loaded ad content Z=1.539, P=.122. This means that H3a has to be rejected.

H3b: Search engine ads with emotionally loaded content will lead to higher conversion

rates among female users vs users.

Table 12

H3b: Sample size, Proportion (Conv. Rate) and t-test of difference between proportions

N Conv. Rate T-test

P-value 2-tailed Female (G4+G6) 1866 1.16%

Male (G3+G5) 951 1.83%

Difference (Female – Male) 2.383 .023

Based on the results it can be concluded that males that were exposed to emotionally loaded ad content had statistically significantly lower conversion rate than females that were exposed to emotionally loaded ad content Z=2.383, P=.023. This means that H3b can be confirmed.

(38)

Below, an overview of all hypothesis is given including the results and the direction of the results.

Table 13

Number Hypotheses Direction Results

H1a Applying gender targeting vs. no gender targeting leads to a higher click through rate.

Negative Not supported

H1b Applying gender targeting vs. no gender targeting leads to a higher Conversion rate.

Positive Supported

H2a Using emotion (joy) in search engine ads vs. no emotion in search engine ads increases the likelihood of a higher click through rate.

Positive Supported

H2b Using emotion (joy) in search engine ads vs. no emotion in search engine ads increases the likelihood of a higher conversion rate.

Positive Supported

H2c Using emotion (surprise) in search engine ads vs. no emotion in search engine ads increases the likelihood of a higher click through rate.

Positive Supported

H2d Using emotion (surprise) in search engine ads vs. no emotion in search engine ads increases the likelihood of a higher conversion rate.

Positive Supported

H3a Search engine ads with emotionally loaded content will lead to higher click through rates among female users vs male users.

Positive Not supported

H3b Search engine ads with emotionally loaded content will lead to higher conversion rates among female users vs users.

(39)

4.3 FURTHER ANALYSIS

During the execution of the statistical tests for the predetermined hypothesis, several hypothesis-related analyses came about. These additional analyses have the same type of statistical test (t-test for difference between proportions), as the dataset and the level of the variables is still the same.

The original results for the first hypothesis, H1a, showed that there was a significant difference between the application of gender targeting and no application of gender targeting on click through rates. But this result had a negative Z-value, which presumes the opposite. Namely, that applying no gender targeting causes higher click through rates. In this further analysis, the difference between male and female targeting on click trough rate is examined. The extra analysis for hypotheses H1a present that males are a little affected by targeting (Z=.087, P=.397), while females are totally not affected by gender targeting

Z=-4.203, P=.000), (outcome tables can be found in appendix A).

For H1b, the data showed a significant result on the t-test for differences between proportions, and a positive value for Z. The further analysis on H1b is again concerned about the difference between male/female targeting on conversion rates. The extra analysis for hypotheses H1b implies that females are affected by gender targeting in terms of conversion rates (Z=9.117, P=.000), (outcome tables can be found in appendix A).

For hypotheses H2a and H2b, additional analyses have been executed in order to discover if there exists a difference between the application of emotions in advertising (joy) and no emotion in advertising. For both H2a and H2b further tests have been done to find if there are also differences between joy and surprise towards search engine advertising outcomes. Analysis show that surprise does not causes higher click through rates than joy (Z=-3.859,

(40)

P=.000). While surprise does score higher on conversion rates than joy (Z=5.459, P=.000),

(see appendix A).

Hypotheses H3, were centred around the question if search engine ads will lead to higher click through rates or conversion rates when emotion in advertising is applied. The further analysis goes one step beyond and analyses if there is a difference between how males and female perceive different types of emotion (joy, surprise). After the execution of a t-test for difference between proportions, results indicate that there is no significant result between females and males with joy on click through rates (Z=1.927, P=.062). For surprise there was an insignificant difference found for click through rates. The z-value even indicates a negative number, meaning that males are more effected by surprise than females (Z=-.133,

P=.395). For conversion rates, a negative significant difference is found between males and

females with the emotion joy (Z=-2.673, P=.011), meaning that males are more likely to convert when have seen an ad with the emotion joy. For the emotion surprise, no significant difference between males and female is detected on conversion rates (Z=-.129, P=.396). (see appendix A).

(41)

5. DISCUSSION AND CONCLUSIONS

5.1 DISCUSSION

With the rising popularity of search engine advertising and the still unexamined effects of gender targeting and the use of emotions in advertising, this thesis aimed to answer the main research questions:

v What is the impact of a gender targeting strategy on the effectiveness of search engine advertising in terms on click through rates and conversion rates?

v How can search engine advertising be designed by using emotions to increase effectiveness?

v Does the interaction between gender differences and the use of emotions produce a differential impact on search engine advertising effectiveness?

In this chapter, answers to the main research questions will be given by discussing the main findings, theoretical implications, and practical implications. Next to that, limitations & future research will be discussed, and a conclusion will be given.

5.1.1 MAIN FINDINGS

This study was designed in order to analyse if the existing theory about gender targeting and the use of emotions in advertising influences search engine ad effectiveness. Earlier literature stresses the fact that search engine advertising is growing in business importance, as it helps to extend the business and win market rivalry (Indartoyo, Rahayu, Budiwan, Bismo, & Sadeghifam, 2016). Search engine advertising allows marketers to highly target specific users anytime, anywhere (Google AdWords, sd), which can be seen as an opportunity as targeting ads might lead to higher ad effectiveness (Bergemann & Bonatti, 2011). What becomes clear in this study is the opposite, as applying gender targeting does not lead to higher click through rates, especially among females. This could mean that females find it too intrusive when ads are highly targeted based on their gender. On the other side, results imply that females are affected by gender targeting in terms of conversion

(42)

rates, which contradicts the prior statement. It could mean that most of the females do not feel attracted by ads that were targeted based on gender (low click through rate). But once a female is appealed to the ad, she feels connected to the ad and converts more often than ads without targeting. This causes a higher return on ad spend, as marketers only have to pay per click. And in this case the click trough rate is low, while the conversion rate is growing because of gender targeting. For males, other outcomes have been detected. Despite the insignificant results, males have a positive relation with gender targeting when it comes down to click through rates. But when investigating the conversion rates, males did not convert at all. This could also be a feature of the tourism industry, as in general data showed that more females searched, clicked and converted on the search ads.

To answer the first research question: What is the impact of a gender targeting strategy on

the effectiveness of search engine advertising in terms on click through rates and conversion rates? It can be stated that, in general, applying gender targeting does not affect click trough

rates in a positive way. In addition, the contribution to conversion rates is not significant. This contradicts previous studies that say that targeting ads might lead to higher effectiveness as it improves the quality of the connection between the consumer and the advertising message (Bergemann & Bonatti, 2011).

The theory of the Elaboration Likelihood Model specifies that information processed via the central route, has a positive impact on the attention and purchase intent as it is perceived as interesting (Ho & Bodoff, 2014; Lancée, 2014)). Therefore, it is important to know which search ad characteristics get attention. In this dissertation, the use of emotions in advertising is taken into consideration as being an ad characteristic. Emotions that have been studied are joy and surprise because these emotions belong to pleasure. According to Li, Walters, Packer, & Scott (2017) within the dimensional theory, pleasure is the best predictor of

(43)

effectiveness in toursism advertising. The outcomes of the filed experiment find support for this, as for the ads with the emotion joy, a significant click tourgh rate and almost significant conversion rate (P= .52) is found compared to the control group (no emotion).

When investigating the results of the emotion surprise, both click through rates and conversion rate were significanly higher than the for ads with no emotional ad content. The further analysis on H2 show the difference between performance of the two emotions. Analysis show that the application of the emotion joy causes higher click through rates than surprise. While surprise score higher on conversion rates than joy.

The answer to the second main research question: How can search engine advertising be

designed by using emotions to increase effectiveness? is: the answer to the second main

research question is depending on the two sub-questions (2a and 2b). First of all, to answer research question 2a*, results show that the use of emotions in search engine advertisements is able to increase both click through rates and conversion rates. Secondly, research question 2b** can also be answered. This question is concerned about the effect of the different emotions. This study investigated joy and surprise. There are indeed different results found between the two emotions. So, the answer to research question 2 is: in order to generate more clicks, it is best to design the search ad with including the emotion joy. The emotion surprise can used when conversions are the objective of the campaign. In general, both emotions have a positive impact on click through rates and conversion rates. This study is confirming the existing literature, which says that the use of emotions in advertising increases behavioural intentions (click through rate) and purchase intentions (conversion rates).

* is the use of emotions able to increase click through rates and conversion rates? ** if yes, are there differences taking into consideration different emotions? Which specific emotion is more effective?

(44)

The answer to the third research question: Does the interaction between gender differences

and the use of emotions produce a differential impact on search engine advertising effectiveness?The interaction between gender and the use of emotions show that for click

through rates, both males and females have a positive outcome. So, there is no differential impact found between the two genders on click through rates when emotions in search engine advertisements are applied. For conversion rates, a slightly different result has been found. Indicating that females are significantly more affected by the use of emotions in search engine ads than males in terms on conversion rate. The last conclusion, confirms the literature

5.1.2 CONTRIBUTION TO LITERATURE

This research presents a model of the effectiveness of search engine advertising in terms of gender targeting and the use of emotions in advertising. In addition, this study contributes to existing literature in multiple ways.

First of all, over the past years, different researchers studied this growing type of online advertising, named: search engine advertising. All existing studies investigated topics that were related to ad content. As there are not many different visual options with search engine ads. Studies before examined (e.g.) effectiveness of evidence types in search ads (Haans, Raassens, & Hout, 2013). Some researchers studied the impact of emotions in online advertising (Lohtia, Donthu, & Hershberger, 2003; Goldfarb & Tucker, 2011) or the role of gender targeting in search advertising (Banerjee & Dholakia, 2012; Kemp, Bui & Chapa, 2012; Teixeira & Wedel, 2012; Jansen, Moore & Carman, 2012). However, to the best of knowledge, no researcher before investigated the use of emotions in advertising and its interacting effect with gender targeting in search engine effectiveness. This study therefore contributes to the further identification of search engine advertising effectiveness.

(45)

Secondly, the research design was innovative for this type of research question. Up to now, it was only possible to execute a lab designed experiment (based on questionnaires). Basically, until now, it was not possible to measure the effect of gender targeting and the use of emotions in advertising in a field experiment with real-time Google Analytics data, because Google had no feature available for this. Nearing the end of 2016, Google introduced Demographics for Search Ads, this new tool made it possible to specify target groups and measure their online behaviour. In this study, this new type of data collection is applied, which separates this study from many others that tried to study similar topics.

Finally, this study opens up a topic of discussion. As it opposes some statements that strongly have been stressed in prior literature about traditional advertising. Namely statements about how important and influential targeting is on marketing effectiveness. This study found a poor evidence for that statement. Also, one other study by Jansen, Moore & Carman (2012) found that neutral phrases performed better than gender specific phrases in search engine marketing. This could mean that there is a significant difference in how traditional offline marketing and online marketing should be approached. Maybe this is especially the case for search engine marketing. Search engine advertising is very complex as it is limited to a definite amount of characters. Besides that, search engine ads only contain textual content and no other visuals, which makes it harder to meet all users’ needs and wants. It is worth investigating if other online marketing channels are more positively influenced by targeting. But this will be further elaborated in chapter 5.1.4.

Search engine advertising : the impact of emotions and gender differences on search engine advertising effectiveness : a field experiment