Course: SKOM12

Term: Spring 2022

Supervisor Nils Homberg
Examiner

Factors Influence Purchase Intention in Marketing Communication

An Experimental Investigation of Human Images in digital display advertising
XINWEI ZHANG

Lund University
Department of strategic communication
Master’s thesis

I
Abstract

Factors Influence Purchase Intention in Marketing Communication

An Experimental Investigation of Human Images in digital display advertising

With the number of digital display advertisements increasing, consumers have become
accustomed to ignoring or skipping internet ads. Previous research has investigated the
use of different types of human images (like celebrities and social media influencers)
in advertising to improve advertising effectiveness. However, existing studies neglect
the role of normal people and examine the presence of human images in advertising in
general. This study aims to contribute new knowledge about marketing
communication. Using the AIDA model and the theory of planned behaviour, the study
proposes a new model to explore factors that could influence consumers’ purchase
intention. The analysis confirms that there is a significant interaction effect between
‘attention’, ‘interest’, ‘attitude’, and ‘social norm’ with purchase intention.
Additionally, an AB testing experiment approach is applied to test the effect of human
images on consumers' purchase intention. The experiment suggests that human images
appearing in advertising will positively influence the above four factors as well as
consumers’ purchase intention. The result from the experiment also provides evidence
that for consumers who can identify the social media influencer shown in the ad, ads
that include social media influencer images could lead consumers to generate higher
purchase intent than ads that include normal human images.

Keyword: Marketing Communication, Digital display advertising, Human images,

Social media influencers, AIDA model, TPB model, Purchase intention

Word count: 18906

II
Table of contents
1. Introduction ........................................................................................................................... 1
1.1 Background ........................................................................................................................... 1
1.2 Aim and Research question ................................................................................................... 2
1.3 Relevance .............................................................................................................................. 3
1.4 Disposition ............................................................................................................................ 4
2. Literature review ................................................................................................................... 5
2.1 Digital display advertising .................................................................................................... 5
2.2 Human images in digital display advertising ........................................................................ 7
2.3 Influencer Marketing ............................................................................................................. 9
2.4 Influence of visual and verbal factors in digital display advertising................................... 10
2.5 Synthesis.............................................................................................................................. 11
3. Theoretical framework........................................................................................................ 12
3.1 Purchase funnel ................................................................................................................... 12
3.2 Theory of Planned Behaviour ............................................................................................. 14
3.3 Hypotheses development..................................................................................................... 16
3.31 Attitude........................................................................................................................... 16
3.32 Human image................................................................................................................. 17
3.33 Social norm .................................................................................................................... 17
3.34 Knowledge ..................................................................................................................... 18
3.4 Research model and Hypothesis ......................................................................................... 19
4. Methodology ......................................................................................................................... 21
4.1 Post-positivism paradigm and research design ................................................................... 21
4.2 Sample selection and data collection .................................................................................. 22
4.3 Survey design ...................................................................................................................... 24
4.3.1 Why Samsung................................................................................................................ 25
4.3.2 Experimental design ..................................................................................................... 26
4.3.3 Measurements ............................................................................................................... 28
4.4 Ethical considerations ......................................................................................................... 29
5. Result and analysis .............................................................................................................. 31

III
 5.1 Descriptive analysis............................................................................................................. 31
5.2 Control variables and AIDA factors testing ........................................................................ 36
5.3 Hypotheses testing............................................................................................................... 39
5.4 Experimental stimuli testing................................................................................................ 40
6. Discussion and conclusion ................................................................................................... 44
6.1 Interpretation ....................................................................................................................... 44
6.11 Hypotheses discussion ................................................................................................... 45
6.12 Experiment result discussion ......................................................................................... 47
6.2 Conclusion ........................................................................................................................... 49
6.3 Implications ......................................................................................................................... 50
6.4 Limitations .......................................................................................................................... 51
6.5 Suggestions for future research ........................................................................................... 52
7. References............................................................................................................................. 54
8. Appendix............................................................................................................................... 61
Appendix 1- Ad images............................................................................................................. 61
Appendix 2 - Questionnaire ...................................................................................................... 62
Appendix 3- SPSS output.......................................................................................................... 71

IV
1. Introduction
1.1 Background

Every day, through social media, online platforms, or websites, consumers can gather and share
billions of information, post and customize their preferences, and communicate with other
consumers, product producers, or service providers (Pavlou & Stewart, 2000; Berger, 2014). As
these social media, online platforms, and websites are used by many consumers, advertisers will
choose to place their ads on these platforms. Digital advertising shows strong growth in all regions
and the global digital advertising revenue reach 465,5 billion US$ in 2021 (Statista, 2021a). As a
result, the number of online advertisements continues to grow each year, combined with the huge
amount of information that consumers will receive every day, consumers have become accustomed
to ignoring or skipping internet ads (Resnick & Albert, 2014). Therefore, consumers will make a
quick decision about whether to stay on the website within a short period (Bucher & Schumacher,
2006; Lindgaard et al., 2006). In this context, the challenge for advertisers is to attract the attention
of consumers. Visual elements could influence consumers' purchase decisions and an aesthetically
pleasing design could increase consumers’ interest and trust (Townsend & Shu, 2010; Lindgaard,
Dudek, Sen, Sumegi, & Noonan, 2011). Thus, to get consumers’ attention, more and more ads are
following the trend of the times and becoming more and more aesthetical (Huang, 2018).

Moreover, collaborating with traditional celebrities as well as social media influencers (SMIs) to
capture the attention of consumers is also one of the ways advertising is often used today (Schouten,
Janssen & Verspaget, 2020). SMIs refer to people who have become famous for their presence on
social media, as opposed to traditional celebrities famous for sports, movies, music, and TV shows,
before they participate in advertising activities (Khamis, Ang & Welling, 2016). As opinion
leaders, SMIs can have a purposeful impact on members of a particular community who gather
around common interests. Moreover, there are many influencers on social media that are focused
on digital product reviews and collaboration with SMIs is more cost-effective and easier to
generate content than traditional media ads or celebrity endorsements (Lou, Tan & Chen, 2019).
Thus, the strategy of collaborating with SMIs is recognised by most companies (Lou & Chen,
2019).

1
The smartphone is an important communication tool for today's generation and almost no one can
live without it. According to the survey, the average frequency of consumers replacing their mobile
phones is 33.6 months in 2020 (Statista, 2020). So, in order to attract new consumers, mobile
phone manufacturers are partnering with many traditional celebrities and SMIs to promote their
products. Previous research has looked at differences in how consumers perceive different types
of people endorsing the same product and the impact of different types of endorsers on consumer
perceptions. Existing research related to SMIs focuses on the role of SMIs and how to work with
SMIs in the digital marketing communication field, as well as compares SMIs with traditional
celebrities. Previous studies neglected to examine the effect of the presence of human images in
general, and the effect of the presence of normal people in advertising has been ignored.

Therefore, this study wants to explore the effectiveness of two similar and comparable sets of ad
content - ads with human faces vs ads without human faces. And explore the impact of the presence
of normal people images in advertising as well as the difference in the impact of images of SMIs
and images of normal people on consumers' purchase intentions.

1.2 Aim and Research question

Following the background, the main aim purpose of this study is to expand the knowledge of
marketing communication research. This study suggests two groups of advertising that are similar
and comparative. More specifically, this study wants to compare the influence of ads with human
images and ads without human images on consumers' purchase decision process. As well as the
influence of ads with normal people images vs ads with SMIs images. The author will draw upon
existing studies on digital display advertising and consumer purchase intention to gain a deeper
understanding of the effect of the human image. As a result, provide insights into how companies
strategically use SMIs or normal people images in ads to increase market acceptance of their
products. In order to achieve this purpose, the author seeks to answer the following research
questions:
RQ: How to use human images in digital display advertising to increase the effectiveness of
advertising and the purchase intention of consumers?

2
By answering the research question, this thesis contributes to previous studies in three aspects.
First, with the help of an experimental method, it observes the impact of different types of human
images on consumer purchase intention towards a product. Second, it enhances and extends the
application of purchase funnel theory and theory of planned behaviour (TPB) in the context of
digital display advertising. Finally, this thesis provides further insights and builds a deeper
understanding for academics and practitioners of how human images in advertisements are
perceived by consumers. As a result, this thesis will enrich the research field of marketing
communication.

1.3 Relevance

Digital display ads are a kind of information that can be seen everywhere. By placing
advertisements, companies hope to attract people's attention and in turn get potential consumers to
make a purchase decision. Display advertising can have a huge impact on a brand's products and
brand image (Bart, Stephen & Sarvary, 2014). Moreover, good use of display adverts can help a
company achieve its goals or even upgrade its brand (Scott, 2015). How to improve the
effectiveness of ads has also been a popular focus in marketing communication in recent years.

Using experimental methods to explore the differences in the effectiveness of different elements
in created scenarios as well as the impact of different elements on consumer perceptions, casual
relationships can be explored which is something that observational studies cannot do. By
providing empirical evidence, marketers, organisations, academics, and professionals will be
helped to recognise the impact of the human image that is included in advertising on the
effectiveness of advertising and consumer purchase intention.

Furthermore, the analysis will expand the field of influencer marketing as well as digital display
advertising by comparing advertising that contains human images with ads that do not include
human images; comparing ads that include SMI images with ads that include normal people
images. By finding out if the human image factor has an impact on purchase intentions and to what
extent human image will affect the purchase intention. The results of this study can help companies
not only in the smartphone industry but also in all industries to improve the effectiveness of their
display advertising which is a contribution to the existing marketing communication knowledge.

3
Hallahan et al. (2007) define strategic communication as a way that helps organizations reach their
objectives and mission. Marketing communication is one of the research areas under strategic
communication. From a strategic communications perspective, this paper will enhance knowledge
in marketing communication which also enhances knowledge in strategic communication.
Moreover, the result of this study can provide insight into the strategic use of SMIs or human
images in ads. This could help the company to improve the effectiveness of its ads and achieve its
goals. Therefore, a contribution to this knowledge is a contribution to strategic communication.

1.4 Disposition

The structure of this study is as follows. In the next chapter, a literature review related to the topic
will be presented to have a better understanding of digital advertising, and map out the contribution
of this study to the existing literature. In the “theoretical framework and hypotheses development”
chapter, the AIDA model and TPB model will be presented. Following that is the “methodology”
chapter, the methodology, experiment design, sample selection, and data collection will be
described. Afterwards is the “results and analysis” chapter, all results from the experiment will be
shown and discussed. The "discussion and conclusion" chapter follows, which will compare,
discuss, and contrast the statistical findings further with the research question and previous studies.
This chapter will also include conclusions, implications, limitations, and suggestions for future
research.

4
2. Literature review
This chapter seeks to offer a comprehensive review of the core research conducted in the field of
digital marketing communication to obtain a better grasp of the field and to be able to map out the
research gap and how current research contributes to the field. First, the main concepts, definitions,
and previous focus of digital display advertising are provided. Following this is a comprehensive
review of previous studies about human images in digital display advertising. Afterwards is the
influencer marketing section, demonstrating the background, the effectiveness of influencer
marketing, and the focuses of previous research. After this section, the impact of visual and verbal
elements in digital display ads on consumer attention is reviewed. Lastly, a synthesis is provided
with the problem which has not yet been sufficiently analysed, and the contribution are defined.

2.1 Digital display advertising

Digital (Online) display advertising is defined by Goldstein et al. (2014) as “advertisers pay
publishers (websites) to run display ads that users (website visitors) see alongside other content”
(p. 742), and the graphic elements in the advertisement can differ in size, shape, animation,
duration, etc. Moreover, digital display advertising is mostly displayed in online media which are
not searching engines, and the form of digital display ads includes banner ads, text ads, media-rich
ads, and video ads (Goldfarb, 2014). Among all the forms of digital display ads, banner ads are
the most widely used form (Sundar & Marathe, 2010; Lambrecht & Tucker, 2013). In the scope
of the present research, the author adheres to the definition of Digital display advertising defined
by Goldstein et al. (2014).
Figure 1. Different types of Digital display ads

5
Since the early days of online advertising, marketers and researchers have sought to improve the
effectiveness of advertising by targeting people with relevant browsing and search histories
(Pavlou & Stewart, 2000; Sherman & Deighton, 2001). Digital display advertising provides a
bunch of factors that can be used to target advertising (Goldfarb, 2014). Through collecting past
search, view, and purchase history of consumers which enables advertisers to better understand
consumer preferences, personalised product-based banners will be released to the target consumer
groups (Sundar & Marathe, 2010; Lambrecht & Tucker, 2013). Personalised ads also include a
display of the specific products previously viewed by the consumer (Lambrecht & Tucker, 2013).
But previous studies show that personalized ads don't always work well. Goldfarb and Tucker
(2011) demonstrated that personalised ads can have a negative impact on display ad performance
when combined with highly visible material. Similarly, Lambrecht and Tucker (2013) found that
retargeted personalised ads (i.e., display a particular product that the customer has previously
viewed) are typically underperforming compared with general ads. However, this relationship will
reverse when consumers visit specific online review websites (Lambrecht & Tucker, 2013). It also
shows that consumers’ reactions and attitudes to ads can change depending on the platform on
which ads are displayed. Therefore, to guarantee the accuracy of the results, in this study, the
questionnaire will display the advertisements without showing any platform features to avoid
consumers' preconceptions influencing the final results.

Previous research on digital display advertising has explored a number of different aspects, such
as the content of digital display advertising as well as factors that affect the recognition process of
ads. In the early days of research about the content of digital display advertising, Drèze and
Hussherr (2003) find that consumers avoid looking at the ads during their online activities which
refer to the “banner blindness”. Despite the fact that consumers purposefully avoid looking at
display ads, this study concludes that display ads still have a positive impact on brand recognition
and advertising recall (Drèze & Hussherr, 2003). Burke et al. (2005) compare different online
banner ad types (animated ads, static ads, cyan with flashing text ads, and blank ads) and the result
demonstrated that both animated and static ads reduce the visual search speed of consumers and
the fact that animated ads are harder to remember compared with static ads. In the animated ads
range, Yoo and Kim (2005) compare the effects of different speed animations (no animation,
moderate animation, and fast animation) and find moderate animation ads can have a positive

6
effect on ad recognition as well as brand attitude. There is no significant difference between
animated and static ads in terms of getting consumers' attention (Burke, Hornof, Nilsen & Gorman,
2005). In an investigation into display ads and purchase behaviour, Manchanda et al. (2006) point
out that digital display advertising will not only have an impact on consumer recognition and brand
attitude but also could have a positive impact on the purchase frequency of existing consumers
(Manchanda, Dubé, Goh & Chintagunta, 2006). In addition, with the popularity of smartphones,
studies within the field of digital display advertising on mobile phones have been explored by
many researchers. Andrews et al. (2015) mention that mobile provides a range of target variables
(i.e., time, weather, location, crowdedness) that can be used for purchase impressions which means
personalized ads can be more precise. Regarding the impact of mobile digital display advertising,
Bart et al. (2014) point out that a significant proportion of consumers who were exposed to mobile
digital display advertising reported being more likely to subsequently purchase the advertised
brand, and mobile digital display advertising is more effective at changing brand awareness,
attitudes and intent than comparable traditional online display ads. Because of the high efficiency
and impact of mobile digital display advertising, it has been interesting for researchers to explore
factors that influence the effect of mobile digital display advertising.

2.2 Human images in digital display advertising

Endorsements are commonly used by marketers to promote products, brands, and services, as well
as improve the credibility and effectiveness of the advertising (Schouten, Janssen & Verspaget,
2020). An endorser is usually a celebrity with a positive image, such as actors, models, athletes
etc. Trustworthy, attractive, and popular celebrities have a positive impact on brand ratings
(Bergkvist & Zhou, 2016). By adding endorser images to the ad, the positive image and
characteristics of the endorser are then transferred to the product and even the brand (Schouten et
al., 2020). Many previous studies have confirmed that celebrity endorsements dramatically boost
the effectiveness of advertising (Amos, Holmes & Strutton, 2008; Bergkvist & Zhou, 2016;
Schouten et al., 2020). Likewise, Friedman et al. (1976) suggest that advertisements with endorsers
increase consumer credibility compared to advertisements that do not use endorsers.

7
In online display advertising, the human face is valuable in understanding the characteristics,
personality, intentions, and emotions of others, which is considered to be a uniquely effective
stimulus for attracting visual attention (Vuilleumier & Schwartz, 2001). Moreover, behavioural
data suggests that when the face is present in a visual scene along with other stimuli, the face can
be more likely to attract attention than other objects (Vuilleumier, 2000; Kuisma, 2015). Thus,
human faces are likely to be a powerful means of attracting and retaining viewers' attention in an
online advertising environment (Kuisma, 2015; Schouten et al., 2020). Xiao and Ding (2014) point
out that faces can not only attract viewers' attention but also influence their reactions to
advertisements, and people show fairly constant preferences regarding faces. Similarly,
Sajjacholapunt and Ball (2014) conducted that compared to the advertisements that did not contain
faces, the advertisements with human faces will make viewers stay longer, and viewers will
generate a higher level of attention which leads to better memory for advertising content. In the
ad, it is thus clear that the face is a key factor influencing the impact of human images.

Many aspects can influence the effectiveness of human images in advertising. First, the
effectiveness of the appearance of human images in advertising is influenced by the product
category. Usually, presenting an image of an expert in the relevant field in an advertisement can
enhance the effectiveness of the advertisement when the product is technically complex, or when
the consumer needs to be confident in its functionality (Munnukka, Uusitalo & Toivonen, 2016).
Second, when a human image appears in multiple advertisements about different products, thus
inflating the associations with the specific product it leads to a reduction in the credibility not only
of the individual but also of the advertisement (Munnukka et al., 2016). Third, Schouten et al.
(2020) point out that the direction of the eye gaze of the faces in an advertisement can also affect
the effectiveness of the advertisement. Comparing to the averted gaze condition, the mutual gaze
condition could lead to better memory of ad content for the viewer (Schouten et al., 2020). Lastly,
when the image of the individual does not match the image of the brands, it can lead to negative
comments about the individual and even the brand, which in turn can affect the effectiveness of
the advertisement (Munnukka et al., 2016; Schouten et al., 2020). Therefore, to ensure the
effectiveness of your advertising, marketers need to be careful in selecting the individuals who
appear in their advertisements as well as the present form (Munnukka et al., 2016).

8
2.3 Influencer Marketing

Influencer marketing is a marketing strategy that uses the influence of social media influencers
(SMIs) to increase consumers’ brand awareness and/or their purchasing decisions (Scott, 2015).
Usually, SMIs promote the brand's products or services on their account, either directly shown on
the advertisement or more subtly, such as by placing a product on a table which means they will
promote the content to both the followers of SMIs and the target consumer groups of the brand
(Lou & Yuan, 2019). Luo and Yuan (2019) define SMIs as “a content generator; one who has a
status of expertise in a specific area, who has cultivated a sizable number of captive followers—
those are of marketing value to brands—by regularly producing valuable content via social media”
(p. 59). It is clear that SMIs refer to people who have become famous and influential for their
presence on social media, as opposed to traditional celebrities famous for sports, movies, music,
and TV shows, before they participate in advertising activities (Khamis, Ang & Welling, 2016).
SMIs have previously established themselves by focusing on a certain field which means that
consumers are more inclined to accept or believe an influencer's opinions when they work with
businesses that fit their unique area of expertise (Khamis, Ang & Welling, 2016). And a recent
study also points out that compare with collaboration with traditional media ads or celebrity
endorsements, it is more cost-effective and easier to generate content to collaborate with SMIs
(Lou et al., 2019). Reports in recent years show that people trust SMIs almost as much as they trust
their friends; nearly 40% of Twitter users have made a purchase at some point because of an
influencer's recommendation; and 70% of teenage YouTube users report that they think
YouTubers (influencers on YouTube) are "more like one of us and close to us" and they have a
closer relationship with SMIs than traditional celebrities (Lou et al., 2019). Based on what was
mentioned above, it is very feasible to use social media influencers in marketing.

Previous research looked at the aspects that related to the efficacy of influencer marketing in a
variety of situations. De Veirman, Cauberghe, and Hudders (2017) demonstrated that when
developing an influencer marketing strategy, SMIs' follower numbers, followers/followees ratio,
and product kind (i.e., divergence level) should all be taken into account. Lu, Chuan, and Chang
(2014) mention that except the type of sponsorship and product type, brand awareness of
consumers could affect consumers' attitudes towards influencer marketing. In addition, Djafarova
and Rushworth (2017) have examined the impact of the credibility of SMIs’ profiles in influencing

9
the purchase decisions of young female consumers, and the result shows that the credibility and
trust of SMIs could impact the purchase intention of young female consumers. Therefore, lots of
aspects or factors could impact the efficacy of influencer marketing.

2.4 Influence of visual and verbal factors in digital display advertising

According to the dual-coding theory, the cognition of humans consists of visual and verbal systems
(Paivio, 1986). When an external stimulus is received, the information is converted into a
representation, but visual memory and verbal memory are stored separately (Paivio, 1986). There
is a reference relationship between visual and verbal memory, and the combination of the two can
enhance cognition and improve memory (Paivio, 1991). When visual and textual elements are
present at the same time, the consumer gets more information than with a single text or visual
element (Paivio, 1991). In the cognitive system that extends from dual-coding theory, everyone
acquires and uses information in different ways and will perceive it differently, for example, some
people will use more verbal cues to get information and others prefer to use more visual cues to
get information (Richardson & Sheikh, 2020). It has been suggested that visual factors and verbal
factors could affect each other and when an ad has concrete images, the imagery instructions in
the caption can more effectively stimulate a consumer's mental imagery association (Walters,
Sparks & Herington, 2007). Likewise, Mayer (2005) concludes that ads containing both visual
elements and verbal elements can strengthen consumers’ relevant memories. Numerous eye-
tracking experiments on print ads have also proven that photo/image type ads are more eye-
catching than text-only ads (Goodrich, 2011). Thus, most display advertising today includes a
visual element - the image, and a verbal element - the caption.

The photo/image element is often considered to be the main element that attracts the consumer's
attention (Dreze & Zufryden, 1997; Rossiter & Percy, 1997; Blackwell, Miniard &Engel, 2006;
Wells, Burnett & Moriarty, 2008). Rossiter and Percy (1997) argue that “the picture is the most
important structural element in magazine advertising, for both consumer and business audiences,”
and they suggest that “the straightforward rule for magazine ads, therefore is “the bigger the picture,
the better” (p. 295). Similarly, Wells, Burnett, and Moriarty (2008) conclude that the larger the

10
illustration in the advertisement, the greater the attention it will receive. Moreover, Shaouf, Lü and
Li (2016) found that consumers are only willing to buy if they like the design of a website
advertisement. The visual elements of an advertisement can therefore influence consumers'
attention and willingness to buy and even their attitude towards a brand (Sundar & Noseworthy,
2014; Shaouf, Lü & Li, 2016). And verbal elements are also key to attracting consumers' attention;
good taglines and headlines can help to communicate the content of an ad (Pieters & Wedel, 2004).
According to a prior study on the size, format, text, and image in online adverts, online display
banner ads with larger images and limited text content are more aesthetically attractive (Huang,
2018). In this study, all advertisements will take the same ad text content to ensure the accuracy
of the experimental results.

2.5 Synthesis

As shown, a growing body of literature has investigated how to effectively use SMIs in
advertisements to improve the effectiveness of ads as well as how to design and select verbal and
visual factors to attract consumers’ attention. But as the literature demonstrates, the extent of the
earlier research focuses on the effectiveness of a single type of human image used in advertising
and links the type of character to the intention. As well as focused on the differences in the
effectiveness and impact on consumers of the use of SMIs and traditional celebrities in advertising.
Existing research on online display advertising lacks studies on the effect of human images in
advertising on consumer purchase intentions in general and compares the effect of SMIs with
normal people appearing in advertising. Therefore, based on the previous literature and research
suggestions from scholars, this research intends to address the literature by exploring the human
image in an advertisement in general, and further investigate how human image influences
consumer's purchase intention as well as the differences in the impact on purchase intention
between SMIs and natural people.

11
3. Theoretical framework
This chapter sets out to explain the present research's theoretical framework. The section begins
by outlining the framework of the theory of purchase funnel, afterwards is followed by the theory
of planned behaviour (TPB). The goal of using the purchase funnel and TPB is to explore factors
that could impact the purchase intention of consumers. And based on that investigate the
differences of impact between digital display ads with normal human images or SMI images and
digital display ads without human images on the purchase intention of consumers. After the
theoretical framework is presented, based on the AIDA model and TPB model, the hypotheses
development is presented. Lastly, the research model for this study is demonstrated.

3.1 Purchase funnel

The purchase funnel is a marketing model that focuses on consumers and this theory is deeply
rooted in the traditional hierarchy of effects models (Barry & Howard, 1990). The hierarchy of
effects models suggests that advertising is a driving force that propels consumers over time through
different discrete stages and ultimately leads them to make a purchase (Barry & Howard, 1990).
There are many different types of purchase funnel models, the traditional linear purchase funnel
model - AIDA purchase funnel model will be used in this study. And the acronym of the name
refers to attention, interest, desire, and action. AIDS model highlights the four-step cognitive
processes that an individual goes through after receiving a new idea or purchasing a new product
(Michaelson & Stacks, 2011). The four-step formula is to get attention, attract interest, and create
desire, and the last step is to take action (Heath & Feldwick, 2007). This model is highly effective
in analyzing the influence of advertising because it controls every stage of the psychological shift
that occurs from the moment an individual sees an advertisement until the individuals involved
make a purchase (Kojima et al., 2010).

In the online advertising world, although the AIDA model was proposed centuries ago, the basic
principles of the model are still relevant. Consumers still need to be aware of the existence of

12
products or services, show interest in products or services based on the relevant information
obtained, and express a desire or intention for these products or services because these products or
services meet the needs, wants and interests of consumers, and finally, take action to make a
purchase decision or other relevant behaviour (Michaelson & Stacks, 2011). The AIDA model has
different levels. The first level is the cognitive level which is about getting the attention or
awareness of consumers. Digital display advertisements can draw consumers’ attention and
awareness, and let consumers know the existence of the product or service. The second level is the
affective level, consumers are interested in the services being offered and would like to learn more
about them. Digital display advertisements are also relevant at this level, by providing attractive
information to let the consumer generate consideration or even intention. The last level is the
behaviour level, the action takes place and consumers will think of the service or the product as a
valued resource.
Figure 2. AIDA model

It has been argued previously that the AIDA model has been used extensively in the online
marketing field and is suitable for studying the impact of advertising (Heath & Feldwick, 2007;
Kojima et al., 2010). However, it is also argued that there is a lack of research into how the AIDA
model can be applied to online marketing. Thus, this study will base on the AIDA model and

13
adapts the model with other theory to analyse the impact of advertising on consumer purchase
intention.

3.2 Theory of Planned Behaviour

The theory of planned behaviour (TPB) is a theory that extends from the theory of reasoned action
(TRA) which was first proposed in 1980 by Martin Fishbein and Ajzen (Ajzen, 1985). The TPB
theory is organically rooted in psychological science but now has been applied to a wide variety
of fields, including advertising, communication, public relations, and healthcare. The reason why
TPB theory has been been used in so many studies is the fact that it provides a thorough
understanding of the determinants of consumer intent and behaviour (Vermeir & Verbeke, 2008).
The strength of TPB theory is that it links beliefs to behaviour. In TPB, the most immediate
predictor of human actual behaviour is the behavioural intention, combine with other factors which
could impact behavioural intention in the theory, TPB provides a causal structure for explaining
and predicting a wide range of human behaviours including consumer behaviours (Ajzen, 1985;
Madden, Ellen & Ajzen, 1992). As a result, it is possible to explore how and what different variable
factors under personal dynamics influence consumer behaviour through the TPB theory model. In
the context of advertising, the goal of advertising is to create consumer awareness, to create an
impression of the brand and product, and finally to convince the consumer that their product or
service is the best to motivate them to make a purchase decision (Trosslöv, 2001). Based on TPB
theory, the impact of human images in ads on consumer intention and purchase behaviour can be
explored by controlling variables in advertising.
Figure 3. Theory of planned behaviour

14
As visualized in Figure 3, attitude, subjective norm, and perceived behavioural control influence
an individual's intention, and both the individual's intention and perceived behavioural control
could influence the individual's final behaviour. Usually, the stronger the intention to engage in
behaviour, the more likely it is to be carried out (Vermeir & Verbeke, 2008). In the theory model,
the individual's intention is the critical construct since it mediates variables between personal
dynamics and behaviour. According to Ajzen (2011), intention is the immediate antecedent of
behavior and indicates a person's preparedness to do a certain behavior, and the intention is the
direct function of attitude, subjective norm, and perceived behavioural control. And there is no
assumption that attitude, subjective norm, and perceived behavioural control are all formed in
rational, unprejudiced or all of them reflect reality as it is, regardless of how people arrive at their
views about behaviour, norms, and control, their attitudes toward behaviour, subjective norms, and
perceptions of behavioural control all follow their beliefs automatically and consistently (Ajzen,
2011). Thus, a realistic response is what the TPB model requires.

Moreover, attitudes, subjective norms, and perceived behavioural control also influence each other.
Attitude refers to the level to which a person's opinion or appraisal of behaviour is favourable or
unfavourable (Ajzen, 1985; Vermeir & Verbeke, 2008). The attitude towards a certain behaviour
reflects the individual's evaluation of whether a certain behaviour is good or bad (Ajzen, 1985;
Vermeir & Verbeke, 2008). In general, the more positive the attitude towards the behaviour, the
stronger the individual's willingness to perform the behaviour (Armitage & Conner, 2001).
Subjective norm is the perception of one person towards a specific behaviour that can be influenced
by others, especially the judgement of significant others (Ajzen, 1985; Amjad & Wood, 2009). If
a person agrees (or disagrees) with others' behaviour, they are more (or less) likely to engage in
the behaviour (Hegner, Fenko & Teravest, 2017). Perceived behavioural control (PBC) is about
the perception of one person towards performing a particular behaviour and the perception could
be easy or difficult (Ajzen, 1985). In general, perceived behavioural control is comprised of two
components: self-efficacy (concerning the ease or difficulty of carrying out a behaviour) and
controllability (the extent to which performance is up to the actor) (Ajzen, 2001). If a person feels
they lack opportunities to perform behaviours, or they feel it's quite hard to get the information or
resources they need, they are less likely to form strong intentions to perform behaviours (Vermeir
& Verbeke, 2008). In consumer purchasing behaviour, perceived behavioural control reflects past

15
experience, as well as whether the consumer could buy the product easily (Vermeir & Verbeke,
2008). In general, when an individual has a favourable attitude toward a behaviour, the attitude is
consistent with relevant norms, and the individual sees a high level of behavioural control, the
individual is likely to have a strong desire to conduct the activity in the issue. Finally, if the
individual has a sufficient level of actual control over their behaviour, they should be able to carry
out their intentions when the opportunity arises (Ajzen, 2001; Godin, 2006).

3.3 Hypotheses development

Guided by the AIDA model and the TPB model, a testable hypothesis and research model are
created. The main objective of this section is to develop a hypothesis based on previous research
and a theoretical framework. The testable research model consists of five components: attitude,
human image, social norm, and knowledge, which are all predicted to influence consumers’
purchase intention. The research model will be presented at the end of this chapter.

3.31 Attitude
Attitude refers to the degree to which a person's opinion or appraisal of behaviour is favourable or
unfavourable (Ajzen, 1985; Vermeir & Verbeke, 2008). It represents people's positive or negative
evaluations of participation in a specific behaviour. In other words, people prefer to embrace
behaviour that could bring them good outcomes or experiences (Fishbein & Ajzen, 2011).
Individuals are more inclined to adopt a behaviour toward which they have a favourable attitude
(Armitage & Conner, 2001). Manchanda et al. (2006) point out that digital display advertising has
an impact on consumer recognition and brand attitude, which demonstrates the importance that
having good display advertisements. Previous literature found that there is a positive association
between attitude toward advertisement and attitude toward the brand (Amos et al., 2008; Silvera
& Austad, 2004). In other words, good advertising can lead to good consumer attitudes towards
the brand and enhance the brand's position in the minds of consumers. Moreover, Reed et al. (2012)
suggest that consumers use the brands they used to show their identity to others, and likewise, they
evaluate others based on their consumption behaviour and the brands of the products they use.
Consumers will therefore prefer to get products from brands that have a better brand image in their
minds (Reed et al., 2012). Thus, if consumers have a positive attitude towards the ad, it is likely to

16
generate purchase intention. The advertisement image that the brand conveys is therefore very
important. Based on the preceding arguments the following hypotheses are created:
H1: Consumers' attitudes toward the ad that has a positive effect on purchase intentions

3.32 Human image

Under the literature review about visual and verbal factors in digital display advertising, visual and
verbal elements could have an impact on consumers’ attention. In the visual media understanding
process and in the consumer purchase funnel model, attention refers to the choice of the individual
to focus on specific items while ignoring others (Martinet, Lablack, Lew & Djeraba, 2009).
According to the AIDA model, only by getting the consumer's attention will there be an
opportunity for the consumer to consider whether to buy the product or not (Michaelson & Stacks,
2011). When consumers see advertisements, they initially evaluate them as a whole, and once the
stimulus has been detected, they will respond by noting the object's characteristics, relevance, and
value (Just & Carpenter, 1976; Barden, 2013). Barden (2013) suggests that the specific item that
consumers are used to seeing, can help consumers create attention and identify new items. Previous
research show that the human image, especially the human face, is a uniquely effective stimulus
for attracting visual attention and identifying new items compared to other objects (Vuilleumier &
Schwartz, 2001; Vuilleumier, 2000). The human image encompasses different types of characters,
and this study will focus on the effect of SMIs and normal people's images on consumer purchase
intentions. Moreover, Sajjacholapunt and Ball (2014) conducted that compared to the
advertisements that did not contain faces, the advertisements with human faces will make viewers
stay longer, and viewers will generate a higher level of consideration which leads to better memory
for advertising content. The high level of consideration also helps to drive consumer purchasing
decisions. Thus, human faces are likely to be a powerful means of attracting and retaining viewers'
attention as well as generating purchase intention in an online advertising environment (Schouten
et al., 2020). Therefore, it is hypothesized that:
H2: Human image has a positive effect on purchase intentions

3.33 Social norm

According to the TPB model, the subjective norm is the perceived social pressure to participate or
not participate in behaviour that was impacted by normative beliefs (Fishbein & Ajzen, 2011).

17
Normative beliefs refer to the perceived pressure to engage in a certain behaviour and that pressure
is from significant referent individuals or groups such as the person’s friends, family, or can be
anyone (Fishbein & Ajzen, 2011). Consumers are more likely to behave in a particular way when
they are under pressure from their environment or other individuals (Rhodes & Courneya, 2003).
A referent individual encourages to perform a behaviour, or the referent individual is likely to
perform the behaviour by him or herself, these two situations will promote the consumer to perform
the behaviour (Fishbein & Ajzen, 2011). Moreover, as social beings, the behaviour of a person is
surrounded by social norms which have an impact on how we think, what we do, and a variety of
other behavioural decisions (Cialdini et al., 1990; Reno et al., 1993). And previous studies also
highlight the fact that the perceptions of social norms can affect behaviour, not the actual social
norms (Cialdini et al., 1990; Reno et al., 1993). In other words, perceived social norms can impact
the whole society and everyone’s behaviour in society. This also means that perceived social norms
can influence the behaviour of those around the consumer which can ultimately affect normative
beliefs as well as subjective norms. According to TPB, the intention is a function of attitudes
towards behaviour, subjective norms, and perceived behavioural control (Fishbein & Ajzen, 2011).
As the perceived social norm can determine the subjective norm, subjective norms are expected to
be positively related to purchase intention. Consequently:
H3: Perceived Social norm has a positive effect on purchase intentions

3.34 Knowledge
Perceived behaviour control refers to people's perceptions about their capacity to do a specific
behaviour (Fishbein & Ajzen, 2011). It is assumed that perceived control is determined by the
control belief (Ajzen, 2011; Fishbein & Ajzen, 2011). Control beliefs are associated with the
perceived presence of factors that might help or impede behavioural performance (Fishbein &
Ajzen, 2011). At the individual level, control belief is influenced by several factors such as age,
gender, education, knowledge, income, emotions, etc (Ajzen, 2011; Fishbein & Ajzen, 2011). In
the context of advertising, if consumers could not get enough trusted information, or if they could
not evaluate the quality of the product or service, uncertainty will arise (Moorman et al., 1993).
The presence of uncertainty can hinder the performance of the behaviour. As digital display
advertisements cannot change the age, gender, income, intelligence, values, stereotypes, or
education level of the viewer, ads can enhance the knowledge of the audience by providing useful

18
information. If a person gains sufficient knowledge about the behaviour from the advertisement,
then it will reduce uncertainty and increases control beliefs towards specific behaviour. And
according to Ajzen (2011), the control belief determines perceived behaviour control. Moreover,
as to the TPB model, perceived behaviour control is expected to be positively related to purchase
intention. Thus, the last hypothesis is:
H4: Knowledge has a positive effect on purchase intentions

3.4 Research model and Hypothesis

Figure 4. Research model

The conceptual research model is constructed based on the proposed hypothesis, as visualized in
Figure 4. Since there are so many different types of human images, this study focuses on two types
of human images which are SMI images and normal people's images. This study will explore the
effect of human image on consumer purchase intentions and get insight into the difference between
the effect of SMI images and the effect of normal people images. So, in the human image
hypothesis, “1” refers to the SMI images and “2” refers to the normal people's images. According
to TPB theory, attitude, subjective norm, and perceived behavioural control could affect intention
towards behaviour. Based on previous literature and background factors about TPB theory, except
attitude, the social norm replaces the subjective norm, and knowledge replaces the perceived
behavioural control. Moreover, according to previous research about human images in digital
display ads and visual and verbal factors in digital display advertising, the human image could

19
have an impact on consumer purchase intention. Thus, the human image was introduced into the
research model. The first hypothesis (H1) suggests a positive relationship between the independent
variable attitudes and the dependent variable purchase intention. The second hypothesis (H2)
predicts that human image could positively influence purchase intention. The third hypothesis (H3)
is similar to hypothesis one but tests the independent variable social norm. The last hypothesis (H4)
predicts that knowledge could also positively impact consumer purchase intention. Thus, this
research model will use these four hypotheses to identify how to use human images in digital
display advertising and explore the differences in the impact of different types of human images.
All theories are based on purchase funnel theory and the TPB model and will be investigated
further in subsequent sections.

20
4. Methodology
This chapter explains the methodology used in the thesis, which includes the research paradigm,
research design, survey design, sample selection and data collection, and finally ethical
considerations on methodology. The study employs a quantitative approach, with primary data
gathered using online surveys that contain two different questionnaires. These two questionnaires
in combination constitute a controlled AB-test experiment. The aim of this section is to explain
and describe the methods adopted for the current research. The section begins with an outline of
the research paradigm and an overview of the research design. Then comes a more in-depth look
at the survey design. This includes a detailed discussion of the questionnaire part and the
experiment part of the research design, example brand, and measurements. Following that, sample
selection and data collection will be explored, and ethical considerations will be present at the end
of this chapter.

4.1 Post-positivism paradigm and research design

The current research is undertaken within post-positivism as a philosophy for scientifically

observing and understanding reality. Developing from positivism, post-positivism philosophy is
increasingly noted as a basis for contemporary empirical research activity (Clark, 1998). The
positivist perspective assumes that objective reality is an existence that can be accurately perceived
through human senses, and positivist researchers believe that via experiment and observation, they
could gain complete knowledge (Ryan, 2006). Concepts and knowledge are seen to be the result
of direct experience, which is then understood by logical deduction (Ryan, 2006). But positivism
paradigm ignores the importance of subjective, social, spiritual, cultural, social, or experiential-
based biases (Clark, 1998). It was the criticism and limitation of this paradigm of positivism that
led to the creation of post-positivism as well as the emergence of the post-positivist paradigm.
Post-positivist research argues that absolute truth is nowhere to be found; it is not about
discovering the truth, but about the scientific endeavour to explore reality and represent it (Muijs,
2004). Moreover, researchers and their perceptions are not seen as completely removed from

21
inquiry, and personal process and participation are considered to be features of human inquiry
(Clark, 1998; Ryan, 2006).

This study aims to contribute knowledge about how consumers perceive the presence of human
faces in advertising and the impact of different types of human images on consumers' intention to
buy smartphones. The ambition of this study is to test the set of hypotheses to understand the
differences in the perceived impact on consumers of the presence of human images in digital
display advertising and to test the impact of the differences on consumers' purchase intentions.

Traditional hypothesis construction involves using existing literature and theory to formulate
hypotheses (Muijs, 2004). In this study, all hypotheses are guided by the existing literature and
theories. Thus, the deductive method is used in the study, and all hypotheses are generated from
the theory. A causal design was used for the study to measure the impact of specific changes on
existing norms and assumptions. Causal effects mean when changes in the independent variable
lead to changes in the dependent variable. In this study, the independent variable is the attitude,
human image (both social media influencer images and normal people images) in advertising,
social norm, and knowledge, while the dependent variable is the consumer's intention to purchase.
Since in the post-positivism research tradition, statistics and experiments are methods of
attempting to represent reality (Clark, 1998; Ryan, 2006), to explore the effects of human images,
this thesis will use the experiment method combined with a statistic to represent the reality.

4.2 Sample selection and data collection

This thesis aim to gain and deepen knowledge about the effect of human faces especially SMIs in
adverts on consumer purchase intention. If a person can use a smartphone or computer and access
the internet, regardless of their age, education level, and country, when they got to a website or
social media platforms, they could see digital display adverts. Thus, to ensure that participants
understand the context of the study, one of the most important criteria for the participants is that
they could use smartphones or computers to access the internet. As the questionnaire was
distributed online, it was ensured that all participants have access to the internet through a mobile
phone or computer. Originally, people who are not familiar with social media and SMIs should be

22
excluded from this study. As it is unlikely that all participants were able to identify the social media
influencers used in this study. If participants could not give recognition to the social media
influencer shown in the ad, then the presence of the social media influencer in the advertisement
appeared the same to them as the presence of a normal person in the advertisement, both being
unfamiliar people appearing in the advertisement. Despite this concern, these participants were
included in this study because their data may be useful in certain hypotheses and compare the
difference in the impact of identifying social media influencers on consumer purchase intentions.
Moreover, to detect potential bias in the results after the data collection, participants were asked
whether they recognize the social media influencer shown in the ad as control questions.

The sampling approach in this study is convenience sampling which is a nonprobability sampling
approach, and the participants were not selected randomly, but rely on readily available subjects -
those that are nearby or easily accessible (Van de Ven, 2007; Lune & Berg, 2017). In particular,
this study implicated an online convenience sample which means everyone that came across the
survey link could take part. One advantage of online convenience sampling is that participants
could easily access the survey. Data collection through the network can also effectively improve
the speed of data collection and reduce the time cost of data collection. Moreover, the biggest
advantage of convenience sampling makes it is possible to reach many people in a short period
with little effort and cost (Lune & Berg, 2017).

However, the convenience sampling approach also has disadvantages. Bias is the biggest
disadvantage of convenience sampling (Etikan, Musa & Alkassim, 2016). Since the distribution
of the questionnaire is not completely random, certain groups will dominate the sample and the
sampling is subjective, so the convenience sample cannot be representative of the whole social
group (Lune & Berg, 2017; Etikan, Musa & Alkassim, 2016). Although the survey will be
distributed through a variety of channels, distribution through personal social networks will expose
participants to the strong influence of the author's sociodemographic characteristics and may lead
to bias. For example, younger generations especially the millennial generation are familiar with
social media and are more receptive to SMIs, and many younger users will tend to think of SMIs
as like their close friends and trust the products recommended by SMIs (Lou et al., 2019) Younger
generations are used to receiving promotions from SMIs. Those who are unfamiliar with social

23
media and social media influencers, especially older people, are not used to receiving promotions
from SMIs which may have a different attitude towards purchasing products recommended by
SMIs. To address this, many sociodemographic factors (such as age and education level) were
asked in the questionnaire and accounted for in the data analysis. In addition, as this survey requires
respondents to provide personal information, respondents could deliberately conceal their true
thoughts for the purpose of protecting their personal information. To prevent this bias, at the
beginning of the start page, a short information content emphasized that all the personal
information and answer would be collected anonymously and kept confidential, only the author
has access to this information.

Despite these disadvantages, convenience sampling was chosen for this work. Firstly, the target
population was too large to allow for randomisation. In addition, the work had to be completed in
a very short period, which is why convenience sampling was suitable because of the high speed of
data collection and the ease of access to subjects. As the study was written as part of a master's
thesis and was not funded by the company, the cost-effectiveness of convenience sampling is
another reason to choose the convenience sampling approach.

Data were collected for three weeks, from March 16th to April 4th. The poll was distributed via
all social media platforms (Facebook, Facebook Groups, Facebook Messenger, WhatsApp,
LinkedIn, Instagram Messenger) with a shareable link. This strategy was chosen since it allowed
for the access of both older and younger sectors of the population. The goal was always to stick to
population quotas.

4.3 Survey design

An experimental quantitative research design is adopted in this study and a two-folded study
design will be used to fulfil the aim of the study. A questionnaire, using Google form create,
consisted of different sections. This study will have two surveys and each survey include six
sections. These six sections can be divided into three parts. The first part of the survey is conducted
to explore the background information of participants about their attitude towards SMIs, attitude
towards Samsung brand, attitude towards smartphones, attitude towards technology, and their use

24
habits of social media. The relevant sections of the first part are shown in section one and section
five of the questionnaire. Section one of the questionnaire is conducted to monitor the extent to
which respondents were similar and to control for standard deviation and normality, as the study
was based on a convenience sample. Thus, section one focused on collecting demographic
information about the respondents and asking them about their frequency of use of social media
and whether they know SMIs. Consists of demographic questions including age, gender, education
level, occupation, annual household income, and the degree of familiarity with and use of social
media. Section five relates to the attitude of participants towards the Samsung brand, smartphones,
technology, and SMIs. This section enables participants' attitudes to be investigated in preparation
for data analysis.

The second part is conducted to do the experiment which consisted of three sections (section two,
section three, section four). And each section will display one advertisement, in which participants
were asked to answer six questions based on their thoughts after viewing the images. The questions
under all advertisements are the same and include questions related to consumers’ attention,
interest, purchase intention, attitude, social norm, and knowledge. The detail of the experiment
will be explained in the following experimental design section.

The last part is about recognition and manipulation checks. This section is to confirm whether
participants identified the social media influencer images that have been used in the experiment,
gave feedback on the survey, and whether they wanted to know the final results of the survey. All
the measurements are presented in the following part and the survey can be found in Appendix 2.

4.3.1 Why Samsung

Samsung as a global multinational manufacturing conglomerate was founded in the year 1969 by
Lee Byung-Chui who was driven by the mission of “supporting people to be their best” (Samsung,
n.d.). Samsung has many businesses, the most well-known of which is its smartphone business.
After 50 years of growth and expansion into a global business, Samsung has become the world's
largest mobile phone manufacturer (Statista, 2021b). Since the success of its smartphone business,
Samsung is an internationally recognized leading company in technology and a Top 10 global
brand (Samsung, n.d.). As a leading smartphone brand manufacturer, Samsung is well known to

25
Europeans and has a stable smartphone market share of around 30% in Europe (Statista, 2021b).
Moreover, Samsung spends a huge amount of money on digital advertising every year. Not only
celebrities, but Samsung has also cooperated with many social media influencers even normal
people to promote their products. Samsung's collaboration with a wide range of people makes
Samsung an interesting case study for exploring knowledge about marketing communications.
Recently, Samsung has just launched a new generation of phones, the Galaxy S22 series.
Consumers are not yet fully aware of the new product, even some of them do not know that the
new product has been launched, which ensures that they are unlikely to have any bias towards the
product and thus facilitates the conduct of this study. Therefore, the new Samsung-branded Galaxy
S22 series was chosen as the example product to create ads in this study.

4.3.2 Experimental design

The second part of the survey is to generate new knowledge about whether human images could
impact participants’ purchase intention towards Samsung smartphones. This section will provide
a detailed explanation of the experiment and the designed stimuli. The manipulated advertisements
were created by the authors themselves and all of them are about the latest generation of Samsung
Galaxy 22 series phones. Since the aim of this experiment was to test the impact of the presence
of a human image, and consumers’ reactions and attitudes to ads can change depending on the
platform on which ads are displayed, all participants were not informed about the advertisements
posted on a specific website or social media platform account.

The objective of the current study is to examine how human images in advertisements influence
consumers' purchase intention toward smartphones. Moreover, this study hopes to explore not only
the impact of human images on consumer intentions but also explore the differences in the impact
of social media influencer images and normal people images on consumer intentions. According
to previous research, human faces are likely to be a powerful means of attracting and retaining
viewers' attention in an online advertising environment (Schouten et al., 2020; Xiao & Ding, 2014).
Therefore, this experiment chose to place the human face as the main factor in the advertisement
to test the effect of the human image on consumer intention.

26
To achieve the study objectives, six different advertisements for Samsung Galaxy S22 series
phones were created. The six advertising images are based on three Samsung phone posters which
can be divided into three groups, each group contains two advertising images with the same
background. For each group, both ads have the same dimensions, the same caption, and the same
background image. To test the impact of social media influencers on consumer intentions, the first
group of ads differed in whether included images of SMIs. For the images of SMIs, since this
research is about mobile phone products, Mkbhd, an influencer focusing on technology products
with over 10 million followers on several different social media platforms, was selected. The
second group of ads was designed to explore the differences in the effect of SMI images and
random people images on consumer purchase intentions, so the ads differ in that one contains
images of SMI and the other one contains images of random people. Regards to the images of
random people, as the SMIs were male, female images were chosen to explore whether the gender
of the characters in the advertisement would have a different effect on participants. To avoid
disputes, a random AI-generated image was downloaded from the Generated photos website (the
image is free to use for educational research purposes). To explore if the image of random people
would have an effect on consumer purchase intention, the difference in the third group of ads was
whether the image of ordinary people was included. Three groups of ads are shown in Appendix
1. The six ads are divided into two ranges, each containing three different ad images.

Two questionnaires (control group questionnaire and treatment group questionnaire) are included
in this survey, each containing the same questions, the difference is that the two questionnaires
show images of advertisements from different ranges. A start page for the survey was created via
the google app script. Each participant was randomly assigned to take the survey on one of two
questionnaires with a random probability of 50% on control group questionnaire, and 50% on
treatment group questionnaire. Participants were asked to answer all the mandatory questions and
state their degree based on their thoughts after viewing the images of the advert. By comparing the
data from the participants' answers to the two questionnaires, it is possible to investigate the effects
of manipulation on participants who are exposed to different advertisements. Before the official
launch of the survey, a small pre-test was conducted in which ten random consumers were invited
to see if all the ad images and questions were clear and understandable. Based on the opinions of

27
the participants in the pre-test, small adjustments were changed before the survey was published
online, including phrasing, grammar, the placement of the caption, and the size of human images.

4.3.3 Measurements
The five constructs of interest in this study are all measures on Likert scales. All 38 - item
questionnaire was elaborated based on the TPB model as well as the AIDA model (Fishbein &
Ajzen, 1980; Michaelson & Stacks, 2011). All factors are included in the questionnaire in order to
explore the effect of the human image in the ad. Guerrero (1996) argues that the number of
questions in the questionnaire was related to the credibility of the responses, with fewer questions
obtaining reliable information from the same participants. Therefore, both questionnaires used in
this survey were 38 questions. The exact number of questions can be seen in Table 1.
Table 1:

Construct No. of Items Sources

Attitude 3 Sparks et al. (1995) and

Huang (2018)
Human image 0
Social norm 3 Chetioui, Benlafqih, and
Lebdaoui (2020)

Knowledge 3 Sparks et al. (1995)

Purchase intention 3 Huang (2018)

Attitude: The operationalisation of the dependent variable ‘attitude’ includes attitude towards ads,
Samsung brand, smartphones, technology, and SMIs. The seven-point Likert scale (1= strongly
disagree to 7= strongly agree) is used to measure the participant’s attitude. All questions related to
‘attitude’ are developed from the previous study by Sparks et al. (1995) and Huang (2018).

Human image: The questionnaires did not include questions about the influence of human images
on purchase intentions since the results of the two questionnaires were used to compare the
influence of human images on consumer purchase intentions.

28
Social Norm: Social norm measurement is developed from the previous study done by Chetioui,
Benlafqih, and Lebdaoui (2020) operationalized a seven-point Likert scale (1= strongly disagree
to 7= strongly agree) to measure perceived knowledge of Ads and SMIs.

Knowledge: Knowledge is based on the Sparks et al. (1995) scale which is a five-item scale
measured on a five-point Likert scale. To ensure the coherence of the questions as well as to
streamline them, this study was adjusted to a seven-point Likert scale and choose only two items.

Purchase intention toward the Smartphone: The purchase intention toward the smartphone is
measured using a seven-point Likert scale (1= strongly disagree to 7= strongly agree) based on a
previous study done by Huang (2018).

4.4 Ethical considerations

To ensure ethical standards, ethical concerns were made throughout the study. On the start page
of the survey, informed consent was provided to all participants in the experimental survey, which
is a requirement for agreeing to begin the survey. The informed consent describes the purpose of
the study as well as how anonymity and confidentiality will be maintained. It was also mentioned
that all data would be kept confidential. The data will be used exclusively to answer the research
question, and no personal information will be shared with third parties. Only question number 36
may collect sensitive personal data, in which participants are asked if they would like to leave their
email address to receive short results of the study. However, this question is not mandatory, and
participants can simply skip this question. Question number 35 on whether participants have
questions about the survey can also be skipped. In addition to the above two questions, all
remaining questions with an asterisk (*) in the survey are mandatory that all participants need to
answer and cannot skip. The purpose of this is to avoid missing data that could make data
unavailable and cause problems for analysis. Pallant (2016) points out the pairing deletion in
multiple regression will exclude cases of specific analysis and induction, which processes data by
replacing missing data with average values, is not considered a substitute for real answers, and the
results may tend to look more refined. Therefore, there may be a risk that some cases will be
excluded when the analysis is run, and the results may be reported incorrectly due to the small

29
number of items on the scale. Finally, the survey could be cancelled at any time, participants were
told that they could drop out of the survey at any time. Therefore, if any question makes the
participant feel uncomfortable or unsuitable to answer, the participant has the option to cancel the
survey. On the start page of the survey, participants were informed of all these ethical criteria. If
they agree, they can press 'start experiment' to start the actual survey.

30
5. Result and analysis
In this chapter, the quantitative analysis and the results of the experiment will be presented. The
study aims to provide insight into how human image usage impacts the purchase intention of
consumers. Based on the survey result and combining the AIDA model with TPB theory. This
section will use the IBM SPSS Statistics program to analyze all the data and thereafter present all
the experiment results. This will provide insight into the impact of human images in ads as well as
the difference between SMI images and normal human images. The first section sets out the
analyze the descriptive data, followed by multiple analyses to test the control variables, AIDA
factors, and hypothesis. The last section explores the result of the experiment and compares the
purchase intention mean values among all the stimuli.

5.1 Descriptive analysis

In order to examine and analyse the answers to the survey, different analyses appropriate to the
different variables were carried out. The first step was a descriptive analysis, which explored the
means and standard deviations between the different variables. After the data cleaning, 283 valid
responses remained, of these, 141 respondents answered the questionnaire for the control group
and the remaining 142 respondents answered the questionnaire for the treatment group. The gender
composition of the respondents was as follows, with a majority of 51.6% female, 46,6% male, and
1.8% non-conforming. The mean age of respondents is 28.42 years (SD = 6.744), highly skewed,
with the lowest age of 16 years and the highest age of 66 years old. The sample trend to have a
good academic background, 62.9% of the respondents had already obtained a bachelor’s degree
and 20.5% of the respondents have obtained a master’s degree. 57.7% of the respondents are
currently in school, 40.6% of respondents are currently working, and the rest respondents are
unemployed. The distribution of annual household income was relatively skewed to the right
which is shown in Figure 5. The majority of respondents have a high level of social media use,
280 respondents will use social media every day, 2 respondents use social media twice a week and
only 1 respondent uses social media once a month. And the mean value of time respondents spend
on social media every day is 2.08 hours, 50.2% of respondents will spend two to three hours on

31
social media every day. Regarding which type of phone respondents use, 45.9% of respondents
use Android smartphones, 52.7% of respondents use Apple smartphones, and the rest 1.4% of
respondents use other types of smartphones. Statistics on the following number of SMIs show that
only 5 respondents did not follow any SMIs, and the majority of respondents follow 6-10 SMIs on
social media. A demographic profile of the sample is presented below.

Figure 5. Annual Household Income Before Taxes

Table 2. Demographic profile of the sample

Frequency Percentage

Gender
Female 146 51.6%
Male 132 46.6%
Non-conforming 5 1.8%
Age
Under 25 119 42%
25-30 63 22.3%
31-35 69 24.4%
Over 35 32 11.3%
Education Level
High School or less 44 15.5%
Bachelor’s degree 178 62.9%

32
 Master’s degree 58 20.5%
Doctorate degree 3 1.1%
Occupation
Student 163 57.6%
Working 115 40.6%
Unemployed 5 1.8%
Annual household income
Less than 10.000 60 21.2%
€10.000 to €29.999 111 39.2%
€30.000 to €49.999 74 26.1%
€50.000 to €69.999 23 8.1%
€70.000 to €89.999 8 2.8%
€90.000 to €99.999 2 0.7%
€100.000 or more 5 1.8%
The frequency of visiting social media
Every day 280 98.9%
Twice a week 2 0.7%
Once a month 1 0.4%
Time spent on social media every day
Less than 1 hour 17 6%
1 hour 44 15.5%
2-3 hours 142 50.2%
3-4 hours 60 21.2%
More than 5 hours 20 7.1%
Smartphone type
Android 130 45.9%
Apple 149 52.7%
Other 4 1.4%
The number of how many SMIs followed
NO 5 1.8%
1-5 69 24.4%

33
 6-10 139 49.1%
11-20 hours 39 13.8%
More than 20 31 11%

Regarding the attitude toward smartphones and technology. The mean value of attitude toward
smartphones is 6.31, and the mean value of attitude toward technology is 6.00 which shows the
positive attitude that respondents have towards smartphones and technology. Regarding the
attitude toward the Samsung brand, the mean value is 4.43. Regarding respondents' satisfaction
with their current skills in using mobile phones, the mean value is 6.1. When looking at the
perception of SMIs, the mean value is 9.98. All values are shown in Table 3.

Table 3.

Variable Mean SD

Attitude-Smartphone 6.31 0.725

Attitude-Technology 6.00 0.853
Attitude-Samsung 4.43 1.360
Phone use skill 6.10 0.715
Perception-SMIs 9.98 2.357

(N=283)

For the control group and treatment group, the independent variable 'attitude' is negatively skewed
(-1.054) which indicates clustering of scores at the high end. A positive kurtosis value indicates
that distribution is peaked. 'attitude' has a positive kurtosis value (2.387) which indicates the
distribution is peaked. Regarding the mean, the 5% trimmed mean and the original mean are very
similar (13.06 and 13.25). Given this, it can be shown that the values are not too different from the
remaining distribution and the mean is not significantly affected by the extreme data. Thus, all
data will be retained. Similarly, 'social norm' and 'knowledge' are also negatively skewed, 'social
norm' is approximately symmetric (-0.33) with a positive kurtosis of 0.612, and 'knowledge’ is

34
highly skewed (-1.842) with a positive kurtosis of 5.429. For both 'social norm' and 'knowledge',
all data will be kept as the 5% trimmed mean and the original mean are very similar ('social norm':
5% trimmed mean 10.70, original mean 10.65; 'knowledge': 5% trimmed mean 14.63, original
mean 13.34). The dependent variable ‘purchase intention’ is also negatively skewed (-1.482) and
has a positive kurtosis of 3.108. As the 5% trimmed mean and the original mean are very similar
(5% trimmed mean 13.52, original mean 13.27), all data will be retained.

For the treatment group, all three independent variables are negatively skewed, and the skewness
value of 'attitude', 'social norm', and 'knowledge' are -1.185, -.394, and -1.331. Same as the control
group, all three variables have a positive kurtosis value ('attitude':1.772, 'social norm': 0.103,
'knowledge': 2.171). Moreover, for all independent variables, as the 5% trimmed mean and the
original mean are very similar, all the data will be kept ('attitude': 5% trimmed mean 13.08, original
mean 12.86; 'social norm': 5% trimmed mean 10.42, original mean 10.37; 'knowledge': 5%
trimmed mean 14.37, original mean 14.15). The dependent variable ‘purchase intention’ is also
negatively skewed (-1.341) and has a positive kurtosis of 1.289. Similarly, the 5% trimmed mean
and the original mean are very similar (5% trimmed mean 13.17, original mean 12.90), and all the
data about purchase intention will be kept. All means, standard deviations, and correlation are
shown in Table 4.

Table 4. Descriptive Results and Correlation Matrix

Control group

Variable Mean SD Scale Range 1 2 3

1 Attitude 13.06 2.69 1-7

2 Social norm 10.65 3.25 1-7 0.650
3 Knowledge 14.34 2.80 1-7 0.625 0.468
4 Purchase intention 13.27 2.90 1-7 0.752 0.758 0.612

(N=141)

35
Treatment group

Variable Mean SD Scale Range 1 2 3

1 Attitude 12.86 3.10 1-7

2 Social norm 10.37 3.16 1-7 0.750
3 Knowledge 14.15 2.82 1-7 0.690 0.553
4 Purchase intention 12.90 3.37 1-7 0.792 0.715 0.637

(N=142)

5.2 Control variables and AIDA factors testing

To explore the effect of control variables as well as AIDA factors on the dependent variable, a
regression analysis is conducted. With the help of multiple regression, it is possible to presume a
causal relationship between the dependent and independent variables. And the result of the
regression analysis will show which independent variable is the best “predictor”. The first block
contains the control variables which include gender, phone type, attitude towards smartphones,
attitude towards Samsung, attitude towards technology, and skills for using the smartphone. The
second block pulls in the AIDA factors which are attention and interest. The third block adds the
three main predictors which are 'attitude', 'social norm', and 'knowledge' to the regression model
and the results will be presented in the following hypotheses testing section.

All data must be checked before running a regression analysis to ensure that the regression analysis
is suitable, and that the data fulfils all assumptions. Firstly, Pallant (2020) points out that the
dependent variable must be measured on a continuous scale, and the independent variables must
be categorical or continuous. Although it is widely discussed whether the Likert scale is an ordinal
or continuous variable. In this study, the Likert scale is treated as a continuous variable. By
recoding the non-conforming gender and other types of smartphones to system missing, gender
and using phone types these two variables are changed to dummy variables. Except for these two
dummy variables, the rest variables are measured by the seven-point Likert scale. Thus, all

36
variables can be taken into regression analysis. Secondly, residuals must have normality, linearity,
independence, and homoscedasticity (Pallant, 2020). Through normal p-polt can inspect whether
the residuals are normally distributed or not and can check the linear relationship between the
dependent relationship and the independent variable. Moreover, through normal p-polt can also
inspect whether the data show homoscedasticity. After testing, the data from this study met the
above requirements. Thirdly, before the regression analysis, multicollinearity also needs to be
checked, to ensure that the independent variables cannot be highly similar to each other. A
correlation analysis will be conducted among all the continuous independent variables. About the
result, on every model, all values are below r=.9 which indicates that the independent variables are
not highly correlated (Pallant, 2020). Additionally, the VIF (‘Variance Inflation Factor’) and
Tolerance value can help control multicollinearity, if the VIF values are greater than 10 and
Tolerance less than 0.1 indicates multicollinearity (Pallant, 2020). All values meet the assumptions,
hence there is no multicollinearity in this study. Finally, the sample size needs to be more than
50+8M (M=independent variable numbers) and should not include any significant outliers or any
influential points (Pallant, 2020). The regression analysis did have 8 independent variables, and
the minimum sample size for 8 independent variables is 114. There were five respondents who did
not identify their gender and four respondents who used the rest of the phone types. Because of
the conversion to a dummy variable excluding these nine respondents, the valid sample size is 272
which is still much larger than the minimum sample size request. As previously mentioned, all the
data will be kept as there are no significant outliers or any influential points that could significantly
influence the mean. In summary, the data from this study can be used for regression analysis. For
a simpler and clear overview, the results are shown in Table 5 below.

Table 5. Coefficients results for control variables and AIDA factors

Unstandardized Coefficients Standardized Coefficients T Sig

Beta Beta

Model 1 △R2 = .185, p <.000

Gender .714 .113 1.999 .047
Phone Type -.334 -.053 -.992 .357

37
 Attitude-Smartphone .590 .136 2.237 .026
Attitude-Samsung .447 .193 3.182 .002
Attitude-Technology .887 .240 3.578 .000
Skill-Phone use -.901 -.204 3.229 .001

Model 2 △ R2= .694, p <.000

Gender .335 .053 1.520 .130
Phone Type .116 .018 .517 .606
Attitude-Smartphone .052 .012 .314 .753
Attitude-Samsung .148 .064 1.689 .092
Attitude-Technology .020 .005 .125 .900
Skill-Phone use -.377 -.085 -2.174 .031
Attention .331 .319 4.763 .000
Interest .537 .513 7.837 .000

In the first model, the regression model explains 18.5% of the variance in the dependent variable,
and the result is also significant (p < 0.00) with an adjusted R Square of 0.167. The regression
result shows that attitude toward technology makes the biggest contribution to predicting the
'purchase intention', with a beta coefficient of β = 0.240 (t=3.578, p<0.00). The sig value for the
phone type is more than 0.05. Thus, using phone type does not make a significant and unique
contribution to predicting consumers' purchase intentions. The other five independent variables
contribute to predicting consumer purchase intention. Since the model R Square value is less than
0.3, which shows all these five independent variables have a none or very weak contribution to
predicting consumer purchase intention (Moore & Kirkland, 2007).

In the second model, ‘attention’ and ‘interest’ from the AIDA model are taken into the regression
model. The new model explains 69.4% of the variance in the dependent variable, and the result is
also significant (p < 0.00) with an adjusted R Square of 0.685. The regression result shows that
‘interest’ makes the biggest contribution to predicting the ‘purchase intention’, with a beta
coefficient of β = 0.513 (t=7.837, p<0.00). And ‘attention’ has fewer contributions to predicting

38
the 'purchase intention', with a beta coefficient of β = 0.319 (t=4.763, p<0.00). The skill of using
the smartphone is the one that makes the least contribution, with a beta coefficient of β = -0.085
(t=-2.174, p<0.031). The sig value for the rest independent variables is more than 0.05, which
means all other independent variables turn out to be non-significant. Thus, when taking AIDA
factors to the regression model, gender, phone type, attitude towards smartphones, attitude towards
Samsung, and attitude towards technology, these six independent variables do not make a
significant and unique contribution to predicting consumers' purchase intentions.

5.3 Hypotheses testing

There are four hypotheses in this study, three of them which are about 'attitude', 'social norm', and
'knowledge' can be answered by regression analysis, and the remaining one hypothesis about
“human image” needs to be answered by comparing the results of the control group with treatment
group which will show in the below section. Hypothesis 1 states that higher levels of perceived
attitude will generate significantly more purchase intention toward smartphones. Similarly,
hypothesis 3 states that higher levels of the social norm will generate significantly more purchase
intention toward smartphones, likewise, hypothesis 4 states that higher levels of the knowledge
will generate significantly more purchase intention toward smartphones. Therefore, to know how
well the independent variables predict the dependent variable, a regression analysis is conducted.
With the help of multiple regression, it is possible to presume a causal relationship between the
dependent and independent variables. And the result of the regression analysis will show which
independent variable is the best “predictor”. Since the research model aims to investigate causal
relationships, multiple regression analysis is the best approach.

By adding 'attitude', 'social norm', and 'knowledge' to the third block, 'purchase intention' was
regressed onto all the independent variables. According to the regression result, the overall model
explains 77.1% of the variance in the dependent variable, and the result is significant (p < 0.00)
with an adjusted R Square of 0.762. When the sample size is small, the adjusted R square is the
most suitable and optimistic value to provide (Pallant, 2020). The scatterplot and normal
probability plot showing the distribution of the residuals can be found in Appendix 3. Regarding
the hypothesis support, the regression result shows that 'social norm' makes the biggest

39
contribution to predicting the 'purchase intention', with a beta coefficient of β = 0.277 (t=6.436,
p<0.00). And 'attitude' has fewer contributions to predicting the 'purchase intention', with a beta
coefficient of β = 0.171 (t=3.109, p<0.002). However, the p-value for 'knowledge' is more than
0.05 which means 'knowledge' turned out to be non-significant. Thus, based on the result from the
regression analysis, hypotheses one and three are supported, but hypothesis four is not supported.
Among three main independent variables, 'social norm' makes the strongest unique contribution to
explaining the model. All results of the new model are shown in Table 6 below

Table 6. Coefficients result for the final model

Unstandardized Coefficients Standardized Coefficients T Sig

Beta Beta

△ R2= .771, p <.000

Gender .436 .069 2.259 .025
Phone Type -.176 -.028 -.891 .374
Attitude-Smartphone -.091 -.021 -.628 .531
Attitude-Samsung .016 .007 .206 .837
Attitude-Technology .021 .006 .150 .881
Skill-Phone use -.198 -.045 -1.299 .195
Attention .191 .185 2.983 .003
Interest .308 .294 4.728 .000
Attitude .186 .171 3.109 .002
Social Norm .273 .277 6.436 .000
Knowledge .068 .061 1.443 .150

5.4 Experimental stimuli testing

As in line with the aim of the thesis, this section aims to explore the efficacy of human image in
advertisements can positively or negatively influence consumers’ purchase intention, and do the

40
images of SMIs and ordinary or normal people have a different impact on consumers? To answer
these two questions, all respondents were divided into two groups according to whether they
identified the image of the SMI used in the survey or not. In the control group, 104 respondents
did not identify the image of the SMI and 37 respondents identified the SMI; in the treatment group,
106 respondents did not identify the image of the SMI and 36 respondents identified the social
media influencer. Since for respondents who did not identify with the SMI, the image of the SMI
could be considered as an ordinary person. Thus, for respondents who did not recognise the SMIs,
the images of the SMI image and the AI-generated human image were almost the same - images
of unrecognisable characters. Moreover, by analysing data from respondents who identified the
SMI image that was used in the advertisement, it is possible to explore the differences in influence
between social media influencers and ordinary personas. By applying a bivariate analysis in SPSS,
it is possible to compare the mean values of variables in different conditions.

In the first group of advertisements (Shown in Appendix 1), the control group's advertisement did
not include human images, while the treatment group's advertisement included an SMI image. The
results of respondents who did not identify SMIs showed that the respondents exposed to the ad
with the human image have a mean value toward the 'purchase intention' of 4.61. While the
respondents exposed to the ad without the human image have a mean value toward the 'purchase
intention' of 3.32. For respondents who can recognise SMIs, the mean value of the ad with an SMI
image is 5.13 and the mean value of the ad without any human images is 3.89. This shows that
regardless of whether respondents identify SMIs or not, ads that include images of people can lead
to a higher mean values about consumer purchase intention than ads that do not include images of
people.

Similarly, the difference between the two ads in the third group of advertisements (Shown in
Appendix 1) is also whether the human image is included or not. The control group’s ad includes
a normal human image and the ad in the treatment group did not include any human images. The
results of respondents who did not identify social media influencers also show that the mean value
(4.56) toward the 'purchase intention' from the ad with the human image is higher than the mean
value (3.60) from the ad without the human image. For respondents who can recognise SMIs, the
mean value of the ad with a normal human image is 4.56 and the mean value of the ad without a

41
normal human image is 3.50. The above results also demonstrate the positive impact of human
image on motivating respondents to make a purchase intention.

In the second group of advertisements (Shown in Appendix 1), the control group's advertisement
included the SMI image, while the treatment group's advertisement included an AI-generated
image of people. For respondents who can recognize the SMI in this study, the results showed that
the respondents exposed to the ad with the SMI image have a mean value toward the 'purchase
intention' of 5.27. While the respondents exposed to the ad with the AI-generated human image
have a mean value toward the 'purchase intention' of 4.38. Therefore, the result indicates that the
respondents that saw the advertisement with the SMI image will generate more purchase intention
toward the product compared to the ones that were exposed to the advertisement with the AI-
generated human image. Hypothesis 2 states that the human image has a positive effect on purchase
intentions, based on the results from the experiment, hypothesis 2 is supported. All the results are
presented in Table 7 below.
Table 7. Descriptive results.
Respondents who cannot recognize the SMI

Purchase intention Range Ads with human image Ads without human image
Mean Mean
The first group ads 1-7 4.61 3.32
The third group ads 1-7 4.56 3.60

Respondents who can recognize the SMI

Purchase intention Range Ads with human image Ads without human image
Mean Mean
The first group ads 1-7 5.13 3.89
The third group ads 1-7 4.56 3.50
Purchase intention Range Ads with SMI image Ads with AI human image
Mean Mean
The second group ads 1-7 5.27 4.38

42
Lastly, the human image has an impact on attention, interest, attitude, social norm, and knowledge.
In both the control group and the treatment group, for respondents who cannot recognize SMIs,
the mean values of all independent variables for ads that include human images are higher than the
mean values for ads that do not include human images (All data was shown in Appendix 3).
Moreover, in the second group of advertisements (Shown in Appendix 1), the control group
advertisement used an image of a male and the treatment group used an image of a female. Thus,
it is possible to explore the effect of demographic variables like occupation, and education, on
consumer purchase intention. Through analysis of the data from respondents who cannot recognize
the SMI in this study. The mean value of male respondents toward the ad with a male image is
5.02, while the mean value is 4.91 when male respondents saw the ad with a female image. For
female respondents, when they were exposed to advertisements containing an image of a male the
mean value is 4.71, and when they were exposed to the ad with a female image the mean value is
4.50. The differences turn out to be small. Through the compare means analysis between purchase
intention and occupation, the result shows the differences among different occupation statuses also
turn out to be small, thus there is no significant relationship between occupation and purchase
intention. The same situation also happens when comparing the means between education level
and purchase intention.

In summary, the results of all three groups above demonstrate the positive impact of human images
on purchase intention which means hypothesis 2 is supported. The respondents that saw the
advertisement with the human image will generate more purchase intention toward the product
compared to the ones that were exposed to the advertisement without the human image included.
Moreover, for the person who can recognise the SMI, compare to normal people, advertisements
with SMI images could have a greater impact on consumers' purchase intentions than those that
do not. All results of the hypothesis test and the experimental stimulus test will be discussed further
in the next chapter.

43
6. Discussion and conclusion
This chapter first discusses the result of the previous chapter and what they mean for the
hypotheses as well as a further discussion of the results and findings of the experiment. This section
will discuss what the results mean in terms of the research questions, the issues observed, and how
it relates to the theoretical framework and previous literature. Then comes with conclusion section.
This section will answer the research question and summarise the results of the study. Following
that, implications, limitations, and suggestions for future research will be present.

6.1 Interpretation

The descriptive statistics show that all respondents in fact have a positive attitude toward the
smartphone, technology, skills to use smartphones, and SMIs. The mean value of their attitude
toward smartphones and technology are 6.31 and 6.00 respectively, with a minimum value of 1
and a maximum of 7, which indicates the attitude amongst the respondents. And the test on
respondents' skills in using mobile phones also shows that respondents are satisfied with their
current skills in using mobile phones with a mean value of 6.10. Moreover, more than 98% of
respondents use social media on a daily basis and follow more than one SMI, of these respondents,
more than 73% follow more than six SMIs. This illustrates that the majority of respondents to this
survey are familiar with SMIs. In terms of respondents' perception of SMIs, the mean value is 9.98
with a minimum value of 2 and a maximum of 14. This result shows that respondents trust SMIs
and are influenced by SMIs when making purchase decisions. But the attitude of respondents
toward the Samsung brand is more neutral with a mean value of 4.43. This indicates that there is
no clear preference or dislikes for the Samsung brand, and the neutrality of the survey results is
beneficial to its accuracy.

The first block of the hierarchical regression analysis includes gender, phone type, attitude towards
smartphones, attitude towards Samsung, attitude towards technology, and skills for using the
smartphone. The regression model only explains 18.5% of the variance in intention to buy
Samsung smartphones, which shows this model has a very weak contribution to predicting
consumer purchase intention (Moore & Kirkland, 2007). However, through add ‘attention’ and

44
‘interest’ from the AIDA model to the regression model, the new regression model could explain
69.4% of the variance in consumer purchase intention. Compare to the first model only explains
18.5% of the variance, adding ‘attention’ and ‘interest’ to the model improve the regression model
significantly. ‘Attention’, ‘interest’, and ‘skills for using the smartphone’ turn out to be significant.
In the new model, ‘interest’ makes the strongest contribution to predicting the 'purchase intention';
and ‘attention’ makes the second-place contribution; ‘skills for using the smartphone’ makes the
minimal contribution. The results are consistent with the previous research by Ghirvu (2013),
which confirmed that high levels of attention or awareness and interest could positively impact
purchase intentions. Moreover, when taking ‘attention’, ‘interest’, ‘attitude’, ‘social norm’, and
‘knowledge’ into the same regression model. The regression result shows that ‘interest’ still makes
the strongest unique contribution to predicting ‘purchase intention’, with a beta coefficient of β =
0.294. This shows that interest levels have a greater impact on consumers' purchase intentions.
How to increase consumer interest in the future is a question that advertisers should consider.

6.11 Hypotheses discussion

To answer all hypotheses, the third block of the hierarchical regression analysis was used. The
third model explains 77.1% of the variances in consumer purchase intention. Compare to the
second model, adding factors related to the TPB model improves the regression model further. The
first hypothesis state that there is a positive significant relationship between the dependent variable
'purchase intention' and the independent variable ‘attitude’. Among the three main independent
variables, the regression result indicates that ‘attitude’ is the second-place contribution to
predicting the dependent variable, with a beta coefficient of β = 0.171. It has been discussed by
Ajzen (2011) that ‘attitude’ is an important predictor of behaviour. Moreover, Armitage and
Conner (2001) stated that the more positive the attitude towards the behaviour, the stronger the
individual's willingness to perform the behaviour. Thus, the results of the final regression model
are consistent with previous research done by Armitage and Conner. In line with the hypotheses,
‘attitude’ makes a unique contribution to predicting ‘purchase intention’.

Regarding the second hypothesis, the results of the experiment provide exploratory results to
support hypothesis 2. For respondents who did not recognise the SMI used in this experiment, in
the first group of ads, the mean purchase intention value of respondents towards the advertisement

45
with the human image is 4.61. In comparison to the advertisement without the human image, the
mean difference is -1.29. Similarly, in the third group of ads, the ad that includes images of human
receive a higher mean of purchase intention, and the mean difference between the ad with the
human image and the ad without the human image is -0.96. For respondents who can recognize
the SMI, in the first group of ads, the mean purchase intention value of the ads containing SMI
images is 5.13. In comparison, the mean value of purchase intention for the ad that does not include
an image of SMI is 3.89, with a mean difference of -1.24. Likewise, in the third group of ads, the
mean value of purchase intention for ads that include AI-generated personas is 4.56, while the
value without personas is 3.50, a mean difference of -1.06. In both the control group and the
treatment group, all result above entails that whether lirespondents recognise the human image
used in the advert, human images could help them generate a higher level of purchase intention.
As the second hypothesis state that human image has a positive effect on purchase intentions,
which means hypothesis 2 is also supported. These results are consistent with research by Xiao
and Ding (2014), which found that if an image of a person in the advertisement matches viewer
preferences, the human image can not only attract viewers' attention but also positively affect
viewers' response to advertisements.

Regarding the third hypothesis, the perceived ‘social norm’ has a positive effect on ‘purchase
intention’. According to the regression analysis, the contribution of ‘social norm’ makes the
strongest contribution to predicting the ‘purchase intention’, with a beta coefficient of β = 0.277.
The result indicates that ‘social norm’ could positively affect ‘purchase intention’. And this result
is consistent with research from Belgiawan et al. (2017) which found that social norms are an
important predictor of purchase behaviour and could significantly correlate with purchase intention.

The fourth hypothesis stat between the that there is a positive relationship between ‘knowledge’
and ‘purchase intention’. Compared to the other two variables, ‘knowledge’ is the only one that is
non-significant. Suki (2016) found that knowledge is the most significant determinant of consumer
purchase intention, and knowledge about the product positively relates to purchasing intention.
But according to the regression result, ‘knowledge’ did not contribute to predicting the dependent
variable, with a beta coefficient of β = 0.061, p=0.150. Since ‘knowledge’ failed to provide a

46
significant contribution to this study, hypothesis 4 in this study is rejected. This may be consistent
with adding factors from the AIDA model as the independent variables to the regression model.

Overall, the finding from the regression analysis indicates that ‘attitude’ and ‘social norm’ are all
positively related to ‘purchase intention’. Compared to the other main independent variables,
‘social norm’ made the strongest contribution to predicting the dependent variable. Except for
Hypothesis4 is rejected, Hypothesis 1, Hypothesis 2, and Hypothesis 3 of this study are supported.

6.12 Experiment result discussion

Based on the result of the experiment, it is possible to analyze the differences between the effect
of SMIs and ordinary people by analyzing the data of respondents who recognized social media
influencers. In the second group of ads, the mean value of purchase intention for the advertisement
that includes the SMI image is 5.27, and the mean value of the ad that includes an AI-generated
normal people image is 4.38, the mean difference is -0.89. For the same respondents in the control
group, the mean value of purchase intention when they saw an ad containing SMI images is 5.27,
while the mean value of purchase intention when they saw an ad containing the normal people
image is 4.56. Similarly, in the treatment group, the mean value of purchase intention when seeing
an ad containing SMI is 4.61, while the mean value of purchase intention when seeing an ad
containing a normal person is 4.38. In sum, in both the control group and the treatment group, all
the above result shows that advertising with the SMI image leads to higher levels of purchase
intention. Compare to normal people's images, SMI images can have a greater impact on
consumers’ purchase intention. The result of the experiment is consistent with previous research
done by Lou et al. (2019), which stated that consumers trust SMIs almost as they trust their friends
and they are willing to buy products that are recommended by SMIs. Therefore, compared to
normal people, SMIs could have a better impact on consumers' purchase intention.

From the perspective of the research model in this thesis, consumers' purchase intention is
influenced by ‘attitude’, ‘human image’, ‘social norm’, and ‘knowledge’. According to the
correlation analysis, in both the control group and the treatment group, there is a positive
correlation between attitude, social norms, and knowledge. Moreover, in the control group, except
for the correlation between social norm and knowledge, the rest correlation between independent

47
variables are large correlation. And in the treatment group, all correlations are large correlation.
This result is consistent with the original TPB model in which all dependent variables affect each
other (Ajzen, 2011). Thus, the three independent variables of attention, social norm, and
knowledge can influence each other, and all three independent variables have a positive
relationship with each other.

In addition, by checking the mean value of all variables as well as attention and interest, it was
found that the mean value of the advertisements with a human image was higher than the mean of
those that did not contain a human image. In the control group, the first ad does not include the
human image, and the other two ads include the human image. For respondents who did not
recognize SMIs, towards the first ad, the mean value of ‘attention’, ‘interest’, ‘attitude’, ‘social
norm’, and ‘knowledge’ are 3.97, 3.71, 3.17, 2.49, and 3.5. While, towards the second ad, all mean
values are 5.19, 4.88, 4.73, 3.95, and 5.04, which are all higher than the mean value for the first
ad. Likewise, in the third ad, all the mean values are 5.34, 5.14, 4.86, 4.06, and 5.72 which are also
higher than the mean value for the first ad. The same result is found in the treatment group data,
where the third ad without a human image has a lower mean value than the two ads with a human
image. Moreover, by checking the correlation between ‘attention’ and ‘interest’ with other
independent factors, all relationships are largely correlated. Based on the experimental result from
both the control group and the treatment group, human images have a positive impact on other
independent variables in the research model as well as attention and interest. The above results are
in line with previous studies (Vuilleumier &Schwartz, 2001; Vuilleumier, 2000; Schouten et al.,
2020) on function of human images in ads.

Overall, the findings from the experiment indicate that all the respondents will generate a high
level of purchase intention when they are exposed to the ad with human images. When the sample
is divided by whether they recognize the SMI that shows in the ad, for respondents who recognize
the SMI, the presence of an SMI in an advertisement creates a higher purchase intention than the
presence of a normal people in an advertisement. In addition, there is a positive influence
relationship among attitudes, human image, social norms, and knowledge. And human images in
advertising can also influence consumers' attention and interest. Ultimately, these results

48
demonstrate evidence of combining the TPB model with the AIDA model, as well as the positive
impact on consumers of the use of human images in advertising.

6.2 Conclusion

This study aimed to contribute new knowledge about how human images impact consumers’
purchase intention after exposure to advertisements, besides, to contribute knowledge about the
impact of using SMI images and normal people images in advertisements. A new research model
was proposed based on the AIDA and TPB models, and the respondent's purchase intention was
defined and analyzed using the research model. Moreover, by using an experimental approach,
this study was able to contribute to a better understanding and new understanding of human images,
SMI images, and normal people images affect consumers’ purchase intention. The study was set
out to answer the following research question: How to use human images in digital display
advertising to increase the effectiveness of advertising and the purchase intention of consumers?

The result of the regression analysis provides evidence that attention, interest, attitude, and social
norms are positively correlated with consumers' purchase intention; human images can positively
influence all the above factors, including the consumer's purchase intention. Among all significant
independent variables, ‘interest’ is the best predictor of purchase intention; among the three main
predictors, ‘social norm’ is the best predictor. Additionally, by exploring the adjusted R Square,
the interaction among ‘attention’, ‘interest’, ‘attitude’, and ‘social norm’ explains 71.1% of the
variance in purchase intention. This implies that the result is consistent with the majority of
previous studies, which have been discussed earlier in the thesis.

Moreover, the experimental part of the study shows that consumers will have different levels of
purchase intentions when faced with different advertisements, and the images of the people shown
in the advertisements can influence consumers' purchase intentions. For all respondents who
participated in the experiment, whether respondents recognised the human images used in the
advertisements, the presence of the human image in the same advertising context can increase
consumers' purchase intention. In comparison to a normal people image, the SMI image was likely
to derive a higher level of purchase intent among consumers when respondents recognized the SMI

49
image included in the advertising. And based on the results of the experiment, it was also found
that human image was also positively related to the other independent variables in the research
model as well as ‘attention’ and ‘interest’ which are from the AIDA model. Through analysis,
there is a large correlation between factors from the AIDA model and research model factors which
are relate to the TPB model. Thus, factors in the research model do not only affect the ‘desire’
stage in the AIDA model but also affect the two stages before the ‘desire’ stage which are the
‘attention’ stage and the ‘interest’ stage.

Given the above, this study has provided insight and contributed to the knowledge about how
human images in advertisements impact consumers’ purchase intention. By applying and
extending the AIDA model and TPB model, it is possible to test whether and how involvement
factors correlate with ‘purchase intention’. In addition, to contribute knowledge about influencer
marketing, the study has provided evidence that compared with normal people’s presence in
advertisements, the presence of SMIs in advertisements could have a greater positive impact on
consumer purchase intention.

6.3 Implications

The finding of this thesis enriched the existing body of knowledge and bring previous SMI and
digital display advertising research as well as the TPB model and the AIDA model into new content.
The results of this study confirm that human images in advertisements could help consumers
generate a higher level of attention. Furthermore, this study finds that the presence of a human
image could lead to higher levels of purchase intentions. For strategic communication academics,
this study provides the basis for future research that delves into the impact of the presence of
personas in advertising on consumers. Previous research has focused too much on SMIs and
traditional celebrities, neglecting the impact of the presence of ordinary personas in advertising.
To further develop the research about marketing communication especially digital display
advertising research, it is necessary to study many types of characterisation.

Moreover, this research showed that ‘attention’, ‘interest’, ‘attitude’, and ‘social norm’ are useful
factors to predict consumers’ purchase intention. However, ‘knowledge’ did not work significantly

50
as the predictor. In this study, ‘knowledge’ is derived from PBC which is the main difference
between the TRA model and the TPB model. This result suggests academics further elaborate and
derives PBC variables to better predict consumers’ intention. Possibly, attention and interest in the
AIDA model could work as a predictor of purchase intention. Apart from that, it stands to reason
to explore purchase intention through different theoretical backgrounds and explorative
approaches to enrich the field. Except for the effect of human images, the experimental
investigation further explores the difference between SMI images and normal people images.
Previous research has neglected to look at the relevance of normal personas, but many brands have
started to use normal consumers in their advertising. Thus, it is relevant for strategic
communication academics theoretically to further study this phenomenon and expand its
perspective. As the results of this study found that generic personas have a positive impact on
consumers' attention, interest, and purchase intention, it could address the problem of online
advertising being habitually ignored by consumers, a phenomenon that promises to enrich the field
of marketing communication research as well as strategic communication research.

Lastly, the public's perception of an organisation's reputation can be influenced by advertising

(Lloyd-Smith & An, 2019). Therefore, the findings are relevant to the field of brand
communication, where the choice of persona in advertising can influence consumer perceptions,
which in turn affects the perception of an organisation's reputation. Furthermore, understanding
individuals' perceptions of the use of personas in advertising can provide valuable insights for
strategic communicators working in marketing and advertising, which can improve the
effectiveness of online display advertising. Ultimately, with the help of an experimental method,
this study provides personal insight into human images in digital display advertising.

6.4 Limitations

Although this thesis explores some interesting and valuable findings that contribute to the existing
body of knowledge on marketing communications as well as strategic communication, it is
important to note the limitations. Firstly, the limitations of the convenience sample method. As
discussed, the use of a convenience sample has several limitations that limit the generalisability
and representativeness of the population in this thesis. The results show that most of the

51
respondents who responded to the survey were under the age of 35, academically well educated,
use social media every day, familiar with the internet, and followed SMIs. And this experiment
ignores the impact of the aesthetics of advertising design on consumers. Therefore, the results of
this study are not fully representative of all people in society as a whole and are less ecologically
valid. Second, this thesis follows a post-positivist research tradition and applies a quantitative
approach. As a result, another study and research paradigm limitation is that the results and
attitudes are totally based on numbers. In-depth comparisons might have helped to better
understand individual attitudes if a mixed-methods approach had been used. Moreover, as the
experiment was conducted online and the survey platform used did not have access to the time
spent by each participant in the experiment, there was no guarantee that each respondent had taken
the experiment seriously. Third, this study did not consider cultural and other types of bias. The
cultural perceptions and aesthetic preferences of different regions can affect consumers' attitudes
towards advertising and, in turn, the results of this study. Focusing on a particular region or cultural
group may make the findings of the study more accurate. Finally, the consumer purchase decision
encompasses a complex and multi-stage consumer journey, and although this study extends the
TPB theory further, there is still the assumption that the consumer has been given the opportunity
and resources to successfully perform the behaviour, ignoring the remaining variables that
influence behavioural intentions and motivations.

6.5 Suggestions for future research

The limitations of this thesis lead to new questions that could be of interest for future research.
Firstly, convenience sampling was used in this study and the respondent population was too
concentrated to be representative of the entire consumer population. Future research could use
random sampling to conduct surveys in a particular region, country, or cultural group to obtain
more accurate findings. Furthermore, future research could move away from mobile phones as a
product category and explore the impact of using personas for other products on consumer
purchase intentions. Secondly, the literature recognises the lack of research and the need for a more
comprehensive understanding of the process of consumer purchase decisions as well as exploring
factors that influence the consumer decision process (Kojima et al., 2010; Lu, Chuan, & Chang,
2014; Sundar & Noseworthy, 2014; Munnukka et al., 2016). The purpose of this thesis is to explore

52
and contribute to a better understanding of how the use of human images in advertisements can
influence consumers' purchase intentions and contribute new knowledge that the use of human
images can influence consumers' intention toward the product that is shown in advertisements. As
discussed, the findings suggest that one hypothesis about the PBC is not supported, and this result
conflicts with previous research. The reason for this may be constant with adding AIDA factors to
the regression model with TPB factors. Future research could further explore how to better
combine the AIDA model with the TPB model to better predict consumer purchase intention.
Moreover, further research aimed to test the AIDA model and the TPB framework with online
display advertising, but with other factors would be meaningful, such as the price of the product,
the specific platform on which the ad was posted, etc. Through further expanding TPB theory and
purchase funnel theory to understand if other factors functioned as peripheral cues for advertising.

Finally, due to the quantitative approach to research, all consumer attitudes, social norm, etc. are
analysed based on numbers. Future research could use qualitative research, for example, in-depth
interviews to gain insight into consumers' perceptions of the presence of characters in advertising
and the impact on their attitudes towards advertising. Moreover, previous literature has examined
the impact of using celebrities or social media influencers on brand image, and future research
could also explore the impact of using images of ordinary consumers in advertising on the brand
image through qualitative research.

53
7. References
Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In Action control (pp.
11-39). Springer.
Ajzen, I. (2002). Perceived behavioral control, self‐efficacy, locus of control, and the theory of
planned behavior 1. Journal of applied social psychology, 32(4), 665-683.
Amjad, N., & Wood, A. M. (2009). Identifying and changing the normative beliefs about
aggression which lead young Muslim adults to join extremist anti‐Semitic groups in
Pakistan. Aggressive Behavior: Official Journal of the International Society for Research
on Aggression, 35(6), 514-519.
Amos, C., Holmes, G., & Strutton, D. (2008). Exploring the relationship between celebrity
endorser effects and advertising effectiveness: A quantitative synthesis of effect size.
International journal of advertising, 27(2), 209-234.
Andrews, M., Luo, X., Fang, Z., & Ghose, A. (2016). Mobile ad effectiveness: Hyper-contextual
targeting with crowdedness. Marketing Science, 35(2), 218-233.
Armitage, C. J., & Conner, M. (2001). Efficacy of the theory of planned behaviour: A meta‐
analytic review. British journal of social psychology, 40(4), 471-499.
Barden, P. P. (2013). Decoded: the science behind why we buy. John Wiley & Sons.
Barry, T. E., & Howard, D. J. (1990). A review and critique of the hierarchy of effects in
advertising. International journal of advertising, 9(2), 121-135.
Bart, Y., Stephen, A. T., & Sarvary, M. (2014). Which products are best suited to mobile
advertising? A field study of mobile display advertising effects on consumer attitudes and
intentions. Journal of Marketing Research, 51(3), 270-285.
Belgiawan, P. F., Schmöcker, J.-D., Abou-Zeid, M., Walker, J., & Fujii, S. (2017). Modelling
social norms: Case study of students’ car purchase intentions. Travel Behaviour and
Society, 7, 12-25.
Berger, J. (2014). Word of mouth and interpersonal communication: A review and directions for
future research. Journal of consumer psychology, 24(4), 586-607.
Bergkvist, L., & Zhou, K. Q. (2016). Celebrity endorsements: a literature review and research
agenda. International journal of advertising, 35(4), 642-663.
Blackwell, R. D., Miniard, P. W., & Engel, J. F. (2006). Consumer behavior. South-Western Pub.

54
Bucher, H.-J., & Schumacher, P. (2006). The relevance of attention for selecting news content. An
eye-tracking study on attention patterns in the reception of print and online media.
Burke, M., Hornof, A., Nilsen, E., & Gorman, N. (2005). High-cost banner blindness: Ads increase
perceived workload, hinder visual search, and are forgotten. ACM Transactions on
Computer-Human Interaction (TOCHI), 12(4), 423-445.
Chetioui, Y., Benlafqih, H., & Lebdaoui, H. (2020). How fashion influencers contribute to
consumers' purchase intention. Journal of Fashion Marketing and Management: An
International Journal.
Cialdini, R. B., Reno, R. R., & Kallgren, C. A. (1990). A focus theory of normative conduct:
Recycling the concept of norms to reduce littering in public places. Journal of personality
and social psychology, 58(6), 1015.
Clark, A. M. (1998). The qualitative‐quantitative debate: moving from positivism and
confrontation to post‐positivism and reconciliation. Journal of advanced nursing, 27(6),
1242-1249.
De Veirman, M., Cauberghe, V., & Hudders, L. (2017). Marketing through Instagram influencers:
the impact of number of followers and product divergence on brand attitude. International
journal of advertising, 36(5), 798-828.
Djafarova, E., & Rushworth, C. (2017). Exploring the credibility of online celebrities' Instagram
profiles in influencing the purchase decisions of young female users. Computers in Human
Behavior, 68, 1-7.
Drèze, X., & Hussherr, F.-X. (2003). Internet advertising: Is anybody watching? Journal of
interactive marketing, 17(4), 8-23.
Dreze, X., & Zufryden, F. (1997). Testing web site design and promotional content. Journal of
Advertising Research, 37, 77-91.
Etikan, I., Musa, S. A., & Alkassim, R. S. (2016). Comparison of convenience sampling and
purposive sampling. American journal of theoretical and applied statistics, 5(1), 1-4.
Fishbein, M., & Ajzen, I. (2011). Predicting and changing behavior: The reasoned action
approach. Psychology press.
Ghirvu, A. I. (2013). The AIDA model for advergames. The USV Annals of Economics and Public
Administration, 13(1 (17)), 90-98.

55
Godin, G., Conner, M., & Sheeran, P. (2005). Bridging the intention–behaviour gap: The role of
moral norm. British journal of social psychology, 44(4), 497-512.
Goldfarb, A. (2014). What is different about online advertising? Review of Industrial Organization,
44(2), 115-129.
Goldfarb, A., & Tucker, C. (2011). Online display advertising: Targeting and obtrusiveness.
Marketing Science, 30(3), 389-404.
Goldstein, D. G., Suri, S., McAfee, R. P., Ekstrand-Abueg, M., & Diaz, F. (2014). The economic
and cognitive costs of annoying display advertisements. Journal of Marketing Research,
51(6), 742-752.
Goodrich, K. (2011). Anarchy of effects? Exploring attention to online advertising and multiple
outcomes. Psychology & Marketing, 28(4), 417-440.
Guerrero, L., Colomer, Y., Guàrdia, M. D., Xicola, J., & Clotet, R. (2000). Consumer attitude
towards store brands. Food quality and preference, 11(5), 387-395.
Hallahan, K., Holtzhausen, D., Van Ruler, B., Verčič, D., & Sriramesh, K. (2007). Defining
strategic communication. International journal of strategic communication, 1(1), 3-35.
Heath, R., & Feldwick, P. (2007). 50 Years using the wrong model of TV advertising. ADMAP,
481, 36.
Hegner, S. M., Fenko, A., & Teravest, A. (2017). Using the theory of planned behaviour to
understand brand love. Journal of Product & Brand Management.
Huang, Y.-T. (2018). The female gaze: Content composition and slot position in personalized
banner ads, and how they influence visual attention in online shoppers. Computers in
Human Behavior, 82, 1-15.
Just, M. A., & Carpenter, P. A. (1976). Eye fixations and cognitive processes. Cognitive
psychology, 8(4), 441-480.
Kojima, T., Kimura, T., Yamaji, M., & Amasaka, K. (2010). Proposal and development of the
direct mail method PMCI-DM for effectively attracting customers. International Journal
of Management & Information Systems (IJMIS), 14(5).
Kuisma, J. (2015). Consumer perception of online advertising-the effects of animation, ad
characteristics, repetition and task relevancy on attention and memory.
Lambrecht, A., & Tucker, C. (2013). When does retargeting work? Information specificity in
online advertising. Journal of Marketing Research, 50(5), 561-576.

56
Lou, C., Tan, S.-S., & Chen, X. (2019). Investigating consumer engagement with influencer-vs.
brand-promoted ads: The roles of source and disclosure. Journal of Interactive Advertising,
19(3), 169-186.
Lou, C., & Yuan, S. (2019). Influencer marketing: how message value and credibility affect
consumer trust of branded content on social media. Journal of Interactive Advertising,
19(1), 58-73.
Lu, L.-C., Chang, W.-P., & Chang, H.-H. (2014). Consumer attitudes toward blogger’s sponsored
recommendations and purchase intention: The effect of sponsorship type, product type, and
brand awareness. Computers in Human Behavior, 34, 258-266.
Lune, H., & Berg, B. L. (2017). Qualitative research methods for the social sciences. Pearson.
Madden, T. J., Ellen, P. S., & Ajzen, I. (1992). A comparison of the theory of planned behavior
and the theory of reasoned action. Personality and social psychology Bulletin, 18(1), 3-9.
Manchanda, P., Dubé, J.-P., Goh, K. Y., & Chintagunta, P. K. (2006). The effect of banner
advertising on internet purchasing. Journal of Marketing Research, 43(1), 98-108.
Martinet, J., Lablack, A., Lew, S., & Djeraba, C. (2009). Gaze based quality assessment of visual
media understanding. 1st International Workshop on Computer Vision and Its Application
to Image Media Processing (WCVIM) in conjunction with the 3rd Pacific-Rim Symposium
on Image and Video Technology (PSIVT), Tokyo-Japan,
Michaelson, D., & Stacks, D. W. (2011). Standardization in public relations measurement and
evaluation. Public Relations Journal, 5(2), 1-22.
Moore, D. S., & Kirkland, S. (2007). The basic practice of statistics (Vol. 2). WH Freeman New
York.
Moorman, C., Deshpande, R., & Zaltman, G. (1993). Factors affecting trust in market research
relationships. Journal of marketing, 57(1), 81-101.
Muijs, D. (2004). Doing quantitative research in education: With SPSS. Sage.
Munnukka, J., Uusitalo, O., & Toivonen, H. (2016). Credibility of a peer endorser and advertising
effectiveness. Journal of Consumer Marketing.
Paivio, A. (1986). Mental representations: A dual coding approach. Oxford University Press.
Paivio, A. (1991). Dual coding theory: Retrospect and current status. Canadian Journal of
Psychology/Revue canadienne de psychologie, 45(3), 255.

57
Pallant, J. (2020). SPSS survival manual: A step by step guide to data analysis using IBM SPSS.
Routledge.
Pavlou, P. A., & Stewart, D. W. (2000). Measuring the effects and effectiveness of interactive
advertising: A research agenda. Journal of Interactive Advertising, 1(1), 61-77.
Percy, L., & Rossiter, J. R. (1997). Advertising communications and promotion management.
McGraw-Hill.
Pieters, R., & Wedel, M. (2004). Attention capture and transfer in advertising: Brand, pictorial,
and text-size effects. Journal of marketing, 68(2), 36-50.
Reed II, A., Forehand, M. R., Puntoni, S., & Warlop, L. (2012). Identity-based consumer behavior.
International Journal of Research in Marketing, 29(4), 310-321.
Reno, R. R., Cialdini, R. B., & Kallgren, C. A. (1993). The transsituational influence of social
norms. Journal of personality and social psychology, 64(1), 104.
Resnick, M., & Albert, W. (2014). The impact of advertising location and user task on the
emergence of banner ad blindness: An eye-tracking study. International Journal of
Human-Computer Interaction, 30(3), 206-219.
Rhodes, R. E., & Courneya, K. S. (2003). Investigating multiple components of attitude, subjective
norm, and perceived control: An examination of the theory of planned behaviour in the
exercise domain. British journal of social psychology, 42(1), 129-146.
Richardson, A., & Sheikh, A. A. (2020). Individual differences in imaging: Their measurement,
origins, and consequences. Routledge.
Ryan, A. B. (2006). Post-positivist approaches to research. Researching and Writing your Thesis:
a guide for postgraduate students, 12-26.
Samsung. (n.d.). Our mission & values. Retrieved 2022-03-15 from
https://www.samsung.com/ca/about-us/company-info/
Sajjacholapunt, P., & Ball, L. J. (2014). The influence of banner advertisements on attention and
memory: human faces with averted gaze can enhance advertising effectiveness. Frontiers
in psychology, 166.
Schouten, A. P., Janssen, L., & Verspaget, M. (2020). Celebrity vs. Influencer endorsements in
advertising: the role of identification, credibility, and Product-Endorser fit. International
journal of advertising, 39(2), 258-281.

58
Scott, D. M. (2015). The new rules of marketing and PR: How to use social media, online video,
mobile applications, blogs, news releases, and viral marketing to reach buyers directly.
John Wiley & Sons.
Shaouf, A., Lü, K., & Li, X. (2016). The effect of web advertising visual design on online purchase
intention: An examination across gender. Computers in Human Behavior, 60, 622-634.
Silvera, D. H., & Austad, B. (2004). Factors predicting the effectiveness of celebrity endorsement
advertisements. European Journal of marketing.
Sparks, P., Shepherd, R., Wieringa, N., & Zimmermanns, N. (1995). Perceived behavioural control,
unrealistic optimism and dietary change: An exploratory study. Appetite, 24(3), 243-255.
Statista. (2021a). Digital Advertising report 2021. Available: https://www-statista-
com.ludwig.lub.lu.se/study/42540/digital-advertising-report/ [2022-02-05]
Statista. (2021b). Samsung's smartphone shipment share in Europe from the 1st quarter of 2019
to the 2nd quarter of 2021. Available: https://www-statista-
com.ludwig.lub.lu.se/statistics/1232278/samsung-smartphone-market-share-in-europe/
[2022-02-05]
Suki, N. M. (2016). Green product purchase intention: impact of green brands, attitude, and
knowledge. British Food Journal.
Sundar, A., & Noseworthy, T. J. (2014). Place the logo high or low? Using conceptual metaphors
of power in packaging design. Journal of marketing, 78(5), 138-151.
Sundar, S. S., & Marathe, S. S. (2010). Personalization versus customization: The importance of
agency, privacy, and power usage. Human communication research, 36(3), 298-322.
Townsend, C., & Shu, S. B. (2010). When and how aesthetics influences financial decisions.
Journal of consumer psychology, 20(4), 452-458.
Vermeir, I., & Verbeke, W. (2008). Sustainable food consumption among young adults in Belgium:
Theory of planned behaviour and the role of confidence and values. Ecological economics,
64(3), 542-553.
Vuilleumier, P. (2000). Faces call for attention: evidence from patients with visual extinction.
Neuropsychologia, 38(5), 693-700.
Vuilleumier, P., & Schwartz, S. (2001). Emotional facial expressions capture attention. Neurology,
56(2), 153-158.

59
Walters, G., Sparks, B., & Herington, C. (2007). The effectiveness of print advertising stimuli in
evoking elaborate consumption visions for potential travelers. Journal of Travel Research,
46(1), 24-34.
Wells, W., Spence-Stone, R., Moriarty, S., & Burnett, J. (2008). Australian Advertising Principles
and Practice. Pearson Education Australia.
Xiao, L., & Ding, M. (2014). Just the faces: Exploring the effects of facial features in print
advertising. Marketing Science, 33(3), 338-352.
Yoo, C. Y., & Kim, K. (2005). Processing of animation in online banner advertising: The roles of
cognitive and emotional responses. Journal of interactive marketing, 19(4), 18-34.

60
8. Appendix
Appendix 1- Ad images
Group 1.

Group 2.

Group 3.

61
Appendix 2 - Questionnaire
Start page.

62
Control group.

63
64
65
66
Treatment group.

67
68
69
70
Appendix 3- SPSS output
Correlation Matrix among AIDA factors and Main predictors
Interest Purchase Attitude Social Knowledge
Intention Norm
Attention .851 .775 .742 .598 .644
Interest .809 .767 .640 .620
Purchase Intention .775 .733 .624
Attitude .701 .659
Social norm .511

P-plot: H1, H3, & H4.

IV: Attention, Interest, Attitude, Social norm, & Knowledge
DV: Purchase intention

71
Scatterplot: H1, H3, & H4.
IV: Attention, Interest, Attitude, Social norm, & Knowledge
DV: Purchase intention

Control group-Mean value for all factors

Respondents who recognize the SMI:

72
Statistics

Atte
ntion
1_1_ Attitude1 Socialnorm Knowledge
R_S Interested1 Purchase1 _1_R_S 1_1_R_SM 1_1_R_SM
MI _1_R_SMI _1_R_SMI MI I I

N Valid 37 37 37 37 37 37

Missing 246 246 246 246 246 246

Mean 4.43 4.0541 3.8919 3.9189 2.8649 3.8108

24

Std. 1.04 1.15340 1.12506 1.25562 1.27284 1.24360

Deviation 191

Statistics

Atte
ntion
2_1_ Attitude2 Socialnorm Knowledge
R_S Interested2 Purchase2 _1_R_S 2_1_R_SM 2_1_R_SM
MI _1_R_SMI _1_R_SMI MI I I

N Valid 37 37 37 37 37 37

Missing 246 246 246 246 246 246

Mean 5.70 5.4865 5.2703 5.3514 4.4324 5.4865

27

73
 Std. 1.22 1.21613 .96173 1.20684 1.48213 1.19307
Deviation 168

Statistics

Atte
ntion
3_1_ Attitude3 Socialnorm Knowledge
R_S Interested3 Purchase3 _1_R_S 3_1_R_SM 3_1_R_SM
MI _1_R_SMI _1_R_SMI MI I I

N Valid 37 37 37 37 37 37

Missing 246 246 246 246 246 246

Mean 5.29 4.7568 4.5676 4.6216 3.7568 5.2162

73

Std. 1.05 .95468 .98715 1.03686 1.11568 1.13370

Deviation 053

Respondents who can not recognize the SMI:

Statistics

Atten
tion1
_1_N Interested1 Purchase1 Attitude1 Socialnorm Knowledge
otR _1_NotR _1_NotR _1_NotR 1_1_NotR 1_1_NotR

N Valid 104 104 104 104 104 104

74
 Missing 179 179 179 179 179 179

Mean 3.971 3.7115 3.3269 3.1731 2.4904 3.5096

2

Std. 1.065 1.04902 .84120 .91844 .90302 1.10599

Deviation 40

Statistics

Atten
tion2
_1_N Interested2 Purchase2 Attitude2 Socialnorm Knowledge
otR _1_NotR _1_NotR _1_NotR 2_1_NotR 2_1_NotR

N Valid 104 104 104 104 104 104

Missing 179 179 179 179 179 179

Mean 5.192 4.8846 4.8462 4.7308 3.9519 5.0481

3

Std. 1.149 1.27186 1.26006 1.16778 1.44397 1.12673

Deviation 73

Statistics

Atten
tion3
_1_N Interested3 Purchase3 Attitude3 Socialnorm Knowledge
otR _1_NotR _1_NotR _1_NotR 3_1_NotR 3_1_NotR

75
 N Valid 104 104 104 104 104 104

Missing 179 179 179 179 179 179

Mean 5.346 5.1442 4.9423 4.8654 4.0673 5.7212

2

Std. 1.320 1.34683 1.42670 1.33698 1.49604 1.42418

Deviation 26

Treatment group-Mean value for all factors

Respondents who recognize the SMI:

Statistics

Atte
ntion
1_2_ Attitude1 Socialnorm Knowledge
R_S Interested1 Purchase1 _2_R_S 1_2_R_SM 1_2_R_SM
MI _2_R_SMI _2_R_SMI MI I I

N Valid 36 36 36 36 36 36

Missing 247 247 247 247 247 247

Mean 5.80 5.3889 5.1389 5.4444 4.6389 5.7778

56

Std. 1.21 1.53582 1.72631 1.42316 1.75910 1.26742

Deviation 466

76
Statistics

Atte
ntion
2_2_ Attitude2 Socialnorm Knowledge
R_S Interested2 Purchase2 _2_R_S 2_2_R_SM 2_2_R_SM
MI _2_R_SMI _2_R_SMI MI I I

N Valid 36 36 36 36 36 36

Missing 247 247 247 247 247 247

Mean 4.75 4.3333 4.3889 4.3056 3.3056 5.0833

00

Std. 1.46 1.30931 1.47895 1.45051 1.19090 1.46141

Deviation 141

Statistics

Atte
ntion
3_2_ Attitude3 Socialnorm Knowledge
R_S Interested3 Purchase3 _2_R_S 3_2_R_SM 3_2_R_SM
MI _2_R_SMI _2_R_SMI MI I I

N Valid 36 36 36 36 36 36

Missing 247 247 247 247 247 247

Mean 4.22 3.9167 3.5000 3.5000 2.7778 3.6111

22

77
 Std. .897 1.18019 .97101 1.15882 .98883 1.33690
Deviation 97

Respondents who can not recognize the SMI:

Statistics

Atten
tion1
_2_N Interested1 Purchase1 Attitude1 Socialnorm Knowledge
otR _2_NotR _2_NotR _2_NotR 1_2_NotR 1_2_NotR

N Valid 106 106 106 106 106 106

Missing 177 177 177 177 177 177

Mean 4.98 4.7642 4.6132 4.5660 3.5566 5.0943

11

Std. 1.46 1.50271 1.58907 1.45418 1.33866 1.33473

Deviation 047

Statistics

Atten
tion2
_2_N Interested2 Purchase2 Attitude2 Socialnorm Knowledge
otR _2_NotR _2_NotR _2_NotR 2_2_NotR 2_2_NotR

N Valid 106 106 106 106 106 106

78
 Missing 177 177 177 177 177 177

Mean 5.08 4.8679 4.6509 4.6698 3.7736 5.3396

49

Std. 1.49 1.53732 1.52467 1.51640 1.39580 1.47284

Deviation 359

Statistics

Atten
tion3
_2_N Interested3 Purchase3 Attitude3 Socialnorm Knowledge
otR _2_NotR _2_NotR _2_NotR 3_2_NotR 3_2_NotR

N Valid 106 106 106 106 106 106

Missing 177 177 177 177 177 177

Mean 4.21 3.8302 3.6038 3.5000 2.9245 3.6132

70

Std. 1.05 1.05534 1.07508 1.26679 1.22434 1.06521

Deviation 112

79