MASTER THESIS
How to understand brand equity, involving the network of strong, favourable and unique brand associations, by using visual user-generated content as source of information?
“Visual user-generated content as a measure for brand equity” A case study on the brand Iamsterdam
University of Amsterdam Faculty of Economics and Business Master of Science in Business Studies
Track: Marketing
Under supervision of: B. Rietveld
Student: Sissy Zwienenberg Student Number: 10003088 Date of submission: 24 June 2016
Statement of originality
This document is written by Student Sissy Zwienenberg who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
TABLE OF CONTENTS
1. INTRODUCTION ... 4
LITERATURE GABS AND RESEARCH QUESTION ... 5
SCIENTIFIC AND MANAGERIAL CONTRIBUTION ... 6
2. LITERATURE REVIEW ... 9
CUSTOMER-BASED BRAND EQUITY ... 9
Brand awareness ... 11
Brand image ... 12
MEASUREMENT MATTERS... 13
THE IMPACT OF E-WOM ON BRANDING ... 17
VISUAL USER-GENERATED CONTENT AS A MEASURE FOR BRAND EQUITY ... 19
Measuring brand equity from a consumers’ perspective ... 20
Measuring associations ... 20
Measuring the consumers’ associative network structure ... 21
Measuring the behavioural component ... 22
The measure is quick to conduct ... 22
The measure is objective ... 23
CBBE is measured in an unconscious and un-verbalizable manner... 23
The measure is easily applicable on a large sample size ... 23
The measure can measure brand equity of 'non-monetary’ brands ... 24
Measuring the social component ... 24
CONCLUSION ... 25
3. THE CASE OF IAMSTERDAM ... 26
AMSTERDAM MARKETING ... 26
THE CITY AS A BRAND ... 27
BRAND EQUITY OF A CITY ... 28
RELEVANCE OF THE CASE ... 30
4. RESEARCH DESIGN ... 31
DATA COLLECTION PROCEDURE ... 31
REDUCING THE DATASET TO A SUBSET ... 32
CONCEPTUAL FRAMEWORK AND OPERATIONALIZING OF THE VARIABLES ... 33
5. ANALYSIS AND RESULTS ... 35
DESCRIPTIVE STATISTICS ... 36
Associations ... 36
Strength ... 38
Favourability ... 39
Uniqueness ... 41
MAPPING CONSUMER-BASED BRAND EQUITY ... 42
6. DISCUSSION ... 50
7. CONCLUSION ... 54
8. LIMITATIONS AND SUGGESTIONS FOR FUTURE RESEARCH ... 56
LIMITATIONS ... 56
SUGGESTIONS FOR FUTURE RESEARCH ... 57
Abstract
Understanding brand equity involves identifying the network of strong, favourable, and unique associations in a consumers’ mind. This research introduces a methodology, based on visual user-generated content, for eliciting brand associations and measuring and mapping brand equity. In this research a model is proposed, that uses visual and textual data from visual user-generated content, to understand brand equity. It is argued that this new methodology remedies problems that occur with exciting brand equity measurements. To illustrate the new method, city branding has been used as an application by measuring brand equity of the brand Iamsterdam. The analyses show that brand associations and their strength, favourability and uniqueness are measurable by investigating visual user-generated content. Based on the conducted analyses, and consequently the mapping of these results, valuable brand related information can be provided to marketing practitioners.
1. Introduction
Brand equity is regarded as a very important concept in business practice as well as in academic research. Brand equity is about identifying the value of a brand and therefore it is an important measure of the brand and a marketeers’ performances. From a consumer’s perspective, Keller (2003) states that brand equity occurs when the consumer is familiar with the brand and holds some favourable, strong and unique associations in mind. This network of associations plays an important role in successfully differentiating a brand and securing competitive advantages (Aaker, 1991; Keller, 1993, 2003). Many researchers have studied the impact of brand equity on business performances. They state that high brand equity increases shareholder value (Madden, Fehle & Fournier, 2006), profit and share prices (Till, Baack & Waterman, 2011), margins (Keller, 1993), and returns (Barth, Clement, Foster & Kaszkik, 1998). Furthermore, brand equity facilitates price premiums (Starr & Rubinson, 1978), raises
levels of customer satisfaction (Pappu & Quester, 2006), improves the efficiency of marketing efforts (Smith & Park, 1992) and increases loyalty (Taylor, Celuch & Goodwin, 2004),
Several measures of brand equity exist, however considering a dynamic business environment, new research techniques and new information sources available, it is important to constantly review existing measures and look for improvement. This study aims to propose a new measure for brand equity and thereby remedying some of the problems that existing measures do coop with. One of the main issues with the current survey and interview based approaches is that these methods are not standardized which makes this type of measurement prone to subjectivity. Furthermore, most measures are time consuming to conduct, they do not investigate brand associations as a network and assume that brand equity is formed in
isolation.
The method proposed in this research, aims to understand brand equity, involving the network of strong, favourable and unique brand associations, by using visual user-generated content (VUGC) as source of information. In order to illustrate the measure, city marketing has been used as an application by measuring the brand equity of the brand Iamsterdam. This research has three general objectives namely (1) investigate a new measure of brand equity, (2) apply this measure by investigating the brand equity of the brand Iamsterdam and (3) provide an effective and useful way to map the outcomes of this measure.
Literature gabs and research question
In the literature review in section two, it is outlined what brand equity is and why it is valuable for companies to get insight in the brand equity of their brand(s) and the associations that consumers have towards these brand(s). Several brand equity measurements exist, but this research shows that all of these measures have one or more shortcomings. Therefore, this
research proposes a new measure which can remedy some of the problems that exist with current measures.
By studying visual user-generated content posted on Instagram, this study aims to investigate a new way of measuring brand equity. It suggests a quantitative approach, which is undertaken in a standardized, simple and quick manner. Instead of gathering new data, visual user-generated content that has already been made available by users of the social media network, is used as source of information in an unconscious manner. To illustrate the measurement, it is applied by investigating the brand equity of the brand Iamsterdam. Furthermore, this research provides an effective and useful way to map the outcomes of the measurement.
With the recent emergence of social media networks, it has become popular to indirectly leverage user-generated data on online communities. Beneficially, the resources on such social media are obtainable instantaneously and inexpensively from a large crowd of potential customers. However, no study has explored visual user-generated content as source of information when measuring customer-based brand equity. This altogether results in the following research question: “How to understand brand equity, involving the network of
strong, favourable and unique brand associations, by using user-generated content as source of information?”
Scientific and Managerial contribution
By developing a new method for measuring brand equity, applying this method to measure the brand equity of the brand Iamsterdam and finally mapping the results of this measure, the outcomes of this research have significant scientific and managerial implications.
The new method of measuring brand equity makes a contribution to the literature regarding earlier identified ways of measuring brand equity by overcoming limitations that
existing measures do coop with. Brand equity, from a consumer viewpoint, it is traditionally measured by analysing consumer responses to a survey. A contribution of this research is to introduce a novel source of data, namely that of visual user-generated content.
The ubiquitous, dynamic and real-time interaction, enabled by social media, has changed the landscape for brand management and substantially influenced the performance of brands (Gensler, Völckner, Liu-Thompkins & Wiertz, 2013). Due to the new dynamic networks of consumers, and the ease of sharing brand experience in such networks, consumers have become authors of brand stories and brand managers have lost their pivotal role as author of these stories (Gensler et al., 2013). This means that brand equity is being shaped and sculpture by social media more and more. User-generated content now shapes what a large mass of consumers think about a brand. These customer-generated brand stories are able to determine a brands image and general associations (Gensler et al., 2013). When knowing this, it is important to no longer ignore consumer-generated brand stories as source of information when investigating the value of a brand.
Rather than thinking of consumers as passive absorbers of brand information and brands as controllable knowledge structures, it is important to understand brands as a “repository of meanings for consumers to use in living their own lives” (Allen, Fournier & Miller, 2008, p. 782). Therefore, all stakeholders of the brand, including consumers, are active co-creators of brand knowledge and brand meanings. The construction of brands can thus be seen as a collective process involving several brand authors/stakeholders who all share their brand stories (Gensler et al., 2013).
Corresponding with the second motivation of Keller (1993) as stated later (“a strategy–based motivation to use the outcome as a strategy to improve marketing productivity”), marketers are continually under pressure to justify the impact of their marketing activities. Therefore, there is renewed interest in measures of marketing
performance (O’Sullivan & Abela, 2007). The current research aims to show a new way to give brand managers and marketers strategic, comprehensive and consumer-driven insight into their brand associations and associative network. By identifying the strength, favourability and uniqueness of the brand associations, guidelines are provided to (brand)managers about how to maintain a brands strong image. The measure provides information about which core associations should be protected from erosion or dilution.
Marketers can exploit this new measure for a variety of purposes. First, it can be used diagnostically to check brand meaning and the brands overall health. Second, managing the strength of key associations can also help to ensure that the brand sustains and achieves its desired position in the competitive market. Third, the measure provides scope to monitor the effectiveness of communication activities and their impact on how consumers perceive the brand. Fourth, it is suggested that insights into the consumers’ brand associations can help to determine the strength of the negative impact on a brand following a crisis situation (Dawar & Pillutla, 2000). Finally, managers can exploit the insights provided by this new measure to maintain or adjust their brand management strategy accordingly.
In order to provide a useful and attractive way to present a brand’s brand equity to marketing practitioner and other stakeholders, the outcomes of the research are visualized. This visualization includes mapping of the brand associations and the consumers’ associative network structure. Furthermore, visualizations are created for all the different dimensions of brand associations as indicated by Keller (1993). Altogether, these insights into the measurement, and improved understanding of brand associations and their network, strength, favourability and uniqueness, represents a significant contribution to brand equity literature and managerial practice.
This research is divided into eight sections. The next section is the literature review which concludes by outlining why visual user-generated content functions as good source of
information when measuring brand equity. The third section introduces the case of Iamsterdam, which is used to applicate the measure. The fourth section proposes the research design and the fifth section elaborates on the analysis and results. In the sixth section results are discussed, the seventh section provides the conclusion of the research and in the last section, limitations and suggestions for further research are given.
2. Literature review
In order to structure this literature review, it is divided in five parts. The first part of this literature review provides a comprehensive elaboration about what brand equity is and its different components. The second part elaborates on brand equity measurement and compares the different methods. The third paragraph explains the strong impact of eWOM on branding. The fourth part explains, according to different desiderate for brand equity measurement, why visual user-generated content serves as good source of information when measuring brand equity from a consumers’ perspective. The final part provides a summary of the literature review and gives a sub-conclusion of this research.
Customer-based brand equity
Brand equity is considered as a critical part of brand building (Keller, 1993) and therefore it is an important concept to outline when doing research in the field of marketing and branding. Though the terms ‘customer-based brand equity’ and ‘brand equity’ have been used interchangeably, the present research focusses on customer-based brand equity (CBBE). Measuring CBBE differs from other type of brand equity measurements since it does not investigate financial assets or a products performance in the market place. Instead, it involves investigating customers’ reactions to an element of the brand. So when measuring customer-based brand equity, managers are able to track brand equity at a customer level.
CBBE can be defined in terms of the marketing effects that are uniquely attributed to the brand (Keller, 1993). Furthermore, customer-based brand equity is the differential effect of brand knowledge on consumer response to marketing of the brand. Consequently, understanding the structure and content of brand knowledge is important because this influences what comes to a consumers’ mind when they think about a brand. The consumer response to marketing is defined in terms of consumer perceptions, preferences and behaviour arising from marketing mix activity. Keller (1993) states that building customer-based brand equity requires the creation of a familiar brand that has favourable, strong and unique brand association (see figure 1).
Figure 1: Dimensions of Brand Knowledge (Keller, 1993, P. 7)
Another important element in Keller’s (1993) study is understanding the structure of brand equity. He outlines a node based memory approach, meaning that (brand) knowledge consists of a set of nodes and links. Nodes are associations stored in memory that are connected by links that vary in strength. Consequently, the strength between the activated node and all linked nodes determine the extent of ‘spreading activation’ (Keller, 1993, p. 2),
which is a consumers’ retrieval from memory. The same associative network memory model as outlined by Keller (1993), serves as the basis for determining the consumers’ associative network according to the new method proposed in this research.
Brand equity is an important and frequently studied subject, which is not strange if one considers the consequences of the creation of a strong brand. Dacin and Smit (1994) argue for example that a brand is one of the firms most valuable assets. Also Keller (2003) listed many benefits of having a strong brand such as (increased) customer loyalty, increased marketing communication effectiveness, and being less vulnerable to marketing crises. Furthermore, the creation of a valuable brand can strongly influence a company’s bottom line result. Fehle, Fournier, Maddon and Shrider (2008) provide empirical evidence for the fact that strong brands cause its stock value to increase and therefore brand value may be a useful tool for fundamental stock analyses. Moreover, Keller (1993) outlines two general motivations for studying brand equity namely (1) a financially based motivation, to estimate the value of a brand more precisely for accounting purposes and (2) a strategy–based motivation to use the outcome as a strategy to improve marketing productivity.
According to Keller (1993), brand equity is conceptualized along two dimensions which are brand awareness and brand image. The current research mostly focusses on brand image, nonetheless, for the comprehensive understanding of band equity, both concepts are outlined in the following sections.
Brand awareness
Brand awareness can be described as the strength of the brand node in memory. It reflects the consumers’ ability to identify the brand under different conditions (Rossiter & Percy, 1987). Brand awareness of a brand name relates to the likelihood that the name will come to mind and the ease by which it does so.
According to Keller (1993), brand awareness can be sub-divided in two concepts, namely brand recognition and brand recall performances of consumers. Brand recognition can be conceptualized as ‘Have you ever heard of brand x?’. Brand recognition requires that consumer correctly discriminate the brand as having seen or heard of before. Brand recall is the consumers’ ability to retrieve the brand when the product category is given. This means that it requires of consumers to correctly generate the brand from memory.
Brand image
Brand image is the set of ideas, beliefs, and impressions that a person holds regarding an object (Kotler, 1997, p. 607). Therefore, brand image refers to the set of associations linked to the brand that consumers hold in memory (Keller, 1993). Understanding this mental picture is important because people’s attitudes and actions towards a brand are highly conditioned by this image (Kotler, 1997, p. 607; Jaffe & Nebenzahl, 2006, p. 15). Understanding and measuring brand image, as the most important determiner of brand equity, is the main focus of this research.
The key components of brand equity are the associations consumers have with the brand (Aaker, 1991; Keller, 1993). Brand associations have been called “the heart and soul of the brand” (Aaker, 1996, p. 8), and “fundamental to the understanding of customer-based
brand equity”. The central role of brand associations in the creation and maintenance of brand equity is widely accepted (e.g. Hsieh, 2004; Walvis, 2008; Wansink, 2003). Practically, high equity brands are more likely to have positive brand associations (i.e. brand image) than low equity brands (Krishnan, 1996). Brand associations serve to differentiate a brand and to create meaning for brands. These associations (both intended and unintended) give meaning to the brand and are an important component of brand equity. Therefore, a better understanding of brand associations is a fundamental role of brand managers (Till et al.,
Brand associations can be captured on different relevant dimensions. Keller (1993) distinguishes between the following dimensions of brand associations: favourability, strength and uniqueness. When brand associations are favourable it means that the consumer believes that the brand has benefits and attributes that satisfy his/her wants and needs in a way that a positive overall brand attitude is formed (Keller, 1993, p. 5). When a consumers’ brand associations are mostly favourable, they are more likely to form overall positive brand judgements. Secondly, strong brand associations can be characterized by the connection to the brand node. Strength is a function of both the quantity and the quality of the associations (Keller, 1993, p. 5). A brand association will be stronger, the deeper a consumer thinks about the association and is able to relate them to existing brand knowledge. Finally, unique brand associations are associations that are not largely shared with competing brands. Building brand equity is about positioning the brand in a way that it has unique selling points and thereby a competitive advantage over the long run (Keller, 1993, p. 6). Strong and favourable brand associations are critical for brand success, however when these associations are largely shared with competitors, the brand is not likely to gain a competitive advantage from these associations.
Measurement matters
Brand equity has become more important over the last 15 years as the key to understanding the mechanisms, objectives and net impact of the integrated impact of marketing (Reynolds & Phillips, 2005). Knowing this, it is not surprising that measures capturing brand equity, or certain aspects of brand equity, have become part of a set of marketing performance indicators (Ambler, 2003). Therefore, many academic research has been conducted on identifying measures for brand equity and brand image (e.g. Aaker, 1996; Lasser, 1995).
These studies have offered relevant insight into the processes of consumers evaluating and choosing brands within a given product category.
Academics and practitioners who gathered at an MSI (1999) workshop on brand equity metrics, summarized the following purposes for measuring brand equity: (1) to guide tactical decisions and marketing strategy, (2) to determine the extendibility of a brand, (3) to evaluate the impact of marketing decisions, (4) to track the brand’s health compared with that of competitors and over time, and (5) to assign a financial value to the brand.
In order to review the current measures of brand equity, use is made of Keller and Lehmann’s (2001) division into three different categories. The categories between which these authors distinguish are the following: (1) ‘product-market’, (2) ‘financial market’ and (3) ‘customer mind-set’ measures. In the next part of this literature review, every category of measurement is explained. Finally, a table is provided in which the strengths and shortcomings per measurement are identified.
The first category are measures related to ‘product-market’ variables. These measures conceptualize product-market-level brand equity as the incremental revenue that the brand earns over the revenue it would earn if it was sold without the brand name (Ailawadi, Lehmann & Neslin, 2002). The most common measure within this category is the price-premium measure; that is the ability of a brand to charge a higher price than a private label equivalent (Ailawadi, Lehmann & Neslin, 2003). Overall, measures in this category are simple, intuitive, and easy to calculate from public sources. Moreover, they have a strong construct and face validity. According to Ailawadi et al. (2003), product-market measures offer an attractive middle ground between financial and customer mind-set measures in terms of relevance to marketing and objectivity.
The second category of measures assesses ‘financial market’ variables. From the perspective of financial markets, brand equity is the capitalized value of the profits that results
from associating a certain brand name with particular services or products (Simon & Sullivan, 1993). These measures asses the value of a brand as a financial asset. These type of measures focus on the outcomes or net benefit that a firm derives from the equity of its brands. Measures in this category are easy and quick to conduct, however, only financial data is being considered.
Measures in the final category ‘customer-mindset’ are focusing on assessing the consumer-based sources of brand equity and investigate a consumers’ brand knowledge. These type of measures are rich in the way that they asses several sources of brand equity such as associations, attitudes and awareness. These type of measures can also be used to predict the potential of a brand. However, these type of measures are mostly based on consumer surveys. Meaning that these measures are time consuming to compute and they do not provide an objective and simple measure of brand performance. Measures in this category have been the focus of much academic research (e.g. Aaker 1991, 1996; Keller 1993, 2003).
In table 1, an overview is provided of brand equity measurements in the three different categories. Examples within all different categories are given, according to measurement methods that have been proposed in frequently cited articles. To provide a comprehensive overview, also industry measurements such as Interbrand and Millward Brown are included. Exact methodologies of the industry measurements are obviously kept confidential, but broad parameters have been published. Interbrand’s methodology is based on the net present value of the earnings the brand is expected to generate in the future (Kumar & Shah, 2015) and Millward Brown’s method ‘BrandZ’ combines consumer research with financial analyses.
Both the academic and industry measures are rated according to a list of desiderata for the ideal measure. The table provides further evidence for the idea that visual user-generated content offers an attractive way of measuring brand equity in comparison to earlier identified methods.
Table 1: Overview of brand equity measurements
The impact of E-WOM on branding
Earlier in this literature review the concept of CBBE has been outlined and the importance of this concept has been stressed. The existing (categories of) measurement methods of brand equity have been reviewed and criteria have been stated which are important when creating a new measure for brand equity. In this paragraph, the impact of electronic word-of-mouth (eWOM) is discussed and the relationship between eWOM and branding is outlined.
Recently, a strong relationship has developed between electronic word-of-mouth and branding. This because brand stories are no longer only told by the marketer but social media gave a strong voice to consumers. Consequently, it is possible for consumers to share their brand stories with peers. These stories can help to build awareness, comprehension, recognition, recall, and to provide meaning to the brand (Singh & Sonnenburg, 2012, p. 189). Firms need to pay attention to these consumer-generated brand stories to ensure a brand's success in the marketplace (Gensler et al., 2013). The brand stories that are told by consumers of a brand can determine a brand’s general associations and the image of the brand (Holt,
2003). Because of this finding, there nowadays exists a critical question for marketers and brand managers on how to successfully coordinate consumer-generated brand stories. To highlight the important relationship between brand equity and eWOM; it is found that eWOM is one of the most effective factors influencing brand image (Jalilvand & Samiei, 2012).
Peres, Shacher and Lovett (2013) further explore the relationship between branding and (e)WOM by presenting a theoretical framework whose fundamentals are consumers and what stimulates them to engage in (e)WOM. They find that brand characteristics, above and beyond its category or product type, play an important role in generating (e)WOM. They further argue that consumers spread the word on brands as a result of functional, social and emotional drivers. On the contrary, this research resonates that this link exists in the opposite direction since it is argued that eWOM properties make up CBBE.
User-generated content (UGC) has been, and will likely be, increasingly changing the way that people search, find, read, gather, share, develop, and consume (brand related) information. UGC is “the media impression created by the consumers, usually informed by relevant experiences and shared or archived online for easy access by other impressionable consumers” (Zheng & Gretzel, 2010). UGC may serve as the new form of word-of mouth (Ye, Law, Gu, & Chen, 2011). Although UGC is closely aligned and often confused with eWOM, the two differ depending on whether the content is generated by users (UCG) or the content is conveyed by users (eWOM). To be successful, eWOM depends on the spread of content, and UGC has less influence without eWOM (Cheong & Morrison, 2004, p.3).
Through the Internet, individuals can make their ideas and opinions more easily accessible to other Internet users (Dellarocas, 2003). Furthermore, the majority of consumers reported that they trusted the opinions which were posted online by other consumers (Gretzel & Yoo, 2008). One of the reason, why this research proposes that UCG is very useful as information source when determining the brand equity of a brand, is because of the strong
influential and persuasive characteristics of UGC.
When talking about user-generated content, a distinction can be made between textual (words) and visual (photographs, images and video) content. The combination of social media with the integrated mobile technology makes the capturing of activities easier and more enjoyable. Therefore, social platforms supporting visual content, such as Instagram (http://instagram.com), Tumbler (http://tumblr.com), and Pinterest (http://pinterest.com) are rising to the top of social media channels (Thomas, 2012). Since the current research focusses on Instagram, the following numbers are useful to illustrate the impact of this specific social media network: on average, 80 million pictures are posted per day and these pictures are generating on average 3.5 billion likes per day (Our story, a quick walk through our history as a company, 2016).
In conclusion, it is proposed that eWOM is becoming one of the main drivers of consumers (brand related) behaviour. This brand behaviour/purchase behaviour comes forth from the consequences of the consumers’ brand knowledge which is the consumers’ associative brand network. Therefore, in a brand related context, eWOM is becoming more relevant than ever.
Visual user-generated content as a measure for brand equity
In this paragraph it is proposed that using visual user-generated content as source of information when measuring CBBE, fills up a large gap of important desiderata that current measures do not address. The argumentation is structured according to the most important desiderata as listed in table 1. In the final conclusion of this literature review, a summary and conclusion is given about the idea that VUCG is a valuable, new source of information when measuring CBBE.
Measuring brand equity from a consumers’ perspective
Like all other measures in the category ‘consumer-mind set’, this method also aims to measure brand equity from a consumers’ perspective. Measures in the category ‘customer mind-set’ might be most appropriate when aiming to measure customer-based brand equity because these measures investigate awareness, attitudes, associations, and loyalties that customers have toward a brand. When measuring CBBE, it is believed the value of the brand resides in the minds of the customers and not in the brand itself. Product-market and financial-asset based measures are both outcome based measures, meaning that they are a result of the knowledge structure and its impact on subsequent consumer behaviour. As a consequence, these measures provide little diagnostic value. By using visual user-generated content as source of information, the proposed method is able to identify the underlying processes of consumer-based brand equity and has the ability to predict the brands potential.
Measuring associations
Brand associations can be anything which is seated in the consumers’ mind and related to a brand. All brand associations together will form the consumers’ overall brand image. To argue whether visual user-generated content is a good way to understand brand associations, it first needs to be investigated how consumers actually learn brand associations. Van Osselaer & Alba (2000) conducted a series of experiments to illustrate a learning process that enhances brand equity at the expense of quality determining attributes. They argue that when the relation between brand name and product quality is learned prior to the relationship between product attributes and quality, inhibition of the latter may occur. This phenomenon is described as ‘blocking’ of consumer learning. In other words, consumers value brand cues at the expense of quality cues. This finding provides evidence for the idea that blocking of consumer learning is a phenomenon that has implications for brand equity. The results of this
research are in line with theories that view learning as a forward-looking process. It suggests that learning associations is dominated by a satisficing process aimed at establishing accurate prediction of future consumption benefits (van Osselaer & Alba, 2000, p. 13).
It is argued that associations measures can provide diagnostics to practitioners that other (outcome) measures cannot. Only a few existing methods assess brand associations, all by using qualitative research techniques. This research aims to introduce a measure that elicits brand associations by using a standardized and quantitative method.
Measuring the consumers’ associative network structure
While actual purchase behaviour and financial assets may be used as a measure of equity, delving into consumers’ knowledge structures may provide information on a brand’s potential that may not be captured by past behaviours (Krishnan, 1996). Delving deeper into consumers’ association structures may provide new information on the brand’s weaknesses and strengths. Most of the current measures do not identify brand associations as a network, which is one of the mayor shortcomings of these measures. Measuring the consumers associative network structure is desirable since it is believed that consumers store information in memory in the form of networks (Anderson & Bower, 1973).
Moreover, an influential network model is developed by Collins and Loftus (1975). These authors provide evidence for the concept of spreading activation, meaning that when someone is reminded of a stimulus, the activation of the nodes (associations) corresponding to this stimulus occurs. Also this theory investigates the importance of analysing the consumers’ associative network structure, rather than investigating isolated brand associations.
Moreover, the hashtags used can be seen as the description and/or the categorization of the content (Nam & Kannan, 2014). Therefore, user-generated content does not only give insights into the brand associations as isolated concepts, it also allows to see relationships
between these associations and gain information about the consumers’ associative network. In this research, brand associations are analysed by the use of network analytics techniques. It is believed that these analyses yield rich information about the linkages and relationships between brand associations. A part of the analysis that is conducted determines the centrality of associations within the network. Some nodes are more connected than others. According to the spreading activation theory, the most connected nodes are the strongest since these nodes facilitate a better spreading.
Measuring the behavioural component
It is argued that the new method is capable of measuring the behavioural component of customer-based brand equity. Most methods only measure consumer attitude towards a brand (i.e. ‘What do you think of a brand?’), and thereby not measuring actual consumer behaviour. When using VUGC as a source of information, not only intentions but the actual behaviour of consumers is measured.
The measure is quick to conduct
Surveys are time consuming to conduct and therefore the proposed method aims to provide a quick alternative to elicit brand associations from a consumers’ perspective. Furthermore, strength, favourability and uniqueness of the associations is also investigated by using systemized, quantitative techniques. For most existing measures (especially in the consumer mind-set category), it will cost additional time to enlarge the investigated sample. For the proposed measure, this is not the case and the sample can be as large as all VUCG available. Instead of obtaining new information for the purpose of brand equity measurement, the proposed measure makes use of already available information. There is no need of collecting new data, which makes the measure less time consuming to conduct.
The measure is objective
Almost all measures in the ‘consumer mind-set’ category use interviews and surveys to elicit consumers’ brand associations. These techniques are sensitive to several types of bias due to the social nature of these methods. Three major types of bias that are likely to occur are interviewer bias (i.e. the interviewer can have prejudices), respondent bias (i.e. the respondent may lie) and the actual interview situation (i.e. the social setting). The new measure, as proposed in this research, is standardized and does not make use of any interview technique in order to elicit brand associations. Therefore, the outcomes will not depend on the competence of the interviewer which makes that the newly proposed is more objective than existing ‘customer mind-set’ methods, while still measuring brand equity from a consumers’ perspective.
CBBE is measured in an unconscious and un-verbalizable manner
Methods that measure brand equity in an unconscious and un-verbalizable manner all make use of financial or product-market related data and therefore do not investigate brand equity from a consumers’ perspective. However, measuring brand equity in an unconscious way is important in order to overcome social desirable responses. Therefore, the proposed method aims to measure brand equity from a consumers’ perspective in an unconscious way. The VUGC is created without preliminary knowledge of the purpose of brand equity measurement and therefore it investigates associations that are elicited by consumers in an unconscious manner.
The measure is easily applicable on a large sample size
Another advantage of the new standardized approach is that it can make use of an extremely large dataset in order to investigated brand associations. All brand related VUCG is taken into
account and used as source of information. Moreover, enlarging the sample will not cost extra time but only extent the amount of information.
The measure can measure brand equity of 'non-monetary’ brands
Many of the existing brand equity measurements make use of a monetary approach by investigating actual purchase behaviour or the company’s revenues and/or profit. It needs to be realized however, that brands can be much more than selling branded products or services at a (fixed) price. Think for example of personal brands (e.g. celebrities), place brands (destinations), NGO’s or media brands. By investigating consumers brand knowledge (structure) through VUGC, the proposed measure does not make use of any financial data and can therefore measure brand equity of nearly every brand type.
Measuring the social component
Many existing measures assumes that brand equity is formed in isolation, meaning that consumers are not socially influenced when determining their opinion about how much they value a brand. The currently proposed measure rejects on this assumption since for VUCG on Instagram it is proposed that the hashtags used and the content that is featured, reflect the social interpretation of the brand (Fu, Kannampallil, Kang & He, 2010).
The description and/or categorization of user-generated content is filtered through the lens of an individual user’s knowledge structure as well as through the lens of others’ social tags. This process of assigning particular tags to brand related content provides a social interpretation of the content. Therefore, it is argued that the hashtags and content of VUGC provide insight into a person’s beliefs and will therefore be a good representation of the favourability of consumers’ brand association and thereby of the consumers’ overall evaluation of the brand (Nam & Kannan, 2014).
Nam and Kannan (2014) state that people have motivations in two categories when engaging in social tagging, namely (1) content organization and (2) social communication. This latter motivation shows that the social tagging processes provides a social interpretation of the content which will in turn influence other users in forming their brand associations and brand related knowledge structures.
Conclusion
This literature review started by explaining the concept of brand equity and the importance it has gained as a key concept in understanding effects of marketing practice. Capturing and/or measuring brand equity have become part of a set of marketing performance indicators and therefore many measures of brand equity have been identified. An overview, including strengths and shortcomings, is shown of the most important (categories of) brand equity measurements. As a result of the increased impact of eWOM on branding and the shortcomings that exist with current brand equity measurements, a new brand equity measurement is proposed. This new brand equity measurement uses VUCG as a source of information.
Both qualitative and projective measurement tools offer opportunity in the direction of assessing brand equity. However, all these measures coop with a mayor downside since they all make use of questionnaires, surveys, and attitude scales in order to elicit consumer-based brand equity. This means that most of the time, these measures lack standardization and the outcomes depend on the competence of the interviewer. Moreover, these methods require a lot of time and resources, and are usually applied on small samples so the outcomes cannot be generalized. Furthermore, these methods are measuring consumers’ attitudes instead of consumers’ actual behaviour. Overall, attitude scales which are considered open methods, as well as all the closed methods, assume that brand image is a conscious and fully verbalizable
construct (Cian, 2011). Consequently, they do not allow for a full investigation of the brand image, which is in part un-conscious and un-verbalizable (Ballantyne, Warren, & Nobbs, 2006).
On the other hand, product-market and financial asset measurements are mostly standardized, systemized and thereby more objective. Most of these measures also make use of already available information which makes that these type of measures are less time consuming conduct. However, a mayor short coming of these type of measures is that they are all outcome measures meaning that they provide little diagnostic value from a consumers’ perspective and they do not provide valuable information regarding the brands potential.
The measure that is proposed in this research combines the strengths of all three types of measurements (i.e. consumer mind-set, product-market and financial assets) by accessing brand equity from a consumers’ perspective and using a standardized, quantitative approach.
3. The case of Iamsterdam
The new measure is applied and illustrated by measuring the brand equity of the brand Iamsterdam. In this section, an introduction about Amsterdam Marketing and their mission and vision is provided. Furthermore, it is described whether a city can be regarded as a brand and consequently how brand equity of a city(brand) is formed. Lastly, it is elaborated on the relevance of the case as illustration for applying the proposed brand equity measurement.
Amsterdam marketing
This study is undertaken in collaboration with ‘Amsterdam Marketing’, the city marketing organization of the Amsterdam metropolitan area. Their ambition is to put the Amsterdam metropolitan region on the map as one of the five most attractive metropolitan areas of Europe and therefore they aim to positively influence the city’s public image internationally
2016). Recently, Amsterdam Marketing is moving from a ‘sales’ to a ‘guide’ function, focussing on reputation management. This will enable them to emphases different aspects in the city marketing activities that benefit the balance in the city. Amsterdam marketing helps facilitating the distribution of visitors in time and space, with a focus on different type of visitors (Strategic plan 2016-2020, 2016, p. 13). Amsterdam Marketing will guide residents and tempt visitors to visit unknown neighbourhoods in the city and metropolitan area (Strategic plan 2016-2020, 2016, p. 14). Amsterdam marketing owns the motto ‘Iamsterdam’ and uses this motto to brand the city.
The city as a brand
According to the definition of Hankinson and Cowking (1993) a brand is defined as ‘a
product or service made distinctive by its positioning relative to the competition and by its personality, which comprises a unique combination of functional attributes and symbolic values’. Furthermore, these authors state that the key to successful branding is to establish a relationship between the brand and the consumer, such that there is a close fit between the consumer’s own physical and psychological needs and the brand’s functional attributes and symbolic values. Like brands, cities satisfy symbolic, functional and emotional needs (Rainisto, 2003). Another comparison between cities and brands is that the attributes that satisfy those previously mentioned needs, need to be orchestrated into the city’s unique proposition (Ashworth & Voogd, 1990). Moreover, Morgan and Pritchard (2000) state that ‘the battle for customers in the tourism industry will be fought not over price but over the hearts and minds — in essence, branding . . . will be the key to success’.
The previous mentioned definitions of branding and brands outline a strong comparison between branding versus the goals of city marketing and managing the city’s image as identified in the literature (e.g. Ashworth & Voogd, 1990, 1994; Kotler et al., 1999).
Therefore, it is stated that branding provides a good starting point for city marketing (Kotler et al., 1999) and as a consequence, cities can be considered as brands.
Brand equity of a city
In the previous paragraph, it is established that there exists a strong overlap between product/service branding and city branding. Because of this, it is accepted to consider a city as a brand, and cities or destinations as the product category. Consequently, it is stated that like it is possible to measure the brand equity of a product/service related brand, it is also possible to measure the brand equity of a city or city brand (e.g. Iamsterdam).
Initially, brand equity was conceptualized as consisting of consumers’ brand associations that include brand awareness, knowledge and image (Keller, 1993; 2003). From all these elements, a crucial role within the city marketing mix is played by image formulation and image communication (Kavaratzis, 2004). This because it is accepted that encounters with the city take place through images and perceptions. The probability of including and choosing a specific destination in the process of decision making will get higher when people have a strong positive image about the destination (Alhemoud & Armstrong, 1996). City image is consecutively the starting point for developing the city’s brand (Kavaratzis, 2004). The approach of image marketing in the field of city branding also emerges from the realisation that images can be effectively marketed while the products to which they relate remain vaguely delineated (Ashworth & Voogd, 1994).
In order to communicate a city’s image and evoke brand associations, Kavaratzis (2004) outlines three forms of communication, namely primary, secondary and tertiary communication. Primary communication relates to the communicative effects of a city’s actions (actions in four areas namely: (1) landscape (2) structure (3) infrastructure and (4) behaviour). Secondary communication is the formal, intentional communication that most
commonly takes place through well-known marketing practices like indoor and outdoor advertising, public relations, graphic design, the use of a logo etcetera (Kavaratzis, 2004, p. 68). Finally, tertiary communication refers to word-of-mouth.
Figure 2: Image communication (Kavaratzis, 2004, p. 67)
Whereas primary and secondary communication are both partially controllable by marketeers, tertiary communication is not. As can be seen in figure 2, the entire branding process, including the two controllable types of image communication, have as goal to evoke and reinforce positive tertiary communication. Frequently, tourists trust more in the images and opinions of other tourists than in those provided by companies and destinations management organizations (Tussyadiah & Fesenmaier, 2009).
Previous findings further stress the importance of word-of-mouth when determining, measuring and/or influencing a city’s image. Traditionally, destination management organizations and private sector businesses controlled the formation and dissemination of a desired destination image (Lo, McKercher, Lo, Cheung, & Law, 2011). However, through the internet, destination management organizations and industry marketing bodies are now also being advised to consider the implications of Web 2.0 and independent user-generated content created on their activities. The rapid grow of the use of social media networking creates a new
phenomenon in promoting and creating awareness on the existence of the destination (Hanan & Putit, 2013).
Relevance of the case
It is proposed that this new measure of brand equity can be used for measuring brand equity of nearly every brand in every product category as long as consumers engage in creating visual brand related content. However, in this section it is outlined why the new measure is particularly relevant in the case of destination/city branding.
Destination photography has a strong influence on potential visitors and residents of a city and therefore it is valuable to study the brand image that is created by visual user-generated content. Especially in tourism, visual content (i.e. photography) plays a dominant role because tourism is a uniquely visual experience (MacKay & Fesenmaier, 1997). Moreover, photos have the ability to ‘tell’ desired stories about a place or destination (Jenkins, 2003). The development of marketing through posting a photo on a social media network influences (new) interests to social media users to travel to the specific destination (Hanan & Putit, 2013). The advent and rise of a range of online photo and image sharing media has democratized the image creation and dissemination process of destinations and places (Lo et al., 2011). Destination management organizations must now compete with a wide range of non-commercial content, posted by users. These information providers are now felt to exert a significant influence on the tourist’s decision-making behaviour (Akehurst, 2009).
For example, Instagram, the photo sharing social media network which is the focus of this research, initially only served as a medium for online photography. Recently it evolves effectively in advertising, promotion, marketing, distributing ideas/goods and providing information services fast, precise and accurate (e.g. Doolin, Burgess & Cooper, 2002;
Sweeney, 2000). The function of the postcard as the traditional tool of marketing continuous by photo sharing on Instagram, where the traveller can easily upload pictures and share the experience on the place of interest that lead to the development of express marketing for the tourist destinations (Hanan & Putit, 2013, p. 472). The contribution of user-generated content shows an increasing tendency in shaping a destination brand (Bronner & de Hoog, 2011).
4. Research design
In this section it is first outlined how the data collection is undertaken. Thereafter, it is explained how the total dataset is reduced to a representative subset to conduct analyses. Finally, the conceptual framework of this research is proposed and subsequently, the variables of the framework are operationalized.
Data collection procedure
In order to obtain data for the research, information is extracted from the social media network Instagram. It is chosen to make use of the Social media network Instagram since this network incorporates only visual content, mostly in combination with meta-data such as a caption, (a) hashtag(s) and a location. Moreover, in contrary to for example Facebook, many users have a public profile, meaning that the content is publicly accessible and therefore available to use for the purpose of this research. When analysing the content posted on Instagram, both visual and textual information are taken into account, meaning the picture itself (i.e. what is in the picture?) and the hashtags that users use to describe and categorize the content they post.
All pictures posted on Instagram, referring to the investigated brand by being tagged with the brand name, serve as the population to make inferences about. The population that is investigated for this research are images, which are tagged using (at least) the hashtag
#iamsterdam. This implicates that the total population of the current research consists of more than 400.000 pictures.
Data mining techniques have been used to extract the needed content from Instagram. Once this gathered data is transformed into an understandable structure, visual and textual analyses are conducted in order to obtain information about the brand associations, the associative network structure of the brand and the favourability, strength and uniqueness of the identified associations. Please see table 2 below for an overview of the used methods and corresponding data collection procedures.
Analysis Method Why?
Data mining Use the Instagram API in order to extract content from Instagram and transform this data into an understandable structure
To collect visual user-generated content posted on Instagram, featuring at least the hashtag of the brand name (#iamsterdam in the case of the current research)
Concept analysis
Use a publicly available concept detection API called ‘ImageNet’
To determine which concepts are featured in the pictures
Embedding analysis
Use a tool called ‘word2Vec’ to analyse the embeddings between the identified associations
To investigate relationships between associations and to consequently determine the consumers’ associative network structure
Sentiment/ favourability analysis
Use a publicly available sentiment detection API developed by ‘SentiBank’ to determine the favourability of the visual information and ‘SentiText’ is used to determine the textual favourability
To analyse the favourability of the brand associations.
Hashtag analysis
Use a script in order to determine which hashtags have been used in the researched dataset
To determine the most frequently used hashtags in the researched dataset
Table 2: Overview of data collection procedure per method
Reducing the dataset to a subset
The entire dataset consists of more than 400.000 images, therefore, it needs to be reduced in order to obtain a subset from which analyses can be conducted, which is still representative for the complete dataset of ‘#iamsterdam-images’ and that aims to measure the behavioural component of brand equity. Furthermore, SPAM posts (e.g. advertising/promotions) need to
be reduced from the dataset, since this content is not considered as being user generated. As a consequence, the dataset is reduced based on the following two criteria:
(1) Only considering pictures posted between 01/01/2015 – 31/12/2016. The analysis incorporates pictures posted throughout one year in order to overcome a bias related to different seasons. However, since this research aims to measure the behavioural component of brand equity, only the most recently posted pictures are taken into account.
(2) SPAM posts (advertising, promotions etcetera) are deducted from the dataset. In order to do so, 2500 images have been categorized manually to indicate whether they are considered to be a ‘user-post’ or a ‘SPAM-post’. Based on this information, an algorithm is created to exclude SPAM-posts from the dataset. Approximately 2.5% of the entire data set is considered to be a SPAM-posts.
When the entire dataset of more than 400.000 pictures has been reduced according to the two criteria above, a subset containing 88.137 posts is left. This set of 88.137 posts is used for the purpose of conducting analyses for this research, as described in the next section.
Conceptual framework and operationalizing of the variables
Below, the conceptual framework is proposed. The starting point of the analysis is visual user-generated content. Both textual (hashtags) and visual (the picture) information is extracted from the content. This data is used to analyse the different parameters that
Figure 2: The conceptual framework
As mentioned throughout this research, Keller (1993; 2003) states that brand equity occurs when the consumer is familiar with the brand and holds some strong, favourable and unique associations in mind. In order to measure brand equity according to these variables, the variables are operationalized/measured according to the following descriptions:
0. Familiarity/awareness: this research proposes that everyone who is using the hashtag of the brand name (#Iamsterdam), has (good) knowledge of the brand. It is assumed that the users who post content, and simultaneously using the hashtag of the brand name, are recognizing the brand and associate it with the right product/service. Meaning that they are familiar with the investigated brand.
1. Brand associations: in order to determine the brand associations that are evoked by the investigated brand, the text and picture of the brand related content is analysed. It is investigated which hashtags consumers are using and which concepts are detected in order to see which brand associations consumer have.
associations are investigated. Associations are being processed with the purpose of grouping vectors of similar associations together in a vector space. Conducting this analysis shows which associations are grouped together and do consequently determine the consumers’ associative network structure.
2. Strength: The strength of brand associations is determined by analysing most frequently used hashtags and most frequent identified concepts. Although consumers may identify many things with a brand, it is the core brand associations that should be the focus of management efforts (John, Loken, Kim & Monga, 2006). Moreover, the strength of the associations can also be identified within the consumers’ associative network structure by investigating the associations’ centrality.
3. Favourability: In order to determine the favourability of brand associations, textual and visual sentiment analyses are conducted. The visual analysis identifies which concepts are present on the picture and connects these concepts with the corresponding sentiment. The textual analysis looks for words in the metadata that express emotions and feelings (i.e. favourability).
4. Uniqueness: The uniqueness of the brand associations is defined by comparing brand associations of the investigated brand with associations consumers have with another brand in the same product category. In this specific case, brand associations of another destination brand are investigated in order to compare brand associations and consequently find unique brand associations for the researched city brand.
5. Analysis and results
In this section, the different analyses are conducted and consequently the results of these analyses are demonstrated. Outcomes of the different analysing techniques are given in order to provide information regarding the brand equity of the brand ‘Iamsterdam. First the
descriptive analyses are being discussed and second the mapping of consumer-based brand equity is outlined.
Descriptive statistics Associations
First, as what is earlier explained as the essence of brand equity, the brand associations of the brand Iamsterdam are investigated. Brand associations need to be investigated first, in order to later determine whether they are strong, favourable and/or unique. Brand associations are investigated according to the analysis of the different hashtags used and the concept analysis.
All posts in the researched dataset featured a total number of 1.455.755 hashtags, meaning that on average users use about 16 hashtags to describe the content they post. Only 103.666 of these hashtags are unique, meaning that the group of 1.352.089 hashtags is compost of hashtags that are used at least two times. In order to determine the strongest brand associations, a top 100 is defined for the most frequently used hashtags. In total these top 100 hashtags are being used 654.632 times, explaining that 44.47% of all the hashtags used, is one of the hashtags that is in the 100 most frequently used hashtags. This finding provides evidence for the idea that every brand has strong, core association which are shared by a large group of consumers. The top 100 of most frequently used hashtags can be found in column A of appendix A. Number of posts Number of hashtags Average number of hashtags per post
Unique hashtags Non-unique hashtags
88.137 1.455.755 16 103.666 1.352.089
Table 3: Descriptive statistics hashtags analysis
In order to analyse the content of the pictures, use is made of two different types of analysis, namely ‘ImageNet’ (Deng, Li, Do, Su, Fei-Feiand, 2009) and ‘SentiBank’ (Borth, Ji,
Chen, Breuel, & Chang, 2013). First the ‘ImageNet’ analysis is discussed. According to this analysis, 15.293 ImageNet concepts, which can be present on the picture, are identified. Therefore, the outcome of this analysis is an 88.137 (pictures) x 15.293 (concepts) matrix. For all pictures, a nonnegative vector for each dimension is indicated. This value is a probabilistic estimation of the concept presence in the picture. A script is built in order to identify the most likely present concept in each picture. This results in 5289 unique concepts which are highly likely to be on at least one of the pictures. From all these 5289 concepts, it is identified which 100 concepts are detected the most. This results in a top 100 of concept associations which can be found in column A of appendix A.
A third way to identify brand associations is by making use of SentiBank, a novel visual concept detector library that is used to detect the presence of 1200 adjective-noun pairs (e.g. beautiful_flowers, disguising_food) (Borth et al., 2013). These adjective-noun pairs are used to identify the favourability of the associations in the picture. However, the nouns detected also provide an identification of the concepts present in the picture. A probability score is given for the presence of each adjective-noun pair (ANP). A script is built in order to identify the most likely present ANP. For the purpose of the association analysis, only the nouns (concepts) have been taken in account. In total the analysis provides scores for 527 different nouns. A top 100 of the most likely present nouns is provided in column A of appendix A.
Finally, it is proposed that this method is extremely valuable in identifying the associative network that is in the consumers’ mind. This network is the relationship and interaction between the different brand associations that consumers have. This network is identified according to the investigation of the relationships that exists between the different associations that are identified. Use is made of ‘Word2Vec’ software (Mikolov, Chen, Corrado, & Dean, 2013). This system can make highly accurate guesses about a word’s
meaning based on past appearances. Those guesses are used to establish the relationship between associations and consequently determine the consumers’ associative network structure. The Word2Vec analysis is conducted for the 300 most frequent associations, based on the hashtag, concept and ANP analysis. A score is indicated for the embedding between one association and another. Later in this section, an elaboration is given on the consumers’ associative network structure by providing visualizations of this structure.
Strength
In order to determine the strength of the brand associations, a frequency analysis of the concept, ANP and the hashtag analysis is undertaken. It is proposed that a brand association is strong when the same brand association is highly repetitive in both textual and visual analyses, and therefore shared by many consumers. From the analyses it results that 92% of the hashtags used, and 91% of the objects identified, is not unique. Meaning that this brand association is shared with at least one other user/consumer. When looking at the total dataset, 47% of the hashtags used is one of the hashtags that is in the top 100 most frequently used hashtags, 43% of the most likely detected concepts is in the top 100 of most likely detected concepts and 91% of the most likely detected noun is in the top 100 of detected nouns. These numbers provide evidenced for the idea that certain brand associations are widely shared with other consumers. For each association, a score for weight is indicated, based on the frequency of this node. These numbers represent the strength of the individual brand node, relatively to the other identified associations. An overview of these scores can be found in column C of appendix A.
Another way of identifying the strength of the associations is within the consumers’ associative network. By investigating the centrality of the brand nodes it is argued which
associations are strong based on their central role within the network. The centrality score of each association can be found in column I of appendix A.
Favourability
The favourability of the brand associations is investigated through the use of sentiment analysis. The analysis of sentiment is conduct both visually and textually.
The visual content analysis of the pictures is undertaken by making use of SentiBank. The advantage of using adjective-noun pairs (ANPs), is the capability to turn a neutral noun into and ANP with a sentiment. This makes concepts more easy to detect and furthermore, this method provides a useful analysis for the favourability of the concept. All the posts have been analysed which resulted in an 88.137 (number of posts) x 1.200 (number of ANPs) matrix. So each image is linked to a 1.200-dimensional vector in which each dimension shows the probability of the ANP presence in the image. A high probability score therefore means that the APN is highly likely to be on the picture, whereas a low score indicates that the ANP is unlikely to be in the picture. For each image, the ANPs with the highest score is identified and taken into account when conducting the analysis determining the top most frequently occurring ANPs. From all these ANPs detected, the top 100 of overall highest scoring ANPs is identified. Every identified ANP has a sentiment score based on the valence of the ANP (e.g. happy and awesome have a positive score, dirty and bad have a negative score). The sentiment value of each ANP is created by merging the adjective sentiment value and the noun sentiment value together. The sentiment values are ranging from -2 (negative) to + 2 (positive). The total analysis results in 88.137 sentiment scores of which 32% is negative (i.e. have a sentiment score below 0) and 68% is positive (i.e. have a sentiment score above 0). Furthermore, the sentiment score of the most frequently identified ANPs is determined. The top 100 identified ANPs have an average sentiment score of 0,75. From these top 100,
