Reaching for Customer Centricity—Wine Brand Positioning Conﬁgurations

: This study set out to uncover brand positioning conﬁgurations by presenting state-of-the-art brand management literature and applying a novel, mixed-methods approach to examine the under-researched wine industry transformation towards open innovation in branding. German winery brands were analyzed using a multimethod approach leaning on a novel netnographic methodology and multiple sources. The sample included 572 wineries from all 13 German wine regions with website text data and online review text data from each winery. The study identiﬁed nine prime words used to describe both brand identity as well as wine brand image. It revealed word–price clusters of brand identity and image. The results offer insights into communication and pricing opportunities for wine brand identity as well as image, thereby contributing to open brand innovation.


Introduction
The wine industry is currently in a transition from being producer-driven to reaching open innovation and consumer centricity. In parallel to the paradigmatic shift in focus, the digital transformation impacts business model design of the players with increasing co-creation of customers, also from a brand management perspective. Brand management is expected to grasp emerging opportunities presented by the transition and a trend for open innovation and co-creation with novel models needed for managing resulting wine brand duality between brand identity and brand image. Brand identity and brand image can only be managed by covering both perspectives simultaneously [1][2][3][4]. In addition, previous research has identified a lack of practice-oriented brand management models that are grounded in brand data metrics while moving away from financial and psychological metrics [5]. This research aims at filling the identified research gap while leaning on open innovation and exploiting the notion of co-creational brand design that takes into account netnographic artefacts created by (brand) managers and consumers.
The main difference between brand identity and brand image is that brand identity is on the message sender's side, while brand image is on the message receiver's side [6]. Distinctive brand identity indirectly strengthens the market power of a brand [7]. Understanding of how to deploy brand orientation as a strategic resource in an SME is a prerequisite and an antecedent of creating a strong brand identity [8]. On the other hand, brand image is considered to be a tool for controlling the realization of the created brand concept in action and facilitating corrective action [9]. It is therefore a market-oriented brand positioning approach [2].
Brand positioning needs to be aligned with customer needs and interests while creating demand and increasing willingness to pay [10,11]. Brand positioning can be achieved through definition of brand core, brand function, and brand personality [12]. Urde [1] proposes a similar approach with an additional element of appealing to emotions. On the other hand, Kapferer [6] classifies brand personality as one of six elements of brand core, along with brand physique, relationship, culture, popular reflections of brand customers, and customer self-image. Further approaches to brand positioning include the differentiation between the market-oriented school of positioning vs. brand-oriented school of positioning as well classifying the wide variety of brand positioning approaches found in practice (value congruency models, quantitative brand positioning models, brand value and equity models, brand strategy models, qualitative brand positioning models, and brand communication models) [2,5,13].
In contrast to research on brand positioning, wine brand positioning has not been able to establish itself as a relevant research field, lacking rigorous theoretical and statistical methods in order to be considered as a legitimate research field. Mora [14] has made a significant contribution by examining in detail 30 international case studies and by including wine brand identity, image, and personality as well as segmentation. The pioneering work of Flint et al. [15] proposed to position wine brands along four dimensions: innovative, modern, classic and traditional. Brand identity has since been deployed as integral part of brand positioning, while brand image and personality have been excluded from this research.
The research field on novel communication technologies in the wine industry has focused on interactive social media as a separate and disconnected phenomenon from one-way communication channels, such as websites. It should be noted that wine 2.0 includes all the functionalities of the wine 1.0 framework (websites, blogs, and other one-way communications) adding interactive functionalities for submitting reviews or leaving comments, which happens primarily through social media [16,17]. Therefore, it is rather surprising that no previous research has dealt with this increasingly diversified media landscape for wineries by deploying an approach that includes all the elements of the wine 2.0 (e.g., both websites as one-way communication media, as well as SM as interactive communication media). One major advantage of the approach presented in this study is that it can be deployed to analyze the wine brand through an unlimited number of Web 2.0 media by dividing between Web 1.0 media sources for band personality and wine 2.0 interactive media for brand image, thereby contributing to open innovation. The research approach thereby proves to be a useful tool and a model to respond to conceptual and managerial issues related to the increasing diversification of online communication channels in their function, type, and content [18,19]. This research approach is designed to advance the wine brand personality and image research but also contribute to wider brand research efforts in other industries. Previous research has identified the research gap on useful models for supporting open innovation in wine branding in the sense of enabling the deployment of multiple data sources for analysis of wine brands [20]. This research sets out to address this research gap.
Digital communication technology is changing brand communication. Researchers in wine marketing need to trace and analyze the changing communication and buying habits. Novel research on branding and consumer behavior in a rapidly evolving communication environment invites the deployment of novel methodological approaches. For example, text mining techniques have come a long way from being used only for books, patents, and scholarly articles to nowadays covering social media and other online content, as well as brand-related online sentiment [21][22][23][24][25]. Modern social media research builds on web scraping tools like Facepager and Netvizz [26] as well as a variety of text processing techniques, software packages, and algorithms (with statistically demanding calculations based on latent Dirichlet allocation (LDA) [24,27,28] to text mining packages for automated text mining, such as "NVivo", "QDA Miner", the "qdap" package in R and "AntCont" [23,26,29,30]). Keyword extraction algorithms, e.g., PositionRank, TextRank, and RAKE, thereby rely on predefined dictionaries of keywords (rapid automatic keyword extraction) [31,32]. However, novel approaches dealing with digital communication envi-ronments need to be methodologically robust and lean on a variety of methods offered in basic sciences, such as anthropology [33,34], ethnography [35], sociology [36,37], and philosophy [38,39]. Marketing and branding in the online communication context can thereby benefit from the variety of methods across research disciplines extending the methodological toolbox and allowing academics and practitioners to provide valuable contributions by analyzing and understanding online phenomena in commercial settings [40][41][42][43][44]. One of the most prominent approaches that integrates ethnography, technology, and social media is "netnography" [45][46][47]. It can be deployed both by basic science research as well by commercially oriented sciences, such as marketing and branding. Despite still vague methodological procedures, data validity problems, and potential data overload, the important advantages of the netnographic approach are accessibility, speed, and low cost of obtaining large quantities of data [48].
Considering the aforementioned frameworks, the research goal to create a classification of dominant brand identities and brand images of wineries served to uncover prevailing brand positioning types. Apart from deploying a novel, mixed-methods research methodology, the approach can satisfy an identified research gap, providing needed orientation for practitioners. The research has been guided by the following research questions: • RQ1: What are the major wine brand identity types from a supplier perspective regarding keywords used as well as price level? • RQ2: What are the major wine brand image types from a consumer perspective regarding keywords used as well as price level?
The introduction to this article provides a short overview of brand positioning and data mining. It moves on to netnographic research for wine communications 2.0 with a special focus on websites and social media as the two prominent communication channels. Three major research streams are identified as the basis for the applied netnographic research. The article than moves to wine branding research with an overview on major concepts and critical investigation of existing methodologies. Major wine branding concepts investigated in the literature review section are wine brand identity, wine brand image, and wine brand personality. The methodology section sets out to present methodological considerations in detail. When presenting the results of the cluster analysis, the characterizing keywords are explored in regard to brand identity and brand image of wineries. An analysis of the average price levels by cluster serves as a proxy for profitability. Indeed, the researched wineries jointly build on more premium wines, and therefore price premium can be turned into profitability. The discussion section then reflects scientific and managerial implications not neglecting limitations of the study approach. In conclusion, this research demonstrates the deployment of a netnographic approach coupled with statistical clustering for wine brand research. The approach can prove to be useful in a modern communication research environment with an abundance of data from different data sources.

Digital Communication Technologies as a Novel Research Field in Wine Branding Research
The wine branding research has been increasing in the last years, primarily spurred by wide availability of data [49,50]. However, the digital wine branding and marketing research, similarly to other digital-related research areas, lacks methodological rigor and consistency in order for methodologies to be easily replicated and interpreted. Three major research streams were identified (also presented in Appendix A) in wine netnographic research: (1) website, (2) social media, and the (3) research combining both, which appears to still be in its infancy. They present three distinct subsets of the winery-related digital communication research, with different data sources as well as different methodological approaches. The research stream dealing with winery websites is more advanced and more international [51][52][53][54]. Methodologically, this research stream relies predominantly on large sample sizes and observation of website functionality, communication effectiveness, and consumer loyalty to the website, with some attempts to use content analysis. The second research stream deals with winery social media research [55][56][57][58] and is characterized by an increasing usage of public data on wineries, online services for social media metrics, and online questionnaires. In addition, the role of new actors, such as wine influencers, has been researched along with their importance for wine brand online communications [20]. Advanced statistical techniques are increasingly applied in this research stream, spurred by evolving sample sizes and data volumes. It appears that the large abundance of data that can be obtained for analyzing social media requires specific analytic data skills, in contrast to the research stream focusing on websites. The third research stream has received less attention as it requires a unified approach to winery brand communications, exploiting all data sources, websites, as well as social media [59][60][61]. Advanced theoretical frameworks necessitate integrating these completely different data sources, demanding advanced statistical "big data" skills similar to the research stream on social media.

Wine Branding: Identity, Personality, and Image
The research on wine brands is dominated by three major methodological approaches, presented in Appendix B: the wine brand identity, wine brand image, and wine brand personality approach. The research stream on wine brand identity demonstrates high versatility, ranging from interviews to questionnaires to netnographic approaches [62,63]. Some wine brand identity literature relies on standard definitions of brand identity [6], while other authors go on to explore other, wine-specific brand identity variables [64,65]. This research stream has thereby confirmed the importance of places, both as names as well as images, in crafting strong wine brand identity [66,67]. In that sense, wine brand identity research should build upon findings from place-and tourism-based brand personality literature, where digital brand identity, culture, and place have been researched in detail [68][69][70][71]. Wine brand identity literature builds on the findings from SME and digital branding literature, where the most important brand identity components are brand values, brand vision, and brand positioning [72,73]. In addition, identity in SMEs is conceptualized as having strong ties to brand-building behavior of SMEs as well as brand orientation (consisting of brand norms, brand values, and brand artefacts) [8].
The research stream on wine brand image is rather small and lacks recent contributions. However, it involves some significant contributions and multi-study approaches [74,75]. Similarly, as with wine brand identity literature, the focus in wine brand image literature is on regional, spatially related aspects of brands. However, there are also insights that in the case of sparkling wines, brand image is an important purchasing decision and is influenced by expert reviews and recommendations, perceived reputation and prestige, as well as online reviews [76]. Apart from the research in the wine industry, brand image is often researched in food industries and fashion industries, usually in an international context [77][78][79][80][81].
Wine brand personality literature relies to a large extent on instruments developed by Aaker [82] and often requires large samples to research wine exporting countries, wine regions, as well as bottled wines as brands [83,84]. Wine brand personality research has brought about interesting insights into inclinations of green wine consumers as well as certain age groups, such as millennials [85,86]. However, the aforementioned reliance on timeworn theoretical frameworks seems to be the major weakness of this research approach, thereby possibly missing out on some important developments in the wine industry but also some newer brand personality models such as the one developed by Davies et al. [87]. Indeed, measurement of brand personality is the most active field of brand personality research, but other relevant fields, such as effects of brand personality as well as dynamics of brand personality dimensions, should not be neglected. Therefore, the research method in this study employed a bottom-up approach in order to reveal new dimensions that could be of relevance in the research of wine brands that have not been captured by general brand personality and identity frameworks. The approach therefore builds upon methodological findings of Rojas-Méndez, Hine, and Rod [84] and takes into account the current tendencies towards brand personality and anthropomorphism in the digital world [88].

Methodology
Netnography was used as the data collection method, suitable for researching digital artefacts in the online, Internet environment [89]. The quantitative data analysis focused on typology creation as a classification technique [90]. The focus of the research on classification of the online text content and identification of the relevance for the existing wine branding theories is an important intersection of the research at hand with the grounded theory. Although the grounded theory is almost exclusively qualitative in nature, it is rooted in a pragmatic-oriented school of thought and often deploys classification as a preparation for building a theory [91][92][93]. Similar, netnographic classification and interpretation of online content in wine branding has been previously conducted by Ingrassia, Altamore, Bacarella, Columba, and Chironi [20]. Previous research on branding and marketing is based on the well-established methodological framework for brand positioning regarding (A) price as well as brand core, consisting of a dichotomy between (B) brand identity (internal orientation) and (C) brand image (external, customer orientation) [1][2][3].
In the first stage of the research, a database of 885 German wineries was created where websites of all the wineries were accessed in the search for primary branding terminology. No specific software was used for data scraping, which was performed manually. In the initial stages of variable identification by extracting the wineries' most used words, MAXQDA v18 (VERBI GmbH, Berlin, Germany) was used, while for extracting the number of word occurrences per winery (by assigning binary values for top extracted words), an MS Excel sheet was deployed. The dataset was analyzed with the deployment of k-means cluster analysis in IBM SPSS v23 (IBM Corporation, New York, NY, USA), with an initial exploratory dendrogram created through the hierarchical clustering method in SPSS. This helped to determine the optimal number of clusters for k-means clustering.
Furthermore, data on bottled wine prices were extracted from a famous wine guide and an average price calculated for each winery brand. This value was used as a first clustering variable. In the next stage, text data were collected online from August to December 2019 in order to examine netnographic artefacts relating to: (a) brand identity, by collecting the text data from the "about us" section of the website; and (b) brand image, by collecting the text data from up to 10 most relevant Google reviews. This process is presented in Figure 1. After cleaning up the dataset to include only entries with no missing data, a final set of 572 wineries served the analyses. All the wineries in the final dataset were rated by the wine experts, had a website, and possessed at least one Google review with text data but usually more than five reviews. Wineries in the sample are geographically spread among all 13 German wine regions.
The word count option in MaxQDA (MaxDictio plugin tool) (VERBI GmbH, Berlin, Germany) was set to extract words with at least four characters. The algorithm for word extraction in the MaxDictio tool is able to separate words from spaces, a standard in Latin languages. In addition, identified words were lemmatized in German, given the German dataset. Lemmatization means that the algorithm returns results of the same meaning regarding their basic form, regardless of their declination or case. For example, the word "service" can take many forms in German: "die Leistung", "leiste", "leistet", "leisten", "leistete", "leistetest", "geleistet", etc. All these forms were categorized into appropriate categories. No case sensitivity was applied, and a stop-list was extended iteratively to exclude unnecessary words, leaving predominantly nouns and adjectives while excluding words like "wine" and "winery", as wine and winery brand are independent variables in this research.
words like "wine" and "winery", as wine and winery brand are independent variables in this research. Nine extracted keywords enabled a clustering on the sample of 572 entries. According to literature [94,95], the maximum number of variables for clustering can be determined through the 2 n formula, where n is a maximum number of variables for a sample size that is equal to or larger than the result of this squaring function. Besides providing robust guidance, this serves as a necessary tool for understanding the actual limitations of a clustering procedure. For clustering purposes, keywords were coded as ordinal variables for each case (winery), capturing the number of times each word shows up for each winery. The final step was to normalize the keyword variables by creating descriptive output in SPSS and saving standardized values as variables (therefore, all used variables have Z as the first letter of the name).
The clustering procedure combined inputs from hierarchical clustering and k-means clustering. As a first step, a hierarchical clustering algorithm was deployed in order to generate a dendrogram, so as to determine an optimal number of clusters for k-means clustering in the next stage. In both cases, the centroid clustering method was deployed, as well as the interval type of average linkage measurement. The centroid algorithm first computes the geometric centers of each cluster, while the distance between clusters equals the distance between the centroids [94]. Based on the information from the dendrogram for brand identity, an optimal number of clusters was therefore set to 6, while the dendrogram for brand image suggested the solution with 3 clusters. In the next step, a k-means clustering algorithm was deployed to extract the clusters, where brand identity clustering succeeded in 17 iterations, while brand image succeeded in 16 iterations, speaking for a slightly higher stability of the brand image clustering solution. As a final step, the six brand identity clusters, as well as the 3 brand image clusters, were validated by conducting one-way ANOVA to examine the price levels of the extracted clusters. This was achieved by saving cluster membership of the conducted k-means cluster analysis as a new variable and then using it as a factor for ANOVA, while the average winery price was used as a dependent variable. Both k-means clustering solutions were checked for stability through pairwise comparisons in ANOVA Bonferroni post-hoc analysis. Regarding pairwise comparison in the brand personality cluster, 96 out of 270 were not statistically significant, speaking for somewhat lower stability of the solution. The dendrogram visualizes a significantly higher stability of a solution with two clusters but obviously lacking relevant insight or meaning. Therefore, the six-cluster solution, despite the relative Nine extracted keywords enabled a clustering on the sample of 572 entries. According to literature [94,95], the maximum number of variables for clustering can be determined through the 2 n formula, where n is a maximum number of variables for a sample size that is equal to or larger than the result of this squaring function. Besides providing robust guidance, this serves as a necessary tool for understanding the actual limitations of a clustering procedure. For clustering purposes, keywords were coded as ordinal variables for each case (winery), capturing the number of times each word shows up for each winery. The final step was to normalize the keyword variables by creating descriptive output in SPSS and saving standardized values as variables (therefore, all used variables have Z as the first letter of the name).
The clustering procedure combined inputs from hierarchical clustering and k-means clustering. As a first step, a hierarchical clustering algorithm was deployed in order to generate a dendrogram, so as to determine an optimal number of clusters for k-means clustering in the next stage. In both cases, the centroid clustering method was deployed, as well as the interval type of average linkage measurement. The centroid algorithm first computes the geometric centers of each cluster, while the distance between clusters equals the distance between the centroids [94]. Based on the information from the dendrogram for brand identity, an optimal number of clusters was therefore set to 6, while the dendrogram for brand image suggested the solution with 3 clusters. In the next step, a k-means clustering algorithm was deployed to extract the clusters, where brand identity clustering succeeded in 17 iterations, while brand image succeeded in 16 iterations, speaking for a slightly higher stability of the brand image clustering solution. As a final step, the six brand identity clusters, as well as the 3 brand image clusters, were validated by conducting one-way ANOVA to examine the price levels of the extracted clusters. This was achieved by saving cluster membership of the conducted k-means cluster analysis as a new variable and then using it as a factor for ANOVA, while the average winery price was used as a dependent variable. Both k-means clustering solutions were checked for stability through pairwise comparisons in ANOVA Bonferroni post-hoc analysis. Regarding pairwise comparison in the brand personality cluster, 96 out of 270 were not statistically significant, speaking for somewhat lower stability of the solution. The dendrogram visualizes a significantly higher stability of a solution with two clusters but obviously lacking relevant insight or meaning. Therefore, the six-cluster solution, despite the relative lack of stability, was favored. Only six out of 54 pairwise brand image clusters were not statistically significant, speaking for a high stability of the identified solution.

Results
In order to enable clustering, top ranked words that could be used as relevant variables for brand identity and brand image had to be identified. This was achieved by initially extracting 15 top ranked words (see Table 1 below) for both brand identity and for brand image. The k-means clustering for both brand identity as well as for brand image, with the number of cases in each cluster, final cluster centers, as well as a bar chart of the final cluster centers, is shown in Table 2. All of the variables in the ANOVA tables were statistically significant, indicating that all dependent variables contributed to the clustering solution.

Brand Identity
The six profiles created by a six-cluster k-means solution and depicted in Tables 2 and 3 and Figure 2 present different winery brand identities in terms of attributes used by winery owners and managers to describe their winery. Brand identity profile is moderately large regarding size, with 56 wineries belonging to this cluster. It is characterized by significantly elevated use of the attribute "Riesling" as well as moderately decreased use of attributes "since" and "family" as well as "generation" and "viticulture", all pointing to an increased reliance on tradition. Brand identity profile two, with only 11 wineries, represents the smallest of the six clusters and has extremely elevated use of attribute "generation" pointing to tradition as well as highly elevated use of attributes "vineyards" and "quality", as well as moderately elevated use of attribute "Riesling". Brand identity profile three represents 77 wineries, and it is characterized by a moderately elevated use of attributes "viticulture", "since", and "vineyards" as well as slightly decreased use of attribute "generation", making it a midway or a balanced profile. Brand identity profile four is the second largest cluster, with 96 wineries belonging to this cluster. It is a cluster marked by tradition, as there is a highly elevated use of attributes "generation" and "family" as well as moderately elevated use of attributes "since" and "today". Slightly decreased use of attributes "grape" and "Riesling" are also to be noticed. Brand identity profile five is the second smallest cluster with only 25 wineries belonging to this cluster. However, it is characterized by an elevated use of all nine attributes: "quality" and "viticulture" are highly elevated, closely followed by attributes "vineyards", "today", and "since". This points to a mix of quality and tradition in creating brand identity. Brand identity profile six is the largest cluster of all six, and it is characterized by a decreased use of all nine attributes. The highly decreased use of attributes "since", "family", and "viticulture" is a most pronounced characteristic of this profile, followed closely by "vineyards" and "Riesling". This directly points to a possible use of other differentiating brand identities in this profile.
In order to validate the six clusters in terms of the average price levels of the winery, serving as a proxy for profitability, ANOVA was conducted. The results of the statistical significance test are presented in Table 4. Bearing in mind that the significance level was set to p = 0.05, while our results were below this value, with p = 0.028 and F = 2.52 with 5 df, the zero hypothesis (i.e., no statistically significant difference between the clusters regarding price levels of the wineries) is rejected. The hypothesis that there is statistically significant difference between brand personality clusters in terms of bottled wine prices can be deemed true.
As presented in Table 4, Table 5 and Figure 3, the highest price level of around 26 EUR per bottle is characteristic for cluster one, followed by clusters two and three, both being around 21 EUR. The lowest price level characterizes brand personality type 4 (around 17 EUR), while clusters five and six represent the middle range market with about 18 EUR on average. It should be noted that this middle market, represented by clusters five and six, is also the largest in terms of overall share (332 out of 572 wineries).

Brand Image
The three profiles created by a three-cluster k-means solution and depicted in Table  2, Table 6 and Figure 4 present different winery brand image positionings in terms of attributes used by customers to describe their winery experience. Brand image profile one is also the largest one, with 321 wineries belonging to this cluster. It is characterized by a decreased use of all attributes and especially "good", "pretty", and "friendly". This find-

Brand Image
The three profiles created by a three-cluster k-means solution and depicted in Table 2,  Table 6 and Figure 4 present different winery brand image positionings in terms of attributes used by customers to describe their winery experience. Brand image profile one is also the largest one, with 321 wineries belonging to this cluster. It is characterized by a decreased use of all attributes and especially "good", "pretty", and "friendly". This finding suggests that customer reviews are rather diverse in terms of keywords, making differentiation challenging. Brand image profile two is the second largest one, characterizing 190 wineries of the sample. It is characterized by a disproportionately lower use of attributes "food" and "price" when compared to all other attributes. The brand image attributes mostly used in this profile are "good", "pretty", "gladly", and "friendly". Brand image profile three is the smallest of the three, with 61 wineries belonging to this sample. This profile is dominated by a very high use of attributes "tasty" and "food" and high usage of words "good", "super", and "kind". It is worth noting that this profile has a high usage of all of the examined attributes.  In order to validate the three brand image clusters in terms of the average price set by the winery, ANOVA was conducted. The results of the statistical significance test are presented in Table 7. Bearing in mind that the significance level was set to p=0.05, while our results where above this value, with p = 0.069 and F = 2.69 with 2 df, the zero hypothesis that there is no statistically significant difference between the clusters regarding price of bottled wine in wineries is accepted. In order to validate the three brand image clusters in terms of the average price set by the winery, ANOVA was conducted. The results of the statistical significance test are presented in Table 7. Bearing in mind that the significance level was set to p = 0.05, while our results where above this value, with p = 0.069 and F = 2.69 with 2 df, the zero hypothesis that there is no statistically significant difference between the clusters regarding price of bottled wine in wineries is accepted. Although not statistically significant, there are differences regarding price levels of different brand image positionings. As presented in Table 8 and Figure 5, the highest price of around 20 EUR is characteristic for cluster one, which is also the largest one (321 out of 572 wineries), followed by cluster two with around 19 EUR. The cheapest cluster is cluster three with around 14 EUR per bottle of wine, representing at the same time the smallest cluster (61 out of 572 wineries).  By cross-examining the results on brand identity and price level, the following six brand identity profiles as well as three brand image profiles can be identified.
Brand identity profile one has the highest price of around 26 EUR and a significantly elevated use of attribute "Riesling" as well as moderately decreased use of attributes "since" and "family" as well as "generation" and "viticulture", all pointing to a decreased reliance on tradition. Brand identity profiles two and three have a high price of around 21 EUR. Profile two has an extremely elevated use of attribute "generation", pointing to tra- By cross-examining the results on brand identity and price level, the following six brand identity profiles as well as three brand image profiles can be identified.
Brand identity profile one has the highest price of around 26 EUR and a significantly elevated use of attribute "Riesling" as well as moderately decreased use of attributes "since" and "family" as well as "generation" and "viticulture", all pointing to a decreased reliance on tradition. Brand identity profiles two and three have a high price of around 21 EUR. Profile two has an extremely elevated use of attribute "generation", pointing to tradition, and an elevated use of attributes "vineyards", "quality", and "Riesling". On the other hand, brand identity profile three is characterized by a moderately elevated use of attributes "viticulture", "since", and "vineyards" as well as slightly decreased use of attribute "generation", making it a midway or a rather balanced profile. Brand identity profile four has the lowest price of all six clusters with around 17 EUR, and it is marked by an elevated use of attributes "generation" and "family" as well as moderately elevated use of attributes "since" and "today". Slightly decreased use of attributes "grape" and "Riesling" are also to be noticed, pointing to positioning through tradition. Brand identity profiles five and six have a similar price of just below 18 EUR. Brand identity profile five is characterized by an elevated use of all nine attributes: "quality" and "viticulture" are highly elevated, closely followed by attributes "vineyards", "today", and "since". This point to a mix of quality and tradition in creating brand identity. Brand identity profile six is the largest cluster of all six, and it is characterized by a decreased use of all nine attributes. The highly decreased use of attributes "since", "family", and "viticulture" is a most pronounced characteristic of this profile, followed closely by "vineyards" and "Riesling". This directly points to a lack of clear differentiation in this "middle market" segment.
Brand image profile one is the dominant one and has a price premium of up to 20 EUR. It shows a decreased use of all attributes and especially "good", "pretty", and "friendly". Brand image profile two, with a slightly more moderate price of around 19 EUR, is characterized by a disproportionately lower use of attributes "food" and "price" when compared to all other attributes. The brand image attributes mostly used in this profile are "good", "pretty", "gladly", and "friendly". Brand image profile three is the smallest of the three and has the lowest price of around 14 EUR. It is dominated by a very high use of attributes "tasty" and "food", with high usage of words "good", "super", and "kind".

Discussion: Uncovering Brand Positioning and Open Innovation
This research proposes a novel, unified, qualitative research approach, which was applied to a multi-source data collection from websites, social media information, and thirdparty evaluations. The resulting combination of a qualitative and quantitative methodology as well as the reliance on different data sources takes advantage of data availability in the digital space to move wine branding research to open brand innovation, where brand identity and brand image are deployed in an open and mutually beneficial manner. The approach thereby provides a promising future avenue of research for combining an unlimited number of "netnographic" textual artefacts (social media outlets, websites, forums) and analog data sources. The methodology partially builds on the approach developed by previous research [96,97], where lexical analysis has been deployed to research netnographic artefacts. The present study contributes to fine-tuning wine branding vocabulary regarding the top ranked words for both wine businesses as well as wine consumers and creating powerful brand positioning, which takes into account brand identity and brand image at the same time. In this sense the study closes the gap identified by [98] on differing vocabulary and communication styles by the wine industry and wine consumers.
Regarding the wine price, no previous study has attempted to research the mutual interaction between winery brand identity and brand image traits on the one hand and winery portfolio pricing policy on the other. Wine price has previously been researched from a variety of other perspectives: the impact of third-party opinion on price [99], the impact of weather and bundles of characteristics on wine quality and price [100][101][102], identification of price market segments and their boundaries [103][104][105], as well as price elasticity of bottled quality wine [106].
Understanding brand image in social media is a crucial step in enhancing a brand identity. This research presents a novel methodology for co-creating brand identity with the inclusion of brand image, as an open innovation approach. In this sense, previous research has confirmed the importance of brand image as a mediating variable in the creation of brand attachment and brand preference [107,108]. Social media content can also be a source of data related to brand image for making design-related decisions for existing as well as new offers. In design-oriented new product development (NPD), brand image seems to be playing a crucial role in structuring design language [109][110][111]. Searching online consumer content and related discussions is a powerful tool for the identification of novel ideas and responding to potential problems in brand image [112,113]. The process of branding therefore goes far beyond the consumption value of the offer to include various augmented aspects of consumption, such as emotional, psychological, cultural, and social aspects of consumption [114][115][116][117][118]. These aspects can be uncovered by the methodology deployed in this article, which is deployable both in general social media posts, as well as for brand community members as a they are a rich source of information. Previous literature has confirmed that brand community members willingly participate in collective activities and help create brand identity, but this participation is dependent on trustworthiness and empathy of the brand identity in question [119,120].
The relation between the brands and the society is that of a mutual influence. Social media have especially intensified this two-way communication between SME brands and consumers, as the cost for this communication has decreased significantly [121]. Brands have also been shown to have the power to create new, previously non-existent global markets, which makes them an crucial actor in open innovation [122]. This quality of brands has traditionally been an under-researched field. It would be important for the future research to examine the effects that powerful territorial wine brands can have in creating new regional wine tourism markets as the most common extension of wine business. In this sense, creation of tourism brands has been identified as the way forward for the development of agriculture [123].

Implications
The netnographic research approach deployed proved instrumental to research phenomena related to wine communication 2.0 and modern wine branding. Content analysis of online texts allowed clustering of brand identity and brand image, coupled with validation of clusters regarding winery prices. This is a novel research approach in wine brand positioning research. The identified gap between predominantly fact-based supplier communication (e.g., "viticulture", "grape", "quality", and "vineyards") and emotionallydriven brand image from the consumer side (e.g., "pretty", "friendly", "kind", "tasty") indicates that the transition from producer focus to customer centricity is an ongoing issue. This is consistent with the previous findings on customer-centric offer design in the wine business, where wine offers are created based on a multitude of possible options [124]. In addition, brand practitioners can profit from matching their pricing strategy, brand identity, and brand image. Building a winery brand identity predominantly on "quality", "generation", and "today" limits the consumer willingness to pay for a winery's wines. As such, brand management needs to address the customer perspective by creating powerful emotional utility with corresponding brand image attributes.
The presented technique of clustering most used words from a company brand's communication content together with a brand's price category represents a novel technique for data mining and semantic clustering of online content. Semantic mapping is increasingly used to map social media debates by mapping words, co-words, and their wider contexts [25,125,126]. The technique deployed in the present research graphically represents a semantic-marketing categorization of brand positioning configurations. This typology of word-price combinations can enable brand managers to better position their brand inside certain groups by emphasizing unique words, not common to certain brand positioning configurations.
This study paves the ground for future studies by creating a comprehensive dictionary of wine brand identity as well as wine brand image identified for the German winery landscape, also providing the most relevant word combinations in different price categories. Previous literature (presented in Appendix B) deals predominantly with partial issues of wine brand personality and image, with no exhaustive lists of positioning strategies. The present study closes this research gap and provides the basis for extensive future research dealing with wine brand identity and image.

Limitations and Future Research Directions
Bearing in mind the clear distinction between the content created by the marketer and content created by the consumer inside the Web 2.0 brand communications, there is a need for future research to more clearly define the two perspectives. This research contributes to closing the gap in the brand positioning literature regarding the intersection between inside-out and outside-in perspectives of brand management and creating usable models spanning the boundaries of these two approaches inside open innovation. In addition, each winery brand should consider so-called point-of-difference positioning and point-of-parity positioning aspects identified by Keller in relation to the six identified brand identities as well as three brand image positionings [127]. Future research should also deploy larger samples in order to enable deployment of a larger number of words as clustering variables, thereby allowing for identification of additional brand positioning strategies.
Further brand personality research fields of relevance for wine research are brand personality type and brand extensions, the effects of iconic brand colors on brand personality judgements, impulsiveness of brand personalities, as well as brand personality as a basis for brand forgiveness after brand failure [128][129][130][131]. Since strong wine brand identity depends on spatial aspects of branding [66,67], it would be of interest to expand the research approach to aggregate text analysis of vintners and other wine-and winetourism-related stakeholders in understanding spatial aspects of (territorial) wine brand development. Pioneering work in this field has dealt with netnographic analysis of Tri-pAdvisor reviews of wine tour organizers [132]. The methodology used in this article could be deployed to research the brand identity, brand image, and price premium in famous and geographically protected areas versus regular wine regions and those with no geographical protection. Future research on branding should include a multitude of social media sources, as well as some qualitative criteria for evaluating websites for millennials. It has been demonstrated that brand loyalty, as well as recognition of winery websites, is generally very low among millennials, whereas lively, novel, and interactive websites seem to catch their attention [59,133,134]. Enriching the research on online debates, communities and influencers with further wine marketing cases present a promising future field of research [20,125,135].
This study has several limitations. The results are not representative for the industry since the data sources (websites and wine guides) do not justify a representative dataset. Moreover, the set of wineries includes mostly the high-quality and high-price part of the market. In addition, only one country has been analyzed, while the methodology is still in its infancy. The sample membership was determined by using different sources, digital and analog information, and a mix of qualitative and quantitative variables. In the process of cleaning up the dataset, as many as 133 wineries of an initial sample of more than 800 wineries had to be excluded from the final sample. In addition, one major limitation of the k-means clustering approach was the maximum number of variables in relation to the sample size and directly influencing cluster stability. This leaves much space for further research with different methods and approaches. By focusing on a netnographic and branding approach, no social, demographic, or economic factors were taken into account in this study. Furthermore, the generalizability is limited because the wine industry is characterized by high emotional value of their products. The study also ran into a few linguistical limitations due to the different interpretation of certain words in German as opposed to English. Some notable examples include words "since" and "generation".
While "since" can be both a preposition and a conjunction, its German counterpart "seit" can only be used as a preposition referring to a time period. It is typically used to denote a time of firm establishment, thereby clearly referring to company tradition and history. While "generation" can signify both a peer group as well as the creation of something, its German counterpart "Generation" has the sole meaning of a peer group.

Conflicts of Interest:
The authors declare no conflict of interest. Questionnaire with a combination of quantitative and qualitative data collection

Appendix A
The effect of regional brand image on consumer's quality perceptions 570 engaged wine consumers [76] Focus groups and thematic analysis Purchasing preferences, personal taste, brand image, reputation, and symbolism 56 sparkling wine consumers Wine brand personality [85] Application of Aaker [82] brand personality framework to access the popular wine brands Brand awareness and brand personality Convenience sample of 330 green wine consumers [86] Descriptive statistics, ANOVA of quantitative data, number of mentions of personality traits Wine label design and wine brand personality inclinations 404 millennial wine consumers [83] Experiment with questionnaire, MANOVA for examining personality dimension scores Wine personality image formation through interaction of wine personality, country personality, expected price, and willingness to engage 695 wine consumers [143] Scraping and text analysis of entire website material according to Aaker [82] brand personality framework Brand personality dimensions 48 wineries from five international wine regions famous for wine tourism [84] Data collection through questionnaires and correspondence analysis with symmetric plots Brand personality traits of largest wine exporting countries 757 respondents from 22 countries