Promotion of urban tourism: insights into user engagement on social media

Interactions between tourism and social networks are among the most notable phenomena of recent times, generating new approaches, in terms of both analyses of and policies for tourism promotion. Public authorities have been forced to become involved in these new realities, adapting their promotion channels to tourists’ new behaviour patterns and carefully cultivating interactions with them. It is becoming ever more important to create and transmit an image capable of stimulating high levels of engagement. This article analyses the role of one of the most booming social networks, Instagram, applied to the case of Berlin, a leading tourist city. All posts generated over the course of a year on the German capital’s official Instagram account were encoded, and the characteristics of those that generated the most interaction with users in the form of likes and comments were analysed. Our study reveals that posts more directly intended as advertising generate more negative results, while there are differences between the elements capable of generating more likes and more comments, respectively: likes are more general in nature, while comments are more specifically linked to the Berlin brand. These findings suggest important conclusions for the more efficient development of strategies to promote interaction with users.


Introduction
The social networks have revolutionised the way in which the tourism industry communicates with tourists, as well as how tourists seek, obtain, share and experience their trips (Xiang and Gretzel 2010;Barbe et al. 2020). And, needless to say, the public authorities have also been forced to engage with these new realities, adapting their promotion channels and methods to people's new patterns of behaviour (Dogra and Kale 2020;Yi et al. 2020;Iglesias-Sánchez et al. 2020). Since public authorities increasingly use social media as a way of generating engagement with potential tourists, it seems necessary to analyse how (and how effectively) digital platforms are being used, Wozniak et al. (2017), is an interesting approach to the return on tourism organizations' social media investments. That is why this study aims to determine the elements that make posts attractive, a useful quality for generating interaction among followers of cities' official Instagram accounts (Ge and Gretzel 2018). From the perspective of consumers, information prior to tourists' travel and sightseeing is increasingly available on the Internet and social networks. Both during and after the experience, the use of social media-posting photos, comments, etc.-has become a powerful dimension for generating satisfaction/experience, and makes relevant information and advice available to other potential users (Minazzi 2015;Huertas and Marine-Roig 2016;Nilashi et al. 2018;Kim et al. 2019).
The main research question that this paper addresses concerns how to quantify these user responses in the form of interactions that reflect, in one way or another, that posts have captured or attracted their attention. This question is considered in greater detail than is the case in the prior literature, as we focus on statistical analysis of the interactions generated by all posts published over the course of a year on the official Instagram account of the city of Berlin. We should immediately add the caveat that the capacity to generate interactions does not guarantee that these necessarily translate into more sales. However, they can be considered as a significant initial barometer of their ability to "arouse interest". Further clarification and evaluation is needed of the effectiveness of different strategies for the selection of content, with future implications regarding how the powers that be should go about shaping the image and personality of the tourist destination. Besides a more in-depth study of the different types of content in posts, our paper also explores two additional dimensions. Firstly, the use of posts as explicitly advertising-related elements (not just providing information about the city or promoting its image). And, secondly, an aspect that is acquiring ever greater importance on social networks: the role of interactions with users. We also consider it necessary to deepen in this aspect the role of interactions both through the prism of a more traditional indicator (the capacity of posts containing images made by other users to generate more responses) and a more novel approach: whether the number of followers of users whose posts are reused on the official account has any effect or not on the interactions generated, what we call the bandwagon effect. The strategies for "selling" the city's tourist attractions and communication strategies aimed at optimising interconnections with social network users as a whole, need to be clarified by deepening the role of content, advertising interactions in the communication strategies of the official accounts. An integral approach, based on statistical analysis of all posts published on the official Instagram account of the city of Berlin in 2018 offers new ways for evaluating the effectiveness and impact of this key channel of communication for tourist attractions, with relevant findings for profiling and improving their effects.
Of all the social networks, this study focuses on Instagram, because this is a platform that is acquiring more and more followers every day and has a very high growth rate. In 2018, nearly 714 million users accessed the platform monthly, and this figure is expected to reach over 988 million by 2023 (Statista 2020). What is more, Instagram is a social network where image plays a central role, and image is also a key, central element in the tourist trade (Paül i Agustí 2018; Balomenou and Garrod 2019;Huerta-Álvarez et al. 2020). In a 2018 WeSwap (2018) survey of 2000 people between the ages of 18 and 34, some 37% of respondents stated that their choice of holiday destination was influenced by social media, and 31% claimed that posting holiday photos on Instagram was just as important as the holiday itself.
The study analyses engagement levels generated on the official Instagram account of the city of Berlin (@visitBerlin). The German capital occupies a prominent position in rankings of the most visited destinations (Lonely Planet 2020), and there were 13,486,202 arrivals at tourist establishments in the city in 2018 (Knoema 2019). Ease of access to city destinations provided by low-cost airlines, improved infrastructure and the provision of attractive products and experiences for visitors has generated exponential growth in visits to cities (Fyall and Garrod 2019), including, of course, Berlin.
The initial choice of Berlin for a visit is explained by the attraction that this destination has exercised in recent decades for a variety of reasons, among which emotional and symbolic motivations occupy an important place. The changes that the city has seen since 1989 have given it particular significance and appeal, and a personality that combines specific features of recent changes with the status of a traditional European capital. Berlin occupies a leading position in all the main rankings of European cities' attraction to tourists and can well be considered one of the cities with the highest communication capacity (de Rosa et al. 2019).
This paper is organised as follows: the Literature Review section examines the existing literature, studying both cities' e-branding and user engagement in the tourism industry; the second section describes the research procedure and the hypotheses posited; the third section provides details on the method used to collect and encode the data and the model developed to analyse it; the fourth section contains a discussion of the hypotheses put forward; and, finally, the paper closes with conclusions, theoretical and practical implications, and proposals for new lines of research.

Cities' e-branding as a tool for tourism promotion
The meteoric rise of social media and its many implications for tourism is so obvious that it hardly requires further explanation.  document both the use of social networks and an increased focus on this phenomenon in research activities. The dimensions are multiple from the perspective of users, and obtaining information before choosing a tourist destination has become an essential activity, as has taking the opportunity provided by trips to narrate experiences in text and graphic messages on the social networks (Kiráľová and Pavlíčeka 2015;Klostermann et al. 2018;Capolupo et al. 2020). Studies have focused on the key role played by user visions in certain areas of tourism management and planning (Balomenou and Garrod 2014). From the perspective of tourist destination managers, creating an image necessarily involves the use of social networks. This image has become an integral and influential part of travellers' decision-making processes (Choi et al. 2007) or, to cite Vinyals-Mirabent et al. (2019, one of the ways that destinations "talk" about themselves. That is why destination marketing organisations (DMOs) should make these communication and promotion tools a top priority.
The need to optimise the use of social networks has become more pronounced as competition among tourist destinations intensifies. Mariani et al. (2018) note that this is especially important for a European context in which, despite maintaining a leading position in the statistics, we are beginning to note that the powerful emergence of alternatives has globalised competition in the sector.
The theoretical rationale that establishes the importance of these mechanisms combines several elements. At a more general level, in a circular and dialoguebased process, these include analysis of both the creation of a brand image with specific, differentiating elements of personality or identity, and of interactions between brand identity generated by those responsible for promoting each destination and the image perceived by users/customers (de Rosa et al. 2019). At a more specific level, research into these areas focuses on analysing content created by both DMOs and users and, particularly, content generated with some type of interaction between the two.
Indeed, consumer-to-consumer communication in tourism is becoming more and more widespread, while traditional sources of information, including DMOs and the media, are declining in popularity (Hays et al. 2013). Although creating the image of destinations has always depended on a combination of organic and seller-induced images (Gunn 1972), the joint creation of destination images is now occurring on a much larger scale (Lim et al. 2012).
Studies which are now considered classics, such as that by Choi et al. (2007), focus on both the textual and visual ways in which users characterise their tourist experiences, highlighting the similarity between the two formats: Stepchenkova and Zhan (2013), Michaelidou et al. (2013) and Mak (2017) make interesting comparisons between tourists' posts on social networks and the images of the 1 3 Promotion of urban tourism: insights into user engagement… official accounts, and their implications. As explained below, the present practice reposting users' photos on official accounts establishes a "bridge" between both sources (user and official). To cite just one additional study referring to each social network, Giglio et al. (2019) Acuti et al. (2018) centre their work on Instagram. Our analysis focuses on Instagram due to the fact that this social network is being used increasingly by the official tourism agencies of countries, regions and cities. Other aspects of interest include the prominent role played by interactions between DMOs and user-generated content (UGC). UGC is sometimes used by public authorities in their own posts, giving an added dimension to followers' comments and likes.
Similarly, there has been a growth in the number of detailed analyses of interactions. Here, a key element is the description of types of content included in each message or post. Taking into account the various dimensions that may influence users' decisions to visit (or not) a destination, the literature emphasises both the role of symbolic factors, which help to shape the personality or uniqueness of a place, including explicit references to their authenticity (Michaelidou et al. 2013;Kim et al. 2019;Chen et al. 2020), with others of a more functional or practical nature. Listings of traditional elements applied to cities include those formulated by Kladou and Mavragani (2016)

Consumer engagement with tourism social network accounts
The capacity to generate consumer engagement is, of course, another key area of research. This concept is examined in such diverse fields as psychology (i.e., task engagement), organisational behaviour (i.e., employee engagement), sociology (i.e., civic engagement) and marketing (i.e., customer engagement; Ahn (2019) and Taheri et al. (2014), among others, have pointed out, the dynamics of consumer engagement in specific environments such as tourism is a field of research that remains very much open. Certain studies measure consumer engagement by numbers of likes, comments and (on certain social networks) shares. However, many of the references mentioned in the previous section highlight the visual dimension. Vivek et al. (2014) define consumer engagement as a multidimensional construct composed of three dimensions-cognitive, emotional and behavioural-while others prefer to see it as one-dimensional or two-dimensional. Table 1 shows the dimensions used by various authors to describe consumer engagement, and helps to identify the common characteristics and most significant differences in the way these dimensions are presented.
Our paper focuses on Instagram, a social network that highlights, precisely, the visual dimension. The aim of the study is to determine, apart from aggregate numbers of interactions (likes and comments), those elements whose presence in posts  (2015) Likes, comments, shares, interaction,

3
Promotion of urban tourism: insights into user engagement… most induce such interactions, establishing a statistically significantly difference for the presence of each element compared to its absence. Accordingly, an initial specific dimension in this paper is that it combines the line of study described in the previous paragraph (classification of the attributes of each post) with the line described in this paragraph (generation of interactions). Additionally, since comments and likes indicate different degrees of engagement, the two categories are differentiated throughout the entire paper.

Development of hypotheses
Multimedia content is important on all social networks but, obviously, much more so on Instagram, because the image is the main element shared on this platform, making it a decisive element in generating engagement with users. Consequently, developing content on Instagram that can connect with the motivations that persuade Instagram users to follow an account is a key consideration when the goal is to increase both numbers of followers and online impact (Barbe et al. 2020;Rietveld et al. 2020). Providing useful, pleasant, enjoyable content that is perceived as entertaining can be crucial to generating engagement. To analyse the impact on engagement of the image content in this study, thirteen such elements were defined (see Table 2) based on our adaptation of earlier studies by Choi et al. (2007), Stepchenkova and Zhan (2013) and Mak (2017).
To evaluate the extent to which the selection of these characteristics generates (or not) more interactions, Table 2 includes a selection of elements that takes into account many accredited previous studies and which we have used to categorise the posts on the official account: both originals created by the account itself, as well as those "borrowed" from private users to be reproduced on the official account.
Our work applies the analysis linked to the studies cited above to the case of Berlin, and from them we can formulate the hypotheses that we are going to test. The series of hypotheses proposed enable us to describe relevant aspects that have potential effects, thereby helping to guide strategies for presence on the social networks implemented by tourist destinations. Figure 1 summarises the proposed research model.
Analysis of the content of posts enables us to formulate the following hypothesis: H1. Different image content leads to different levels of engagement.
Another important aspect of posts on the official account of a DMO is the introduction of explicit advertising elements, including publicity for specific establishments in the city, directly and explicitly for commercial purposes (see Table 3). The more explicit presence of advertising elements has generated a discussion about whether it was appreciated by users or whether it could, on the contrary, provoke a certain rejection (Johnson et al. 2019 Promotion of urban tourism: insights into user engagement… H2. Non-promotional posts generate more engagement than those which contain explicit advertising.
DMOs can use various formulas to try to elicit or induce more engagement from users by explicitly asking for feedback or by including images of users in posts on official accounts as a form of "proximity" to their audience. These could be called "Engagement Tactics". Table 4 presents two features of the posts along these lines: on the one hand, those posts on the official account that explicitly ask for feedback from users and, on the other hand, those posts in which the official account reproduces an item created by a user.
Our third hypothesis about the two mentioned dimensions of "Engagement tactics" is bifurcated therefore, in two (sub)hypotheses.
H3a. Explicit feedback requests lead to higher engagement. H3b. Reposting posts of consumers leads to higher engagement.
Regarding the role of the presence of users' posts, some of the literature has focused on how influencers could play a leadership role over other users (Casaló et al. 2020; Femenia-Serra and Gretzel 2020). Our analysis of reuse of images generated by private users refers to all users, as a starting point for examining a wider bandwagon effect, to assess whether or not the number of followers of the user whose item has been reproduced influences (or not) the generation of more engagement (in the form of likes or comments) when these items are used in the official account.
This gives rise to a new (sub)hypothesis concerning this potential bandwagon dimension of the "Engagement tactics" approach:

Methodology
For the purpose of our research, the total number (279)  In this paper, we will focus on the two simplest interactions that Instagram permits: numbers of likes and comments generated by the different posts. In the study of social networks, likes and comments are the dimensions most commonly used to measure levels of user engagement by observing the numbers of each (Lee et al. 2018). Although Instagram also allows other interactions (shares, etc.), in this paper we will focus on the two most basic interactions as the starting point for meeting the objectives of identifying aspects of posts that generate differences in them. This approach will allow us to extend our analysis to other social networks in future research work. These two dimensions have complementary dimensions, as likes reflect a more immediate response, expressing initial sensations, whereas comments imply a somewhat greater degree of elaboration and engagement (John et al. 2017).
As mentioned in the previous section, the posts were classified according to three non-exclusive dimensions: Content, Advertising, and Tactics of Engagement The number of reactions that posts generated in the form of likes and comments was also counted. In the latter case, we should note that comments were counted numerically.
The encoding was conducted by a group of three people who did not form part of the research team. They were first trained for the task, and the reliability of the encoded data was subsequently verified by measuring the Scott's Pi and Cohen's Kappa coefficients, which compare the encoded data with the values that would be obtained randomly. In the following tables, this result is identified by + or − signs, and three levels of statistical significance are distinguished concerning the capacity of the presence of a variable to generate differences with respect to the posts in which that characteristic was absent. Although it is more common to use the means, among other things to analyse a t test, the tables present the medians to avoid distortions produced in those cases in which a minority of posts generate special impact, which can affect the mean much more than the median. To assess the impact of each of the features, data were obtained on the mean, median and interquartile ranges generated in the number of comments and likes by the presence or absence of each of the elements mentioned and described above. Both a bivariate analysis and a multivariate analysis were carried out to identify the most relevant statistical significance. The three levels were: the most demanding, with 1 3 Promotion of urban tourism: insights into user engagement… a value of p < 0.0001 (designated by ***p); an intermediate level with a value of p < 0.01 (represented as **p); and a third, weaker level, with a value of p < 0.05 (denoted as *p). A marginally significant value p < 0.1 is also showed.
As there is no standard for the number of cases to be encoded by applying intercoder reliability, this was applied to 50 analysis units, resulting, in all cases, in values of over 0.80, and in most cases a value of 1. However, this was the result expected in all the descriptive categories, with little or no subjective need for evaluation. Figure 2 provides an example of a post that was classified among the support variables as a "photo" and in the content category as "civil architecture", "public way", "nightlife" and "most visited".

Results and analysis
To ensure the maximum reliability of the conclusions, the data was initially processed by bivariate analysis of the impact that each variable generated in comments and likes. The data was then processed again by multivariate analysis to obtain maximum information from the encoded data. In order to evaluate the first hypothesis H1, about the impact of the image content, Table 5 shows the results from the bivariate analysis for each of the features included in the category of "content", as well as indicating the p-value for both comments and likes, that is, whether the presence of each of the elements in image contents makes significant differences and, if so, whether in doing so it generates more or fewer interactions (comments or likes).
Panoramic views, parks and gardens and public ways generate positive impact on likes. Civil and religious architecture and tourism facilities and infrastructure information also generate positive impact. On the other hand, significant negative differences (fewer likes) are found in those containing only information and, perhaps surprisingly, those featuring museums and other cultural content.
As for comments, the results are somewhat less clear, with similarly positive impacts generated by other categories that received more likes, such as nightlife, requesting user feedback and most visited places. Alternative art also generates a significant number of comments (significantly more than likes, comparatively speaking). This is an aspect of the bivariate analysis that is reinforced by the multivariate analysis.
Another aspect worth mentioning is that some elements show differences according to whether measurement of their impact is based on the median or the average. An example is "parks and gardens". This is another variable with insufficient aggregate significance to make a difference, one that generates more likes on average when present than when absent (5700 compared to 5602). However, the average number of likes is lower when the category is present compared to  Promotion of urban tourism: insights into user engagement… when it is absent (5052 compared to 5430). The presence of a limited number of posts with significant impact, which raises the average higher than the median, is an explanation that argues for research, in such cases, into the specificities of these particularly successful posts. In order to evaluate the H2, Table 6 presents the results associated with the advertising elements, which include more explicit elements for the direct advertising of commercial establishments, mainly restaurants. We should note that, although the initial list of elements to be studied included "shopping centres" and "hotel advertising", this type of content was not found in Berlin's institutional account.
One point worth highlighting is what we might call the "penalty" for advertising uses, particularly as regards private activities such as restaurants and other establishments. Although, needless to say, the effect of generating substantially fewer likes (and also, though less significantly, a smaller number of comments) does not necessarily mean that it is not effective as a way of attracting a subgroup of users as customers, it does question the use of the city's institutional account for private-sector purposes. This issue should be treated with caution, since similar results in terms of generating significantly fewer likes also occurs in the case of "Museums, Exhibitions, Theatres", with a cultural dimension largely linked to dimensions of "public interest". This may (still) be a more minority interest, but it is not considered less worthy of promoting for all that (in fact, perhaps, quite the opposite). The positive response (in terms of likes) to posts in which widely used public services, such as public transport, are present, also points to this potential impact of public services.
We should also highlight the dimensions that are most explicitly linked to interactions via the social network. As previously noted, posts containing users' photos (58 of the total 279 during year 2018) and those that request feedback obtain, overall, very good results in terms of engagement. Given that Instagram users do not form a particularly heterogeneous group, some having many followers while others have fewer, the question raised is whether using the photos of users with more followers on the network can attract more engagement than others. In other words: do photos used on the official network generate more comments/likes if they are reposted from users with more followers? In order to evaluate H3a and H3b and adopting the same format as the previous tables, Table 7 summarises the results from responses linked to these dimensions of Engagement tactics.
Particularly noteworthy here is the positive and highly significant impact of the inclusion of users' photos in generating both comments and likes, and positive effects (although with a lower level of significance to "make a difference") generated by posts that request feedback. These results reflect the positive dimension of user participation.
It is also worth mentioning the important effect in terms of generating likes and comments of using photographs taken by users (and not from official sources) and the good results achieved by posts that invite feedback. This contrasts with the lower interactions generated by "mere information" posts, probably due to their "one-way" nature.
In order to evaluate the H3c, Table 8 shows the correlations between the interactions generated by users' photos and the followers of each of the authors of these photos posted on the official account. The first line shows Spearman Correlation Coefficients between the number of followers of the account from which the official account borrows a post and the engagement generated; the second line: indicator of statistical significance that the probability of the number of followers of the original account affects engagement; and the third line: post number of the official account used by private user's post.
The evidence shows, overall, weak correlations in terms of likes and even negative correlations as regards comments.
This would appear to show that the intrinsic merit of the photo outweighs its author's popularity on the social network. However, it could be argued that, although the official account has posted the photos of users with large numbers of followers, the presence of superstars/influencers on social networks is not detected, as these probably exert their much-sought-after influence through other channels (Femenia-Serra and Gretzel 2020). In any case, the number of followers of users whose photos are reposted on the official account neither created a bandwagon effect nor generated more engagement. Figure 3 shows the weak positive correlation (for likes) and the negative correlation (for comments) between the number of followers of a user whose image is published (ordinate axis) and the number of likes and comments generated. Additionally, in order to emphasise the differential aspects of the engagement dimensions of Comments and Likes, Tables 9 and 10 show the results obtained from the multivariate analysis, which includes only the elements with at least marginally  significance in the bivariate analysis. These results ratify and (in some cases) modulate those obtained in the bivariate analysis. Table 9 focuses on likes, while Table 10 refers to factors that generate comments. Overall, as was to be expected, these results ratify the conclusions already presented for the categories with the most impact. We should note that the position of certain aspects, such as alternative art, is reinforced-in this way linking up with one of the features attributed to Berlin-while the role of interactivity items such as users' photos and requesting feedback, and other posts with content such as panoramic views and most visited places, is consolidated. It is also notable that our study ratifies another of the features attributed to Berlin: the role of its "nightlife". Posts that refer to this aspect generate significantly larger numbers of comments and likes.
In any case, the importance of fame is noted in the high numbers of interactions with most visited places, reinforcing the idea of focusing on what are considered to be the best-positioned tourist attractions.
Finally, by way of summary or synthesis, Table 11 shows the main results in terms of the most effective elements for "making a difference", whether positive or negative, with regard to both comments and likes, with indications denoting the level of significance previously defined (Sect. 4).

Discussion
The analysis conducted on the posts of the official Instagram account promoting tourism in Berlin enables us to draw some conclusions which suggest answers to the questions raised. Our results have allowed us to explore more deeply the relevance of the elements identified in previous studies by advancing along three lines: (a) as a more systematic typology (image content, advertising variables and, with a novel approach, Engagement tactics); (b) in a statistical analysis applied to the official account of one of the most significant European capitals and; (c) to obtain implications about the role of the different elements as a mechanism for generating user engagement, thus offering added value when reviewing and reformulating Promotion of urban tourism: insights into user engagement… tourism marketing communication strategies. The main results are discussed below, and in the following section some comments are made regarding implications for management. Firstly, as regards the question as to which categories of posts generate the most user engagement, our study identifies those items whose presence significantly affects interactions in the form of likes and comments. H1 is confirmed, as certain image content generates significantly more interactions, in line with the results of Choi et al. (2007) and Stepchenkova and Zhan (2013), although our study allows for a broader analysis of content typology and engagement dimensions (comments and likes). It highlights the role of both general aspects, like most visited places and panoramic views, and others that are more specific and closely related to the image of Berlin, such as nightlife and alternative art.
In addition, H2 is clearly confirmed by a negative response: more explicit/direct advertising posts generate less engagement. The results thus confirm users' wariness of advertising messages presented as informative, in line with Chen et al. (2020). The confirmation of H3a and H3b is also noteworthy: posts generate substantially more engagement when they explicitly "play" with the tactics to try to stimulate user responses by inducing or stimulating Engagement. This "play" includes, for instance, using users' photos and making requests for feedback. In this regard, we can highlight the important role played by users' photos (not photos from "official sources") and that, in principle, whatever the number of followers that each user has, exercise a modest (at best) bandwagon effect beyond the intrinsic value of each photo. This novel result thus complements previous contributions more focused on the role of prominent influencers (Erz et al. 2018;Ge and Gretzel 2018;Casaló et al. 2020) in assessing the potential role of contributions from "ordinary users" with different numbers of followers. That is why the results of the test of hypothesis H3c, about a potential bandwagon effect are modest but positive for likes (positive correlation between number of followers of the user whose photo is the subject of repost) and modest but negative for comments, are nuanced but a starting point for further examination. The set of sub-hypotheses H3 carry the caveat that, while it is true that the photos of other users have significant impact on both likes and comments, the number of followers that the authors of photos posted have does not seem to generate a clear significant bandwagon effect.
In addition, it is worth highlighting how the results throughout the study show specificities between the elements that generate the most interactions do so parallel to more impulsive responses in the form of likes and more elaborate responses in the form of comments, in line with John et al. (2017). Table 11 identifies three features with positive impact for both types of interaction. However, certain differential impacts are also detected: there are characteristics that have a positive impact on interactions through likes (panoramic views, public ways and civil architecture) and others that exercise this positive impact on interactions through comments (alternative art, nightlife and people). The former (more likes) are more "general" in nature, while the latter-especially those with a higher level of significance in comments (alternative art and nightlife)-appear to be more specific and resemble certain possible characterisations of the "Berlin brand". This, in addition to the asymmetric aspect that certain advertising items clearly generate fewer likes.

Managerial implications
The implications of this study for managing tourism in Berlin are significant, while the research also confirms the usefulness and sensitivity of Instagram as a tourism promotion tool. The asymmetries detected in the different types of content aimed at eliciting interactions seem to recommend lines of action aimed at revising the selection of images to favour those that show a greater capacity to induce responses. Going beyond, it raises the need for a more strategic communication approach and less based on trial-and-error.
The divergences in the degrees of response to different types of post help to outline the communication strategy more precisely, highlighting those aspects that best contribute to defining the personality of the city.
Promoting explicit interactions with users is a way of generating greater user engagement that should be used more intensively. It offers the possibility of creating a community based on users' photos-without the need for these users to be particularly important influencers-as the diffuse influence of all participants on the Internet is significant. There are interesting elements of complementarity-"bridges"-between the users and managers of official accounts. Providing content that generates the most responses is important, and this suggests that combining content linked to broader, more general themes (panoramic views, most popular attractions) with more specific items featuring unique aspects of Berlin that are already well known, such as alternative art and nightlife, would have a positive effect. The negative results generated by more direct or explicit advertising should also provide food for thought. Aspects that remain open include the question of how to enhance the attractiveness of information about museums and other cultural dimensions, increasing the popularity of what is perhaps still a minority dimension and integrating it more fully into tourist assets to which broader audiences can have access.
Recommendation of a continually adjusted, more proactive and interactive strategic plan for the use of social networks are clear implications of our research. Our paper offers more of a starting point than an end point for what could/should be an almost "real-time" communication strategy. Upgrading the role of those responsible for social networks in DMOs deserves particular attention especially in the current situation in which recovering visitors after the problems of 2020 has become a central aspect of economic revitalisation.

Conclusions and limitations to the research
Taken together, the conclusions drawn from our analysis point to a potentially even more important role for Instagram as a tourism promotion tool, taking advantage of the dynamic of interactivity with its users, combining the most firmly-established general tourist attractions with certain peculiar characteristics of the city of Berlin (as an alternative art capital, as a city that is interesting by day and magical at night), and leaving open questions about the use of this social network as a more direct form of advertising, both for private and public sector activities, from restaurants to museums, taking into account the heterogeneous nature of the growing numbers of Instagram users.
We should note that that the "new realities" will have a significant impact on the aspects analysed as regards the elements with the greatest capacity to arouse interest. Aspects related to a safe city, including health and sanitary conditions, will gain in importance after the events of 2020 and will force changes to strategies.
This study could be expanded by analysing the official accounts of other European capitals or international cities that are major tourist attractions, and it would also be interesting to compare the different accounts of these. Expanding the study's scope to cover several years would enable transversal analysis of the evolution of engagement, and this could also be most interesting. Finally, linking the image of the city that users project through their accounts to the image that the official account seeks to transmit would expand the spectrum of the research and generate potentially interesting results.

Conflict of interest
The authors declare that they have no conflict of interest.