Social media are popular platforms for many types of information dissemination. Their wide adoption and low cost have turned them into powerful marketing tools (Jansen et al., 2009; Leung et al., 2015) that companies have adopted as a part of their marketing strategy (Ashley & Tuten, 2015; Constantinides, 2014; Stelzner, 2020; Tsimonis & Dimitriadis, 2014). The same is true for book publishers (Grima, 2017; Throsby et al., 2018). Nevertheless, to be successful in their social media strategies publishers must remain active (Kaplan & Haenlein, 2010) and this requires time and resources that not all companies have. Social media marketing is more difficult for small publishers because of these factors (Martens, 2016; Nolan & Dane, 2018).

Social media campaigns can help book publishers to improve their online visibility, manage their brand and reputation, promote their books and authors, and engage with readers (Li, 2018; Lis & Berz, 2011; Mrva-Montoya et al., 2019; Thoring, 2011; Wang & Zuccala, 2021). Readers can use social media to gather and share opinions about books (Gruzd & Rehberg Sedo, 2012; Nguyen et al., 2019) and social media mentions help publicise a new title (Criswell & Canty, 2014). A convenience sample survey of Bulgarian readers showed that 86% of respondents had purchased a book after seeing it or reading about it on social media (Krumova, 2017), for example. Sweepstakes (free prize draws), giveaways or contests also tend to generate social media consumer participation (Berger & Schwartz, 2011; Jung et al., 2020), so they are frequently used as marketing tools (Teichmann et al., 2005). User engagement and participation have therefore become essential in marketing.

Social media marketing initially aimed to attract new audiences, but is now more concerned with the management of audiences by extracting data about reader preferences (Nolan & Dane, 2018). Through social media, a publisher can get valuable information about who its readers are and what they are interested in; however, this issue has rarely been investigated. One exception is a convenience sample survey showing that Bulgarian readers who are active on social media tend to follow book authors (73%), bookstores (66%), publishing houses (60%) and book bloggers (50%), and they follow publishers and authors to keep up to date with the latest book news (75%) and to get book recommendations (50%) (Krumova, 2017). A survey of 313 customers of four Vietnamese bookshops that used social media to seek books showed that most of these customers used publishers’ social media to look for reviews, comments and posts generated by readers and customers (81.2%), information about promotions (68.4%), new books or titles (51.1%) and book fairs or events (50.2%) (Nguyen et al., 2019).

Neuromarketing, a novel field using neuroscience techniques and biometric instruments, has emerged as a tool to better understand customers’ behaviour, preferences and decision making (Morin, 2011; Stanton et al., 2017). Neuromarketing research has shown that entertainment content with an emotional charge generates more reactions from the consumers (Vences et al., 2020).

Whilst Facebook and Twitter are the social media platforms currently most used for marketing by book publishers (Gómez López & Hellín Ortuño, 2019; Grima, 2017; Magadán-Díaz & Rivas-García, 2020; Martín Becerro, 2016), Twitter seems to generate more individual interactions (Criswell & Canty, 2014). Book publishers seem to use this platform to disseminate content not provided on their websites or other social media sites (Thoring, 2011).

Launched in 2006, Twitter is a microblogging service where each user has a profile page with a name and a @username and optionally can provide a photo/logo, a short description, a location and a webpage address. Users can post messages or tweets of up to 280 characters in length, which are public by default and displayed on their profile page. When a user tweets, the tweet will appear on their followers’ timelines and can be found by other users searching Twitter. Connections among Twitter users are often not reciprocal, and it is normal to follow accounts for news without expecting interactions (Gruzd & Rehberg Sedo, 2012; Kwak et al., 2010).

Book publishers use Twitter to announce, promote and recommend their books, attract new authors, interact with other publishers and professionals, and post news about authors and illustrators (Campos Moreno & Fernández Cuesta, 2019; Martín Becerro, 2016; Thoring, 2011; Wang & Zuccala, 2019). Thoring’s (2011) early study of 48 UK trade publishers found that only 42% of the sampled publishers had a Twitter account. It also reported that publisher’s size influenced its general Twitter use, but publisher size was less influential in terms of patterns of use and the content tweeted. Research has shown that books mentioned by their publisher’s Twitter account get significantly more Twitter mentions as compared to books mentioned by non-publishers Twitter accounts. Thus, by engaging on social media activity book publishers could increase both Twitter mentions and their books’ visibility (Wang et al., 2021).

Through its Twitter account, a publisher can get valuable information about who its followers (audience) are, when they comment, like or share (retweet) tweets, which can help with developing an effective marketing strategy. Tweets can contain mentions of other users (@usernames), hashtags, external hyperlinks (URLs) and emoticons. A hashtag is a word prefixed by the # symbol (#hashtag) that serves to identify a topic and find posts around that specific topic, and a URL points to outside content. As an example, a book publisher can tweet to announce a new book, mentioning its author by including the author’s @username and a link to the sales page. This may help to promote that book on Twitter (González-Arenas, 2013).

A retweet is a re-posting of someone else’s tweet or one’s own tweet, broadcasting the message to the retweeter’s followers (Kwak et al., 2010). This is a type of information sharing an engagement with content (Boyd et al., 2010; Gruber, 2017; Puschmann, 2015; Vargo, 2016). Retweeting has been investigated from different points of view; for instance, research has explored what is retweeted (Chung, 2017; Metaxas et al., 2015), who retweets (Chung, 2017; Luo et al., 2013), reasons to retweet (Boyd et al., 2010), and factors that affect retweeting (Suh et al., 2010). Retweeting often indicates interest in the message or topic, trust in the message and its creator, and agreement with the tweet content (Majmundar et al., 2018; Metaxas et al., 2015).

There are many reasons for retweeting (Macskassy & Michelson, 2011), such as spreading the information, publicly agreeing with someone, commenting on someone’s tweet, saving tweets for future personal access, showing friendship or loyalty, or participating in promotions (Boyd et al., 2010; Majmundar et al., 2018; Recuero et al., 2011). Factors that may affect retweeting include the presence of photos or images (Chung, 2017), the inclusion of usernames, hashtags and URLs, the number of followers and followees of the original tweeter, the age of their account (Naveed et al., 2011; Suh et al., 2010), the perceived informational value of the tweet (Rudat & Buder, 2015), the emotional charge of the tweet (So et al., 2016; Stieglitz & Dang-Xuan, 2013) and the user’s sentiment (Chen et al., 2020). Nevertheless, these motivations and factors can vary between contexts. For instance, whilst the use of URLs was positively associated with retweeting during World Environment Day (Pang & Law, 2017), it was not during a Breast Cancer Awareness Month campaign (Chung, 2017).

Features of tweets posted by book publishers on Twitter operating across different languages have not yet been investigated in detail. The present study gives insights into this gap through a content analysis of the most popular (retweeted) messages posted by Spanish and non-Spanish book publishers, and by comparing the content of different publishers’ tweets. Comparing Spain to the rest of the world gives an interesting contrast that may offer additional insights. Spain has a large publishing industry; Spanish publishers may target a more national audience (although Spanish is an international language) and they may follow different cultural conventions. The research questions are as follows.

The Scholarly Publishers Indicators (SPI) website (http://ilia.cchs.csic.es/SPI) was used to identify book publishers. This website ranks some Spanish (including from Latin America) and non-Spanish publishers that are relevant to humanities and social sciences researchers. The SPI portal includes two rankings: by thematic specialization, which lists book publishers according to the number of titles published in each discipline, and by editorial prestige, where prestige is based on the opinion of 515 Spanish scholars accredited as professors during 2010-2016 (Giménez-Toledo, 2018). More specifically, the SPI index includes all publishers recorded as relevant to their discipline by at least one Spanish scholar, with each scholar allowed to list ten publishers (Giménez-Toledo, et al., 2017).

The editorial prestige ranking includes a general ranking and a discipline-based ranking. The 2018 general ranking was used for this study, comprising 385 Spanish and 425 non-Spanish publishers, a total of 810 organisations comprising of scholarly and trade book publishers, as well as other institutions such as universities, cultural or academic/scientific institutions and governmental organizations for which book publishing is not a core business activity. Whilst editorial prestige is based on the opinion of a small set of Spanish scholars, a previous study using this ranking found that the most prestigious book publishers also had the highest educational impact, based on syllabus mentions, and the highest research impact, based on Microsoft Academic citations (Mas-Bleda & Thelwall, 2018).

The SPI portal was selected because it includes publishers judged to be prestigious by academics and allows data collection from it. Whilst there are more comprehensive databases, such as the Spanish book publishers’ database, they don’t let data be collected automatically and so would need slow manual data extraction.

The first author [AMB] manually checked whether the 810 book publishers had a Twitter account during November-December 2019, collecting their usernames and Twitter URLs. She also checked the main language of the non-Spanish publishers’ accounts based on their location and the main language of their tweets. The Spanish publishers were also manually searched for in the Spanish book publishers’ database (https://www.culturaydeporte.gob.es/webISBN/tituloSimpleFilter.do).

This database is managed by the Spanish ISBN Agency and gives information about all Spanish publishers that are, or have been, active since 1972, such as their current name, other names previously used, their type, location, website and whether they are active. This database was used to verify the Twitter accounts of the publishers, based on their location (postal address) and their website address. Forty (10.4%) publishers were not found in this database and eleven (2.9%) were reported as being inactive.

Half (189, 49%) of the Spanish book publishers had a Twitter account (see Table I), of which 27 (14%) were university presses. Similarly, 236 (56%) of the non-Spanish publishers had a Twitter account (see Table I). These were mainly in English (122, 52%), but also in French (38, 16%), Spanish (33, 14%), Italian (24, 10%), German (12, 5.1%), Portuguese (6, 2.5%) and Dutch (1, 0.4%). A quarter (60, 25%) were university presses.

An in-depth analysis of Spanish publishers revealed some reasons that explain the low proportion with a presence on Twitter. For instance, a few publishers belong to publishing groups with a collective Twitter account for the group rather than the publisher, so these accounts were excluded. Some universities also had a general institutional Twitter account but not a specific account for its publication service, so these accounts were also omitted. Moreover, many publishers listed on the SPI portal were not named book publishers but institutions that publish books of interest for humanities and social sciences areas. In particular, 112 (57%) out of the 196 Spanish publishers that did not have a Twitter account were cultural or academic/scientific institutions (e.g., Casa de Velázquez, Museo del Prado, Sociedad Española de Estudios Medievales, Fundación de Investigaciones Marxistas) or governmental organizations (e.g., Ministerio de Empleo y Seguridad Social, Xunta de Galicia, Instituto Vasco de Administración Pública). Almost all had a general institutional Twitter account, but not a specific account for their publication service.

The social media data analysis software Mozdeh (http://mozdeh.wlv.ac.uk/) was used to collect, on 15 February 2020, all tweets from the book publishers, separating out the single most retweeted tweet from each publisher (if any tweets had been retweeted) using the retweet count information from Twitter. This retrieved 271 most retweeted tweets posted by the Spanish publishers, and 164 tweets posted by the other publishers. These 164 tweets posted by non-Spanish publishers were in English (99, 60%), French (52, 32%), Spanish (12, 7.3%) and Italian (1, 0.6%). This study focused on Spanish-language tweets posted by the Spanish publishers and the English-language tweets posted by the non-Spanish publishers; therefore, as there were 99 English-language tweets, the 99 most tweeted Spanish tweets were selected, to simplify the comparison between them (see Table II). The most popular Spanish tweet had 499 retweets whereas the most popular English tweet had 4461 retweets. For RQ1, an analysis of the most popular tweets was carried out.

For RQ2, inductive content analysis was applied to the 99 most popular tweets (Spanish language) posted by the Spanish publishers and the 99 most popular tweets (English-language) posted by the non-Spanish publishers, 198 in total (Table II). Content analysis is a “research technique for making replicable and valid inferences from texts (or other meaningful matter) to the context of their use” (Krippendorf, 2018, p. 24). It reduces content into categories to enhance understanding of the data and reflect the subject of the study in a reliable manner. In inductive content analysis, the categories are derived from the data rather than prior information or theory (Elo & Kyngäs, 2008). Standard trustworthiness criteria help to assess the reliability of the results (Elo et al., 2014).

The first author read and re-read a random sample of 300 tweets to get familiarised with them and created an initial coding scheme with four facets (topic, #BookTitle, attached image/video, and included link). Two coders [AMB and MM] then conducted a pilot study content analysis of a random sample of 60 tweets in order to test the reliability of the coding scheme, and discussed ambiguities to ensure that the descriptions were clear, with a particular focus on the topic facet. The coders then performed a second pilot study of 60 new random tweets, based on the revised coding scheme, achieving a Cohen’s Kappa score above 0.8 for all facets (Freelon, 2010; http://dfreelon.org/utils/recalfront/recal2/). The coders then carried out a content analysis of the final sample of 198 tweets, based on the final coding scheme, achieving a Cohen’s Kappa score above 0.95 for all facets. This very high agreement level reflects standardised clear tweets written for a narrow range of purposes. The few discrepancies were discussed and resolved by reaching a consensus.

This facet records the topic of the tweet.

RQ3 was addressed with a word frequency test using Mozdeh. The test helped to identify words that occur more often in one publisher’s Twitter account (all tweets, not just the highly retweeted tweets) than in the other publishers’ accounts (Thelwall, 2021). This is a statistical test that compares the proportion of tweets from a publisher containing a word with the proportion of tweets from the other publishers containing that word, producing a z value, that is (approximately) from the normal distribution. A higher value gives stronger statistical evidence that the difference between the proportions is not due to chance, so the top words are the most likely to be publisher-specific terms. This approach systematically identifies differences between publishers at the level of words rather than broad strategies, but strategic differences presumably translate into different words used, so the results may also hint at wider differences between publishers.

This word association analysis was initially applied to the 189 Spanish publishers and the 122 non-Spanish English-language publishers that had a Twitter account, and the top five words from each publisher with the highest z score were picked for the analysis. Tests were separated by language so that, for example, Spanish language tweeting publishers were compared only to other Spanish language tweeting publishers. Four Spanish publishers and three non-Spanish publishers were duplicates; for instance, the Universidad de Valladolid appeared in the SPI portal as ‘Universidad de Valladolid’ (listed in order 69) and ‘Universidad Valladolid’ (listed in order 98). These duplicates were removed and, consequently, the final word association analysis was based on 925 words from tweets posted by 185 Spanish publishers and 595 words from tweets posted by 119 non-Spanish English-language publishers (see Table III).

The 99 most retweeted Spanish tweets were posted by 12 Spanish book publishers (Table IV) and the 99 most retweeted non-Spanish tweets were posted by 19 non-Spanish book publishers (Table V). The Spanish publisher Alianza Editorial created 72% of the popular tweets, followed by the Editorial Espasa (8%). The non-Spanish publisher Frontiers posted almost half (46%) of the retweeted non-Spanish tweets, followed by Macmillan Publishers (11%), Bloomsbury UK (8%) and Penguin Books (8%). The Spanish publisher with the most retweeted tweets also had the most followers and tweets, although the same was not true for the non-Spanish publisher.

Four features were examined in the content analysis: topic, book title hashtags, images, and URLs. The topic of the tweets was examined for 198 tweets (99 posted by Spanish publishers and 99 posted by non-Spanish publishers). The other three features were not examined for tweets coded as “Non-Applicable” on the topic. Therefore, the inclusion of hashtags, images and URLs were examined in 140 tweets (99 posted by Spanish publishers and 41 posted by non-Spanish publishers).

As Figure 1 shows, most (80%) of the popular tweets posted by the Spanish publishers are general literary quotes (73%) or quotes from books published by the publisher (7%), while the remainder are mainly general news (8%), publisher sweepstakes (6%) and literary award announcements (4%). Unexpectedly, the majority (59%) of the popular tweets posted by the non-Spanish based publishers are not related to literature or the book industry. The reason is because almost all of them are posted by Frontiers, a large publisher that not only publishes books but also journals. Its Twitter account (@FrontiersIn) is a general account rather than a book-related account and all its sampled tweets are about research findings published in its scientific journals. The remaining popular tweets posted by the non-Spanish based publishers are mainly sweepstakes (23%), general news (9%), announcements of new books (5%) and tweets in which a book is featured or recommended by its publisher (3%). Tweets related to literary awards are only provided by Spanish publishers. None of the sampled tweets refer to a book launch, book signing, book talk, book fair or book review.

An in-depth analysis shows that each publisher seems to have a particular successful social media strategy. Almost all quotes (both general quotes and quotes from the publisher’s books) were posted by Alianza editorial. Union Editorial posted only general quotes; Tusquets Editores, Hart Publishing, Hodder & Stoughton, Macmillan Publishers and Thames & Hudson mainly posted sweepstakes; Duke University Press only posted featured books; and Trotta and Penguin Books mainly posted general news. Other publishers, such as Editorial Espasa and Bloomsbury UK have posted popular tweets with different content.

Almost all literary quotes are classic quotes from well-known writers, novelists or philosophers, such as Fiódor Dostoyevski, Franz Kafka, Albert Camus, Jean-Paul Sartre, Søren Kierkegaard, Arthur Schopenhauer, Friedrich Nietzsche, Oscar Wilde and so on listed in the order of frequency. A few quotes (seven) are presumably extracted from a book published by the publisher, since the book title, its author and a picture of that book cover or a link to the publisher website are also attached in the tweet. These books are republished classic works from well-known writers rather than recently published books by contemporary authors. The shared literary quotes are somewhat inspirational, reflective or emotional sentences about freedom, solitude, emotions (anger, sadness, love, hope, hopelessness), knowledge and ignorance, critical thinking, passion, disobedience, or ways of living. Table VI displays the top 10 most retweeted quotes.

As shown in Table VII, the inclusion of hashtags of book titles, images/videos and URLs is much more frequent in popular tweets from non-Spanish book publishers than in those posted by the Spanish publishers. Out of 65 tweets that include images or videos, three are videos. The images are usually pictures of book covers or pictures related to a sweepstake. The sweepstake pictures usually illustrate the book or product being given away, often together with competition terms and conditions. Some images, especially in tweets posted by Spanish publishers, are author pictures (Figure 2).

Almost a quarter (31, 22%) of the highly retweeted tweets included a URL. Because of the Twitter length limitation, it is common to use a URL shortening service when including a URL in tweets (Suh et al., 2010), with t.co, bit.ly, ow.ly and goo.gl used in the sampled tweets. Out of 31 tweets including a URL, the majority linked to the publisher website (61%) but some also linked to media sites (13%), such as newspapers, magazines, TV programmes and radio stations. Links to a publisher’s other social media (particularly to blogs) and links to booksellers or book recommendation sites were only included by two non-Spanish publishers (see Figure 3). Links pointing to the publisher website mainly target: 1) a page of information about the book and its authors, and usually the option of buying the book online, and 2) a page with the sweepstake terms and conditions. Links pointing to booksellers or book recommendation sites also point to an author or book webpage. Regarding the four links pointing to media sites, three point to an online newspaper and one to a book industry magazine.

The word association analysis shows which terms occur more often in a Twitter feed of a specific publisher than the rest. Two sets of word frequency tests were carried out to identify the five most publisher-specific terms for each publisher. The first test generated 925 words from 185 Spanish publishers’ Twitter accounts and the second test generated 595 terms from 122 non-Spanish English-language publishers’ accounts. About half of the top words were hashtags or usernames (see Table VIII). Overall, the results suggest that the publishers mainly tweet about themselves. This is unsurprising since the purpose of each account is presumably to market publisher’s books, directly or indirectly. In this context, there are largely predictable differences in the people and books covered by the different publishers, rather than illuminating strategic differences between their tweeting strategies.

The most common type of term found for the Spanish book publishers was a publisher’s name: 135 terms (15% of all terms), of which 60 (44%) were their Twitter usernames and 19 (14%) were hashtags. Twitter accounts of authors (@author) associated with the publisher were also common (84, 9% of all terms) and there were a few other professionals related to the publisher, such as illustrators, translators and publisher directors/editors. Including the name of an author can help to market a book by reaching Twitter users that search for the author, and by attracting the author’s attention in the hope of a retweet to their followers. Some of the publishers included in the study are university publishers, so it is not surprising that some publisher-specific terms were university-related Twitter accounts (for a university, university faculty, university library or university research group). Rare terms included publisher news (13, 1.3%), book titles (11, 1.2%) and book recommendations (8, 0.9%), almost all as hashtags, and the city where the publisher is located (8, 0.9%). There were some references to book fairs (4) bookstores (4), awards (4), publisher adverts/news (3), book signings (2) and a publisher’s anniversary (1).

The results were similar for English-language publishers. The publisher-specific type of term most common in their tweets was a publisher’s name, with 149 (25%) occurrences, of which 45 (30%) were their Twitter usernames and 9 (6%) were hashtags. The results included Twitter accounts of associated authors (38, 6.4% of top terms), publisher subsidiaries, publisher imprints, journals, series, collections and projects related to the publisher (25, 4%) and university-related Twitter accounts (university, university faculty, university library or university research group accounts), the city of the publisher (13, 2.2%), events (conferences), publisher advents/news (6, 1%) and publisher anniversaries (6, 1%). There were also a few mentions of book titles (4), associations (4), book fairs (2), bookstores (2) and book signings (1).