Wrinkles in a CSR story: mismatched agendas in fast fashion service brands ’ CSR reputation Wrinkles in a CSR story

Purpose – Corporate social responsibility (CSR) is an increasingly important issue for service brands in fast fashion retailing, as consumers ’ negative impressions about retailers ’ CSR activities influence brand experience. Consumers ’ impressions of CSR efforts arise based on agendas communicated through many channels from different sources. The paper unravels the ‘ wrinkles ’ , i.e. possible mismatches in CSR communication around servicebrandsbystudyingdifferencesbetweenthethreemainsourcesoffastfashionbrand-relatedCSRagendas:Autonomouscompanycommunication,newsmediaandsocialmediapostingsbyconsumers. Design/methodology/approach – The authors use structural topic modeling (STM) to analyze a corpus of textsfocusingontheCSReffortsofthreemajorfastfashionservicebrandsoverthreeyears.Thetextsincluded 89 items of company communication (CSR reports and press releases), 5,351 news media articles about the brands ’ CSR efforts and 57,377 consumer generated tweets about the brands. Findings – The STM analysis extracted 26 different CRS-related topics from the texts. Results showed differences in how much the three sources emphasized topics. The brands ’ own communication puts emphasis on environmental responsibility. Newsmedia tendedto report on economicissues,treatment of employees and specific CSR-related events. Twitter showed more activity in discussing incident-based and emotionally charged topics. Research limitations/implications – The results feed into the ongoing discussion about how companies ’ CSR communication relates to communication in the press and among consumers. The authors highlight themes in the individual topics that are emphasized by the three sources, and discuss how CSR themes emerge in the overall transformative agenda. Practicalimplications – The paper highlightshow fast fashion service brands can identify and understand different CSR agendas arising around their brand. Insight into such agendas can be used to tailor the brands ’ communication strategies. Originality/value – Thepapercontributestotheunderstandingofthefactorsbehindfashionservicebrands ’ CSR reputation, highlighting how the three main sources of CSR reputation (companyreports, news and social media) emphasize different types of agendas.

groups' CSR impressions. We will frame our study using Hewett et al.'s (2016) "echoverse" model. However, whereas Hewett et al. (2016) focus on the volume and valence of communication, we contribute by focusing on meaning and content of communication. Agenda-setting theory (McCombs and Shaw, 1972) provides us with a logic to explain how communication from stakeholder (actor) groups (companies, news media and consumers) serves to set the agenda for which topics become important for the public. To our knowledge, there have not been any studies about differences in CSR agendas between all three of Hewett's et al. (2016) communication sources.
We use computational content analysis (Chae and Park, 2018) to discover and compare topics within the three types of sources. We thus also contribute to service research showing how this type of methodology can be used in a service brand context. In the empirical part of this study, we focus on three leading fast fashion brands (H&M, Zara and Primark) during a three-year period (2018, 2019 and 2020). We collected data from this period about: (1) The brands' official CSR communication in terms of CSR reports and press releases, (2) news media articles focusing on the brands' CSR, and (3) Twitter posts that talk about the brands' CSR. We then compared the salience of the discovered topics within the three sources, aiming to identify differences between them.
The study contributes to research on transformative agendas in service communication (Tsiotsou and Diehl, 2022) by uncovering mismatches in CSR agendas between fast fashion service brands' CSR communication, reporting on the same issues in news media and in social media communication. We characterize such mismatches as "wrinkles" in the overall CSR story of the service brand. We discuss reasons for such wrinkles, and highlight how they can be indicative of CSR communication problems for the service brand. Moreover, we contribute to literature on CSR communication by showing how company communication relates to other types of communication beyond the company's control. CSR communication literature typically studies how the company can influence stakeholders, but is less focused on the role of external stakeholders in overall CSR communication (Crane and Glozer, 2016;Verk et al., 2021).
The paper is structured as follows: We begin by characterizing the meaning and role of CSR for fast fashion service brands. Next, we introduce agenda-setting theory as a way to understand the emergence of CSR reputation, and discuss different sources for CSR agendas. We then present a topic modeling study where we use texts from the three sources of CSR agendas identified in the previous chapter to discover a set of topics related to the CSR efforts of three fast fashion brands. We present the differences in how the sources emphasize these topics, and conclude by discussing the significance and implications of our findings for service brand communication strategies and suggest topics for further research in this domain.
The CSR reputation of fast fashion service brands CSR has been characterized as organizations' acts of assuming a responsible stance within strategic planning and core operations, "so that the firm is managed in the interests of a broad set of stakeholders to optimize value over the medium to long term" (Chandler, 2016, p. 248). A popular way of understanding CSR is in terms of the "three pillars" of social, economic and environmental responsibility (Księ _ zak and Fischbach, 2017;Purvis et al., 2019). A review of CSR literature characterizes CSR as a multidimensional concept that incorporates economic, ethical and social dimensions, the interests of different stakeholders as well as sustainability and discretionary dimensions (Sarkar and Searcy, 2016). A review focusing specifically on CSR in fast fashion identified similar themes, but also highlighted additional subjects such as advertisement-, consumption-and culture-related issues, as well as company performance-and supply-chain issues (Thorisdottir and Johannsdottir, 2020). In sum, the issues can be seen as reflecting instrumental, political, social or ethical approaches to CSR (Garriga and Mel e, 2004).
Arguably, the fast fashion industry is currently moving towards more ethical and sustainable practices (Rutter et al., 2017). This is partly driven by pressure from customers. Research has found that there is a link between consumers' perceptions of fast fashion brands' CSR efforts and their purchase intentions as well as the brand equity (Sun et al., 2014;Chang and Jai, 2015). Moreover, if consumers can attribute brands' CSR efforts to altruistic motives, this will further serve to support CSR perceptions, brand trust and purchase intentions (Miotto and Youn, 2020). For service brands, CSR performance has been shown to have a positive relationship with customer satisfaction and brand loyalty (He and Li, 2011;Cha et al., 2016). Thus, it is important that service brands communicate about CSR issues with consumers to build their CSR reputation (Lee, 2016). However, consumers may interpret fast fashion brands' communication as hypocritical if their CSR claims appear incompatible with the brands' business model (Chun and Giebelhausen, 2012;Wei and Jung, 2021). News media reports can serve to bring such hypocrisies to public attention (Seele and Gatti, 2017). Consequently, fast fashion brands' CSR reputation emerges through interactions between various agendas that originate from many different sources (Kim and Ferguson, 2014).

Agenda setting theory
Researchers have argued that news media reporting influences the public's perceptions of societal issues (McCombs and Shaw, 1972). Agenda-setting theory describes how news media "sets the public agenda" by altering the salience of topics and their images among the general public (Carroll and McCombs, 2003;Kim et al., 2017). This is usually divided into first-level agenda setting, which refers to what topics become salient, and second-level agenda setting, which refers to how various properties and qualities are communicated within these topics (Wu and Coleman, 2009). Later on, agenda setting theory has been augmented with intermediary agenda-setting, which refers to how issues are transferred from one medium to another, for example between corporate reports and news media stories (Tam, 2015;Cheung et al., 2020).

Sources of CSR agendas
Research has highlighted two primary sources that influence the public CSR agenda: corporate reports and news media (Pollach, 2014;Tam, 2015). Businesses try to build their own CSR reputation by means of CSR reports and press releases (Lee, 2016;Pollach, 2015). Simultaneously, the news media highlight various CSR topics through news reporting (Zhang and Dong, 2021). The salience and tone of this news reporting has a positive relationship with corporate reputation (Vogler and Eisenegger, 2021). In general, issues expressed in news media tend to mirror those in corporate reports (Pollach, 2014;Tam, 2015). However, while news media agendas may influence corporate agendas, the inverse is not likely to happen (Pollach, 2014). A reason for this may be that news actors do not consider corporate CSR topics to be very newsworthy (Pollach, 2015). Thus, company communication in the form of CSR reports and website communication has an important role in raising public awareness about CSR topics that the businesses themselves want to emphasize, serving as a counterpoint to the news media agenda. This means that it is important to understand whether there are differences in how the two sources emphasize CSR topics.

RQ1
. Are there differences in the salience of brand-related CSR topics communicated by traditional news media and in fast fashion brands' own corporate CSR communication?
Recent research has also highlighted social media as an important third CSR agenda-setter (Feezell, 2018). Social media agendas and traditional news media agendas are often intertwined and reflect each other (Yang et al., 2016). However, while news media agendas have a clear effect on social media agendas, the inverse is less likely (Su and Borah, 2019). Crucially, social media differs from traditional news media through its interactive nature, allowing every user to contribute to the general agenda (Albalawi and Sixsmith, 2015). This sometimes leads to differences in the agendas that news media and social media emphasize (Landis and Allen, 2022). Moreover, communication about CSR on social media tends to be more emotional and less formal than within traditional media (Chae and Park, 2018). Thus, it is important to understand how the agendas around fast fashion brands' CSR efforts differ between traditional news media and social media.
RQ2. Are there differences in the salience of fast fashion brand-related CSR topics communicated in traditional news media and in social media?
Finally, there are often mismatches between companies' CSR agendas and the CSR expectations of social media users (Colleoni, 2013). Companies frequently fail to engage in any meaningful dialogue with social media users (Colleoni, 2013;Martin and Gr€ ub, 2016). Moreover, social media environments are heterogeneous, and each stakeholder group on social media will interpret CSR communication through their own prism of important issues (Dutot et al., 2016). Within the fast fashion context, research has also indicated that celebrity influencers can use their social media presence to influence public opinions about sustainability and CSR in fashion (McKeown and Shearer, 2019). This means that it is relevant to understand the differences in what fashion brand CSR topics are prominent on social media compared to the ones emphasized by the brands themselves in their own communication.
RQ3. Are there differences in the salience of brand-related CSR topics communicated in social media and in the fast fashion brands' own CSR communication?

Themes in CSR agendas
The three previously highlighted sources of agenda-setting mirror Hewett's et al. (2016) "echoverse"-model, which summarizes the main sources of messages about brands as (1) company communication, (2) traditional news media and (3) social media communication.
The three sources interact and reflect each other, ultimately influencing consumer sentiment and decision-making (Hewett et al., 2016). Thus, each source contributes to the public transformative agenda (Tsiotsou and Diehl, 2022). However, the different actors involved in the three sources are likely to emphasize different CSR domains. In general, companies strive to raise awareness about their CSR efforts (Chaudhri, 2016), news media actors aim to report on newsworthy stories (Luo et al., 2012), and social media users wish to communicate about issues that they feel are important (Boyd et al., 2016). Understanding any underlying themes in the differences between the sources may have implications for how service companies' can plan their CSR communication (Andreu et al., 2015;Tsiotsou and Diehl, 2022). Thus, it is relevant to understand if there are any underlying CSR themes that are specific or common to each of the three sources.
RQ4. Are there differences in how the three sources' (company communication, news media and social media) emphasize underlying themes in their communication about fast fashion brands' CSR?
Next, we describe our methodology for studying the content and salience of topics about fast fashion-brands' CSR within the three sources, and so answering the four proposed research questions.

Methodology
Service brands incorporate many touchpoints and interactions (Leroi-Werelds and Matthes, 2022) with a focus on relational exchanges (Tsai, 2011). Many fast fashion companies are highly store-and interaction-focused, and can thus be seen as a type of service brand. We chose to focus on studying brands in the fast fashion industry. In order to study the content and salience of fast-fashion brand related CSR topics within corporate communication, news media and social media, we employed a mixed methods approach, combining manual data collection, data scraping and topic modeling with a qualitative and quantitative analysis of the results. Topic modeling is a form of natural language processing (NLP: Liddy, 2001) that is used to extract a set of topics from a text corpus. To represent the three information sources, we collected three types of data: (1) a set of fast fashion brands' official communication about their CSR efforts (CSR reports and press releases), (2) published newspaper articles about the brands' CSR efforts and (3) consumer communication about the brands' CSR on Twitter. To make the amount of data manageable, we limited our data collection to three high-profile fast fashion brands, H&M, Zara and Primark. These are three of the four largest fast fashion brands in terms of revenue (Smith, 2021). Moreover, we restricted our data collection to a recent three-year period (2018, 2019 and 2020). Zhang and Dong (2021, p. 269) state that "as CSR has rapidly developed in the past decade, up-to-date data must be incorporated into the analysis". Data were collected using a set of keywords based on Sarkar and Searcy's (2016) review of CSR topics (Table 1). This roots the data collection in existing conceptualizations of CSR. The keywords include Sarkar and Searcy's (2016) original keywords representing the six dimensions of CSR, namely the economic, ethical, social, stakeholders, sustainability and discretionary dimension. In addition to the original keywords, we included a set of general CSR-related terms, such as "responsibility", "CSR" and "impact". This approach expands the scope of data collection beyond previous studies of CSR communication, which typically have searched for texts using only the term "CSR" (e.g. Chae and Park, 2018). The keywords were also stemmed (e.g. "*sustainab*") to capture variations of the words. This means that we also capture antonyms for most of the words (for example, "ethical" also covers "unethical"). However, the terms dealing with well-being found under the social dimension do not generate antonyms. To address this shortcoming and make results in the social dimension equivalent with the other ones, we drew on Stone and Mackie (2013) to include the antonyms "suffering" and "abuse", thus covering negative aspects related to well-being.
Using the dimensions and keywords as a guide, we first manually collected CSR reports and CSR-related press releases that the three fast fashion companies had archived on their public websites. Next, we scraped Twitter for consumers' tweets about the three fast fashion companies using Python and the snscrape package (JustAnotherArchivist, 2021). We combined the selected brands' names with the search terms in Table 1 to automatically identify and collect relevant tweets. Finally, we collected CSR-related news articles about the three companies from the LexisNexis database (www.lexisnexis.com) by using a similar search logic. We removed all non-English texts from the corpus, and then removed HTML tags, URLs, punctuation, special characters, numbers and stop words from the texts. Words were then stemmed, converted to lower cases and infrequent words were removed. The resulting text corpus consisted of 89 CSR reports and press releases (H&M: 52; Zara: 19; Primark: 18), 57,414 tweets (H&M: 19,050; Zara: 15,445; Primark: 22,919) and 5 351 news articles (H&M: 2,227; Zara: 734; Primark: 2,390) adding up to a total of 62,836 documents.
We analyzed these data using structural topic modelling (STM) with latent dirichlet allocation (LDA; Blei, 2012;Blei et al., 2003). LDA uses a "reversed generative process" answering the question: "What is the hidden structure that likely generated the observed collection?" (Blei, 2012, p. 10). The documents and its content (words) are observed, but the topics, per-document topic distributions, and per-document per-word topic assignments are hidden structures. The model tries to deduce the hidden topic structure, based on the observed documents and words inside the documents. For an in-depth explanation of LDA we refer you to Blei et al. (2003) and Chae and Park (2018). STM builds forth on LDA by facilitating the use of two types of covariates: topic prevalence and content prevalence covariates. This study solely used topic prevalence covariates, which makes it possible to see how the proportions of topics differ for different levels of the covariate. This makes it an excellent tool for researchers in social sciences and business. For an in-depth explanation of STM, see Roberts et al. (2014) and Chae and Park (2018). The underlying STM model is an unsupervised machine learning technique, but requires the researcher to decide on the number of topics that will be extracted. According to Roberts et al. (2020), in general around 60-100 topics works well for corpora with 10,000 to 100,000 documents. After running diagnostic tests on various evaluation metrics (held-out likelihood, residuals and semantic coherence) and setting the number of topics by the algorithm by Lee and Mimno (2014), we found that 67 topics seemed to be suitable for this particular data set.
Next, we needed to interpret the meanings of the extracted raw topics, and then decide whether they dealt with CSR or not. Two independent coders analyzed the extracted topics and interpreted them in order to formulate topic descriptions. This was done based on the extracted topic keywords as well as exemplary texts for each topic (see Table 2, "topic keywords" and "topic interpretation"). The coders agreed about the interpretation of the topics in 62 of 67 of the cases. Conflicts were resolved by discussion until consensus was reached. However, the CSR keywords in Table 1 cast a wide net for data collection, and so a majority of the extracted topics did not directly reflect CSR issues, instead representing, e.g. customer complaints. This is a consequence of how, for example, "customers" is a CSR keyword (Sarkar and Searcy, 2016), but the inclusion of this word also highlighted Twitter communication between customers and company representatives. After manually screening the 67 extracted topics using Sarkar and Searcy's (2016) dimensions as a guide, we found that 26 of the topics were directly related to CSR issues (see Table 2).
These 26 CSR-related topics were then analyzed to find each topic's proportional point estimate for each level of the covariate in each of the three sources (company reports, Twitter and news). The estimateEffect function of the STM package performs a regression analysis with the topic-proportions as the dependent variable and the document meta-data as covariates [1]. Using ANOVA analysis for each topic separately, we compared these topic proportions to see whether a given topic was more prevalent in one data source compared to another. Finally, using inductive categorization (Gioia et al., 2013) we categorized the 26 topics into a set of five emergent, underlying CSR themes arising from the topics (Table 2). Findings Table 2 summarizes the findings from the topic model, displaying all discovered topics that concern the selected fast fashion companies' CSR efforts. The coefficients in the table indicate how prominent the topics were in each of the three sources (company communication, tweets and news stories). The table presents (from left to right) the overarching themes, the topics, number of documents where the topic is dominant, the topic proportions for each source, the F-test statistic of the ANOVA test and the results of the post hoc tests where we compared the coefficients of the sources with each other for each topic. Asterisks indicate cases where there were significant differences between the three sources, allowing us to address research RQ1, RQ2 and RQ3. The table also displays overarching themes representing groups of topics, addressing RQ4. Next, we present the discovered topics and themes.
Theme 1: responsibility towards employees and workers The first clear CSR theme arising from the data concerned the treatment of employees, covered by topics 1-5. For example, topic 1 focused on sharing company profits among employees. A list of additional unique keywords supported this interpretation, as they included hashtags such as the ones in the following tweet: @storebrand_no you care about others' well-being. In the next @hm AGM it is in your hands to support our citizens' proposal to use this year's dividends to pay #LivingWageNow. I count on you to #TurnAroundHM! #ShareYourProfits Topic 2 concerned the furloughing employees during the pandemic, whereas topic 3 reflected the story of how H&M was fined for spying on its employees in Germany. Another aspect of this theme was found in topics 4 and 5, which concerned the treatment of workers in the companies' factories, for example in terms of labor rights in South Asian factories, as well as general forced labor-and minority issues, such as the treatment of the Uighur people in China. All of the five topics received similar attention in all three sources (company communication, news media and social media), except in the case of topic two (furloughing), which clearly was covered to a higher degree by news media than by the other two.
Theme 2: profits and economic responsibility Topics 6-8 focused on economic themes, chiefly dealing with the companies' profits, as indicated by the keywords in Table 2. Topic 7 concerned the government and the payment of taxes, including for example an event where manufacturing businesses in Myanmar urged the government to lower their taxes. Within this theme, there was a clear tendency for the news media to communicate more about the topics, at least concerning financial news.
Theme 3: sustainability and environmental protection Topics 9-14 reflected themes of sustainability and environmental protection. For example, topic 9 highlighted fast fashion companies' boycott of Brazilian leather due to the cattle industry's impact on the rainforests, whereas theme 12 focused on actions for a circular economy.
A clear trend could be spotted within this theme, as companies communicated to a higher degree about sustainability and environmental protection than the other sources did. In every occasion where a significant difference could be found (topics 9, 11, 13 and 14), it was companies that communicated about the topics more. The following excerpt from Primark's press release about sustainable cotton (topic 11) on August 28, 2019 serves an example of such communication: Theme 4: legal and ethical responsibility This theme encompassed a wide range of topics that nevertheless all highlighted the treatment of external groups. All of topics 15-21 concerned situations where fast fashion companies had either made a mistake or corrected their course in cases such as screening for offensive designs in mass produced fashion, breaching the copyright of street artists, improving animal rights, or facing consequences for an offensive ad.
This was the only theme where a topic had been most prominent on Twitter, as the accusations of racism against an H&M ad (topic 20) generated much discussion among Twitter users. However, the ad later featured more prominently in the news media than in the other sources due to an event where members of the South African Economic Freedom Fighters (EFF) group attacked H&M stores in response to the ad (topic 21). Similarly, the general topic of various harmful designs was more prominent in news media than in other sources (Topic 15).
Theme 5: responsibility to the community and customers Topics 22-26 all dealt in various ways with the companies' responsibility towards their customers and the community, mostly in terms of safety rules for COVID-19, but also by warning customers about illegal resellers, as well as the impact of delayed shipments.
There were two topics where a difference could be found between the sources. Topic 22 dealt with various consequences of COVID-19 for the companies, and was more prominent in the news media compared to the other sources. Topic 24 dealt with the same topic, but focused on the companies' contributions to the fight against COVID-19, and was the most salient in the fast fashion brands' own communication. Figure 1 further illustrates the findings by visualizing all the identified differences in topic salience between the three sources. The width of the bar indicates the error margins.

Discussion and implications
The results indicate that there are differences in how the three sources (firm communication, news stories and social media) approach CSR topics relating to fast fashion brands. Interestingly, over half of the topics (14 of 26) were equally prominent in all three sources. These topics included both high-magnitude ones, such as Topic 25 ("Delayed deliveries because of Covid19 disrupted supply chains") and low magnitude ones, such as Topic 15 ("Harmful designs"). This supports findings from earlier research on agenda setting, which has shown that the agendas in the three sources are likely to reflect each other (Pollach, 2014;Tam, 2015).
However, the results also highlighted several differences in agendas between the sources, allowing us to address our Research questions 1-3. Most prominently, the fast fashion brands themselves stressed environmental sustainability more than the other two sources did. Previous research has identified environmental engagement as a core theme in companies' CSR communication, a theme that is especially emphasized by companies involved in manufacturing (Tang et al., 2015). It may thus be easier for fast fashion companies to communicate about environmental sustainability in relation to their production processes than about other types of sustainability. In terms of responsibility towards the community, the companies tended to communicate more about their own efforts in fighting COVID-19 than the other sources did.
The second sourcenews storiesdiverged from the other two by putting more emphasis on topics that had a link to recent events or developments. According to Harcup and O'neill (2017), news organizations select what news stories to report based on a set of "news values", i.e. factors assumed to raise interest among readers. Examples of such factors are conflict, surprise, magnitude, drama and the power elite (Harcup and O'neill, 2017). Such news values may explain why our results show how traditional news media emphasize CSR topics such as company profits, COVID-19 safety measures, protests against the companies, or the furloughing of employees. Thus, our results can be used to address RQ1: When compared to news media, fast fashion service brands tend to emphasize environmental CSR topics in their communication more, whereas news media put more emphasis on CSR topics related to surprising events and conflict.
The third source for CSR agendas was social media. There was only one topic that was more prominent within this source than in the other ones: Discussions about the inflammatory ad by H&M. This is in line with earlier research by Chae and Park (2018, p. 15), who point out that communication in social media often is "less formal and more emotional" than in traditional media. In addition, activists often use social media to challenge companies' CSR messages (Tench and Jones, 2015). This may help explain the prominence of this particular Figure 1. Topic proportions between the three sources topic, as the controversial H&M ad inspired many strongly worded responses among Twitter users. Thus, a partial answer to RQ2 and RQ3 may be that social media differs from the other two sources by providing greater salience to event-based, emotionally charged topics.
When viewed in combination, our findings address RQ4 by suggesting general themes in the topics that the three sources emphasize. Research has indicated that fast fashion companies wish to communicate topics that can contribute to a positive impression of their own CSR efforts (c.f. Chaudhri, 2016). Thus, they tend to report on what they have done in terms of acts of responsibility or philanthropy. News media is driven by their quest for important or interesting news stories (Luo et al., 2012), and so they gravitate towards viewing CSR in terms of remarkable events, important developments or injustice. Finally, social media users' communication is driven by their personal view of important issues (Boyd et al., 2016), and so they might on the aggregate level engage more with emotionally loaded topics.

Theoretical implications
The observed differences in agenda themes between sources reveal the underlying potential for "wrinkles" in the service brand's CSR story, that is, possible discrepancies between topics prevalent in the three "echoverse" communication sources. Our paper contributes to research on CSR communication by providing extended context for corporate communication about CSR themes, relating it to how other stakeholders within Hewett et al.'s (2016) "echoverse" communicate about the same topics. Earlier CSR communication research has predominantly focused on communication carried out by the company itself (Crane and Glozer, 2016;Verk et al., 2021). Consequently, the paper contributes by showing that company CSR communication takes place within the context of the two other overarching echoverse communication sources, which then ultimately combine to set the public agenda on CSR issues. In addition, whereas Hewett et al. (2016) focused on the sentiment and volume of messages, we introduced content and textual narratives to the model and showed how CSR communication can be studied on a large-scale using topic modeling techniques. Moreover, the paper contributed to service research by demonstrating how advanced text analytical methods can be used in service research to extract meaning from large volumes of text data. The paper contributes to transformative service research by shedding more light on the roles of different CSR domains in transformative service communication (Tsiotsou and Diehl, 2022). The concept of the "wrinkle" helps us study how CSR topics and themes differ between the communication sources that set the transformative public agenda, and how this may have implications for transformative marketing communications. In the fast fashion industry specifically, wrinkles that originate from brand communication tended to relate mostly to the third pillar of CSR-environmental responsibility (Księ _ zak and Fischbach, 2017;Purvis et al., 2019). We argue that this type of wrinkle is not a problem for service brands as such, as they result from the brands' efforts to impose their own transformative agenda onto the public agenda. News media and social media, on the other hand, focused more on the negative effects of the brands' activities and deficiencies in their CSR efforts. This led to another type of wrinkle, where the service brand is either out of tune with the public agenda, or has chosen to abstain from certain topics. In the fast fashion context, the wrinkles that emerged from the news and social media mostly reflected social and ethical dimensions of CSR (Sarkar and Searcy, 2016;Garriga and Mel e, 2004). Thus, social and news media often highlight problems, deficiencies or mistakes made by the service brands, which then can influence the public CSR agenda.

Implications for business
The paper highlights how fast fashion service brands can identify and understand different CSR agendas arising around their brand. By studying the differences in agendas between the three sources, businesses and organizations can recognize how their own brands' agenda aligns with the two other sources' agendas. In general, fast fashion brands need to decide what topics to include into their own CSR communication. Elving et al. (2015) argue that CSR communication requires a shared understanding between the business world and the lifeworld. The identification of topics that are prevalent for different groups contributes towards creating this shared understanding. By being mindful of the agendas in other sources, the brands can tailor their own CSR communication to consciously counter, respond to, or ignore the external agendas and when appropriate align themselves with the demands of customers and the community as a whole. Our outcomes challenge managers to rethink their service brand communication strategies, which often risk relying on myopic assessments of corporate brand communication effects. Based on the wrinkles in the CSR story service brands will be able to improve their CSR communication and practices by assessing and understanding the possible effects it might have in the echoverse.

Limitations and further research
The scope of this paper was limited by focusing on only one industry making use of a limited set of communication sources. Further research could deepen the insights provided in the paper by including other industries and include other modes of CSR communication, such as other social media channels and in-store communication. Moreover, future research could take into account the valence and volume of topics and how they evolve over time. In this paper we only compare the topic prevalence covariates between the sources. Further studies could also introduce content covariates (Roberts et al., 2020), which would facilitate an examination of the vocabulary used to talk about a particular topic within various sources, allowing researchers to, e.g. compare the differences in the words used to describe the same topic in news media, tweets and company communication. Furthermore, it might be interesting to incorporate competition and competitive interactions into the analysis of CSR communication. Another limitation is that the data collection was based on keywords from academic CSR research sources. It is likely that consumers discuss CSR-related issues in language that is less formal. Thus, future research could incorporate a wider set of search terms to provide a richer and possibly more accurate set of topics. Given the fast development of text analytic tools this method could be extended to more service industries, more brands and over longer periods of time at the same time.