Emojis analysis at international trade shows in five countries: ex- and post-COVID-19 Open Access

This research examines the role of emojis in B2B digital marketing across countries. Emojis enhance message clarity and engagement in online interactions (Luangrath et al., 2022; Ge and Gretzel, 2018). In formal B2B communication, understanding their impact on eWOM volume (e.g. retweets on Twitter) is key, especially post-COVID when digital engagement grew. By exploring cross-cultural differences in emoji interpretation, this study addresses how emojis affect brand perception and purchasing intentions globally.

In this scenario, the aims of our study are both to predict eWOM volume based on emoji presence, amount of emojis and time frame (posting date ex ante COVID-19 or posting date ex post COVID-19) in five countries and to identify differences between the samples and a moderation effect of the country on the relationship studied.

This study will contribute to marketing literature in B2B contexts differently. First, marketing literature has recognized social media’s utility in promoting international trade shows (ITSs) (Lapoule and Rowell, 2016), improving their performance (Singh et al., 2017).

Second, this paper focuses on the uses of emojis, which are less analyzed than other social media tools in B2B contexts (Balaji et al., 2023). In this sense, despite B2B firms’ expanding use of social media, they still face significant challenges in creating compelling messages that effectively engage customers (Balaji et al., 2023).

Third, the study adds value by identifying whether there are differences among samples in emoji presence, amount of emojis, time frame, eWOM volume and whether there is a moderation effect of the country on the relationship studied. Given that the eWOM studies have been focused on a single country (Donthu et al., 2021), our study amplifies knowledge about this variable.

At the general level, this study responds to the recent call for research on social media (Dwivedi et al., 2021). In addition, the measurement of the eWOM volume by companies is another recent research call (Dwivedi et al., 2021). Moreover, researchers have called for studies about emoji usage in both the real-world communication field (Bai et al., 2019) and among different contexts (McShane et al., 2021). In addition, there is also a recent call for research about the differences between countries in the use of social media (Dwivedi et al., 2021).

In this sense, this paper studies these differences in a particular B2B context as there is a lack of studies on this topic (Bai et al., 2019). Then, our findings will enhance our understanding of cross-cultural differences in the usage of emojis among countries with three different languages (English, Spanish and French).

To reach the objectives, the data was collected from X (formerly and still commonly known as Twitter), from 10 ITSs in France, Spain, the UK, Mexico and the USA (considering two ITSs per country). The tweet was used as the unit of analysis. In total 9,329 tweets were analyzed: 3,566 tweets ex ante COVID-19 and 5,763 tweets ex post COVID-19.

After a literature review from 1996 to 2019, Babić Rosario et al. (2020) define eWOM as consumer-generated, consumption-related communication using digital tools, primarily aimed at other consumers. Suggest that Social Capital Theory helps investigate eWOM formation and spread, particularly how online social network features influence eWOM. This theory emphasizes that connections within social groups generate relational resources, shaped by the valence (nature and quality) and volume (quantity) of shared content. eWOM thrives on social networks like Twitter, where users connect with people or organizations they do not know, spreading new information (Phua et al., 2017). Creating and sharing relevant brand content builds strong brand relationships (Hollebeek and Macky, 2019). Twitter, widely used by firms (Sridevi et al., 2020), is powerful for fast information dissemination (Barhorst et al., 2020) and promoting eWOM through retweets, which share content within users’ networks (Kim et al., 2014; Rivadeneira et al., 2021).

Grounded in Social Capital Theory, our study examines eWOM within the context of Twitter, defining eWOM volume as the number of retweets (Kim et al., 2014; Soboleva et al., 2017). Retweets are considered “another form of engagement” (McShane et al., 2021), making them a relevant metric for this analysis. This variable was selected as a dependent variable to be explained, given that it is an outcome of the online communication process in social media contexts such as Twitter (Son et al., 2019); the effect of the eWOM volume has a stronger impact on sales than eWOM valence (Babić Rosario et al., 2020) and researchers have called for studies about emoji usage in the real-world communication context (Bai et al., 2019) in different contexts (McShane et al., 2021).

In network communication, the emoji is used frequently (Bai et al., 2019), is part of digital content marketing (Holliman and Rowley, 2014) and represents a new form of interaction. Not in vain, in the academic field, the research about emojis is growing and is related to areas of communication and marketing, among others (Bai et al., 2019). Using emojis as a language transcends borders, being the fastest-growing language globally (Kerslake and Wegerif, 2017). Specifically, emojis increase customer engagement as a visual stimulus, complementing or replacing the written language (Valenzuela-Gálvez et al., 2022).

Many contents can be expressed using emojis (Bai et al., 2019). Specifically, eight categories are included in the emoji reference website: Smileys and People, Animals and Nature, Food and Drink, Activity, Travel and Places, Objects, Symbols and Flags (https://emojipedia.org/).

Furthermore, emojis have been used to analyze emotional associations with food and beverages (Jaeger et al., 2017).

Emojis have been analyzed following diverse theories. McShane et al. (2021) apply the elaboration likelihood model (ELM), suggesting that emojis act as peripheral cues that enhance message impact without requiring deep cognitive processing. Mladenović et al. (2023) combine the Dual Process Theory and emotional contagion to examine whether exposure to emojis increases purchase intention and campaign effectiveness. Almaguer et al. (2024) integrate Sign Theory with linguistics to position emojis as a significant evolution in modern language. This paper follows the ELM (Petty and Cacioppo, 1986), as used in previous studies (McShane et al., 2021; Luangrath et al., 2022).

Acknowledging social media’s role in promoting ITSs (Lapoule and Rowell, 2016) and the use of emojis in digital marketing (Holliman and Rowley, 2014), we predict emoji usage significantly influences eWOM volume in B2B contexts like ITS. Three key drivers of eWOM volume have been identified:

• emoji presence (McShane et al., 2021);

• emoji quantity (McShane et al., 2021); and

• time frame (pre- or post-COVID-19) (Das, 2021; Liu et al., 2022).

Based on this, we defend that emojis impact eWOM volume, influenced by these factors.

More and more brands use emojis (Cavalheiro et al., 2022) because, at a general level, their presence in promotional communications increases purchase intentions, at least for hedonic products (Das et al., 2019). Even more, emojis are positively related to brand engagement (McShane et al., 2021). Specifically, McShane et al. (2021) found that emojis in a tweet increase the eWOM volume. In addition, Cavalheiro et al. (2022) noted a greater acceptance of brands that use emojis when advertising on social media. Moreover, Liu and Jansen (2018b) proved that emojis can increase responses. In the food arena, Indwar and Mishra (2022) found a significant effect of emoji presence on eWOM, confirming the effectiveness of emoji presence in communication in the digital age without face-to-face communication. In the same line, Luangrath et al. (2022) demonstrated that the emojis’ presence acts as a visual stimulus that enhances information processing, which increases the probability of a message being shared in a B2B context (eWOM volume). Also, on Facebook, some authors found that emojis positively influence content sharing (Osorio Andrade et al., 2020).

In sum, based on the previous literature, we propose the following hypothesis:

Literature has proved that emoji quantity positively influences eWOM volume on Twitter (McShane et al., 2021). In other words, the more emojis a tweet has, the more retweets it will have (McShane et al., 2021).

In this sense, Pınarbaşı and Kırçova (2021) found that tweets in the Turkish food and beverage industry contained an average of 1.94 emojis. Pereira and D’Urso (2021) suggest that more emojis enhance message persuasiveness by making it more relatable and engaging. Similarly, Luangrath et al. (2022) showed that a higher emoji counts boosts emotional impact and engagement. Previous research also indicates that a larger number of emojis increases message attention.

In sum, the influence of emoji usage on the eWOM volume on Twitter can be analyzed through the amount of emojis (McShane et al., 2021). So, based on the previous literature, we propose the following hypothesis:

In the B2B context, the pandemic has posed challenges for firms, particularly in communication during crises (Crowe, 2012). Accurate information dissemination has become crucial, given the increasing influence of social media (Sharma et al., 2020). Although COVID-19 has driven eWOM (Dang and Raska, 2021), emojis and eWOM were already significant before the pandemic. Thus, the third objective examines the time frame, defined by the tweet’s publication date (Villamediana et al., 2019). Specifically, we compare tweets posted ex ante COVID-19 and ex post COVID-19, as Das (2021) demonstrated that emoji usage per tweet changed before and after the pandemic. Liu et al. (2022) also found a reduction in emoji use ex post COVID-19 in China and Ye and Ho (2023) noted that the COVID-19 time frame affects how messages are received and shared, impacting eWOM.

Based on literature and the justification of the pandemic situation, we propose the following hypothesis:

Cultural differences in Twitter users can be seen in emoji frequency (Li et al., 2019), and in the business field, culture affects how companies use social networks (Cummings and O’Neill, 2021; Dwivedi et al., 2021). eWOM and emoji usage vary across cultures (Dang and Raska, 2021), with studies showing that emoji use depends on culture and country (Krejtz and Borkowski, 2020; Cummings and O’Neill, 2021; Levi et al., 2024). Kejriwal et al. (2021) suggest some languages and countries use emojis more. Krejtz and Borkowski (2020) explore cultural influences on emoji usage, whereas Wang and Wang (2022) show how cultural context affects the relationship between emojis and eWOM volume.

In sum, the literature has analyzed the data in two ways: a global analysis of emoji usage and emoji usage by country (Das, 2021). Given this, this study will try to answer the research gap identified by experts, related to differences between countries in their use of social media (Dwivedi et al., 2021; Wang and Wang, 2022). We expect that differences exist between the countries in emoji presence, amount of emojis, time frame and eWOM volume. We also suppose that the country moderates the relationship between predictors and responses in our study. In this sense, our second and third hypotheses are the following:

Figure 1 represents all specific hypotheses in our study.

The data was collected from Twitter, using a tweet as the unit of analysis (Sridevi et al., 2020). Twitter is an appropriate platform for research because it is public, it is possible to investigate each tweet, the tweets can be taken simultaneously and it provides enough data for thorough analysis (Leek et al., 2019). ITSs were identified on Twitter through their handles beginning with the @ sign. All tweets and the retweet count for each tweet were collected from Twitter.

COVID-19 was declared a pandemic on March 11, 2020, by the World Health Organization (Chakraborty et al., 2020). The tweets were collected from June 21, 2019, to January 31, 2022, using the Twitonomy subscription service (Soboleva et al., 2017) offered by the website www.twitonomy.com. The periods analyzed were two, first, Period 1 from June 21, 2019, to March 10, 2020 (posting date ex ante COVID-19). Period 2 from March 11, 2020, to January 31, 2022 (posting date ex post COVID-19). Period 2 is 2.6 times that of Period 1 because the pandemic impacted the global service markets (Kabadayi et al., 2020).

Due to the characteristics of the data, we used non-parametric techniques to test the hypotheses, including contingency tables (chi-squared), the Kruskal–Wallis test and regression analyses (optimal scaling). For H1a–H1c, we used regression analyses with optimal scaling using the CATREG algorithm (Gifi, 1990), which quantifies categorical variables and optimizes regression coefficients for categorical data (Kooij, 2007). This method has been effectively applied in marketing studies (Villamediana et al., 2019) and offers advantages like not requiring residual normality, linear relationships or normal data distribution (Hartmann et al., 2009). For H2, to examine the relationship between emoji presence, quantity, time frame and eWOM in each sample and to analyze country-specific differences and moderation effects, we conducted multiple regression analyses with optimal scaling, following prior research (Wang and Chung, 2020; Hayes and Rockwood, 2017).

The food and beverage ITSs were collected from the website, www.feriasalimentarias.com, allowing free and up-to-date information on international food fairs. The ITSs were published on March 13, 2022, on the www.feriasalimentarias.com website. Convenience or directed sampling was used to select the ITSs from the www.feriasalimentarias.com website (Geldres-Weiss et al., 2021). In sum, the final sample included 9,329 tweets, from 10 ITSs in five countries (France, Spain, Mexico, the UK and the USA), considering the following criteria: ITSs with Twitter accounts; ITSs with a minimum of 300 new tweets; non-annual ITS were excluded; countries with less than two ITS were excluded; and it was selected the two ITSs with more emojis per country.

Following McShane et al. (2021), tweets were coded by emoji presence and the amount of emojis. To code the emoji presence, we adopted Unicode Full Emoji Data v3.0 as our list for coding emoji.

As shown in Table 1, we analyzed 9,329 tweets from 10 ITSs across five countries. We identified whether the tweets used emojis and the number of emojis used. Specifically, we found 370 emojis in the French sample, 1,060 in the Mexican sample, 527 in the Spanish sample, 740 in the British sample and 2,916 in the American sample.

We have also classified tweets into two categories according to their publication date (time frame): before the pandemic (ex ante COVID-19) and after the pandemic (ex post COVID-19).

Data was analyzed using SPSS, v.27. A statistical significance level of p-value < 0.05 (α) and a 95% confidence interval were considered to accept or reject the hypotheses.

First, we have analyzed the characteristics of our data set. To describe our data, we have divided the study’s variables into two categories: categorical and numerical. Table 1 shows the frequencies of emoji presence, the mean (average) of the amount of emojis by tweet, eWOM volume and the frequencies of time frame by country. France and Spain are the countries that most use emojis and the USA is the country that less uses emojis in tweets.

Then, we conducted an exploratory data analysis to check the data distribution of the response variable in the five countries. Table 2 shows the results of the Kolmogorov–Smirnov test, probing that the normality test was significant (all p-values are less than 0.01) and that our data is not normally distributed. Considering these results and the research objectives, we have selected the most suitable statistical analysis for testing our hypotheses.

As explained in the methodology section, analyses regression with optimal scaling, using the CATREG algorithm, was used to study the influence of emoji presence, amount of emojis and time frame on eWOM volume. The statistical assumptions of CATREG were checked (predictors = 3, n = 9,329, n > 3 + 1).

Before interpreting the results, we evaluated the intercorrelations among the predictors for both the untransformed and transformed predictors, checking the matrix of correlations and the tolerance before and after the transformation ( Appendix). As we expected, multicollinearity is not a concern in this study. In general, a low tolerance is indicative of multicollinearity. The tolerance of all predictors was higher than 0.21, with values above 0.10 (Yang et al., 2022).

According to the results, the regression model with optimal scaling, for the sample that included all countries together (Table 3), was statistically significant (p-value < 0.01). Specifically, the results revealed a highly significant and low correlation (R = 0.197) between eWOM volume and the best combination of emoji presence, amount of emojis and time frame. In total, 3.7% of the variance in eWOM volume is explained by its predictors [R² adjusted = 0.037, F (0.961) = 23.590, p-value < 0.01]. According to these statistically highly significant results, emoji presence, amount of emojis and time frame explain the eWOM volume on Twitter in the sample that included all countries together. Thus, we accept H1. More specifically the standardized coefficients (Table 3) and the mean graphs (Figure 2) reveal the following:

First, the emoji presence (β = 0.048, e = 0.021, F-value = 5.331, p-value < 0.01) influences positively the eWOM volume, supporting H1a. It means that tweets with emojis have more eWOM volume in the sample, including all countries.

Second, the amount of emojis (β = 0.143, e = 0.042, F-value = 11.725, p-value < 0.01) influences the eWOM volume (Table 3). Specifically, the categories with more emojis have more eWOM volume in the sample, including all countries. Therefore, H1b has been accepted.

Third, the time frame (β = 0.098, e = 0.011, F-value = 85.544, p-value < 0.01) increases the eWOM volume in the sample that included all countries together, supporting H1c. It means that tweets published before the COVID-19 pandemic have more eWOM volume.

Regarding a cross-cultural analysis, we run a regression model for each country to test H2. First, we have analyzed whether the behavior of the study’s variables is different in the five countries’ samples. To do the cross-cultural analyses, we followed the previous process. We have divided our variables according to the kind of data into two categories: categorical (emoji presence and time frame) and numerical (amount of emojis and eWOM volume) variables. Then, we have selected the suitable statistical analyses for each kind of data. Specifically, as Table 4 shows, we have run a contingency table analysis to evaluate the differences between groups (countries) in the categorical variables following previous research (Francesco and Roberta, 2019). In this sense, there are highly significant differences between France, Mexico, Spain, the UK and the USA in emoji presence (chi squared = 1333.86, p-value < 0.01) and time frame (chi squared = 707.01, p-value < 0.01). It means that emoji presence and time frame vary among countries. Thus, H2.1a and H2.1c have been accepted (Figure 1 and Table 9).

As Table 1 and Figure 3 show, France and Spain use more emojis than the other countries. Comparing the amount of emojis after COVID-19 and before COVID-19, France and Spain have published more tweets with emojis after COVID-19 than tweets without emojis. In addition, Figure 3 shows that only France decreased tweet frequency during the pandemic, whereas the other countries raised it during the same period. Thus, H2.1b has been accepted (Figure 1 and Table 9).

Then, we run a Kruskall–Wallis test to evaluate the differences between countries in the numerical variables (Sheraliev and Ślepaczuk, 2023). As Table 5 shows, there are differences between France, Mexico, Spain, the UK and the USA in eWOM volume (p-value < 0.01) and amount of emojis (p-value < 0.01). Specifically, Figure 4 shows that France and Spain behave similarly but differ from the other countries. In detail, France and Spain are the countries with the highest eWOM volume and the highest amount of emojis. Thus, H2.1d has been accepted (Figure 1 and Table 9).

The next table (Table 6) shows the results from regression models for each sample to test the specific hypothesis related to H2.1. The standardized coefficients (Table 6) and the mean graphs (Figure 5) reveal the following: First, the emoji presence, amount of emojis and time frame explain the eWOM volume on Twitter in the samples from France, the UK, Spain and the USA. More specifically, the eWOM volume is explained by its predictors in the samples from France with 7.0% of the variance [R² adjusted = 0.070, F (0.919) = 7.313, p-value < 0.01], the UK with 7.0% of the variance [R² adjusted = 0.070, F (0.923) = 12.297, p-value < 0.01], Spain with 1.2% of the variance [R² adjusted = 0.012, F (0.977) = 2.003, p-value < 0.05] and the USA with 0.6% of the variance [R² adjusted = 0.006, F (0.992) = 4.275, p-value < 0.01]. However, the sample from Mexico was not statistically significant (p-value > 0.05).

Second, regarding the impact of emoji presence on eWOM volume in each country, the results showed that only in the American sample did emoji presence significantly influence eWOM volume (β = 0.124, e = 0.036, F = 12.189, p < 0.01). This indicates that tweets with emojis generate more eWOM than those without emojis in the American sample. In contrast, emoji presence did not affect eWOM volume in France, Spain or the UK.

Third, regarding the influence of amount of emojis on each country, the results have revealed that this variable affects the eWOM volume in the cases of the French sample (β = 0.299, e = 0.068, F-value = 19.350, p-value < 0.01), the Spanish sample (β = −0.163, e = 0.023, F-value = 50.232, p-value < 0.01) and the American sample (β = 0.073, e = 0.026, F-value = 8.143, p-value < 0.01). In the British sample, the number of emojis does not affect eWOM volume.

Fourth, regarding the impact of time frame, the results show that this variable affects eWOM volume in the French (β = 0.065, e = 0.028, F = 5.541, p < 0.05), American (β = 0.070, e = 0.018, F = 16.729, p < 0.01) and British (β = 0.267, e = 0.028, F = 92.626, p < 0.01) samples. The behavior in France and the USA appears similar (Figure 4), with tweets from both countries published before COVID-19 showing an increase in eWOM volume. In contrast, the British sample shows a stronger and opposite trend, with tweets published after COVID-19 resulting in increased eWOM volume. In the Spanish sample, the time frame did not affect eWOM volume.

Moderation, or interaction, occurs when the effect of X on Y is influenced by W (Hayes and Rockwood, 2017). To test H2.2, we examined the interaction between country (our moderator) and predictors (emoji presence, emoji count and time frame) using multiple regression analysis with optimal scaling. We ran seven regression models, including country as an interaction term, one for each predictor and sample (Table 7). This approach identifies whether country moderates the relationship between predictors and eWOM, as shown in previous studies (Wang and Chung, 2020; Hayes and Rockwood, 2017). The results indicate significant differences in most relationships, confirming that country moderates these associations. Only the relationship between time frame and eWOM is similar in France and the USA. Therefore, H2.2b and H2.2c are accepted (Figure 1 and Table 9).

Tables 8 and 9 summarize the findings. As can be seen, the totality of hypotheses was accepted in the sample that included all countries together (n = 9,329). However, these relationships were not always significant in the subsamples (France, Mexico, Spain, the UK and the USA).

Our results align with previous studies showing that emojis can increase eWOM on social media (Cavalheiro et al., 2022; Indwar and Mishra, 2022; McShane et al., 2021; Osorio Andrade et al., 2020; Liu and Jansen, 2018b). Specifically, in France and Spain, more emojis led to more retweets (McShane et al., 2021). However, for Mexico, the USA and the UK, the relationship is non-linear, following a “U” shape – more emojis increase eWOM up to a point, after which it decreases. This supports Osorio et al.'s (2020) findings that excessive emoji use can overwhelm the message. The cut-off points for these samples are eight for Mexico, six for the UK and three for the USA. These results match Liu and Jansen's (2018a) suggestion of an optimal emoji use level, which varies by country.

Finally, consistent with previous research (Levi et al., 2024; Kejriwal et al., 2021; Li et al., 2019), we conclude that emoji use on Twitter varies by country. Some languages and cultures are more inclined to use emojis (Kejriwal et al., 2021), with the UK showing less emoji usage compared to other samples. As Hall (1976) suggests, high-context cultures (e.g. Spain and Mexico) rely more on contextual cues, which may explain the positive relationship between emoji use and eWOM in the French and Spanish samples. Indwar and Mishra (2022) also found that culture influences how consumers interpret and respond to emojis. For instance, Mexicans are more tolerant of expressive content, leading to a more favorable response to emojis, whereas UK and US consumers, with more restrained preferences, may experience quicker saturation.

The aims of this research were, first, to identify if emoji presence, amount of emojis and time frame (posting date ex ante COVID-19 or posting date ex post COVID-19) influence the eWOM volume at ITS organizers on Twitter in five countries. Second, to identify whether there are differences between countries in emoji presence, amount of emojis, time frame and the eWOM volume and moderation by country. The present research extends the digital content marketing literature by studying the emoji influence on eWOM volume in the B2B context in five countries, on Twitter. We tested 22 specific hypotheses (14 of 22 were accepted) (Figure 1).

From a theoretical approach, we can affirm that content quantity is shared in B2B contexts in social media. From a practice approach, we suggest that B2B companies incorporate emojis to drive digital marketing content nationally and globally. In this process, the B2B firms should consider the time frame as a predictor and the culture as the moderator of eWOM volume.

This research highlights the relevance of emoji presence, amount of emojis and time frame (posting date ex ante COVID-19 or posting date ex post COVID-19) in the production of eWOM volume in a social network such as Twitter. The findings show that all predictors positively influence the eWOM volume in B2B contexts. Furthermore, the study confirms the existence of differences between countries in terms of emoji presence, amount of emojis, time frame and eWOM volume.

From a general point of view, the findings show that emoji presence, amount of emojis and time frame explain the eWOM volume on Twitter in the sample that included all countries together (n = 9,329), supporting the emoji effectiveness in the B2B social network communications (Indwar and Mishra, 2022; McShane et al., 2021), contributing with one research related to the use of emojis in the real-world communication context (Bai et al., 2019) in different countries (McShane et al., 2021) and in the business contexts (Indwar and Mishra, 2022). In specific terms, the main contributions to the scientific literature are as follows:

Regarding emoji presence in the sample that included all countries together, the study confirms the effectiveness of emoji presence on eWOM in the business arena (Cavalheiro et al., 2022; Das et al., 2019; McShane et al., 2021; Osorio Andrade et al., 2020). Thus, this research highlights the relevance of emoji presence in the production of eWOM volume in social networks, confirming the power of Twitter to disseminate information (Barhorst et al., 2020); emojis like visual stimuli (Liu and Jansen, 2018b; McShane et al., 2021; Valenzuela-Gálvez et al., 2022); emoji presence in the food and beverages contexts (Jaeger et al., 2017); and emojis in social interactions, stimulating responses from users (Ge and Gretzel, 2018).

Regarding the amount of emojis in the sample that included all countries together, the results support that the amount of emojis in a tweet is relevant to produce eWOM volume (McShane et al., 2021). From a general point of view, it was found that the more emojis a tweet has, the more retweets it will have (McShane et al., 2021; Osorio Andrade et al., 2020) confirming the effects of different amounts of emoji in a message (Das et al., 2019; Osorio Andrade et al., 2020). In sum, For French and Spanish samples, more emojis are associated with more retweets. However, this relationship was non-linear for Mexico, the USA and the UK, where eWOM increased with emojis up to a point before decreasing. The cut-off points were eight emojis for Mexico, six for the UK and three for the USA.

Regarding the time frame in the sample that included all countries together, the study confirms the effectiveness of the time frame to increase the eWOM volume, given that the number of emojis per tweet changed due to COVID-19 pandemic confirming previous research (Das, 2021). Specifically, tweets published before the pandemic COVID-19 have more eWOM volume; on the contrary, the number of emojis decreased ex post COVID-19 (Liu et al., 2022).

Regarding the cross-cultural approach, from a general point of view, the results confirm the differences by country between emoji presence, emoji quantity, time frame and eWOM volume on Twitter. Thus, even emojis cross borders (Kerslake and Wegerif, 2017) in the business domain and influence the use of social networks by companies (Dwivedi et al., 2021); the specific context or culture influences the results of this nonverbal communication. Specifically, the UK used less emoticons than other samples.

Our study also found a country-moderating effect on almost all (except one) significant relationships. This country’s interaction was statistically significant. This means that the effect of emoji amount and time frame on eWOM varies among the five regions analyzed: France, Mexico, Spain, the UK and the USA. Only eWOM in France and the USA behaved similarly during the pandemic (both showed less eWOM than before).

Only in the USA sample, all predictors (emoji presence, number of emojis and time frame) had a significant influence on eWOM volume, even though the US sample had the least emoji presence, only 21% of tweets had emojis. In the case of the French sample, two predictors significantly influenced the eWOM volume, the number of emojis and the time frame. However, in two countries, only one predictor influences eWOM volume: in the case of the Spain sample, it was the number of emojis, and in the case of the UK sample, it was the time frame.

Regarding the time frame, the behavior of France and the USA were similar; that is, tweets published during the COVID-19 period increased the eWOM volume. However, only France reduced its tweet frequency during the pandemic, whereas the other countries increased theirs. In contrast, the British sample shows a stronger, opposite trend, with tweets published after COVID-19 resulting in higher eWOM volume.

Generally, we can conclude that emoji presence, amount of emojis and time frame are relevant predictors of eWOM volume in B2B contexts. Nevertheless, the culture moderates the eWOM volume behavior in each country (Dang and Raska, 2021). Specifically, in the business field, culture impacts how companies use social networks (Dwivedi et al., 2021) and affects eWOM volume. More importantly, the country’s influence is significant in B2B contexts. These findings highlight the effectiveness of emojis in driving eWOM in B2B settings. More specifically, the country’s moderation is significant in B2B contexts. Overall, these findings demonstrate the usefulness of emojis in generating eWOM in B2B contexts.

At a general level, the results show that B2B companies should integrate emojis into social media communications, given the effectiveness of the presence of emojis and quantity on the eWOM volume production (Indwar and Mishra, 2022; McShane et al., 2021), confirming the usefulness of social networks to promote ITSs (Lapoule and Rowell, 2016).

Regarding the presence and quantity of emojis, our results suggest that it is useful to introduce emojis into messages on Twitter, at least up to a certain quantity that varies between countries. Specifically, more emojis are always merrier for French and Spanish audiences. However, there are thresholds beyond emojis becoming less effective for Mexico (six), the UK (six) and the USA (three). These findings are helpful for practitioners.

In terms of the time frame, it was found that it is significant to the production of eWOM for France, the UK and the USA, but its effects differ among regions. Specifically, only France and the USA behaved similarly during the pandemic, but other relationships changed depending on the sample. This means that considering differences by time frame (the context in which communication occurs), culture and country is very important for increasing the effectiveness of strategies on Twitter. The results showed that a crisis like COVID-19 hurts eWOM in many countries. As COVID-19 stages also hide other variables (such as the differences between formats, presence vs online and the use of different strategies to approach audiences), these hidden factors could explain the results. In this sense, the data could suggest that ITS clients prefer face-to-face trade shows.

The data was carefully collected, coded and analyzed in our study. Our findings are valid and other authors can replicate the study. However, our research also has limitations. The main limitation is that only one industry and one social network were analyzed. Consequently, the findings should not be generalized to contexts with different settings from this study. We suggest new studies applied across diverse sectors, platforms and countries. Finally, the usage of emojis in other B2B firms can be studied.

Furthermore, Table 10 illustrates conclusions, theoretical and managerial implications.

Declarations. There is no conflict of interest.

This study has not received any funds.

All the authors have contributed to its elaboration.

Data can be available under the petition.

Table A1