Keywords

1 Introduction

Investment in streaming or over-the-top (OTT) services has grown exponentially in recent years [1] with the USA being the industry leader with more than 200 companies offering these services [2], although the industry is actually led by companies such as Netflix, Amazon Prime Video, Hulu and Disney+ [3].

An important part of the success of these platforms has been the way they communicate and relate to their users by achieving the integration of online and offline experiences with brands [4] and creating relevant content [5], being the most demanded, series, films and TV programs [6]. In this sense, one of the keys has been the implementation of promotion and positioning strategies [7], which allow digital platforms to change the forms of relationship and interaction of consumers towards brands [8] enabling their target audience to go from being mere observers to active participants [9] which allows engagement [10].

Social networks have become the main platform for brands to interact [11] and generate engagement with their consumers [12,13,14] by providing relevant content [15] on social networks [8] that get customers to interact with them and contribute in their dissemination and advocacy [16].

Regarding engagement, this can be understood as those behaviours that the customer has with the brand, which go beyond the purchase of the product [17], where brands should also predefine what type of Customer Engagement Behaviour they want to obtain from their customers through the generated content [18], since as it is known, this can vary [19].

Its research aims to contribute to an academic research priority in marketing and the social sciences in general [20]. In the particular case of the present research, the study of engagement is focused on its generation through social networks called Social Media Engagement Behaviour (SMEB) [21] under the view that brands create content that motivates users to be engaged [22] since as is known, there is a relationship between the types and variables of content generated on social networks with the type of response or engagement behaviour of users [21]. Specifically, this research goes beyond previous studies, as it aims to compare whether the Social Media Content (SMC) created by Netflix on its corporate page and the SMC created, also by Netflix, on the Facebook pages of its own series, generate different SMEB in their followers [23]. This article is structured as follows. First of all, it contains an introduction. Then a literature review and hypothesises are based. The third section includes the methodology. Then the results and the discussion are developed.

2 Materials

2.1 Social Media Content (SMC)

Companies create different contents to attract or retain users [24], SMC being the main element to achieve a successful digital marketing [25]. These contents, which are expressed in different forms such as photos, videos, and texts [26] can be classified into informative content, entertaining content and relational content [27] being rational or emotional in nature [22].

Posts with informative content are characterised by being rational in nature [28], as they aim to communicate functional characters about a brand, product, or service relevant to users [29], such as general product information, product mentions, events, or information about product availability [22], motivating users to interact with the brand and facilitating engagement [30], which generates one of the highest possible benefits for a brand's online communities [31].

Regarding entertaining posts, these seek to be fun for users [21] and to be perceived as expressive in their communication [32], appealing to their emotional side [22] to generate greater interaction, attention and a better digital experience [33] to achieve greater engagement [34]. It should be noted that these contents often influence brands’ online communities [35] through memes, colloquial jargon, images with celebrities or post descriptions with emoticons and/or interesting images linked to the brand that generate fun facts [22].

While relational posts may be similar to entertaining posts [29], since both appeal to emotions by using posts that encourage activities among users [22], it is worth highlighting that relational content focuses on attracting users to brands [36] by fostering users’ social and psychological integration satisfaction through content generated by brands in social, by filling the need to gain a place of belonging in communities related to them [34].

2.2 Social Media Engagement Behaviour (SMEB)

The SMEB is behavioural manifestations that users externalise in social networks, generated by motivations commonly represented through participations and interactions [8] that can be seen not only by brands, but by other users within the same online space [19] and generate a link between brands and consumers [34] and finally increase the purchase behaviour [37].

Within the SMEB typology, co-creation is the highest and most beneficial level for a brand [22] as users increasingly seek to exert their influence on every aspect of the brands they follow [38] through the creation of content related to the brand and other members to the brand [21] such as creating stories, comments or content that reflect interactive behaviours that facilitate people’s learning about brands [39] with the aim of manifesting their creativity and expressing their opinion or ideas about a brand [49].

On the other hand, contribution, another SMEB typology, is defined as sets of moderate, positive and active behaviours, characterised by contributing to the diffusion of a brand in social networks [22], which can be generated between user and brand as user-user about the brand [31], due to users leaving their role as “observers” and becoming “collaborators” [41].

Meanwhile, the consumption typology corresponds to the minimum positive and passive level, since users consume the content generated by brands [21], limiting themselves to using social networks with the sole intention of obtaining some benefit from the information provided, without any form of participation and interaction with it [42]. This type of behaviour is evidenced by actions such as reading posts and viewing images or videos about a brand [43].

Also, recent literature demonstrates the existence of a significant influence on outcome analysis when contrasting different contexts in social networks [44], including other countries [45] and cultures [46]. Moreover, some research confirmed how the SMC is directly related to engagement behaviour [47], impacting brand health [35, 48], using different social media platforms such as LinkedIn [49], Facebook [50] or Instagram [51]. Moreover, previous findings conclude that the source of the Social Media Content influences users’ behaviour to a variety of engagements [52], willingness to share recommendations [53], purchase intention, and eWOM [23]. Furthermore, the relationship between social media engagement and firm content was supported in the media industry [54]. Based on the above, two hypotheses are formulated.

Hypothesis 1—H1. The Social Media Content generated by Netflix's official Facebook page has a positive influence on the Social Media Engagement Behaviour.

Hypothesis 2—H2. The Social Media Content generated by the official pages of the Netflix series on Facebook has a positive influence on the Social Media Engagement Behaviour.

3 Methodology

To validate the research hypotheses, data collection was carried out through two online surveys, using two self-administered online questionnaires for one month in the second half of 2021. The study population was composed of Netflix subscribers in Peru who follow Netflix’s official Facebook or any of the different official Facebook of the Netflix series. Facebook was chosen because it is the social network where people usually search, share, and recommend content about series and streaming platforms [55]. Data was collected through rigorous ethical protocols to guarantee the respondents’ anonymity and the answers’ reliability. Furthermore, a filter question in each questionnaire was used to ensure that only followers of the official Netflix Facebook page or the official Facebook pages of the Netflix series answered the surveys. The convenience sample (non-probabilistic sampling procedure) for the present study was 284 people in total. 122 users in the sample indicated that they followed the official Netflix Facebook—Sample 1, while 162 people said they followed the official Facebook of Netflix series—Sample 2. The results of an independent t-test confirmed that the sample characteristics of both samples were similar. The questionnaires were divided into two sections. In the first section, demographic data from the followers as gender and age was collected. In sample 1, 45% of the respondents were male, 55% were female, and 92% were under 35 years old. In sample 2, 99.4% of the respondents were under 35 years old, 56% were female, and 44% were males.

Surveys were conducted online because many people can be accessed at a low investment, response time is shorter, survey length is more accurately addressed, and there is great flexibility and efficiency in management [56].

In the second part, two scales were constructed based on previous research. The first was used to assess the constructs in Netflix’s official Facebook followers (Sample 1). In contrast, the second instrument was used to collect responses from official Facebook followers of the Netflix series (Sample 2). The scales used to measure the variables came from previous literature, adapting scales other authors had already tested in previous studies. These scales were written in English, so they were translated into Spanish so that they could be applied to the study population. All the items in the questionnaire were questions based on a Likert scale from 1 to 5 (1 being totally disagree and 5 totally agree). The scales were pretexted to confirm that the items were correct and only minor corrections were made. The pre-test sample was 15 respondents.

The multidimensional scale of Vale and Fernandes [34] and Schivinski et al. [43] was adapted to measure Social Media Engagement Behaviour (SMEB), based on the dimensions of consumption (four items), contribution (four items), co-creation (four items) validated by Schivinski et al. [41]. Also, the dimensions of entertaining (one item) and informative (one item) content types, originally designed by McQuail [34], were adapted from Vale and Fernandes [57]. One item was taken from Azar et al. [36] on entertaining content, which was adapted from Shu and Chuang [58]. The Fernandes and Castro [42] scale was used to measure the dimensions of informative content (two items) and entertaining content (one item); these were previously adopted by Baldus et al. [59] and Gummerus et al. [51]. Similarly, the approach in De Vries et al. [61] was used with the dimensions of informative content (one item) and entertaining content (one item) based on Muntinga et al. [31]. Relational content was measured employing the three-item multidimensional scale proposed by Helme-Guizon and Magnoni [62], which was adapted from Allen and Meyer [63].

For the present investigation, a PLS-SEM analysis was performed using the Smart-PLS 3 tool [64]. Due to the ability to analyse highly complex, multivariate models with many indicators, in addition to its great statistical scope, PLS-SEM is positioned as the appropriate method for the analysis of social networks [65].

Considering that the two study constructs are multidimensional, since the independent variable (SMC) is formed by the dimensions of informational content (IC), entertaining content (EC) and relational content (RC) and the dependent variable (SMEB) by the dimensions of co-creation (CC), consumption (CS) and contribution (CTR), a second-order analysis was carried out, where the indicators of the first-order latent variables were used to construct the variables SMC and SMEB [66], both of second order. SMC and SMBE being multidimensional variables of reflective character, the model validation was done in three steps, the measurement model of lower-order reflective constructs, the measurement model of higher-order reflective constructs and structural model assessment [67].

4 Results

The PLS-SEM results were evaluated according to the embedded two-stage approach of hierarchical models with Mode A [67]. Likewise, two different samples were used on the structural model; these differ only in the approach of the independent variable. Sample 1 focused on content generated by Netflix's official Facebook page, while Sample 2 focused on content generated by the official Facebook pages of the respondents’ favourite Netflix series.

The first step followed was to check the measurement model of lower-order reflective constructs [68], for which Cronbach’s Alpha (α), Rho_A, composite reliability and the convergent validity (AVE) were calculated. Analysing the results of the composite reliability assessment [69], the IC, EC, RC, CTR, CC, and CS constructs were used for Sample 1—Netflix (Table 19.1) and Sample 2—Series (Table 19.2). The results of the analysis showed that both constructs and samples satisfactorily reached the level of composite reliability, given that they are within an acceptable range between 0.70 and 0.95 [70]. Continuing with the reliability test, Cronbach’s Alpha, of both samples (1 and 2), yielded values between 0.80 and 0.89, being considered acceptable, because an Alpha value between 0.7 and 0.9 is a good internal consistency [65, 67]. Regarding convergent validity, for both samples, the IC, EC, RC, CTR, CC, and CS constructs were used. The AVE values are above 0.5, which indicates a convergent validity, by exceeding 50% of the variation of its elements [67].

Finally, the discriminant validity results were analysed using the HTMT criterion, which shows that the values are below the established threshold [71, 72].

The second step was the measurement model of high-order reflective constructs. The analysis results indicate that the reflective measurement model meets the relevant criteria required for both samples. The model was built, initially, based on the analysis of the factor loadings of the indicators related to each of the dimensions in order to assess the reliability of the internal consistency of the latent variables, and subsequently, the causality between the exogenous and endogenous constructs to test the hypotheses raised [73]. Regarding the external loads, Sample 1 (Fig. 19.1) obtained external loads between 0.810 and 0.925, while Sample 2 (Fig. 19.2) obtained external loads between 0.785 and 0.897.

Both samples obtained results above the threshold of 0.7, which reveals that the indicators show a satisfactory level of reliability [67, 68]. The next step is the internal consistency reliability; first of all, following the results of the composite reliability assessment [69], the SMC and SMEB constructs were used for both samples. Sample 1 showed the values of 0.915 and 0.899, respectively, while Sample 2, the values of 0.884 and 0.875. These results represent the degree of satisfaction in the level of reliability with respect to the composite reliability, since they are in the satisfactory range of between 0.70 and 0.95 [66, 73]. Likewise, continuing with the internal consistency reliability test, both samples (1 and 2) were subjected to the Cronbach's Alpha test, whose results were found to be within the range of 0.785 and 0.861, which is considered acceptable, since an Alpha value between 0.7 and 0.9 indicates good internal consistency [74]. For the elaboration of the convergent validity, both samples used the SMC and SMEB constructs, as a result, the corresponding AVE values of Sample 1 were 0.717 and 0.700, respectively, while for Sample 2, they were 0.783 and 0.749. These results are above 0.5, which indicates that there is a convergent validity, since both constructs show more than 50% of the variation of their elements [65, 70, 73] (see Table 19.1).

Table 19.1 Reliability and validity statistics—high-order reflective constructs
Full size table

Accordingly, the discriminant validity was worked through the HTMT criterion; for this, the different indicators of the research model were grouped, thus applying a second order for both samples. In this way, Sample 1 obtained a value of 0.890, while Sample 2 obtained one of 0.836. With these results, the discriminant validity reaches a suitable level of satisfaction, since it has not exceeded the threshold of 0.90 [71] (see Table 19.2).

Table 19.2 HTMT criterion—high-order reflective constructs
Full size table

Subsequently, the development of criteria for the structural model was preceded with. Regarding the collinearity of the model, the PLS-SEM analysis had the results by means of the quotient of the variance of this or VIF. Sample 1 obtained values within a range of 1865 and 2798, while Sample 2 obtained values between 1401 and 2369. As a result, it is concluded that the levels of collinearity of the indicators of the research model are not critical, since they are below 5 [65, 71, 73]. Continuing with the evaluation of R2, the SMEB dependent variable of the research model was submitted in both samples (1 and 2). Sample 1 obtained an R2 of 0.595 in its dependent variable, while Sample 2 obtained an R2 of 0.450. According to these results, the proportion of the variation of the dependent variable SMEB, explained by its independent variable SMC, can be considered moderate [70, 71, 75] (see Figs. 19.1 and 19.2).

Fig. 19.1
A flowchart for the P L S, S E M of sample 1 results. S M C N points to C E N via 0.922, to C I N via 0.835, and to C R N via 0.896 on one side and points to S M E B N 0.595 via 0.772 on the other side. S M E B N 0.595 points to C C N via 0.810, C N via 0.858, and C T R N via 0.925.

Results of the PLS-SEM of sample 1. Followers of the official Netflix page

Full size image
Fig. 19.2
A flowchart for the P L S, S E M of sample 2 results. S M C S points to C E S via 0.897, to C I S via 0.796, and to C R S via 0.846 on one side and points to S M E B S 0.450 via 0.671 on the other side. S M E B S 0.450 points to C C S via 0.824, C S S via 0.785, and C T R S via 0.897.

Results of the PLS-SEM of sample 2. Followers of the official pages of Netflix series

Full size image

Afterwards, the data was subjected to a Bootstrapping of 5000 samples, without the option to change the sign, Bootstrapping confidence interval (BCa), with a significance level of 0.001. The results of each sample validate hypotheses H1 and H2, proving that both samples comply with a value less than 0.001 because they have a p-value (ρ-value) of 0.000 (see Table 19.3). The H1—SMCNETFLIX → SMEBNETFLIX had a β of 0.772. Meanwhile, the H2, that analysed the relationship of SMCSERIES → SMEBSERIES showed a β of 0.671 (see Table 19.3).

Table 19.3 Standardised path coefficients, R2 and hypothesis testing
Full size table

The next step was to establish the effects of f2 for both samples; Sample 1 obtained an f2 of 1.472, while Sample 2 obtained an f2 of 0.818. These values indicate that there is a higher effect of the latent exogenous variable, since the value is well above 0.35 [76]. Regarding the evaluation of Q2, the dependent variable SMEB was subjected to this test in both samples (1 and 2). Sample 1 obtained a Q2 of 0.427 in its dependent variable, while Sample 2 obtained a Q2 of 0.304; these results indicate that the prediction accuracy of the model is acceptable, since the values are greater than zero [65, 67, 71].

5 Discussion

This research shows several contributions to the research framework on consumer behaviour in the context of social networks, as suggested by previous research from Cao et al., Creevey et al., Dolan et al., Schivinski et al., Vale and Fernandes [22, 77], but on this occasion, a reliable entertainment technology company such as Netflix has been taken as the main context within a growing industry which is highly valued by users such as video streaming. The first contribution of the study has been the comparison that the impact of the SMC on the SMEB, analysing two types of followers of the same company, is different; unlike previous research that evaluated the possible causes and consequences of the impact of the SMC on the SMEB in other industries [8, 34], this research has addressed the importance of the context in social networks in a new way, since the comparison of the SMEB achieved through the SMC of the products, in this case, the series, was made against that obtained by the company, Netflix. The first group of followers evaluated is made up of followers of Netflix series fan pages, while the second group is made up of followers of the official Netflix page; the study was carried out on Face-book. The second contribution refers to the verification that, indeed, the SMEB is different depending on the entity that generates the content. According to the results of this research, there is a greater impact of the SMC on the SMEB in users who interact with the official Netflix page on Facebook, than in those users who follow the content of their favourite Netflix series. Taking into account that both the SMC of the series pages and the official page itself are created by Netflix, it could be deduced that users prioritise the general and diverse content created by Netflix as a brand, over the content created for one of its series in particular, probably because users value Netflix’s SMC as a whole and not in a specific way when creating SMEBs.

This is an important finding because although some Netflix series are top-rated, the research results show that the content created by Netflix on its official fan page is more effective in getting followers to generate SMEB. This may be due to the fact that although some of the series become very popular, their popularity is limited to brief periods of time [78], commonly coinciding with the time that elapses from the launch of the new chapters until the followers watch the entire season, which makes it challenging to generate SMEB with the series in the long run. On the contrary, the results allow us to affirm that the constant generation of digital content on Facebook by Netflix makes its followers more predisposed to generate SMEB, sharing, cocreating, or consuming the content offered by Netflix on its official Facebook fan page. This may be because content consumption on streaming platforms, such as Netflix, is not limited to a single series; on the contrary, the subscribers of these platforms seek extensive and varied content, which is constantly renewed. This statement is especially true of users who generate SMEB since they are heavy users of Netflix in most cases.

Furthermore, the research also supposes a methodological contribution to the study of said variables [67] since the research proposal was analysed as a second order since both constructs were multidimensional, contrary to the previous researches where the dimensions of SMEB and SMC were considered constructs and studied as a first PLS-SEM.

Also, the results have straightforward practical implications for companies that offer streaming services, such as Netflix. As verified in the results, there is a clear relationship between SMC and SMEB. This implies that the strategies carried out by Netflix through social networks, specifically Facebook, are effective in getting users to be willing to create content, share information or consume it. As is known, nowadays, getting users to generate SMEBs is crucial for companies in general since achieving engagement increases the profitability of companies [79] and the purchase intention, mainly because users share their resources selflessly with the company, in the specific case of this investigation, with Netflix. At the same time, the results allow us to corroborate the importance for Netflix to frequently generate informative content while entertaining the followers and fostering relationships. This implies that the SMEB is not generated spontaneously; on the contrary, Netflix users react to digital marketing actions on social networks through the SMC. It should be noted that these findings allow companies that offer streaming services to understand how to optimise their digital content strategies by demonstrating the close relationship between SMC and SMEB [80], which can undoubtedly be of great help in an increasingly competitive environment and with smaller marketing budgets.

It is also worth noting that this research clarifies the debate on whether it is more important to have one or two high-impact series or, on the contrary, Netflix users prefer vast and diverse content to view. The results allow us to understand that although SMEB can be achieved through SMC strategies in the official fan pages of the most popular Netflix series, it is undoubtedly more effective to implement SMC strategies on the official Netflix fan page promoting the full content of the streaming platform. In this sense, it can be affirmed that the users of a streaming platform are willing to generate SMEB and share their resources. Still, for this, platforms such as Netflix must have an immense catalogue of series and programs that allow them to create informative, entertaining content with the ability to go viral and reinforce brand equity.

The present investigation has a series of limitations, which can be corrected in future investigations. In the first place, the information gathering of the study was carried out clearly in Peru, which, although correct, limits the extrapolation of the findings to other realities; in this sense, the research could be replicated in other South American countries in order to universalise the results. Second, the research focused on the most important streaming brand, Netflix, but it is suggested that other researchers expand the number of streaming platforms, as it would serve to obtain a broader view of the impact of SMC and SMEB on the context of companies capable of producing high-value audio-visual content for users in general. In third place, it would be significant to expand the range of social networks of the study because, currently, digital platforms such as Instagram, TikTok, or the future web 3.0 [81], are becoming more popular than Facebook, and SMC and SMEB are likely to be produced differently. Furthermore, the expansion of the complexity of the model with variables related to the content generation strategy is suggested, such as the time or day the content was published, since it could generate interesting results on engagement behaviour [22]. Likewise, the expansion of the research with more current measurement instruments in the context of social networks and the digital field, in general, is suggested [41], such as Big Data or the blockchain. Finally, studying the role of influencers in the generation of Social Media Content can be considered a future research priority.