Social Links vs. Language Barriers: Decoding the Global Spread of Streaming Content

TL;DR

This study investigates how social connections and linguistic similarity govern the global diffusion of streaming content across Netflix (video), Spotify (audio), and YouTube (user-generated video) using two-year trending data from 10 countries. It operationalizes socio-cultural distance with Facebook SCI and language lexicon similarity, and models content lifetime with five distributions via maximum likelihood, revealing platform- and content-type–specific diffusion rules. The key finding is that audio-oriented content tends to diffuse via social ties, while video-oriented content is more constrained by language, with YouTube displaying a dual pattern due to its mixed content and prosumer dynamics. These insights have practical implications for localization, recommendations, and virality forecasting across global streaming platforms.

Abstract

The development of the internet has allowed for the global distribution of content, redefining media communication and property structures through various streaming platforms. Previous studies successfully clarified the factors contributing to trends in each streaming service, yet the similarities and differences between platforms are commonly unexplored; moreover, the influence of social connections and cultural similarity is usually overlooked. We hereby examine the social aspects of three significant streaming services--Netflix, Spotify, and YouTube--with an emphasis on the dissemination of content across countries. Using two-year-long trending chart datasets, we find that streaming content can be divided into two types: video-oriented (Netflix) and audio-oriented (Spotify). This characteristic is differentiated by accounting for the significance of social connectedness and linguistic similarity: audio-oriented content travels via social links, but video-oriented content tends to spread throughout linguistically akin countries. Interestingly, user-generated contents, YouTube, exhibits a dual characteristic by integrating both visual and auditory characteristics, indicating the platform is evolving into unique medium rather than simply residing a midpoint between video and audio media.

Figures (13)

• Figure 1: Numbers of shared trending content between countries: (a) YouTube, (b) Spotify, (c) Netflix Film, and (d) Netflix TV show. Each point is colored according to the number of shared contents in a log scale (see the color bar). The labels of the color bar correspond to the quartiles for each platform.
• Figure 2: Cross-platform comparison for the numbers of co-trending contents between countries. The orange solid line represents a linear regression line between two platforms measured in a log-log scale ($y \sim x^k$), where we also measure the coefficient of determination ($R^2$) and Spearman correlation ($\rho$). The annotated p-value represents statistical significance tests for both $R^2$ and $\rho$. (a)--(c) While YouTube shows a high $R^2$ ($>0.6$) with every other platform, (d)--(e) Spotify and Netflix have a relatively lower coefficient of determination between them. (f) It's interesting to note that the relationship between Netflix's Film and the TV show has a lower $R^2$ than their relationship with YouTube. We observe similar patterns when we take into account the Spearman rank correlation.
• Figure 3: The correlation between the number of shared trending contents and two proxies of social similarity: (a)--(d) Facebook Social Connected Index bailey2018social and (e)--(h) language similarity (see Methods). The orange solid line represents a linear regression line between two platforms measured in a log-log scale ($y\sim x^k$), where we also measure the coefficient of determination ($R^2$) and Spearman correlation ($\rho$). The annotated p-value represents statistical significance tests for both $R^2$ and $\rho$. Spotify shows a comparatively stronger $R^2$ for social networks (Facebook SCI) than linguistic similarity (compare (b) with (f)), yet linguistic similarity displayed a greater $R^2$ for Netflix (compare (c)--(d) with (g)--(h)). YouTube shows high $R^2$ for both proxies of social similarity ((a) and (e))
• Figure 4: Interrelationship between the correlation of the trending videos on YouTube with the Facebook SCI and the language similarities across categories. For (a), the x-axis represents the $R^2$ of the number of shared trending videos to the Facebook SCI between countries, whereas the y-axis represents the $R^2$ between the number of shared trending videos to the language similarity. Panel (b) shows similar relations using the Spearman rank correlation instead of $R^2$. For panels (a) and (b), the color of circles represents the category groups: Musics categories (green), Games categories (purple), Sports categories (blue), Visual arts (red), and others (grey). The abbreviation and full name of each category are described in the legend. See Supplementary Table \ref{['tab:tab1']} for the full list of categories and Tables \ref{['table:s6']} and \ref{['table:s7']} for the detailed statistics. The diameter of the circles corresponds to the number of country pairs that have mutually shared trending videos. The dashed line in (a) and (b) represents a median value of language lexicon similarity and SCI of all categories in the data. Panels (c)--(f) display the scatter of the number of shared trending videos with Facebook SCI ((c) and (e)) and language similarity ((d) and (f)), respectively, for two example categories: (c)--(d) Music and (e)--(f) Film.
• Figure S1: The dendrogram for the complete-linkage cluster of Figure \ref{['fig:fig1']}: (a) YouTube, (b) Spotify, (c) Netflix Film, and (d) Netflix TV show. The cluster was optained using the Scipy's scipy.cluster.hierarchy.linkage with following parameters: method='complete', metric='euclidean', optimal_ordering=False)