“When pigs fly”: Resources swapping,affordable marketing,and the transformation of Douyin from short video sharing to online shopping

Abstract

With 600 million daily users in 2020, Douyin faced monetization challenges after increased advertising led to a decline in user engagement. The app pivoted successfully to “shoppable videos” and livestream shopping by operating as a retail infrastructure. This article analyzes this transformation, arguing that Douyin strategically limits data availability, creating an artificial scarcity for capturing advertising and infrastructural rents. This approach is sustained by Douyin’s carefully crafted vision for career and business success derived from data’s speculative value. The app established reciprocal relationships with music creators and later video creators and sellers through reward programs, contractual deals, and online workshops. However, success often remains unattainable due to the uneven growth between continuous influx of creators/sellers and stalled user growth. This disparity allows Douyin to incorporate paid traffic into its business model, perpetuating data’s manufactured scarcity and obscuring exploitative practices by making paid traffic an essential component for success.

Introduction

In China, the expression “even a pig can fly if it stands at the center of a whirlwind” (zhan zai feng kou shang de zhu dou neng fei qi lai) is used to explain unlikely success in the rapidly changing landscape of the Internet. First mentioned by Lei Jun, the founder of Xiaomi, the narrative was used to describe Xiaomi’s ascent to the world’s most valuable technology start-up within 4 years of its founding (Price, 2015). Dubbed the “flying pig theory” (fei zhu li lun) by media commentators, the wide-spread belief relates to the entrepreneurial spirit surrounding big tech platforms among ordinary people and small businesses.

Since their founding in 2016, Douyin and its twin platform TikTok have brought a considerable number of early-career artists and creators success through their interactive features, including lip-syncing, dances, duets, and covers. Douyin attained 600 million daily active users in 2020 (ByteDance, 2021). User engagement activities and time spent (i.e. digital traffic) on the app have been celebrated as an Internet whirlwind for creators and businesses alike. When looking at Douyin since 2020, one can observe that short videos are no longer the app’s primary offering. Livestreams, digital payments (Douyin Pay), and “Shop” constantly appear alongside the “For You Page.” While media hype surrounding Douyin’s fortunes abound, replicating such success has become increasingly difficult due to the commercialized data supply that is driven by growing influx of new entrants from various sectors.

This article argues that the carefully crafted vision of investing in future growth compels businesses and careers to transition into big tech platforms. This shift is largely motivated by data’s speculative value based on anticipated growth at the expense of current gains (Davis, 2018). Scholars have argued that revolutionary visions surrounding the Internet are historically rooted in its long-standing business model that monetizes free resources (Srnicek, 2017) and unpaid user labor, which have been theorized as the audience commodity (Smythe, 1977), and later the Internet prosumer commodity (Fuchs, 2014). Propelled by ideologies such as individual freedom and community, as well as cultural values like participation, creative expression, and openness (Fuchs, 2014; Stevenson, 2018; Van Dijck, 2013), the portrayal of big tech platforms as transformative forces remains relevant in understanding their data-driven business model.

Using the case of Douyin’s transformation from short video sharing to online shopping, this article examines the company’s carefully crafted vision of career and business success associated with platform data (including audience data, content data, and transactional data), the supply and demand of which are now controlled and commercialized through its stratified traffic distribution mechanism, informally known as the “traffic pool.” This indicates that platform data have become means of extracting value in the circulation market through creating advertising space and business infrastructure (Fuchs, 2014; Sadowski, 2020; Srnicek, 2021). On Douyin, newly published videos undergo an initial assessment with a limited number of viewers (i.e. initial traffic pool). If the engagement surpasses certain thresholds, the video enters a larger traffic pool. Conversely, if the performance falls short, the app stops recommending the video entirely, even for a small group of viewers whose preferences may align with the content. Douyin now directs a substantial amount of viewing traffic toward content that is “popular among a wider range of audiences” (TikTok, 2020).

While scholarly attention has focused on personalized recommendation algorithms, the highly opaque, and by implication, manipulative trending algorithms that Douyin uses for building audience reach have been sidetracked. Trending algorithms center on the measure of popularity, thus they involve a lot of guesswork (Gillespie, 2016). For Douyin, although the “traffic pool” is organized by trending algorithms, it also requires constant human decision-making to select content that can appeal to mass interests, on top of individual-based, more dispersed recommendations that reflect personal preferences. This is because the scale effects of mass popularity (i.e. virality) accelerate user acquisition and retention.

Since Douyin’s viewing traffic is concentrated on select content, competition for edging into the “For You Page” has intensified (Zhang et al., 2021). The intensified competition allows Internet companies to control data supply for rent capture (Howson et al., 2023; Van Doorn and Badger, 2020). In the case of Douyin, the app strategically captures value by manipulating traffic availability, compelling creators and businesses to purchase viewing traffic. The traffic progression system, designed for data allocation for the circulation market, serves as a central mechanism to coordinate automated and human control for value capture. In contrast to the extensive scholarly examination of the automated aspects of the platform economy, the operational dimension—the human control behind the technical interface—has been largely underexplored, in part due to a lack of information from business insiders. This dimension, however, is crucial for understanding the demand-driven, commercialized data supply. This is because data’s monetary value is often measured by their (more controllable) circulation rather than their (more automated, due to machine learning) production (Fuchs, 2014).

The article makes two main theoretical contributions. First, by examining human control enhanced by automated technologies (i.e. proprietary algorithms) in transforming data’s speculative value into monetary value, it complements the algorithm-centric research paradigm in existing platform economy scholarship. Second, it advances understanding of the audience commodity theory in the context of Douyin’s transformation: audience commodity now takes on a technical form of “traffic” which has been the object of automated and manual control. This technical form of audience commodity reinforces both the speculation of data value and the vision for career and business success while trivializing the inherent exploitative nature of the platform economy.

Based on qualitative data of Douyin’s business-facing subsidiary platforms that sell traffic, along with interviews with Douyin staff, sellers, MCNs (i.e. multi-channel networks), marketing firms, and influencers, this article delineates the development of Douyin into three key stages, each characterized by a crafted vision for career and business success and paired with reward programs, contractual deals, and online workshops. The first stage is to make copyrighted music freely available for user acquisition through its musician program “Douyin Spotlight” (kan jian yin yue ji hua). The second is to energize a creator culture by means of the Creator Reward Program (chuang zuo zhe cheng zhang ji hua). After successfully onboarding a significant number of talented creators, Douyin started to scaffold their audience reach to align with businesses with varying marketing budgets. As time spent on the app declined when advertising content exceeded 8% in a user’s “For You Page” (Shi, 2022), Douyin introduced “shoppable videos” and livestream shopping in the third stage (TikTok, 2021).

The speculative value of data and human control behind the interface

Although most tech start-up companies have no trading history or assets, and produced no profits or dividends, the estimation of future growth over the present accrues speculative value for platform data in the venture market (Davis, 2018; Srnicek, 2017). As explained by Van Doorn and Badger (2020), data’s speculative value is an important collateral value produced alongside a platform’s monetary value from the services it provides. I argue that this speculative value not only attracts venture investors, but it has been extrapolated into a vision, an ideology or, more figuratively, a whirlwind for individuals and small and medium-sized businesses to invest in the belief that big data create opportunities that they cannot afford to overlook. In this way, the speculative value of data accelerates the accumulation of monetary value for platform companies (Srnicek, 2017).

Scholarship on the relationship between data and the market has focused on trade flows among platforms, businesses, and data-focused entities, such as data brokers and data analytics companies, which primarily serve advertising, marketing, publishing, and those in media industries (Gerlitz and Helmond, 2013; Helmond and Van Der Vlist, 2023; Kotliar, 2020). However, not all platforms rely on advertising, and among those that do, their dependence varies (Srnicek, 2021). In the Western platform ecosystem, the reliance on advertising is prominent in Facebook and Google. While Amazon is establishing its presence in the advertising domain, its dependence is more significantly related to cloud computing and its e-commerce platform (Srnicek, 2021). For this reason, Srnicek (2021: 32) argues that the valorization of data in finely targeted advertising space does not amount to “the sale of data to another entity in any meaningful sense.” Srnicek goes on to explain that, instead of being the source of all digital value, data are more a means to capture rents from other sectors of the economy.

China’s major platforms exhibit this value capture behavior. The Chinese market is distinct from the one in the West in that data brokers and data analytics companies do not play a significant role. This is in large due to China’s Internet conglomerates—represented by Alibaba, Tencent, and ByteDance—privatizing data analytics into proprietary business territories through subsidiary platforms, despite the government’s intervention to facilitate cross-company data trade (Fei, 2023). Monopoly ownership over personal data allows them to create traffic-selling platforms of their own. By creating rentier relations with the retail sector, these platforms operate as “the Internet of landlords” that controls and profits from access to platform data while creating technical systems and trade protocols for value capture at scale (Sadowski, 2020).

For algorithms employed in the platform economy, existing scholarship often conflates human and algorithmic agents without critically reflecting on how these two enhance each other’s performance in exerting control (Daugherty and Wilson, 2018). Consider, for instance, social media influencers. Much of the emphasis has been put on the pursuit of visibility by increasing algorithmic literacy (Bishop, 2018; Cotter, 2019). Visibility is not merely an automated outcome, however. TikTok moderates visibility by manipulating the audience reach through algorithmic and regulatory means (Zeng and Kaye, 2022). This means that despite individual variances in algorithmic visibility, there is a holistic human design for viewing traffic. Analyses centered on algorithmic automation risk separating visibility from the market conditions and imperatives that both produce and commoditize it. This is because algorithms serve the dual purpose of personalizing experiences and augmenting human control in the interest of platform companies. As denoted by TikTok (2020) in its Newsroom:

. . . sometimes you may come across a video in your feed that doesn’t appear to be relevant to your expressed interests or have amassed a huge number of likes . . . By offering different videos from time to time, the system is also able to get a better sense of what’s popular among a wider range of audiences to help provide other TikTok users a great experience, too. Our goal is to find balance between suggesting content that’s relevant to you while also helping you find content and creators that encourage you to explore experiences you might not otherwise see.

This excerpt reveals how personalized “For You Page” algorithms and trending algorithms work together in Douyin’s transformation into e-commerce. Although recommending content that may not align with a user’s interests is framed as “diversity” discourse by TikTok, scholars have critiqued this framing as part of the underlying ideology for platforms such as Facebook (“making the world more open and connected”) and Netflix (“inclusion strategy”) to gain cultural legitimacy and profits (Hoffmann et al., 2018; Khoo, 2023). In this light, algorithmic and data-driven technologies are “more akin to bureaucratic or administrative mechanisms than intelligent systems” when integrated into industries (Caplan and boyd, 2018: 2). While algorithmic assessment has been widely discussed, human control pertaining to the operational dimension receives less acknowledgment. This discrepancy has led to a metric-obsessed economy that reduces the dynamic market factors at play into computational technicalities.

The primary reason platforms control data supply is that, contrary to assumptions that data are an inexhaustible resource, the userbase of a platform is finite, and so is the data traffic they produce. Regulating data supply hence becomes central to transforming speculative value into monetary value. Platforms determine prices of data through control mechanisms by creating high demand through short supply, leading to a demand-driven circulation market. The significance of algorithms thus not only resides in their automated capabilities but also in their capacity to complement, extend, and augment human abilities in exerting control in the interest of platform companies (Daugherty and Wilson, 2018).

It is crucial to understand that platforms capture “rent” value within the existing market of which they are a part, rather than creating entirely “new” market demands (i.e. value creation), as often claimed by platform companies (Sadowski, 2020; Srnicek, 2021). As creators and sellers stake their future on data’s speculative value by moving their careers and businesses onto big tech platforms, they are simultaneously enrolled into the platform value capture cycle as their demands shift from the existing market domain to a platform-based one. These carefully crafted visions have formed the deep-seated ideology of the Internet since its inception and this ideological force has been continually revitalized in various technological stages (Fuchs, 2014; Stevenson, 2018; Van Dijck, 2013). By marketing the technical and business capacity of traffic, these visions augment data’s speculative value while obscuring institutional factors (i.e. ownership and control) that both constrain and foster businesses (Caplan and boyd, 2018; Napoli, 2014).

Methods

Considering the stacked nature of the research object—economic exchanges in various dimensions and involving various participants (Caliskan, 2020)—this study de-centers Douyin as the primary focus of research. Placing platforms at the center of analysis might overlook the ways in which they are integrated into broader environments and relations (Pink et al., 2016). This study draws upon two primary sources of data: livestream transcripts from Douyin’s traffic-selling platforms and semi-structured, in-depth interviews with Douyin employees and market participants. A thematic analysis is then conducted to identify quotations, keywords, and themes that can inform the transformation of Douyin (Naeem et al., 2023).

Douyin operates four traffic-selling platforms: Dou+ (https://doujia.douyin.com/), which allows creators and sellers to artificially boost their views; The Influencer Platform (Juliang Xingtu, https://www.xingtu.cn/), which facilitates partnerships among sellers, creators, and MCNs; OceanEngine (Juliang yinqin, https://www.oceanengine.com/), which offers data traffic strategies for sellers; and ShoppingAds (Juliang qianchuan, https://www.shoppingads.cn/), which provides integrated marketing and sales solutions for sellers. These platforms, except for The Influencer Platform, operate official accounts on Douyin, where they offer livestreamed tutorials to assist businesses in selecting the most suitable data service.

From 28 February 2023 to 17 April 2023, a student research assistant was hired to watch and transcribe daily livestreams by Ocean Engine, Dou+, and ShoppingAds. The observation included not only the tutorial content that constructs career and business visions but also interactions between streamers and creators/sellers, with a particular focus on sellers who expressed a keen interest. Three transcription documents, each with a page length between 13 and 17, were produced as research data. Throughout the observed period, the same or similar content repeated daily, with the exception of 5 April (China’s Qingming Holiday) and occasional weekends. This consistency enabled the creation of well-organized transcriptions.

Semi-structured, in-depth interviews were conducted with 12 Douyin employees, 9 influencers, and 7 representatives from sellers, MCNs, and marketing firms. Detailed information about the participants can be found in Appendices 1 to 3. The employees who participated in this study worked in user growth, business development, product development, interactive entertainment (including music), and the AI lab. However, the names of their workgroups may change as ByteDance constantly reorganizes its corporate structure for business development. All interviewees were recruited through personal contacts and snowball sampling. On the recruitment of interviewees, I established contacts with key staff members in the interactive entertainment unit who were participating in and speaking at a high-level forum on the music industry in Beijing in 2020, back when music-driven videos were considered Douyin’s primary features. Based on several preliminary interviews with them, key factors that marked Douyin’s development emerged, including “reward programs,” “traffic pool,” “algorithms,” and “AI lab.” A purposive sampling and search for employees who were responsible for the above-mentioned factors was initiated through the author’s personal contacts and their referrals. This prolonged process spanned 3 years (December 2020 to December 2023), involving multiple fieldtrips to Beijing to establish contacts and conduct interviews.

Informed consent was obtained prior to the interviews. For Douyin employees, key questions included strategies for acquiring new users, mobilizing market actors such as creators, MCNs, marketing communication firms, and sellers, protocols for monetization, and terms and rules for regulating market participants. For employees working in business development, product development, and the AI lab, the questions were focused on the production (the automated dimension) and circulation (the operational dimension) of platform data. For creators and influencers, the questions revolved around their experiences and their involvement in monetization activities organized by Douyin and its Influencer Platform. For market participants such as sellers, MCNs, and marketing firms, questions related to traffic-selling platforms, their intended purposes, the outcomes, and how they coordinated logistics and warehouse inventory to support livestream sales designed to convert viewers into customers.

A thematic analysis entails selecting quotations and keywords, developing themes, and conceptualizing through interpretation of keywords and themes (Naeem et al., 2023). For the selection, this study adopted Owen’s (1984) three criteria: repetition, recurrence, and forcefulness. First, I marked words and phrases that appeared in both livestream and interview transcripts (repetition). Then I marked different wordings (quotations) used to indicate the same or similar things. This step shows how Douyin, creators, and market actors communicate meanings of these words and phrases (recurrence). Finally, I followed Orbe and Kinefuchi’s (2008) application of forcefulness in their thematic analysis of student reactions to a movie by identifying thematic insights that possessed interpretive significance for the subject under study rather than merely counting frequency. This is of great importance for understanding Douyin’s transformation. As pointed out by Stevenson (2018: 70), media technology does not develop in a linear fashion but is constantly shaped by various actors with varying interests making decisions about “which technologies or media to promote and use, how to use them, which ones to invest in, and so on.” These choices are rarely rational outcomes but are influenced by “culturally specific values, beliefs and practices, political and commercial interests, as well as the material constraints of available technology” (Stevenson, 2018: 70). In this light, it is problematic to use code counts only to determine relevance and importance of themes in thematic analysis. As Owen (1984: 274) succinctly explains, themes are “less a set of cognitive schemas than a limited range of interpretations” that are used to conceptualize research findings. This article thus focuses on thematic insights by putting identified themes back into the context and conversations in which they emerge to examine how Douyin employees and different market participants communicate the meanings of these key themes. In this process, frequently mentioned keywords such as “music,” “content,” “data indicators,” “traffic (pool),” and “conversion,” and quotations containing these keywords, were identified. Themes that inform the transformation of Douyin subsequently emerged. In interpreting these themes, I focused on visions crafted by Douyin for various market participants at each stage of its transformation.

Copyrighted music for free: accelerating user acquisition through music trends

ByteDance is known as a “super app factory,” meaning that the company uses internal competition to allocate key resources to a few priorities based on estimates of user growth (Zhai, 2019). Between 2018 and 2020, ByteDance had at least 140 apps competing for the company’s “selective focus” (Chen and Ma, 2022). As a music-driven platform, the high royalties charged by large music labels like Universal, Warner Music, and Sony hindered Douyin’s growth at the early stage. To gain music content, Douyin turned to small- and medium-sized studios through a “resource swapping” contractual deal: small and medium labels offered their songs free-of-charge to Douyin in exchange for top positions on the app’s trending chart.

This deal was internally known as “starve the big, feed the middle and small.” Taking advantage of their desperation for market access, the deal created a vision for small music companies and early-career artists. In January 2018, this deal was formalized into Douyin Spotlight, a musician program pledged to assist fledgling music enterprises and independent/early-career artists in building communities of fans (ByteDance, 2019a). Through this program, Douyin formed partnership with more than 800 small and medium-sized labels and copyright holders (ByteDance, 2019a). The vision of investing in the future and nurturing fan growth successfully fostered participation, with more than 30,000 artists debuting their original work on the platform in the first half of 2020 (Douyin, 2020).

Free music laid the foundation for Douyin’s growth. The increasing number of the songs published on Douyin allowed the platform to profile user engagements against music trends (colloquially known as shen qu, meaning “viral music”), which helped grow and retain users. Chen and Ma (2022) conceptualize ByteDance’s strategy as the “shared-service platform” model, where business, technology, and operational functions are centralized within the company to more effectively provide support to product teams in fostering growth. In the case of music, the team customizes the company’s “traffic pool” system to identify emerging music trends. According to Douyin’s music team and Dou+ livestreams, the traffic pool establishes a progressive audience reach model for content. While preliminary analyses of a song’s engagement data are conducted through the traffic pool, staff members fine-tune the cursory algorithmic outputs to finalize the recommendation list. This is because the traffic pool system is not yet intelligent enough to assess a song’s lyrics, melody, and arrangement, nor is it able to parse the feelings and emotions of a song. For the music team, rather than an exact science, the traffic pool involves experimenting with algorithms to identify music trends, a process filled with ambiguity and “hard to explain” outputs (see also Gillespie, 2016).

The interview with R6, product manager in automation at Douyin, further confirms the ambiguous qualities of the traffic pool system. According to R6, the company’s AI lab does not have their own business. Instead, they transfer the lab’s computation abilities to business development teams. For instance, when the music team is tasked with identifying music trends, the AI lab would provide data intelligence support. Despite its strong technical base, the traffic progression mechanism was trained according to prior business experiences subject to individual perspectives and interpretations rather than objective realities. In his own words:

Data are not as complex as one may imagine. For us (staff in audience research), data are not technical; they are information about consumption habits, viewing habits, and so on. It’s not that we use specialized tools to analyze data, technical professionals can do that better than we can. For us data may appear subjective (zhu guan). What we look at is what these data say, what they reflect, and how this information can be applied (into business), and how can we automate this process for better productivity.

This suggests that automated technologies are invested with underlying human decision-making based on prior experiences and knowledge, which has seldom been publicly acknowledged but is essential in shaping how these technologies operate and impact outcomes. This has resulted in a conflation of the dimensions of data production and circulation in the platform economy scholarship. Data operate under a dual logic in the platform economy: “value is a measure of the production process, price a measure of the circulation process” (Fuchs, 2014: 131). In the case of Douyin, platform data first conjure speculative value for music creators, which motivates them to voluntarily contribute. However, this value is more of a carefully crafted vision than an inevitable reality. Second, as exemplified by the stratified traffic pool, it begins to establish a circulation system to transform this speculative value into monetary value. As discussed later, this value conversion has extended into both areas of creative labor management and livestream shopping.

As Douyin provides market access to music creators, it simultaneously devalues their creative labor through contractual and technical designs. In a sample contract provided by R2, a staff member at Douyin’s interactive entertainment, self-publishing musicians not only cede their rights of reproduction, adaption, and performance of their work to ByteDance, they also allow their music to be used in all apps owned by ByteDance through its “shared-service platform” (Chen and Ma, 2022). This authorization exempts Douyin from copyright infringement in its e-commerce in which music-driven short videos and livestreams are repurposed for shopping. The Internet has a long-standing business model of monetizing free resources (Srnicek, 2017). The case of Douyin illustrates that when the necessary resources for growth are not free, the company strategically makes them freely available and cashes in on their monetary value in various initiatives later.

This approach is detrimental to early-career artists. The situation is further exacerbated by the technical design of “challenges,” which disconnects original artists from their own music due to a proliferation of covers. For Douyin, these “challenges” serve as one of the most prominent features in generating engagement. Douyin concentrates its traffic on mass popularity, on top of personalized recommendations, especially during its initial stage when user acquisition was the primary objective. This was achieved through principles of mimesis: imitation and replication (Zulli and Zulli, 2022). Incorporating artists while simultaneously marginalizing them is Douyin’s systematic way of acquiring free music content on the one hand and facilitating user acquisition on the other. The rapid growth of music content creates a competitive audience attention market, which allows the launch of the Dou+ platform to formalize artificial boosting of views into its value capture ecosystem.

While copyrights were once a barrier that hindered its growth, Douyin now weaponizes them to punish music creators who challenge the platform’s power of control. In 2018, a song named “The Taishan Mountain Next Door” went viral on Douyin. However, its copyright holder declined to sign the above-mentioned contract with Douyin. Trending songs bring in new users and traffic, and, in theory, their owners can direct users and traffic elsewhere. When Douyin’s music team found “The Taishan Mountain Next Door” appearing on its competitor platform Kuaishou, they quickly capped the song’s audience reach. The continuous emergence of similar cases compels Douyin to design an “olive-shaped” structure (explained below) for music creator management, and by extension, for video creators: the app now scaffolds audience reach by increasing view counts (within controllable bounds) for those with a middle-sized following while undercutting audience reach for those at the top and the bottom.

The olive-shaped structure: affordable influencers for businesses of all sizes

In 2019, predicting that Douyin’s daily active users would continue to grow (it had reached 320 million by July 2017), Douyin launched a Creator Reward Program (chuang zuo zhe cheng zhang ji hua), aiming to assist more than 10 million creators in monetizing their content (ByteDance, 2019b). The company’s music team uses the term “olive-shaped structure” to describe Douyin’s creator management. This systematic design makes those with a moderate-sized following the majority, while giving less visibility to those at the top and at the bottom. During the interview, R1, then Head of Music at Douyin, said, “Top creators are very likely to ‘kidnap’ the platform with their unregulated growth.” Complementing algorithm-centric analyses of influencer visibility at the individual level—often described as “playing the visibility game” against automated systems (Bishop, 2018; Cotter, 2019; Liang, 2022)—the Douyin case reveals a more holistic, human-controlled approach to visibility.

According to ShoppingAds livestreams, small businesses—mostly national brands and homegrown consumer goods—account for the majority of sellers in Douyin Shop, where a customer spends on average, between 100–150 CNY (14–21 USD) for cosmetics and 9.9–39.9 CNY (1.4–5.7 USD) for food in a single purchase. This spending behavior conditions price ranges that influencers charge for endorsements. To make creative labor affordable for small businesses and brands, Douyin reduces creators’ bargaining power by controlling their audience reach through the traffic pool. While growing a follower base remains a goal for influencers, it has become increasingly challenging as “audience reach” is sold as a commodity through Douyin’s traffic-selling platforms, including Dou+, Ocean Engine, and ShoppingAds.

The perceived “threat” posed by creators with large followings was systematically addressed through Douyin’s Influencer Platform, which was released in November 2020. This platform connects sellers, creators, MCNs, and marketing firms and monitors their transactions. Based on asking price and follower counts, influencers are stratified into a structure consisting of “head” (i.e. top, expensive), “shoulder” (moderately expensive), “waist” (middle, moderate price), and “tail” (bottom, meaning cheap). Although this classification is based on calculated assumptions, it has been an instrumental measure adopted by sellers, MCNs, Douyin, and marketing firms in e-commerce communications.

By 2022, the Influencer Platform had incorporated more than two million influencers, 1500 MCNs, 1000 marketing communications firms, and 1.9 million sellers (Toutiao, 2022). According to interviews with influencers, sellers, and marketing firms, they are all requested to place their orders through the Influencer Platform. Failing to comply might result in a shadow ban, another way Douyin manages audience reach to regulate influencers. The gatekeeping power allows Douyin to control and oversee all transaction data on its platform.

Influencers with fewer than 100,000 followers—the threshold for Douyin’s Influencer Platform—usually need to participate in seller-sponsored marketing activities, known as “branded missions,” to grow followers. Influencers submit their branded creations in exchange for perks, such as free traffic, complimentary samples, or modest compensation. Douyin prioritizes recommending videos with product placements because traffic has already been paid for by sellers through Dou+ or ShoppingAds. However, the Influencer Platform often offers compensation lower than creators’ asking prices, which has discouraged several of our interviewed influencers from participating in branded missions.

Early-career creators, however, depend on branded missions for visibility, as branded content benefits from paid traffic. According to an interview with S2, a 21-year-old female influencer who has a follower base of 23,000, some branded mission invitations “don’t make sense at all, finding a suitable match is time-consuming and exhausting.” Although she occasionally finds missions that align with her profile, most offer complimentary items with little monetary value, such as a sachet or a lipstick, or a nominal payment of approximately 200 CNY (roughly 29 USD). Interviewees S4 and S7 expressed similar frustrations about the mismatch between branded missions and their profiles. This suggests that the algorithms matching sellers and influencers may be overrated or require improvement. This deficiency warrants human intervention and control from both Douyin and MCNs to optimize the transaction process. One month before the interview, S2 became contracted with an MCN. MCNs have specialized business teams to match relevant products or services for influencers to represent and to negotiate favorable deals with sellers and marketing firms on their behalf. In exchange, S2’s channel was partially taken over by the MCN.

The regulation of audience reach and the organizational incorporation of creators demonstrate that labor exploitation is a systemic design facilitated by platform governance, market demands, and institutional takeovers (Howson et al., 2023). Contrary to assumptions that algorithms are responsible for exploitation, the case of Douyin reveals covert human control remains central to automated systems. The market-based allocation of viewing traffic organizes both individual experiences and collective conditions in the creator market. Creators sometimes add “the limited audience reach edition” (Xian liu ban) after their username to indicate their awareness that Douyin has capped their viewership. The hurdles of acquiring viewing traffic have prompted influencers to transition into livestream shopping, because viewing traffic purchased by sellers can concurrently boost influencers’ visibility. As this practice becomes normalized, the size of an influencer’s follower base becomes increasingly irrelevant. This has diminished influencers’ negotiating power even further, as they become increasingly dependent on sellers’ investments in traffic. The next section discusses how Douyin captures another form of value: infrastructural rent in its transformation into online shopping.

“Let every good brick-and-mortar store be seen”: interfacing data traffic with retail

ByteDance launched the Douyin E-commerce Platform in June 2020. Although it is an independent operational unit separate from Douyin’s content, the E-commerce Platform is built upon data traffic generated from content by making it a “newfound” source of growth for the e-commerce sector. At the early stage, Douyin adopted an advertising-based e-commerce model that allowed sellers from other online retail platforms such as Taobao and JD.com to embed purchase links in short videos and livestreams. Interested viewers completed a purchase by being redirected to the partnered shopping apps and sites. This business partnership was referred to as “traffic referrals,” meaning that Douyin earned commissions from each redirection (Shi, 2022).

However, user retention and time spent declined when advertising content exceeded 8% on the “For You Page” (Shi, 2022). To balance monetization and retaining users, Douyin coined the terms “shoppable video” and “social commerce” to make e-commerce engaging and convertible, while keeping entertainment at its core. In doing so, Douyin introduced e-commerce algorithms on top of its content algorithms, which modified “For You Page” recommendations. According to OceanEngine livestreams, its e-commerce algorithms center on two criteria: conversion rates and interaction metrics. The conversion rates are calculated by sales numbers, the ratio of viewers to buyers, repeat customers, and the return/refund rate. The interaction metrics consider time spent in a livestream (deemed satisfactory if viewed for more than 1 minute and better if longer than 3 minutes), interactions (viewers who like, share, and comment during a livestream), and the conversion ratio of viewers to brand/shop members. These measurements are not scientifically determined but are based on prior business experiences and are therefore subject to modification according to market changes.

Algorithmic modifications were accompanied by changes in Douyin’s interface (the newly added “Shop” tab) and operational planning in support of e-commerce transformation. The strategic goal was to attract sellers to relocate their online shops from other retail platforms to Douyin. Since October 2020, Douyin no longer allows purchase links from external sites in its livestreams (with the exception of paid ads). Instead, it put forward the vision of “Let every good brick-and-mortar store be seen.” The early traffic referrals provided intelligence on viewer-to-buyer conversion data for Douyin to craft a vision of data-driven sales. To accelerate the seller recruitment, Douyin advertised on the “For You Page” that “You can open a shop on Douyin as long as you have a business license, irrespective of your business type.”

This campaign was coupled with a series of business showcases that started in 2021. Three frequently cited cases are all from the clothing industry: an athleisure brand, a women’s clothing label, and a fashion brand. Douyin’s business development team crafted a vision of retail growth by highlighting how these sellers thrived by recreating offline shopping experiences in livestreams. To incentivize sellers, Douyin reduced infrastructural service charges, including platform commissions (reduced from 20% to 5%) and payment handling fees (Douyin Pay; Shi and Gao, 2020). In January 2022, these initial endeavors culminated in the launch of the program “Retail in the cloud” (yun ling shou), which incorporated livestreaming from physical stores as a key component of its e-commerce. This program shifted the sales focus from foot traffic in brick-and-mortar stores to digital traffic, leading to increased dependence on platform data allocation.

The interface and algorithmic changes, along with the no-external link policy and platform campaigns and incentives, demonstrate that platform economies can hardly be viewed as merely automated processes but are coordinated with human control in order to materialize the algorithm’s productivity and data’s monetary value. These factors together facilitated Douyin’s transformation into an independent shopping platform that turned its former e-commerce partners Taobao and JD.com into its main competitors. In May 2022, Douyin incorporated a “Shop” tab next to its prominent “For You Page.” Since then, its recommendation algorithms regularly push pages such as “Good stuff you may like” (ni ke neng xi huan de hao wu) and “You may want to search” (cai ni xiang sou) in video feeds to direct viewing traffic from “For You Page” to “Shop.” Altogether, it reflects that Douyin is dividing share of the existing e-commerce market that was once dominated by Taobao and JD.com. Srnicek (2021) argues that platforms redistribute demand instead of creating new demand to capture value. Considering Douyin’s transformation into e-commerce, neither shifting demand from offline to online (as in the “Retail in the cloud” program) nor shifting demand from one retail platform to another can be seen as new value creation.

To convince sellers that digital traffic can boost sales, Ocean Engine and ShoppingAds promote the idea that “paid traffic leverages organic traffic.” ShoppingAds recommends that beginners concentrate on offering benefits, such as digital coupons, to viewers during their first three livestream sessions. This approach enables sellers to profile their shop by monitoring viewer numbers and analyzing view peaks and troughs. These data insights, generated from organic traffic, provide information about a shop’s prospective customers based on factors such as age, gender, and other social categories, and therefore inform decisions on traffic purchase. In contrast to the traditional advertising model that builds brand awareness by making a product or service known to the public, Douyin’s livestream shopping emphasizes immediacy by converting audience reach and attention into real-time sales through interactive entertainment.

As flows of entertainment and goods become regularly interfaced on the “For You Page,” Douyin increasingly influences user behavior by converting online views into paying customers through constant exposure to a commodity logic (Fuchs, 2014). Coupled with its built-in payment method Douyin Pay and Shopping Cart (i.e. xiao huang che), Douyin now operates in a dual infrastructure consisting of both advertising (exposure-oriented branding) and retail (sales-oriented conversion). Thus, Douyin accrues two types of rent from e-commerce: advertising rent from advertising spaces embedded in the “For You Page” (Srnicek, 2021) and infrastructural rent (e.g. Shopping Cart and Douyin Pay) from livestream sales by strategically limiting data traffic supply to forge high demand. The scarcity of platform data becomes the primary commodity that Douyin manufactures for the retail market. As the retail sector becomes ever more digitized, “the owners of these platforms gain more control over the fees that can be charged to access them” (Srnicek, 2021: 38).

Conclusion

This article has analyzed the transformation of Douyin from an entertainment-focused, short video sharing platform into an e-commerce site that mixes entertainment with online shopping. In its three key transformative stages, Douyin crafted visions for career development and business success through contractual deals, online workshops, and business development presentations. Media platform visions are very much a part of the cultural, technological, and economic history of digital media as they “set expectations, galvanize communities of producers and users, create inequalities of attention, steer financial speculation, etc.” (Stevenson, 2018: 70). In Douyin’s carefully crafted visions, the value of data is seldom open to question and is at best speculative. A lack of critical understanding of platform visions runs the risk of subsuming time, labor, and other costs involved in adapting careers and businesses to big tech platforms into voluntary contributions. It also obscures the process by which platform data are synthesized with other factors of production (e.g. creative labor) and market actors (e.g. MCNs, marketing firms, and sellers) in transforming their speculative value into monetary value.

The olive-shaped management regulates audience reach of music creators and influencers to match with a market structure in which small businesses are the majority. While Douyin provides businesses at various stages of growth with affordable advertising, marketing, and e-commerce solutions through its traffic-selling platforms, it also means that organic growth becomes increasingly difficult due to the control of viewing traffic as a scarce commodity. As Srnicek (2021: 38) puts it, “their scarcity is a product of these dynamics.” Controls on the supply of and demand for data traffic exacerbate competition within the creator market, where inequalities are not merely a consequence as expressed in personal experiences but more a collective condition and the result of the systematic value capture. To secure more gigs, creators are compelled to join large MCN organizations at the expense of their professional autonomy, which in turn consolidate organizational power over individual agency.

These findings suggest that the advertising-focused data trade research paradigm is no longer adequate for describing how platform companies operate. In line with Caliskan (2020), a data-centric analysis falls short in accounting for the dynamism, multiplicities, agency, and performativity of economization practices employed by a diverse range of market participants in various geographical contexts. Platform-based economic practices often extend beyond data automation and marketization. This is particularly evident in the case of Chinese platforms, which demonstrate territorial behavior and carefully designed human control enhanced by automation technologies. In converting data’s speculative value into profits, Douyin demonstrates a shift from an earlier focus on user acquisition and retention (i.e. audience data production) to a dual focus on both production and circulation. The traffic commodity is produced, and the price of data is subsequently determined in the sphere of circulation (see also Fuchs, 2014). Instead of creating value, platform companies focus more on value capture and rent accumulation (Sadowski, 2020; Srnicek, 2021).