Rethinking Platform Capitalism: Mapping the Algorithmic Logics and Labor Control Models of Uber Eats and Fantuan

Abstract

The authors examine how platform companies structure labor control in the food-delivery sector by comparing two major apps operating in the United States and Canada: Uber Eats and Fantuan. Although existing research on algorithmic management emphasizes the emergence of “algorithmic despotism”—opaque, data-driven systems that shape worker behavior through surveillance, nudges, and individualized pay schemes—there is a need to map out the variations in platform control. Drawing on a yearlong ethnography that combined app walk-throughs, participant observation, and semistructured interviews with 34 delivery workers in the greater Toronto/Hamilton area, the authors analyze the key organizational “nodes” of each platform, from onboarding and order dissemination to pay structures and worker surveillance. Our comparative analysis demonstrates that labor control under platform capitalism does not operate as a single, monolithic system. Uber Eats deploys a “dispersive-extractive” model that extracts profit through algorithmic opaqueness and performance metrics throughout the driver’s workflow. Fantuan, by contrast, operates through an “ethnocultural” model in which ethnocultural norms as well as human managerial interventions work alongside algorithmic systems to enforce hierarchy, efficiency, and compliance. By mapping these distinct regimes, the authors develop a comparative framework for understanding the variations within algorithmic management and calls for more research across ethnically diverse platform markets.

Keywords

gig economy labor control platform control algorithmic management algorithmic despotism digital labor platforms Uber Eats Fantuan platform work

Introduction

Platform work, in which cloud-based technology and algorithms match workers with consumers (Vallas 2019), has fueled extensive discussion over the past decade. The widespread use of food delivery applications, such as Uber Eats (UE), DoorDash, and Skip the Dishes, in particular, has prompted scholars to conceptualize this new economy as platform capitalism (Howcroft and Bergvall-Kåreborn 2019; Liang, Aroles, and Brandl 2022). Although consumers enjoy the convenience these technologies provide and companies promote the freedom and autonomy they claim to offer delivery workers, many labor scholars question how apps shape the labor process and establish labor control, often at the expense of working conditions.

In the fields of sociology of work and labor studies, scholars have produced a substantial body of research tracing how labor control evolved from the scientific management of mechanized industrial work (Braverman 1998) to the emotional, affective, and relational forms of control that emerged in the service economy (Hochschild 1979; Kang 2010; Padios 2018). In today’s knowledge economy, researchers have only recently begun to theorize new models of labor control that operate through platform applications. In these models, algorithms—the coded rules and computational sequences that structure digital systems—often replace traditional managers and owners. This phenomenon, also known as algorithmic management, differs from earlier forms of labor control insofar as managerial techniques no longer appear through observable human practices but instead operate through opaque technical systems, making it difficult to determine who holds control and how the underlying logics of labor governance function.

Emerging literature on algorithmic control suggests that food delivery apps govern workers through algorithms, data-driven surveillance, and gamified “choices” that together create a flexible but intrusive system of labor control (Gandini 2019; Rosenblat and Stark 2016; Veen, Barratt, and Goods 2020; Wood et al. 2019). Still, research on algorithmic management remains in its early stages and a number of scholars have called for more comparative studies that map key variations in how human, technical, and organizational dynamics of labor control intersect across platforms (Ametowobla and Kirchner 2025; Davis and Sinha 2021; Griesbach et al. 2019; Lei 2021). Such variations in labor management become more pronounced as smaller “ethnic niche” food delivery apps such as Fantuan (a Chinese-based app that caters to East Asian restaurants and consumers in the United States, Canada, Australia, and the United Kingdom) look to scale up and compete with mainstream apps such as UE by developing new models and products of algorithmic control. Moreover, algorithms do not operate independently; humans build, design, and oversee platform apps. As a result, sociocultural dynamics continue to play a crucial role in how companies construct and enforce systems of labor management. Contributing to the literature on algorithmic management that explores the diverse typologies of labor extraction and control, our study systematically compares two major food delivery apps in the United States and Canada, UE and Fantuan, to illuminate how algorithmic control varies across platforms and how sociocultural dynamics structure the logics of delivery work.

Using ethnography and “digital ethnography” (Murthy 2008) as our primary methods of investigation, we analyze the distinct phases, or what we call “nodes,” of both UE and Fantuan, to reveal how both companies deploy different managerial techniques. Similar to Cameron’s (2022) concept of “touchpoints,” which refers to the “interactions or points of contact with the work,” we use the term nodes to denote the crucial junctures in a labor management system in which norms, values, and policies converge to shape the work experience. In our analysis, we focus on four key nodes through which workers encounter labor control and surveillance, algorithmic or otherwise: (1) onboarding and access structure, (2) order dispatch, (3) competition and earnings, and (4) surveillance and discipline.

By comparing the various nodes of UE and Fantuan, we argue that labor control under algorithmic management is non-monolithic and that ethnocultural values can significantly shape how a platform operates. Our findings show that UE relies on what we call a “dispersive-extractive” model of labor management, in which the company instrumentalizes worker-generated data, employment guidelines, and competition among drivers throughout the app to maximize profit. In contrast, Fantuan adopts what we term an “ethnocultural” model of labor control, in which profitability is maximized through formalized coordination, tighter labor governances, and ethnoculturally mediated managerial oversight. Specifically, ethnocultural refers to shared cultural identification, including language, norms, values, beliefs, and historically sedimented practices of a particular ethnic group. Together, these findings map out two platform labor control models and offer a comparative framework for future scholarship on platform delivery work.

We organize our analysis of UE’s and Fantuan’s labor control models into four sections. In the first section we review scholarship on how algorithmic management establishes labor control, with particular attention to the concept of algorithmic despotism (Griesbach et al. 2019) as an influential yet limited theorization. In the second section we discuss our research methodology and rationale for selecting UE and Fantuan. In the third section we compare four nodes across the two platforms to illustrate the dispersive-extractive and ethnocultural models of labor control for UE and Fantuan, respectively. In the concluding section we consider the broader implications for theorizing labor control under algorithmic management and identifies directions for future research that can deepen our understanding of platform capitalism.

Labor Control in the Platform Economy

Although some scholars argue that the proliferation of the platform economy and its “flexible” work arrangements might lead to possibilities for increased worker freedom and liberation, an expanding body of research shows that work flexibility often produces precariousness and new forms of labor control (Griesbach et al. 2019). Rather than acting as neutral intermediaries connecting consumers and delivery workers, digital platforms actively impose a form of “algorithmic management” that is both continuous with and diverges from earlier modes of worker surveillance (Glavin and Schieman 2022; Joseph-Goteiner 2024; Reynolds et al. 2024; Wood 2021; Wood et al. 2019).

Michael Burawoy’s (1982) early account of the labor process laid the foundation for later scholarship on labor control in the platform economy. For Burawoy (1982), labor control is not merely directed by a boss or manager but involves political and ideological dimensions that frame workers’ lived experiences. Burawoy’s (1982) insights incorporated the more “subjective” lived experiences of workers, introducing the notion that worker “autonomy” (or the appearance thereof) could function less as liberation and more as a means of securing consent; workplace systems can absorb worker choice into their models of labor control (Burawoy 1982). Richard Edwards (1979) likewise expanded analyses of labor control processes beyond arbitrary managerial authority by demonstrating how technical elements, such as machines on the Fordist assembly line, enact and enforce workplace discipline. Burawoy’s (1982) framework encouraged scholars to engage with a more extensive range of issues concerning the labor process, such as immigration, gender, local labor markets, and spatial organization (Braverman 1998; Burawoy 1985; Wu 2024), while Edwards directed attention toward the technical elements of the labor process.

Arbitrary Despotism and Algorithmic Authority

Drawing centrally on Michael Burawoy’s (1985) and Richard Edwards’ (1979) work, Griesbach et al. (2019) explore how algorithmic labor control maintains continuity with earlier mechanisms of labor control. Following Burawoy, Griesbach et al. scrutinize how platforms incorporate worker choice into systems of control, describing how the “arbitrary control of the foreman was replaced with a system through which workers became actively invested in the game of production” (Griesbach et al. 2019). Although workers encounter market mechanisms (such as pricing and delivery choices) throughout the labor process, these mechanisms are embedded within broader technical control systems of the platform and serve to influence workers’ choices and workplace decision-making. Under this model, platforms are able to obscure the “rules of the game,” including the algorithmic logics that govern pay and order distribution (Rosenblat and Stark 2016). Griesbach et al. therefore argue that workers experience algorithmic management as “algorithmic despotism,” an arbitrary authority similar to that of a traditional human supervisor, wherein the “boundary between technical and [arbitrary] control blurs, for it is the technological system itself that exerts what workers experience as unjust and personalized discretionary authority.”

One of the decisive methods by which algorithmic despotism derives a sense of arbitrary authority is through the information asymmetries built into the very nature of platform work. Workers typically confront the algorithmic management embedded in such platforms as an opaque “black box,” unable to concretely verify how the system calculates assignments or pay, all while the platform collects a vast amount of data on each worker’s digital clicks and movements throughout the city. Griesbach et al. (2019) demonstrate how platforms weaponize these data to impose “individual-level pay discrimination,” whereby the algorithm “learns” the lowest wage a specific driver will accept in a given time and place and repeatedly offers that rate. Although the algorithm is able to offer gigs to workers with an intimate understanding of their economic decision-making, it simultaneously conceals important information that would allow workers to assess the full economic value of a given gig (Veen et al. 2020). Thus, the presence of information asymmetries within platform work ultimately allows companies to maximize the economic value extracted from its workers.

Techniques of Control

Information asymmetries not only intensify economic extraction but also enable new forms of labor control (Gandini 2019; Griesbach et al. 2019; Rosenblat and Stark 2016; Veen et al. 2020; Woodcock and Cant 2022; Wood et al. 2019). Several scholars have expanded on how platforms transform worker autonomy into a mechanism of control by constraining, nudging, and channeling aspects of worker “choice.” Rosenblat and Stark (2016), for instance, outline how platforms gamify work through “personal bests” on daily earnings or distances or through “surge pricing” on “hot spot” areas, functioning as a form of technonormative control (Franke and Pulignano 2022). Platforms rely on continuous data collection to generate detailed knowledge of worker behavior, allowing algorithms to induce particular behaviors at the lowest economic cost. Tironi and Albornoz (2022) develop the notion of “friendly surveillance” to classify these veiled nudges, describing how platforms strategically reveal incentives and promotions to keep workers engaged and guide their behavior, while simultaneously concealing information necessary for fully informed decision-making. These various techniques of “soft control” work in tandem: platforms withhold key information in a despotic fashion while selectively granting limited forms of autonomy. Together, these practices reveal what Tironi and Albornoz describe as “algorithmic governmentality,” a labor structure in which platforms regulate individual behaviors through programming. Although platforms do not impose explicit obligations on workers, their digital landscapes guide and personalize options, encourage the feeling of autonomy, and subtly direct workers toward more profitable behaviors (Tironi and Albornoz 2022). The result is an illusion of freedom embedded within a mode of management that continuously recalibrates worker conduct toward the platform’s economic goals.

In addition to “soft control” techniques to direct worker behavior, Griesbach et al.’s (2019) notion of “algorithmic despotism” also elaborates on workers’ persistent feeling of managerial supervision. In a similar vein, Woodcock and Cant (2022) develop an account of the “algorithmic panopticon,” in which platforms create the impression of constant managerial oversight through periodic supervision and disciplinary acts. Although scholars originally coined the panopticon metaphor to theorize factory labor management, platform regimes have adapted the dynamic to digital systems which are able to operate without the direct presence of human supervisors. In an algorithmic panopticon, platforms govern their workers through both formal disciplinary measures as well as via the social power of algorithms, wherein workers clearly see evidence of detailed monitoring, but the exact machinations of decision-making remain opaque. Beyond economic nudges and friendly surveillance, algorithms exert an additional layer of social control insofar as the constant presence of data collection encourages workers to regulate their own behavior.

Beyond Algorithmic Despotism

Although these scholars present a comprehensive view of algorithmic despotism as an organizing principle of the labor process, Newlands (2021) points out that algorithmic knowledge of the labor process remains incomplete. According to Newlands, algorithms rely on processes of “datafication,” whereby the material realities of labor are translated into quantitative, often binary data that risks overlooking “messy” qualitative dimensions of work and simplifying lived realities. The demand for standardized data limits the platforms’ ability to capture situated knowledge and produces an “epistemological gap” between workers and the organization in which the “irreducible remainders” of subjective work experience are left untended (Newlands 2021). To address the challenges inherent to datafication, platforms supplement algorithmic management with other forms of human control. Newlands thus theorizes the platform economy as a “tripartite multimodal surveillance assemblage,” that combines algorithmic, managerial, and customer surveillance.¹ Gandini (2019) likewise expands on the human forms of control dispersed throughout platforms. By interrogating how platforms embed metrics such as ratings and rankings directly into their interfaces, Gandini shows how platforms compel workers to engage in emotional labor as a routine and central part of their job, and how such labor ultimately sustains the gig economy’s dispersed and decentered structure. Indeed, the “voluntary” emotional labor undertaken by workers within the platform’s tightly structured rating systems reinforces capital-labor relationships. For these reasons, although algorithmic despotism often dominates discussions of platform management, Newlands and Gandini caution against reducing the capital-labor dynamic to algorithms alone. To be sure, a full account of platform control must also examine the human interactions, managerial interventions, and sociocultural expectations that continue to organize and structure the labor process.

Thus far, the literature reveals an amorphous and flexible regime of labor control in the platform economy that relies on extensive data collection to respond to workers’ individual characteristics and maximize value extraction, what we term a “dispersive-extractive model.” On the one hand, platforms transform worker choice into a mechanism of control through information asymmetries, economic nudges, gamification, and friendly surveillance (i.e., “soft control”). On the other hand, platforms rely on the illusion of constant surveillance to establish an “algorithmic panopticon” that polices workers’ behavior through the social authority of algorithms. Together, these forces sustain a dialectical relationship, linking the illusion of freedom to the perception of constant surveillance, enabling platforms to impose increasingly intimate and personalized forms of control.

At the same time, scholars increasingly question whether algorithmic despotism operates uniformly across platforms (Ametowobla and Kirchner 2025; Davis and Sinha 2021). Although much of the existing literature documents personalized surveillance and panoptic control, researchers are now beginning to examine how algorithmic despotism interacts with human forms of management in ways that differ across platforms and local contexts. In addition to Newlands’s (2021) study on surveillance, Lei’s (2021) framework likewise resists reducing platform labor to algorithmic control alone and instead identifies legal, technological, and organizational dimensions of labor governance. Although these dimensions remain analytically distinct, platforms combine and prioritize them differently, producing varied implications for labor control and worker experience. In this light, scholars now call for further engagement with the diversity of platform structures and point to “varieties of platform capitalism,” in which labor organization reflects domestic institutions, local contexts, and competition dynamics. Indeed, Griesbach et al. (2019) themselves qualify their theory of algorithmic despotism by noting that “although all food delivery platforms use algorithmic management to assign and evaluate work, there is significant cross-platform variation,” and that further studies must distinguish the varieties of algorithmic control as well as their effects on workers’ earnings and experiences in order to inform legal and political debate regarding labor.

Our research responds to these calls for further scholarly engagement with the diverse and multimodal varieties of platform capitalism by closely evaluating the specific configurations of labor control on two gig delivery platforms. By analyzing multiple nodes across the labor process to develop a detailed mapping framework, our research is a comparative contribution to the emerging literature on typologies of labor extraction under platform capitalism. Our comparative study of UE and Fantuan shows how social and cultural dimensions shape the differential use of algorithmic despotism and human management, and how these distinct models, in turn, produce different effects on labor. Although we conceptualize UE as a dispersive-extractive model that maximizes value extraction across each node of the labor process, we find that Fantuan embeds a human-mediated logic of labor control that mobilizes ethnocultural values at multiple points along that process.

Importantly, Fantuan’s ethnocultural labor control model raises the question of the extent to which racial, linguistic, and migratory dimensions shape labor control in the platform economy. Recent scholarship has begun to engage with this issue. For example, Schaupp (2022) shows that algorithmic systems can rely on workers’ precarious immigration status in order to discipline and govern migrant labor more effectively. Relatedly, in proposing the framework of racial platform capitalism, Gebrial (2024) argues that racialization and migration politics are central to producing the platform economy and its management of workers. Although a thorough discussion of this growing body of literature lies beyond the scope of this study, much of the scholarship that addresses race and migration in algorithmic management, however, pays little attention to ethnicity and ethnocultural modalities of labor control. In this study, we treat race and ethnicity as related but analytically distinct, as race often operates as a broad, catch-all social classification through which human difference is generalized, hierarchized, and governed, whereas ethnocultural more specifically refers to forms of shared cultural identification, including language, norms, values, beliefs, and historically sedimented practices of a particular ethnic group. In Fantuan’s case, our analysis centers on Chinese (primarily migrant) workers, who are differently positioned within a racialized labor market, and for whom labor control is mediated through ethnocultural dimensions such as shared linguistic competence, moral expectations, reputational codes, and culturally specific understandings of worker conduct. Our use of the term ethnocultural therefore does not displace race but rather, specifies a distinct mechanism of labor control that racial categories alone do not fully explain.

Methodology

We used a qualitative, multimethod design to examine the models of labor control in platform food-delivery work in the greater Toronto/Hamilton area. Our comparative focus on UE (English language, nationwide, mainstream) and Fantuan (Chinese language, localized, ethnic market) enabled us to investigate the diverse set of algorithmic logics in order disbursement, customer-courier contact, dispute resolution, and disciplinary practices, without presuming a universal model of “algorithmic control” encoded in either platform (Griesbach et al. 2019; Huang 2023).

We selected UE and Fantuan as theoretically illuminating contrasts because UE currently serves as the largest English-language delivery app in the United States and Canada, whereas Fantuan, which originated in Vancouver, targets primarily Chinese and broader East Asian diaspora communities. Fantuan drew strong influence from China’s food-delivery ecosystem, particularly from Meituan, the country’s largest food-delivery platform. Fantuan now operates in more than 40 cities across Canada, the United States, Australia, and the United Kingdom and has become the largest Asian-focused food-delivery platform in North America, competing with other diaspora-facing platforms such as HungryPanda and other regional rivals.

By comparing the two apps in a shared metropolitan labor market, we were able to observe how each platform deploys distinct logics to shape work experiences, and how differences in scale or platform size (measured by total users) and market competition influence emerging models of labor control. Although UE is significantly larger than Fantuan (~95 million users worldwide vs. ~2 million users worldwide), Fantuan’s growing success may encourage cross-pollination between the two platforms models and generate new regimes of algorithmic control that require complex regulatory mechanisms.

By studying UE and Fantuan comparatively, we tested whether observed divergences in labor control arise from the platform model itself or from the imprint of specific ethnocultural norms and beliefs embedded in the platform. Methodologically, this pairing lets us hold constant a metropolitan labor market while varying platform logics and managerial practices, thereby improving causal inference around how ethnocultural values are translated into routinized forms of algorithmic control.

Between July 1, 2024, and June 30, 2025, our team conducted ethnographic fieldwork that combined semistructured interviews, participant observations, and digital ethnography that included app walk-through notes, screenshots of platform interfaces, and textural reading of company manuals. Our analytic strategy was iterative and comparative. We treated each platform as a case within a common urban labor market and systematically compared app processes (onboarding, order assignment, pay and fees, ratings, and grievance handling) via worker narratives, job shadowing walk-throughs (wherein researchers on our team accompanied delivery drivers for a few hours during their daily shifts), and our own experiences as part-time delivery drivers.

We conducted semistructured interviews with 34 adults who had worked on either UE or Fantuan for at least three months and could comfortably communicate in either English or Mandarin. Requiring a minimum of three months’ experience ensured that participants were adequately familiar with workflows and governance practices (Lee et al. 2015). Our recruitment proceeded through multiple channels: a web-based survey inviting preliminary screening and scheduling of interviews; targeted outreach and advertisements on Facebook, Reddit, and Instagram; snowball sampling; and community-based recruitment. We provided honoraria of $40 to $75 to participants who completed their interviews depending on interview length and format. On average, interviews lasted about an hour and were conducted either in person or online through virtual teleconferencing (Zoom, WeChat). When we required clarification or additional details, we conducted follow-up interviews with selected participants to deepen our understanding of specific platform practices. All participants provided their informed consent prior to participating in the study. Interviewees were also informed that pseudonyms would be used and that any identifying details would be removed from all reports and publications.

Data Collection and Analysis

Our nonrandom sample, although purposive, remains statistically nonrepresentative of all platform delivery workers in the greater Toronto/Hamilton area. In a labor market characterized by opacity, high turnover, and limited administrative transparency, however, probabilistic sampling is often ill equipped to capture lived governance effects. Official data offered by UE and Fantuan only provide partial insight into workforce composition and day-to-day managerial practices, thus making qualitative, community-engaged research essential for understanding how labor control operates in “hard-to-reach” groups and contexts. In this light, the heterogeneity within our sample, combining both worker and managerial perspectives, offers a substantively grounded account of labor control in a rapidly expanding and structurally opaque segment of the platform economy.

The 34 delivery workers we interviewed represent a broad range of ages, immigration backgrounds, and levels of dependence on platform work (Tables 1 and 2). Participants ranged in age from 19 to 51 years and included both part-time and full-time couriers. Although most Fantuan drivers identified as Chinese and spoke Mandarin, UE drivers reflected a more diverse mix of ethnic and linguistic backgrounds (Figures 1 and 2).

Table 1.

Basic Sociodemographic Statistics, All Interviewees (n = 34).

Variable	Interviewees
Uber Eats (0/1)	19 (0.56)
Fantuan (0/1)	15 (0.44)
Age (years) (range = 19–51 years)	28.47 (7.53)
Male (0/1)	26 (0.76)
Female (0/1)	8 (0.24)
Student (0/1)	12 (0.35)
Nonstudent (0/1)	22 (0.65)
Full-time workers (0/1)	19 (0.56)
Part-time workers (0/1)	13 (0.38)
Employment status unspecified	2 (0.06)

Note. Binary variables are reported as proportions (mean of 0/1 indicator); standard deviations are shown in parentheses.

Table 2.

Descriptive Demographic Statistics, Uber Eats and Fantuan Delivery Worker Interviewees.

Variable	Uber Eats (n = 19)	Fantuan (n = 15)
Age (years)	28.95 (7.03)	27.87 (8.33)
Male (0/1)	0.63	0.93
Female (0/1)	0.37	0.07
Student (0/1)	0.32	0.40
Nonstudent (0/1)	0.68	0.60
Full-time workers (0/1)	0.47	0.67
Part-time workers (0/1)	0.42	0.33
Employment status unspecified (0/1)	0.11	0.00
Canadian citizen (0/1)	0.47	0.00
Canadian permanent resident (0/1)	0.21	0.20
International student work permit (0/1)	0.11	0.33
Work permit/open work permit (0/1)	0.11	0.47
Immigration status unspecified (0/1)	0.11	0.00
Chinese ethnicity (0/1)	0.27	0.93
Non-Chinese ethnicity	0.73	0.07

Note. Binary variables are reported as proportions (mean of 0/1 indicator); standard deviations are shown in parentheses.

Figure 1.

Immigration status of interviewees by platform, Uber Eats versus Fantuan.

Figure 2.

Ethnicity of interviewees by platform, Uber Eats versus Fantuan.

In addition to interviewing drivers, we also spoke with managerial staff connected to Fantuan’s local operations, including two operation specialists who had previously worked as drivers before transitioning into managerial roles based in Hamilton and Mississauga. Including these former drivers allowed us to capture both worker-level and supervisory perspectives on Fantuan’s platform governance. Moreover, these conversations helped us discern how company policies are applied in practice, especially with respect to the ability of managers to override algorithmic order assignments and shift scheduling. We further supplemented our interviews with participant observations: researchers on our team created courier accounts, navigated onboarding, accepted and completed orders, documented payments, ratings, and the dispute-resolution process. These app walk-throughs functioned as a means to triangulate and verify claims regarding workers’ lived practices. To capture the situated practices which drivers endure, we briefly conducted job-shadowing practices (wherein a researcher on our team would accompany the delivery driver on their shift) during peak periods and informally engaged couriers at restaurants and delivery hotspots.

Our audio-recorded interviews were transcribed (and, where applicable, translated) as well as coded for information across three larger domains: (1) operational logics, (2) communications, and (3) surveillance and regulations. We conducted a cross-case comparative analysis contrasting UE and Fantuan with regard to signup processes, pay and fee structures, route assignment, customer-worker contact, and grievance handling. We analyzed our interview data using a grounded theory approach that moved abductively between empirical material and relevant theory, consistent with qualitative research traditions (Corbin and Strauss 2014). After transcribing the interviews, we engaged in iterative thematic analysis, beginning with close, line-by-line coding to preserve participants’ own language and experiential framing. Through inductive reasoning, we identified first-order “tags” that captured recurring descriptions of platform governance, worker coping strategies, managerial discretion, and algorithmic opacity. We then grouped these into broader second-order thematic categories through comparison across our field notes, interviews, and platforms, refining and collapsing codes as patterns emerged.

Many of our ex ante tags were not static and new tags were created by our team if we identified an analytical or thematic pattern during our initial analysis stages. Our coding schema evolved and included but were not limited to the following tags: entrepreneurial futurity, interethnic tension (i.e., ethnic clustering, negative remarks about other ethnic groups), ethnoculturally managed (i.e., Fantuan’s assigning low-value orders to non-Chinese drivers or to another third-party platform), app discontent, wage discontent, job motivations, worker solidarity, algorithmic opaqueness (i.e., workers’ complaining about not understanding how the platform works), and employer dereliction. Throughout our data analysis, we oscillated between emerging empirical patterns and existing scholarship on algorithmic management; our iterative movement between data and theory enabled us to generate a conceptually grounded account of platform labor regimes while remaining attentive to participants’ situated experiences.

Comparing the Labor Control Models of UE and Fantuan

In this section, we comparatively analyze four nodes in the delivery processes of UE and Fantuan: (1) onboarding and access structure, (2) order dispatch, (3) competition and earnings, and (4) surveillance and discipline. By mapping each of the four nodes, we demonstrate how both UE and Fantuan showcase differing principles in their methods of labor control.

UE, on the one hand, relies on an automated, algorithmically opaque, and metrics-driven model that disperses profit extraction across many micro-stages of the driver’s workflow, exhibiting what we identify as a dispersive-extractive labor model. Fantuan, on the other hand, organizes control through a human-mediated, tiered system in which managerial discretion and ethnocultural norms shape access to work, order allocations, enforcement measures, and driver penalties, exhibiting what we identify as an ethnocultural labor model (Table 3).

Table 3.

Argument at a Glance: Comparison between Dispersive-Extractive and Ethnocultural Models of Labor Control.

	Labor Control Model
	Dispersive-Extractive(Uber Eats)	Ethnocultural(Fantuan)
“Managerial” power	Applied primarily through algorithms	Applied primarily through humans
Onboarding and access	Automated checks; low barriers to eligibility but no guarantee of work; costs and risks are externalized to drivers	Manual review by local operation specialists; tiered access (“Ricer”/“Baller”)
Order dispatch	Individualized by algorithmic dispatch, reoffer ratchets, and hidden performance thresholds, culminating in race-to-the-bottom dynamics	Formalized through shifts and worker ranks; algorithmic scheduling is reviewed and finalized by human managers
Competition and earnings	Drivers are provided with estimate-only offers; income is fragmented into multiple volatile components (base pay, boosts, batching totals, tips) and revealed only partially at acceptance and adjustable after the fact	Income and high-value orders are mediated by drivers’ reputation, interpersonal workplace relations, reliability, linguistic and cultural fit
Surveillance and discipline	Hidden thresholds and vague categories trigger penalties and deactivation; limited, slow, or automated recourse	Peer reporting and manager patrols; visible, interpersonal discipline in online chatrooms (WeChat) and on the road; penalties mediated by rank

By human mediation, we refer to the discretionary authority that operation specialists exercise alongside the algorithm. Rather than simply following automated instructions by system, these specialists act as intermediaries who can manually adjust orders, scheduling, or penalties in response to real-world situations. Although UE’s algorithmic code is not “valueless” or neutral—as all technologies encode human norms and values that indirectly shape the labor process—Fantuan’s human-mediated model has more channels for ethnocultural values to directly affect workers’ day-to-day experiences. In this context, values specifically refers to ethnocultural norms and expectations. Effectively, UE’s labor control model produces broad participation with low-price guarantees and minimal worker protections, whereas Fantuan’s model produces tiered worker access with higher-price, selective guarantees that hinge on drivers’ reputation, interpersonal workplace relations, and linguistic or cultural fit.

Node 1: Employee Onboarding and Access Structure

By participating in the employee signup process ourselves, we observed clear differences from the outset in how UE and Fantuan structure the onboarding process as well as regulate access to shifts and gigs. Although the processes on both apps function as forms of labor control, the UE model operates by foregrounding highly automated and algorithmic mechanisms that shift risk to workers, whereas the Fantuan model functions by managing workers through a human-mediated process in which supervisors exercise discretion in assessing the merits of individual workers through interpersonal judgment and culturally specific expectations embedded in Fantuan’s Chinese-speaking coordination system.

In its hiring process, UE follows a largely automated and relatively frictionless model that onboards drivers as independent contractors, while quietly loading them with the obligations of an employee. Prospective drivers have no contact with human managers during the hiring process: they simply download the app, supply the platform with extensive documentation (i.e., vehicle ownership, insurance, work eligibility, and third-party background checks) and wait for the platform to issue a final decision through an opaque review process. Although response times vary, most applicants receive approval or rejection within the same day. In practice, UE admits drivers as long as they meet basic eligibility requirements and submit complete documentation; the platform has limited standards for evaluating worker reliability, commitment, or competency at the entry point to employment.

Unlike UE’s automated and standardized system checks, Fantuan’s sign-up model is visibly manualized and often individualized toward each applicant, particularly in smaller local markets. Although applicants are still required to submit their identification and work eligibility documents (such as a phone number, a copy of their driver’s license, and proof of a work permit), through the online system, this information is ultimately reviewed and verified by a local operations specialist. Prior to account activation, drivers typically complete an in-person interview with an operation specialist and are required to watch a set of Fantuan guide videos as part of the sign-up process. The manual review process used by Fantuan creates space for both managerial discretion and bias. For instance, some international students without valid work permits may nonetheless gain access to the platform in practice, depending on internal ties. This discretion also operates through cultural and linguistic expectations embedded in Fantuan’s Chinese-speaking coordination system. In practice, workers’ access to shifts, support, and fuller participation in the platform often depends not only on formal eligibility, but also on their ability to navigate these cultural and communicative norms.

Once workers have been formally employed, UE and Fantuan diverge further in how they structure access to shifts and delivery orders. With UE, drivers can simply log in to the app at any time to potentially receive delivery opportunities. Although drivers have no obligation to accept any delivery opportunity sent to them, UE explicitly states that it does not guarantee the availability of delivery requests while drivers are online.² UE sustains this arrangement by classifying drivers as “independent contractors” rather than employees, allowing the company offload capital expenditures (e.g., vehicle costs, fuel, maintenance, mobile data plans) onto the driver while retaining unilateral control over pricing, profit, incentives, and access to the app. UE’s model creates an appearance of “equal opportunity” by lowering significant barriers to entry, however, eligibility does not translate to availability, as UE makes no guarantees about order availability.

In contrast to UE’s hands-off model that grants relatively equal opportunity for drivers to be eligible for deliveries, Fantuan’s access structure is highly human-mediated and hierarchical. According to internal regulations, Fantuan divides drivers into two categories, “Ricer” (full-time drivers) and “Baller” (crowdsourced drivers). Ricer drivers work scheduled shifts and must remain “online” to execute assigned orders, and receive priority in order allocation along with higher hourly pay, which places them at the center of Fantuan’s scheduling network. By contrast, Baller drivers are crowdsourced externally (from companies such as DoorDash and UE), to complete orders the Ricers cannot. Furthermore, Ricer drivers are usually Chinese drivers directly managed by the Fantuan algorithm and the operation specialist, while most Baller drivers do not speak Mandarin and rely entirely on the automatic management of the system. In driver group chats across the greater Toronto area, nearly all announcements, emergency notifications and operational discussions regarding Fantuan occur in Mandarin. Non-Chinese-speaking drivers are unable to fully access these communications and thus remain excluded from Fantuan’s human-mediated coordination network. The distinction between Ricers and Ballers is therefore not only organizational, but also cultural, insofar as fuller incorporation into Fantuan’s labor process depends on proximity to its Chinese-speaking communication practices and shared norms of coordination. As a result, Fantuan’s access structure functions as a tiered system in which language proficiency and cultural familiarity significantly shape drivers’ access to work.

As Table 4 shows, Baller drivers hold lower priority than Ricers, though the platform still registers them as Fantuan drivers. Beyond the absence of shift restrictions and the flexibility of working hours, many Ballers describe Fantuan as a “backup platform,” using it only to pick up additional orders when convenient. These Ballers primarily rely on other platforms, such as UE or DoorDash, for their main source of income. As Ballers do not participate in Fantuan’s internal communication and support channels because of information gaps created by language barriers, these drivers also report lower expectations for earnings. Although some Ballers express confusion about platform rules and systems, they often lack a clear channel through which to seek clarification. Effectively, Ballers’ marginal position is reflected both in their lower dispatch priority and in their exclusion from the cultural and communicative channels through which Fantuan provides it workers with support, circulates information, and resolves everyday problems.

Table 4.

Fantuan Driver Composition and Order Dispatch Operations.

Category	“Ricer”	“Baller”	Third Parties
Type	Dedicated drivers with scheduled shifts	Crowdsourced drivers with “on-demand” deliveries	Drivers sourced from Uber Eats or DoorDash
Racial composition	Predominantly Chinese drivers	Predominantly non-Chinese drivers (primarily of South Asian or Middle Eastern backgrounds)
Delivery priority	Allocated the highest priority for order dispatch	Allocated second priority for order dispatch; not assigned orders exceeding $40.00 CAD	“Bad” orders are manually allocated to third-party drivers

Node 2: Order Dispatch

For our analysis of the second node, we primarily relied on data collected through our own participant observations and interviews. Once drivers are “online” and eligible for orders, clear structural differences emerge in how UE and Fantuan dispatch deliveries. UE relies entirely on algorithmic order assignment, whereas Fantuan uses its algorithm to generate a preliminary driver assignment which managers then review and finalize. UE’s automated dispatch system enables the company to maximize value extraction by offering the lowest fee a worker is likely to accept and “nudging” drivers toward low-paying orders. Fantuan’s human-mediated structure, by contrast, allows managers to override algorithmic assignments on the basis of their own perception of each drivers’ competence and reliability.

UE’s order dispatch structure individualizes access to revenue-bearing tasks for drivers while simultaneously ensuring the cost and risks associated with earning on the platform falls completely on the workers. Interviewees commonly shared how UE’s order allocation logics double as control measures, nudging couriers to accept low-value deliveries in order to protect the platform’s bottom line. Couriers shared how the platform “send[s] out orders to people who have higher acceptance rates,” though UE “won’t tell us what that percentage is.” Peter, a part-time driver, describes how the platform punishes drivers who repeatedly reject (low-value) orders by offering fewer and worse-quality orders in the future, noting how “it took a really long time to assign the next [delivery] . . . no tip . . . 10-15km for $3.00.” As the platform’s algorithm ‘learns’ about the driver’s behavior and subsequently disseminates orders through priority criteria geared toward proximity, ratings, and reliability (i.e., acceptance and cancellation rates, on-time delivery history), drivers who repeatedly decline low-paying orders ultimately risk minimizing potential future work. By gatekeeping order dispatch behind opaque algorithmic metrics and performative thresholds, UE makes refusal costly and compliance financially necessary; what appears as worker “choice” becomes instrumental to the platform’s managerial power.

Although UE’s dispatch structure is managed by algorithms, Fantuan operation specialists possess broad authority to intervene in the delivery process, and their internal systems grant supervisors considerable flexibility (Wood 2018). In practice, Fantuan’s algorithmic assignments function more as a draft, and the final decision on the distribution of orders rests in the hands of operation specialists, who can modify, revoke, or even disregard the algorithmic dispatch results depending on the situation. For example, if the system assigns a new order that conflicts with a driver’s ongoing route, a specialist may reassign that order to another available driver or outsource the order to a different platform, such as UE. In one instance, an operations specialist described a “very long distance” order requiring pickup from a remote area with little likelihood of return trips. After Fantuan drivers complained via WeChat and refused the assignment, the specialist redirected the order to UE. The fact that only drivers on WeChat—effectively Chinese drivers—can negotiate assignments and have their requests honored demonstrate the peer digital network and a common sense of appropriate demands that are specific to the Chinese-speaking community. Through such interventions, Fantuan managers aim to maintain on-time delivery rates and sustain order volume in the region. To ensure efficiency, specialists combine algorithmic evaluations with their own judgments of each driver’s reliability and capability, demonstrating how human discretion continually shapes the system. Through such flexibility, Fantuan’s delivery process operates as both a flexible and people-oriented system, where interpersonal relationships frequently influence dispatch decisions.

Node 3: Competition and Earnings

For our analysis of the third node, we relied primarily on data collected through interviews, including multiple sessions with two operation specialists. Although gig delivery work inherently produces competition among drivers, given the subcontractual nature of employment and fluctuating order volumes, UE and Fantuan manage this competition in distinct ways. Where UE’s system capitalizes on opaqueness, strategic incentives, and race-to-the-bottom economic logics, Fantuan’s food delivery system largely relies on logics of human judgement and interpersonal competition (Table 5).

Table 5.

Comparison of Pay Structure and Earnings, Uber Eats versus Fantuan.

	Uber Eats	Fantuan
Base pay	Upward of $2.50 CAD (median of $5 CAD) per delivery before tip, with the possibility of surge pricing	Fixed rate based on distance; total earnings fluctuate according to tip
Tips	Inconsistent; risks of tip-baiting	At least a 10 percent tip is guaranteed^a
Hourly earnings	Income varies widely	Typically, $17–$25 CAD per hour, with the potential for more under peak conditions
Delivery expenditure(s)	Drivers are responsible for covering operating expenses such as fuel, maintenance, and insurance	Drivers are responsible for all fuel consumption costs and vehicle damage repairs; Fantuan does not offer any commercial insurance for drivers to purchase
Additional costs	Drivers are charged in two installments after their first week working for Uber Eats, to cover the cost of their background check	A deposit of $200 is required
Income transparency	Only estimated totals are shown to drivers; a detailed earnings breakdown (i.e., tip amounts, tip presence, base delivery supplement) is provided only after the delivery	The actual earning is displayed before the order is accepted and will not fluctuate after the order is completed

After July 2025, customers could set a custom tip, including $0, while the platform charged a higher mandatory service fee from customers. The 10 percent tip was no longer guaranteed.

UE’s compensation is fragmented across multiple, shifting micro-components within the platform (base pay, “surge pricing,” tips, promotions) that fragments earnings and disperses risk downward onto drivers. Interviewees repeatedly noted that base offers often fail to cover operating costs and that hourly earnings frequently fall below minimum wage. Couriers reported hourly returns ranging from approximately $10 to $12 among the lowest observed earnings and occasional highs near $28. However, drivers who reported higher earnings emphasized that higher peaks heavily depended on surge pricing or tips. After deductions in gas and maintenance expenses, part-time couriers reported average earnings of $18.50/hour, whereas full-time couriers averaged roughly $16/hour. In practice, most part-time drivers typically earn about $10.00 per order, and compensation is dependent on various contingencies (time of day, weather, promotions, tips).

The lack of information provided by UE to drivers in their offer screens, payouts, and earning breakdowns enables extraction both at the moment of decision (i.e., when the driver accepts an order) as well as after the fact. Several interviewees noted how UE shows only a total payout upfront—“we can only see the total fare . . . we cannot see what fare Uber is paying and what is the tip”—which ultimately blunts drivers’ capacity to evaluate risk, reject low-tip orders, or target better base-pay work. Drivers also reported that the platform provides only approximate destination details, and that apartment deliveries impose unpaid time costs for parking and elevator access. Many interviewees described “tip-baiting” (i.e., when customers initially promise high tips and later reduce or remove them altogether) as a common practice among UE customers, leaving drivers undercompensated after completing the delivery all while the platform has secured the transaction. Through these mechanisms, UE’s earnings system is obscured by an opaque algorithmic process, denying drivers’ abilities to make informed decisions and therein contributing to wage volatility and insecurity.

For many drivers, the volatility of UE’s pay structure, described as “not like a fixed fare . . . it’s just fluctuations,” leads them to seek out higher volumes of orders and gratuities rather than relying on a predictable rate. In other words, the opacity of earnings pressures drivers to accept low-fee orders simply to maintain income flow, even when such strategies undermine long-term viability (Schor, Tirrell, and Vallas 2024). These dynamic foster a race-to-the-bottom work environment, wherein the economic pressure and algorithmic opacity mitigates drivers’ abilities to prioritize decent paying orders, which in turn leads to decreased fees for all drivers because of the inherent competitiveness of subcontractual work. By concealing key parameters within the platform’s pay structure and weaponizing algorithmic informational opacity as a technique of wage governance, UE preserves unilateral flexibility to drive down wages for all of its drivers.

As previously described, Fantuan embeds its dispatch process within interpersonal negotiation networks inside the Chinese driver community, and operation specialists influence assignments through their personal preferences and emotional bonds. Beyond expressions of personal liking, Fantuan managers’ discretionary preferences operate as informal judgements about which drivers are most dependable, responsive, and easiest to coordinate, particularly under time-sensitive conditions; and these judgements themselves are shaped through ongoing Chinese-language WeChat interactions within Fantuan’s driver community.

When orders surge beyond capacity during peak demand periods, such as winter breaks or during heavy snowfall, specialists prioritize drivers who respond quickly on WeChat, remain easy to contact, and follow rerouting instructions efficiently. Although specialists meet short-term area targets by prioritizing these drivers, this practice obscures the criteria for success as a Fantuan driver and can generate perceptions of favoritism in the workplace. Indeed, several Fantuan drivers reported that maintaining a strong personal relationship with a manager increases their chances of receiving priority scheduling and higher tip orders. These advantages certainly reflect the relationship between Fantuan’s platform design and managerial discretion, but moreover, they are produced through a cultural environment where ongoing Chinese-language interaction shapes who is regarded as reliable, trustworthy, and easy to coordinate.

Fantuan’s ranking system further structures this competitive environment. According to the Fantuan driver manual, the top-ranked drivers earn an additional $0.50 CAD per order and can accept multiple orders simultaneously, incentives designed to encourage compliance with platform standards. The platform frames these perks as benevolent rewards, but they also function as governance tools that tie material benefits to rule adherence (Healy and Pekarek 2025). Formally, the algorithm calculates rankings according to punctuality, attendance and customer ratings, and these metrics determine access to bonuses and order volume. In practice, however, operation specialists retain discretion to override these rankings and can assign high-value or convenient orders to lower ranked drivers whom they trust or perceive as cooperative. Drivers who cultivate positive relationships with specialists usually secure better routes and higher earnings. This pattern of discretionary assignment also creates cumulative structural advantages for the platform: favored drivers receive better trips, allowing them to improve their formal ranking more quickly. As such, Fantuan’s earning structure fuels internal competition, reflected in significant hourly wage disparities that ranged from approximately $17/hour to more than $40/hour. To maximize earnings, drivers must invest in both performance metrics but also in cultivating relationships with supervisors to improve their perceived reliability.

For Fantuan, WeChat serves as a key site where these informal competitions unfold. Drivers widely use the platform for work communication, often blurring the line between work and nonwork (He 2025). Within WeChat groups, drivers have developed an informal reciprocal practice known as “red envelopes” exchanges, which have now become a common convention. As Fantuan’s algorithm does not formally allow drivers to swap or relinquish shifts within the app, drivers use WeChat to coordinate replacements. If a driver cannot work because of an emergency, they can locate a substitute in the group and transfer a “red envelope” payment directly to that driver, bypassing algorithmic mediation and relying entirely on interpersonal networks. As a result, drivers also feel pressure to maintain strong relationships with fellow drivers in order to secure unscheduled shifts and supplemental “red envelope” income. Interpersonal competition therefore stands at the center of Fantuan’s earnings structure. Because these exchanges take place through WeChat groups and within a largely Chinese-speaking driver community, shift replacement also reflects the cultural and communicative environment in which Fantuan’s work is organized.

Node 4: Worker Surveillance and Discipline

The final node in which we notice a difference in the apps’ operational logics is in worker surveillance and discipline. For our analysis of the fourth node, we primarily relied on data collected through textual reading of driver manuals combined with interviews and app walk-throughs. Both platforms create ambiguities that generate a panoptic sense of surveillance, however, UE produces ambiguity through flexible and opaque algorithmic metrics that obscure how the platform evaluates and punishes behavior, whereas Fantuan generates ambiguity through cultural expectations and interpersonal judgment. In both systems, unclear standards surrounding discipline shape work to fashion the drivers’ job compliance.

UE mobilizes a sense of ambiguity directly into its delivery agreement and terms of service, with the frequent use of terms such as unreasonable and fraud that have a significant impact on employee discipline and termination but lack clear definitions for drivers. For instance, although UE makes clear that “there are generally no consequences for accepting or rejecting orders,” the delivery agreement includes the caveat that “repeat cancellations may also trigger fraud-monitoring processes.” The guidelines further specify that the platform may reduce or cancel delivery fees for a number of reasons, including if “(iii) the route chosen by you was unreasonably inefficient, (iv) the time to complete the Delivery Service was unreasonably lengthy, or (v) suspected Fraud or misuse has been detected.” Under section 13.2, UE stipulates that “fraud” may lead to termination, yet the agreement never specifies how many cancellations trigger suspicion or how the platform determines that a route or delivery time qualifies as “unreasonable.” By leaving these thresholds undefined, UE is able to (mis)classify drivers as subcontractors by claiming worker autonomy and “choice” over deliveries while simultaneously incorporating disciplinary mechanisms to police driver conduct by threatening pay reduction or termination. The ambiguities in the terminology of the UE employment documents produces an information asymmetry between the drivers and the company about the terms of their employment, allowing greater value extraction by retaining termination discretion on the basis of flexible factors such as local markets.

When thresholds are unclear and recourse thin, drivers hedge against penalties by accepting low-value orders to maintain their ratings and eligibility, however, in doing so, they reinforce UE’s algorithmic model that normalizes such low-value dissemination. Together, these practices individualize access to revenue while preserving platform discretion and algorithmic governmentality. Thus, the environment UE creates for its workers is one which demands that couriers be highly visible (to both the platform and its algorithmic logics) but that the criteria and the consequences of this “high visibility” remain largely invisible to those couriers.

Importantly, Fantuan’s surveillance is not purely interpersonal or outside algorithmic control. Like UE, it also relies on top-down, platform-based monitoring. For example, the platform tracks the routes and real-time locations of all online drivers within the area. For Ricer drivers, the system also issues random pop-up facial recognition checks during delivery, requiring them to verify their identity within a specified time. The difference is that, whereas UE uses algorithmic data collection and shifting thresholds to observe and maintain worker compliance, Fantuan’s surveillance is further reinforced through lateral observation, peer reporting, and interpersonal enforcement.

One prominent example appears in dress-code enforcement. Unlike UE, Fantuan encourages drivers to display company branding by wearing Fantuan vests, using branded delivery boxes, and placing logo stickers on their vehicles. Although the contract does not formally mandate these practices, the app offers a “promotional bonus” for following these expectations that accounts for roughly 8 percent of a driver’s total income. One driver remarks, the platform is “quite cunning” insofar as it “gives me what is originally mine as a promotional bonus.” Although the company frames this payment as brand promotion, several drivers interpret it as a reclassification of ordinary delivery pay rather than an additional reward.

To enforce these expectations, Fantuan relies on peer and on-road surveillance that has greatly expanded the management scope of the platform. Managers encourage workers to report drivers who fail to display company logos, effectively expanding managerial reach through mutual oversight and internal community pressure. Several interviewees described instances of “mutual reporting,” including one case in which a driver photographed another driver without a uniform and forwarded the image to an operations specialist. Some supervisors privately endorse such practices, suggesting that reporting noncompliant drivers improves one’s standing and may influence the allocation of higher paying orders. These practices are rooted in Chinese-speaking peer networks wherein surveillance is carried out by both managers and fellow workers observing and evaluating one another. As a result, discipline operates not only through formal platform monitoring, but also through community-based expectations about proper conduct within Fantuan’s driver community.

Fantuan reinforces this lateral surveillance through geolocation-based tracking and “on-road observation”, making workers feel permanently trackable (Heiland 2021). Fantuan’s regional managers regularly patrol areas dense with Chinese restaurants across the greater Toronto area in order to inspect driver vehicles for stickers, verify whether drivers wear uniforms, and confirm that couriers carry branded boxes. When violations are identified, managers may even photograph the driver and post the image in internal group chats as a public warning. Other drivers often interpret such incidents as cautionary “lessons,” reinforcing a culture of vigilance and self-restraint. Fantuan, therefore, cultivates a lateral panopticon that contrasts with UE’s black-box algorithmic oversight. Rather than relying primarily on invisible code, Fantuan leverages the interpersonal and communal dimensions of work to enforce discipline.

Conclusion

Our comparative analysis of UE and Fantuan demonstrates that labor control under platform capitalism cannot be reduced to a singular or universal model of algorithmic management. Although the existing literature documents the emergence of “algorithmic despotism” (Griesbach et al. 2019), our findings show that platform governance emerges not only from datafication, opacity, and economic nudges, but also from human practices and ethnocultural values embedded in platform design. By systematically tracing four key “nodes” of labor management—from onboarding and order dissemination to compensation structures and dispute resolution—we show that UE and Fantuan enact two distinct regimes of labor control that rely on algorithmic governance while drawing on different sociocultural dynamics.

UE’s dispersive-extractive model extends and intensifies dominant logics in the platform economy: individualized competition, information asymmetry, fractured pay systems, and the threat of algorithmic surveillance (Gandini 2019; Rosenblat and Stark 2016; Veen et al. 2020; Wood et al. 2019). These mechanisms fragment worker decision-making and maximize value extraction by pushing risk onto drivers. Although the dispersive-extractive model can be understood as one model of algorithmic despotism in which arbitrariness is the key element of control, we argue that it is the dispersive nature of the app that needs to be emphasized. Prior to the advent of platforms when labor management was done by humans, workers can encounter different level managers and departments (e.g., human resources) and therefore had more opportunities to express discontent. Under a platform economy, though, algorithms now absorb many of these nodes, including onboarding and employment agreements, reducing opportunities for direct negotiation between employees and employers. (Glavin and Schieman 2022; Joseph-Goteiner 2024; Reynolds et al. 2024; Wood et al. 2019) As a result, platforms produce a form of control that remains opaque and structurally insulated from challenge, making worker dissent more difficult to articulate or enact (Duggan et al. 2020; Yao 2025).

In contrast, Fantuan’s ethnocultural model reveals how platform governance can be shaped by embedded ethnocultural norms that manage expectations, facilitate communication, and streamline work routines, but also reinscribe ethnoracial and linguistic hierarchies within the labor process. Rather than treating algorithmic control as a value-free system, Fantuan’s model highlights how platform management can integrate cultural values into its regulatory logic—sometimes to the benefit of efficiency and sometimes to the detriment of workers positioned differently within the platform’s hierarchy. The ethnocultural model advanced in this article complements the framework of “racial platform capitalism” by urging greater attention to a broader range of social identities and cultural dynamics in platform analysis. Although racial and ethnocultural dimensions are intertwined, they are not interchangeable. The former often functions as a broad classificatory category, whereas the latter refers more specifically to language, norms, values and other forms of cultural identification associated with a particular ethnic group. Fully understanding Fantuan’s labor management model therefore requires an analytic approach that accounts for this wider constellation of social factors. Beyond theoretical implications, Fantuan’s ethnocultural model raises some regulatory challenges for policy makers as they seek to address labor issues in the platform economy. For example, New York City recently passed the Intro 1332 Bill, which provides protection against unjust deactivation of driver accounts. Similarly, Ontario’s Digital Platform Workers’ Right Act 2022, which came into force on July 1, 2025, requires platforms to provide two weeks’ notice before suspending drivers for 24 hours or longer. Fantuan drivers, however, are subject to temporary suspensions that last less than 24 hours and that operation specialists administer at their discretion. Although these temporary suspensions may seem inconsequential to outsiders, they greatly affect drivers’ daily income, especially when they occur during peak hours. As such, Ontario may need more nuanced legislation to address these gaps before other mainstream platforms adopt similar practices.

In sum, these two cases call for a more relational theorization of algorithmic management, one attentive not only to the technical workings of algorithms but also to the cultural, linguistic, and organizational infrastructures that shape how algorithmic authority is exercised and experienced. Rather than treating platform capitalism as a uniform system, we argue that it operates as a heterogeneous field composed of multiple labor control models shaped by interactions between algorithmic architecture, local labor markets, and the social contexts from which platform firms emerge. Our research contributes to the literature on labor management on platform apps by further interrogating the diverse typologies of algorithmic management and demonstrating that ethnocultural dynamics can have significant impacts on the operations of platform models.

Future research would benefit from extending this comparative framework to additional platforms, markets, and ethnocultural contexts. Such work could further clarify how platform governance varies across geographic, regulatory, and demographic contexts, and how workers navigate platform logics in their daily practices. As food-delivery platforms continue to expand and diversify, understanding the social foundations of algorithmic control will remain crucial for scholars, policymakers, and labor advocates seeking to address the implications of gig work for worker rights, economic inequality, and the evolving nature of labor in the digital age.

Footnotes

ORCID iDs

Tommy Wu

Junyi Zhang

Zachary Gan

Ayra Thomas

Ethical Considerations

This study has been reviewed and received ethics clearance from McMaster University.

Consent to Participate

Participants were invited to participate in a research study exploring the experiences of food delivery app drivers in Canada. Participation in this study was voluntary, and participants were given the right to withdraw at any time without penalty. Participants were provided with letters of information and consent forms prior to the study; informed consent to participate was obtained both verbally and in writing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Financial support for this project was provided by the Social Sciences and Humanities Research Council of Canada.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

1

For instance, if a delivery driver fails to complete an order because of bad driving conditions, the algorithm can capture only the binary result of that delivery as a success or failure, and it is up to the driver to contact a dispatcher, evoking managerial surveillance ().

2

As stated in the UE delivery agreement, provided upon registration as a driver.

Author Biographies

Tommy Wu is an assistant professor in the School of Labour Studies at McMaster University. His research lies at the intersection of political economy, the sociology of work, and race and ethnic studies, with a focus on migrant labor. Drawing on ethnography and community-based research, his work examines labor mobility, racialization, and technology under contemporary regimes of labor control. His scholarship has been published in Labour/Le Travail, Labor Studies Journal, and the Journal of Asian American Studies.

Junyi Zhang completed an MA in globalization and the human condition at McMaster University. Her research lies at the intersection of media and cultural studies, political economy, and affect studies, with a focus on platform-mediated creative labor in East Asia. She is developing a project on affective labor and commissioning practices on Anime, Comic, and Games (ACG) commission platforms in East Asia.

Zachary Gan is a graduate student in human geography at the University of Toronto and a fellow at the Institute for Environment, Conservation, and Sustainability. He researches the political geography of extractive capitalism, the infrastructural politics of Canada’s hegemonic green transition, and state efforts to establish a critical minerals industrial strategy. His writing on eco-Marxism and new materialism has been published in Capitalism Nature Socialism.

Ayra Thomas received her BA(Hons) in English and philosophy at the University of Toronto and her master’s degree from McMaster University’s Department of English and Cultural Studies. She remains dedicated to analyzing literature, ethnography, cultural artifacts, and visual iconography through the intersecting lenses of feminist, sexuality, late colonial, and critical race studies. Her recent and forthcoming publications are available in (Un)Disturbed: A Journal of Feminist Voices, The Oracle, and Pivot: A Journal of Interdisciplinary Studies and Thought.

References

Ametowobla

Dzifa

Kirchner

Stefan

. 2025. “Varieties of Platform Capitalism? Competition, Regime Types and the Diversity of Food Delivery Platforms across Europe and North America.” Socio-economic Review 23(2):899–931.

Braverman

Harry

. 1998. Labor and Monopoly Capital: The Degradation of Work in the Twentieth Century. New York: Monthly Review Press.

Burawoy

Michael

. 1982. Manufacturing Consent: Changes in the Labor Process Under Monopoly Capitalism. Chicago: University of Chicago Press.

Burawoy

Michael

. 1985. The Politics of Production. London: Verso.

Cameron

Lindsey D.

2022. “‘Making Out’ While Driving: Relational and Efficiency Games in the Gig Economy.” Organization Science 33(1):231–52.

Corbin

Juliet

Strauss

Anselm

. 2014. Basics of Qualitative Research: Techniques and Procedures for Developing Grounded Theory. Washington, DC: Sage publications.

Davis

Gerald F.

Sinha

Aseem

. 2021. “Varieties of Uberization: How Technology and Institutions Change the Organization(s) of Late Capitalism.” Organization Theory 2(1).

Duggan

James

Sherman

Ultan

Carbery

Ronan

McDonnell

Anthony

. 2020. “Algorithmic Management and App-Work in the Gig Economy: A Research Agenda for Employment Relations and HRM.” Human Resource Management Journal 30(1):114–32.

Edwards

Richard

. 1979. Contested Terrain: The Transformation of the Workplace in the Twentieth Century. New York: Basic Books.

10.

Franke

Milena

Pulignano

Valeria

. 2022. “Labor Control and Commodification Strategies within a Food Delivery Platform in Belgium.” Pp. 135–48 in Digital Platforms and Algorithmic Subjectivities, edited by Armano

Briziarelli

Risi

London: University of Westminster Press.

11.

Gandini

Alessandro

. 2019. “Labor Process Theory and the Gig Economy.” Human Relations 72(6):1039–56.

12.

Gebrial

Dalia

. 2024. “Racial Platform Capitalism: Empire, Migration, and the Making of Uber in London.” EPA: Economy and Space 56(4):1170–1194.

13.

Glavin

Paul

Schieman

Scott

. 2022. “Dependency and Hardship in the Gig Economy: The Mental Health Consequences of Platform Work.” Socius. doi:10.1177/23780231221082414.

14.

Griesbach

Kathleen

Reich

Adam

Elliott-Negri

Luke

Milkman

Ruth

. 2019. “Algorithmic Control in Platform Food Delivery Work.” Socius. doi:10.1177/2378023119870041.

15.

Tianjiao

. 2025. “Managing Hybrid Social Media: A Case Study of Employees’ Boundary Management Strategies on WeChat.” New Technology, Work and Employment 40(3):588–99.

16.

Healy

Joshua

Pekarek

Andreas

. 2025. “The Triangular Relationship in Platform Gig Work: Consumers, Platform Beneficence and Worker Vulnerability.” New Technology, Work and Employment 40(2):265–84.

17.

Heiland

Heiner

. 2021. “Controlling Space, Controlling Labor? Contested Space in Food Delivery Gig Work.” New Technology, Work and Employment 36(1):1–16.

18.

Hochschild

Arlie Russell

. 1979. “Emotion Work, Feeling Rules, and Social Structure.” American Journal of Sociology 85(3):551–75.

19.

Howcroft

Debra

Bergvall-Kåreborn

Birgitta

. 2019. “A Typology of Crowdwork Platforms.” Work, Employment and Society 33(1):21–38.

20.

Huang

Hui

. 2023. “Algorithmic Management in Food-Delivery Platform Economy in China.” New Technology, Work and Employment 38(2):185–205.

21.

Joseph-Goteiner

David

. 2024. “From Degrees to Dimensions: Accounts of Workers’ Socioeconomic Dependence on Platforms.” Socius. doi:10.1177/23780231241275412.

22.

Kang

Miliann

. 2010. The Managed Hand : Race, Gender, and the Body in Beauty Service Work. Berkeley: University of California Press.

23.

Lee

Min Kyung

Kusbit

Daniel

Metsky

Evan

Dabbish

Laura

. 2015. “Working with Machines: The Impact of Algorithmic and Data-Driven Management on Human Workers.” Pp. 1603–12 in Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems. New York: Association for Computing Machinery.

24.

Lei

Ya-Wen

. 2021. “Delivering Solidarity: Platform Architecture and Collective Contention in China’s Platform Economy.” American Sociological Review 86(2):279–309.

25.

Liang

Yin

Aroles

Jeremy

Brandl

Bernd

. 2022. “Charting Platform Capitalism: Definitions, Concepts and Ideologies.” New Technology, Work and Employment 37(2):308–27.

26.

Murthy

Dhiraj

. 2008. “Digital Ethnography: An Examination of the Use of New Technologies for Social Research.” Sociology 42(5):837–55.

27.

Newlands

Gemma

. 2021. “Algorithmic Surveillance in the Gig Economy: The Organization of Work through Lefebvrian Conceived Space.” Organization Studies 42(5):719–37.

28.

Padios

Jan M.

2018. A Nation on the Line: Call Centers as Postcolonial Predicaments in the Philippines. Durham, NC: Duke University Press.

29.

Reynolds

Jeremy

Quintero

Daniel Felipe Pinzón

Aguilar

Julieta

Kincaid

Reilly

. 2024. “Work Schedules, Finances, and Freedom: Work Schedule Fit and Platform Dependence among Gig Workers on Amazon’s Mechanical Turk Platform.” Socius. doi:10.1177/23780231241286933.

30.

Rosenblat

Alex

Stark

Luke

. 2016. “Algorithmic Labor and Information Asymmetries: A Case Study of Uber’s Drivers.” International Journal of Communication 10:27.

31.

Schaupp

Simon

. 2022. “Algorithmic Integration and Precarious (Dis)Obedience: On the Co-constitution of Migration Regime and Workplace Regime in Digitalised Manufacturing and Logistics.” Work, Employment & Society 36(2):310–27.

32.

Schor

Juliet B.

Tirrell

Christopher

Vallas

Steven Peter

. 2024. “Consent and Contestation: How Platform Workers Reckon with the Risks of Gig Labor.” Work, Employment and Society 38(5):1423–44.

33.

Tironi

Martin

Albornoz

Camila

. 2022. “Surveillance and the Ecology of Frictions in Platform Urbanism: The Case of Delivery Workers in Santiago de Chile.” Tapuya: Latin American Science, Technology and Society 5.

34.

Vallas

Steven P.

2019. “Platform Capitalism: What’s at Stake for Workers?” New Labor Forum 28(1):48–59.

35.

Veen

Alex

Barratt

Tom

Goods

Caleb

. 2020. “Platform-Capital’s ‘App-Etite’ for Control: A Labor Process Analysis of Food-Delivery Work in Australia.” Work, Employment and Society 34(3):388–406.

36.

Wood

Alex J.

2018. “Powerful Times: Flexible Discipline and Schedule Gifts at Work.” Work, Employment and Society 32(6):1061–77.

37.

Wood

Alex J.

2021. “Algorithmic Management: Consequences for Work Organisation and Working Conditions.” European Commission. Retrieved May 21, 2026. https://publications.jrc.ec.europa.eu/repository/handle/JRC124874.

38.

Wood

Alex J.

Graham

Mark

Lehdonvirta

Vili

Hjorth

Isis

. 2019. “Good Gig, Bad Gig: Autonomy and Algorithmic Control in the Global Gig Economy.” Work, Employment and Society 33(1):56–75.

39.

Woodcock

Jamie

Cant

Callum

. 2022. “Platform Worker Organising at Deliveroo in the UK: From Wildcat Strikes to Building Power.” Journal of Labor and Society 25:1–17.

40.

Tommy

. 2024. “Spatial Labour Control: Experiencing Labour Mobility in the Chinese/Asian Restaurant Industry in the United States.” Labour/Le Travail 93:115–34.

41.

Yao

Wei

. 2025. “Delineating the Multifaceted Consequences of Algorithmic Management for Gig Workers—A Literature Review.” Selected Papers of the IRIS 16(2025):1.