Delivering Solidarity: Platform Architecture and Collective Contention in China’s Platform Economy

Escalating grievances and perceived injustice

Sociolegal and social movement scholars point out the importance of analyzing the interpretive process of constructing grievances and injustice frames to understand the emergence and escalation of contention (Felstiner, Abel, and Sarat 1980; Gamson 1995; Levitsky 2008). They use the term “injustice frames” to refer to interpretations of experiences or conditions that show moral principles have been violated and ought to be readdressed rather than tolerated (Gamson 1995). Research shows that developing and sharing injustice frames is a necessary, albeit insufficient, condition for collective mobilization.

Building on social movement literature, scholars of industrial relations argue that labor’s collective action requires constructing a sense of injustice from grievances. A work situation must be understood as wrong or illegitimate rather than simply being dissatisfying (Beck and Brook 2020; Kelly 1998). The focus on constructing a sense of injustice resonates with labor research that analyzes the interplay between the labor process, morality, and solidarity (Beck and Brook 2020). Perceived injustice often manifests in collective action frames, which comprise a set of interconnected ideas about the existence of problems, causes of and responses to the problems, and rationales for labor collectivism (Gahan and Pekarek 2013). Although some research analyzes labor collectivism from the perspective of rational analysis, labor scholars find collective action frames go beyond the calculation of benefits and costs, and they highlight the moral imperatives for collective action (Morgan and Pulignano 2020).

Critically, labor scholars point to the need to study moments of escalation, as they often become moments of collectivism and generate “cultures of solidarity”—emergent oppositional practices and meanings—as Fantasia (1988:17) defines the term. At unusual moments of escalation, workers develop new repertoires of action, network ties, and moral valuations and suspend customary practices of daily life (Atzeni 2009; Fantasia 1988).

Enhancing the appeal of collective contention

Even when grievances are escalated and when collective action frames are constructed, workers still have various options regarding whether and how to respond to perceived injustice. The literature suggests a collective action frame will not prevail if workers see problems as inevitable or soluble only by individual exit, not collective voice (Kelly 2018). Indeed, research on low-paid workers in the United Kingdom finds that, although problems at work are widespread, satisfactory resolutions are rare because workers tend to discuss issues only with managers or just find another job (Pollert and Charlwood 2009).

Sociolegal scholarship offers the concept of the “dispute tree” to capture the wide range of options a grievant can pursue, from formal legal disputes in court to institutionalized non-court alternatives, individual informal legal mobilization, and extra-institutional collective action (Albiston, Edelman, and Milligan 2014). Indeed, workers can choose one or a few options from the “dispute tree.” Arguably, when workers develop a strong sense of injustice and consider institutionalized channels or individual informal legal mobilization ineffective, extra-institutional collective action can become a more appealing option.

Making available free spaces for collective contention

Scholars also argue that collective action requires certain free spaces to develop. Free spaces—whether physical or virtual—are “small-scale settings within a community or movement that are removed from the direct control of dominant groups, are voluntarily participated in, and generate the cultural challenge that precedes or accompanies political mobilization” (Polletta and Kretschmer 2013:1). Such spaces are critical as they provide a relatively autonomous environment to develop networks, skills, and solidarity that assist collective action. In the context of industrial relations, free spaces mean spaces with relatively less oversight by management that are less influenced by fear of managerial retribution, normative control, and surveillance (Thompson and van Den Broek 2010). Free spaces allow workers to construct grievances and a sense of injustice, develop collective action frames, and evaluate various dispute resolution options. Labor’s collective contention does not necessarily require formal organization, but it does require rudimentary organizing efforts (Kelly 2018)—efforts made possible in such free spaces.

The multidimensionality of platform architecture

Most of the literature on labor control in platform economies emphasizes algorithmic control (Griesbach et al. 2019; Lee et al. 2015; Rosenblat and Stark 2016), but recent research has examined the multidimensionality of labor control and management (Moore and Joyce 2019; Veen et al. 2019). Indeed, legal scholarship theorizes the co-existence of multiple models of control and regulation in society—law, technology, and social or organizational norms (Lessig 1999). These insights can be applied at the level of labor process. As such, I seek to analyze technological, legal, and organizational dimensions of control and management, as well as the relations among these dimensions. I use the term “platform architecture” to capture these multidimensional relations of control and management holistically.

Technological control and management refers to control and management that uses artifacts or “non-physical, systematic methods of making or doing things” (Hecht 1998:11). Prior studies show platform companies use algorithms not only to “match” consumers with service providers, but also to manage and control workers’ behavior, converting workers’ general capacity for work into a specific and quantified amount of labor (Goods, Veen, and Barratt 2019; Griesbach et al. 2019; Moore and Joyce 2019; Rosenblat 2018; Rosenblat and Stark 2016; Wood et al. 2019). For example, Rosenblat and Stark (2016) detail how Uber utilizes algorithmic control—characterized by information asymmetries and opacity of technological design—to shape drivers’ behavior. Griesbach and colleagues (2019) analyze how food-delivery platforms use algorithms to calculate prices, influence workers’ choices, and control workers’ activities.

Platform economies also require legal arrangements that govern work and delimit the rights and obligations of various parties involved. Most studies of platform economies concentrate on the contractual classification of workers: platform companies often seek to classify workers as independent contractors rather than employees (Kenney and Zysman 2016; Vallas 2017; Veen et al. 2019). This classification is critical, for it shapes the basic rights and obligations of both parties and the degree of precarity and autonomy of work.

Building on Suchman’s (2003) work that analyzes contracts as social artifacts, I examine contractual design and processes in the platform economy rather than focusing on classification only. Legal scholars note the problems and inherent unjustness of contractual design in asymmetric contracts—that is, contract relationships between a dominant business and another market player with unequal bargaining power (Roppo 2009). Because platforms tend to have enormous market power and resources to make legal arrangements (Collier et al. 2017), they can use asymmetric contracts as a means for labor control and management. As legal scholars theorize, a contract is a contingent relational process rather than a static transaction (Macneil 1999). Thus, problems can result from contractual design long after the initial consent.

The organizational dimension of control and management refers to the extent to which management personnel can exercise normative influence and surveillance on platform workers. In traditional workplaces, management personnel tend to have a close supervisory relationship with workers and influence how workers respond to grievances (Pollert and Charlwood 2009). Platform workers, however, such as Uber drivers, often do not have a human supervisor (Rosenblat and Stark 2016). Rather than hiring management personnel to supervise workers, some platform companies invest in professionals who develop technologies and analyze data to automate the supervision of platform workers. In such scenarios, management personnel manage professionals rather than workers.

How platform architecture influences collective contention

The technological, legal, and organizational dimensions of control and management can influence key mechanisms for collective action in several ways. First, studies of the platform economy show that the opaque, disciplinary, and arbitrary characteristics of algorithmic control and management can generate considerable grievances (Rosenblat and Stark 2016; Tassinari and Maccarrone 2017, 2020). When platforms exercise stringent algorithmic control of workers’ time and activities, workers not only feel dissatisfied with work but consider the platforms unfair (Griesbach et al. 2019). Extant literature thus suggests certain characteristics of technological control and management can generate grievances and a sense of injustice.

Second, the legal design of platform architecture can affect how workers construct grievances and injustice frames and, in turn, the perceived appeal of collective contention. Certain contractual designs can deviate from workers’ moral norms. Legal scholars consider asymmetric contracts unfair (Roppo 2009). Workers can share similar moral judgments when they experience enormously unequal power relations. Previous research on labor contention in China finds that workers developed a shared injustice frame that challenges the gap between “law on the book” and “law in action,” and they invoked labor laws to demand resolution to their grievances (Gallagher 2017). I expect that when a platform’s legal design is not based on labor law, workers are more likely to construct injustice frames based on a more abstract moral principle. Furthermore, when injustice frames are based on an abstract moral principle rather than a specific legal clause, grievants are more likely to choose collective contention because it is difficult for them to turn injustice frames into a concrete legal claim in the institutionalized legal process (Fritsvold 2009). In contrast, when injustice frames concern the gap between “law on the books” and “law in action,” grievants can, relatively easily, make a legal claim through the institutionalized legal process or individual negotiations.

Third, the organizational design of platforms can influence the construction of grievances, the appeal of collective contention, and the availability of free spaces. Previous research on low-paid workers shows that communications between managers and workers often prevent grievances from escalating (Pollert and Charlwood 2009), and regular interactions can provide management opportunities to exercise normative influence on workers (Thompson and van Den Broek 2010). Recent research on the service sector further shows that when supervisors perform the emotional labor of managing workers’ feelings and expressions, it can improve workers’ emotional labor and increase their work satisfaction (DiCicco-Bloom and DiCicco-Bloom 2019). Accordingly, lack of regular interaction between management and platform workers under certain platform architectures could make grievances more likely to develop and escalate. The perceived implausibility of grievance resolution through management could also increase the appeal of collective contention. Finally, the absence of supervisory relationships between management personnel and workers can make certain free spaces available for mobilizing solidarity and collective action (Tassinari and Maccarrone 2020).

Overlap of control and management

The three dimensions of labor control are analytically distinct, but they can operate simultaneously under a platform. Such overlap does not negate the need to distinguish the three dimensions analytically, as how the dimensions interact can vary across platform architectures and have different implications for collective contention.

The overlap of technological and legal control and management often results from the mutually enabling relationship between technologies and contracts under certain platform architectures. Technologies can mediate the formation and implementation of contracts. Platform companies often design a “clickwrap agreement,” a digital prompt that allows individuals to accept or decline a digitally-mediated agreement. People cannot use a platform app unless they accept the clickwrap agreement. Because clickwrap agreements take the form of a non-negotiated take-it-or-leave-it choice, they tend to give the party designing the agreement tremendous power (Hubley 2016; Smith 2016) and are often asymmetric contracts (Roppo 2009).

Contracts can also enable and facilitate technological control and management. Algorithms operate according to a set of rules that can be changed to achieve certain goals, for example, to improve their optimization. Certain contractual designs, such as the right to change contractual terms unilaterally in a clickwrap agreement, allow companies to easily change contractual terms and work policies (Lessig 1999). This enables management to enact legal and technological updates to improve platforms’ performance. Existing literature emphasizes that technological control and management occurs in black boxes and remains opaque to platform workers (Moore and Joyce 2019). However, certain legal updates have to be overt to take effect and be followed by platform workers; when changes in technological control and management intersect with such legal changes, workers can clearly see these changes. As platforms’ power becomes more visible, moments of simultaneous legal and technological change can become moments of grievance escalation (Atzeni 2009; Fantasia 1988).

Technological and organizational dimensions of control and management can also overlap because platform companies use technologies to monitor workers, whether or not workers have a human supervisor. Technological and organizational control can interact in various ways. When a manager uses technologies to supervise workers, the two dimensions of control can have opposite effects on collective contention. Certain methods of technological control and management can generate grievances, but supervisors’ mediation could reduce grievances. When supervision is completely automated through technology, workers’ problems with technological control might be exacerbated by the absence of managers to respond to the problems. Ultimately, automated supervision remains limited in its ability to exercise normative influence and perform emotional labor (Hochschild 1983).

In summary, my theoretical framework analyzes how platform architecture influences collective contention, which requires workers develop a sense of grievance, consider collective contention appealing, and have access to free spaces for collective action. The technological dimension of platform architecture can generate considerable grievances. The legal dimension affects grievance construction and the appeal of collective contention. And platforms’ organizational design can influence the construction of grievances, the appeal of collective contention, and the availability of free spaces. When different dimensions of control and management reinforce each other in ways that escalate grievances and perceived injustice, enhance the appeal of collective contention, and make free spaces available for collective contention, collective contention is more likely to arise.

Architecture of the service platform

The architecture of the service platform aims to establish a reliable workforce. As Liu, an in-house lawyer at Meituan, explained, building a dependable workforce requires a high level of supervisory power over couriers, but such power also requires bearing the legal responsibilities associated with conventional employment relations.

Both Meituan and Ele.me franchise their business to other companies to create a reliable workforce. Franchisees operate service stations in a local jurisdiction, and they employ and supervise district managers, service-station supervisors, and delivery couriers. Meituan and Ele.me provide their franchisees with technology—a platform system for managing delivery orders and couriers. The two franchisors oversee their franchisees through a series of key performance indicators (KPIs). These indicators are aggregated from the performance of individual couriers through the platform system. Cai, a district manager, described how KPIs work:

Platform companies set many KPIs, such as the on-time rate, customer review scores, the number of complaints, etc. If our KPIs are not good, Meituan can terminate our contract or give us a fine. Therefore, we ask everyone to have good KPIs.

Meituan and Ele.me’s franchisees use the technology provided by the two franchisors—specifically, the platform system—to assist in managing and controlling the labor process. Technology is thus a tool used by management personnel, rather than something that replaces them. On the one hand, franchisees and station supervisors are constrained by the platform system. Station supervisors mostly rely on the platform system and the algorithms embedded in it to dispatch couriers, but they can also manually dispatch couriers. Dispatch algorithms consider couriers’ performance metrics when assigning tasks. The platform system incorporates various disciplinary rules, such as fines, and it sets the maximum time for completing each delivery. Franchisees can change certain parameters in the system, such as the amount of fines imposed on couriers.

Meituan and Ele.me’s technological control and management also includes dimensions of “gamification” (Deterding 2019). The mobile apps are designed to induce couriers’ participation in the “game,” with a lively graphic design reminiscent of video games. Similar to popular video games, Meituan and Ele.me’s apps classify couriers according to their performance metrics. Higher-level couriers earn certain privileges, such as priority in taking an order. The classification is dynamic, so couriers have to maintain their scores to avoid a downgrade. One SPC said, “I was amazed by the fact that working can be like video gaming. I competed with other couriers and tried to raise my level.”

Legally, Meituan and Ele.me’s franchisees, rather than Meituan or Ele.me, form a labor contract with SPCs that classifies couriers as employees. How franchisees establish a labor contract with couriers varies across service stations, but such contracts must establish a salary structure, both parties’ obligations, and the supervisory relationship. All of the labor contracts calculate salaries based on piece rates and the number of delivery tasks. Some contracts include a minimum monthly wage, but some do not. ⁶ In most service stations, supervisors display disciplinary rules on the wall.

The organizational design of the service platform emphasizes the role of management personnel in supervising couriers. Management personnel aim to maximize KPIs through managing couriers. Huang, a service platform supervisor with a high school education, explained: “The platform system is a useful tool because it spares my time from assigning tasks manually, but ultimately my job is to manage people, who have emotion.” Cai, a district manager, told me that when hiring supervisors, he considers people with prior experience as couriers to be desirable candidates who can work with SPCs to ensure excellent KPIs.

Station supervisors and couriers have regular face-to-face interactions. The daily interaction begins with a morning meeting. Yang, an Ele.me station supervisor, described the routines:

We begin our morning meetings from 10 a.m. Couriers sanitize their delivery boxes. Then we go over statistics. Our company has three service stations in this district. We announce how many orders, good reviews, and bad reviews each station received yesterday. I teach couriers how to behave in a civilized way. We practice how to apologize to customers. I have to do ideology work to better educate couriers, so they would have excellent KPIs. Many customers are crazy, so I often have to lift couriers up.

Note that “ideology work” is the term used by the Chinese Party-state’s propaganda system to indoctrinate people. This quote shows station supervisors’ effort to influence how couriers behave, deal with emotion, and generate excellent performance metrics through normative control and emotional labor (Hochschild 1983).

Architecture of the gig platform

The technical, legal, and organizational design of the gig platform aims to establish a flexible and disposable workforce. According to Liu, a lawyer at Meituan, gig platforms achieve this goal by surrendering most of the supervisory power of management personnel over couriers: “If management personnel are visibly involved in supervising couriers, the court will consider work relations as employment relations. That means platform companies do not have flexibility to adjust workforces. Also, platforms will have to take more legal responsibility.”

Consequently, gig platforms automate most management and supervision tasks. Similar to technological control and management under the service platform, Meituan and Ele.me rely on technology to discipline and dispatch GPCs. Disciplinary rules are built into and implemented by the system, such as fines and deactivation of an app account. Dispatch algorithms incorporate rules that decide which couriers are given priority in choosing assignments. Wong, a system development engineer at Meituan, explained the company’s algorithm:

When a consumer makes an order, the system calculates the distances between the consumer and available gig couriers and the distances between the restaurant and available gig couriers. The algorithm also considers couriers’ performance statistics. The system gives priority to gig couriers with better statistics.

The two gig platform systems classify couriers into different levels. In addition, the platforms use credit points to measure a courier’s credibility—how reliable the courier is—based on their behavior, such as couriers’ records of accepting or rejecting dispatch assignments. Similar to the service platform apps, the gig platform apps integrate elements and dynamics of “gamification” to induce couriers’ participation in the “game” (Deterding 2019).

The two gig platforms also use algorithms to set remuneration. Under the service platform, piece rates are generally fixed and stated in a contract by Meituan and Ele.me’s franchisees. In comparison, Meituan and Ele.me’s gig platforms use algorithms that factor in several factors (e.g., distance, condition of the destination, weather) to calculate piece rates. The pricing model is similar to Uber Eats’ model (Veen et al. 2019). Local managers who work for gig platforms can change some parameters in the system that influence remuneration, for instance, the basic rate for a kilometer.

Both Meituan and Ele.me use technology to confirm the identity of gig couriers and monitor couriers’ compliance with rules. Because gig couriers do not have a human supervisor, both platforms require couriers to scan their face before they can take an order and to take a picture of themselves during their work. As Xue, a GPC, described, “Ele.me often ask me to upload a selfie when I’m on the road. They want to see whether I wear the company’s helmets and vests and carry the company’s delivery box.” Failure to comply with such rules has disciplinary outcomes.

Meituan and Ele.me meticulously design their contracts to maximize the gig platforms’ flexibility. Liu, a lawyer at Meituan, explained the legal work the two platforms have done:

The legal relationship is complicated as there are many parties involved: platforms, couriers, restaurants, consumers, and people on the street. Our task is to minimize the company’s liabilities and maximize flexibility. We avoid having an employment relationship with gig couriers. We don’t want to be responsible for couriers’ traffic negligence. We’ve left some space for the company to adjust contractual terms, policies, and, importantly, technologies.

Indeed, Meituan’s clickwrap agreement stipulates that China’s Labor Contract Law does not apply to the contractual relationship between Meituan or its subcontractors and GPCs. This legal classification design is similar to counterparts in gig economies outside China (Vallas 2017; Veen et al. 2019). Meituan’s clickwrap agreement also stipulates that GPCs must accept every clause in the agreement and any rules published on Meituan’s platform, including disciplinary measures. Meituan and Ele.me are similar to conventional companies that employ contingent workers and attempt to avoid liabilities resulting from employment (Kalleberg 2009). But unlike conventional companies, Meituan and Ele.me endow themselves with the power to change contractual terms unilaterally through a non-negotiated clickwrap agreement (Stewart and Stanford 2017). Asked whether he considered Meituan and Ele.me’s contracts asymmetric, Liu—likely influenced by his awareness of my legal training—replied frankly:

The power of platforms and that of couriers are enormously asymmetric. We only have two major platforms in China. Tencent is the largest stockholder of Meituan, while Alibaba independently owns Ele.me. The two platforms have a lot of capital, political connections, and lawyers. What do couriers have?

The organizational design of gig platform architecture minimizes the role of management personnel in managing and supervising couriers. Although Meituan’s subcontractors and Ele.me set up offices in cities where the two companies’ gig platforms operate, Peng, a manager who works for Meituan’s gig platform, explained that his office’s main focus is monitoring and managing data and the platform system itself, not overseeing couriers. When I visited Peng’s office, I noticed that most of the employees were either analyzing data or identifying irregular patterns in front of their computer screens. Similar to Wong, an engineer whom I interviewed, Manager Peng saw the operation of the delivery platform as a mathematical problem of “multiple goal optimization,” that is, “how to maximize revenues and customer satisfaction through dynamically optimizing the matching of orders and couriers with the best metrics.” Couriers’ experience was not something either Peng or Wong sought to optimize.

The number of gig platform offices in a city is much smaller than the number of service stations, as the former do not have regular interactions with couriers as service stations do. To activate an app account, both Meituan and Ele.me require couriers to complete online training. They further require offline training, which is often outsourced, as a prerequisite to receive more than a certain amount of delivery tasks or to reactivate an app account that was blocked when GPCs violated platform rules. I attended one such offline training, where an instructor talked about the platform company’s rules, including fines and other forms of punishment, and ended with an exam that tested whether couriers could memorize the rules. Table 2 summarizes the platform architectures of the service and gig platforms.

OPEN IN VIEWER

Table 2.

Platform Architecture

	The Service Platform	The Gig Platform
Guiding Principle	The purpose is to create a reliable workforce. The design requires a higher level of supervisory power over couriers, at the expense of having legal responsibility based on employment relations.	The purpose is to create a flexible workforce. The design avoids legal responsibility based on employment relations, at the expense of having less supervisory power over couriers.
Technological Dimension	Supplementing the task of management personnel – Using algorithms that consider performance metrics to dispatch couriers – Using algorithms to monitor and discipline couriers – Incorporating “gamification”	Automating management and supervision – Using algorithms that consider performance metrics to dispatch couriers – Using algorithms to monitor and discipline couriers – Incorporating “gamification” – Using algorithms to set and adjust remuneration – Using technologies to confirm couriers’ identity and monitor whether couriers wear the company’s helmets and vests
Legal Dimension	Establishing legal relations based on a conventional employment contract – Based on a labor contract that classifies couriers as employees – With a stable salary structure – Regulated by labor law	Avoiding legal relations based on a conventional employment contract and maximizing flexibility – Based on clickwrap agreements, which are asymmetric contracts – Not classifying couriers as employees – Falling outside of labor law – Endowing the platform with the power to change contractual terms and platform rules unilaterally
Organizational Dimension	Management personnel in charge of managing and supervising couriers – Emphasizing management of couriers – Human supervisors can override certain algorithmic decisions – Calculating the KPIs of supervisors and service stations from SPCs’ KPIs – Imposing normative influences and surveillance on and performing emotional labor vis-a-vis couriers	Management personnel generally not in charge of managing and supervising couriers – Emphasizing management and monitoring of data and the platform system – Testing couriers on whether they can memorize important platform and disciplinary rules in online and limited offline training sessions

Technological control and management

SPCs did complain about their work and named their injurious experiences, especially KPIs and disciplinary measures built into the platform system. Several SPCs expressed that one complaint can cost 150 RMB (around US$22.3). On average, SPCs made 5 RMB (around US$.74) for each delivery, and the most productive couriers completed 50 to 60 delivery tasks each day. Thus, a fine could be a considerable proportion of an SPC’s daily earnings. Many SPCs complained about the stress resulting from disciplinary measures built into the app. They talked about speeding in chaotic traffic and severe weather conditions to deliver food on time. Similar to platform workers in the United States who are stressed by performing emotional labor (Raval and Dourish 2016; Rosenblat and Stark 2016), SPCs described feeling immense pressure to get excellent reviews.

SPCs also complained about dispatch algorithms. Chen, a 26-year-old SPC, explained why he thought Meituan’s algorithm was unfair:

It’s clear that I am closer to the restaurant, but the system does not assign the order to me. Instead, the system assigns the order to a courier with higher KPIs. Top five couriers keep getting orders, but others do not get many orders. This is unfair.

Wang, a 27-year-old SPC, had similar complaints: “The algorithm is stupid. When I am delivering an order, it always assigns me another within short delivery time. That decreases my on-time rate.” Dissatisfaction with algorithmic control and management was prevalent among interviewees.

Legal control and management

Despite SPCs’ complaints about technological control and management, I found their dissatisfaction seldom evolved into a grievance or a sense of injustice, due, in part, to their understanding of the legal relationship. SPCs believed their work conditions were expected based on their labor contracts. Several SPCs mentioned they knew the salary structure and disciplinary rules from the very beginning, and the rates and rules remained the same. As Dong put it: “You know how much you are going to get and what the service station expects you to do. If you are unhappy, don’t take the job. You can leave the job anytime.” Similarly, another SPC remarked: “The disciplinary rules are harsh, but the service station needs to ensure the quality of service. And we know that from the contract and rules on the wall.” These quotes show the legal design of the platform architecture—specifically, the concrete and stable terms in a labor contract—helps stabilize labor relations.

In addition, SPCs distinguished dissatisfying from unacceptable aspects of work. They saw salary as the most important component of their contracts. As long as their employers paid their salaries, they did not have serious issues. Wang’s view was typical among SPCs:

I’m dissatisfied with the dispatching algorithm, but it’s not the most important issue. As long as my boss pays me salary, I’m fine. I got my contract down in black and white. If my employer does not pay me, I’ll have evidence to prove his violation of the contract and labor law. I can talk to supervisors. If they don’t respond, I will go to the court.

Asked whether going to the labor bureau or the court was difficult, Wang said, “Couriers on WeChat say there are free templates available. We only need to fill out a form to file a complaint. You don’t need a lawyer. Some couriers got money back successfully.” When I asked Wang whether he would consider a strike or a protest if he did not receive his salary, he said he would not do so unless none of the other methods worked.

Wang’s answer, which was common among SPCs, shows the relationship between injustice frames and the appeal of collective contention. Because SPCs’ injustice frames were based on a highly established legal claim regarding violation of the labor contract and labor law, they believed they could solve problems through individual informal legal mobilization and institutionalized dispute resolution channels. They considered collective contention only when other branches of the “dispute tree” did not work (Albiston et al. 2014).

SPCs’ focus on one contractual element—salary—was also reflected in how they thought about social insurance. Even though service stations did not entirely follow the Labor Contract Law, very few SPCs were angry about this. Due to the often weak enforcement of labor law, SPCs had low expectations regarding social protection (cf. Vallas 2017). In fact, only one SPC said his employer bought social insurance for him. However, the SPCs and supervisors I interviewed said that because the turnover rate of SPCs is high, it is difficult for service stations to provide SPCs with social insurance. My interviewees told me that not even station supervisors receive social insurance. None of the SPCs I interviewed considered social insurance a critical component of their labor relations. Fang, a 26-year-old SPC, said:

Almost no employers in the service sector would buy social insurance for employees. Also, we don’t think social insurance useful, so we don’t ask the station to buy it for us. I bought social insurance when I worked in an automobile factory. I don’t feel I’m in a better situation with social insurance because of my household registration status.

SPCs cared mainly about wages. In most cases, it was only when a grievance about salary emerged that other issues, such as social insurance and disciplinary measures, were incorporated into the discussion.

Organizational control and management

SPCs told me they communicate with their supervisors when problems arise, and this makes them feel more informed about the situation even when their complaints are not solved. Many couriers understood that supervisors themselves are constrained by the platform system and Meituan or Ele.me. For example, although Chen and Wang complained about dispatch algorithms, they told me it would be unreasonable to blame their supervisors because their supervisors are obligated to use the algorithms. Chen explained:

I told my supervisor the platform system assigns orders unfairly. I received fewer orders than people with good KPIs. He told me he couldn’t change the system, but he manually assigned me some orders from time to time after I talked to him. That made me feel better.

This quote also shows that even when a complaint occurs—in Chen’s case, from dissatisfaction with algorithmic control—supervisors’ intervention could prevent problems from worsening.

Wu, a 30-year-old SPC, told me about a dispute in his station that could potentially have led to a strike. When Wu and his fellow couriers did not receive their salaries on time, they discussed the issue on WeChat and talked to their supervisors, who knew SPCs could initiate a strike. Under pressure, Wu’s supervisor asked the district manager to explain the situation to the couriers. The manager said his company had problems with Ele.me and would distribute salaries once they received payment from Ele.me. In the end, the SPCs received their salaries. Wu remarked: “Because the management solved our problems, we didn’t take further action like strikes.” This instance shows effective management can reduce the appeal of collective action.

Most SPCs told me they had no serious problems with their supervisors, although two complained about supervisors’ unfair manual assignment of tasks. I wondered why the supervisory relationship did not seem to be tension-ridden. Chen explained that his supervisor’s KPIs and the station’s KPIs are the aggregation of SPCs’ KPIs. He and his colleagues can retaliate against their supervisor by lowering their KPIs and then leaving their job if their supervisor is mean to them. Chen also said it is easy for SPCs to transfer their work to another service station if they have conflicts with supervisors.

In interviews with management, I also found organizational constraints on supervisors’ behavior. Huang, a station supervisor, said: “It’s better to educate and communicate with couriers nicely. We are in the service industry, not manufacturing. I need their smile, politeness, and good attitudes to make a living.” Similarly, Cai, a district manager, told me the key is “managing the hearts and minds of both customers and couriers” to get good KPIs. Cai pointed out that according to the franchise contract, the platform could terminate his company’s right to operate the delivery business if a strike or protest occurs; franchisees also have obligations to oversee SPCs’ behavior in social media. Essentially, the organizational constraints through KPIs and franchise contracts motivate supervisors to perform emotional labor and exercise normative influence—and surveillance—on SPCs.

Indeed, supervisor Huang said it is not difficult for management to discern what is going on among SPCs as he sees SPCs every day and is in their social media groups. “You know who is doing what. . . . I do extra ideological work on unruly couriers.” One aspect of ideological work, according to Huang, is telling SPCs that protests or strikes are illegal. Huang’s quote corresponds with how Wu, an SPC, described his supervisor’s surveillance: “He knew what we were discussing. It’s quite difficult for us to discuss things. Also, couriers with a closer relationship with the supervisor might betray us.” Huang’s and Wu’s quotes reveal that couriers’ workspace is not free but monitored by management—an established challenge to collective action (Pollert and Charlwood 2009).

Collective contention

Despite organizational control and management, collective contention still arises on rare occasions. Three SPCs who reported participating in collective action provided similar accounts of why collective action happened. Ming described one situation:

We were supposed to receive our salaries. Our supervisors didn’t know why the company [franchisee] hadn’t paid us, but he agreed the company violated its obligation. Since some couriers said the company would go bankrupt soon, we decided to take action [a strike]. We didn’t have time for the court.

Ming’s quote reveals that the injustice frame for SPCs was violation of contractual terms. After the strike, Ming’s company told SPCs the company had financial problems. SPCs eventually received around 80 percent of their salaries.

I also interviewed station supervisors and a district manager about collective action. Yang, a service-station supervisor, stated:

Couriers rarely initiate a strike or protest as supervisors already solve most of the problems. Strikes occur only when franchisees have problems with Meituan or Ele.me or when there is an internal dispute among investors. These problems create financial difficulty for franchisees and cannot be solved by supervisors. Under such circumstances, franchisees try to pay less. Couriers and franchisees have more or less similar understandings about what is right and what is wrong.

Yang’s quote corresponds to something district manager Cai told me, that it would be difficult for franchisees to deny their obligation to pay salaries. The focus of negotiation from the perspective of franchisees, Cai explained, is to “let couriers know the company’s difficulty and accept what the company can afford.” Together, my interviews with SPCs and management suggest that, even when a collective action arises, couriers and employers do not have fundamentally different views on contractual terms and law. Rather than challenging the legitimacy of contracts, SPCs were demanding that employers honor the contracts.

Within the service platform, technological control and management can cause dissatisfaction among SPCs, but the legal and organizational design of the platform architecture often serves to contain grievances and make collective action less appealing. Collective contention seems less appealing to SPCs because their injustice frames (i.e., violation of contracts and labor law) are based on well-established legal claims and they can communicate with management personnel. Finally, the platform architecture’s organizational design—particularly the constant presence of supervisors—also restricts the amount of free space for organizing collective contention.

Technological control and management

Similar to SPCs, GPCs were critical of dispatch algorithms and disciplinary rules built into the platform system, but they also had problems with the calculation of remuneration, decreasing remuneration, and changing dispatch and disciplinary rules. Several GPCs accused the gig platforms of “stealing money” from them. Qiang, a 29-year-old GPC, explained: “Meituan is scamming us. They know the real delivery distance but use the linear distance between two points to calculate remuneration. We are not flying an airplane!”

Like gig workers in Europe (Tassinari and Maccarrone 2017, 2020), GPCs overwhelmingly criticized how platforms have decreased remuneration. They believed platforms change algorithms to reduce remuneration, although the two platform companies have never openly acknowledged that they lower the piece rate. Similar to Rosenblat and Stark’s (2016) study of Uber in the United States, gig platforms and couriers in China have asymmetric information about the platform system. The actual practice of algorithmic pricing remains a “black box,” hidden from workers’ view, so GPCs complain about the decreasing rate based on their observations. Although gig platforms use algorithms that consider many factors to calculate piece rates, couriers said it is evident that piece rates under similar circumstances have decreased over time.

GPCs often take screenshots of their tasks and share them online, so they believe they have evidence of the decreasing piece rates. For example, a GPC said that in the past, he received 5 RMB (around US$.74) for a delivery task within 1.5 km, but the rate went down to 3.5 RMB. There was consensus among my interviewees that Meituan and Ele.me had changed their algorithms and lowered the piece rate stealthily, hoping couriers would not notice. GPCs complained that decreasing piece rates had a significant effect on their lives. According to Shi, a 22-year-old GPC: “I was able to make 9,000 to 10,000 yuan [around US$1,278 to 1,420] per month. Now I make around 6,000 yuan [around US$852] a month.” Many GPCs reported having to work two more hours every day to maintain the same level of income.

GPC interviewees emphasized that along with decreasing piece rates, the platforms have also been changing the dispatch and disciplinary rules within the system to restrict GPCs’ freedom to decline a dispatch assignment. This is similar to Rosenblat and Stark’s (2016) study of how Uber constrains drivers’ freedom to reject a request. When a GPC declines a dispatch assignment, Meituan’s platform system deducts the courier’s credit points. Couriers with low credit points are less likely to be assigned orders with higher remuneration. In some cities, Meituan has even made new dispatch rules that require offline training and restrict a courier’s eligibility to take an order after they decline a certain number of dispatch assignments.

In online communities, I saw a lot of criticism: “Given that couriers do not have freedom, Meituan should stop using freedom in its advertisement to recruit gig couriers”; “The platforms promise to give us freedom, but now they treat us like their slaves. We are forced to take orders with low pay.” These comments critiqued the contradiction between gig platforms’ promise of freedom at work and how dispatch rules and algorithms negate such freedom. When GPCs realized gig platforms increase technological control to reduce the promised freedom, they saw the platform-courier relationship as a master-slave relationship and felt betrayed.

Legal control and management

Certain updates of dispatch and disciplinary rules in the gig platform system require corresponding legal arrangements, thus leading to the overlap of technological and legal control and management. Unlike changes in pricing practice, changes to dispatch and disciplinary rules are often transparent. Before implementing new dispatch and disciplinary rules in the platform system, local offices notify GPCs of the new rules and justify the changes by stating their purpose is to “better regulate delivery behavior.”

Local offices further affirm that they have the “ultimate legal rights” to change platform rules according to contracts between the platform and couriers by pointing to the clickwrap agreement that couriers accept when installing the delivery app. However, none of my gig platform interviewees reported reading the legal texts when they installed the app and accepted the clickwrap agreement, and none knew that platforms had the right to change the content of the agreement and platform rules unilaterally. They all expressed anger about such clauses and other rules enacted by platforms unilaterally. GPCs called such contractual terms and platform rules “despotic clauses” (bawang tiaokuan) and said they were evidence of exploitation.

I was curious as to why the companies were so secretive when it came to changing pricing practices, but open about changes in dispatching and discipline rules. Manager Peng explained the companies have to announce the latter to shape GPCs’ behavior: “We use the carrot and stick approach. If the new rules remain unknown to couriers, couriers will not change their behavior even after we incorporate the rules into the platform system.” He emphasized the need to adjust platform rules dynamically:

The purpose is to enhance the efficiency of dispatch assignments and customers’ experience. Under the service platform, couriers must accept every order assigned by his station. Under the gig platform, couriers don’t have such obligation, but the company wants every order to be taken by an excellent courier. Achieving this goal needs continuing improvement of dispatch and disciplinary rules. Now our platform system is powered by artificial intelligence and machine learning. It learns from the big data it collects and gives us feedback. These tools help us improve the system’s efficiency by changing the rules dynamically.

Manager Peng’s remark reminded me of my observations of offline training sections, in which GPCs were required to memorize platform rules and tested on the content of the rules. Gig platforms, I realized, seek to turn couriers into a part of the platform system that automatically implements the rules. When companies change the rules, new rules have to be installed not only in the platform system but also in the minds of couriers themselves. When this can be done successfully, human workers effectively become part of the platform technology.

Yet unlike SPCs, GPCs do not internalize the rules through normative influence or emotional labor. Gig platforms just announce the changes in platform rules and corresponding disciplinary measures, assert their “ultimate legal rights,” and test GPCs on the rules. In these moments, technological and legal control overlaps, and the platforms’ power becomes salient. Deng described such a moment:

When I saw the platform’s notice about the new dispatch and disciplinary rules in the system [the platform system] and the company’s legal rights, I felt furious. Then I understand why couriers complained about despotic rules on WeChat. . . . Couriers were saying everything—from platform rules to legal clauses—is determined by the platform but not us.

Deng’s quote shows that the revelation of platforms’ unbridled legal and technological power turns moments of change into moments of escalation (Atzeni 2009; Fantasia 1988). Although previous literature has focused on the legal classification of workers as employees or individual contractors (Vallas 2017; Veen et al. 2019), GPCs are most concerned about platform companies’ unilateral and unconstrained legal and technological power. In the interviews, many GPCs emphasized that the platforms’ exploitation is blatant. For example, one GPC remarked:

Now the platforms implement all kinds of despotic clauses. If you don’t follow their rules or terms, they deduct money from you. They force you to accept their dispatch assignments with low pay. If you don’t take their order, they deactivate your account. It is unfair. Platforms blatantly exploit couriers.

In general, gig couriers believed platforms used unfair legal arrangements, making the inputs and return of labor totally disproportional. They understood that the two platforms may not technically violate any existing law, but they felt there were unfair loopholes in the legal system. Here, we can see how the legal dimension of the gig platform architecture generates grievances and heightens perceived injustice.

Asked whether they can address their grievances through the labor bureau or the court, most GPCs said no because they simply do not know which law platform companies are violating. Importantly, they understood that their legal situation differed from conventional employment relations, but they could not use legal terms to describe their status. As Zeng said,

We don’t have a contract. I mean a labor contract. I don’t know what kind of status we have. I only know we’re not their employees. I don’t think we can go to the court or the labor bureau. That’s for people with a labor contract. We don’t know what kind of law the platforms have violated. We just feel they are wrong and things are unfair.

In social media group discussions, most GPCs shared the same legal understanding as Zeng. As one GPC shared with a social media group composed of other GPCs, when he went to a labor bureau for legal advice, he was told the bureau could not help because it could only assist people with a labor contract. Consequently, GPCs see a disjuncture between their injustice frames that challenge the legitimacy of contracts and claims that can be accepted by dispute resolution institutions. Put another way, the institutionalized resolution “branches” in the “dispute trees” do not seem viable to GPCs (Albiston et al. 2014). As such, the legal dimension of the gig platform architecture makes collective action more appealing than institutionalized resolution channels.

Organizational control and management

The organizational design of the gig platform further intensifies grievances and increases the appeal of collective contention. As the gig platform largely automates control and management, GPCs do not develop social relationships with management personnel. Instead, they interact with the “cold” technological interface. GPCs use the apps to interact with platform companies. Similar to Uber drivers in the United States (Rosenblat and Stark 2016), they encounter difficulties trying to speak to platform representatives. As Shi told me:

I tried many times, but they’d never responded to me. Last time when they changed platform rules, I contacted them to complain. I was not the only one. In the end, couriers became angrier and angrier. We had to do something else to get our voices heard.

Indeed, GPCs shared similar frustrations about not being able to solve their problems through apps or local offices. Frustration and anger escalated grievances. In most of the collective actions I analyzed, GPCs contacted platform companies but did not get any meaningful responses; they soon realized this approach would not address their grievances.

The disengagement of platform companies also means GPCs have relatively free spaces to mobilize solidarity and organize collective action, mostly online but also offline (Polletta and Kretschmer 2013; Tassinari and Maccarrone 2020). GPCs expressed their anger and sense of injustice in social media groups. I saw GPCs rename their social media groups as a “labor union” even though they could not form formal labor unions according to Chinese law. Many couriers discussed the importance of solidarity with online statements like: “We must have solidarity to defend our interests, brothers. Otherwise, we’ll continue to be exploited by the platform.” Some couriers began to organize collective action offline. To be sure, the platforms try to undermine online organizing. In some cases, after local offices found GPCs were organizing a protest via social media groups, they deactivated organizers’ app accounts and asked social media companies to block those groups. Nonetheless, GPCs were still able to form new groups.

Collective contention

Collective action occurred mostly when GPCs believed gig platforms had drastically decreased remuneration by changing algorithms or when platform offices notified GPCs of new platform rules and simply cited the platform’s contractual right to do so. Usually, a core group of GPCs—those with existing interpersonal relationships—initiated a strike or protest. They did so in the absence of assistance from grassroots labor NGOs or government-organized trade unions. Consistent with findings in previous sociolegal studies (Levitsky 2008), experience and knowledge mattered. GPCs with experience working in the manufacturing sector played a crucial role in organizing such collective actions. Core organizers tended to be those who completed a huge amount of orders every day. Decreasing rates and changing rules cost such workers hundreds of RMB daily. Core organizers’ attempts to improve their work situation was understandable, given that so many did not want to return to the manufacturing sector and did not have a feasible alternative career plan.

Here, I use a series of strikes and protests in Chongqing in 2018 to illustrate a typical mobilization process in “free spaces” online and offline. The core organizers belonged to a close-knit group. Tang, a 23-year-old GPC, recounted how his group organized a strike after Meituan announced a change in dispatch and disciplinary rules:

We only had four people in the beginning. We became friends through a WeChat group. Four of us hang out after work twice to three times a week. When we saw couriers got angry in social media groups, we discussed in a BBQ place and decided to organize a strike. . . . I saw how people in my factory went on strikes. They got what they wanted. [Me: Why didn’t you talk to Meituan or go to the court or the labor bureau?] A lot of couriers already expressed their discontent to Meituan, but Meituan didn’t respond. No one knows how to file a legal complaint under such circumstance. What we write would not be accepted by the court.

Tang’s quote is consistent with my previous analysis about how the legal and organizational dimensions of control influence GPCs’ perception of different “branches” in the “dispute trees” (Albiston et al. 2014).

Tang and his friends made banners and signs featuring messages such as “despotic clauses,” “resist exploitation,” and “stop taking orders.” They chose these terms because other GPCs had used them in WeChat groups, where Tang and his friends recruited participants. They then visited several restaurants, because restaurant owners also suffer from asymmetric contractual relations with Meituan, especially increasing commission fees. Three restaurant owners lent their support, not only by hanging banners saying “resist monopoly, resist Meituan” and “resist despotic clauses,” but also by providing GPCs with free meals to show solidarity. The core group sent out a message warning couriers that if non-striking gig couriers were seen delivering food during a strike, their tires would be slashed. One core member organized a 10-person patrol team to this end.

After the strike and protest began in a plaza, GPCs uploaded screenshots of their delivery app to social media groups to further mobilize participation. They proclaimed their own “offline” status with comments like: “Hundreds of brother couriers are offline. Don’t be a bad apple. Our patrolling team is sanctioning bad apples on the street” and “We’ve determined not to be exploited by evil capitalists. Long live the working people!” GPCs often join multiple social media groups, so information spread quickly online and among GPCs who worked for different companies. Some non-participants encountered violence for disregarding warnings. Faced with potential sanctions, several GPCs decided not to work and instead went to protest sites as bystanders. Tang was surprised by what happened:

Initially, we had a strike in our district and drew about 100 couriers, but it soon spread to five other districts and lasted for a week. In each of the subsequent strikes, I saw 50 to 80 couriers. We initiated a strike against Meituan, but the strikes also attracted Ele.me couriers.

Most participants were full-time GPCs, who were more dependent on platforms. One part-time GPC told me, “[the protest] isn’t my business because I have a real job.” He also said he was not as angry as full-time couriers as his stake was low. This finding resonates with literature that shows how economic dependency influences work experiences (Schor and Attwood-Charles 2017; Schor et al. 2018).

Eventually, GPCs sieged one of Meituan’s local offices in Chongqing for hours. The core members told Meituan’s local office specific demands concerning platform rules and remuneration. Zhang, a 20-year-old GPC with a high school education, recalled that moment:

Finally, we saw those people [the management] who’ve never responded to us. Four to five people came out from the office. They are all in their 20s, just like us. You can tell they are college graduates. The manager is a woman. They were so scared and cannot say anything to couriers. Although they did not tell us, we found they increased the rate. They also sent out a notice to withdraw the new platform rules. . . . After the protest, Meituan did not decrease the rate for eight months, but they deactivated a few participants’ accounts.

Zhang’s quote also reveals class and gender boundaries between management and couriers and the awkward absence of interaction when GPCs eventually encountered management.

Their singular focus on optimization of the platform system made gig platforms’ local offices ill-equipped to address grievances and resistance. This tendency was evidenced by Meituan’s attempt, in September 2019, to install more restrictive dispatching rules in Chongqing, despite the strikes in 2018. When many GPCs threatened to initiate another strike, Meituan finally gave up trying to install the new rules. GPCs were happy to see Meituan change its plan under such pressure, but they also recognized their inability to oversee the platform and actualize such threats regularly.

I was interested in why GPCs joined a collective action, particularly considering potential police suppression, rather than simply finding another job. Several GPCs told me they joined the collective action because the platform companies are so despotic and their rules are so unfair. Anger and a strong sense of unfairness, rather than an instrumental calculation about the potential pros and cons of collective action, prompted some GPCs who were not core organizers to join a collective action. Fang told me how such sentiment drew him into a protest: “I went there because the relationship is so unfair. I didn’t know if we can change the situation, but I had to go because I was so angry.” In fact, several GPCs were puzzled when I asked why they protested instead of finding another job. They said they might find a job sooner or later because they did not plan to be a GPC for a long time, but they still wanted to teach platform companies a “lesson.” My finding thus resonates with research that shows collective action frames can go beyond rational calculations and are often based on moral imperatives (Morgan and Pulignano 2020).

A few couriers told me they were just bystanders. One GPC remarked: “When other couriers and I knew some gig couriers went on a strike in a plaza from social media groups, many of us immediately dashed to the site. We felt curious and wanted to see what happened.” Interviewees who described themselves as bystanders shared the same sentiment as other GPCs, but they emphasized their need to make money to sustain their lives. Nonetheless, the presence of bystanders did escalate collective action. Not surprisingly, GPCs who decided not to join a collective action talked about needing to make money and thought such collective action would not work.

Asked why they were not afraid of police suppression, many GPCs said they were just bystanders or pretended to be bystanders; they did not think the police would arrest them as there were a lot of people there. Participants who joined the protests in Chongqing in 2018 said the police took away only a few participants and told them organizing a strike or protest was illegal. Most interviewees did not seem intimidated by the police.

Under the gig platform, technological control and management leads to grievances and perceived injustice. This dimension of control and management overlaps at times with, and is reinforced by, legal and organizational control and management, generating moments of escalation. GPCs’ injustice frames (i.e., the illegitimacy of contracts) and the organizational design of the platform architecture make institutionalized dispute resolutions and individual informal legal mobilization seem implausible to GPCs. The platform companies’ focus on managing the platform system rather than couriers further renders free spaces available for GPCs to turn moments of escalation into moments of collective action and solidarity.