Workers and Their Foes: Customer Scapegoats in the Service Triad

Abstract

The author reconceptualizes the service triangle as a Simmelian triad and draws on 16 months of fieldwork to analyze restaurant servers’ experiences of income and interactional precarity. Precarious work frustrates servers. However, triadic dynamics position customer scapegoats as responsible for servers’ frustrations. Employers passively benefit from this dynamic, which continually minimizes opposition to the structure of employment. This is a new mechanism for the reproduction of precarious service work that can explain why service workers consent to exploitation. These findings apply to workplaces in which workers experience low levels of autonomy and encounter customers who are interactionally salient and can control workers’ behavior. In such workplaces, the triadic dynamics identified in this article could weaken the potential for worker-customer alliances and complicate the promise of anti-tipping campaigns.

Sociologists have long been drawn to the question of why workers do not resist unjust and exploitative conditions (e.g., Braverman 1974; Du Bois [1935] 1975; Marx [1867] 2011). When it comes to the contemporary context of precarious service work (Kalleberg and Vallas 2018; Schneider and Harknett 2021), sociologists argue that part of the answer lies in the shift from the dyadic employer-worker interactions of manufacturing to the triadic employer-worker-customer interactions of interactive service work (Leidner 1993). Drawing on Burawoy (1979), these sociologists (e.g., Sallaz 2002; Sherman 2007) show how workers’ sense of autonomy in interactions with customers keeps workers invested in their work. However, as documented in the voluminous literature on low-status service work, workers in the service industry do not always feel a sense of autonomy (Glavin, Bierman, and Schieman 2021; Van Oort 2019; Vargas 2021). Thus, we need an additional mechanism to explain why service workers agree to their exploitation in those workplaces in which a sense of autonomy does not predominate. When workers do not feel like they control their own labor, what interaction-level dynamics minimize service workers’ opposition to increasingly “bad” jobs (Kalleberg 2011)?

In this article, I reconceptualize Leidner’s (1993) service triangle (the relationship among workers, employers, and customers) as a Simmelian triad (Simmel 1950), a generic pattern or “social form” (Vaughan 2014; Zerubavel 2020), to explain how workers’ opposition to employer control is minimized in service settings where workers do not feel autonomous. Using data from 16 months of participant observation with servers at a restaurant I call “Juniper,” I show how precarious work frustrates servers, even as the restaurant’s triadic dynamics channel this frustration toward customers and away from employers. Customers’ interactional control leads servers to blame customer scapegoats—not employers—for their problems. The service encounter thus continually minimizes opposition to employers, who profit from a precarious work system free of the threat of employee complaints. These findings may apply to other workplaces in which workers experience low levels of autonomy and encounter customers who are interactionally salient and can control workers’ behavior. In such workplaces, these dynamics could weaken the potential for worker-customer alliances and complicate the promise of antitipping campaigns.

Consent, Control, and the Customer

Dominant explanations for workers’ tolerance of unfair treatment by employers draw on Burawoy’s (1979) classic study of factory labor and rest on workers’ experiences of autonomy. Burawoy asks why the workers he observed consented to their own exploitation, and he finds the answer in the organization of the labor process. Workers strove to win the game of “making out,” playing the piece rate system to their own advantage by varying output. These efforts gave them a sense of autonomy: they felt like strategic actors in control of their own labor. This sense of autonomy helped their employer secure their hard work and obscured the structural relations by which the employer (not workers themselves) controlled workers’ labor.

Researchers taking Burawoy’s theory from the shop floor to the service counter explore customers’ implication in the generation of worker consent (e.g., Sallaz 2002; Sherman 2007). The service industry involves the production of interactions and emotions more than the production of material goods. Therefore, management must allow service workers some autonomy in their interactions with customers so that they can perform realistic emotional labor (managing one’s emotional displays for pay) tailored to customers’ desires (Hochschild 1983; Leidner 1993; Sallaz 2002; Sherman 2007). As workers perform customer-focused labor, they feel a sense of strategic autonomy, becoming, for example, clever entrepreneurs manipulating customers to increase their tips (Sallaz 2002; Sherman 2007). Again, chief to the production of consent is workers’ sense that they control their own labor. Like the factory workers Burawoy studied, these workers are not coerced into exploitation; they agree to it.

Burawoy’s theory of worker consent, as translated to the service industry by Sallaz (2002) and Sherman (2007), continues to influence exciting research on service work (e.g., Cameron 2021; Halpin 2015; Ranganathan and Benson 2020). Still, some researchers point to the limits of this perspective. Mears (2015), for example, argues that researchers who center workplace relations in the construction of consent miss how work is shaped by relationships outside of the workplace; meanwhile, Lopez (2010) and Subramanian and Suquet (2018) argue that focusing on relations between workers, managers, and customers comes at the expense of exploring relations within each group; finally, Otis (2016:158–59) argues that researchers focused on how employers recruit workers’ “hearts and minds” into participating in the labor process have neglected to explore “the role of the body and bodily aesthetics” in generating worked consent. However, the literature has done less to explore an additional limitation to the consent-autonomy model: not all service jobs allow workers to feel autonomous (Glavin et al. 2021; Van Oort 2019; Vargas 2021). If a sense of control is a hallmark of being a blackjack dealer (Sallaz 2002) or a luxury hotel concierge (Sherman 2007), an acute sense of being controlled by others marks other service industry jobs (e.g., Reich and Bearman 2018; Rosenblat and Stark 2016; Vargas 2017). Customer-facing workers in the service industry may be just as likely to feel subject to customer demands as they are to feel a sense of autonomy.

Three streams of research on customer-facing workers in the service industry highlight times when workers feel subject to others’ demands, those “others” often being customers. First, research on precarious work—work that is “uncertain, unstable and insecure” (Kalleberg and Vallas 2018:1)— demonstrates how, as employers shift risks to employees, workers exercise less control over their labor (Kalleberg 2018; Lambert 2008). This objective lack of control can translate into a felt lack of autonomy. For example, when employers minimize labor costs with short-staffing, workers may feel stressed and outmatched by seemingly endless customer demands (Ikeler 2016; Misra and Walters 2022; Reich and Bearman 2018; Vargas 2017). Second, research on workplace emotions emphasizes how workers cannot fully control their interactions with customers, as their emotional labor must conform to employers’ “emotion rules” (Hochschild 1983). Although workers sometimes engage in emotional labor strategically (Bolton and Boyd 2003), there is little dispute that they are not always able to do so (e.g., Ikeler 2016; Misra and Walters 2022). Third, research on work control demonstrates how employers seek to control workers indirectly through customers with tipping and commission systems, mystery shopper reports, customer rating systems, and customer abuse (Fuller and Smith 1991; Gandini 2019; Korczynski and Evans 2013; Maffie 2022; Mills and Owens 2021; Rosenblat and Stark 2016).

The point here is not to argue for a black and white conceptualization of the presence or absence of workers’ sense of autonomy. Indeed, such an understanding may misrepresent workers’ experiences: Siciliano (2016), for instance, showed how workers in the digital publishing industry worked with technology that both afforded and constrained their sense of creative autonomy. Given that frontline service workers do not always feel autonomous, however, we need explanations for why workers do not resist their own exploitation in workplaces in which the assumption of a sense of autonomy is not fully borne out. In service workplaces characterized by a lack of felt autonomy, what interaction-level dynamics minimize service workers’ opposition to precarious work? Taking Leidner’s insight in a Simmelian direction makes it possible to move beyond the consent-autonomy paradigm and begin to answer this question.

Triangle as Triad

Leidner (1993) argues that service work replaces “the two-way struggle between management and labor with a triangular pattern of shifting allegiances and interests among workers, managers, and customers” (p. 41). She offers a rich portrait of workplace interactions that takes customer presence seriously, helping us understand how on-the-ground life in the service sector is distinct from that in manufacturing. Her idea of the service triangle has wielded great influence in sociological research on service work (Lopez 2010), but it is not without limitations. One major limitation is that the triangle thesis speaks primarily to the level of face-to-face interaction, what Goffman (1983) calls the interaction order. Leidner’s theorization does not distinguish this interactional level from social structure, so it runs the risk of conflating the two. This lack of specificity may minimize the continued importance of structural relations to workplace dynamics by painting power in the triangle as equilateral (Bélanger and Edwards 2013; Taylor and Bain 2005).

Interactional relations between workers, employers, and customers may belie the structural relations between workers, employers, and customers. The solution to this problem is not, however, to assume that social structure is the base that shapes all workplace experiences (e.g., Bélanger and Edwards 2013). As Goffman (1983) argues, interactional and structural domains can operate fairly independently: “what one finds, in modern societies at least, is a nonexclusive linkage—a ‘loose coupling’—between interactional practices and social structures” (p. 11). Instead of giving social structure or the interaction order primacy, we can turn to Simmelian formal sociology to re-envision the service triangle so that it speaks to interaction-order dynamics without ignoring or obscuring social structure.

For Simmel (1950), social relationships have both form and content. Content refers to “what something is about,” while form refers to something’s “shape or outline” (Broćić and Silver 2021:89–90). The content of the service triangle, for example, is about interactions between employers, workers, and customers in the service workplace. Following Simmel (1950), the social form of the service triangle is a “triad,” a social relationship involving three “elements” (people, parties, countries, etc.). Simmel (1950:145) identifies three “typical group formations” that emerge with triadic interactions but cannot occur in others. In one type, a third element is added to a dyad as a “non partisan or mediator” that calms the passions of the other two elements. In another type, divide et impera (divide and rule), a third element actively divides the other two elements, preventing their unification to gain advantage. In a final type (which maps onto Juniper’s dynamics), a third element—the tertius gaudens (the third who benefits)—passively draws advantage from the other two elements’ conflict. Conflict holds the first two elements “in check,” effectively paralyzing them and allowing the third element to “make a gain which one of the two would otherwise deny him” (p. 154).

Because the Simmelian formal approach gives primacy to social patterns (Vaughan 2014; Zerubavel 2007), subsuming content to form (Broćić and Silver 2021; Simmel 1909:299), in principle three parties in any context at any scale can be discussed as elements of a triad (Vaughan 2014). We could just as easily use the triad to describe relations among the United States, China, and Russia as we could to understand relations between two parents and their child. However, whereas Simmelian sociology typically conceives of any social form as existing at only one level of analysis, I argue that some social forms may exist at multiple levels of social life simultaneously. The service triad, for instance, exists at both the structural and interactional levels at the same time: there is both a structural relationship uniting workers, customers, and employers and an interactional one.

Reconceptualizing the service triangle as a formal triad allows us to explore both its structural and its interactional reality, and to see how, in each case, triadic control operates in its own way. At the structural level, we might observe how employers’ control over the labor process shapes the structural relationship between workers and customers: employers, for instance, institute the tipping system, and they require employees to perform their work according to specific rules. This is essentially the mediator triad, in which one element (employers) shapes the relationship between the other two (workers and customers). At the level of the interaction order, things appear differently. Here, we might observe how customers’ interactional control over workers shapes the interactional relationship between workers and their bosses: customers determine the work that workers must perform for their managers, and they encourage workers and bosses to work together to serve them. This triad, too, could be considered a mediator triad, in which one element (customers) shapes the relationship between the other two (workers and employers). Still, although each of these analyses may be correct, neither tells the full story on its own.

If we examine multiple levels of triadic dynamics simultaneously, we can observe how these two single-level dynamics are part of a larger multilevel dynamic of control based in the interaction of social structure and the interaction order. Employers shape the structural relationship between workers and customers, and this has implications in the interaction order. The tipping system and employers’ rules do not just determine the structure of workers’ relations with customers, they also empower the customer in social interaction: the employer-instituted tipping system compels servers to cater to customers interactionally (to earn their pay), and employers’ rules stipulate how workers should interact with customers (as subordinates). Customers’ control in the interaction order, then, is at least in part a product of employer control at the level of social structure. Interaction-order dynamics are likelier to hold our focus than structural dynamics, as we experience these dynamics directly in everyday life (unlike social structure, we can observe and participate in interactions directly). This means that, to workers, customer control should be more salient than employer control. As a result, although employers are structurally responsible for workers’ experiences of precarity, customers are likelier to be blamed for workers’ frustrations that are tied to precarity. This is the tertius gaudens triad, in which a third element (employers) passively draws advantage because of the other two elements’ (workers and customers) conflict. Recasting the service triangle as a social form makes it possible to understand how labor control can operate at multiple levels of social life simultaneously without relying on workers’ experiences of autonomy or giving primacy to either social structure or the interaction order.

In the next section, I describe my methods before turning to an analysis of how the interplay of structural and interactional triadic dynamics encourages workers to direct frustrations toward customer scapegoats rather than management.

Methods

From October 2014 to March 2016, I collected data through participant observation at a small, midrange restaurant in a small midwestern city. I worked as a server at Juniper for approximately 20 hours each week and regularly socialized with Juniper staff members outside of work. I jotted short notes on an order-taking notepad while working but wrote most field notes after work. Notes focused on interactions between managers, customers, and front of house (FOH) employees who interacted directly with customers. ¹ Participant observation allowed access to workers’ experiences of interaction rather than reports about those experiences (Jerolmack and Khan 2014). As I was concerned with the situated dynamics inside the triangle, this method was crucial. ² Many of Juniper’s serving employees were students at nearby colleges. Some workers were college graduates, some had dropped out or taken a “break” from college, and some had no educational experiences beyond high school. Workers often had additional jobs on the side. Servers were predominately white and in their late teens or early 20s, though several bartenders were closer to 30 years old. Managers at Juniper had a similar profile, though they were typically older (in their late 20s or early 30s). As in many American restaurants (Wilson 2020), Juniper’s staff was de facto segregated. Compared with FOH workers, back of house (BOH) workers were more likely to be Latinx, have immigrated from Mexico or Central America, and speak Spanish as a first language. Customers were predominately white, but they were somewhat more diverse racially than the FOH staff and the city itself. They included professionals, college students, weekend brunch-goers, families with young children, business travelers, and parents visiting their kids at college. The restaurant and its workers are not representative of any population. Rather, the processes I observed are theoretically generalizable to other situations (Stolte, Fine, and Cook 2001), potentially found elsewhere under similar conditions. Still, descriptions in others’ research (e.g., Leidner 1993; Rosenblat and Stark 2016) and popular accounts (e.g., Brecht [1928] 1994; Cardosa 2016) make me confident that the dynamics I observed have occurred elsewhere.

Any service setting in which the dominant social form is triadic, where work is precarious, and where workers do not feel a strong sense of autonomy would have been a useful research site. Juniper fit this bill. Still, Juniper was a particularly useful case. First, Juniper FOH workers and customers were from a wide variety of socioeconomic backgrounds, mostly white, and mixed in terms of gender. This makes me more confident that the dynamics I observed were not simply about race, class, or gender differences. Second, many years of restaurant serving experience made a restaurant a good site for me to conduct this research, as I was able to focus on interactional dynamics rather than learning how to be a server.

Data analysis was abductive (Timmermans and Tavory 2012). As I collected data, I wrote memos about emerging themes and processes; these developed in an analogical and iterative fashion as I moved among written fieldnotes, memos, and fieldwork (Vaughan 2014). My focus shifted as I observed features of the Juniper workplace that seemed surprising in light of sociological research and theory, in particular, how the key divide was between workers and their customers rather than between workers and management. This became the focus of the research. The article reflects this analytic choice; it is not intended as a description of all modes of engagement present at Juniper or in restaurant work generally. (There were, for example, some confrontational interactions with bosses and enjoyable interactions with customers). It is for analytic purposes that I isolate this dominant aspect of Juniper’s multifaceted interactional domain.

Before I discuss my findings, two points around terminology are worth mentioning. First, for simplicity, I use service industry, service work, and service workers to broadly refer to interactive service work and interactive service workers, though I recognize that the service sector includes work involving little customer interaction. Second, I also use words such as managers, employers, and bosses somewhat interchangeably. Again, this is a simplification. Managers are often mediators between workers and employers (or other managers), may have little power, and may not profit directly from workers’ labor (Bolton and Houlihan 2010; Misra and Walters 2022). However, as managers at Juniper act as employers’ representatives at the level of social interaction and my analysis focuses on the interaction order, this simplification seems justified in the interest of clarity.

Channeling Frustration: Income Precarity

Triadic structural and interactional dynamics position customers as the cause of frustrating income and interactional precarity. I begin here with income precarity.

Juniper, like many restaurants, shifts income control to customers through the tipping system. Employers structure this system. They pay servers the federal minimum wage for tipped employees of $2.13 per hour (less than a third of the standard minimum wage), but they leave it to customers to make up the rest of servers’ wages with tips. Servers anticipate taking in 15 percent to 20 percent of each customer’s tab this way. If servers do not earn the minimum wage ($7.25/hour), the restaurant must make up the difference, but by calculating that wage biweekly rather than hourly or by the shift, the restaurant rarely contributes additional funds. The result of this wage system is that, in servers’ experience, working life is directed toward avoiding customer ire and maximizing customer pleasure.

Customers who require a lot of work relative to how much they tip are undesirable, and workers heuristically construct categories of these bad tippers. Such naming practices have been well documented in research on service industry workers (e.g., Erickson 2009; Paules 1991). At Juniper, these categories reflect the power servers feel customers have over them. ³

Take “coasties.” Coasties are people from the East or West Coast. They are often college students or their family members, and workers feel that they are wealthy (they can afford out-of-state tuition), needy, and unreliable tippers. According to Ben, a server, although it is normally “tacky” to “auto-grat” a table (to apply an 18 percent gratuity to large parties’ checks), it is okay to auto-grat coasties. Servers, including Arlene, do not relish waiting on coasties:

A middle-aged woman and her husband sit down in Arlene’s section. They are staying in a hotel nearby and awaiting their college-aged son; it’s his birthday. The woman orders an iced latte, but with soymilk. Later, she also orders her salad with dressing on the side, and she requests a side plate for the bread that usually comes in the salad bowl. According to Arlene, this typical coasty behavior makes her job harder.

As Arlene’s suggests, coasties are irritating because they require a large amount of work for an uncertain payout. Workers do not assume coasties will be either particularly good or bad tippers in terms of absolute percentages or amounts, but they do assume that coasties will not tip nearly enough to compensate for the extra work they require (even with a 20 percent tip, waiting on coasties might not be worth it). This makes coasties risky customers.

The “U-Card student” is another type of unappealing customer. Juniper accepts a form of payment called the “U-Card,” a part of a student meal plan through a local university that allows parents to load money onto cards that students can use at local restaurants. U-Card holders bring extra work because of the tedious U-Card payment process. The restaurant has a single, notoriously slow U-Card reader, and servers must learn a complicated and unintuitive procedure to operate it. Waiting for a chance to use the machine and then for the machine to process payments can put servers behind during busy shifts, threatening their ability to take care of other customers. What’s more, if workers do not properly enter U-Card tips into the U-Card reader (in addition to the regular computer system), they do not receive these tips.

Workers also assume that U-Card students are bad tippers, so the work of processing U-Card payments is not felt to be worth the reward:

One afternoon, Megan stood by the computer terminals entering tip amounts from U-Card receipts: “I’ve never gotten anything more than two dollars on a U-Card. Like ever.”

Megan goes on to say that, basically, these customers are kids whose parents just give them money. She’s indignant: “It’s not even their money, and they still can’t tip.” She doesn’t have a U-Card. She just uses her debit card for food when she’s not on campus, and she pays for all her food. “So it must be that their parents are just giving them money, because why wouldn’t they just use their debit card?”

Students who pay with U-Cards make servers do extra work but do not compensate them for it. Megan’s indignation at her low pay is compounded by the perception that students who use U-Cards do not even lose anything by tipping well (servers assume the cards are filled with parents’ money, rather than the students’ own). This dramatizes the arbitrary power that servers feel U-Card students have over their income.

Workers’ use of typologies at Juniper reveals their lack of felt control in the workplace and their sense that customers, not employers, determine their income. An alternative interpretation of these dynamics could be that servers are wresting back a sense of autonomy from customers: servers might use typologies to ensure that they will be tipped well, to avoid risky customers, or to minimize the work they put into bad tippers. However, servers do not believe that their tips have much to do with their job performance: tips seem contingent upon customer type or the vagaries of customer mood. What is more, servers do not get to choose which customers they have; this is up to restaurant hosts. Finally, and most important, servers cannot use these typologies to minimize their work because the categories are revealed only after servers have committed to or already completed the extra work. Coasties become coasties when servers are already being sent to fetch another ramekin of béarnaise or to retrieve a Diet Coke with just a little less ice, while U-Card students materialize when they are ready to pay.

Although restaurant management structures this wage system, customers act it out in the interaction order by distributing tips. The wage system instituted by employers at Juniper channels workers’ frustrations with income precarity toward customers rather than management, as evidenced by workers’ use of typological categories to make sense of challenging customers.

Channeling Frustration: Interactional Precarity

Triadic dynamics position customers as responsible for servers’ interactional precarity as well as their income precarity. In this section, I describe the company myth that sets up Juniper’s rules for customer-worker relations and analyze how it translates into customers’ interactional dominance.

The Lettuce Myth

In service workplaces, customers are commonly framed as powerful sovereigns who must be pleased despite their sometimes difficult or even abusive behavior (Bishop and Hoel 2008; Hochschild 1983; Korczynski and Evans 2013; Korczynski and Ott 2004; Mills and Owens 2021). Juniper sets up this dynamic with a company myth, which trainers share with new servers during their first shifts and management recites during biannual staff trainings. This training instructs servers to put customers’ needs above all else.

The myth goes like this: Juniper’s owner, Mike, who owns several restaurants in addition to Juniper, was at his vacation home. While preparing a BLT (bacon, lettuce, and tomato sandwich) for lunch, he realized he lacked one important ingredient: the lettuce. Mike strolled down the road to a small restaurant, where he asked if he could buy some lettuce. The person at the counter told him no; they needed the lettuce to make their sandwiches that day. Mike was floored! How could they not give him lettuce? What awful service! Mike wondered how the workers at his restaurants would act in this situation, so after his vacation he sent an “old woman” to attempt a lettuce purchase at each of his restaurants. Before the next management meeting, Mike surreptitiously placed a head of lettuce on the chair of each of his general managers whose restaurants had passed the test. To everyone’s relief, all the managers were greeted by lettuce.

This tale evolved into the company philosophy: “Give them the lettuce!” The idea is that the key to hospitality is giving customers what they want, whenever they want it. Of course, in practice, workers, not Mike or managers, execute this philosophy. This counters worker/boss structural dynamics: whereas bosses set up and benefit from a traditional employment relationship, it is customers to whom workers are most subordinated in the interaction order.

This subordinating dynamic is especially noticeable when servers deviate from the rules. In the following example, Rachel, an earnest, overworked, new, and busy server, felt she had failed her customer:

Usually the consummate polite, agreeable, and chipper server, Rachel seemed to be getting to her breaking point. Both of us had been working since 7:45 am, and it was now approaching 9 pm. A man walked over from one of the tall bar tables. His puffed-up chest filled his black T-shirt and his bushy white eyebrows closed in at the center of his forehead, one higher than the other. Expecting a “yes,” he asked Rachel: “Can we order from the menus over here? We only have one waitress, and she’s slow and hasn’t come back with our cocktails.”

Rachel responded quickly with a strained smile: “Sir, we’re very busy.”

“Are you?”

“The restaurant is full.”

He stared for a moment, then cocked his head with a slight jerkiness that suggested he’d already had a few cocktails. He wandered back over to the bar area.

Rachel’s face filled with shame. “Was that okay? I didn’t—we’re supposed to say ‘yes.’”

Rachel’s bosses set up the interactional rules that empower customers over servers, pushed her to a breaking point with an 11-hour shift, and did not staff enough people to cover an influx of customers. Rachel was too plagued by the needs of the customers she already had to handle a new request. Unable to please the customer, Rachel was initially frustrated with herself. Over time, like the rest of her coworkers, she began to blame customers for bad shifts such as this one. She did not direct her frustration to the bosses who put her in impossible situations.

Interactional Entrapment

Customers’ dominant interactional position put workers at continuous risk of what I call interactional entrapment. Workers surrender control to customers, so customers always have the power to start and end conversations with servers, regardless of servers’ own needs. A classic example of entrapment is when customers (often men) snag servers (often women) in conversation. Work control is often gendered (Crowley 2013), and women servers’ interactional subordination to customers is underscored by gender dynamics in addition to workplace rules. I had such an experience one Saturday evening:

I was at a booth near the back of the restaurant rolling silverware into napkins when a white man in his late sixties stumbled in silently with his son. They sat down across from me. For the next fifteen minutes, while I smiled and tried to roll silverware so I could go home to snatch a few hours of sleep before my shift early the following morning, the son asked me personal questions while his father silently stared, occasionally letting his head nod down toward his neck then yanking it back up when the movement woke him up. I wanted to finish my work quickly and leave, but I was stuck.

Servers are always on reserve for customers, so servers’ concerns must be backgrounded. On a busy weekend morning, Olivia was training Tania, a new server:

It was Tania’s last day of training, so Tania worked by herself while Olivia watched her from the corner. A host sat a new customer at one of her tables, and Tania hesitated before going up to him. Olivia instructed Tania to take the table. Tania presented the young white man in a sweatshirt with a glass of water and wrote something down on her notepad; then she came back to Olivia. The man, it transpired, was her ex-boyfriend. She’d broken up with him a few months before, and now here he was, at her new job, trying to get back together. She had to serve him. While Tania asked him kindly, but firmly, to leave once he had eaten and paid for his meal, she was unable to do so until the conclusion of the economic exchange—and then only because she knew him.

Later, Laura, a bartender who had witnessed the encounter, summed up the situation’s power dynamics: “She was trapped there.”

The risk of interactional entrapment means that servers’ lives are always in danger of customer capture. Whether a server is engaged in a work task (such as greeting a new table) or a personal one (such as hurrying to the bathroom), a customer might suddenly wrench her onto a new course. In a Goffmanian sense (Goffman 1959), customers were able to keep servers frontstage or prevent servers from retreating to backstage spaces to repair and prepare for more. The experience of serving customers at Juniper was not one of autonomously controlling one’s own labor; rather, it was one of being buffeted about by other people’s whims. This interaction-order experience is at odds with social structure. Servers experienced customers as the main aggressors they faced in the Juniper workplace, even as management, the third element to this triadic form, structured and passively benefitted from this arrangement. Below, I show how servers responded to this experience of customers, demonstrating the conflict dynamic of Simmel’s (1950) second triadic type, in which two parties of a triad clash, “hold[ing] each other in check” (p. 154).

Servers and Their Foes

As servers experienced income and interactional precarity at customers’ hands, customers emerged as the source of their workplace frustrations. This is evidenced by servers’ comments about customers, whose appearance and behavior they sometimes mocked. For example, servers might disparage a Ted Cruz lookalike or point out customers on bar stools whose behinds are exposed by low-cut pants. This vision of the customer was also apparent in servers’ interactional fantasies. One snowy weekday evening, management closed the restaurant for a staff party:

A sign on the front door indicated Juniper was closed for a private event, but the door was unlocked so that employees could get in. I sat in a leather booth with three of my coworkers, and our conversation turned to what we would say if a customer happened to come in during the party. My co-workers suggested responses along the lines of “Get the fuck out!” or “We’re fucking closed!”

Eventually a young woman carrying a backpack did come in the front door. She waited at the host stand, looking around blankly to see if anyone was going to get up and help her to a table. We craned our necks toward the door, and a few servers shut her down in gleeful unison: “We’re closed!!”

We talked about this triumphant event, and the stupid customer who couldn’t read a sign, multiple times over the rest of the evening.

Servers could not be so direct with customers on the clock. They could only challenge customers’ elevated status after work or out of earshot.

The frustration servers felt toward their customers could boil over into anger. During a busy morning filled with rude and demanding customers, Sophie, a bartender, came up with an imaginary “game” that shows just how far this rage could go (in fantasy). Unlike the game of making out documented by Burawoy (1979), this fantasy underscored Sophie’s lack of autonomy:

Sophie frowned and complained about a customer. I didn’t catch everything she said, but I could tell she was upset. Expressing sympathy, I asked her, “Do you want me to go around giving death stares to all of them?”

“Yes. That would be great,” Sophie said. Then, after a brief pause, with a loud and maniacal laugh: “Or we could just kill them.”

Laughing back, I replied, “We’ve definitely got enough, uh, kitchen supplies to really do some damage.”

“We can just throw knives at them,” Sophie said, deadpan. We fantasized about a new “game.” We’d grab handfuls of steak knives from the plastic bin by the kitchen window and simply throw them at the crowd. They would hit whomever they hit, though we would try not to aim for people who had not personally caused us harm.

Sophie never played the “knife game” (by traditional definitions, it was not even a real game). This rebellion only existed in fantasy. Nonetheless, the knife game shows how, in Juniper’s interactional domain, customers, not bosses, were workers’ “foes,” the cause of their workplace frustrations. The identification of customers rather than bosses as workplace foes was not surprising considering that, because of how bosses had structured the labor process, customers ultimately controlled just how much workers could earn in an hour, how busy they were, or whether they would be able to use the bathroom. Workplace resistance was resistance to customers, not management, and it was toothless both because it existed as a fantasy (workers were not permitted to take their complaints up with customers) and because it addressed the interactional cause rather than the structural cause of their frustrations. I consider how management draws advantage from this “paralyzation of forces” (Simmel 1950:154) in the following section, where I briefly contrast the worker-customer divide with the more congenial relationship between workers and their managers.

Servers and Their Managers

Thus far, I have focused on one relationship in the service triangle—that between servers and customers—but, as Simmel (1950) reminds us, one cannot look at one leg of the triad without also looking at the whole thing; one must consider the form in its entirety rather than thinking solely about one of its component parts. Servers’ relationship with management shows one of the interactional alternatives that existed at Juniper. Servers were not simply hateful, mean people. Rather, interaction-order dynamics fueled their negative feelings toward customers.

Managers at all levels structured the labor process that transformed customers into servers’ interactional foes. They hired and fired servers, scheduled servers’ shifts, decided when servers could leave for the day, assigned servers’ sections, and granted or denied their requests for time off. However, when customers are servers’ foes, managers can be servers’ friends.

Management and workers got along well. For example, Kate, a server, organized a group gift for Sebastian, an assistant manager, on his birthday: everyone chipped in for a trip to a new aquarium exhibit. Sebastian used the money to pay for a keg of beer at an informal work party a few weeks later. In an apartment courtyard, Sebastian joined the serving staff for rounds of badminton over the beer he bought. Managers often attended gatherings such as this, where they were usually treated as one of the crew. Additionally, four of the managers had been long-time (over a year) servers and bartenders at the restaurant before taking on additional management duties. These four continued to work in their old capacity as servers and bartenders for some shifts each week, and they maintained friendly relations with coworkers both during and after work. That managers and servers continued these friendly relations after the promotions (when the new managers could hire, fire, and schedule servers, decide when they could go home, assign their sections, and grant or deny time-off requests) only makes these relations more notable. ⁴

During busy times, managers often stepped in to help servers. They would seat new customers, bring food to tables, mix drinks at the bar, clear empty plates, and run customers’ credit cards, always making sure that cash tips ended up in the appropriate servers’ pockets. Together, servers and managers would beat back the flow of customer demands, sharing complaints, stress, hunger, thirst, and aching feet at the end of the day (though managers could always step away for breaks at will). In contrast, workers and managers often spent slow times at work (the rare moments when workers had enough autonomy to direct their own behavior) amicably messing around together. Research has found that workplace games and fooling around are key to passing time and generating consent (e.g., Burawoy 1979; Roy 1959; Sherman 2007). What is interesting about workplace entertainment at Juniper, though, is that it was often a joint production of management and workers, rather than centering on interactions between workers and their machines (Burawoy 1979), workers and their customers (Sherman 2007), or workers and other workers (Roy 1959). Goofiness and frivolity cemented friendship dynamics between workers and managers as they told each other jokes, gossiped, looked up weird facts on the Internet, or evaluated Taylor Swift albums.

Not everyone got along. Staff resented one of the managers, Tim. He could be sarcastic and harsh when busy or in a bad mood. One such morning, he saw a new server’s phone pecking out between her jeans and the bottom of her blue work shirt. Quietly, he hissed, “Get that phone out of your back pocket. It looks like fucking shit.” Still, although Tim could be intimidating, his anger was usually directed toward workers who were blatantly breaking rules. Some workers did not appreciate his jabs but still expressed their respect for him, feeling that he made the workplace run more smoothly and that, as the person in charge and answerable for Juniper’s day-to-day operations, he was only doing his job. Other workers contradicted negative assessments of Tim. Charles, a server, explained how he was fun when they went out drinking once, and Tim and Olivia would sometimes talk and laugh together at the host stand. Many workers were friends with him on social media. In any case, workers who did dislike Tim seemed to resent him for his sarcastic personality, not for enforcing rules (i.e., not for being a manager).

The trouble with Tim was not the same as the trouble with customers. Like customers, Tim’s personality could frustrate servers. However, Tim could always be counted on to take out his anger on people who were breaking rules. In that sense, his behavior was predictable even if his tone was not. Crucially, too, servers experienced Tim as an individual rather than as a part of an anonymous block (as “Tim,” who had both good and bad qualities, rather than as “customers”).

Management was spared criticism as a rule. The only anger workers directed at management was anger about managerial tone and volatility, not the structure of the labor process, and it was directed toward individuals, rather than indiscriminately toward “management.” In Juniper’s interaction order, managers were largely servers’ friends and teammates; occasionally they were jerks; never were they the primary source of workplace frustration. Keeping critiques of the employment structure off the table meant that most of workers’ frustrations would flow toward customers.

Formalizing the Service Triangle

To summarize, Juniper customers have control over workplace interactions and appear through the tipping system to be responsible for workers’ wages, whereas employers’ control lies in the structural power of the employment relationship. As workers become frustrated with precarious elements of their work as set up by employers, they direct their frustration toward customers. Employers benefit because they are not blamed for workers’ frustrations and because worker resistance becomes symbolic resistance to customers’ interaction-order power rather than resistance to employers’ structural power.

Recognizing the social form, a triad with a limited number of possibilities for interaction, enables us to make this situation more abstract (see Figure 1). There are three parties to the triad at Juniper: employers (A), workers (B), and customers (C). Because C has salient control in the interaction order, while A has structural control, B blames C for her problems and A benefits.

Figure 1.

Formalizing the tertius gaudens dynamic.

Simmelian formal sociology thus illuminates formal scope conditions for the tertius gaudens dynamic identified here: it applies in situations in which B has low autonomy, C is interactionally salient and can control B’s behavior in the interaction order, and A is less interactionally salient but controls social structure. Where else might we find this dynamic?

Below, I use formal analogy to extend this social form to other workplaces and identify its content-based scope conditions (Vaughan 2014; Zerubavel 2020); I then scale up the discussion to address the practical implications of these triadic dynamics for labor organizing. Simmelian formal sociology aims to uncover “common formal patterns across substantively diverse contexts” (Zerubavel 2020:39). Because formal sociology is transportable, scalable, and relevant to multiple contexts, we can use case comparison and analogy to create and refine falsifiable propositions that extend beyond one social context to contexts that seem divergent (Broćić and Silver 2021; Land 2001; Prus 1987; Snow, Morrill, and Anderson 2003; Vaughan 2014; Zerubavel 2020).

Formal Scope

The tertius gaudens dynamic would apply in other workplaces. Consider rideshare drivers who interface with management through phone applications. Rosenblat and Stark’s (2016) study showed that these workers, beholden to the instructions of a phone application, experienced little autonomy. Their most salient interactions were with customers—the only people workers encountered in person—who directed and disciplined them. Employers structured the employment relationship, but customers were blamed for frustrating experiences. When management becomes more interactionally salient, however, this dynamic shifts. In Maffie’s (2022) study of drivers working for both Uber and Lyft, using two platforms with two different customer rating systems highlighted employer control practices and thus also highlighted employer responsibility for working conditions. These drivers blamed management, not customers, for work frustrations.

The tertius gaudens dynamic is likely operative in workplaces in which workers have low autonomy and tips or commissions effectively comprise workers’ wages. This might include other restaurants such as Juniper, where customers come in waves and servers earn a subminimum wage. We should also expect the tertius gaudens dynamic in low-autonomy no-tip workplaces in which customers exercise other forms of control like making requests that prevent workers from completing tasks, leaving work, and taking care of their own needs (e.g., Reich and Bearman 2018; Vargas 2021). For example, researchers have found that retail workers can become frustrated with customers who mess up their product displays (Ikeler 2016; Misra and Walters 2022).

We would not expect this dynamic in places where workers take time to make service a precisely tuned production, such as high-end restaurants. In such high-autonomy workplaces, workers likely play the consent-generating games documented elsewhere (Sallaz 2002; Sherman 2007). For example, Ikeler (2016) found that Macy’s workers who applied specialized knowledge to persuade customers to buy products both experienced a sense of autonomy and an adversarial relationship with management. Meanwhile, workers at Target who engaged in routinized work and scripted interaction with customers did not often express strong management opposition and described frustrations with customers. Differences in workers’ sense of autonomy are not the only cause of variation in management and customer attitudes, but this pattern aligns with the tertius gaudens dynamic.

There are also low-autonomy workplaces in which the dynamic at Juniper would be absent: workplaces where customers are interactionally salient and management control is obvious, but where customers have little interactional power. For example, fast food work requires customer interaction and is directly controlled by management, but customers’ access to workers is routinized and controlled by lines and counters (Leidner 1993; Ritzer 1993). We would expect these workers to direct frustration to employers. Widespread strikes by fast food workers may reflect these dynamics (Luce 2017). Or, consider certain care workers. The most vulnerable care recipients are interactionally salient but nondominant. Coupled with health care aids’ internalization of norms surrounding care recipients (Stacey 2011), this nondominance may prevent aids from pinning their frustrations on clients themselves, making employers’ actions a handier explanation for workplace troubles. ⁵ As Brown and Korczynski (2010) found, when speedup interferes with health care workers’ effort to provide “meaningful care,” workers may ally with their clients and distance themselves from management. Here, we can begin to see how cultural content shapes triadic dynamics, as norms around care shape who gets blamed for work-related frustrations.

Content-Based Scope

Simmel’s formal sociology does not attend to content, but particular work contexts, economic regimes, regulatory structures, and cultural environments shape triadic dynamics. Content-attuned analysis makes Simmelian formal analysis more powerful by helping explain two things Simmel does not: how a triad comes to be and the type of triad it becomes.

For example, the tertius gaudens dynamic depends upon employers having more structural power than workers because of the employment relationship (workers must sell their labor to survive), tipped minimum wage laws, union decline, and, more generally, the rise of precarious work. More employee power might lead to the mediator type of triad (Collett 2011), in which servers mediate between the conflicting interests of bosses (who seek high profits) and customers (who seek good deals). Alternatively, in the context of strong “waitress unionism,” servers might respond to mistreatment by allying with customers against employers (Cobble 1991).

This dynamic also depends on the culture of customer sovereignty granting customers interactional power (Korczynski and Ott 2004). We would not expect customer scapegoating in workplaces shaped by a history of customer subordination to restaurant workers (Spiegel and Rosin 2016), where workers “talk back” to customers (Whyte 1948); or where workers and customers are constructed as equals (Besen 2006; Wilson 2016). Alternatively, in high-status restaurants, “chef sovereignty” might eclipse customer sovereignty, with servers answering to chefs and managers as one group and customers as another (Borkenhagen and Martin 2018; Bourdain 2000; Leschziner 2015). Customers’ attitudes matter, too. In Otis and Wu’s (2018) study of Beijing retail workers, customers disproportionately directed abuse toward rural-origin compared with urban-origin workers, so rural workers felt antipathy toward customers while urban workers identified with customers.

As Otis and Wu’s (2018) research suggests, intraworker boundaries are relevant to the tertius gaudens dynamic. At Juniper, the divide between FOH and BOH workers shaped who ended up in the triad to begin with. Four factors maintained this divide. First, orders were submitted to the kitchen via computers, and the kitchen typically ran smoothly, making extensive FOH-BOH interaction unnecessary. Second, Juniper’s physical structure separated BOH and FOH workers and made BOH workers peripheral to customer interactions. Third, two different wage systems (BOH workers earned hourly wages and FOH workers earned tips) meant that BOH and FOH workers had different relationships to customers and to employers. Underlying these factors is a fourth: the racial/ethnic boundary separating FOH and BOH workers. Cultural and language divides discouraged high levels of interaction between BOH and FOH workers, and spatial segregation kept BOH and FOH workers apart and signified BOH workers’ “peripheral or subordinate position” (Vallas 2003:385). A racial/ethnic hierarchy enabled a dominant group (white, non-Latinx FOH workers) to construct a subordinate group (Latinx BOH workers) as incidental to everyday life (Guiffre and Williams 1994). As a result, for FOH workers, BOH workers were simply not as interactionally salient as managers and customers. If Latinx workers were mostly in the FOH and white non-Latinx workers were mostly in the BOH, I would have observed different dynamics. Alternatively, absent BOH-FOH divisions, we might see all staff as one element.

Although Simmel does not describe triadic formation, triadic membership is dependent on social context and not inevitable. There are numerous other formations we might find in a restaurant, given different conditions (Whyte 1948).

Scaling Up to Labor Relations

Scaling up the triad allows us to see how workplace relations are themselves part of a larger set of labor relations: the tertius gaudens dynamic, repeated many times over, might play out among whole groups of workers, bosses, and customers. That is, we can connect the service labor process to practical issues of labor organizing, a neglected effort in studies of service work (Ikeler 2016; Lopez 2010). Three implications for labor organizing follow from this observation.

First, if workers do not experience their employers as meaningfully contributing to bad working conditions, they are not likely open to challenging those employers (Ikeler 2016). Organizers who do not address the possibility that workers feel customers and not employers are their primary workplace foes may be ineffective. Second, the customer-as-foe dynamic could preclude worker-consumer alliances that have often been part of effective organizing tactics (Chun 2005; Cobble 1991; Lopez 2010). Efforts to form worker-customer alliances may need to address worker-customer tensions. Third, this analysis provides mixed support for anti-tipping movements. Tipping shifts blame to customers. It also helps maintain a boundary between FOH and BOH workers, so ending tipping could make powerful FOH/BOH alliances possible. Still, eliminating tipping would not eliminate the culture of customer sovereignty. Furthermore, given the miniscule pay many servers receive from employers, servers accustomed to tips might oppose giving employers more control over their income. One solution could be to organize for the elimination of the tipped minimum wage but not tips, so that FOH and BOH workers are paid in similar form, tips become less meaningful, customers become less powerful in the interaction order, and employers’ structural control over workers’ income becomes more apparent.

Sociologists, organizers, and workers may disagree about the key force of domination in workers’ lives, with sociologists and organizers pointing to social structure and employers, and workers gesturing to the interaction order and customers. Rather than calling out the interaction order for its lack of resemblance to social structure, however, we could instead use models such as Simmel’s to nimbly account for both dimensions of social life. This would lead to richer understandings of social life and advance our joint efforts to create a world in which people resist rather than consent to poor treatment and exploitation.