Supermarket ‘dark jobs’ and rapid grocery delivery: Transformations in labour,technology and logistics

Abstract

As demand for rapid grocery delivery surged during the COVID-19 pandemic, Australia’s supermarket duopoly set about transforming relations of labour, technology and logistics to secure dominance in the growing sector. I consider the rise of ‘dark jobs’ of the supermarket and what this means for affected workers. My research encompasses in-depth interviews with 17 supermarket workers, including personal shoppers and dark store workers. I also draw upon corporate documents and media statements released during the pandemic, as well as insights gleaned through my work within Australia’s union movement. I argue that the transformations underway result in three major developments. First, the intensification of supermarket work to adhere to the temporal logics of the on-demand economy, enabled by the increasing use of digital devices. Second, the absorption of precarious workers via digital labour platforms into the traditional firm which may undermine existing job security. Third, a blurring distinction between retail and logistical spaces with consequences for union representation that may erode workplace conditions. Taken together, the supermarket becomes a powerful testing ground in which the logics of on-demand platforms are transforming conventional employment settings and practices.

Introduction

So ubiquitous is the supermarket that at first glance it can appear banal. Behind the mundane exterior, however, ‘the entire network of infrastructures, technologies, spaces, workers, and violence that makes the circulation of stuff possible remains tucked out of sight for those who engage with logistics only as consumers’ (Cowen, 2014: 1). For many supermarket workers, the surge in demand for rapid grocery home delivery accelerated by the COVID-19 pandemic is radically changing the nature of their work. Australia’s supermarket duopoly – Coles and Woolworths – are steadily transforming technological and labour relations to both generate and meet demand for rapid grocery delivery. This includes a substantial increase in ‘personal shoppers’, who follow algorithmically determined instructions to pick orders from supermarket shelves; formal partnerships with last mile digital platforms including DoorDash and UberEATS; and the expansion of purportedly semi-automated dark stores and fulfilment centres.

The economic effects of the pandemic have been unevenly felt, particularly across ‘winner-takes-all’ industries such as grocery and e-commerce (Oxfam International, 2021). While smaller independent grocery stores suffered losses due to decreased foot traffic during lockdowns (Giles, 2022) and an inability to quickly pivot to online services (Stewart, 2021), profits boomed for the supermarket giants. By leveraging their unrivalled logistics networks, control over the supply chain, and real estate dominance of brick-and-mortar stores, Coles and Woolworths were able to exploit the chaos of the pandemic to further entrench their duopoly position in one of the most tightly concentrated grocery industries in the world (Merrett, 2020). The transformations discussed in this article are significant, because as Australia’s largest private sector employers (Pash, 2018; Woolworths Group Limited, 2021) they wield considerable political influence and the power to reshape or erode industrial norms. Unlike other many other employers, Coles and Woolworths have historically operated within the paradigm of Australian corporatism (see Humphrys, 2019). This involves cooperation and/or collaboration with the state and unions, and a commitment to shoring up public legitimacy through community outreach and favourable media narratives. As a result, both Coles and Woolworths rank very highly as the most trusted brands in Australia (Roy Morgan, 2022). Although imperfect in many ways, for decades Coles and Woolworths have provided modest yet secure long-term jobs. Recent developments may signal a shift towards more precarious and on-demand employment practices.

On the importance of supermarkets, Joseph Turow has argued that as powerful institutions they can reinforce the ‘hidden curriculum’ – implicit norms and values that condition citizens to accept and accommodate certain ways of organizing public life. Supermarkets have also pioneered foundational technologies of logistics and tracking such as the barcode scanner and customer loyalty cards (Turow, 2017), RFID, GPS, voice picking (Kanngieser, 2013), and even the humble shopping basket and trolley (Cochoy, 2009, 2015; Grandclément, 2009). Vast infrastructures are essential for the lubrication of

. . . movement as much as possible . . . because static merchandise is a liability. . .the ideal condition for a retail logistics network is to have all of its merchandise suspended in a constant state of circulation, a humming network of movement with no backups, no bottlenecks, and no accumulation of storage. (LeCavalier 2016: 6)

Indeed, the just-in-time imperative that characterized twentieth century ‘Toyotism’ and continues to underpin ‘supply chain capitalism’ (Tsing, 2009) took initial inspiration not from the auto-industry, but from the American supermarket (Ohno, 1988).

The importance of the ‘logistics revolution’ (Bonacich and Wilson, 2008) has been taken up as an important area of research (Allen, 1997; Chua et al., 2018; Cowen, 2010, 2014; Lichtenstein, 2009), highlighting systemic issues faced by workers across supply chains (Tsing, 2009) and a global mode of production, which is increasingly logistical in character (Moody, 2019). As Rossiter (2016) shows, from ‘warehouses, ports, intermodal terminals, container yards, and data centers’ the world we inhabit is designed to meet the demands of ‘24/7 capitalism’ (p. xiii). Although logistics has historically been linked with the planning and coordination of military operations, wars and global trade (Cowen, 2014), today some of the biggest retail firms in the world, including supermarkets, are logistics companies (Lichtenstein, 2009; Moody, 2020). Pandemic era demand for on-demand and rapid grocery delivery has further accelerated this logic with implications for logistical labour.

Important recent scholarship has considered the failings of US-based start-up digital platforms promising rapid grocery delivery, and the consequences for urban geographies in the aftermath (Shapiro, 2022). I depart from this focus to consider not the impact of failure, but the supermarket duopoly’s likely success and the possible deterioration of hundreds of thousands of jobs across the country. Put simply, Australia’s supermarket giants are actually doing what start-ups have promised but are unable to deliver, partly due to the very presence of the duopoly power. I take up Vincent Mosco’s (2004) provocation that ‘when technologies . . . cease to be sublime icons of mythology and enter the prosaic world of banality – when they lose their role as sources of utopian visions – that they become important forces for social and economic change’ (p. 6). In comparison with start-ups, the banality of the established supermarkets has provided cover from much public and media scrutiny. Although product shortages and empty shelves were figured as aberrations against the backdrop of frictionless, just-in-time logistics of the supermarket, it is more accurate to say the pandemic revealed the permanent state of crisis that is supply chain capitalism (Chua et al., 2018; Tsing, 2009). Supermarket changes characterized by the pandemic are not, then, a point of exceptionalism, but rather offer a jumping off point from which to analyze broader structural changes that have been legitimized during a ‘state of emergency’ and are likely to endure long after the pandemic.

My research encompasses in-depth interviews with 17 current and former Coles and Woolworths workers, including personal shoppers (n = 12), dark store workers (n = 3) and delivery drivers (n = 2). Interviewees are male (n = 9) and female (n = 8) spanning 21–70 years of age, and located across the cities and regions of every state and territory except Tasmania. I interviewed workers of various levels of seniority from casuals to department managers, with tenure ranging from approximately 12 months to 30 years. For several interviewees, Coles or Woolworths has been the sole employer of their working life. This study was advertised by the Retail and Fast Food Workers Union (RAFFWU) via a single social media post, which directed potential participants to a recruitment website. Interviews were approximately 1 hour in length and took place online or by phone at the request of the interviewee. Although the study was originally advertised via a union, word of mouth snowballing resulted in a majority non-union sample. Several participants were encouraged to take part in the study by their managers and were not themselves familiar with the original social media post. I attribute this to the high rates of employment at Coles and Woolworths and the US$50 Visa gift card incentive that was offered to all participants. Interview data was further contextualized by close reading of corporate documents and media statements, as well as insights gleaned through my professional experience with United Workers Union (UWU) which has members across the supermarket supply chain.

I argue that the transformations underway are resulting in three major developments. First, the intensification of supermarket work to adhere to the temporal logics of the on-demand economy, enabled by the increasing use of digital devices in the retail space. Second, the absorption of precarious gig workers via digital labour platforms into the traditional firm which may undermine existing job security. Third, a blurring distinction between retail and logistical spaces and technologies with consequences for union representation that may further erode workplace conditions. Taken together, the supermarket becomes a powerful testing ground in which the logics of on-demand platforms are transforming conventional employment settings and practices. I conclude by reflecting on the impact of duopoly power and what these shifts may signal for the supermarket of tomorrow.

The personal shopper and automated decision-making

Following the pandemic and Australia’s experience of the world’s longest lockdown (Miller, 2021), online departments and the role of ‘personal shopper’ have boomed within the major supermarkets. These are the workers who pick and pack items from supermarket aisles for home delivery. Although the home delivery role long predates the pandemic, they have since taken on a new dimension and scale. Personal shoppers can be seen moving at speed between aisles pushing large vertical trolleys with individual totes to fulfil as many as eight orders at once. A large touch screen or handheld barcode scanner is mounted to the trolley, which delivers algorithmically determined instructions for which items to pick and in what order, which tote to place the item in, how long each run should take to be completed and the optimized ‘pick path’ through the store for greatest efficiency. In doing so, the device seeks to rationalize the worker’s movements and use of space, which, unlike the regular shopper, become predictable and quantifiable. As one personal shopper describes:

It’s basically idiot proof if you’re doing it right. Because it tells you what you want, it gives you item module, shelf, position, aisle. So [for example] you go to aisle one, module one, shelf three, position two, pick it up and scan it. 99 times out of 100 that would be the correct item, it would light up green, show you what crate to put it in, you open the drawer, put it in, hit enter and repeat. (Personal shopper)

At first glance it may be tempting to characterize this process as straightforward algorithmic management whereby the use of the algorithm becomes ‘an invisible authority that determines the organization of work’ (Massimo, 2020: 133). Although typically associated with digital platforms and the gig economy (Chan, 2019; Lee et al., 2015; Wood et al., 2019; Woodcock and Graham, 2020), algorithmic management is also operationalized in conventional employment settings such as the supermarket (Gent, 2018; Orlikowski and Scott, 2014; Roose, 2019). Collectively, scholars have offered a powerful critique of algorithmic management and the ways in which asymmetries of information intensify precarity and erode foundations of the standard employment relationship (Griesbach et al., 2019; Shapiro, 2018; van Doorn, 2017; Veen et al., 2020). In practice, however, an ‘ideal’ form of algorithmic management in which the algorithm can assume all decision-making functions, and human management is no longer needed, is rare (Ivanova et al., 2018). As Li (2022) has argued, a narrow focus on algorithmic management can risk overlooking historical continuity with other workplace technologies as well as the complexities of organization-based mechanisms of control. At Coles and Woolworths the labour process of the personal shopper is mediated by both algorithmic and human management decision-making, digital devices as well as worker discretion as required. For instance, the algorithm that determines product location is updated constantly by store-level management through a process called ‘sequencing’ to ensure changes in seasonal variation of goods and shelf rearrangements due to promotions are accounted for. In some cases, workers give feedback that is used to maintain the algorithm, and workers mostly report this as being a useful tool in their work.

Personal shoppers must also exercise personal discretion when making product substitutions, a very common requirement amid supply chain disruptions and widespread product shortages. While swapping out one item for another may seem trivial, substitutions are important to the supermarkets for two reasons. First, if the personal shopper does not take the time to substitute and instead marks the requested item ‘out of stock’, this results in lost revenue that multiplies at scale. Workers report the out-of-stock KPI is closely monitored by management as it represents the discrepancy between what the customer wanted to spend, and how much they actually spent. Second, if the personal shopper makes a substitution that is not to the customer’s liking, they may refund it, which also results in lost revenue and additional work for the online department. Customers are notoriously fussy – particularly over items of varying quality such as produce – and this is considered as a major barrier to online grocery shopping (Otis and Zhao, 2016). Although some supermarkets have introduced algorithmic suggestions for substitutions, workers report they are often ignored and instead personal shoppers intuit what the customer might want. For instance, one worker recalls a berry shortage and their substitution thought process, explaining:

If somebody wants strawberries I try and give them raspberries or mixed fruit. But I was scratching my head last week, the closest I could get was a ‘summer mix’ with bloody pineapple. But I figured they’re probably going to make a smoothie or something, so they want the fruit. It’s really tricky to know, you’ve got to have a bit of imagination. (Personal shopper)

Careful consideration of substitutions takes time and must be weighed up against time pressures and another important KPI – the pick rate.

Pick rates, speed-up and work games

Despite some advantages, the introduction of digital devices to instruct and monitor work is not a neutral force. While useful in some respects, it also deskills and devalues the expertise of longer serving staff and has ushered in explicit monitoring and work speed-up. For instance, both Coles and Woolworths implemented a ‘progress bar’ that is displayed on the tablet or barcode scanner during each run. Much like in a warehouse setting, it delivers real-time feedback on the personal shopper’s speed vis-a-vis KPIs determined by management in the backend. As one worker explains:

It’s a traffic light system. If you’re green you’re picking fast enough, if you’re yellow you need to go faster, if you’re red you’re too slow. (Personal shopper)

Meeting daily productivity targets, commonly referred to as ‘pick rates’ is a KPI typically associated with warehouse work, although the term was also used in relation to cotton picking on plantations (Rosenthal, 2018) and shares some similarities with the piece rate system. For wage-labouring personal shoppers, ‘picking’ refers to retrieving the correct items for a particular order as part of the first steps in the fulfilment process. The ‘rate’ refers to how many items are picked in a given amount of time, typically per hour. Pick rate time pressures are widely reported in

Amazon studies and news articles, often depicted as ‘oppressive, capricious, opaque and mostly unattainable’ (Briken and Taylor, 2018: 453) and part of ‘ritualistic humiliation’ which plays an important role in disciplining workers and speeding up the pace of work (Barr, 2018). It is certainly the case that some firms – such as Amazon – integrate pick rates and algorithmic management to embolden authoritarian managerial regimes. Delfanti’s research with Italian Amazon fulfilment centre workers highlights how the outsourcing of some managerial functions to algorithms transforms the role of the human manager, which becomes more political rather than organizational, taking on a form of ‘augmented despotism’ that is enhanced by the use of technology to surveil and discipline workers (Delfanti, 2021).

Like all workplace technologies, however, pick rates and the technologies that instrumentalize their use are embedded within a particular political economy of work (Winner, 1980). The technology is not universally imposed from above, but is rather a ‘contested terrain’ (Edwards, 1979) influenced by many factors, including managerial regime, union density and historical struggles over the implementation and use of such technologies (Dörflinger et al., 2021). This is not to suggest that pick rates are not specifically designed to monitor and speed-up the pace of work – they are – but that employers approach this aim with different strategies. Feenberg’s (2017) critical approach to technology highlights how ‘technology is not a thing in the ordinary sense of the term, but an “ambivalent” process of development suspended between different possibilities’ (p. 15). While workplace technologies are biased towards the interests of management, realizing these interests requires management to constantly negotiate with workers along a continuum of coercion and legitimation. This is because labour contains a double indeterminacy – management can engage workers for a given period of time (labour-power) but cannot predetermine or fix the amount of effort expended by the worker during this time (labour-effort) and ultimately where workers choose to sell their labour (mobility-power). This structural antagonism gives rise to management strategies to control the labour process more tightly, and subsequent worker resistance (Smith, 2006). Michael Burawaoy’s (1979) seminal work Manufacturing Consent illuminates an important distinction between a firm such as Amazon, which is ‘market despotic’ (favouring coercion) and Coles and Woolworths, which are more akin to a hegemonic regime (favouring legitimation). Rather than apply brute force coercion, workers’ cooperation is manufactured and captured through the combination of ‘bureaucratic techniques (real time control and performance evaluation) with a particular type of corporate culture that depends upon . . . meritocracy and diversity discourse, corporate welfare, and soft authoritarianism’ (Massimo, 2020: 129–130). Here, Burawoy’s (1979) concept of ‘work games’ provides a useful framework for understanding how pick rate legitimation is secured through workplace culture and positive team spirit, a shared commitment to customers and local community (heightened during the pandemic), and a sense of pride and personal responsibility in the performance of essential labour.

Work games tend to develop organically by workers from below, as opposed to strategies of gamification which are imposed from above (Woodcock and Johnson, 2017). Work games make the experience of work more enjoyable, but may foreclose more radical opportunities to agitate for a greater share of profit when ‘winning’ the game becomes an end in itself. As Wood (2021) explains, ‘work games provide a cultural-cognitive control function as they produce consent by focusing workers’ attention away from their exploitation and onto winning these games, with success being measured by profit produced for the employer’ (p. 122). For the personal shopper, work games seem to have developed in response to the time pressures of the role and the social responsibility felt during the pandemic to get essential items into people’s homes. The speed of work is a common theme raised by workers:

You’re under a severe time crunch. You’ve got a certain amount of minutes to complete the round. And you get weekly performance reviews. And I would say it’s the most performance reviewed department in the whole of Woolworths. It’s very intense. (Personal shopper)

Despite this, most personal shoppers viewed pick rates as challenging yet mostly achievable, and a legitimate part of the job, with one explaining:

It depends on your personality how you handle [pressure]. I always like to try and beat the clock. I don’t do it too often but when I do, I always tell the boss and say ‘yay!’ Because it adds a little bit of excitement to what can be pretty mundane. (Personal shopper)

Another described how the progress bar suits their personality, which is very goal oriented:

I personally find it quite motivating . . . I know a lot of people . . . they’ll come back and be like, ‘I was green that whole trolley’ and they were so excited . . . but then there are the other people . . . it never changes from red anyway, so they don’t see it as a goal. (Personal shopper)

An online manager – who also does the work of a personal shopper as required – expressed a genuine desire for their team to have a positive experience of work. As an experiment they reversed the algorithmic picking sequence so that personal shoppers began their pick in room temperature goods (the fastest area to pick). They explained:

It’s seen a massive efficiency improvement with people starting off in the green rather than ending in it. So it keeps them shopping quicker . . . I do try to reward the faster shoppers, but there’s no disincentive for the slower ones. (Personal shopper and Online manager)

Indeed, most workers confirm that pick rates are not enforced in an overtly punitive way, yet at the same time, there is no material reward for working above the pick rate (which some do). Instead, the pace of work is legitimized through work games, managerial style and workplace culture. This process is uneven, however, and on the continuum of coercion and legitimation, some managers do favour more top–down approaches:

[Management] will have their own expectations on how fast they want their team to work in their stores. I know this one in another regional town that says if you’re not picking over 200 articles [per hour] you no longer work in that department.

You’re not a good fit. So off you go, we’ll find someone else to do the 200 articles. (Personal shopper)

Worker cooperation is also uneven, and like any workplace, consent is more readily secured in some workers, while others will resist or challenge methods of legitimation. In my interviews, I found that younger men with management aspirations were more accommodating of the time pressures, whereas workers of longer tenure who recall a time ‘before’ explicit work speed-up did not take part in work games:

I think I’ve learned to just ignore it [the progress bar]. But like I said earlier I’m a permanent employee, so I feel some security and I don’t have to constantly push myself to compete with my co-workers and such. (Personal shopper)

Some workers are antagonistic to methods of work speed-up and voice their discontent.

You’re not a person when you walk in the door, you’re a machine. (Personal shopper)

Coles and Woolworths’ use of algorithmic management, monitoring technologies and pick rates within a hegemonic firm highlight the nuance of different management regimes that must constantly be renegotiated in an effort to secure workers’ consent (O’Neill, 2017).

Asynchronous temporalities of the supermarket

Supermarket workers race to meet performance metrics alongside slow-moving customers, resulting in a site of considerable temporal tensions. Supermarkets were designed to promote a slow and meandering pace whereby customers must traverse the entire store to collect staple items, encouraging impulse purchases and higher sales (Turow, 2017). This runs in direct conflict with the temporal logic of the personal shopper. Customers often complain personal shoppers are a hindrance, and ‘in the way’, suggesting their labour is unbefitting of the retail floor and more suited to after hours, behind closed doors or in a warehouse. Indeed, this was the norm until quite recently. Cochoy’s (2007) influential work on supermarkets explored the ‘two temporally differentiated scenes’ whereby night-fill workers ‘replace and tidy up products while consumers pick and mix them up’ (p. 113). Cochoy explains how these conflicting logics required supermarket labour be hidden from customers and completed outside business hours. He notes, ‘supply is obviously the mirror of demand. But the rule is that neither one nor the other should meet each other directly’ (Cochoy, 2007: 113). Consequently, ‘backend’ labour remained mostly hidden, reinforcing the commodity fetishism of items that appear on supermarket shelves as if by magic, completely disembedded from the messy realities of global supply chains.

The rise of the personal shopper has fragmented Cochoy’s temporal binary and further blurs the distinction between worker and consumer. Here the labour of the personal shopper is not only visible but stands in as proxy for the customer. In fact, some personal shoppers referred to their labour as a form of automation, whereby they automate the labour of the customer through efficiency gains. As one explained:

While [the customer] is walking around the store, sometimes taking up to an hour to do a grocery shop, I’ve done six people’s groceries in less than an hour. (Personal shopper)

This sentiment is consistent with theories of automation that highlight how labour is not eliminated by technology but displaced from one group onto another who may be more compliant or invisibilized (Atanasoski and Vora, 2019). It also echoes earlier developments, such as the introduction of the shopping trolley and self-service, which displaced the labour of the clerk and placed this responsibility with the customer (Grandclément, 2009), with self-service also considered as a form of automation (Cochoy, 2015). Although the customer performs unpaid labour – unlike the personal shopper – both are examples of reconfiguring the supermarket environment to accommodate new technologies that can monitor and accelerate the work of both customers and workers. Decreasing customer foot traffic was considered as an important public health strategy during the pandemic, but some workers are concerned it will legitimize reductions in the labour force. In such an instance the supermarket as a place of leisure would give way to the imperatives of worker efficiency required by on-demand services. One personal shopper explained:

They push their online departments a lot more than what they used to because it takes a lot of foot traffic off the store. We as workers are more efficient than customers to pick. So they can make . . . every department a little bit smaller, and they can make online a bit bigger because we’re more efficient than the customers. (Personal shopper)

The personal shopper must also juggle asynchronous temporal demands generated by on-demand rapid orders that ‘drop in’ on top of home delivery orders that are placed the day before. Rapid orders attract a higher delivery fee, a cap on item quantity and are delivered by gig workers. It is not uncommon for personal shoppers to be picking multiple orders at once, each with their own distinct time constraints. Here, there is a conflict between competing temporal imperatives within the labour process that are both unpredictable and overlapping, resulting in a kind of ‘rhythmic arrhythmia’ (Lefebvre, 2004). As one worker explains:

We get this weird dinging sound [on the barcode scanner] that everyone dreads. You have to pick that order within the half hour or within the hour . . . it can drop in at any time. So if you’re sitting there having lunch for an hour, you still have to go do it because you’ve got that KPI to hit. (Personal shopper)

Once the item is picked the next step is delivery. In what follows, I focus specifically on rapid delivery services that partner with gig workers and last mile delivery platforms. In this model, the personal shopper hands over the items either at the customer service desk of the supermarket, or in the carpark where they place the items directly in the gig worker’s personal car.

Absorption of precarious labour into the supermarket

On-demand rapid grocery delivery facilitated by the gig economy is primarily an ‘urban phenomenon’ requiring population density to capture both customers and workers (Artioli, 2018). As Sadowski explains, diverse cities provide greater opportunities for the platform economy to mediate social relations and extract economic value (Sadowski, 2020). Until recently, slow delivery times have been the Australian norm due to low population density and relatively higher wages that, before the expansion of gig-based digital platforms, rendered last mile delivery too costly to be worthwhile. However the integration of digital labour platforms with the logistical power of the major supermarkets, alongside broader processes of ‘platform urbanism’ (Barns, 2020), has transformed customer expectations. Notably, ultra-rapid services (on-demand delivery in 1 hour or less) are, for now, available only in ‘select’ suburbs of Melbourne and Sydney. These tend to be middle-class areas with above average household incomes and property values. Some personal shoppers living outside these major cities referred to rapid delivery as a distinctly ‘East Coast’ trend that was not important to their local community. One personal shopper light-heartedly commented that:

If you ever want a good laugh, go through the Woolworths community Facebook page. About 50% is people complaining about delivery times, and you see where they are in the country and most of it is eastern states. Victoria and New South Wales. . .I think it’s more of an issue over there. (Personal shopper and Online assistant manager)

While the availability and demand for rapid services is uneven across the country, in places where popularity has surged, the work of the personal shopper must increasingly adhere to the temporal logics of the on-demand economy. For direct-employed traditional employees of the supermarket, gig workers are neither colleagues nor customers, and yet they are essential in the provision of on-demand services. Many workers reported awkward interactions between supermarket workers and gig workers, particularly when a problem arises and there is no clear demarcation of accountability. For example, supermarket workers field complaints for items damaged in transit, and gig workers can be financially penalized via the app when supermarket workers are running behind schedule. This is not to suggest that gig workers are the problem. Much like work games discussed above, the issue is that the resolution of worker tensions at the interpersonal level becomes an end in itself rather than questioning the structural forces that normalize the very reliance on precarious workers to begin with. Given that the vast majority of gig workers are migrants, racialized divisions of labour also remain unquestioned (Gebrial, 2022). Most personal shoppers suggested that on-demand delivery necessitates the use of cheap and precarious gig labour due to very low profit margins, risks and driver insurances, and logistical challenges of last mile delivery. Put simply, it is not feasible or desirable for the supermarket to take on this responsibility.

Such a scenario typically legitimizes outsourcing, or the ‘fissuring’ of the firm whereby non-core business (such as last mile delivery) is outsourced to a company that specializes in this service (Weil, 2017). I suggest, however, that gig workers are not outsourced bur rather subsumed by the supermarket duopoly, enrolling them into an industrial grey area where they are deemed ‘part of the team’ but not a formal employee. This is a form of institutional layering in which new institutional arrangements are added and interact with existing arrangements, leading to significant transformations over time (Streeck and Thelen, 2005). An equivalent example from the local context is Australia’s two-tiered health system which incentivizes the private model through measures, such as tax breaks, in an effort to undermine and erode the public system gradually over time. In this instance, the concern is that precarious labour is legitimized and layered alongside traditional employment security in an effort to erode the latter. One worker raised their concern that the gig economy is ‘getting closer’ and that they felt:

A little dismayed when the whole DoorDashing started because it’s like, oh no, the gig economy is getting closer and closer. Gig stuff always . . . makes me uncomfortable . . . It’s all this whole long-term ploy to destroy some existing industry or place, or eliminate worker protections. (Personal shopper)

Another considered the long-term ramifications, stating that:

My biggest worry is that they start outsourcing the actual shopping procedure. I think that would be the next logical step similar to what America has with Instacart. (Personal shopper)

The process of subsuming precarious labour into a traditional firm is not new. The capitalist mode of production relies upon the integration of diverse forms of labour exploitation, where diversity is central to labour mobilization in global supply chains (Tsing, 2009). As Banaji (2010) emphasizes, ‘capitalist relations of production are compatible with a wide variety of forms of labour, from chattel-slavery, sharecropping, or the domination of casual labour-markets, to the coerced wage-labour peculiar to colonial regimes and, of course, ‘free’ wage-labour’ (p. 359). The supermarket articulates this point quite clearly. To draw generalizations from the Australian context, agricultural work often mobilizes undocumented workers and modern-day slavery practices; meat and fishing industries rely on migrants from low-income countries; delivery and security workers are often international students working within punitive visa restrictions; and warehouse workers skew male and unionized. Put simply, the labour relations of the supermarket are not reducible to a singular form of exploitation. Given the many labour and temporal tensions of the supermarket, the ‘frictionless’ dark store looms large in the imagination of not only the public, but of personal shoppers as well, discussed further below.

Fantasies of the ‘frictionless’ dark store

While ‘dark stores’ have long been a feature of e-commerce, the pandemic heightened consumer interest in supply chain visibility, fueling media interest in sites of logistics. A proliferation of media stories depicted these small warehouses, or micro-fulfilment centres, as regular supermarkets closed to the public and repurposed for personal shoppers and home delivery (e.g. see Kushner and Lindsay, 2021). This seemingly contradictory space has captured the public imagination because it is both familiar yet elusive. Dark stores are typically located within or near residential areas to facilitate speedy home delivery and mitigate ‘last mile’ delivery costs which are considered as the most expensive leg of logistics (Alimahomed-Wilson and Ellen Reese 2020). Shapiro (2022) considers how ‘going dark’ may negatively impact urban environments and public life. Others have cautioned that the proliferation of dark stores will transform ‘downtowns into dark cities’ (Kushner and Lindsay, 2021).

This analysis is particularly important in the context of pandemic-fueled retail closures and the subsequent billions of dollars in venture capital (VC) that rushed to fill this vacuum. Start-ups promising rapid grocery delivery services surged during the pandemic and attracted significant media interest despite many failing in their first year (Pardes, 2022). In Australia, four rapid grocery delivery start-ups emerged amid the pandemic: SEND, Quicko, Voly and Milkrun. All except Milkrun have since failed, although even this much-feted start-up remains unprofitable, has abandoned its central value proposition of rapid delivery and has admitted to burning through much of the US$75 million of VC funding it secured in early 2022 (Bonyhady, 2022a). In the current post-quantitative easing environment of rising interest rates, high inflation, and decreased discretionary spending, it is unlikely further rounds of VC will continue to prop up the failing start-up. These fledgling apps were always more hype than reality and as such their failure does not pose a significant disruption to Australian urban geographies. Zooming out from the start-up space, however, I consider the impact of more enduring models of micro-fulfilment. Woolworths provides a clear example of last mile complexity and how this challenges the more narrow, popular usage of the term dark store.

Understanding online grocery to be of strategic importance, both Coles and Woolworths began expanding their e-commerce capabilities before the pandemic, with each adopting different strategies in this ‘online arms race’ (Mitchell, 2019). Coles is prioritizing large fulfilment centres in partnership with global technology companies such as Ocado, whereas Woolworths is investing heavily in micro-fulfilment and partnerships with the gig economy. During the height of pandemic lockdowns, Woolworths temporarily converted three retail stores to ‘pop up’ dark stores to meet surging online demand, all of which have since reverted to regular retail spaces (Woolworths Group, 2020b). With this exception, Woolworths tends to leverage its considerable retail presence by repurposing existing bricks and-mortar stores to fulfill hybrid functions that include both retail and e-commerce. That is, Woolworths dark stores are not always dark, per se.

In mid-2022, Woolworths launched the Metro60 app which promises 1-hour home delivery to select suburbs in Sydney and Melbourne, for a flat delivery fee of AU$5 (Bonyhady, 2022b). These orders are fulfilled from existing Woolworths ‘Metro’ stores, which are smaller urban and inner-suburban supermarkets that carry basic grocery items as well as convenience goods and alcohol. This means that last mile delivery partners, such as Uber drivers, interface with a public retail store, often lining up alongside regular customers. Similarly, in 2019 Woolworths partnered with Boston-based tech company Takeoff Technologies to develop three hybrid sites dedicated to e-commerce. To achieve this, a section of a retail store’s footprint is cannibalized and transformed into a small fulfilment centre with tall and narrow vertical racking that is concealed from public view (Woolworths Group, 2019). Although this section of the store is ‘dark’, it is a transformation of Woolworths’ existing real estate rather than a new site. More recently, Woolworths launched yet another model of dark store that functions as a fulfilment centre but with an outdoor public area for customers to collect their on-demand purchases. The Australian Financial Review reported that Woolworths now ‘brings customers straight to its warehouse’ (Schlesinger, 2022). Although customers cannot enter the warehouse itself, the public dispatch and pick-up area is visibly branded and staffed with supermarket workers who engage with customers directly. In addition, Woolworths has an extensive network of dedicated fulfilment centres which the company refers to as ‘online warehouses’, from which small, company branded trucks collect orders for home delivery. Although the model of dark store differs somewhat architecturally, the labour process of picking is relatively standard across each site.

Here, it becomes clear that media interest in dark stores is coloured by the start-up experience, but this does not align with the reality of a logistics giant like Woolworths. More surprisingly, however, is that personal shoppers also have a somewhat distorted view of the dark store, and yet these are people working for the same employer. In my interviews with personal shoppers, the dark store looms large as a completely ‘frictionless’ site of efficiency and automation, giving rise to both myth and anxiety. Many remarked that as demand for online services continue to grow, the ‘messiness’ of the supermarket would no longer provide a sustainable place to pick from, and that this work will likely move to a dedicated dark store. For some this was a welcomed prospect, with one worker commenting:

I quite love dark stores. I wish there was one in [my area]. (Personal shopper and assistant manager)

Most personal workers assumed the dark stores are automated or at the cutting edge of warehousing technology:

From my understanding dark stores are very streamlined. Peeking at some internal reports, the pick rates that I see when looking at those stores, I’m guessing that they have some very streamlined processes in those locations. (Personal shopper and Online manager)

Interestingly, one personal shopper expressed some relief that the temporal frictions of the supermarket may provide necessary protection from a perceived relentless pick rate in the dark stores:

I think it’s good thing this job is done in an open supermarket because then if I’m running a bit slow or a bit hampered or something, I can say the reason why I didn’t meet targets was because there was a lot of customers asking questions or getting in my way. And I feel this is a blessing because I imagine if I was in a dark store or something, they’ll probably be stressing the targets a bit harder. (Personal shopper)

Interviews with dark store workers tell a different story. They describe a very manual and physically demanding job further challenged by high temperatures in the summer. The pace of work is fast, but similar to retail workers, dark store workers claim pick rates are not enforced in an overtly punitive way. With some exceptions, management tends to incentivize work speed-up through a culture of teamwork. Dark store workers also commented on the lack of sophisticated technologies that are essential to their role – the barcode scanner, refrigeration, pallets and forklifts remain foundational to what is essentially warehouse work. Furthermore, some cited the necessity of human decision-making in product substitutions as a possible barrier to automation:

They’re not automated by robots as of yet. Just very much manual labour, in a sense, everything does require a human to run behind it. There would have to be a pretty good robot. They’d have to be able to pick out good produce and sort damaged items from items that are not damaged. They would have to be able to produce quality produce for customers . . . which I think people can do better than AI at the moment without it being super high tech and expensive. (Dark store worker)

Another adds:

Everything was very manual there, the only time I heard anything of any sort of automation was they were going to try and automate the night pick. They put up a notice in the lunchroom that they were testing out new technologies. That went up three years ago, and I never saw or heard anything about it after that. (Dark store worker)

Dark stores are often imagined as a novel configuration of technological and labour conditions, the ‘sublime’ of which has captured public attention (Nye, 1994). The strong focus on technologies that to a large extent, do not yet exist, has overshadowed the very real yet mundane grievances of ‘dark jobs’ – wages and conditions.

Worker representation and the partner union model

The perception of dark stores as novel retail spaces has allowed the supermarket duopoly to pay dark store workers retail wages, instead of warehouse wages. In Australia, this is significant because the two industries are covered by different legal standards that outline the minimum pay rates and conditions of employment (the Award system) and are represented by different unions. Industrially, the consequences of this are multifaceted, but there is one fundamental point to emphasize – warehouse wages tend to be higher than retail wages because of differences in union approach.

The union for retail workers, the Shop, Distributive and Allied Employees Association (SDA), adopts a partnership model (Forsyth, 2022) in which union-management relationships are prioritized above the interests of workers. The SDA fosters a particularly close relationship with the supermarket duopoly (remembering these are the biggest private sector employers in the country), which generates significant revenue for the union via dues. Within Australia’s union movement it is widely acknowledged that in exchange for unfettered access to new employees and payroll deduction of dues (meaning the worker may not even be aware they are due-paying union member), the SDA will quell membership grievances and not take industrial action against a partner employer. Furthermore, in recent years it has come to light that SDA supermarket enterprise agreements undermine legal minimum standards, meaning that thousands of supermarket workers would be ‘better off overall’ if the union agreement did not exist (Schneiders et al., 2016). The enterprise agreement for Woolworths dark store workers is particularly egregious and pays a full-time worker as much as US$3000 less a year than a Woolworths worker in a regular store (Schneiders and Millar, 2017). Total underpayments across Coles and Woolworths are estimated to have exceeded US$1 billion AUD in wages since 2012 (Schneiders, 2020). The partnership model has allowed the SDA to establish close connections with employers who are notoriously anti-union – such as McDonalds – and this is why the SDA is generally considered as an outlier in the Australian union movement (Schneiders, 2022).

Alternatively, warehouse workers are represented by UWU which adopts an industrial model with a focus on worker organizing and industrial action such as strikes. These members work in Coles and Woolworths distribution centres which replenish retail stores, rather than the dark stores or fulfilment centres which service home delivery. The UWU negotiates ‘comprehensive’ enterprise agreements which build upon the minimum legal conditions to include higher wages and other conditions such as overtime, shift provisions and additional leave provisions. UWU takes regular industrial action, such as in 2020 when Woolworths warehouse workers went on strike to secure above-inflation wage increases and changes to pick rates (Bonyhady, 2020). While retail workers are paid an average minimum of US$23/hour AUD, industry standard warehouse wages are approximately US$34/hour (not including further leave and superannuation entitlements). Simplistic media narratives of dark stores have aided the supermarket duopoly in classifying dark store ‘online pickers’ as retail workers to avoid paying higher union wages and conditions, and to maintain a workforce that is, in practice, unorganized. Since finalizing an earlier version of this article, Woolworths has applied to Australia’s Fair Work Commission to legislate changes that would remove any ambiguity and ensure dark stores and online fulfilment centres are covered by the retail union; the SDA (Marin-Guzman, 2023). Both SDA and Coles have supported Woolworth’s application, with significant ramifications for the entire retail sector, as many businesses are expected to move further into e-commerce.

This discrepancy in wages between retail and warehousing was raised by a dark store worker:

When I started talking to [other dark store workers] there was this sentiment of – why are we not getting paid warehouse wages? Whenever I see reporting on dark sores, people talk about it as if they’re just working in a store but no one’s there, as though it’s this quirky thing. But I think what people misunderstand is that these people are doing pretty intense manual labour at times, very intense warehouse work. It’s a warehouse that tries to give off the illusion of a store. I think it’s a marketing ploy to make it seem more wholesome than it really is. (Dark store worker)

Consideration of Woolworth’s actually existing dark stores reveals diverse nodes of logistics that are not necessarily disrupting urban geographies or innovating cutting-edge technologies. Instead, the myth of the dark store becomes significant for what it conceals – the misclassification of ‘dark jobs’ resulting in downward pressure on wages and conditions. As Mosco (2004) explains, the political economic context that renders myths powerful is multidimensional and includes ‘visionaries promising electronic utopia, the mass media looking for a good story, politicians wanting to be identified with the next new thing, and businesses . . . eager to market the latest promise to transform life as we know it’ (p. 118). In the case of on-demand grocery delivery, the reality is banal yet exceedingly more materially meaningful than popular, hype-filled media narratives of technological innovation.

Conclusion: the human cost of convenience

As the COVID-19 pandemic and subsequent lockdowns began to spread throughout the country, many Australian households began doing something they had never done before: ordering groceries online. Despite some initial stumbling blocks, Australia’s supermarket duopoly accelerated significant transformations to meet the surging demand. For many supermarket workers, the rise of ‘dark jobs’ has radically changed the experience of work with consequences which will endure long after the pandemic. The personal shopper has become near ubiquitous as online departments swell. Their work is mediated by algorithmic instructions, pick rates and temporal imperatives imposed by the on-demand platform economy. Interactions with gig workers generate not only asynchronous temporalities in the workplace, but also concerns that existing secure employment may be undermined by these new partnerships. Beyond the start-up space, the diversity of supermarket dark stores reveals the complexity of last mile logistics. Rather than technological or geographic disruption, however, the key issue at hand may simply be that the myth of the dark store as a novel retail space is obscuring the material reality of dark store work, and applying downward pressure on wages and conditions. Taken together, the supermarket becomes a powerful testing ground in which the logic of on-demand platforms is applied to conventional employment settings, further blurring the distinction between employment and gig work.

Efforts to critique the ways in which essential goods are distributed is often met with defensiveness. Typically, vulnerable groups such as the elderly and people living with disabilities are marshalled – willingly or otherwise – by industry representatives as justification that the ends justify the means. At the beginning of the pandemic Woolworths media statements were explicit in this claim, stating a need to accelerate online delivery capacity to ‘support elderly and vulnerable Australians in self-isolation’ (Woolworths Group, 2020a). To put it simply, one may accept that rapid delivery services are eroding secure jobs, potentially affecting hundreds of thousands of workers across the country, but this is necessary, inevitable and perhaps even desirable. This argument can be challenged on two fronts. First, if it is the case that rapid grocery delivery is essential, then this provides further justification for the protection, and indeed advancement, of workers performing such important work. It is certainly true that in the absence of comprehensive social services, many vulnerable communities rely upon home delivery services for essential goods. Addressing these shortcomings is a matter for the state and should not be entrusted to unscrupulous digital platforms to plug the many holes of a city hollowed-out by privatization. Second, it is not entirely convincing that rapid grocery services, in their current form at least, are in fact essential. Personal shoppers also express some doubt in this claim and highlight a correlation between how quickly something is to be delivered, and the frivolity of the purchase. For instance:

No one used to do it. Now, people buy only five things and they’ll pay that fee to have it delivered soon. It’s more popular for alcohol or cigarettes or something like that. (Personal shopper)

Another adds:

People are ordering . . . a single banana and a Red Bull. It’s really weird the stuff you get. (Personal shopper)

One personal shopper was deeply sceptical of the move to rapid delivery, recalling how partnerships with the gig economy predate the pandemic, with the crisis merely providing an opportunity to manufacture demand for an existing product:

It didn’t seem like it was about meeting the demands of shoppers, that’s made explicit through the article cap for UberEATS . . . you can only order 25 [items] so it wasn’t about regular shopping. Really, I think it was just more for the convenience. Instead of going to the shops yourself, you can just wait at home for it, and someone else can pick it for you. (Personal shopper)

The supermarket has always been a site of contested social and labour relations (Deutsch, 2010), and the human cost of convenience is one such tension that will continue to reverberate. This is because, ‘grocery stores are more than just places to buy food. They are in a broader sense a reflection of our culture . . . and thus a gauge of who we are, supermarkets illuminate what we care about, what we fear, what we desire’ (Ruhlman, 2017: 9). Although reified social institutions such as the supermarket appear natural and fixed, they are always changing and being renegotiated in important ways. As Coles and Woolworths seemingly break with the hegemonic paradigm and move towards something potentially despotic, the issue of duopoly power comes into sharp relief. There is no other store that requires patronage every week and takes a greater percentage of the working family’s total weekly budget (Ruhlman, 2017). Unlike other retail giants such as Amazon, most citizens cannot opt-out from or boycott the major supermarkets, because alternative independent options are unavailable, unaffordable or inadequate. It is often argued that the effect of monopoly capitalism is the stifling of competition among firms. It is more accurate to contend that the supermarket duopoly strengthens cooperation among supermarkets and large employers, the gig economy and partner unions, while intensifying competition between groups of workers. Consumers may for now enjoy the low cost and high convenience of rapid grocery delivery, but the consequences warrant greater scrutiny.