Researching Care Platforms: Methodological and Ethical Considerations in the Broad Field of Domestic Platform Labour

Abstract

This article presents a methodological reflection on the challenges of researching domestic and care work mediated by digital labour platforms. While knowledge production on gig work in the logistics sector has soared, research on care platforms is slow to catch up, especially in very relational forms of home-based work, such as eldercare. We take this uneven development in the literature as a starting point to unpack the empirical conundrums in this field. Drawing on our own experience with trying to recruit care platform workers in Germany between 2019 and 2024, we shed light on the ethical dilemmas we encountered and offer some lessons learnt. The article calls for long-term commitments, multi-sited ethnographies, longitudinal perspectives and mixed-methods designs to study care platforms in the future. Finally, we advocate for researching platform labour beyond the gig.

Introduction

In view of the vast knowledge production on algorithmic management and the experience of gig workers in ride-hailing and logistics work, feminist (labour) researchers across disciplines have decried a dearth of knowledge production on the ‘gigification’ of the home (Flanagan, 2019; Huws, 2019; Kampouri, 2022; Surie and Huws, 2023). Even though the first studies on domestic work platforms date as far back as the earliest work on Uber (see, e.g. Hunt and Machingura, 2016; Schmidt and Kathmann, 2017), research on platformised reproductive services has developed much more slowly. Over the years, this disparity in knowledge of these different sectors within the gig economy has largely remained stable as have the continued calls and pleas to unpack the digital mediation of care and domestic work.

Eager to heed this call, we designed research projects to investigate care platform labour in Germany between 2019 and 2024. In this paper, we reflect on ethical and methodological pitfalls we have encountered in this particular research field. Our understanding of care platform research also builds on insights gained from many conversations with our peers in the field. Learning that our struggles were structural rather than individual inspired us to write this methodological reflection in the hope that our experiences can provide valuable food for thought to others embarking on comparable projects.

Many platform companies and the technologies they employ operate globally, yet platformisation is always historically and geographically specific (Gebrial, 2024). For example, employing domestic servants is still relatively common in some postcolonial societies (Hunt and Machingura, 2016; Nguyen et al., 2024) and thus embedded in historical and regional trajectories that are different from the Western European context we are writing from. Similarly, in contexts without a welfare state regime, such as the United States, the platformisation of eldercare likely functions differently and is monetised in different ways than in Germany as we will explain in detail when we introduce our empirical setting. We offer our lessons learnt with these caveats, knowing that our suggestions may have to be adapted to specific contexts.

Throughout this paper, we advocate for, and follow, well-established principles of feminist and gender research, namely grounding our research in, and understanding it from the point of view of marginalised groups and labour positions (Bailey, 2012; Dupuis et al., 2022). If taken seriously, this approach questions not only the modes and gigification of research participation (McKenzie, 2024) but also what researchers focus on, what they consider as ‘work’, and how they conceptualise research projects.

First, we provide a method-centred review of the platform labour literature at large to present an overview of the main strategies for recruiting interview partners. We then hone in on the subfield of platformised reproductive work and its specific practical and ethical challenges. We offer a critical appraisal of the research designs employed to navigate the issues we see as inherent in the field. Drawing on our own experience of empirical data collection in three research projects conducted in several German cities between 2019 and 2024, we then focus on one particular approach; namely posing as clients on eldercare platforms to recruit workers. Reflecting critically on our use of this method, we offer several lessons learnt and recommend alternative methods and research designs for future research. Namely, the use of follow-the-people/follow-the-subject approaches, employing biographical and repeated interviewing, as well as the use of mixed-methods that situate platform work in larger life and discursive contexts. We conclude with a call to engage with feminist methodologies to help us decentre the platform and instead centre the care environments that shape this work stretching from social welfare to immigration regimes.

Methodological Challenges in Platform Labour Research

Platform-mediated labour tends to be spatially dispersed and is usually short-term, often paid one gig at a time. This fragmentation of the workforce means workers often do not know each other, making snowball sampling of interview partners difficult. Digital labour platforms usually have no centrally located company infrastructure and no communication channels for workers. This renders a classic industrial relations case study approach where researchers can enter the field through key gatekeepers like union representatives or HR departments obsolete (Iphofen et al., 2022). Using the digital interface of the platform to recruit workers is also ethically fraught due to the omnipresent surveillance of workers on platforms. Using platform interfaces for recruitment carries a high risk of inadvertently de-anonymising and thus exposing critical workers to company retaliation (see, e.g. Ivanova et al., 2018; Ravenelle, 2019). Overall, there is thus a broad consensus among platform labour scholars that it is necessary for ethical reasons to find alternative field access strategies (McDonald et al., 2024; Ustek-Spilda et al., 2022b).

First and foremost, researchers resort to accessing workers via important online and offline infrastructural points such as transport hubs, warehouses and city districts with a high density of restaurants and dark kitchens in food delivery (see, e.g. Iazzolino and Varesio, 2023; Raval and Lalvani, 2022). Large-scale research projects with advertising budgets have also paid for ads in digital and public spaces to recruit workers (Ustek-Spilda et al., 2022b). In addition, researchers have booked workers for a gig and then interviewed them on the spot once they had arrived for their job. This technique is common in studies on cleaning platforms (e.g. Gerold et al., 2022; Gruszka et al., 2022) and on ride-hailing platforms (e.g. Holtum et al., 2022; Kuttler, 2023).

As workers have built digital organising infrastructures and mutual aid groups on social media sites and apps such as Reddit, Facebook or WhatsApp and their local variations, researchers have also tapped into these to find interview partners (see, e.g. Niebler and Animento, 2023; Tandon and Sekharan, 2022; van Doorn, 2023). While interviews with workers remain the main qualitative approach to researching the gig economy, co-creating research with workers as participatory action research (see, e.g. Heiland, 2020a; Leonardi et al., 2019) and conducting auto-ethnographic experiments as delivery workers and riders (see, e.g. Richardson, 2020; van Doorn and Vijay, 2024) are also common ways to understand the experience of workers and newly emerging labour processes.

However, as many researchers acknowledge, the outlined research approaches are of limited use when attempting to research platform workers who work inside private homes. Domestic workers neither have branded bikes and equipment nor designated meeting places such as restaurants or popular passenger pick-up sites that would make them recognisable as platform domestic workers to researchers or each other (see methods sections, e.g. Baum and Kufner, 2021; Ticona et al., 2018; van Doorn, 2023). Reproductive work has always been invisibilised and happened in relative isolation without a shop floor. Thus, workers are even less likely to know each other, making snowball sampling much more difficult. Relatedly, there are also fewer attempts to create mutual aid groups and organising drives in this section of the gig economy (Niebler and Animento, 2023) that could be used as an entry point for participatory action research or simply to recruit interview partners.

These additional hurdles in the recruitment of platform workers offering reproductive services result in considerable difficulties in generating interviews in this sector. In light of the small number of interviews that specifically detail the experience of care and domestic workers, delivery and ride-hailing are often used as proxies to theorise platform capitalism’s impact on labour relations. Yet, as digitalisation (re)shapes many sectors with different consequences for differently positioned workers and differently constructed industries, we need to be cautious with this shortcut for several reasons:

First, the gender makeup of workers within the platform economy reflects that of non-platformised industries. Mirroring the wider transport and logistics industries, significantly more men than women work in platform-mediated ride-hailing and food delivery (see, e.g. Holtum et al., 2022; Iazzolino and Varesio, 2023; Zhou, 2022). Similarly, domestic work on platforms is predominantly carried out by women (see, e.g. Kampouri, 2022; Pulignano et al., 2023; Rodríguez-Modroño et al., 2022; Tandon and Rathi, 2021). These worker demographics combined with the imbalance of what types of platforms are being researched thus likely produce gendered data gaps. Second, the industry or sector a platform emerges from plays an important role in how a platform can restructure or influence a field of labour (Au-Yeung and Qiu, 2022; Seibt, 2024). As a result, different sectors produce very different platform models that in turn result in different labour processes and forms of control (Hunt and Machingura, 2016; Ticona et al., 2018). For example, empirical findings that are built on research with platforms that primarily rely on geodata – such as the proximity of workers to restaurants and clients – are difficult to compare with marketplace platforms based on worker profiles that more closely resemble directories (Ticona et al., 2018; Ticona and Mateescu, 2018). In other words, when we assume a universal gig worker experience, we risk reproducing an implicitly male view of labour conditions.

To avoid these biases, we need to study a variety of platforms and workers. Creating wider empirical foundations for our understanding of the impact of platform mediation on labour processes and worker experiences creates both more robust academic knowledge as well as better labour policies and regulations. In the following section, we therefore hone in on studies that have tried to tackle the research field of domestic platform work in particular. We highlight both the advantages of these study designs as well as the areas that require further development.

Research Approaches in Domestic and Care Platform Work

When domestic and care platforms emerged, researchers first had to understand the particularities of domestic and care platforms. To this end, early care platform studies resorted to analysing publicly available information. For example, analysing a platform’s contractual terms and conditions or website (McDonald et al., 2021). Similarly, researchers conducted ‘technical walkthroughs’ to understand the designs, interfaces and operation of care platforms (Gruszka et al., 2022). Often, studies combined this method with a small number of explorative interviews (Hunt and Machingura, 2016; Ticona et al., 2018). As previously outlined, researchers soon found that interviews with care platform workers are extremely difficult to obtain forcing them to explore alternatives to qualitative worker interviews. Broadly speaking, we find four groups of research designs researchers have tried out to understand the labour process at these platforms without speaking to a large number of workers. These include the use of auto-ethnographies, co-researching with platform companies, aggregating interview data across a range of reproductive and care services, and research designs that triangulate a smaller number of interviews with other methods. Recognising the heightened challenges of care and domestic platform research, we now offer a critical appraisal of these studies, their contributions, and remaining gaps.

We start with auto-ethnography since auto-ethnographies are very prevalent in food delivery research projects. The strength of auto-ethnographic research is the ability to observe the affective dimensions of platform work and to gain a firsthand understanding of the labour process (Keller, 2022, 2023). In contrast to the popularity of this method in food delivery research, there are, to our knowledge, only three researchers to date that have attempted an auto-ethnography in the care and domestic sector. Bor (2021) and Keller (2022) have conducted auto-ethnographic work with cleaning platforms and Brauchli (2019) worked as an assistant to a person with a disability as well as two elderly clients contacted via care.com. These researchers describe this approach as particularly challenging for various reasons: all three note their discomfort in trying to approximate the experience of a care worker given their own positionalities as white citizens of the countries where they conducted their research (Keller, 2022: 144). In addition, Brauchli (2019: 69) raises an important ethical concern that is specific to more relational forms of care: onboarding new assistants is emotionally stressful for people needing care. Consequently, clients often prefer a long-term relationship with their carers. This need is at odds with much shorter-term timescopes of research projects. In other words, researchers have to balance a commitment to and relationship-building with elderly or disabled research participants with the limitations they face as researchers needing to finish a project within a limited timeframe. This ethical and practical conundrum likely explains why auto-ethnographies remain rare in the domestic and care platform research field.

A second, and by comparison more common, approach is to tackle the problem of finding interview partners by collaborating with a particular company to facilitate access to workers (Hunt et al., 2019; McDonald et al., 2024; Molitor et al., 2021; Sibiya and Du Toit, 2022). This is a feasible approach as some care platforms are ready to work with researchers, especially in jurisdictions where data protection regulations prevent them from collecting their own user data. In some cases, researchers have even received and analysed quantitative data from the platforms (Hunt et al., 2019; Molitor et al., 2021; Pereyra et al., 2022; Sibiya and Du Toit, 2022). Methodologically, co-research with a platform, however, risks narrowing the research field to those platform companies which welcome researchers. Moreover, this sampling strategy could also inadvertently result in a bias towards satisfied workers as more critical workers may not trust any research the platform facilitates through corporate communication channels (Cornet et al., 2022; Molitor et al., 2021). Lastly, it may be difficult to reach workers engaging in multi-apping and those who only selectively check the platform communication channels. In addition, from an ethical point of view using platforms as a field entry point always entails the risk of de-anonymising workers. Therefore, it remains an approach researchers always need to approach with great caution.

A third very prevalent research design is to work with very broad definitions of ‘reproductive work’ and ‘care work’. Especially large-scale studies tend to group data on any tasks related to the household together. This allows researchers to create samples with enough interviews to compare and contrast care platform dynamics with the operational models of other types of platforms (Rizk et al., 2022; Ustek-Spilda et al., 2022a; Uysal, 2023). In addition, this broad definition allows researchers to incorporate a wide range of household-related tasks from grocery delivery and home repairs to cleaning, tutoring and eldercare to aggregate a smaller number of interviews into larger datasets (e.g. Pulignano et al., 2023). Another similar approach involves sampling specifically for interactional tasks, but interviewing workers from very different care occupations ranging from tutors and domestic workers with household tasks to eldercarers. This also enables studies to pool interview data across the spectrum of work inside or related to the home (e.g. McDonald et al., 2024). While these studies generate insights on the differences in platforms’ operational models and manage to broaden the demographics of gig workers sampled, they do compare two or more vastly different types of work, such as cleaning and eldercare that exist on opposite ends of the spectrum of relational care.

As all of the outlined approaches carry some ethical dilemmas and methodological problems, many researchers follow a fourth approach by combining and triangulating several of the outlined methodologies. Researchers, for example, opt to supplement a smaller number of worker interviews with additional elite interviews, such as interviews with white-collar workers at platform companies or union representatives from specific fields. Triangulating a rather small number of worker interviews with other methods can centre workers’ voices while also validating relatively small interview sample sizes (Floros and Bak Jørgensen, 2023: 10; Hunt and Machingura, 2016: 18; Orth, 2024: 479; Tandon and Rathi, 2023: 218; Tandon and Sekharan, 2022: 692). Using these different approaches to researching domestic and care platforms despite the outlined challenges, researchers have moved the field forward considerably.

To summarise their findings briefly, care and domestic platforms differ from ride-hailing and logistics in three important ways. First and foremost, relational work requires building trust (Ticona and Mateescu, 2018), which often entails a much more long-term scope and thus goes against the gig economy model of extreme flexibility (Baum et al., 2020; McDonald et al., 2021: 885; Molitor et al., 2021: ii). Second, while proximity and thus geodata are the decisive variables in ‘last mile’ logistics work, domestic work platforms function almost exclusively via marketplace and sorting logics (McDonald et al., 2021; Mateescu and Ticona, 2020; Rodríguez-Modroño et al., 2023, 2024; Tandon and Rathi, 2021; Ticona and Mateescu, 2018; Ustek-Spilda et al., 2022a; Williams et al., 2021). Third, to fullfill these trust requirements platforms prioritise detailed worker profiles with many visual elements akin to social media networking sites (Ticona et al., 2018; Mateescu and Ticona, 2020) or cloud work platforms (Williams et al., 2021). The commodification of a worker’s personality transverse digital and material space as good online representation is a prerequisite for getting a job but negative interactions ‘in real life’ can manifest in negative reviews on the profile (Flanagan, 2019). This leads to a paradoxical situation of extreme visibility online that stands in contrast with extreme invisibility at the workplace (Gruszka and Böhm, 2022; Mateescu and Ticona, 2020). These differences in how platforms function affect the language, literacy and digital skills requirements a worker is confronted with and make digitally mediated care work less accessible than platform work commonly is believed to be (Orth, 2024).

Despite these important achievements in understanding different platform operational models, labour processes and the specific experience of a largely female workforce, we find that most studies thus far do not specify what kind of tasks research participants do via care and domestic platforms. Acknowledging this problem, some recent studies have begun to include overviews of their datasets (Pulignano et al., 2023; Rodríguez-Modroño et al., 2024). These overviews corroborate an impression we have had for a few years, namely that most of our empirical care and domestic platform knowledge is based on interviews with childcare workers and workers carrying out comparatively non-relational reproductive tasks. While different types of domestic and care work do share common aspects such as invisibility and spatial dispersion of the workforce, they also differ in important ways. For example, how short- or long-term the commitment to a job is and how close the client-worker relationship will be diverge significantly between grocery shopping, cleaning, meal preparation, and pet sitting at the less relational end of the spectrum and the more relational end of the care spectrum that includes beauty work, childcare, disability care and eldercare. Just as it is problematic to take ride-hailing and food delivery as proxies to understand the gig economy at large, it is inaccurate to extrapolate from cleaning work to elder care. This matters particularly in our contexts, where platforms present themselves as a technological fix to the mounting care crisis posed by ageing populations. If most studies are thus far based on childcare and cleaning work, it is hard to gauge the relevance of platforms for addressing projected care gaps in eldercare.

Therefore, we wanted to explore the role of platforms specifically for eldercare. In the following sections, we reflect on our research projects carried out in Germany between 2019 and 2024. We particularly focus on the recruitment of eldercare platform workers and discuss the creation of online client profiles to contact these workers. A reflection on this method strikes us as timely because it is widely used across the care platform research field but rarely openly discussed or reflected upon in publications. While we primarily draw on our own experiences over the last five years to highlight ethical and methodological pitfalls, our analysis and recommendations are also informed by many informal conversations with fellow care platform researchers who have faced very similar obstacles. In outlining and critically discussing these questions, we hope to contribute to a conversation on how to better design research in this field in the future.

Notes From the Field: The Specific Case of Digitally Mediated Eldercare

Before diving into our specific experiences, we want to briefly outline the context in which our research takes place. Unlike the early studies carried out in the United States and contemporary studies on domestic work in India (Nguyen et al., 2024; Tandon and Rathi, 2021), South Africa (Hunt et al., 2019; Hunt and Machingura, 2016; Sibiya and Du Toit, 2022) and Latin America (Andrada et al., 2023; Pereyra et al., 2022) where there is little to no state-funded welfare system, eldercare – at least in a Western European context – remains partially publicly funded. In Germany, the government runs a means-tested mandatory long-term care insurance scheme (Pflegeversicherung). Once someone has passed the test required to prove they need care, insurance authorities allocate funding. ¹ Similar schemes exist for subsidising temporary rehabilitation care after illness and for the disabled. Concerning childcare, national legislation covers up to 14 months of paid parental leave, and, depending on regional policy, childcare is either entirely free or subsidised for toddlers for between 20 and 60 hours a week.

This difference in the welfare state regime is pivotal in how care platforms function in the context of European welfare states, where the European Union has enshrined the right to access and quality of long-term care (European Commission, Directorate-General for Employment and Social Affairs and Inclusion, 2021). In this context, care platforms are part of a larger shift from public to private funding (Arcidiacono et al., 2022; van Doorn et al., 2021). In Germanz, care platforms are not positioned as yet another of many privatised care agencies but aim to substitute care where existing systems fall short. Consequently, companies do not primarily target private incomes as seems to be the case in the United States and most non-industrialised countries. Instead, care platform business models revolve around accessing and claiming state insurance scheme money, especially through so-called ‘cash-for-care’ schemes (Baum, 2024b; see also MacDonald 2021). In a (declining but still existent) welfare state regime with public–private partnership insurance schemes that pay for some types of care, the business model of care platforms is to try and enter the cracks in the welfare system created by three decades of privatisation.

To understand how this particular operational model affects workers, we began fieldwork in 2019. To generate interviews, we planned to use the platform interface to advertise for interviews but quickly reached the first hurdle: platform companies considered advertisements for research interviews a violation of their terms and conditions and quickly deleted our posts. Being open about the intent of the ad was thus not possible, and the only way to recruit within the platform interface was to register as a client and pose as someone looking for a carer. While posing as a client risks blurring the boundaries between our identities as researchers and as platform clients paying for a service, able to rate and review a worker (Huws et al., 2017: 55), there is an established practice to use this as means to establish first contact with workers. To ensure informed consent, many researchers opt to pay workers for showing up for this first ‘gig’, and use this first offline encounter to explain the aims of their research and offer the opportunity to schedule a separate follow-up appointment for a research interview (see, e.g. Gerold et al., 2022: 17; Ravenelle, 2017: 285).

Registering as a client on the care platforms we researched quickly led to a second conundrum as client profiles require very detailed information on the person in need of care, including the specific hours the client needs attending to, their age and their insurance status. As the need to establish trust and foster a mutual relationship is inscribed in business models of eldercare platforms (Baum and Kufner, 2021), providing this information was mandatory and compelled us to come up with a pseudonym with a full name, complete address, birth date and insurance details. Furthermore, platforms forced us to create an ad that specified why the person needs help, the minimum amount of hours a carer is needed for as well as specific days of the week and a prospective starting date. One platform did not even give us the option to advertise a one-off gig as the programmers had only anticipated long-term arrangements. Owing to these platform designs, both authors had to create a whole imaginary person to be able to fill in the platform’s profile. While we considered this not ideal from the start, we initially assumed that revealing our true intentions once the initial contact had been made would suffice. Yet, as the following examples will show, we learnt that this commonly used approach can quickly run into problems.

During fieldwork in 2021, Author 1 (Barbara Orth) had set up a client profile and a couple of hours after the ad had gone live, an unexpected call came through late in the evening. A woman who had seen the ad was on the phone demanding to speak to her pseudonym. A field diary entry illustrates this encounter:

Even though the woman spoke German natively, our communication was difficult. For twenty minutes, she kept reiterating her demand to speak to the potential client looking for an eldercarer. I tried to explain again and again, that this was not the actual job; that I could not refer her on to this person, [my pseudonym] because they did not exist. In my efforts to try to explain what I was interested in learning about, I kept asking her about other clients but she seemed not to have any. Me asking about it made her even more frustrated, which made me think she did not have any existing clients, and was actually desperate for a job. It was apparent that she had invested some hope in being able to secure a long-term gig through my ad, and was unwilling (or unable?) to believe that this would not materialise. [. . .] I don’t think she understood what I meant by ‘research interview’. (Field diary, July 2021)

Reflecting on this conversation, Author 1 realised that communicating the aim of a research project to a worker who wanted to enquire about a job was a difficult task, and a disappointment to the worker who had hoped for a long-term job. Similarly, Author 2 (Franziska Baum) created a broad ad for a care worker interested in a long-term, recurring job, willing to work up to three times a week for several hours each day. She chose this amount of hours deliberately to approximate the real-life care needs of a client. After a worker expressed interest in the advertised job, the platform shared the worker’s telephone number with Author 2. While Author 2 was thus able to prepare for the conversation with the worker, the interaction was similar to Author 1’s experience: the phone call confused the worker and led him to be increasingly disappointed as he found out the job he had hoped for never existed.

In hindsight, Author 2 realised that the high amount of hours and long-term prospects she had created with this job ad had amounted to an extremely lucrative job offer in the context of freelance work. As a consequence, the worker had already become invested in landing a comparatively long-term job and predictable income. When it turned out that the ‘job’ amounted to a small compensation for a 2-hour interview instead of a recurring ‘gig’, he was understandably suspicious, disappointed and angry. At first, he agreed to interview at a later date but then cancelled the interview appointment. Feeling deceived by the false ad, he cited a lack of trust in the researcher as the main reason for his cancellation. It transpired that he was also worried about the purpose of the study, fearing it aimed to show that self-employment was inferior to a standard employment relationship in institutionalised care settings. As someone who had deliberately chosen to leave this very setting behind, he feared a study critical of the freelance model could contribute to forcing him back into this type of work. Subsequently, the worker also informed the platform about his interaction with the researcher. As Author 2’s client profile became de-anonymised, the platform immediately deleted it and banned Author 2’s email address. These two examples serve as an illustration of our recruiting experiences on several care platforms which have taught us two things.

First of all, care platforms make limited use of technology and algorithmic matching. Rather than receiving app notifications, the worker in the first example obtained Author 1’s number through the platform and directly called. Assuming that it was just a formality and contact with the workers would be via the digital platform interface, Author 1 provided her private phone number when setting up the ad and was caught by surprise. When Author 1 tried to take down the fake ad, it turned out this was also not technically possible within the platform interface. Instead, the only option was to have helpdesk staff remove the ad manually, a process which took several hours on the company hotline. Contrary to narratives about the ‘uberization of care’, there seemed to be little automation or technical intermediation throughout the process.

Second, in approximating real-life care needs and following the affordances of the platforms to enter detailed data of the potential care receivers, both of us inadvertently created expectations and hopes in our prospective research participants that we could not meet. Whether intended or not, this raises important ethical questions. As feminist scholarship reminds us of the need to constantly reassess how we might cause harm to the people researched (Bailey, 2012: 396–399), we will critically reflect on this issue in more detail in the following section.

Reflections on the Recruiting Process

Overall, our experiences served as a reminder to be cautious of our positionality in research and how this positionality shapes our inquiries. The discourse on the precarity of gig work among researchers in our field has made it difficult for us to see that workers do not only take on these jobs as a last resort. As the second example illustrated, the worker did not perceive his platform arrangement as precarious and was invested in keeping his job. In her interactions with this worker, Author 2 was confronted with his fears of critical research on a freelance arrangement that he was happy with. Just as auto-ethnographic research struggles to approximate a worker’s experience who may not speak the local language, have permanent residency in/citizenship of the country they work in and has to live off of their platform work (Keller, 2022, 2023), we also failed to anticipate how much workers would be invested into a potential job that seemed highly precarious to us. Our assumptions about what constitutes a desirable job may be very different from how workers subjectively experience different work arrangements and multiple forms of precarity (Alberti et al., 2018; Pulignano and Morgan, 2023). Especially low-wage workers may take advantage of new options that may make their lives easier, creating unforeseen new relationships of work (Ticona, 2022a). We can best map and understand these from their specific positionality.

This includes understanding that the concept of research itself might be opaque and lead to confusion, especially when the workers we try to reach face many ads that are not real and potentially dangerous. In the context of the ‘ubiquity of scams in the experiences of care workers’ job searches’ (Ticona, 2022a: 1553), workers have to invest time and hope into verifying job ads as well as unpaid labour to get false ads removed as the second case showed. This is not only emotionally stressful but also a waste of workers’ scarce time. This is especially true for the gig economy where piece wages turn time into a ‘calculative asset to the individual who must determine which opportunity best contributes to their livelihood’ (McKenzie, 2024: 247). This means not only that recruitment is challenging because sounding too ‘real’ or ‘too good’ makes workers suspicious (Ticona, 2022b), but it also means that we accidentally created a gig too good to miss. In approximating the long-term commitment of eldercare with a recurring gig and a large number of hours, we inadvertently skewed the workers’ calculation of whether the ad was worth engaging with in our favour.

While care workers have not, to our knowledge, collectively spoken out against being recruited by interviewers this way, a collective of highly researched crowd workers recounts how fake client profiles can cause a lot of disruption to workers: “through what amounted to at least 50 hours of sleuthing over two days, Turkers [verified that the advertised job] was a research project [not a scam]” (Turkopticon Collective, 2020). In their guidelines for researchers, these crowdworkers have therefore asked for researchers to stop blind-recruitment through fake job ads (Turkopticon, Collective, 2020). We have therefore considered whether the method could have been adjusted to avoid some of these pitfalls.

Theoretically, ‘downsizing’ the client ad to a one-time gig with little information on the person could remedy the ethical dilemma. Practically, this is not feasible for several reasons. First, all of the on-demand platforms for eldercare we researched required us to invent a person with specific details before the app design allowed us to specify hours. Hence, the interaction with a prospective interview partner would, in any case, have been founded upon an untrue picture of an elderly person in need that we have conjured. Second, the design and affordances of the platforms are similarly strict concerning the number of hours and the recurrence of the gig. As we outlined, on-demand platforms in eldercare are programmed to cater to long-term care needs and recurring gigs. And third, even if advertising for a one-off gig was possible from a technical point of view, it would potentially skew our sample. Based on the interviews we later conducted with other care workers, we think those who look for long-term relationships and those who would apply for a one-time gig constitute separate segments of the workforce that barely overlap (Baum, 2024a, 2024b). We also found little overlap in the workforces on cleaning platforms and other care platforms suggesting self-employed care work constitutes a subfield of its own (see Baum, 2024a). For example, we could not find migrants from Central and Eastern Europe on care platforms even though they form a majority of eldercare workers off platforms in Germany (see Orth, 2024:485). Relatedly, gig workers in logistics and cleaning may not be invested in establishing a long-term client relationship (Maury, 2023; Newlands, 2024; Stingl and Orth, 2024), relational forms of care require trust and carry high opportunity costs for both parties involved. We therefore think snowball sampling from interview partners interested in short-term gigs to those who look for long-term care work is not a viable solution.

Overall, we think creating a person-in-need was necessary both to emulate the real-life care needs of a client and to fit within the technical design of the platforms. We do not think we could have adjusted the method to avoid the situations we described. However, our research also showed us how ‘low-tech’ these supposed tech companies are, a finding since also been corroborated by many other researchers in the field (Rodríguez-Modroño et al., 2023, 2024; Tandon and Rathi, 2021). Many of the care platforms we researched resemble traditional agencies and there seemed to be little to no algorithmic management. This puzzled us at first and prompted many discussions as to what a platform even is and whether we could even call them ‘care platforms’ if the ‘key ingredient’ of algorithmic management was missing.

Yet, we think it is precisely these moments of irritation in platform research that can be particularly fruitful. These ‘glitches’ in platformisation serve as an important reminder to take a step back from the platform itself and ask whether it is the algorithms that are changing work in a sector. Following Leszczynski’s suggestion to construct ‘minor theory’ from these ‘moments and sites where platforms materialise otherwise or differently than expected, where [they] fall short of their ambitions for capitalist frictionlessness’ (Leszczynski, 2020: 197), we think we need to look for answers beyond platform interfaces that require different research designs and methodologies. In the following section, we suggest what this could mean for designing future research projects in the field of care platform labour.

Designing Future Research Projects Beyond the Platform

As we have laid out, our empirical insights are embedded in the specific context of a declining welfare state, and considering the trajectories and state regulations in domestic and care work differ vastly across countries and regions, other researchers’ experiences may differ. Yet, some of the issues we have outlined – the invisibility of the workforce, work in the private realm, the long-term requirements of relational forms of care – will occur across contexts. Amid these challenges, finding ways to interview care and domestic workers will always require creative solutions. Drawing on our extensive review of the methods and research designs in the care platform labour literature as well as the experiences outlined above, we, therefore, suggest the following strategies for conceptualising further research in this field:

First, we should engage with the wealth of feminist knowledge production on domestic and care labour at large that remains thus far underutilised in digital labour research (Grau-Sarabia and Fuster-Morell, 2021; Huws, 2019; Jarrett, 2016). This research foregrounds the importance of embodiment, the hybridity and fluidity of labour, the diversity of work arrangements and specifically workers’ need to combine paid and unpaid reproductive work.

To incorporate this into research designs means adjusting research questions, focal points and assumptions about what constitutes precarity in and of labour. Rather than focusing on the point of production, the platform, we need to research what happens ‘after’ a gig (Lalvani, 2021; Milkman et al., 2021) and how precarity is unevenly distributed “in and beyond work” (Strauss, 2020: 3). Consequently, the category of ‘worker’ needs to be opened up to understand how precarity is experienced differently by different people. Importantly, if and how an individual experiences precarity can also shift and change over time depending on their life circumstances (Stingl and Orth, 2024).

To this end, ‘follow-the-people’ / ‘follow-the-subject’ or ‘follow-the-thing’ (Marcus, 1995) approaches are more helpful than starting from the platform. We can, for example, find workers by looking for them where they may turn to get their needs met (Ticona, 2022a). As a method, repeated and biographically oriented interviewing seems particularly appropriate. This method allows researchers to understand work as embedded in larger life circumstances and can also trace the unfolding of precarity over time. In addition, repeat interviews can help tackle the issue of small interview sample sizes. Rather than spending their time recruiting as many workers as possible, researchers could create rich biographical case studies in the future. This longitudinal perspective seems especially appropriate to a field with long-term work arrangements.

Second and relatedly, we need to address workers in the profession they identify with. In domestic and care work settings, this means their identities as multiple and committed care workers (Azzarello and Schwenken, 2014; Dupuis et al., 2022). Even if this type of work is mediated by a platform, workers may not understand themselves as platform workers first and foremost; in the case of relational eldercare, it seems people understand themselves as caretakers in a particular ecosystem that is not centred on the platforms at all (Baum, 2024b). Instead, they often foreground their commitments or duty of care to their clients.

Third, for this reason, we caution against aggregating different forms of household-associated work and subsuming different operational models under the ‘care’ label because this risks obscuring very important differences. For example, in our experience, care platforms tend to feature well-settled trained and professional care workers rather than newly arrived migrant workers seeking jobs in any field. Once we acknowledge this stratification of the workforce we can also unpack how platforms seek to monetise different care gaps (e.g. Baum et al., 2020; Mateescu and Ticona, 2020; Rodríguez-Modroño et al., 2023, 2024). In turn, this may allow us to carve out the specifics of how platform models relate to particular gender, immigration and welfare state regimes.

Finally, we may also employ mixed-method approaches to critically evaluate the scope of the phenomenon of care platforms. While tracing the investor capital flows into the field may be an approximation of the phenomenon (Heiland, 2020b; Posada, 2022), we know that investor capital is not necessarily linked to functioning business models in the platform economy. In other words, investors often fund start-ups that do not have viable business strategies, and the phenomenon of ‘gigified’ care may be much smaller than we are led to believe. To validate this hypothesis requires mixed-method and computational approaches that provide quantitative estimates of the number of active care workers on platforms, for example, through the use of web scraping.

To successfully employ these research designs, researchers need time to understand the differences and operation modes of several platforms in a sector. Depending on each platform, they may also need to invest in obtaining specific skills such as learning and recruiting in foreign languages, or learning how to code and deeply engage a platform’s design features, especially when the platforms they investigate are not well-documented in the literature. Similarly, reaching workers within the field of highly relational work – whether on or off platforms – needs long-term commitments.

At the present moment, these commitments are often not compatible with the time frames of research grants and PhD fellowships, and, as a consequence, care questions are often dropped from platform labour studies. As we have outlined, this skews the field towards logistics work and risks biassing our theorising on the future of work. Another alternative, as our literature review has highlighted, is to conduct research on care platforms without interviewing many workers. While this has undeniable merit, we believe there are ways to continue to engage with workers in this field, and we hope this article is helpful for everyone who tries to pursue this course, difficult as it may be.