Social capital creation on professional sharing economy platforms: The problems of rating dependency and the non-transferability of social capital

The sharing economy platforms facilitate collaboration across geographical boundaries and promote service innovation by reshaping traditional business networks. This study takes a Social Capital Theory perspective on how Social Capital (SC) is created on professional sharing economy platforms, with particular attention to the creative services industry. Our in-depth qualitative investigation draws on 35 interviews with freelance designers and platform clients based in 17 different countries. The study demonstrates that SC created outside sharing economy platforms is not readily transferred to these platforms, which represents a major difference from the dynamics of SC in more traditional settings. Furthermore, SC transfer between platforms is difficult. Building platform-specific SC ‘from scratch ’ requires a significant effort and is highly dependent on reputation systems, in the form of ratings and reviews. We argue that the platforms ’ reputation systems force members to become ‘slaves ’ to ensuring their star ratings and reviews are as good as possible. In addition, we explore how platform members learn to build SC on the platforms beyond ratings and reviews. Overall, the study contributes to aca- demic discussions on opportunities and challenges for service innovation within the sharing economy and introduces the application of Social Capital Theory to the context of sharing economy platforms.


Introduction
Sharing economy platforms accelerate the creation of innovative service offerings across geographical boundaries (Sundararajan, 2016) and are of increasing economic importance: in recent years, Airbnb has sold more guest nights than the Hilton chain worldwide, and platforms like Upwork and Freelancer.com are growing at over 20% a year in terms of numbers of users (Lehdonvirta, 2018). The sharing economy has been described as obtaining, giving, or sharing access to goods, services, or information, coordinated through community-based online services (Hamari et al., 2015), typically facilitated by global collaborative digital platforms (Schor, 2014;Sundararajan, 2016;Holland & Brewster, 2020) 1 . Different types of platforms support the sharing economy: some require the geographical proximity of buyers and sellers for service delivery (e.g. AirBnB, Uber, Task Rabbit). Others provide virtual services that can be performed remotely (e.g. Upwork, Freelancers.com) (Howcroft & Bergvall-Kåreborn, 2019), which means that buyers can access a global pool of 'crowd workers' via the 'distanceshrinking network powers of the internet' (Langley & Leyshon, 2017, p. 9).
Another important categorisation distinguishes between asset-based service platforms (e.g. Uber) and professional or online task platforms (Howcroft & Bergvall-Kåreborn, 2019) or online labour platforms (Wood et al., 2019). While asset-based service platforms are economically important, from a service innovation perspective, professional platforms are especially interesting, because they facilitate service innovation whereby a new service or method of service provision is implemented through a process of transformation. Besides clients, the beneficiaries of service innovation can also be the providers, business owners, alliance partners and communities (Ostrom, 2010). Service innovation does not occur in isolation because participants are embedded in networks (de Reuver & Bouwman, 2012). However, there is currently little understanding of how sharing economy platforms disrupt traditional business networks.
The theoretical underpinning for this investigation is Social Capital Theory (SC or SCT), which describes the relationships between individuals and social networks and the norms of reciprocity and trustworthiness arising from them (Nahapiet & Ghoshal, 1998;Arrow, 2000;Lin, 2001;Portes, 1998). SC can take different forms: relational, cognitive, and structural (Nahapiet & Ghoshal, 1998;Li et al., 2013). The different ways in which certain types of SC can be converted into other types are widely discussed in the literature. This leads to the first contribution of this study, which is the identification of problems pertaining to the transferability of SC within professional platforms.
We study the role of reputation systems in the sharing economy, as reputation influences how collaborations evolve in the network and thus affect service innovation (Foroudi et al., 2016). Despite the acknowledged importance of reputation mechanisms such as ratings and reviews (Dellarocas, 2010;Basili & Rossi, 2020), there is limited knowledge about their 'dark side'. While some studies have examined the limitations and downsides of rating mechanisms for asset-based platforms (e. g. Uber, Airbnb) (Rosenblat & Stark, 2016;Chan & Humphreys, 2018;Chan, 2019) and for microwork platforms (e.g. Amazon Mechanical Turk) (Wood et al., 2019), the professional platforms for creative work (e.g. Upwork, Freelancer.com, People-per-hour) have been largely overlooked. The second contribution of the study is that it addresses this gap by exploring some of the effects of sharing economy reputation systems on professional creative work platforms, where providers and clients collaborate in the production of innovative solutions. We argue that current systems are not well suited to the efficient matching of platform members. This hinders creative innovation because the platforms' collaborative potential is not fully utilised. In parallel to this, the third contribution of the study is that it develops an understanding of the network disruption that professional platforms create.

A social capital Theory perspective on the sharing economy
SC is 'the sum of the resources, actual or virtual, that accrue to an individual or a group by virtue of possessing a durable network of more or less institutionalized relationships of mutual acquaintance and recognition' (Bourdieu & Wacquant, 1992, p. 14). SC Theory describes the information, power, and solidarity that an individual can draw on to accomplish goals. It concerns both the relationships an individual has and also the location of that individual's connections in the wider social system (Burt, 2000;Coleman, 1990). The connections between individuals contribute to the roles and procedures that emerge from and are embedded in a network (Nahapiet & Ghoshal, 1998;Ellison et al., 2011). Bourdieu's (1973) theory of social and cultural reproduction explains that SC creation is facilitated by knowledge sharing and can later be transferred to economic and other forms of capital.
Social Capital Theory connects information benefits, power and solidarity that an individual can draw on to accomplish goals (Arrow, 2000;Portes, 1998;Sandefur & Laumann, 1998). Affordances of Social Capital online incorporate both enabling and restricting affordances in relation to Social Capital in an online environment (Nie, 2001;Ellison et al., 2011). This is different from digitally enhanced, yet physically informed social contexts, such as the way social media use influences the emergence of friendship, collegial and wider business networks. Professional platforms form boundaries online and rarely tap into the physical forms of meetings and collaborations. SC can take different forms: relational, cognitive and structural (Nahapiet & Ghoshal, 1998).
Relational forms are characterised by emotions (e.g. personal liking/ disliking), whereas cognitive forms build on a shared understanding in an intellectual/factual sense. Unlike relational and cognitive formats, structural SC is not restricted to an actor's direct relationships, but instead connects actors in an overarching network of relationships (Burt, 2000;Coleman, 1990;Putnam, 1993;Ostrom, 2000). To possess SC, an individual must be related to other individuals, and it emerges from a network of connections (Solow, 2000;Lin, 2001). Due to its collective nature, it cannot be converted into a private good (Fukuyama, 1995). Stocks of capital (such as trust, norms, and shared values) accumulate with usage and weaken if not used.
One of the most important sources of providers' SC is their interactions with clients (a form of structural SC), which acts as a driver for the provider to develop connections through, for instance, trust (a form of relational SC) and a collective mind (a form of cognitive SC) (Fukuyama, 1995(Fukuyama, , 2001Tsai & Ghoshal, 1998). The provider's structural SC can manifest itself in, for instance, the frequency of interactions with the client (Coleman, 1990;Nie, 2001;Villena et al., 2011). Eklinder-Frick et al. (2011 apply the notion of SC to business network settings, noting that the effect generated by businesses cooperating in a network depends substantially on the strength of the social resources of the group and the level of SC in society more widely. Putnam (2000) defines two core functions of Structural SC: bridging and bonding. Bridging is the process of connecting actors that did not know each other previously and bonding makes existing relationships closer and strengthens connections. While functions are relevant to business-tobusiness settings (Eklinder-Frick et al., 2011), most research focuses on bonding.
The digital space can change relationship dynamics and thus the way SC is created. For instance, Boyd and Ellison (2007) demonstrate how social media helps the conversion of latent ties into weak ties, the maintenance of extant ties and the resurrection of past relationships. In this sense, the creation of SC becomes easier online (largely because geographical boundaries are readily overcome) (Enders et al., 2008). The bridging function of SC in digitally enhanced business environments is demonstrated in the work of Nohria and Eccles (1992): '[…the physical] network of relationships serves as a substrate on which the electronic network can float or (…) be "embedded". What the electronic network can do is accelerate (…) the communication flow, but its viability and effectiveness will depend critically on the robustness of the underlying social structure' (p. 304).

The role of reputation systems on sharing economy platforms
The majority of digital platforms use reputation systems (Howcroft & Bergvall-Kåreborn, 2019), in the form of ratings/reviews that allow clients to provide feedback to providers and guide other clients (Pongratz, 2018). Reputation can be defined as a review of one's previous actions within a community that can help other people to make decisions about building a relationship with that particular actor (Dellarocas, 2010). While in tightly knit communities, relevant information is known to members through frequent interaction, in sharing economy platforms, with thousands or even millions of members, central reputation systems are essential, as they greatly facilitate online transactions among actors who are not known to each other (Basili & Rossi, 2020). Reputation systems constitute an integral part of an actor's online presence and have significant implications for network formation on the platforms.
Reputation systems are argued to operate as an 'invisible hand' that is supposed to reward high-quality providers (Howcroft & Bergvall-Kåreborn, 2019). Ratings/reviews have strong signalling effects and play a major role in building initial trust between parties. However, unlike the endorsements that actors may post on their corporate websites, providers have little control over the information being signalled. Nonetheless, uncertainties apply on both sides: clients may experience scarcity risk that refers to the perceived likelihood of resource/service unavailability (Akbar and Hoffmann, 2020), which is especially interesting in terms of the availability and timely response of the providers. Clients on the platforms rate providers' performance by scoring (on a standardised range, such as 1-5 stars) and/or by giving a brief review. The scores frequently include subcategories related to cost, execution time frames, quality of work, response time and general knowledge and skills (Schörpf et al., 2017). These are then translated into metrics, most commonly a reputational score, which are used by clients as proxies for trustworthiness (Gandini, 2019).
Reputation scores tend to be platform-specific and, therefore, reputation systems act as a significant entry barrier. New providers are not readily able to demonstrate the reputation gained on other platforms or outside of the platform economy in general (Nemkova et al., 2019). All providers start with a score of 0 and must work their way up to attract clients. There is a consensus in the literature that without a good online reputation, it is difficult to get new projects (Schörpf et al., 2017). Ratings/reviews are a powerful tool as they replace many of the traditional credibility measures, such as employers' references and personal recommendations. As a result, at the early stages of operating through the platforms, many providers are prepared to execute additional tasks and demand less payment in an effort to gain positive feedback (Demirel et al., 2021). It is argued that for online task platforms (Howcroft & Bergvall-Kåreborn, 2019), for example Upwork or Freelancer.com, reputation in the form of scores is particularly important for the continued flow of work (Barnes et al., 2015), as it influences the selection of a new partner and the employability of actors (Gandini, 2019).
The platforms perform a much broader role than that of a mere thirdparty intermediary connecting clients and providers (Newlands et al., 2018). Actors are typically required to meet a variety of specific metrics in order to operate (e.g. delivery time, availability) (Rosenblat & Stark, 2016). These metrics have been viewed as a form of 'algorithmic control' or 'algorithmic management' which aims to direct the behaviour of its actors, highlighting a 'dark side' of the reputation systems (Shevchuk et al., 2021). They also encourage workers to engage in 'emotional labour' (Newlands et al., 2019) in exchange for a high rating (Chan & Humphreys, 2018). While many platforms do have the facility for providers to review and rate clients, Chan (2019) argues that negative ratings have more severe effects for providers than for clients, because of significant power asymmetries (Shevchuk & Strebkov, 2018). For example, Uber 'deactivates' providers with low rankings, while platforms for creative work typically 'use algorithms to filter work away from those with low ratings, thus making continuing their work on the platform a less viable means of making a living' (Wood et al., 2019, p. 64). In general, the need to comply with strict rules and algorithms at times resembles 'digital Taylorism', where the work is broken down into incremental steps and the efficiency of the providers is closely monitored (Rosenblat & Stark, 2016). That, in turn, can be detrimental for the achievement for innovative and creative solutions where trial and error is an important part of the process. Knoke's (1999) work on Corporate SC regards the processes of forming and mobilising actors' network connections within and between organisations as a prerequisite to gaining access to other actors' resources. Accordingly, reputation systems in the form of ratings/reviews can be conceptualised as 'network mobilisers' representing a new tool that can shape the network, extending the theoretical framing of Mouzas and Naudé (2007) on network mobilisers. Ratings/reviews trigger the mobilisation of information, which is an important resource and facilitates further connections, bridging structural gaps in the network (Rajagopal & Sanchez, 2005).

Network formation facilitated by sharing economy platforms
Service providers with the best scores receive more project requests and proposal acceptances. This has significant implications for network formation on the platforms. According to Wood et al. (2019), projects flow to those providers who maintain a high reputational score over a long period, and who have a broader network of platform contacts. Alacovska (2018) investigated the role of strategic relational work on the platforms, where providers aim to 'build intimate and close relationships' and to secure 'a favourable position in online relational infrastructures' (p.1575). Such investments in the development of personal relationships are transferred into positive reviews and higher ratings. This is in line with the research of Wilkinson et al. (2005) examining how network partners find each other through business mating: developing mental representations of what suitable partners are like, which develop through experience over time, and result in the matching of potential partners.
A platform greatly increases actors' ability to make direct contacts (and build networks of contacts); which in turn facilitates intermediation, as business actors can more readily subcontract their 'gigs' to other actors (possibly newer and/or with lower ratings). Opportunities can thereby be created for new platform members that previously did not exist (Lehdonvirta & Bright, 2015). Intermediation can be particularly useful for large tasks that can be broken down, where the subcontractor coordinates the completion of smaller tasks and ensures the overall quality of the project (Graham et al., 2017). Therefore, platforms can be considered as new forms of intermediaries that disintermediate traditional networks by removing the middleman (Holland & Brewster, 2020) and provide opportunities for intermediation via the emergence of new business actors that extend online value chains.
Based on the above, this article addresses three important issues that remain understudied: (a) (non)transferability of the SC from offline to professional platforms and how it could be facilitated; (b) what the role is of reputation systems in the client-provider relationships; and (c) how professional platforms disrupt traditional business networks.

Methodology
Qualitative exploratory research among creative providers (freelance designers) and their clients was conducted to develop an understanding of the creation and dynamics of SC, the role of reputation systems, and patterns of network formation for collaborative innovation. The focus is on providers on four of the most well-known professional platforms: Upwork, PeoplePerHour, Freelancer.com and 99Designs. All four platforms connect freelancers and platform clients (i.e. individual and organisational providers) to perform a variety of tasks, such as design and content writing. While there are various common traits, there are also some differences between the platforms, as shown in Table 1.
The providers were primarily freelance designers who perform a variety of tasks, ranging from logo or poster design to website design. Purposive sampling was applied to recruit interviewees (Teddlie & Yu, 2007). The provider participants came from 17 countries, and the client participants from 5 countries ( Table 2). As the sharing economy facilitates global inclusion (Sundararajan, 2016), it was important to capture the perspective of the actors located in multiple geographical locations with different economic and social-cultural conditions (Graham and Anwar, 2018).
Freelance designers were approached through a major social network for creative professionals that is used to demonstrate creative projects. One hundred and sixty-six designers were selected if they had an active (rather than dormant) profile, and had posted work in the previous month. Initially, they were sent a private message to ascertain whether they had undertaken any work on at least one of the platforms and whether they would be willing to participate in a study about their platform engagement. While some providers only used a single platform, others worked through multiple platforms. Providers who confirmed their usage and willingness to participate were invited to an interview via Skype. That resulted in 26 interviews with freelance designers. During the interview, they were asked to recommend one of their clients to participate in the study, and this resulted in 9 interviews with platform clients. All the participants of the study were granted anonymity and the ethical guidelines of the first author's university were followed.
Interviews lasted between 45 and 90 min, were audio-taped and transcribed. All interviews were conducted in English. Due to the exploratory nature of the study, semi-structured interviews were conducted, which allowed a higher degree of flexibility (Bryman, 2012). While some of the questions for providers and clients were similar (e.g. those concerning their general experience of the platforms, the initial motivation to participate, and frequency of use), there were some important differences. Providers were asked about their approach to clients and their personal strategies to engage effectively with both clients and platforms. They were questioned about the differences in their relationship with 'platform clients' and other forms of relationship with clients they had experienced. We were also interested how they saw their career prospects both on and off the platforms. Interviews with clients covered the strategies they used to look for a suitable provider, the types of projects they chose to post, and the ways they managed their relationships with the providers.
Thematic analysis with a priori codes (Brooks et al., 2015) was used to process the data, which was then enhanced by open coding. This combination of a priori and open codes draws on a two-level analytical procedure that blends both in an iterative process (Bernard et al., 2016;Patton, 2002). The a priori codes were theory-driven, and the open codes emerged from the data. The literature review on SCT informed the development of the initial (a priori) codes, covering 'social capital', 'network', 'trust', 'norms' and 'values'. Other codes were derived based on the themes arising from the data: 'soft skills', 'communication', 'time pressure', 'level of technical skills', 'bypassing behaviour' and 'barriers to innovation'.
All interviews were coded by two researchers and the codes were cross-checked. The coding process was manual; codes were indicated through highlights with the use of different colours and written comments. Word documents of the interview transcripts were shared between the researchers. Analysis of both a priori and open codes enabled us to establish 30 first-order categories that were then developed into overarching themes. The emerging themes were reviewed interactively and iteratively. Table 3 demonstrates the linkages between theoretical pre-understandings, first-order categories, and second-order themes.

Analysis and findings
Providers reported that the platform enables them to connect with clients directly, without the involvement of middlemen such as traditional ('bricks and mortar') creative agencies. They do not need to work '9 to 5 ′ and enjoy greater flexibility in connecting with clients and collaborators, irrespective of geographical boundaries. They can also avoid the administrative and organisational complexities of agency work. At the same time, clients connecting directly with providers via platforms instead of dealing with agencies enjoy considerable cost savings for their business (1b, 5e) 2 .
The usual matching mechanism on the platforms takes the form of a client posting a project (e.g. a brief, a contest), which multiple providers apply for, one of whom is chosen for the task. Platforms also allow providers to be approached by clients but that happens less often, as there are considerably more providers than clients (2a, 6d). 1 Each platform has its own 'reputation score' that is calculated by the platform based on its own algorithms. The exact methodology of the calculation is not known; the lack of transparency is often reported to be one the downsides of the system (e.g. Wood et al., 2019).
2 Numbers 1a-7d in the text correspond to the coding structure presented in Table 3 Z. Tóth et al.

The non-transferability of social capital on and to sharing economy platforms
The findings offer insights into how the sharing economy reshapes traditional, established network structures, by offering new opportunities and markets. However, it inevitably decreases the SC of certain organisations, especially for some small and medium-sized creative agencies, which are entirely bypassed (1b). Designer Antonia (Portugal/ Brazil): 'When I was in the agency, the designers, they don't have contact with the client… There is an account manager… So, the account manager has this job that he or she speaks to the client.' Acquiring work and building relationships with clients as well as potential collaborators on professional platforms comes with some general and some specific challenges. Creating a good track record and becoming embedded in networks are always time-consuming (3a, 3e).
Designer Patricio (Bolivia): 'The first three months, I was just applying for jobs with no success and he [the first client] was the first after three months. I think three months is quite a good time.' There are typically no word-of-mouth recommendations and the prior professional achievements of the provider are practically invisible to others on the platform (6e). The only evidence of the previous experience 'allowed' by the platforms is that the providers' portfolio that can be uploaded (3c). The majority of providers considered it as an important feature to demonstrate their experience, but saw it as insufficient to communicate their track record. While providers can bring a former client to the platform (for the incentive of a reduction in fees), they cannot incorporate previous testimonials from satisfied clients (3b). The option of bringing a former client was used by only one participant in our sample (Zehra, designer, Turkey). This strategy is not seen as beneficial for clients unless they are planning to engage on the platform in the future. Waiving of fees is not considered enough of an incentive and rarely helps providers to start their engagement on the platforms. That is, existing SC (in the form of reputation) is left behind when starting on the platform. Providers found having to start from scratch especially challenging (3d):

Designer Paul (UK/Poland): 'On portals like Upwork or Behance, it's huge to have some kind of presence because if you want to get clients, they need to know that you've worked with someone else because people like to know that you're reliable and someone else has already paid you money, so you look like a person that can be trusted. It's all about building trust and the only way to do it is to start from practically zero [very low pay] and build it up as you go.'
Designer Angela (Venezuela): 'Every time you finish a job you have the opportunity to give feedback to rate the experience. The ones that hire you always rate you and they rate you about communication, responsibility, quality of the work or something like that.' One strategy that providers used to overcome this barrier was to charge extremely low rates for initial work in order to establish their presence on the platforms (3d, 3e). When the first projects were completed and providers received positive reviews, the acquisition of further projects became easier, as long as a high average rating was maintained. The network formation took off only once providers had managed to receive their first review and rating (1a). They often found themselves working under time-pressure as they discover that clients join the platforms to find speedy solutions to their problems (6c); in reality, providers were ready to accept it to maintain their platform track record. However, after overcoming the initial hurdle, they discovered that there were additional barriers in place (e.g. metrics of hours worked on the platform and percentage of completed projects) that still hindered them from building their online SC (3a).   (Burt, 2000;Coleman, 1988;Putnam, 2000) 1a.  ( Fukuyama, 1995( Fukuyama, , 2001Lin, 2001;Ostrom, 2000) 7a. Cooperative efforts to meet client needs 7b. The value of responsiveness 7c. Trust-building on platforms 7d. Repeated projects Norms of engagement

Becoming a slave to reputation can hinder innovation
Having a history of relevant reviews and good ratings is crucial for building collaborations, requiring a long-term perspective. Nevertheless, it is a difficult task, as providers have limited control over their scores (4c), unlike endorsements posted by companies on corporate websites. Client testimonials published on corporate websites are filtered by the company and only favourable quotes are selected, whereas providers do not have the option of cherry-picking the feedback. There was some anxiety associated with the scores, with providers raising concerns that they do not fully understand what the algorithms are and how their scores are calculated (4a). The lack of control and over-reliance on ratings/reviews makes some providers feel unsure about allowing themselves to 'think out of the box' and become more innovative (6a):

Designer Youel (Israel): 'I'm a little scared to work on things that I don't have a lot of experience in, because I don't want bad reviews …. [The platform] works on the review system, so if you have bad reviews, your work success percentage goes down and it's really hard to find new work. I tend to stay in areas that I know that I have experience in, [so] that I can do a good job.'
There was evidence that providers tend to avoid applying for innovative projects or coming up with unusual solutions in order to 'play it safe' and stay in their comfort zone, largely to protect their ratings (4c, 6a). Ironically, one of the strong motivations to join the platform initially was often a desire to access a larger number of innovative opportunities and to be allowed more creativity (5b, 5d). This was particularly strongly voiced by those providers from the countries where the design market is small and underdeveloped. A bigger pool of clients and opportunities to grow professionally (5a, 5e), to extend the network (1a) and expertise (5d), were seen as the important benefits offered by the platform: While ratings/reviews are meant to signal service quality and help members to find the best possible match, the systems are still far from fully accomplishing what they were designed for, and currently their network facilitation is suboptimal (4b). At the same time, they are powerful mechanisms that can drive an actor away from the platformlow ratings could mean that clients will not employ a particular provider, requiring them to leave the platform to search for work elsewhere (6d).

Reputation systems and client engagement
The clients confirmed that they do pay attention to the ratings/reviews of particular providers, but they also highlighted important challenges associated with the reputation system: while these are supposed to guarantee a certain level of service quality, it is not necessarily the case (2a). Clients reported difficulties in identifying suitable providers and distinguishing them from those who provide lower-quality services, regardless of their ratings (4b). One client indicated polarisation of providers on the platforms: Designer Blagovest (Bulgaria): 'There are two main categories of designers out there. One is the executors who are there to get instructions, do what they're asked and deliver the best result. The other types are closer to consultancy, and they're there to look at things in depth and offer solutions. The majority of designers are on the execution side, and the minority are on the consultant side. (…) To offer them a path we can take so that we get closer to what they have in mind. This is a really challenging thing to do.' In the presence of this polarisation, clients reported difficulties in distinguishing between providers who are able to perform high-quality work and those who offer lower-quality services based on the information provided by the reputation systems (4b). Clients described situations when providers with excellent ratings delivered poor-quality work, and then asked the client not to give a negative review or 'anything less than a 5-star rating'. One of the clients (Robert, UK) explained that 'one needs to be careful and not to take ratings for granted', saying that he had repeatedly been asked not to provide a negative review in return for not paying for the job (4d). That then presented an ethical dilemma: the client was well aware that agreeing to this proposal would contribute to the on-going biases of the reputation systems, while providing honest feedback would reduce the average rating of the provider and likely damage their business.
A particular challenge is that the content of reviews as well as ratings leave little to no space for relational information to be shared; more space would improve the matching mechanism between clients and providers. One client explained that current reputation systems do not enable truly compatible matches for more efficient collaborations to be identified and suggested that, instead, it needs to be done through the 'trial and error' of engaging with a provider (2a):

Client Sofia (USA): 'I think there has to be a personality match and an expectation match, and I don't know if it's through a survey, a personality test or something.'
Regardless of the time-consuming process of identifying a suitable provider for a job, clients indicated that in general the platforms are helpful for 'discovering the talent' (5a). Moreover, once trust is established and a relationship becomes stable (7c), working through the platform becomes less appealing (1d), because of the fees ('up to 20%'). Multiple providers identified 'high fees' as one of the main challenges in working on the platform and believed they did not receive 'anything in return' (e.g. pension, paid annual leave, insurance) (6b).

Designer Nicolas (Colombia):
'…the fee charge that they're taking, the 20% thing and then after 500, 10 and after like $10,000, like 1%…. If they're going to be like that, it's like, after I get graduated and get my pension, no. That's not going to happen.' For clients, engaging with providers off the platform brings unquestionable benefits (principally cost savings), but for providers it can have negative effects (1d). Some providers explained that shifting their relationship with a particular client offline affects their platform metrics (4c) (both the completion percentage and the rate of activity will decrease) which in turn are likely to negatively influence their employability for new projects, and be detrimental to the building of a network (1d). This network dynamic underlines that there is high dependency on ratings/reviews; maintaining a good profile through the reputation systems is important for signalling employability and the creation of SC.
Designer Youel (Israel): 'I was out of the business for a few weeks because I was quite busy, and now it's really hard to get back on track. I have to always look for jobs, always searching for the next job. Once I stopped it was hard to get back. I also work with independent clients and other small studios that need help from freelancers or designers. Because I told them I was busy and now everybody disappeared, and I don't have any work.' The current reputation system is efficient in facilitating client engagement, but leaves providers in a vulnerable position, where they sometimes have to choose between continued ratings/reviews and financially more viable work outside the platform (1d). Platforms also closely monitor if the client-provider relationship has been shifted outside the platform. When these are identified, providers are heavily penalised, either by being charged additional fees or through suspension and/or removal of their account (6d).

Platforms facilitate 'gigs' but there is an increasing preference for building more stable service solution networks
While several of the relationships on the platforms were one-off 'gigs', some clients and providers build long-term relationships and reach various creative solutions over time (7d). There were even instances of providers managing to establish such a strong clientele base that they started subcontracting more simple tasks to other platform members to deal with the volume of projects (1c). Data revealed that a group of platform members consciously invests in building their SC on the platforms by investing into more long-term and trustworthy relationship portfolios (7c). This is especially important in understanding how platform members build SC beyond a reliance on ratings/reviews.
Establishing a mutual understanding between providers and clients on platforms matters, even if the reputation systems provide limited support for this. While some platforms provide a progressive fee structure to incentivise long-term collaborations, providers did not mention it as a reason to invest in long-term relationships. A closer look at information from those who occupied strong positions in the network (i.e. were well-connected and well-paid) revealed that a vital factor in obtaining high ratings and positive reviews is the ability to build interpersonal relationships with clients and to develop communication skills (2c, 7a).

Designer Ahmed (Morocco): 'There are people that won't communicate with the client when they get the contract and only expect talking to him when they finish the job -that's extremely bad. I think many designers do this; they tend not to have good communication skills…. It's not your design skills [that differentiates you]. I actually became a better designer when I started reading books about finance in business.'
Some participants pointed out the important balance between maintaining a high rating (a signal of technical quality) but then strategically investing in signalling relational quality via communications (such as covering letters and asking the right questions). Multiple providers explained that it is their soft rather than technical skills, that allow them to charge higher rates and to get involved in highly creative projects (2b, 2c). These platform members argued that the domain of their expertise is much broader than design, and includes innovative problem solving and solution provision (7b). For this reason, some providers invest in the development of their soft skills (e.g. communication) not in order to receive fee reductions, but for higher-quality relationships with clients, that then tend to be indirectly translated into higher ratings and positive reviews (1a, 7a, 7b). There is an increasing awareness among platform members that a strong network position depends on a combination of technical and soft skills (2a, 2b). Those who have that awareness are better able to create SC on platforms, strengthen collaborative ties and achieve competitive advantage.
Therefore, while some providers learn to navigate the platforms successfully by using their communication skills, they did so not with the 'help of' but 'despite' the reputation systems. The ratings/reviews that can help the actors to navigate on the platforms and facilitate the formation of platform-specific SC often fail to perform their function, and instead create barriers to network building. Therefore, for the reputation systems to evolve appropriately, they need to incorporate more efficient matching mechanisms.

Theoretical contributions
This study focuses on the creation of SC on professional platforms by studying how creative service providers utilize them. It identifies some important constraints that the system imposes on the providers, other than the positive ones identified by Sutherland and Jarrahi (2018), such as the generation of flexibility, building trust and a sense of collective understanding, and matchmaking. Studies characteristically assess platforms as tools of empowerment for innovation (Bouncken et al., 2020) and as catalysers of both improved connections between providers and clients (Huarng & Yu, 2020) and superior customer experiences (Lu et al., 2020).
This study examined a number of issues in further depth, especially those rooted in the creation and management of SC.
The first issue is the non-transferability of SC to sharing economy platforms, which extends the debate on the trade-off between the transferability and specificity of SC (Sturman et al., 2008). The study identifies the non-transferability of SC from off-platform contexts to professional platforms in the sharing economy. Implicitly, Nahapiet and Ghoshal (1998) touch upon the question of convertibility between intellectual and economic capital in the creation of SC. For instance, those with higher intellectual capital may be able to extend this into the creation of greater economic capital through getting well-paid jobs or making profitable investments. Similarly, someone with high economic capital has the option of allocating considerable resources to education and training. Our study demonstrates, however, that the relatively smooth transferability that applies in certain contexts does not apply in the case of most sharing economy platforms. In fact, providers complained that transferring their non-platform-related credentials onto professional platforms is nearly impossible, requiring them to start from scratch. From a wider perspective, our paper contributes to research that has identified SC transferability issues in different contexts, for instance, the less-than-perfect international transferability of qualifications and skills of immigrant workers (Chiswick & Miller, 2009).
It appears that the transferability of any kind of SC to a professional platform is limited; in addition, it is difficult to transfer SC from and between the platforms. For instance, providers who wish to move clients from the platform may face a penalty in the form of missed ratings, which would demonstrate some sort of non-observance of accepted platform norms. Most platforms have penalties to discourage users from engaging with each other outside of the platforms (e.g. account suspension or additional charges) (Newlands et al., 2018). Thus, professional sharing economy platforms display a 'lock in' effect when it comes to SC: when created elsewhere, it cannot be transferred to the platform and any created on the platform is difficult to mobilise elsewhere. This is very different from traditional employment settings, characterised by the ready transfer of skills acquired during earlier employment, and where employers may share detailed information, for instance, through word-of-mouth. Duggan et al. (2020) point out that the 'gig' business model bypasses regular employer responsibilities, with implications for employment relations and human resource management. The nontransferability of ratings created on sharing economy platforms is a reputation transfer issue that appears to be linked to legal requirements (especially those imposed by the GDPR, Article 20) (Teubner et al., 2019) and creates an interesting challenge for policy makers regarding how they could better accommodate the interests of providers and clients in the sharing economy.
Secondly, we argue that reputation systems implemented on professional sharing economy platforms largely do not encourage the development of long-term collaborations. Instead, they encourage shortterm gigs and emphasise the collection of 5-star ratings. The problem with this is that without the opportunity to develop long-term relationships, providers do not use their full innovation potential, as demonstrated by Pérez-Luño et al. (2011). Furthermore, scholars underline the significance of developing strong relationships as a way of supporting innovation (Holmen et al., 2005), as 'socially embedded' exchange relationships have an immediate impact on access to new knowledge and information, and they produce more novel combinations and development prospects (Uzzi, 1999). In cooperative innovation, the transfer of tacit knowledge characteristically requires informal communication methods and in-person contact (Kogut & Zander, 1993), both of which are highly challenging without close connections, as demonstrated in our study.
Finally, we found that professional platforms enable the elimination, or disintermediation, of some middlemen, such as 'bricks and mortar' agenciesfrom the network, which introduces more options and flexibility and reduces some costs. However, those eliminated will include entrepreneurial middlemen who create value through network facilitation (Ellis, 2003). Furthermore, the elimination of middlemen implies that certain tasks, such as administration, account management and negotiations, which would normally be managed within agencies, now fall under the remit of the (mostly) individual providers themselves. From our data it is evident that soft skills are important for providers' longer-term professional and financial growth on platformsbut this is counter to the short-term and technical/hard-skill focus of the platforms' reputation systems. Eller et al. (2020) suggest that soft skills are particularly relevant for entrepreneurs and small and medium-sized enterprises' digital innovationand the present study extends this line of thought with the case of microentrepreneur providers on platforms. In fact, while being a talented designer is obviously important, good communication and project management skills can be equally important, and this is an area in which some providers on professional platforms have significant potential for development. The divide between technical/hard and soft skills is rooted in the differences between the bridging and bonding functions of SC (Eklinder-Frick et al., 2011). While the ratings based on technical excellence may help to bridge structural gaps in the platform, such as the facilitation of new connections, providers still need soft skills for bonding, to strengthen their connections with clients and potentially with other service provides on collaborative projects.

Managerial implications
The study has important managerial implications, related to potential improvements in the transfer of SC to platforms; increased efficiency of the matching facilitated by platforms; and changes to reputation systems. Improvements that can increase the transferability of previously accumulated SC to professional platforms include the option to incorporate external referencing, such as through the integration of a LinkedIn profile and/or a company or personal website in the profile of the provider. Businesses facilitated by the platform would still keep most of its competitive advantage, as the review system would continue to protect work quality and decrease the risk of vendor and client selection. There is the potential for a new cross-platform actor to emerge here, similar to the emergence of independent credit rating providers in the financial sector. A platform-independent 'review master' could develop a 360 • view of providers based on verified reviews. It might be financed by selling this service to professional platforms and clients, like credit ratings in the finance services industry.
Partner selection and the engagement process could be improved by establishing more direct contact between buyers and providers, especially as part of the selection process and before the transaction occurs, for instance by setting up a short, interactive Q&A session via chat or videoconferencing, or by the creation of a short (up to 5-minute) pitch that could be delivered either synchronously or even asynchronously (as a recording). These pitches would be focused on overall style, ways of working and values, instead of testing skills. The interactive Q&A session would improve approachability. The right questions asked at the right time to clarify client needs and to set expectations can bring benefits on both sides. In addition, platform owners and experts could create educational materials on soft skills for freelance providers, for instance pertaining to the management of client communications, active listening, and time management. This could help individual providers to win bids and to maintain their working relationships with clients through the platform on a variety of projects.
Reputation systems can be amended by increasing review credibility and the depth of review content. First, they should allow the posting of more sophisticated reviews, grouped into categories to reflect the separate stages of the service experience, product quality, value for money, and other relevant dimensions specific to the platform. Secondly, they should introduce a weighted review system, where for instance if a client posted only one or two reviews overall, they would be weighted less, to avoid a single bad review having an undue effect. In parallel, frequent buyers and reviews by clients with a long-term relationship might be weighted more. Thirdly, it should be ensured that the review is independent, and the reviewer cannot be influenced by the provider. This also means that the review-avoidance behaviours of providers should be prohibited (illustrated by some providers requesting clients to not give them a rating, even at the expense of forgoing payment for the delivered service). Platform-wide ethical standards should be established, monitored and enforced by the platform. Finally, providers should be allowed to comment on client reviewsto provide clarifications and corrections in response to feedback from clients.

Conclusion and future research directions
With a focus on creative service providers, this study takes a SCT perspective on the sharing economy, identifying the non-transferability of previously created SC to professional platforms. The study demonstrates that while platforms facilitate a direct relationship between providers and clients (e.g. through the disintermediation of established actors such as traditional agencies), the sharing economy's reputation systems can hinder creative innovation, as platform members become 'slaves' to their ratings and reviews. In addition, several providers struggle with soft skills. Future research could focus on creative agencies to understand their views of professional platforms in order to investigate the potential for new business models to emerge.
There are some limitations of the current study that open avenues for future research. For instance, this study did not look in-depth at the views of the clients, and these deserve research attention. This would enable a comparison of perspectives and interests. We acknowledge that platform providers may not be interested in increasing the transferability of SC if they associate this with loss of profits. As this study focuses mostly on the dyadic client-provider relationships, future studies can look more closely into whether and how reputation systems are able to serve the other forms of relationships present on the platforms. To gain more in-depth insights into the use of reputation systems and the accumulation of SC, it would be beneficial to investigate the case when the first reviews received are negative rather than positive, that is, when providers start from an unfavourable position. This would help the development of techniques that providers can utilise to overcome that disadvantage. The interview data reveal potential rating inflation and rating skewness (requests to provide '5-star ratings' may have an influence here); however, there has been no systematic examination of this topic within the context of professional platforms. Future research could investigate the extent of this phenomenon and compare the situation on asset-based platforms (e.g. Uber, AirBnB), which has been documented (e.g. Zervas et al. 2020), with other types of platforms and also the effect of the presence of mutual ratings (i.e. when the provider rates the client too).