Introduction

With a growing internationalisation and privatisation of education provision and governance, education researchers need new tools and methods to make sense of how education policy and public governance are actually created and enacted. New actors participate in the management of public schools, in policy conversation and processes, operating in new spaces and scales - local, national and supra-national. The conceptualisation of the state and governments as bureaucratic, static and rational organisations is no longer viable and does not enable us to understand how new policy actors, such as new philanthropists, venture and impact investors, start-ups, international edu-businesses and international non-governmental organisations (INGOs) act as part of the governance of education. If however, we comprehend the state as constituted in and through complex and dynamic social networks, which include private actors as part of the structure that manages public services, then we can begin to grasp how policy is now being done in practice. This is what authors have been referring to as a shift from government to governance (^{Bevir, 2011}; ^{Jessop, 2002}, 2011; ^{Rhodes, 1996}). Research into this shift requires both theoretical and methodological adaptations.

Network ethnography (NE) is a responsive approach to these shifts, making use of a combination of online research, network mapping and ethnographic tools (such as interviews and field observations) to access and portray the unstable, elusive, mobile, sometimes fleeting and decidedly heterachical nature of new governance. NE is particularly productive when it involves the processes of “following policy”. Drawing from previous studies, this paper describes and discusses the approach and techniques of network mapping and ‘following’.

Changing lenses: a focus on mobilities and not structures

New actors are now participating in the management of public services and, at the same time, there is an increasingly intense mobility of goods, people, ideas around and within policy and the policy process. Policy itself has become subject to privatization and profit-making. The dynamics between the public and private, and the global and local are changing, and the notion that there is a nationally bounded policy process - the “policy-as-government paradigm” (^{Ball, 2012}) - is now outdated. While education researchers often continue to frame their analysis work within the spatial limits of nation states, many policies actually have their origins and influences elsewhere and are mediated by international organisations, mobile consultants, global philanthropists, multi-national companies, and so on. There is an emergence of “new transnational spaces of policy and new intra-national spaces of policy”, with new relationships between them (Ball et al., 2017, p. 12). Nonetheless, the role played by these policy actors is often underestimated in academic research (^{Frumkin, 2008}; ^{Olmedo, 2014}). New research strategies, methods and perspectives are needed to account for the participation of these new actors in the policymaking arena and the concomitant globalisation of policy.

There are several theoretical and methodological challenges here that must be addressed, and three can be highlighted: the perspectives of “educationism”, “statism” and “nationalism” (^{Dale & Robertson, 2009}; ^{Verger et al., 2018}). These approaches analyse education policy without taking into account wider social contexts, and they are typically limited to a focus on the state as a delimited entity - rather than as a complex and messy set of sites of and forms of power. We need to deconstruct and denaturalize the state - and take into account the diverse groups and actors that now partake in policymaking and their diverse techniques and methods of government. That is to say: “we have to study power outside the model of Leviathan, outside the field delineated by juridical sovereignty and the institution of the State” (^{Foucault, 2003}, p. 34). This does not mean that one should not acknowledge the crucial role of the state in the policy process, but at the same time we must attend to the practices or processes of knowledge production and government(alities) that bound together and articulate what we call and understand as the state.

Furthermore, as noted already, it is now necessary to “think outside and beyond the framework of the nation state to make sense of what is going on inside some nation states” (^{Ball et al., 2017}, p. 12). This implies “extending the limits of our geographical imagination”, and “attempting to grasp the joining up and reworking of these spaces in and through relationships” (p. 17). Addressing these limitations implies epistemological and ontological shifts, with less interest in structures and more emphasis on flows and mobilities (Ball, 2012).

To make sense of the shifting terrain in education policy and in order to research these shifts, we have drawn from the literatures concerning governance, networks and policy mobility. These approaches and perspectives focus on new actors that operate within governance, the mundane work of policymaking, the movement of people and ideas, and avoid “methodological territorialism” - that is, “the practice of unreflectively understanding the social world through the lens of territorial geographies” (^{Larner & Le Heron, 2002}, p. 754). They embrace the complexity and messiness and absurdity of policymaking. They also allow for a focus on mobilities and movement, on the labour of connectivity, and the mundane tasks of policymaking and mobility, rather than focusing on structures.

Besides referring to the shifting “reality” of the work of government (and its governmenalities), the concept of governance also represents a change in the theoretical perspectives about the work of the state. So governance also refers to “various new theories and practices of governing and the dilemmas to which they give rise. These new theories, practices and dilemmas place less emphasis than their predecessors on hierarchy and the state, and more on markets and networks” (^{Bevir, 2011}, p. 1). The “reified concepts of the state” are challenged, and “theories of governance typically open up the black box of the state” (p. 1). The governance literature accepts the challenge of revisiting the boundaries between the state, market and society, and explaining the complexity of its functioning with its many different participants.

Drawing from the mobilities literature, policy is considered something that extends beyond the boundaries of the nation state, thus requiring researchers to trace the global movements of policies. There are a variety of approaches and perspectives that can now be drawn upon to analyse the globalisation of education and education policy. For instance, the global convergence of policies and policy transfer. Despite the relevance and contribution of these approaches, here we focus on and use the perspective of policy mobility, in combination with the literature around governance. The policy mobility approach offers a critical perspective on the global movement of policies by drawing attention to the power relations and the labour invested in mobilising policy across countries. The analytical focus is then brought to bear on the “various network strategies through which durability and mobility is achieved, always focusing attention on the tiny, often mundane exchanges going on within the complex commotion of materials and human action that we think of as educational life” (^{Olmedo, 2017}, p. 117). In other words, “we need a more careful tracing of the intellectual, policy, and practitioner networks that under-pin the global expansion of neoliberal ideas, and their subsequent manifestation in government policies and programmes” (^{Larner, 2003}, p. 510). By paying attention to mundane practices and the labour invested in (global) networks of governance, it becomes possible to offer “descriptions of the circulatory systems that connect and interpenetrate ‘local’ policy regimes” (^{Peck & Tickell, 2002}, p. 229), and understand processes of “glocalisation” and the “actually existing neoliberalism” (^{Brenner & Theodore, 2002}).

The policy mobility approach has been developed by researchers from the field of human geography to understand how policies move globally and how they “land” locally (^{McCann & Ward, 2012}; ^{Peck & Theodore, 2010}, 2012). This is part of a “new generation of critical policy studies” (Peck & Theodore, 2010, p. 169). Though heterogeneous, these critical policy studies are rooted in critical epistemologies and assume that

policy actors and actions are understood to be politically mediated and sociologically complex. As such, the beliefs and behaviours of policy actors are embedded within networks of knowledge/expertise (many of which are translocal and transscalar), as well as within more “localised” socio-institutional milieux. (^{Peck & Theodore, 2012}, p. 23)

This perspective opposes itself to “orthodox views of policy transfer” (^{Peck & Theodore, 2010}), which assumes simple rational-choices on the part of policy actors, with a tendency for good policies to drive out bad ones, and they also tend to use positivist research methods (Peck & Theodore, 2010). In contrast, as Peck and Theodore (2010) explain the policy mobility perspective operates with different premises. First, it is assumed that policy formation and transformation are socially constructed processes and fields of power. Second, actors are not seen as “lone learners”, but members of epistemic communities. Third, mobile policies rarely travel as “complete packages”, but are moved about in bits and pieces. Fourth, policies move in a “complex process of non-linear reproduction”, mutating and morphing as they move. And fifth, mobile policies do not move across a flattened space of transaction, but they move in “cross-scalar and interlocal” spaces, increasingly complex within new forms of uneven economic development.

In sum, “in contrast with the orthodox literature on policy transfer, the governing metaphors in critical policy studies are not those of transit and transaction, but of mobility and mutation” (^{Peck & Theodore, 2010}, p. 170). Global policy networks are social and human constructions. They are complex and evolving connections between people objects and technologies across multiple spaces, far and near. Such networks are constantly made and remade, they are always under construction; “always in the process of being made … never finished: never closed” (^{Massey, 2005}, p. 9). While not assuming straightforward and lineal exercises of policy borrowing, the policy mobility approach aims to “identify the surface traces of policy flows or instances of circulating policies” (^{Olmedo, 2016}, p. 57) and the “‘local globalness’ of urban policy transfer” (^{McCann, 2011}, p. 107). The analysis then aims to attend to the “complexities of glocalization-the interplay of global forms [...] with local circumstances” (^{Ball, 2017}, p. 30). This means the approach rests on the need to think both about ways in which the “global” impacts on the “national”, and at the same time, acknowledge the extent to which the national is critical in the formation of global policy agendas. That is, the interdependency of actors and the movement of ideas across local, national and global settings (Ball et al., 2017, p. 11).

This approach is contrasted with a “nationalist” one (^{Verger et al., 2018}). In this case, the analytical framing is not defined by geographic categories and entities (a city, a state, a country...), but by the space configured by the very intersection of global and local elements (^{Ball et al., 2017}). This requires we “acknowledge that global activity can take place at multiple levels and on multiple scales, through complex, evolving network relations, with ‘domestic’ policy actors acting globally in their own right” (p. 12). We need to look at and in the diverse and different spaces of policy. Mobilities, following ^{Ong (2007}), need a “space” to be assembled. And it is in such “spaces” wherein policy takes its shape; in this sense policy is not a predetermined “vector”, but one that “carves” a space in an already crowded terrain. These are “spaces” where floating rationalities travelling around the world can find “an oblique point of entry into the asymmetrical unfolding of emerging milieus” (p. 5). These new “reflexive, self-regulatory and horizontal” spaces of governance are a relatively new histori cal formation that has repercussions for the objects and domains addressed.

The study of the mobility of policies is not abstract. Instead, besides highlighting the power relations present in the movement of policies, the mobility turn also emphasizes the role of mundane activities of policy work, the labour involved in the movement of policy. As ^{Cook and Ward (2012}) argue, “there is labour involved in creating the conditions under which a policy is more likely to be introduced” (p. 143). This work is done by “experts” or policy entrepreneurs, or what ^{Stone (2004}) calls transfer agents, who are “embodied members of epistemic, expert and practice communities” (^{Peck & Theodore, 2010}, p. 170 as cited in Cook & Ward, 2012, p. 140). Thus, “the analytical approach to what is referred to here as ‘the global’ responds to the set of discrete, identifiable and traceable practices (connections, transactions, meetings, travels, influences, and impositions, etc.), through which international economic and political relationships are enacted” (^{Olmedo, 2016}, p. 46).

All of this means shifting “from the study of mobile policy to the study of the practices through which policy is made mobile” (^{Roy, 2012}, p. 35), and “this involves both giving attention to the labour of policy actors and concomitantly thinking differently about the labour of policy researchers-how we research policy” (^{Ball et al., 2017}, p. 43). It is “by paying attention to these agents of neoliberalism [that] it becomes possible to think about how power and ‘expertise’ flow between nations and how policy entrepreneurs, NGOs, think tanks and commercial providers of education ‘do’ globalisation” (^{Exley et al., 2011}, p. 213). In other words, this is “actually existing neoliberalism” (^{Brenner & Theodore, 2002}) and processes of neoliberalisation; “what we are dealing with here are new ways of ‘neoliberalism in action’. That is, a set of practices and processes, structures, and relationships, which constitute what could be understood as ‘doing neoliberalism’” (^{Olmedo, 2016}, p. 59).

In part, the labour of policy revolves around movement, travelling and meetings. The mobility turn emphasises this movement in its multiple configurations and variations, from physical movement of materials and artefacts and people to digital or virtual movements of information, ideas, even of power. It “connects the analysis of different forms of travel, transport and communications with the multiple ways in which economic and social life is performed and organised through time and across various spaces” (^{Urry, 2007}, p. 6). Thus, the movement and labour invested in different networking activities are now a key component of the policymaking process. The “very ‘costly’ meetings, communications and travel through time-space” are central to networks, and are “necessary to ‘form’ and to ‘cement’ weak ties at least for another stretch of time” (p. 231). By “meetings” Urry refers to “both the highly formalized with ‘agendas’, structure and time-tables and the informal to where the specific space and time are planned in advance to where they are negotiated en route” (p. 232). ^{Junemann et al. (2016}) describe these as the “micro spaces” in which (global) education policy is made, which are inhabited by new actors and organisations. In these spaces of meetingness, “network members, from a range of backgrounds, come together, where stories are told, visions shared, arguments reiterated, new relations made, partnerships forged, and commitments made” (^{Ball, 2017}, p. 35). Urry’s mobilities paradigm, although primarily applied to social networks, captures “an alternative theoretical and methodological landscape” (Urry, 2007, p. 18).

All of this requires us to focus our attention on the mobile people, policies and places (^{McCann & Ward, 2012}), or the “whos” and “whats” and “wheres” of policy (^{Ball et al., 2017}). The first, looking at people or “whos”, means asking who mobilises policy, implying that “our work asks how policy actors circulate policies among cities, how they draw on circulating policy knowledge and how and for whom they put these engagements to use as they assemble their own ‘local’ policies...” (McCann & Ward, 2012, p. 42). The study of “whos” can be done by traveling or with documentary evidence of travelling, in “paying attention to the way stories about places and policies are told” (p. 48).

Second, attending to “whats”, means asking how policies are made mobile and what situations, “transit points” and “sites of persuasion” policies move through (^{McCann & Ward, 2012}). This can be done by tracing the movement and mutation of policies with documentary and interview evidence, following policy models or policy actors along the trails they leave in cyberspace, conducting oral histories of the spread of policies and attending “relational situations” like conferences.

Third, the places or “wheres” requires asking how places become attached to models, and attending to the places and events in which the “past, present and potential futures of education co-exist” (^{McCann & Ward, 2012}, p. 48) and thus accessing the mediation and interpretation of policy models at the local and national level. As regards the “wheres”, both “following policies and ‘studying through’ the sites and situations of policymaking” are needed in order to understand how education is assembled “through policy actors’ purposive gathering and fixing of globally mobile resources, ideas, and knowledge” (p. 43).

Following policy “involves close attention to organisations and actors within the global education policy field (and their movement), to the chains, paths and connections that join up these actors, and to ‘situations’ and events in which policy knowledge is mobilised and assembled” (^{Ball, 2017}, p. 32). All of this requires “staying close to practice” (^{McCann & Ward, 2012}, p. 45), in such a way that network ethnographers become what Burawoy calls global ethnographers, that is, they “become the living embodiment of the processes we are studying” (Burawoy et al., 2000, p. 4). As network researchers we travel, we attend, we meet and we network - in order to research networks. Our practice is homologous to/with the networks researched. But it also means that we “examine policy in many forms: written policies, policy models and best practices, policy knowledge, policy responses to specific concerns, and the sociospacial manifestations of policy work”, in a way it “challenges policy as technical, rational, neutral, and apolitical” (McCann & Ward, 2012, p. 42).

To be clear, the perspectives of governance and policy mobility are not only useful to the study of comparative and international education, but in any research concerning education policy in a given setting. They challenge traditional research framings that usually outline a study in a given city or country, and then consider laws or events regarding a specific policy issue (curriculum, financing, teacher training…). Instead, the focus lies on the labour, people, organisations, discourses, and places through which policies move and via which they evolve and change - recognising that often this results in incoherence rather than clarity.

In paying attention to people and policies in movement, researchers need methods that allow them to “observe” social relationships and interactions. This means “observing directly or in digitally enhanced forms mobile bodies undergoing various performances of travel, work and play” (^{Sheller & Urry, 2006}, p. 217). Options include “mobile ethnography”, “time-space diaries” and “cyber-research”, methods that stimulate the memory or methods that would capture the “atmosphere of place” or “transfer points” (Urry, 2007). The method of network ethnography, described in what follows, is a cross-fertilisation of social network analysis (SNA) and ethnography but also draws from and echoes aspects of governance and mobilities research.

Changing methods

Drawing from ^{Ball et al. (2017}), the method of network ethnography (NE) and following policy (^{McCann & Ward, 2012}; ^{Peck & Theodore, 2012}) are intertwined. In practice, the use of NE to attend to the research shifts in perspective and framing entailed in adopting a network conception of policy, requires extensive and exhaustive online searches, interviews, and participation in and observation of events. NE is a toolbox that addresses “the proliferation of policymaking sites and activities around the world and the increasing mobility and flow of education policy” (Ball et al., 2017, p. 18). It is used to investigate the interactions among public and private actors in education governance and global policy (^{Avelar, 2020}; Ball et al., 2017; Ball & Junemann, 2012; ^{Hogan et al., 2015}; ^{Olmedo, 2014}, 2017). It is a responsive and adaptive assemblage of research tactics and techniques that addresses both the organisation and processes of network relations. Attention is given to new configurations of social life and relations, which are increasingly “networked” (^{Urry, 2003}). The method follows the general principles of qualitative research, such as working in natural settings, with studies being subject to on-going design and redesign, a concern with social processes and meaning and data collection and analysis occurring simultaneously (Ball et al., 2017).

More specifically, NE proposes an ethnographic approach to networks, drawing from both classical ethnographic research techniques and Social Network Analysis (SNA). The NE method enables researchers to capture the details and meanings of policy relations, as well as the interactions, practices and meanings policy actors share, together with their direct participation in the policy process (^{Ball & Junemann, 2012}). The NE approach echoes the “inside-out” perspective of ^{Riles (2001}), with both network graphs and an analysis of the social relations that constitute the network, with an awareness of contexts and “an appreciation of the perception of the network from the inside and an appreciation of the content of the ties in terms of quality, meaning, and changes over time” (^{Edwards, 2010}, p. 24). It involves the “identification and analysis” of both the “creation and operation of global education policy networks and the connections that constitute them” (Junemann et al., 2018, p. 458). By identifying, analysing and paying attention to the insider relations and the overall network, NE “allows for greater attention to the power relations that constitute the dynamic flows of material and symbolic resources” (^{Au & Ferrare, 2015}, p. 15). This produces a map of the history of ‘effort’ (Ball, 2016).

In combining elements from formal SNA and an anthropological sensibility, networks present a triple usage: as methodological tools, as metaphors (or analytical abstractions) and as descriptions of empirically identifiable social forms (^{Knox et al., 2006}). First, as a method, the network is used for data collection and analysis. Unlike other tools and objects, the network presents an opportunity to mix qualitative and quantitative methods. Second, as a metaphor, “the strength of the network metaphor has been to encourage us to rethink questions of relatedness, and to consider how the implications of distance(s) of different kinds might be addressed by the network” (p. 134). In the case of current education policy networks, that reach new places, spaces and scales, using networks as metaphors can challenge assumptions of relatedness. Third, as a model,

the importance of the network is that it can be both a model and an object, that it can be turned, as Riles puts it, inside out. The inside of the network (the social relationships of which it is composed) is at the same time the outside (the representation or visualisation). (^{Knox et al., 2006}, p. 133)

Thus, networks support the rethinking of relatedness, distance and space, while also providing a way to go about the complexity of empirical policy settings.

The anthropological approach to networks also employs all three “usages” and aims to record both the narrative and discursive aspects of networks. The process of research is reversed in comparison to formal SNA procedures. Instead of identifying “boundaries” and “populations” first, the researcher begins with the discourses and narratives of a social group. In this sense, “the boundaries for such domains can be identified only through the ‘stories’ which are associated with them, with discourse identifying the ‘insiders’ as those who belong to networks, their roles and identities” (^{Knox et al., 2006}, p. 129). The idiolect of the network defines its membership and its limits. Thus, the discourses and narratives identified both provide a description of the network and constitutes its analysis, as it plays out the complexities and hybridities of social relations among members. From the analytical description, the researcher can also develop a critical account of power relations, or “the place of discourse and narrative through which networks are produced” (p. 132). In this case, the interest is less in defining or delimiting networks, but rather “in tracing these definitions and operationalisations of the ‘network’ as an ideal-type or form of relating” with “detailed ethnographic description of the ways in which people articulate their relationships with one another as network relations, and are able to envisage those relationships through the use of pictorial and diagrammatic representation of networks” (p. 133).

The formal procedures and research concerns of SNA become less important in this approach. Measures, identification of boundaries, appropriate sampling and missing data are not the primary focus. In a structuralist epistemology, missing a node or an edge would jeopardise all measures and explanations of a network analysis. Instead, here meanings and interpretations have primacy. The researcher has access only to traces of unstable, opaque and ever-changing networks, which can only be made sense of by grasping how they work. So networks are addressed “narratively” and “discursively”. There is a focus on the network histories and their evolution. The graphing, so indicative of network analysis, is used as a research tool to support the “following” of people and tracking of policy developments. A network is not a research “product”, “output”, or the end point of analysis in itself, it is not simply a matter of finding network structures and applying measures as in more orthodox versions of SNA (^{Avelar et al., 2018}).

Despite its potential, the use of network methods and analysis presents limitations and dangers. First, there is a long-term tension between formal and anthropological approaches to networks. There is no single and simple theoretical foundation (^{Knox et al., 2006}). For instance, while cultural SNA writers still strive to formalise their understandings of networks with the support of mathematical techniques, anthropological writers do not share this concern (^{Edwards, 2010}). Second, network metaphors present the danger of becoming “descriptors of structures rather than heuristic devices”, or an “explanatory device” (Knox et al., 2006, p. 134). Third, the network, used to explore mobility, might in turn create new fixities and rigidities, in a different form. In this sense, networks “can challenge the received under-standing of the spatial and relational dimensions of social life but, [...] as soon as it stops challenging and starts prescribing, then the productive capacity of the network is diminished” (p. 134).

The practice of network ethnography and following policy

In practical terms, NE involves four main activities, as outlined by ^{Ball and Junemann (2012}): internet searches, interviews, field observation and graph building. These activities do not imply fixed steps in a neat process. Instead, as the studied networks are “always in the process of being made ... never finished; never closed” (^{Massey, 2005}, p. 9), the method must be able to accommodate the complexity of the networks and its constant changes. There is a continual interplay, over time, between the activities. Network ethnography is then “necessarily open and flexible” (^{Olmedo, 2014}, p. 576) in its operation, as the researcher needs flexibility to explore relationships in the network as one finds them during data collection. Indeed, “the method is appropriate inasmuch as it is flexible, evolving and adaptive in bringing ethnographic sensibilities to bear to the portrayal and analysis of the complex, translocal, evolving and multimedia relations that constitute global policy networks” (Ball et al., 2017, p. 19). The flexibility, adaptability and responsiveness of the method is also reflected in its understanding of settings and boundaries.

The process starts with extensive and exhaustive internet searches around the primary actors (or nodes) of the studied network. This is based on the “recognition that as our research settings are multiple, fluid and evolving, they are also to a large extent virtual” (^{Ball et al., 2017}, p. 18). Typically, researchers conduct deep and extensive Internet searches focused on institutes and foundations operating in networks of education governance. This requires visiting countless webpages, personal CVs (especially from actors highly connected and “boundary-spanners”, subjects that move or moved between public and private sectors), and viewing newspapers and related social media, blogs and documents. Internet archives can also be useful⁠ to access previous versions of organisations’ websites. Network graphs are created and worked on throughout this stage and are used to inform decisions concerning data collection and analysis (whilst remaining open to changes are research unfolds).

These Internet searches are a “snowballing” procedure, moving from webpage to webpage, onwards and outwards until connections fade and language changes. At first, exploratory searches are conducted with no specific limitations, searching for traces of relationships. Once something is identified (a person, an institution, a policy, a story, an event...), this is followed in an attempt to gather data, with details, evidence and versions of it. Moving forward on an online search can mean following a link that is available at the page or conducting a new search with an Internet search tool.

Data collection and analysis thus happen simultaneously, by choosing what should be followed and what should not, analytical decisions are made and insights gleaned. Besides taking screenshots, saving PDF files, highlighting these and adding information in spread sheets, there are questions being asked and decisions being made regarding where to proceed next. These network nodes have depth - they are constantly evolving and are not static as graph visualisations can sometimes imply. When visiting webpages, the aim is to understand the organisation at stake - what it does, what are its values, who works for it and with whom they collaborate and how it is funded. In this process, finding and understanding connections and relationships is just as important as understanding the institutions and people themselves. The names of people and institutions may not be very telling at first, as the researcher does not know them a priori. However, as one visits webpages and lists the people and organisations, patterns start to become apparent and key actors start to be repeated. This is one benefit of drawing on SNA. It is important to stress here, however, that the identification of such actors does not necessarily rely on quantitative methods, on identifying, for instance, nodes with higher values of density or centrality - although this can be useful when analysing network graphs. Instead, it is a qualitative perception of relevance that is to the fore. One identifies relevant actors by seeing them repeated in different relevant situations, reports, boards, events, policy stories and so on. This also serves to identify institutions that often work together, and on what type of initiatives, with what type of strategies and mobilising which discourses.

Second, when possible, interviews are conducted with individuals and in institutions identified as highly connected, or relevant and influential. The interviews aim to add meaning and context to the data collected online and clarify network relations. Policy actors are not necessarily asked the same questions, but rather questioning is based on the online findings, about the projects and relationships fostered, each interview is curated in relation to local specifics and personal histories. Online searches are thus vital for the creation of relevant questions and collection of pertinent and meaningful data. Additionally, each interview influences the following one as the researcher progresses with data collection. There is no assumption or intention of keeping oneself unaffected by previous conversations. In fact, some interviews are chosen and carried out with basis on anterior questionings, including recommendations from interviewees about who should be interviewed next. This enables further clarification and confirmation of materials collected previously. As one progresses, some themes become clearer, and the relevant actors are usually confirmed.

Third, when possible, researchers participate in and observe events as key sites of policymaking and network maintenance. The relations analysed are often so opaque and blurry within widespread and evolving networks that one cannot attempt to make sense of them without “entering” them. And one only enters them by partaking in the very spaces and relations one is studying (^{Ball & Junemann, 2012}). Following policy, as part of network ethnography, involves the observation of large events, small meetings and other situations and spaces that may concern the networks being analysed - of course beyond these there are many inaccessible and unknown - backstage- sites that may be of importance. The fieldwork observations provide an ethnographic experience that may allow the researcher to better understand the networks by observing the spaces, the corporate decorations of offices and events, the small talk held in between seminar talks and how people behave in policy spaces. Field notes and pictures can be used in the analysis, together with the data collected online and with interviews.

When participating in meetings and events, the fieldwork involves paying attention to the institutions and people that fund the event, observing the physical organisation of the conference space (with stands, publicity, folders...), listening to the references mentioned in presentations and in informal conversations during the conference. Concomitantly, informal dialogues are also helpful and can support data collection decisions, such as identifying some references and leaders of the network. Coffee breaks can be key moments for data collection. These events also help the researcher to develop a vocabulary that can be fed into interviews: non-state organisations have their very own idiolect to refer to practices and discourses that are named differently in academic research. Again, the use of this insider language can support the creation of a non-threatening environment for interviews, and relatable and intelligible interview questions. Besides physical visits, some events can be “attended” or watched virtually.

However, we reiterate that this does not imply a process without challenges and contradictions in the use of NE. Indeed, the application of NE entails a series of invisible and intricate processes that are often silenced in methodological discussions, which are particularly sensitive in ethnographic fieldwork moments. There involve the power relations in which the researcher is embedded, and their intersectional positionality influences the modes of access, relationship with, and the writing about political elites. The method's flexibility and openness may lead to uneasiness and uncertainties encountered by researchers in their methodological practice (^{Avelar et al., 2021}).

Throughout the three activities, policy network graphs are built as tools to identify relevant individuals, institutions and relationships in relation to specific policies or networks. In these activities, the researcher pays “close attention to organisations and actors within the global education policy field, to the chains, paths and connections that join up these actors, and to the ‘situations’ and events in which policy ideas and methods are mobilised and assembled” (^{Ball et al., 2017}, p. 19). The graphs are fundamental to keeping track of the relationships found between people, institutions, policies and events. Graphs can make possible the visualisation of the emergence of connections and more defined relations and communities in messy and complex networks, which then inform choices as to how to proceed with data collection and analysis. As data collection demands a continuous back and forth movement, the graphs also support the identification of what is relevant and the pursuit of a route, or following these people, institutions and policies.

In network ethnography, the graphs generated are often used flexibily. Unlike formal SNA, in NE graphs are research tools to support and inform data collection and analysis, rather than an end in themselves, they should “be viewed as descriptive devices rather than analytical representations” (^{Hogan et al., 2015}, p. 44). It is worth re-stating the artifactual limits of network graphs. The network is a heuristic for understanding governance and policy mobility and the construction of a new policy dispositive. It is not exhaustive. That is, in good part, as ^{Temenos and McCann (2013}) point out, a consequence of limited resources set over and against the evolving, dynamic and mutating nature of the network. With more time and more money, researchers can follow links and relationships further, through more disparate nodes, to more distant and more local points. Nonetheless, the relationships and connections are continually expanding in number, scope and location, so networks are inevitably outdated by the time of publication.

Although limited, the graphs not only have a pragmatic role of informing data collection and analysis, but they also operate in a symbolic and creative way that “foregrounds the importance of network connections” (^{Hogan et al, 2015}, p. 44). Network graphs remind us how connected and cross-scalar education governance is now, and in spite of limitations, are still useful, especially if compared to text and tables.

Final remarks

The key point here is a simple one - method must fit topic. If we are to make sense of the emergence of network governance, a set of complex heterarchical social relations that extend beyond the state and the nation, then we need to adopt methods that are fit for purpose. The flexibility and agility of network ethnography is well suited to making sense of the elusive and evolving forms of new governance and its diverse locations and participants. Network ethnography can get us to the people and places through which the policy process moves. It is attuned to movement, to instability and to evolution.

However, studies employing NE rely, in many cases, on individual researchers or small and underfunded teams. While much can be achieved with limited funding, establishing better-funded research groups and laboratories would be an appropriate response to the complexities of network governance and the financial aspects involved in this domain and in the “policy work”. The use of more flexible and agile methods, such as NE, to study current educational policy processes and governance, can offer valuable contributions to the field, with empirical, methodological, theoretical, and political implications. In other words, it can assist us in advancing our knowledge about new forms of educational governance, how we study them, how we explain them, and in elucidating the power relations that constitute this new, unstable terrain of educational policy.