Scaling Social Innovation: A Cross-Cultural Comparative Study of School-Based Mentoring Interventions

Social innovation is increasingly recognised as a powerful vehicle to address unmet societal needs. Nonetheless, research into how contexts and agency interact to determine outcomes/risks of its scaling appears still limited. This study draws on structuration theory to fill some gaps. By comparing two school-based mentoring interventions, it first shows that comparable external catalysts can trigger diverging ambitions and paths to scale. Second, it finds that certain strategic/agentic choices (e.g. entrepreneurial, political, coalition-building skills; evaluation), combined with specific contextual features, help to achieve more rapidly a larger scale of expansion. Finally, it highlights which risks, in this field, more strongly relate to agentic and/or contextual factors, providing insights for research, policy and practice.


Introduction
Social innovation, understood as a novel way of working that fosters collaborations between civil society, private and public actors, is increasingly promoted, in public policy debates, as a vehicle to developing innovative and efficient solutions to address societal needs (Sinclair and Baglioni 2014;Steiner et al. 2021). Hence, its scaling, intended as 'the process of expanding or adapting an organisation's output to better match the magnitude of the social need or problem being tackled' (Desa and Koch 2014, 148), has recently become a focus for many (Brandsen et al. 2016;Steiner et al. 2021). Still, given the novelty of the social innovation concept, the body of literature currently available about its scalability remains sparse (Bolzan, Bitencourt, and Volkmer Martins 2019) and the mechanisms behind scalability scarcely researched and understood (Brandsen et al. 2016;van Lunenburg, Geuijen, and Meijer 2020).
A societal issue of increasing concern to several European countries, especially after Covid-19, is combating educational poverty and inequalities of access to learning opportunities among youth (Kraft and Falken 2021). Eurostat statistics (Eurostat 2021a(Eurostat , 2021b show that, in recent years, some European countries had particularly high rates of premature school dropouts and of young people neither in employment, education or training (NEETs). School-based mentoring (SBM) is increasingly widespread as an intervention to mitigate youth educational disengagement and its long-lasting scarring effects. It rests on the establishment of supportive relationships between students and caring non-familial role models who help mentees navigate more successfully their educational and social-emotional development (Spencer 2012). As such, SBM programs are fully-fledged social innovation initiatives. Indeed, among the broad range of coexisting interpretations of the concept of social innovation (Ayob, Teasdale, and Fagan 2016), some particularly focussed on the processes it triggers to lead to transformed social relationships among system agents (Brandsen et al. 2016;Gabriel 2014;Gerometta, H€ aussermann, and Longo 2005). According to such views, social innovations would strengthen system capability, social assets and resource utilisation precisely by promoting the rise of inclusive and collaborative delivery and/or governance networks. This conceptual line fully captures the socially innovative nature of SBM. Indeed, SBM entails a radical departure from the dominant paradigm: with SBM, youth educational and wider developmental needs are no longer exclusive prerogative of schools and families. A third party (namely, a mentor) enters the equation, bridging between two sides of the support system (Jones, Doveston, and Rose 2009). Furthermore, the effective provision of SBM initiatives cannot take place without engaging multiple actors in a collaborative effort. At a minimum, coordinating NPOs partner with hosting schools and persuade communities to contribute volunteer workforce (Jocovy and Garringer 2007). At best, broad-based and internally more varied networks are established to allow the initiative to acquire the visibility and resources critical to its success (e.g. Fassetta, Siebelt, and Mitchell 2014;MacRitchie 2019).
In response to the need for extending research on social innovation scalability, the current study examines how two SBM programmes, respectively promoted in Scotland and Italy, identified and responded to context-specific obstacles and facilitators to scale. More specifically, it draws on structuration theories (Cajaiba-Santana 2014) to explore how the mutual influences between contextual conditions and agentic choices (i.e. scaling strategies) determine which scale up outcomes are attained, as well as which barriers and unintended consequences are faced while scaling initiatives.
The article is organised as follows. First, research concerning social innovation scalability is briefly reviewed, highlighting the advancements promised by structuration theories when it comes to understanding outcomes, barriers and unintended consequences of scaling. Then, the two SBM programmes considered and the methods employed are described. Next, findings are illustrated. First, the article presents how contextual conditions influenced agency for the cases analysed. In this regard, it initially describes those macro or meso-level catalysing factors thatinteracting with the individual attributes, sensibility, motivations, cognitive structures, and ability to think reflexively of the social innovatorssparked the opportunity identification process (Cajaiba-Santana 2014; Edgecombe et al. 2013;Perrini, Vurro, and Costanzo 2010;van Wijk et al. 2019). Next, it documents how scalability was affected by policy and cultural contexts: the set of pre-existing and relatively enduring social and institutional structures (e.g. relationships, normative systems, rules, beliefs, values, habits, routines) which existed prior to the introduction and dissemination of the social innovation (Porter 2015). Second, it provides an account of the agentic choices made by the initiatives to shape their scaling environment. Hence, it outlines the combination of mechanisms adopted to act upon contexts, challenge the pre-established order and promote social change on an increasingly larger scale (Perrini, Vurro, and Costanzo 2010;Porter 2015). Third, the article illustrates how context and agentic responses, in their interaction, resulted in the two scaling organisations encountering and dealing with specific barriers or unintended consequences. Finally, it discusses key study implications for research, policy and practice.

Background
Despite the importance of social innovation scalability for policy and practice, only a narrow body of theoretical and empirical research has explored this topic. Moreover, literature on the subject has long remained dichotomised between individualistic or agentic theoretical perspectives and institutional approaches (Baptista et al. 2019;Cajaiba-Santana 2014;Cuntz, Foray, and Mostovova 2020), limiting our understanding of how the interrelatedness between human agency and institutional and social systems affects social innovation scalability.
The former theories see agents' behaviours, choices, values and attributes as the primary determinants of social innovation. Resulting research primarily focuses on scaling 'pathway choices' or stages (S anchez Rodr ıguez, MacLachlan, and Brus 2021). While some scholars described the varied strategies or tactics used to scale (Dees, Anderson, and Wei-Skillern 2004;Larson, Dearing, and Backer 2017;Moore, Westley, and Brodhead 2012;Mulgan et al. 2007;Sabelli and Harris 2015;Waitzer and Roshan 2011;Westley et al. 2014), others saw choosing a route to scale as just one of the decision-making processes that take place over a multi-stage pathway (Elkington, Hartigan, and Litovsky 2010;Gabriel 2014;Milat and Bauman 2015;Murray, Caulier-Grice, and Mulgan 2010). Although this strand of literature pays attention to the role of some organisational-level aspects, influential contextual conditions remains largely neglected. Conversely, within institutional theories, a prominent role for the causation and diffusion of innovation is attributed to structural or contextual determinants. Hence, too limited attention is given to how agentsembedded in complex institutional environments and social structuresact upon them, either for their validation or to bring about change (Cajaiba-Santana 2014). On the cusp between these two dominant paradigms, structuration theories focus on the iterative and reciprocal influences between individual/collective actions and social-institutional structures (i.e. context), seeking a more comprehensive explanation of the processes of creating and scaling social innovation. According to such more holistic views, both internal and external factors can be conducive to the expansion of social innovations (Bloom and Chatterji 2009;Bloom and Smith 2010;Perrini, Vurro, and Costanzo 2010;Verver, Dahles, and Soeterbroek 2021). Indeed, as some recently argued (Islam 2022;Loukopoulos and Papadimitriou 2022;Roundy and Lyons 2022), not only the context holds potential to affect an initiative's scalability but scaling organisations can also adopt a set of strategies with a view to shape an ecosystem more conducive for an initiative to scale (hence, acting indirectly to create more favourable contextual conditions). Here below, we succinctly overview extant knowledge about how: (i) context influences agency; (ii) agency influences context; (iii) the mutual influences between context and agency translate into scaling barriers and unintended consequences. Then, we highlight why structuration theories may provide a way forward to overcome some research gaps.

Influence of Context on Agency
Structuration views assume that contextual or external catalysts contribute to stimulate transformative changes. Among the stimuli identified, it is worth mentioning: policy mandates, financial pressures (due to shrinking public budgets) or, conversely, available funding opportunities (Edgecombe et al. 2013) and market failures (Cuntz, Foray, and Mostovova 2020). The very existence of persistent social issues and the awareness of the need to address them are also believed to ease the scaling process as well, generating a strong demand for value-creating innovations (Mulgan et al. 2007;Perrini, Vurro, and Costanzo 2010;WHO and ExpandNet 2009).
Existing research has also widely acknowledged the importance of a supportive policy environment for an innovation to scale. A strong political will to address the social need, as well as policy settings that incentivise individuals or institutions to jointly pursue public and private interests and favour experimentation can crucially foster scalability (Cuntz, Foray, and Mostovova 2020;Han and Shah 2020;van Lunenburg, Geuijen, and Meijer 2020;Verver, Dahles, and Soeterbroek 2021). Also, while open, participatory bureaucratic and administrative systems facilitate scalability (WHO and ExpandNet 2009), excessively strict rules, alongside a cumbersome bureaucracy, can frustrate scaling efforts (van Lunenburg, Geuijen, and Meijer 2020).

Influence of Agency on Context
With respect to the ways in which initiatives can act upon contextual characteristics, a first prominent theme pertains to leadership, vision, values and culture. Several contributions have seen the leadership, entrepreneurial, resource-mobilisation and political abilities/skills of the actor who initiates the process as highly influential on scalability (Bolzan, Bitencourt, and Volkmer Martins 2019;Han and Shah 2020;Perrini, Vurro, and Costanzo 2010). Often, as a reflection of the initiator's scale ambition, different scaling pathways ensue (e.g. Westley et al. 2014). Leaders may also define the vision and institutional logic guiding the scaling process, as they can intend their mission as directed to scaling an innovation out and/or deep (Bloom and Chatterji 2009;Desa and Koch 2014;Moore, Riddell, and Vocisano 2015;Zhao and Han 2020). In the case of scaling out, the best way to create social value is 'social spillover' (Zhao and Han 2020): serving as many locations, people and communities as possible. In the case of scaling deep, the aim is a smaller-scale but profound change of 'people's hearts and minds, their values and cultural practices, and the quality of relationships they have' (Moore, Riddell, and Vocisano 2015, 74). Furthermore, leaders' characteristics and mindsets have been acknowledged to influence their resource seeking and contributing behaviours, which in turn can affect the attributes of the surrounding ecosystem. For instance, Roundy and Lyons (2022) argued that some social entrepreneurs are more open to seek the assistance, resources and legitimation of other system players to scale their ventures. Simultaneously, they are more likely to pursue continuous learning (even through mistakes), as well as more eager to give effect to their interdependence-orientation by reciprocating and giving back to their ecosystem.
Another noteworthy family of themes relates to the mechanisms adopted to acquire allies and, consequently, resources. There seems to be broad consensus on the importance of building coalitions with a plethora of stakeholders (e.g. wider local communities, participants, volunteers, employees, investors, other NPOs, large corporations and the government) to enhance scaling (Bloom and Chatterji 2009;Bloom and Smith 2010;Bolzan, Bitencourt, and Volkmer Martins 2019;Han and Shah 2020;Perrini, Vurro, and Costanzo 2010;Scheuerle and Schmitz 2016;Voltan and de Fuentes 2016;Verver, Dahles, and Soeterbroek 2021). Prior research has also pointed to a key role allies can play in going forward on a large scale: securing resources (Han and Shah 2020;Islam 2022;Loukopoulos and Papadimitriou 2022;Scheuerle and Schmitz 2016;Voegtlin et al. 2022;Voltan and de Fuentes 2016). In this respect, some scholars (Islam 2022;Loukopoulos and Papadimitriou 2022) identified a set of actions (e.g. coalition and advocacy work, gaining legitimacy and enhancing organisational learning processes by creating and sharing of knowledgeamong the others) as pursuable strategies to support ecosystem development and indirectly foster innovations' scalability. Among the variety of players, governments deserve greater attention, becauseas some stressedenlisting them as an ally can strongly boost scalability (Baptista et al. 2019;Cuntz, Foray, and Mostovova 2020;Islam 2020;Loukopoulos and Papadimitriou 2022;Pandey, Menezes, and Ganeti 2017).

Unintended Consequences or Barriers Deriving from Context and/or Agency
A growing stream of research investigated the barriers and unintended consequences faced while trying to reach a larger scale. First, scaling can be hindered by the inability to mobilise adequate resources (Braga, Proenc¸a, and Ferreira 2014;Scheuerle and Schmitz 2016;Welty Peachey, Cohen, and Shin 2020) referred to it by Zhao and Han (2020) as a 'shortage of funds for organizational survival and growth' (145). Second, an 'operational risk' can undermine the ability to deliver the desired social impact (Zhao and Han 2020). This pitfall can trace back to an 'internal misperception', wherein implementers misunderstand the model to be scaled and deviate from its constituent elements, implementing it inconsistently. When innovations scale fast and wide, this tension may arise from time-management issues and trade-offs (Braga, Proenc¸a, and Ferreira 2014;Smith, Gonin, and Besharov 2013): the short-term goal of broadening rapidly implementation and operational capabilities conflicts with longer-term objectives of securing an adequately skilled staff (Scheuerle and Schmitz 2016). Tensions of this kind may also relate to the management of wide-ranging partnerships (Smith, Gonin, and Besharov 2013;Voltan and de Fuentes 2016;Westley et al. 2014), resulting in a possible lack of integration, ownership, implementation fidelity and value consistency. Along these lines, some scholars (van Wijk et al. 2019;Voegtlin et al. 2022) argued that an excessive multiplicity/diversity of actors to co-opt can negatively affect the ability of ventures to gain traction and diffuse. This particularly holds true for intervention contexts featuring not only high multiplicity but also high institutionalisation (i.e. established modus operandi so ingrained that there is not sufficient room to innovate). Finally, given the inherently value-laden character of grand societal challenges (Voegtlin et al. 2022), scaling can be limited due to a lack of credibility (Braga, Proença, and Ferreira 2014;Welty Peachey, Cohen, and Shin 2020), 'admission' (Scheuerle and Schmitz 2016) or a 'legitimacy risk' (Zhao and Han 2020), which occurs if stakeholders deny support to widen adoption precisely because the innovation is perceived as incoherent with existing socio-cultural norms and established practices. In this perspective, advocacy work can contribute to undermine the legitimacy of scaling endeavours (Islam 2020) by triggering hostile reactions by governments (e.g. they perceive the initiative as a threat to national interests or counteract the involvement of other actors in welfare provision and political advocacy). By contrast, dependency on the public purse may represent a considerable risk to the detriment of socially innovative initiatives' sustainability (Baptista et al. 2019) or be irreconcilable with cultural/organisational norms that value independence and autonomy (Scheuerle and Schmitz 2016).

Knowledge Gaps and Ways Forward
Overall, prior research is often limited in that little attention has usually been paid to exploring how the interdependencies between systemic-level contextual conditions and scaling efforts themselves affect the outcomes (either positive or negative) of these latter. Furthermore, the role of some exogenous factors (such as, the sociocultural system) or agentic strategies (particularly, monitoring and evaluation -M&E), appears to date largely understudied. Additionally, the area of inquiry concerned with the unintended consequences of and barriers to scaling has overly focussed on the downsides inherent to chosen organisational growth strategies (e.g. Westley et al. 2014;Zhao and Han 2020), overlooking the pitfalls that the broader context or ecosystem growth strategies can in turn conceal. Structuration theories provide a promising lens to explore whether the hindrances to scaling processes, rather than exclusively 'inherent in the chosen change strategies' and 'pathway for scaling up' (Westley et al. 2014, 244), stem out where strategies and context intersect (Islam 2020).
In light of the shortcomings just outlined, our study seeks to explore how the mutual influences between agency and context translate into scaling outcomes, barriers and unintended consequences. The theoretical foundations for the empirical analysis are given the frameworks made available by Koorts et al. (2021) and Zamboni et al. (2019), which both extensively build on the WHO and ExpandNet (2009) framework. The latter encompasses a wide range of elements which, interactively, impact on the chances of attaining scale up goals. These factors cluster in the core areas below, relating to the attributes and features of the: Innovation (such as, being credible and relatively advantageous); Implementers and advocates/champions (e.g. culture or implementation capacity of adopting organisations); Context (hence, restraints or enablers such as political will, bureaucracy or the attitudes, values and needs of adopting communities); Resource team, which engages in purposive efforts to introduce and take the innovation on a large scale (i.e. reliance on a unifying vision or on motivated and credible leaders); Ongoing strategic management (hence, the actions undertaken for advancing and monitoring continuously the scaling up process, such as building coalitions, promoting the contextual adaptation of the innovation, or detecting and managing potential trade-offs or imbalances).
This conceptual device, albeit originated in the field of health innovations, was chosen because of its compatibility with frameworks for innovations in the education field (e.g. Edgecombe et al. 2013;Milat and Bauman 2015) and its exhaustiveness. As such, it promised greater analytical depth and coverage of conceptual dimensions beyond those identified in the review supplied above. In particular, within such framework (see supplementary materials), special emphasis is placed on the potential influence of M&E and the very compatibility of innovations with established socio-cultural paradigms over scalability. Moreover, it stands out for its balanced focus on individual-, organisational-and ecosystem-level factors (Scheuerle and Schmitz 2016;Verver, Dahles, and Soeterbroek 2021;van Wijk et al. 2019) and, hence, on both the agentic and contextual determinants of scalability. In this respect, for instance, great attention is given to the reactions and agency of those adopting an innovation or promoting its diffusion. This was deemed particularly valuable in achieving a fuller view of why certain initiatives succeed in gaining traction and becoming institutionally embedded, while others struggle or fail.

Methods
The study focuses on two SBM interventions, promoted by the Scottish NPO MCR Pathways (MCR) and the Italian NPO Societ a Umanitaria (SU). The MCR programme, set up in Glasgow in 2007, supports secondary school students whose ability to engage in education is impaired by the disadvantage and home instability they experienced, including pupils in the care of local authorities (referred to as care-experienced or looked after children). Its provision involves NPOs, schools, community members, forprofit organisations, local councils andfrom the 2021the Scottish Government. SU, established in Milan in 1893, offers a wide range of socio-cultural activities to promote social dignity through education. The SBM scheme entered this wider portfolio in 2003, targeting pupils in elementary and middle schools of various regions, and is provided mainly with the involvement of NPOs, schools and community members.
In terms of context, the two countries were chosen as presenting vasty dissimilar features. They, in fact, represent different typologies of welfare regimes (Esping-Andersen 1990;Ferrera 1996;Salamon, Sokolowski, and Anheier 2000). Scotland embodies a liberal regime, closer to North America, where SBM originated, while Italy represents a 'Mediterranean' welfare state with a strong family ethos (Reher 1998), wherein the mentor role risks being more easily perceived as illegitimate and in competition with kinship roles (Molpeceres, Pinazo, and Aliena 2012). Moreover, the Italian education system displays greater bureaucracy/institutionalisation than the UK one (i.e. lesser pedagogical autonomy of teachers, stronger academic orientation of the curriculum and path-dependence) and lower multiplicity or pluralism of the educational provision (greater reliance on state-provided education) (Beblavy, Thum, and Veselkova 2011;Mincu 2015).
Furthermore, the routes through which these SBM schemes have been taken to a larger scale are almost antithetical. As Figure 1 shows, MCR started scaling out the programme fairly recently (i.e. 2019) and, since then, expanded fast and wide. Indeed, the expansion involved, by 2021, 10 local authorities and a total of about 1400 volunteer mentors. Also, MCR replicated using branching, with 'the creation of local sites through one large organization' (Dees, Anderson, and Wei-Skillern 2004, 28). Additionally, MCR has recently shifted to vertical scaling (WHO and ExpandNet 2009), since the Scottish Government formally decided, in early 2021, to institutionalise and sustain a nation-wide rollout of the intervention (Scottish Government 2021).
As to SU (Figure 2), the scale out beyond Milan started soon after the onset of the scheme (i.e. 2005) and, until 2013, only branching was used, introducing the programme in its pre-existing headquarters. Afterwards, SU also resorted to affiliation (a 'formal relationship defined by an ongoing agreement between two or more parties to be part of an identifiable network' -Dees, Anderson, and Wei-Skillern 2004, 28) to formally outsource Main stakeholders involved: NPOs, schools, community members, for-profit organisations, local councils, Government

Data Collection
A mixed-method data collection and analysis was carried out. First, nine in depth semi-structured interviews were undertaken, involving five managers from different hierarchical levels and functional areas for MCR (top-level n ¼ 3, middle-level n ¼ 2) and four managers operating in the local teams of different geographic areas for SU (middle-level n ¼ 4). An expert, purposive, heterogeneous sampling was adopted (Etikan, Musa, and Alkassim 2016). The interview guide was designed consistently with the WHO report for the development of scaling up case studies (Fajans et al. 2007) and sought to elicit the participants' views about: How the expansion was conceived and operationalised; Which factors affected the most scaling path and outcomes; Which actions and strategies were undertaken to act upon the contextual factors; The steps/challenges/risks (faced or envisaged) of sustaining the scaling-up.
Additional primary data were drawn from semi-structured interviews with 20 mentors volunteering within the two schemes (MCS n ¼ 12, SU n ¼ 8). Full details on the sampling strategy, and the interview guide's purposes and contents can be found elsewhere (e.g. Bufali, Connelly, and Morton 2023). Although such interviews aimed at grasping mentors' subjective thinking about their wider volunteering experience (i.e. motives, positive/negative outcomes, facilitators and barriers), they could be leveraged to shed light on which factors or strategiesin their perspectiveaffected the most their initial and sustained commitment to the programmes. In both cases, interviews were carried out in participants' native language (either English or Italian), audio-recorded, transcribed verbatim, and subsequently analysed using the qualitative data analysis software NVivo 12.
Additionally, a document review was also conducted, consisting of the thematic analysis of 46 sources (full list in supplementary materials), among which: previous annual reports, institutional web pages, publications (e.g. research articles, evaluation reports, books, magazine articles) and communication materials (e.g. newsletters, videos, podcasts). These documents were used to substantiate aspects of how the programmes were scaled and which factorsat the micro, meso and macro levelaffected their scalability, allowing to incorporate the views of a wider range of stakeholders (e.g. school personnel, programme managers, volunteers, served youth and their families, civil servants, policy-makers, etc.).
Ethical approval was obtained from the University of Strathclyde Department of Management Science Ethics Committee.

Data Analysis
The coding procedureapplied to both primary and secondary dataled to rearrange original sub-themes into three overarching themes: Organisational characteristics of the promoting organisation and responses activated to scale; Contextual characteristics (assets and/or deficits); System agents' responses to the combination of abovementioned factors.
Initially, the first author thematically analysed interview data. In a second round, the second author validated the coding in a blind process. A similar approach was taken for the document review. When illustrating findings, verbatim quotes and/or appropriate bibliographic references are included to substantiate authors' interpretation. For quotes or concepts drawn from reviewed documents, whenever possible, it is specified to which stakeholder they are to be attributed.

Findings
The narrative synthesis of results presents: first, how structural features influenced agentic choices; second, the agentic choices made by the two initiatives to shape a context more conducive to scaling; third, the resulting unintended consequences of or barriers to scaling encountered. The illustration of agents' responses is further structured in three sub-sections: leadership, vision, values and culture; M&E; allies and resource acquisition. Evidence for MCR first and, subsequently, for SU will be presented.

Influence of Context on Agency
The MCR scale-up journey literally originated from failure. Indeed, what triggered its founder was the frustration with the status quo, based on the experience of leading the turnaround of three failing care organisations in England in the early 2000s (MacRitchie 2019). That experience, indeed, led him to realise that the extant support system for care-experienced young people was expensive and, nonetheless, fundamentally unable to prevent young people in the care system from getting poorer outcomes than peers. In fact, as he explains (Programme Founder -MacRitchie 2019, 95): Helping the young people [ … ] often led to them being 'moved on' to new homes. Budgets and the need to reduce costs determined decisions. The cruel reality for the young people was often literally a black bag and effectively another change of parents. Instabilities layered on instabilities. It made no sense.
For these reasons: '[ … ] MCR Pathways came [ … ] from failure, because sorting out those organisations took five years [ … ] and it made no difference to the young people'. [Manager 1]. The sudden realisation of the system failure, alongside the awareness of the attainment gap it fuelled, was at the heart of the decision of the founder to find a way to change the system: 'MCR Pathways was created from frustration and system failure and … just from "It has to be a better way"' [Manager 1]. This trigger ignited a demand-driven expansion, which used a 'greatest need phase in' design (Programme Evaluator -Biggs et al. 2020, 20) to prioritise high-needs schools and local areas where the scheme could make the most impact: '[ … ] you've got to go where you think the programme's most needed' [Manager 4]. Study participants from MCR described a number of enabling factors, rooted in the Scottish policy context, which enhanced scalability and provided opportunities to build up a model that might address the needs of young people. Both 'A Blueprint for Fairness: The Final Report of the Commission for Widening Access' and the Children and Young People (Scotland) Act 2014 were identified as crucial milestones (Programme Founder -MacRitchie 2019). While the former 'set an ambition to close the educational gap between the nation's most disadvantaged young people and their peers' (Programme Founder -MacRitchie 2019, 103), the latter fortified the support system available to care-experienced young people by investing a variety of publiclyfunded organisations with so-called 'corporate parenting' duties. In other words, as reported by Manager 2, this Act has set: 'a legal obligation [ … ] to provide [youth] with an education and to provide them with all the support to give them the best possible start and ongoing chances in life'. Additionally, 'The Promise', a recent report of the Independent Care Review, embodied a strong political will to ensure that the required actions to meet the developmental and educational needs of all children are undertaken. Last but not least, as Manager 3 acknowledged, the Scottish Government further drove change using procurement to incentivise corporate social responsibility efforts from businesses. At the same time, the policy context also hindered the development process, at least in the early stages, due to the lack of adequate and stable investments in new services acknowledged by Manager 1: 'Fundamentally there isn't enough money in the system [ … ]. So that is pressure, pressure on save the money'.
In addition, a climate of scepticism characterised the introduction of the social innovation initiative, with individuals being quite wary of its results, as Manager 5 described: 'I think there was a sort of nervousness about allowing people who aren't trained to teach, you know, to go into schools.
[ … ] How can this possibly make a difference?' [Manager 5]. In the early phase of roll out, part of the teaching staff was resistant as well: '[ … ] there was a lot of teachers like, "What is this? You don't need this"' (Mentored Pupil -Biggs et al. 2020, 55). This general attitude was combined with a fundamental unwillingness of policy-makers to accept system change merely on the basis of anecdotal stories: 'People are inspired by stories. The experiences of those in care charges our emotions to act, to right the wrong. However, policy is driven by statistics and economics'. (Programme Founder -MacRitchie 2019, 95). As Manager 1 suggested: '[ … ] I tried and failed to persuade the local authority for the 35 years prior by sharing stories'. Ultimately, although inspirational stories could make some people passionate about making a difference, to counteract the scepticism and ignite a wider system change: 'you need to show the statistical impact and the savings [ … ]' [Manager 1].
For SU, somewhat similarly, the wake-up call to act came from a 'real need', even more apparent in Southern Italy, where recorded rates of school drop-out and NEETs are 'beyond any logic!' [Manager 8]. Two interviewees explicitly mentioned these dramatic figures as evidence of another system failure, in this case, affecting the wider public education system. This latter, lacking adequate resources to combat education inequalities, was portrayed as in dire need of: '[ … ] any intervention that could help teachers take care of the most unfortunate children, those less supported, or even less backed up by the family context' [Manager 8]. In sum, SU faced an educational system starving for any additional aid that private providers could offer, especially if at zero costs, well planned and executed (Programme Founder -Calogero 2018a).
SU interviewees described a number of hindrances, such as those concerning the surrounding financial and legal policy framework. Next to the Law No. 107/2015, which allowed schools to cooperate with NPOs to contrast early school dropout (Programme Founder -Calogero 2018a), interviewees mentioned the Law No. 285/1997, a policy thanks to which SU managed, in the past, to access funds (Programme Founder -Calogero 2018a). Nevertheless, such experience led to recognition of the downsides of taking advantage of public funding opportunities, which proved fairly unattractive due to the resulting monitoring and reporting (M&R) burdens to be borne: '[ … ] the financing was linked to a continuous monitoring of the activity, a control over the various steps of the project [ … ]' [Manager 6]. Further difficulties were ascribed to the disproportionate focus of current private calls for funding on ground-breaking interventions. This trend was seen as fundamentally incongruent with the capacity-building and staffing necessities of those 'who have been carrying out a project for many years, with professionalism' and who would like to pursue 'a structuring of what works' [Manager 8]. Finally, obstacles to operate and grow were perceived as arising from a lack of responsiveness from public administrators and proliferating legislations (e.g. privacy and child protection) and bureaucracies (Programme Founder -Calogero 2018a), which made managing the programme in a law-compliant manner more demanding.
In addition, interviewees discussed a wide number of barriers related to dominant cultural paradigms, resulting in a widespread misunderstanding of what the mentor role truly entails. The mentor was often erroneously seen as a 'support teacher' or 'educator' (Programme Founder -Calogero 2018a), or as a person that: 'can replace a parent', or there was the assumption 'that mentoring can be an activity that makes up for educational shortcomings' [Participant 6]. Interviewed mentors (three out of eight), also acknowledged the difficulties stemming from interacting with teachers who have not yet fully understood the goals of the interventions, as the following quote exemplifies: [teachers]  Moreover, a misconception that mentoring outcomes should be primarily academic and rapidly attainable was also identified, whereas enhancing mentees' overall wellbeing: 'takes a long time and often does not coincide with the immediate improvement of school production' [Manager 6]. Also, mentoringwhich pulls a student out from face-to-face lessons and the class groupwas often perceived as conflicting with youth's primary educational responsibilities. This created implementation challenges and, at times, a priori rejections by teachers (Programme Founder -Calogero 2018a)who may see mentoring as 'stolen time from teaching' (Programme Evaluator -Aresi, De Luca, and Marta 2020, 148-149)and by parents, such as those Manager 6 described: 'Why should my child leave the classroom to do an activity when, at that moment, an educational activity is taking place?!'.

MCR: Leadership, Vision, Values and Culture
A notable characteristic of MCR is that the founder setfrom the very beginningan ambitious vision, which consisted of permanently and pervasively changing the extant system, achieving a system-level embeddedness of the intervention: 'This is the philosophy and DNA of the MCR model. We will not need to exist in 10 years' time because the MCR model and mentoring will be embedded as a right in Scottish education' (Programme Founder -MacRitchie 2019, 97). Alongside having the 'goal to become "business as usual" within the Scottish education system' (Programme Founder -MacRitchie 2019, 97), the vision was that once programme maturity had been achieved, programme ownership would be transferred to each local authority. Changing the system from the inside out and promoting strong public ownership over the initiative was key to securing its long-term sustainability. Another imperative drove the expansion (CEO -MCR Pathways 2021), as also Manager 1 revealed: '[ … ] any young person that has the misfortune to end up being looked after by social work, either formally or informally, is going to get this support'. Interviewees emphasised that MCR also benefitted from its strong learning culture, which led to introduce the programme in those areas that offered the biggest challenges (Programme Evaluator -Biggs et al. 2020), but also opportunity for learning and chances to persuade external audiences of the merits of the programme, setting a compelling example: '[ … ] you don't do the easy thing.
[ … ] you've got to go to the biggest challenge. Then you might die in the process, but [ … ] you motivate others to get involved'. [Manager 1].

MCR: M&E
The challenges stemming out from operating within a resource-constrained environmentwhere 'budgets and the need to reduce costs determined decisions' (Programme Founder -MacRitchie 2019, 95)made clear that, even if you're trying to change the system you: '[ … ] can't add cost to it' [Manager 1]. In other words: '[ … ] additional expenditure was never going to be the way to sustainable and transformational change' (Programme Founder -MacRitchie 2019, 96). Hence, the model was built upon addressing this contextual characteristic. The reliance on volunteers contributed to enhance cost-effectiveness and overcome resource constraints, since: 'you've got big impact for much less cost than you would if you had to employ all the volunteers' [Manager 1]. Next, it became crucial to make the community perceive this relative advantage: '[ … ] it was very much determined, at the very early stage, that the way to be able to expand significantly was to be able to demonstrate that this works and that it's very, very costeffective' [Manager 3]. Although the programme developed and grew primarily to improve the educational outcomes, career prospects and life chances of disadvantaged youth, long-term sustainability was seen as dependent on the ability to build 'a body of evidence to show the deep and lasting benefits for all participants' (i.e. youth, schools, mentors and their employers, local community) (Programme Founder -MacRitchie 2019, 101), which turned out to be game-changing: 'I went back to the local authority, back to the council and, at that point, they couldn't really argue with me [ … ]. They couldn't argue with the stats' [Manager 1]. The accounts rendered by some mentors interviewed confirm the powerfulness of this strategic choice. For instance, Mentor 2 acknowledged to be drawn to that specific volunteering opportunity because: '[MCR] is run with a kind of business focus, which I really like, as a businessperson. I like to know that every piece of resource, your time is used well'. As others recounted: I was listening to the radio, one day, and there was an interview with the guy who started this, and he was talking about it and there was a particular emphasis on the impact, on how the outcomes are measurable. What actually keeps me is that it has been successful in the amount of kids staying on at school. The people that are in the scheme have higher rates of going to college, of going into work, of staying on at school, of doing well.
[ … ] So, I just registered on the spot. I just went on their website to sign [ … ]. [Mentor 12] Timely engaging in M&E escalated the organisation's bargaining power towards public sector partners as well (Policy-makers -Scottish Government 2021; Wilson 2021), who could provide access to new schools and stable flows of financial resources. However, it also helped convincing employers (e.g. councils, 'Corporate Parents', private firms) to allow their employees to mentor during working hours, given the tangible returns achievable in terms of greater staff skills, wellbeing and motivation (Programme Evaluator, Programme Founder and CEO, respectively in: Biggs et al. 2020;MacRitchie 2019;Morton Fraser Lawyers 2021). Interestingly, almost half of the mentors interviewed mentioned both these aspects (i.e. their employer's buy-in and the job-related rewards personally reaped) among the determinants of their willingness to commit to mentoring: It's wonderful because you can develop human skills that you're not going to get anywhere else [ … ] you could say that you've been able to develop these interpersonal skills, that are going to be so important if you want to do something. It could be business, it could be sports … [Mentor 2] MCR: Allies and Resource Acquisition As exemplified by the following: 'MCR does nothing in isolation and would fail spectacularly without partnerships and collaborations' (Programme Founder -MacRitchie 2019, 97). As some school principals recognised, proposing an intervention that was 'a perfect fit for the ethos of our school' (Deputy Head Teacher -MCR Pathways 2019a) and 'matched the vision they had for the school' (Head Teacher -MCR Pathways 2019b) represented a key enabler for the creation of positive partnerships. Nonetheless, when it comes to building wider, cross-sector coalitions, the MCR managerial team very quickly devised strategies to seise the opportunities provided by the policy context. This was operationalised primarily in two ways. First, mentoring was presented to external organisations as a way of meeting their policy obligations, as: 'an opportunity to contribute to their corporate parenting responsibilities, or their corporate social responsibilities' [Manager 3]. Second, to address the scarcity of resources (Programme Founder -MacRitchie 2019), the organisation opted for a split funding model, where local authorities would only pay for the people employed to run the programme in that specific site, while MRC would raise external funds to cover overheads. The final step to ensure the sustainability of a nation-wide scale-up consisted of making the government provide all the Scottish local authorities with the funding needed to cover variable costs (Policy-makers -Scottish Government 2021).

SU: Leadership, Vision, Values and Culture
Growth was pursued, by SU, according to a vision that was about equipping, at first, its existing branches with this programme. Instead, in those contexts where branching could not be used, as lacking locally established structures, SU scaled through affiliation (Programme Founder and President of affiliate NPO, respectively in: Calogero 2018a; Societ a Umanitaria 2019b). In essence, as opposed to MCR, these additional sites 'have not been chosen, usually there have been requests' [Manager 7]. SU not only undertook a different and supply-driven path to scale but articulated the narrative about the scale-up objectives and results in terms of quality, believing that a growing community participation would naturally descend from it: '[ … ] if there are only three of us but we work well, that's okay.
[ … ] It is not necessary to have many schools! They need to work well for them to talk about it' [Manager 7]. Moreover, a key course of action to establish its credibility consisted of transferring the organisational culture, values and procedural norms into codes of practices, guidelines or operative tools (which are given great prominence by the Programme Founder in: Calogero 2018a, 2018b). As some participants explained, the strengths of the initiative laid in high levels of programme structuring and 'proceduralisation': [ … ] we have an organisation behind useven from a somewhat bureaucratic point of viewwhich greatly helps the induction of the programme in a school. That is, whatever question is asked, by a teacher or a school referent, it has its paper-based instrument [ … ]. [Manager 8]

SU: M&e
Although performed evaluations offered scientific backing to the model and a starting point for its continuous improvement (SU Director and Programme Evaluator -Societ a Umanitaria 2019a, 2019b, 2019c), compared to MCR, the potential of leveraging evidence proving the cost-effectiveness of the scheme seemed relatively untapped. Indeed, so far, the reputation of the parent organisation or its representatives turned out to be sufficient to stimulate the involvement of other community actors: [ … ] I do realise that it [scientific evidence] is an excellent 'letter of introduction' [ … ]. In my case, I didn't have to use these cards, [as] the credibility of the organisation that carries it out opens the way; the knowledge of meas a referent -[ … ] has also facilitated this [ … ]. [Manager 8] Another manager reported thatalbeit used to reassure and persuade the final decision-makers (namely, school principals)evaluations mainly served a crucial internal function: '[ … ] we had in return the fact that we are happy to carry it out because [ … ] it has been scientifically proven that it is effective' [Manager 7].

SU: Allies and Resource Acquisition
As for MCR, the affinities with the ethos and values characterising schools of 'inclusive nature' (Programme Founder -Calogero 2018b) paved the way for the adoption of the programme. Moreover, key internal advocates (i.e. driven teachers or principals) allowed SU managers to enter the school system 'from the front door', as trustworthy spokespeople. In addition, the strong reliance on managers previously employed within the school system greatly eased the interchange. It not only gave an inside view of the transformations involving educational institutions (Programme Founder -Calogero 2018a), but also assured compatibility of mindsets, languages and goals (Programme Evaluator -Aresi, De Luca, and Marta 2020), as recognised by Manager 7: [ … ] So, this head teacher, when he saw me, due to the fact that I was a colleague, he already had a more open listening than with someone that goes there by chance. And so, he accepted.
[ … ] when this programme gave its results [ … ] he called me and said: 'Thank goodness your friend sent you, because I wouldn't even listened to you!'.
In contrast with MCR, SU sought partners for scaling mostly within the non-profit sector. In every local area, SU capitalised on its own or its affiliates' reputation to recruit mentors among their members and, in turn, their members' contact network. For instance, 25% of the mentors interviewed admitted having discovered the mentoring programme in that they were seasoned and loyal members of SU's other and more established activities. Affiliates' strategies developed in the same direction, primarily seeking support from other NPOs to recruit mentors: We were then supported for years by the Rotary group, which gave us its members, advertised within the members, but [ … ] In short, we had, before Covid, about 1000 or more people who gravitated around the headquarters, who then, in turn, had friends [ … ]. [Manager 8] Given the perceived poor accessibility of funding mechanisms, SU's management preferred to resort to a broad-based pool of individual donors and membership fees. Public sector organisations' promotional endorsement was also used, albeit seen with a certain diffidence. Indeed, some managers feared to be perceived as politically connected or to receive more demand than they could meet: I must admit that it also scares me a lot, because we are not a factory of volunteers.
[ … ] Such a thing can be destabilising for a programme that is based precisely on the availability of human resources.
[ … ] so, I haven't advocated this so much anymore [ … ]. [Manager 8] In terms of cultural barriers, parental reluctancies were partly counteracted by leveraging on the site-based nature of SBM and the trust relationship bonding schools and families: [ … ] if the school provides help, it is the school that has to be trusted: [ … ] as a parent, I know that the school [ … ] proposes for my child an intervention that is considered valid, I trust the educational institution and I sign up. [Manager 8] In parallel, to make mentoring seen less as an inappropriate use of pupils' school hours, SU underscored its synergies with the broader educational curriculum and learning outcomes (Programme Evaluator -Aresi, De Luca, and Marta 2020), further promoting the idea that such individualised activity could address young people's educational needs through a bespoke approach: [ … ] demonstrating that the activity being carried out is, in any case, an educational activityand not just a recreational oneis a deal-breaker in the bargaining with teachers … [Manager 6] Unintended Consequences or Barriers Deriving from Context and/or Agency MCR representatives mainly described some unintended consequences of the growth strategy adopted, which resulted in operational risks. More specifically, key informants delineated a number of perils (potential or actual) stemming out from not having adopted a sufficiently phased approach. Indeed, growing too wide too fast could make the organisation give in to the temptation of stretching the infrastructure too much or of matching non-suitable volunteers just for the sake of achieving numerical targets. Moreover, particularly one facet of the coalition work introduced the risk of implementers deviating from the model's constituent elements (i.e. internal misperception and mis-implementation). Although scalability benefitted tremendously from purposely blending the expertise of new recruits drawn from the private, public and NPO sectors, a major drift derived from the difficulties in ensuring that all those taken onboard were aligned with the organisational values or fit for purpose: [ … ] we didn't get the recruitment right. We got some wrongs. And again, that's not fault of the persons we recruited, [ … ] most times they had the values, but just not necessarily the skills and experience.
[ … ] We definitely haven't trained them right. [Manager 1] Interviewees also described some of the mechanisms activated to preserve value consistency and integrity while expanding, so as to safeguard the MCR's customary individualised and people centric approach. This entailed the creation of 'spaces of negation' and 'herding spaces' (Ometto et al. 2019) by making room, within the governance structure, for some intermediate liaison bodies (e.g. local young people advisory groups, management school boards, coordinators' development teams).
These served the purpose of: (i) virally spreading MCR's core values; (ii) allowing key stakeholders, from each local area, to have a say on how the programme is designed and implemented; (iii) develop implementers' sense of ownership and responsibility over the delivery and contextual adaptation of the programme.
Participants from SU, instead, mainly spoke about barriers that constrained the scope and speed of scaling. In particular, understaffing was seen as the primary reason why further growth seems, at the current state, problematic, if not even unfeasible: 'It is clear that [ … ] expanding would requireand will requiregreater attention to monitoring the various steps and results and, therefore, the organisation chart [ … ] will have to be, let's say, revised [ … ]' [Manager 6]. This issue seems to relate to the resource acquisition and partnership strategies adopted, which, in the face of a policy context wherein financing opportunities are scarcely accessible, resulted in a financial self-sufficiency risk. Indeed, investing in staffing and internal capacity-building would allow managing growing numbers of mentors but would need additional funds. This, in turn, would require itself more staff (e.g. to apply and comply with the M&R duties described by Manager 6). Also, given the widespread cultural resistance previously described, the context faced by SU entails risks of lack of legitimacy and organisational misperception. The resulting necessary efforts to convey to relevant stakeholders what SBM is truly about (see 'SU: Allies and resource acquisition'), hence 'overcoming internal resistance and external opposition to the innovation through persuasion' (Cinar, Trott, and Simms 2021, 329), further contributed to the need to scale deep and gradually.

Cross-Case Synthesis
The cross-country comparative study offered valuable insights into the effects that contextual conditions exert on social innovation initiatives' attempts to scale (Table 1). The evidence presented suggests that both initiatives developed to respond to pressing social needs, fuelled by system failures: of the extant support system for looked after children, in Scotland, and of the wider public school system, in Italy. In both the cases, indeed, the inadequate support available for the most vulnerable students resulted in rising educational inequalities, to be tackled. Despite this, MCR and SU Request of hard evidence dealt with two highly differing ecosystems of scaling, presenting distinctive opportunities or obstacles to grow. The Italian legal and financial framework was depicted as generally adverse and not conducive for an effective and rapid scaling. The scarce accessibility of policy instruments (i.e. funders' preference for 'breakthrough ideas'; administrative burdens to access grants), alongside proliferation and unresponsiveness of bureaucracies, impaired scalability (Scheuerle and Schmitz 2016). In Scotland, where scarcity of resources initially represented an obstacle, the policy framework strongly favoured the rollout of the programme by supporting the formation of partnerships and establishing a reward system to incentivise its spread (Cuntz, Foray, and Mostovova 2020). In terms of cultural framework, SU appears grappled with counteracting a wider and more varied combination of barriers (e.g. Scheuerle and Schmitz 2016;WHO and ExpandNet 2009) that have been in a way addressed, by MCR, through the use of research evidence. As regards agentic factors, the analyses highlighted that each initiative pursued distinctiveif not, actually, antitheticalcourses of action to sustain growth ( Table 2). The vision, values and organisational culture emanating from the MCR leader implied scaling wide and fast, according to a 'social spillover' institutional logic, which 'emphasizes the breadth of outreach (efficient and fast expansion to cover as many people as possible) as an appropriate way of creating social impact' (Zhao and Han 2020, 137). In contrast, the greater cultural resistance met by SU, as well as its vision, values and culture, led rather to have the programme depth-scaled, through the promotion of a smaller-scale and gradual shift in rooted mindsets, behaviour patterns and mainstream perceptions (Cinar, Trott, and Simms 2021;Moore, Riddell, and Vocisano 2015). Also, MCR's approach turned out to be successful in being learning oriented, as it provided scope for understanding what works or not in disparate geographic locations. Conversely, SU's approach benefitted from procedural structuring, which made adopting schools perceive the model as credible and not burdensome to implement. Additionally, while SU leveraged on reputation to forge alliances strictly within the non-profit sector, MCR opted to mobilise partners spanning across sectors. This was only possible through the provision of sound evidence that helped persuade policymakers that the programme was worth investing in to achieve policy goals. This, eventually, led to formal government decisions to redistribute resources to promote the adoption of the innovation at the national level (namely, vertical scaling -WHO and ExpandNet 2009).
Finally, the analysis revealed that the fast and exponential growth strategy pursued by MCR and the coalition work carried out concealed some ununited negative consequences, primarily resulting in operational risks. As a response, several mechanisms were activated to assure value consistency and integrity (Ometto et al. 2019). Evidence also suggests that, for SU, a mixture of strategic/agentic decisions and contextual factors created a barrier to scale: a shortage of resources to fuel further growth (financial self-sufficiency risk). Additionally, a cultural and institutional context resistant to change generated a legitimacy risk, which further limited expansion due to widespread perceptions that the model is: (i) incompatible with prevalent paradigms; (ii) expected to yield results limited to rapid improvements in educational attainment. For SU, the barriers encountered simply required greater efforts of persuasion (Cinar, Trott, and Simms 2021), which resulted in adopting a deep and gradual scaling strategy (Table 3).

Contributions to Research, Practice and Policy Development
Overall, in line with structuration theories (Cajaiba-Santana 2014), the study shows that differences in scaling (positive/negative) outcomes and barriers come down to two distinct but interrelated causal forces. First, in accordance with agentic theories, we find the attitudes and capacities displayed by the leadership, then reflected in the modus operandi of each NPO. More precisely, MCR leadership merely proved more predisposed to engage in certain ecosystem growth strategies (e.g. advocacy, coalition, research/publication and learning work) able to ease, even indirectly, the diffusion of the innovation (Loukopoulos and Papadimitriou 2022;Roundy and Lyons 2022). Second, as institutional perspectives would suggest, SU leaders dealt with a field featuring higher institutionalisation and lower multiplicity, which resulted in less scope for breaking away from the established order and a greater aversion to involving multiple/non-state actors in educational provision (van Wijk et al. 2019).
In part, study findings confirm the distinct contribution of some agentic and contextual factors to social innovation creation and scalability. For instance, consonant with some extant views (e.g. Mulgan et al. 2007;Perrini, Vurro, and Costanzo 2010), study findings reveal that the cognition of an urgent need for tackling a social issue was a key external enabler, which triggered the momentum for scaling the initiatives. The paper also strongly reinforces current thinking on the boosting effect of supportive partners and policy environment (e.g. Islam 2020Islam , 2022Scheuerle and Schmitz 2016;Voltan and de Fuentes 2016;Verver, Dahles, and Soeterbroek 2021;WHO and ExpandNet 2009). For instance, for MCR, forging cross-sectoral partnerships and garnering political support heavily affected the successfulness of scaling, especially inasmuch driven to a rapid expansion of geographic coverage. Conversely, SU's scale of expansion appeared capped at that achievable with resources mobilised exclusively through NPOs, individual contributions and volunteering (Verver, Dahles, and Soeterbroek 2021), which resonates with an underlying will to preserve independence from institutional funders (Scheuerle and Schmitz 2016). Nonetheless, the analysis also uncovers novel themes, so far remained on the fringe of current mainstream debate. In particular, as concerns the influence of context over agency, the Italian case makes evident that the existence of institutional structures and socio-cultural paradigms less compatible with the initiative being scaled can hinder a large-scale adoption, diminishing community engagement and contributions of resources (Edgecombe et al. 2013;Gabriel 2014;Scheuerle and Schmitz 2016;van Wijk et al. 2019;Voegtlin et al. 2022;WHO and ExpandNet 2009). Conversely, with respect to the influence of agency on context, the experience of MCR unveils that social innovation M&E should be given greater attention. For the Scottish initiative, indeed, evidencing the cost-effectiveness of the intervention was crucial to overcome a number of contextual barriers. Thus, as some already advocated (Cinar, Trott, and Simms 2021;Edgecombe et al. 2013;Larson, Dearing, and Backer 2017;Milat and Bauman 2015;WHO and ExpandNet 2009), scaling processes may benefit from being initiated only once cost-effectiveness had been established. Also, the use of diversified and farreaching methods of collecting and communicating evidence helped influence stakeholders' opinion on the social issue and persuade them that the innovation is worth adopting and/or supporting. This confirms that communication and knowledge dissemination (Bloom and Chatterji 2009;Bloom and Smith 2010;Han and Shah 2020;Loukopoulos and Papadimitriou 2022;van Wijk et al. 2019) are actions worth investing in to further foster scalability, indirectly acting on the scaling ecosystem.
Finally, the study extends current debates on a hitherto poorly understood research topic: the unintended consequences of or barriers to scaling. In particular, against dominant views, it shows that contextual conditionsjust as agentic choiceshave a strong bearing on the perils faced by scaling organisations, which stem out where strategies and context intersect. In sum, as other stated, 'different scaling strategies might face different challenges in different fields' (Scheuerle and Schmitz 2016, 153) or comparable actions have dissimilar results precisely 'because of the differences in institutional and socio-cultural arrangements' (Islam 2020). As such, scaling organisations are strongly encouraged to consider this when devising their course of action and scholars to corroborate this insight conducting future investigations. Alongside showing that the risks faced vary based on contexts, strategies and their unique interplay, the studyalbeit focussed on a sole intervention fieldalso identified notable differences in how such risks were dealt with. These suggestive results hint at a promising line of enquiry: further exploring the distinctive risks that context, agency or their combination impose on scalability and how these ought to be handled.

Conclusion
The magnitude of current pressing social issues adds momentum to debates about social innovation scalability. This study responds to calls for further research into how the interconnectedness between context and agency affects outcomes, barriers and unintended consequences of scaling efforts. In particular, it demonstrates that similar external catalysts can trigger very diverse and equally worthwhile scale ambitions, that can be successfully met through dissimilar paths to scale. Most importantly, the comparison helps to pinpoint which agentic or contextual factors distinctively enabled MCR to achieve, in a narrow timeframe, such a wide scale of expansion. In part, MCR's success is attributable to a greater organisational ability to detect key opportunities arising in the context and use the proper strategic levers to seise them ('political skills (advocacy, networking, lobbying, partnerships), change management skills (strategic organizational change, organizational sustainability) and resources allocation skills (social-financial-human capital, innovation)'-Loukopoulos and Papadimitriou 2022). However, it seems unlikely that analogous strategic/agentic choices could have been deployed by SU andeven if possibleyielding similar results, exactly due to the role contexts played. In Italy, not only a cultural and institutional framework more resistant to change was found, but policy-makers also proved to be less effective in shaping a conducive environment for intersectoral collaborations and eased access to network resources. The greater institutionalisation and lower multiplicity of the field hampered growth (van Wijk et al. 2019). Ultimately, our case study strongly corroborates the key tenet of structuration theories: just like socio-cultural and institutional structures constrain/empower agentic choices, they are shaped by agents. Both these levels of analysis are required to understand social innovation scalability.

Limitations
First, the analysis largely relies on self-report accounts rendered by programme managers and institutionally channelled materials, which may have limited the ability to achieve an exhaustive and unbiased picture of enablers and barriers. Hence, the conclusions drawn risk to be weighted towards the organisational perspective, as the views of some important downstream stakeholders (i.e. mentored youth, their families, school personnel, policy-makers) were accessed only indirectly. Nonetheless, the use of multiple data sources and, most notably, the primary data gathered among volunteer mentors were expected to mitigate such a bias. Second, the purpose of the study was exploring the process of scaling social innovation (i.e. breadth of outreach), rather than its impacts. Thus, an important question for future research is ascertaining what value is delivered to beneficiaries as interventions are scaled. Third, our study found that certain countermeasures (Ometto et al. 2019) were adopted only to mitigate any unintended consequences of scaling. A promising avenue for future research is investigating whether, in other research settings, comparable actions (particularly 'herding spaces', which connect organisations with their institutional contexts) proved effective in identifying strategic actions aimed at overcoming potential barriers to scaling posed by the context itself. Finally, although the comparative nature of the study may offer stronger grounds for theory generation, future evaluations are encouraged to examine whether similar results ensue from applications in further and diverse research settings.
Despite its limitations, the study sets itself apart from most former evaluations for at least three reasons. First, the authors expressly engaged in a comparative cross-cultural study to contrast scaling processes across contexts. Second, the chosen conceptual framework (WHO and ExpandNet 2009) was borrowed from a cognate field of research: health innovations scalability. This allowed identifying and documenting some influential factors (either contextual or agentic) that are less extensively discussed in extant research on social innovation. Finally, the study draws on structuration theories, focussing on the mutual effects of context and agency over social innovation scalability. It is argued that enquiries of this kind (namely, comparative, cross-cultural, multidisciplinary and underpinned by holistic theoretical perspectives) can crucially enhance the understanding of which practicesgiven certain contextspecific variablespromise an enhanced social innovation scalability.