INNOVATION PLATFORMS IN AGRICULTURAL RESEARCH FOR DEVELOPMENT

SUMMARY Innovation platforms are fast becoming part of the mantra of agricultural research for development projects and programmes. Their basic tenet is that stakeholders depend on one another to achieve agricultural development outcomes, and hence need a space where they can learn, negotiate and coordinate to overcome challenges and capture opportunities through a facilitated innovation process. Although much has been written on how to implement and facilitate innovation platforms efficiently, few studies support ex-ante appraisal of when and for what purpose innovation platforms provide an appropriate mechanism for achieving development outcomes, and what kinds of human and financial resource investments and enabling environments are required. Without these insights, innovation platforms run the risk of being promoted as a panacea for all problems in the agricultural sector. This study makes clear that not all constraints will require innovation platforms and, if there is a simpler and cheaper alternative, that should be considered first. Based on the review of critical design principles and plausible outcomes of innovation platforms, this study provides a decision support tool for research, development and funding agencies that can enhance more critical thinking about the purposes and conditions under which innovation platforms can contribute to achieving agricultural development outcomes.

alternative, that should be considered first. Based on the review of critical design principles and plausible outcomes of innovation platforms, this study provides a decision support tool for research, development and funding agencies that can enhance more critical thinking about the purposes and conditions under which innovation platforms can contribute to achieving agricultural development outcomes.

I N T RO D U C T I O N
Multi-stakeholder alliances or platforms are an increasingly popular approach to enhance collaboration and innovation within the agricultural research for development (AR4D) sector (Dror et al., 2016;Lundy et al., 2005;Neef and Neubert, 2011), as well as in other sectors such as healthcare (McHugh et al., 2016), natural resource management (Faysse, 2006;Misiko et al., 2013;Steins and Edwards, 1999;van Rooyen et al., 2017;Warner, 2006) and infrastructure (Klijn and Teisman, 2003). In the AR4D sector -the focus of this paper -multi-stakeholder innovation platforms (henceforth referred to as 'innovation platforms') are promoted to bring together groups of individuals (who often represent organisations) with different backgrounds, expertise and interests -farmers, traders, food processors, researchers, government officials -and to provide them with a space for learning, action and change (World Bank, 2006). The fact that previously disconnected stakeholder groups come together to diagnose agricultural and broader livelihood problems, identify opportunities and find ways to achieve their goals is among the main benefits of innovation platforms (Klerkx et al., 2012). As the name indicates, innovation platforms have an 'innovation' objective, that is, the introduction and utilisation of any new knowledge (technological or other) in an economic or social process (OECD, 1999).
Depending on the level at which an innovation platform is established (e.g. village, regional and national), and on those initiating the platform, the objective might be to tackle a specific technological, organisational or institutional challenge in a value chain (e.g. access to high quality potato seeds), or a more generic problem that needs to be addressed across different value chains (e.g. farmers' access to credit). Once the innovation platform has achieved its objective, its members may (or may not) decide to take up new challenges (Davies et al., 2018;Hounkonnou et al., 2012). Innovation platforms can start as informal networks and some may transit into more formalised structures, such as public-private partnerships or a cooperative, with the goal of becoming self-sustaining entities (Schut et al., 2017a).
Over the past years, innovation platforms have increasingly been established within the framework of AR4D initiatives (Dror et al., 2016). The innovation platform approach is particularly being embraced as a model for achieving development outcomes through participatory action research (Chave et al., 2012;Ottosson, 2003). The existing literature mainly focusses on providing step-by-step advice on how to implement and facilitate innovation platforms for technological or institutional change. Such implementation and facilitation advice can be found in Adekunle et al. (2010), Makini et al. (2013), Brouwer and Woodhill (2016), Francis et al. (2016) and Nederlof and Pyburn (2012). What currently lacks in the literature are discussions on the usefulness of innovation platforms in overcoming a range of agricultural challenges. In particular, there is a lack of decision support tools that can stimulate critical ex-ante reflection on when and for what purpose innovation platforms are an appropriate mechanism for achieving development outcomes. This poses the risk of innovation platforms being promoted as a panacea for all research and development problems in the agricultural sector, something for which authors analysing multistakeholder partnership models in other sectors have warned for (e.g. Warner, 2006).
The objective of this study is to complement the existing body of scientific literature by focussing on the usefulness question of when and for what purpose innovation platforms are an appropriate mechanism for achieving agricultural development outcomes. In doing so, this study contributes to generating more realistic expectations about what innovation platforms can and cannot achieve in AR4D initiatives. As the implementation of innovation platforms can consume significant time, energy and other human and financial resources, this study is geared towards providing decision support to development practitioners, researchers, funding agencies or farmer unions in determining whether the innovation platforms can help them to achieve their objectives. Data for this study were collected through literature review, and sourced from the practical experiences of the authors, who all contributed significantly to the design and implementation of innovation platforms across different value chains and continents.
The next section elaborates on the rationale for using innovation platforms in AR4D. This is followed by a section that discusses the conditions that should be in place for innovation platforms to be effective. Subsequently, we explore what can realistically be the expected outcomes of innovation platforms. This provides the basis for a decision support tool that can help research, development and funding agencies in more critical reflection on whether and how innovation platforms can truly strengthen their AR4D approaches and programmes. The final section summarises the main conclusions from this study.
Before adopting an innovation platform approach, one should carefully reflect on whether or not innovation platforms are the most useful and cost-efficient vehicle to achieve project or stakeholder objectives. Questions that can guide decision making include (1) what are the expected functions of the platform; (2) what can innovation platforms achieve efficiently and (3) when are innovation platforms particularly useful?

What are the expected functions of innovation platforms?
In an effort to create space for learning, action and change, innovation platforms can fulfil a collated range of functions and related activities in AR4D processes (Table 1). Innovation platforms do not necessarily fulfil -or indeed need to fulfilall of these functions and activities. Depending on the composition of the innovation platform, as well as its specific objectives, specific functions and activities may be more or less relevant (Klijn and Teisman, 2003;Lamers et al., 2017). Furthermore, 578 M A RC S C H U T et al.  Hekkert et al., 2007;Kilelu et al., 2011 the functions and activities outlined can be undertaken in various orders, either sequentially or parallel to one another.

What can innovation platforms achieve effectively?
Innovation platforms aim to counter weaknesses in agricultural innovation systems by building interaction amongst different kinds of actors and their organisations, promoting change in practices, institutions and policies and to effectively deploy available human and financial resources to solve problems and capitalise on opportunities (Davies et al., 2018). Regardless of whether innovation platforms are established at local or higher levels, they can explore technological, organisational and institutional solutions, making them ideal for addressing problems in an integrated manner (Flor et al., 2016;Sanyang et al., 2014;Schut et al., 2016a). In a way, the formation and operation of innovations platforms is an organisational or institutional innovation in itself (Schut et al., 2016a). It entails changes in ways of collaborating, interacting and in relationships between actors and organisations to overcome obstacles and improve the impact of their collective action (Ngwenya and Hagmann, 2011;Swaans et al., 2014).
In AR4D, innovation platforms can support participatory action research. Participatory action research combines both conducting research together with key stakeholders, as well as performing outcome-oriented research (Minh et al., 2014). The involvement of key stakeholders is important for three reasons. First, stakeholder groups can provide various complementary insights about the biophysical, technological and institutional dimensions of the problem, thereby broadening the knowledge base. Thus, by engaging in a social learning process with one another, stakeholders can negotiate what type of innovations are technically feasible, economically viable and socio-culturally and politically acceptable (Esparcia, 2014;Hermans et al., 2011;Schut et al., 2014). Second, through their interaction and participation, stakeholder groups become aware of their different interests, needs and objectives, but also of their fundamental interdependencies and the need for concerted action across different levels to overcome their constraints and reach their objectives (Leeuwis, 2000;Messely et al., 2013;Schut et al., 2013). Third, stakeholders are more likely to support the implementation and scaling of innovations when they have been a part of the design and testing process (Faysse, 2006;Neef and Neubert, 2011).
Doing action-and outcome-oriented research requires flexibility, as innovation platforms operate in dynamic contexts, and themselves aim to promote change. Action research takes the innovation platform members through cycles of designing interventions, testing in practice, observing whether activities bring about desirable change, reflecting on what goes well and what can be improved, which results in a new phase of (re)designing the interventions (Ottosson, 2003). Members, as well as their supporting organisations, need to be prepared to adapt their approach and expectations and, in some cases, for failure.
The effective innovation platform size depends on the specific innovation challenge or opportunity at hand. For farm-level experimentation, working with 20-40 participants may be appropriate (e.g. Andres et al., 2016), whereas innovation platforms aiming at market or regulatory change are more likely to be smaller in size to enhance flexibility and decision-making (Fichter, 2009;Klerkx et al., 2009). Innovation platforms that aim to create economies of scale (e.g. by accessing or distributing inputs or by jointly supplying to private processors) may actually be much larger in size (e.g. Woomer et al., 2016). Furthermore, membership and size may change over time as new farmer groups can join or new members are invited to address specific challenges, or partners may lose interest or leave the platform if their needs have been met (Lamers et al., 2017).

When are innovation platforms particularly useful?
In general terms, innovation platforms are useful when (1) persons or organisations that represent different socio-economic backgrounds, interests and perspectives have a stake in a particular problem or solution; (2) multiple persons or organisations want or need to experiment jointly on aspects that they cannot solve individually or that benefit from synergies; (3) new solutions require a combination of new technologies (technological innovation), effective collaboration (organisational innovation) and/or new rule, funding and incentive structures (institutional or policy innovation) and (4) actors and organisations are willing to share knowledge, resources, benefits and risks, as well as sufficient common interest and trust to engage in collective innovation to address a common challenge (Buerkler, 2013). These conditions are also frequently mentioned in relation to other types of multi-stakeholder approaches such as publicprivate partnerships (e.g. Hall, 2006;McHugh et al., 2016;Van der Meer, 2002).
Additional questions related to (1) the specific problem at stake, (2) the specific type of solution or innovation needed to overcome that problem and consequently (3) the types of innovation partners that should be engaged and (4) the most cost-and time-efficient partner engagement model can further guide decision making on the need for an innovation platform, or for other innovation and partnership approaches (Hermans et al., 2017). If addressing problems does not require the involvement of multiple stakeholder groups and/or organisations, then simpler and more shortterm partner collaborations or formal bilateral arrangements may be more desirable (Head, 2008).
When an innovation platform approach is deemed the most useful innovation and partner engagement model, then it is advised to first make an inventory of existing stakeholder platforms and networks (Boogaard et al., 2013). If the purpose, modus operandi and power dynamics of these existing platforms are in line with the objectives and needs of the involved stakeholders, then building on these existing platforms and networks may be more efficient and quicker than initiating a new innovation platform (Boogaard et al., 2013;Cullen et al., , 2014aCullen et al., , 2014bSchut et al., 2018).
Once the usefulness of an innovation platform has been identified based on the criteria mentioned in the above section, five key conditions for effective innovation platforms need to be met: (1) ability and mandate to pitch the platform at the right level(s); (2) conducive institutional environment for an innovation platform approach; (3) availability of sufficient capacities and skills; (4) organising monitoring, evaluation and learning (ME&L) and (5) adequate funding for innovation platform implementation.

Ability and mandate to pitch the platform at the right level(s)
Innovation platforms can be established at different levels such as village or community level, district level and even province or national level . The guiding question should be 'At what level or levels can a challenge be addressed most efficiently?' For example, a problem of access to good quality planting material may be best tackled at the village or community level, whereas exploring irrigation options would require the involvement of stakeholders at the watershed level (e.g. van Rooyen et al., 2017). As problems at local level are often rooted in, and interrelated with, problems at higher levels (e.g. lack of input certification leading to poor quality fertiliser on the market), the strategic involvement of national level policy actors may be desirable (Schut et al., 2016b). Local innovation platforms might resolve concrete agronomic or organisational issues but, without linkages to decision makers at higher level, will most likely not have enough weight to foster structural changes at higher levels (Lamers et al., 2017). Conversely, higher level innovation platforms may be less relevant for farmers with specific needs but can serve to review and, ideally, shift the rules of the game to make the overall system more amenable to farmer interests and overall public goals such as sustainability, incomes, competiveness, etc. (Hounkonnou et al., 2012;Hounkonnou et al., in press).
Making changes at higher levels often requires more time and is political by nature, which may not well align with the perceived mandate of AR4D organisations (Schut et al., 2016b). Nevertheless, the spin-offs from achieving changes at policy level may lead to the desired agricultural development outcomes. AR4D organisations can find strategic partners who are experienced in engaging high-level decision-makers. Davies et al. (2018) describe how the engagement of an influential representative of a local chamber of agriculture in Burkina Faso developed a basis for gaining support from regional development policy actors, which created an enabling environment for the innovation platform to achieve its objectives.

Is the institutional environment conducive for an innovation platform approach?
Through their demand-driven approach and their capacity to expose and balance existing power inequalities, innovation platforms can create tensions within AR4D establishments (Hounkonnou et al., in press). Innovation platforms may request AR4D organisations to work on themes, commodities or value chains that are outside of their normal mandate or comfort zone (Schut et al., 2016a). Such institutional tensions and the institutional innovations to deal with them are often the unintended consequences of working through innovation platforms, and can have widespread impacts in the sense of how organisations identify demand, work action-oriented and try to be relevant for their next-and end-users . Schut et al. (2016a) pointed out that many AR4D organisations face challenges in supporting and institutionalising innovation platform approaches and principles due to inflexible mandates, incentive structures, procedures and funding mechanisms. They questioned whether in the absence of such an enabling environment or unwillingness of organisations to embrace these tensions, innovation and other multistakeholder platforms can lead to real change, or whether that would just result in a continuation of 'business as usual'. Similar tensions, and institutional and strategic barriers have been observed in other partnership approaches such as public-private partnerships, both within the AR4D sector (e.g. Hall, 2006;Kilelu et al., 2017;Poulton and Macartney, 2012;Spielman and von Grebmer, 2006), as well as in other sectors (e.g. Faysse, 2006;Klijn and Teisman, 2003;Warner, 2006).

Are sufficient capacities and skills available?
Innovation platforms consist of multiple and heterogeneous groups of stakeholders with different interests, ideas and competencies in terms of what they can offer to the platform. Bringing together a group of stakeholders with diverse needs, interests and objectives is likely to lead to tensions, conflicts, manoeuvring to seek advantage and even group displacement, which can hinder collective action towards achieving development outcomes (Hinnou et al., 2018;Kilelu et al., 2013Kilelu et al., , 2017Ruttan, 2008;Thiele et al., 2011). Innovation platforms are known to become arenas of struggle, as solutions for some members may create new obstacles for other members (Leeuwis, 2000). Moreover, power differences exist between different members (e.g. farmer versus government official), and not all members may have equal discussion and negotiation skills (Brouwer et al., 2013;Cullen et al., 2014b). Facilitation of interactions, collaborations, power dynamics and actions is needed to arrive at commonly agreed upon objectives (Tenywa et al., 2011). Innovation platforms are also known to have successfully contributed to prevention of conflict and resolution of disputes (in crop-livestock systems) (Davies et al., 2018). Depending on the specific body of literature, such facilitation has been referred to as 'championing' , 'brokerage' (Madzudzo, 2011), 'boundary spanning' (Fleming and Waguespack, 2007) or 'promoting' (Fichter, 2009).
Researchers and development practitioners engaged in AR4D projects are increasingly called upon to act as facilitators (Cadilhon, 2013b). Stakeholders in an innovation platform that is in its early stage may not feel confident to facilitate; they may look to project implementers and researchers to take the lead (Glin et al., 2016). When researchers act as facilitators, conflicts of interest may arise, and they may confront problems about ambiguity of roles and responsibilities, or they can be viewed by other actors as competitors rather than as neutral or legitimate facilitators (Devaux et al., 2010;Klerkx et al., 2009). Facilitation may also be shared between various people, e.g. through establishment and IP coordination and facilitation team (Hinnou et al., 2018) which can in itself be a strategy to build a collaborative ethos and shared ownership in the innovation platform.
To effectively support innovation platforms, competence and skills of facilitators may need to be strengthened through training, coaching and mentoring (Sanyang et al., 2016). Competencies and skills often go beyond fulfilling solely knowledge brokerage or knowledge management roles, but also require stimulating demand articulation, collective problem analysis with diverse stakeholders, supporting joint decision-making, multi-level network building, mobilising institutional and political support and managing overcoming power inequalities (Kilelu et al., 2011;Schut et al., 2011). In doing so, facilitators do much more than just organise and manage platform meetings. They ensure transparency of discussions and negotiations, and that innovation platforms stay solution-and action-oriented so they can reach their objectives. Additional competencies of facilitators can include (1) bringing about changes in the values, attitudes and self-perception of those who engage in innovation platform activities; (2) keeping an innovation platform functional even without external funding; (3) developing the innovation platform's capacity to move from individual to collaborative activities, with the ability to self-organise and learn; (4) providing mechanisms for accountability and feedback within the innovation platform and (5) establishing lessons with other innovation platforms for learning and collective action. Facilitation requires substantial financial investments (Cadilhon, 2013b;Glin et al., 2016;Swaans et al., 2013b).

Organising effective monitoring, evaluation and learning (ME&L)
Impact assessment of innovation platforms and their effectiveness is a contentious issue, and suitable ME&L tools for multi-stakeholder innovation processes in AR4D are limited (Davies et al., 2018;Swaans et al., 2013c). They produce either qualitative case studies from which data cannot be easily generalised or quantitative impact assessments that do not provide insights into ongoing process dynamics . New tools to effectively monitor, evaluate and learn in innovation platforms have been developed and tested. Cadilhon (2013a) developed a conceptual framework, using quantitative research methods to assess the impact of innovation platforms. The framework has been applied in Ghana and Tanzania to evaluate the impact of innovation platforms on marketing relationships (Adane-Mariami et al., 2015;Pham et al., 2015). Another is the learning system for agricultural research for development (LESARD), which provides integrated quantitative and qualitative data collection and analysis tools to assess the performance of multi-stakeholder processes .
Without appropriate ME&L mechanisms, innovation platforms run the risk of not being able to provide proof of their success or share important learning experiences. ME&L also provides a mechanism to have quick feedback from the innovation platform members so that that a timely adjustment in the innovation platform focus or strategy can be made. This reduces the risk of investing valuable time, energy and financial resources in activities that do not lead to the desired outcomes. The need for continuous reflection, learning and adaptation based on robust ME&L is an essential design principle of innovation platforms (Swaans et al., 2013a).

What are the costs of innovation platforms?
There is very little information on the costs of innovation platforms in an AR4D project, and -to the best knowledge of the authors -no cost or cost-benefit analysis has been conducted on innovation platforms in an AR4D context. As mentioned before, innovation platforms are human-and financial-resource intensive, and research and development donors will require evidence on the return on investments. Innovation platform costs vary, depending on the (1) type of organisation that is implementing/supporting the innovation platform (average staff costs in international organisations are usually much higher than average staff costs in local NGOs); (2) type of innovation that is being explored, e.g. planting distance or intercropping practices (relatively cheap) versus local processing that requires machinery (relatively expensive); (3) level at which the innovation platform is operating (higher level platforms are usually more expensive); (4) number of innovation platform members (more farmers or other members involved can increase operational and support costs); (5) level of platform support functions required (e.g. facilitator, logistics, documentation, and so forth); (6) spin-off activities that emerge as the platform starts to operate (innovation platforms must have flexibility to cover unplanned but important activities); (7) proximity of facilitators to implementation sites (platform facilitation has to be monitored, so having a local facilitator can make a difference to operational costs) and (8) time for preparing, holding and following-up on meetings, and for general exchange, searching for compromises and documentation (transaction costs).
As becomes clear from Box 1, most of the initial costs of innovation platforms are investments in the institutional set-up of the platform or the organisation in general. These are not costs that often quickly result in benefits or outcomes. Buizer (2016) conducted cost analysis of two innovation platforms and its overarching steering committee in Uganda. She concluded that the innovation platform costs are approximately US$83.29 per farmer per year (Box 1). We are aware that we cannot use the Ugandan case to draw firm conclusions on whether innovation platforms provide value for money. To do that, we need to compare the costs of innovation platforms with other approaches of innovation design, testing and dissemination. That is, compare the costs of a platform with the costs of dissemination through extension services or the services provided by agribusinesses (contract farming approaches) to their members. Work by IFAD (1998) published in Quizon et al. (2015) demonstrated that for Farmer Field Schools (FFS) in Uganda, high allowances, transportation costs and several layers of supervision programmes could make extension cost up to US$53 per FFS-trained farmer for a one-season long training, excluding the costs of trainer salaries. Several studies show that diffusion of innovation to non-FFS participants has been disappointing (e.g. Rola et al., 2002). Empowerment and innovation skills of FFS participants were expected to generate economic multiplier effects and more longterm behavioural change of participants, which is similar to the expected outcomes of innovation platforms. When comparing costs of innovation platforms with the costs of government extension services in Uganda, we calculated that the costs per farmer are US$7.36 per farmer 1 . Using an alternative analysis, we could also conclude that the innovation platforms that Buizer analysed, should at least benefit 2357 farmers in order to be competitive with the FFS approach, or even 16 974 farmers in order to be competitive with the incumbent government extension services system.

Box 1. Cost-analysis of two innovation platforms implemented under the CGIAR Research Program on Integrated Systems for the Humid Tropics (Humidtropics) in Uganda.
Buizer (2016) conducted cost analysis of two innovation platforms implemented under the CGIAR Research Program on Integrated Systems for the Humid Tropics (Humidtropics). Two calendar years (2014 and 2015) of innovation platform activities were analysed. One innovation platform focussed on indigenous vegetables and pigs, the other on intercropping soya beans and maize. A national level steering committee was formed to coordinate the work across the two innovation platforms and to link them to policy and other public agencies and the private sector. The two innovation platforms reached approximately 1500 farmers in the areas where they were operating.
To analyse the costs, the study differentiated between (1) basic costs for platform events, coordination of meetings of intervention actors, reflection and preparation for meetings and (2) theme-specific costs for conducting trials, providing training, data collection, etc. The idea behind this separation is that the basic costs will be approximately the same for all innovation platforms, regardless of their specific topic or theme. To organise basic meetings and activities and to hire most of the basic staff, US$71 677 was spent in 2014 and US$64 216 in 2015. Of the total basic costs, expenditure on human resources accounted for the largest part (39% and 42% in 2014 and 2015, respectively). Basic staff include the national facilitator (responsible for facilitating the steering committee as well as the two innovation platforms), a project coordinator, a communications officer, an ME&L expert and drivers and other support staff. When including the theme-specific events and staff costs, the costs were significantly higher: US$109 607 in 2014 and US$140 255 in 2015.
The cost of basic events decreased between 2014 and 2015 for the two innovation platforms and the steering committee. This is mainly because the platform attracted investments from other organisations. Meeting costs were the largest cost category and represented 64% and 59% in 2014 and 2015, respectively. Meeting costs included renting the meeting venue and lunch or transport refunds for participants such as farmers or government officials. Meeting costs decreased after the first year because of a decrease in the number of people attending meeting as the platform's focus had become clear. Fuel costs and participants' transport reimbursements formed the second largest cost category.
In conclusion, establishing and maintaining the two innovation platforms with one overarching steering committee in Uganda, reaching an year was calculated at US$83.29 ((US$249 861/2 years)/1500 beneficiaries). The innovation platform facilitator accounted for the largest share of the basic human resource costs ($1000 per month). The cost per farmer is likely to decrease if the innovation platform is supported by local government and/or local NGOs, instead of being implemented and coordinated by an international agricultural research organisation.
More detailed information can be found in Buizer (2016) and .
Innovation platforms are resource intensive, and research and development donors will require evidence on the return on financial and human resource investments against outcomes and impacts. Mapping the costs of innovation platforms is an important first step towards conducting (long-term) cost-benefit analysis and showing whether innovation platforms can provide value for money. International NGOs and AR4D organisations often provide funding to kick-start innovation platforms. However, this funding is usually available for a limited period and may not be sufficient to meet all the costs associated with the establishment and facilitation of the platform. Continuous support may moreover have a reversed effect on platform ownership, as innovation platform members may not feel fully responsible for the costs and investments. Innovation platforms that are supported through AR4D projects should therefore develop strategies for reducing (financial) dependence on these projects. If engaging in the platform results in obvious benefits, the innovation platform can attract financial resources or other types of support from the private or public development sector. In Bolivia, for example, the private sector took a more proactive role and sought additional funding for the ANDIBOL (Andino Boliviana) multi-stakeholder platform for linking smallholder farmers to value chains (Thiele et al., 2011). Davies et al. (2018) explain that 'In Amantin and Savelugu [Ghana], the registration of the IP as a cooperative was identified as a factor that explained its outcome because this structure was considered to balance the self-interest and shared interest of members.  illustrate this using the case of the Tanga Dairy Platform that successfully lobbied policy makers to reduce value-added tax on dairy inputs and products, and remove limitations on urban dairy farming in Tanga City, Tanzania.

W H AT O U T C O M E S C A N B E R E A L I S T I C A L LY E X P E C T E D F RO M I N N OVAT I O N P L AT F O R M S ?
Innovation platform outcomes should be considered on two levels. The first level concerns the direct beneficiaries (the platform members) and the second level concerns the indirect beneficiaries (the target population or region beyond the platform's direct influence). To reach the second category, some form of scaling is required . The three leading questions that need to be addressed are (1) what are the benefits for innovation platform members; (2) what strategies can support the scaling of innovation platform processes and outcomes and (3) which additional mechanisms and arrangements may be needed to broaden the coverage and impact of innovation platforms?

What are the benefits for innovation platform members?
Innovation platforms can create different types of benefits for its members. These benefits include, but are not limited to (1) a space where each platform member has access to a variety of experts who could enhance their skills, including farmers, researchers, private sector and government; (2) a protected niche where a group of people can experiment, learn and make mistakes without it having huge negative consequences; (3) increased credibility and legitimacy as a result of speaking with a collective voice when the objective is to create change at different levels; (4) a better power and bargaining position as a group for accessing knowledge, inputs, finance, markets and other types of services and (5) network building for developing new initiatives, enterprises and projects (Boogaard et al., 2013).
Among organisations that implement innovation platforms, there is debate on whether and how innovation platform members should be compensated for being part of the innovation platform, and in line with Nederlof et al. (2011) we advise against financial incentives for platform members. Reimbursement of -for example -transport costs can be considered, especially for those participants who are not supported by their constituencies. Benefits should result from the above-mentioned activities and the opportunities that platform membership provides. As membership in innovation platforms is voluntary, members who feel that the platform is not benefitting them sufficiently are free to leave it.
As elaborated earlier, innovation platforms need financial and human resource investments. Facilitation, platform establishment, platform activities and ME&L incur costs that cannot be expected to be carried by the platform members from the beginning. That said, the platform should develop a strategy for becoming independent of permanent outside financial and technical support (e.g. through a development project). Innovation platforms are known to transit into cooperatives (Davies et al., 2018) where platform members make a small financial contribution to the platform's costs. Platforms can also cease to exist once its members feel the mission has been accomplished, or when motivation levels have dropped.

What strategies can support the scaling of innovation platform processes and outcomes?
Innovation platforms initiated through AR4D projects often have the ambition to have impact beyond the initial target area or direct beneficiaries . Such processes of scaling innovation platform processes and outcomes should be an integral design element of innovation platforms and the manner in which they are implemented. The literature (e.g. Hermans et al., 2017) distinguishes between two types of scaling: outscaling and upscaling. Outscaling refers to the horizontal diffusion of innovations to individuals or organisations at the same level (e.g. from one district to another district). Upscaling refers to the embedding of processes or technologies at higher levels (e.g. institutionalisation of new cropping practices in policies). For the innovation platform members (e.g. farmers), the scaling of innovations may not always be beneficial. If there is competition among platform members and secondary beneficiaries, then scaling is challenging. There need to be clear benefits for the platform members to scale out, for example, developing a local brand or production standard that requires a critical mass or a large buyer who requires a minimum quantity produced in one region. An example is the maize innovation platform established by the DONATA Project in The Gambia where women farmers collectively packaged, labelled and sold maize flour and grits in standard bags (Sanyang et al., 2014). Another example is from the Bante and Glazoue rice innovation platforms established under the SARD-SC project in Benin that developed brands for local white and parboiled rice (Sanyang et al., 2016).
Innovation platforms can fulfil an important function in the pathway leading to the scaling of agricultural innovations. Through their participatory approach to identifying and analysing problems, and designing and testing innovations to overcome those problems, they have a higher likelihood to result in solutions that are not only technically sound, but also affordable for farmers and coherent with government policies and objectives. However, if the basic innovation platform features are not respected (e.g. innovation platforms for implementing pre-cooked AR4D projects, with limited participation space for farmers and scaling partners to influence the AR4D agenda), the basis for scaling of innovation may be compromised (Wigboldus et al., 2016). As mentioned before, the involvement of farmers, policymakers and the private sector in decision-making and innovation processes provides an important precondition for supporting the wider use and spread of validated technologies and other types of innovations developed in innovation platforms. Public and private scaling partners can be strategically engaged from the early stages of innovation platform establishment, be allocated explicit roles in the innovation platform (e.g. in an advisory or steering committee) and/or be involved in the developing strategies that align with their core business, values and strategies (Klijn and Teisman, 2003;Lamers et al., 2017).
Whether innovation platforms can support or play a role in large-scale diffusion or scaling of agricultural innovations also depends on their institutional embedding. A recent meta review in Schut et al. (2018) of mature innovation platforms concluded that innovation platforms need to be firmly embedded in private or public mechanisms and broader networks that have the capacity to reach target populations beyond the original scope of the innovation platform. Innovation platforms run the risk of staying solitary initiatives if they are not firmly linked to such existing mechanisms and networks, which reduces the chance of having impact beyond the direct beneficiaries.

Which additional mechanisms and arrangements may be needed to broaden the coverage and impact of innovation platforms?
Once an appropriate combination of stakeholders is defined, it is critical to define the process and associated institutional arrangements for collaboration. Innovation platforms may start off with a very focussed approach to addressing a specific or number of problem(s). In an initial stage, a rather loose arrangement, based on clearly defined goals, roles, activities, results and resource needs, may suffice. As the number of stakeholders and the size of operations increases, more formalised institutional arrangements may be needed to govern the innovation platform, its process and its benefits. This may involve the creation of subgroups with specific contracts within and between them. Over time, it may also be worthwhile to consider obtaining a legal entity (e.g. as cooperative) for an innovation platform to enhance its sustainability in terms of independence, its (monetary and non-monetary) benefits for its members, and its potential to become eligible for donations or credit. A comparative case study paper by Davies et al. (2018) providing examples of improved financial access for innovation platform members as a result of the innovation platform's negotiation with finance providers.

D E C I S I O N S U P P O RT TO O L F O R R E S E A RC H , D E V E L O P M E N T A N D F U N D I N G
A G E N C I E S The above-mentioned design principles and expected innovation platform outcomes are brought together in flow diagram to support project developers, funding agencies and implementers in deciding whether or not innovation platforms are the most appropriate pathway towards achieving their development outcomes (Figure 1). The diagram focuses on the critical questions that research, development and funding agencies need to ask themselves before deciding to embark on implementing innovation platforms in their AR4D projects and programmes. These questions include the following: 1. For what main purpose are innovation platforms being used?
i. For developing and testing new technological innovations (e.g. home vegetable gardens) or institutional innovations (e.g. a contract-farming model or improved market access). ii. For tailoring technological or institutional innovations to the specific needs of end-users or agro-ecological areas (e.g. composition of seed kits for home vegetable gardens for specific households in different districts). iii. For outscaling of existing technological or institutional innovations for the benefit of large numbers of end-users (e.g. disseminating seed kits for home vegetable gardens to thousands of farmers). iv. For upscaling of existing technological or institutional innovations to influence policy, development and business sectors (e.g. embedding distribution of seed kits for home vegetable gardens in nutrition and agricultural policy or markets).
A number of additional questions related to whether innovation platforms would result into the desired development outcomes bring us to the second key question as follows: 2. Do we have sufficient resources as well as institutional support and flexibility to support the implementation of impactful innovation platforms? i. Do we have adequate human and financial resources (facilitator, sufficient and flexible funds) to support innovation platform activities? ii. Is there flexibility in our project to support innovation platforms (e.g. to change focus if the platform feels this is necessary)? iii. Is there institutional support to work in a demand-driven and participatory way with the innovation platform (e.g. is AR4D leadership supportive of the innovation platform approach; are there investments in capacity building of staff)?
If there is sufficient institutional support and flexibility, then an additional third question should be reflected upon as follows: 3. Are there existing multi-stakeholder innovation platforms on which the project could build? i. If yes, is that innovation platform willing and able to collaborate to achieve joint objectives?
The answers to the above questions will to a large extent depend on the specific socio-political context in which the innovation platform is supposed to contribute to achieving agricultural development outcomes (Pamuk et al., 2014). In some countries, for example, it will be experienced as extremely positive that rural actors organise themselves, sit down together around joint constraints and self-organise and implement interventions to overcome these constraints. In other countries, such processes may be viewed with suspicion by governments or other dominant parties, who may feel that these platforms are not needed or undermining their role, mandate and function. In line with Hermans et al. (2017), we conclude that project designers and implementers need to think more critically about how innovation platforms and their principles align with specific governance or socio-political contexts.
C O N C L U S I O N S This study complements the many (case) studies that provide implementation and facilitation principles for, and lessons learned from innovation platforms. By focussing on design principles and setting realistic goals, this article seeks to provide decision support to research, development and funding agencies to think more critically about when, how and in what form innovation platforms can contribute meaningfully to agricultural development outcomes. As the implementation of innovation platforms can consume significant human and financial resource investments, research and development donors will require evidence on the return on investments. This requires investments in structured ME&L, which is missing in many innovation platform initiatives. Furthermore, attaining tangible development outcomes through innovation platforms requires time and flexibility which cannot be taken for granted in the current international AR4D landscape. The study provides decision support to development, research and funding agencies in determining whether and how innovation platforms can help them in achieving their objectives. It is clear in the sense that if the innovation platform approach is not suitable for a specific purpose,