Open Data for Agriculture and Nutrition: A Literature Review and Proposed Conceptual Framework

This paper begins by locating the Global Open Data for Agriculture and Nutrition project (GODAN) in the context of wider debates in the open data movement by first reviewing the literature on open data and open data for agriculture and nutrition (ODAN). The review identifies a number of important gaps and limitations in the existing literature. There has been no independent evaluation of who most benefits or who is being left behind regarding ODAN. There has been no independent evaluation of gender or diversity in ODAN or of the development outcomes or impacts of ODAN. The existing research on ODAN is over-reliant on key open data organisations and open data insiders who produce most of the research. This creates bias in the data and analysis. The authors recommend that these gaps are addressed in future research. The paper contributes a novel conceptual ‘SCOTA’ framework for analysing the barriers to and drivers of open data adoption, which could be readily applied in other domains. Using this framework to review the existing literature highlights the fact that ODAN research and practice has been predominantly supply-side focused on the production of open data. The authors argue that if open data is to ‘leave no one behind’, greater attention now needs to be paid to understanding the demand-side of the equation and the role of intermediaries. The paper argues that there is a compelling need to improve the participation of women, people living with disabilities, and other marginalised groups in all aspects of open data for agriculture and nutrition. The authors see a need for further research and action to enhance the capabilities of marginalised people to make effective use of open data. The paper concludes with the recommendation that an independent strategic review of open data in agriculture and nutrition is overdue. Such a review should encompass the structural factors shaping the process of ODAN; include a focus on the intermediary and demand-side processes; and identify who benefits and who is being left behind.

The review identifies a number of important gaps and limitations in the existing literature. There has been no independent evaluation of who most benefits or who is being left behind regarding ODAN. There has been no independent evaluation of gender or diversity in ODAN or of the development outcomes or impacts of ODAN. The existing research on ODAN is over-reliant on key open data organisations and open data insiders who produce most of the research. This creates bias in the data and analysis. The authors recommend that these gaps are addressed in future research.
The paper contributes a novel conceptual 'SCOTA' framework for analysing the barriers to and drivers of open data adoption, which could be readily applied in other domains. Using this framework to review the existing literature highlights the fact that ODAN research and practice has been predominantly supply-side focused on the production of open data. The authors argue that if open data is to 'leave no one behind', greater attention now needs to be paid to understanding the demand-side of the equation and the role of intermediaries. The paper argues that there is a compelling need to improve the participation of women, people living with disabilities, and other marginalised groups in all aspects of open data for agriculture and nutrition. The authors see a need for further research and action to enhance the capabilities of marginalised people to make effective use of open data.
The paper concludes with the recommendation that an independent strategic review of open data in agriculture and nutrition is overdue. Such a review should encompass the structural factors shaping the process of ODAN; include a focus on the intermediary and demand-side processes; and identify who benefits and who is being left behind.

Introduction
This paper reviews the existing literature on Open Data for Agriculture and Nutrition (ODAN). The review was commissioned as a background paper for a performance evaluation of the Global Open Data for Agriculture and Nutrition programme (GODAN). GODAN was established to help secure the potential development benefits of producing and using open data for agriculture and nutrition. GODAN's founding objectives include (a) empowering the ecosystem of organisations working on open data for agriculture and nutrition by working with policymakers to provide an enabling environment; and (b) increasing the supply and use of open data to stimulate innovation, equitable access, transparency and accountability, service delivery, and economic growth (Carolan 2016). These objectives inform the scope of this review.
This paper begins by first locating GODAN in the context of the wider open data movement. It reviews the literature on open data (OD), then open data for agriculture and nutrition (ODAN), before focusing in on GODAN specifically. The literature review was neither systematic nor exhaustive due to time and resource restrictions. The desk-based research was carried out during February and March of 2020 using an iterative process in which the focus and scope was adapted to reflect the emergent needs of the performance evaluation. Relevant literature was identified in consultation with domain experts, by using snowballing and reverse snowballing techniques, and using Google Scholar to identify more than 100 unique sources. The review identified significantly more existing literature on OD than on ODAN, and very little on GODAN specifically. The review also found that the existing literature on ODAN and GODAN is overdependent on studies and reports produced by the GODAN programme and affiliates, and would benefit from alternative perspectives. The literature review produced a new conceptual framework for analysing open data initiatives, which helps to identify opportunities to improve access, effective use and equity of application. The framework can be readily applied to open data in domains other than agriculture and nutrition. The literature review identifies gaps in the existing research and makes recommendations for future ODAN policy and practice. These include the need to improve the participation of women, people living with disabilities, and other marginalised groups in all aspects of open data for agriculture and nutrition.
The next section begins with some issues of definition before outlining the history and rationale of open data for development.  specific advantages of open data would be to increase transparency, enable accountability, and stimulate innovation and economic growth (Open Data Institute 2018). Where government data has been opened, this has enabled the innovation of mobile apps and online platforms including those that enable citizens to check the voting record of elected representatives, to report problems to local government, to live-track buses and trains, and to check crop prices 4 at multiple markets in real time (Pawelke et al. 2017 In 2007, advocates of open government data met and produced the Sebastopol Principles for Open Government Data, which argued that data should be complete, primary, timely, accessible, machine-processable, non-discriminatory, non-proprietary and licence-free. 5 Since 2011, when the Open Government Partnership 6 was founded, 78 countries have signed up to a process where each member submits an action plan co-created with civil society that includes concrete commitments to enhance transparency, accountability and public participation in government (OGP 2018). However, commitment to gender equity remains a concern. In the most recent Open Data for Development Report (IDRC 2019), despite 3,000 commitments made in country action plans, less than 1 per cent of countries included any gender commitments. The Open Government Partnership has since set an ambitious target of 30 per cent gender-sensitive commitments by the end of 2020 (ibid.). problem-oriented, the open data movement has shifted to focus more around regional, thematic, and issue-oriented initiatives (ibid.). It can be useful to understand this emerging ecosystem using typologies of data sources, geographies, or thematic sectors.

Open data ecosystems
In addition to government data, open data is increasingly drawn from other data sources including private companies (e.g. satellite data), civil society organisations (e.g. funding data), academia (e.g. research data sets) and even 'citizen data' (including citizen science and crowdsourced data) (Pawelke et al. 2017 In contrast to other industries, agriculture has a relatively broad spectrum of stakeholders. Agricultural producers range from parttime, small-scale farmers to high-tech multinational conglomerates, and supply chains begin in some of the world's most remote areas. 'Farm to fork' products go from being raw materials, through processing, trading, hedging and brokering to eventually make it to the customer's table. All of the stakeholders involved in these processes are potential producers as well as consumers of data. (Allemang and Teegarden 2017) The diversity of non-standard and non-interoperable data types and formats in the ODAN sector increases the complexity of the challenge. The theory of change for the Global Open Data for Agriculture and Nutrition (GODAN) programme refers to this diversity, describing the sector as composing 'an immensely complex system of actors from diverse fields, specialisations, jurisdiction and sectors of the economy acting individually and importantly, in concert' (Carolan 2016).
The GODAN programme was established in 2013 in response to these challenges to coordinate stakeholders and improve the application of open data in the sector. The GODAN vision statement reads: We are a group of actors working towards a world where the value chain for agriculture and nutrition is more efficient, innovative, equitable (e.g. by gender, socio-economic status) and accountable; from, for example greater yields and access to markets for farmers, through to more nutritious and safe food on plates. We believe that improving the open availability, use and enrichment of data, and meaningful engagement with stakeholders will enable this vision. We observe that the agriculture and food sector currently suffer from information asymmetries and closed data practices that limits progress, value generation and the fair distribution of resources.
To reverse these information asymmetries and to contribute to more equitable outcomes, GODAN focuses its activities on delivering a range of 'intermediate- level impacts' that include improved interoperability, increased innovation, transparency and accountability, new business creation, and improved service delivery. Also targeted are more accessible information products, data-driven decision-making, and increased access to data by disadvantaged communities. The GODAN vision also states, 'We will proactively seek to address gender balance in our engagement'.
Evidence of progress against GODAN objectives is reviewed in a later section but first, in order to situate ODAN in its broader context, the next section reviews the wider open data literature and builds a conceptual framework for analysing open data for agriculture and nutrition.

Barriers to open data use
In Section 2, we defined open data as 'data that can be freely accessed, used, modified, and shared by anyone for any purpose'. However, not everyone has equal opportunity to freely access, use, modify or share open data. Individuals have different endowments; they operate in different political and regulatory contexts; and they experience a range of different barriers to accessing, using, modifying, and sharing open data. This section reviews the existing literature to produce a conceptual framework for analysing the barriers that shape unequal access and use of open data.
Inequality of ability to make practical use of open data led early practitioners and researchers to pose critical questions. 'Open to whom?' was one such question. Gurstein (2011) was among those who argued that the publishing of open data was not in itself a development outcome and that it could lead to more inequality rather than less. In his paper 'Open data: Empowering the empowered or effective data use for everyone?', Gurstein argued that the primary impact of publishing open data is to empower those who are already advantaged in terms of their access to the digital infrastructure, skills and capacity necessary to exploit such data, leading to the further empowerment and enrichment of the already privileged. Rumbul (2015) has documented how urban men are overrepresented among users of digital governance technologies, and Davies and Perini (2016) have noted that open data use is similarly skewed towards middle class and well-educated users. These findings echo others from a wider review of the digital development literature, which found that the use of digital technologies often reflects, reproduces and amplifies existing patterns of (dis)advantage along lines including gender, ethnicity and caste/class (Hernandez and Roberts 2018). Gurstein (2003) argued that to ensure equitable access and effective use of open data, adequate attention would need to be paid to building awareness of open data and digital literacy abilities among excluded communities. The next sections focus attention on how this can be achieved in practice.   Roberts (2017)  a smaller group for whom affordability is the key barrier to access. A person with little or no disposable income is less able to afford internet connectivity than a relatively wealthy neighbour. Among those for whom the internet is both available and affordable, a lack of awareness can be a third barrier to access. Awareness-raising activities are generally necessary to create visibility of a new open data service as well as its practical relevance to people's lives and goals. Even when awareness exists, a lack of abilities such as digital literacy skills may be a barrier to access. Finally, even for those with the necessary availability, affordability, awareness and abilities, accessibility can be a barrier because the user interface is unavailable in a local language, lacks adaptations for people with visual impairments, or lacks content relevant to local priorities and concerns. As depicted in Figure 4.2, the 5 'A's can be conceived of as a series of concentric barriers to technology access and use (Roberts 2017).

Figure 4.2 The 5 'A's of technology access
Source: Roberts (2017) Roberts and Hernandez (2019) added a sixth 'A': agency. Agency is the capacity to act in the world in pursuit of your goals (Sen 1999). Empowerment can be measured as increases in agency (Ibrahim and Alkire 2007). Sometimes powerful social norms result in individuals lacking the agency to make effective use of a technology in their lives. For example, in some places people are socialised to believe that using mobile phones is inappropriate for respectable women or that the internet is only suitable for male / upper caste / universityeducated people. Open data organisations often seek to empower people to achieve their goals by increasing the supply and effective use of open data. From this perspective, the open data movement is centrally about enhancing the agency of individuals or groups. Doing so requires more than the supply-side provision of data and must include the demand-side increase of abilities and agency. Unless abilities and agency are increased, the benefits of open data will accrue disproportionately to the already privileged and advantaged. Schematically we find it useful to overlay the barriers of the 6 'A's onto the threepart process model (see Figure 4.3) in order to foreground how different barriers to technology access are encountered at different points in the open data process. Building the abilities and agency of marginalised actors in the open data ecosystem is core to activities on the demand-side. The two are often linked because building people's practical skills and competencies is one effective way to boost their self-efficacy and sense of agency (Bandura 1995). If the efforts of the open data movement were to end with the successful production of actionable open data, we would expect the majority of the benefits to accrue to organisations already privileged with the most capital, expertise and technical capabilities (Gurstein 2011). Open data initiatives can avoid the risk of reflecting, reproducing and amplifying existing (dis)advantage by including measures to increase availability and affordability, improve awareness and accessibility, and enhance the abilities and agency of under-represented and marginalised groups to make effective use of open data.

The 6 'A's of open data for development
The In some contexts, the political will may not exist to publish data to combat government corruption. Research by the Web Foundation (2017: 18) found that 'Open data portals do not contain the data people really want (e.g. data on budget, spending, contracting and company registers). These datasets tend to be highly opaque and are often the least open'. In countries such as Ethiopia, a lack of political will to open government data can be combined with poor internet infrastructure and a tendency of government to frequently shut down the internet for political reasons (Netblocks 2019

Awareness:
In order for data to be useful, potential users need to be aware of its existence, its importance and potential applications, and its relevance to their goals and priorities. Zuiderwijk et al. (2015)  Agency: Increasing the supply of material resources like data, devices and digital skills can help build a person's agency, but agency is also importantly about a person's psychological resources, including self-confidence and selfefficacy (Roberts 2015). Two people with precisely the same measurable skills and resources may have very different beliefs about their ability to accomplish goals (Bandura 1995). Bandura's body of research established that a person with an elevated sense of their own ability and entitlement is more likely to succeed than an equally talented or able person who has a low sense of their ability and entitlement. This is true for members of all demographic groups but when interventions aim to improve equity of outcomes, it is especially important for groups who have experienced persistent disadvantage or deprivation and who, as a result, have revised downwards their expectations and beliefs about what they are capable of achieving or deserve (Sen 1999

Gender barriers to open data
Gender inequalities pervade the technology sector in general, and the open data movement is no exception. Data is gender-biased (Criado-Perez 2019), is rarely gender-disaggregated (Buvinic et al. 2014), and large gender data gaps exist. Brandusescu and Nwakanma argue convincingly that 'as long as gender data gaps persist, any open datasets created based on raw data that does not adequately represent women will have limited potential to support transformative action on gender equity ' (2019: 194). Not all women are equally (dis)advantaged. Women from black and minority ethnic communities and women living with disabilities experience overlapping forms of disadvantage (Crenshaw 1989 (Brandusescu and Nwakanma 2019). Women are more likely to lack the capacity to contribute to and use data (Brandusescu and Nwakanma 2019). We know that women are under-represented in Open Government Partnership action plan consultations and that gender considerations rarely feature in national action plan commitments (Feminist Open Government 2019). As a result, issues which women find particularly important, including economic empowerment, political leadership, and violence against women, remain largely absent from Open Government Partnership action plans (Feminist Open Government 2019: 9).
These gender barriers to open data are particularly relevant for the purposes of this literature review as the GODAN vision statement specifically targets agriculture and nutrition systems that are more equitable (by gender and socioeconomic status) and GODAN's theory of change commits to proactively seek to address gender imbalance in its engagements (Carolan 2016).

Power structures
The previous sections reviewed a range of conceptual approaches to analysing open data and used them to apply a critical lens to the existing open data literature. We found the three-phase approaches of Pawelke et al. (2017) and Gurstein (2011)  seek to shape (hence the two-way arrows below). Together these elements provide us with a conceptual framework for analysing the structural context of technology access (the SCOTA Framework), which is illustrated in Figure 5.1.

Open Data for Agriculture and Nutrition
Governments, agribusiness, universities, and individual farmers are increasingly producers and consumers of data to improve yields, innovation, and income. Open data holds significant potential for improved environmental, economic and development outcomes. However, asymmetries of power, information and resources present potential barriers to equitable outcomes. A number of organisations have emerged to strengthen the ODAN ecosystem and increase the supply of open data for development.
The agriculture and nutrition sectors present specific challenges for the promotion of open data due to the volume and diversity of stakeholders. The ODAN ecosystem is composed of organisations of farmers, private sector companies, civil society organisations, governments, and multilateral agencies. It includes many specialist groups including agronomists, meteorologists, data scientists, lawyers, land rights activists, and academics. GODAN is one of several global networks that advocate for open data in agriculture and nutrition. The agricultural and nutrition sectors are composed of many inter-related subsectors and these inter-connect with many other sectors (e.g. food systems, ecology, nutrition, human health, the environment). To make matters even more complex, each related sector and sub-sector is composed of value chains with manysometimes thousandsof diverse stakeholders, both internal (e.g. farmers, producer groups, input suppliers, traders, processors, exporters, etc.) and external (e.g. regulators), who are generating different data types, formats, and ontologies and who themselves have differing data needs (Musker et al. 2018). The Research Data Alliance has more than 200 members and seeks 'to promote good practice in the research domain'. The Research Data Alliance hosts the Agricultural Data Interest Group (IGAD).

12
The Global Forum for Agricultural Research is an open and voluntary multi-stakeholder agri-food research and innovation forum with more than 600 partner organisations. 13 The Global Partnership for Sustainable Development Data is a multi-stakeholder network of more than 150 data champions harnessing the data revolution for sustainable development, including ODAN members such as the Food and Agriculture Organization and GODAN.
14 AgriCord is a global alliance of more than 200 farmer organisations from more than 50 developing countries that are part of the Farmers Fighting Poverty programme. It partners with CTA and others to provide support, seeking to strengthen farmer organisations' use of data and ICT.  weather data in the US and the UK generated substantial innovation resulting in economic and social impact through the creation of a multi-million dollar industry of weather tools and information products. Availability of real-time open weather data has enabled some farmers to improve their yields (Mey et al. 2019). GODAN has been active on the supply side encouraging government to publish open data sets. GODAN found that governments are often reluctant to open up agricultural and nutrition data for political reasons. For example, in Ethiopia, the ability of private companies to develop applications and tools using weather data is limited by a law that states that only the government is allowed to broadcast localised weather information . Tanzania recently introduced and revoked a law that made it illegal for anyone to publish data that invalidates or questions government statistics ). There may be economic reasons to keep land data sets closed in order make it easier to sell off plots of land to foreign investors (Mey, Miguel Ayala and Lokers 2018). In fact, 70 per cent of land data is unregistered (Mey et al. 2018). However, in many cases, land data cannot be made open because land ownership has never been documented and the cost of land registry maintenance is unaffordable .
Governments are important not only as producers of their own data. They are also key because of their power to require private sector and academic actors to open their data (Schaap et al. 2019). Governments can help to provide an enabling environment for open agricultural data through the creation of laws, policies, and regulation, but also by leading a public discourse on the benefits of ODAN. GODAN has successfully leveraged this power by including national governments in the ODAN ecosystem. GODAN played a central role in facilitating the 2017 Nairobi Ministerial Conference on Global Open Data for Agriculture and Nutrition at which 15 African ministers signed a declaration of commitment to publish open data and to raise awareness around open data for agriculture and nutrition. The Nairobi Declaration has inspired Francophone countries to develop their own network, the Conference d'Afrique Francophone sur les Données Ouvertes (CAFDO) (Schaap et al. 2019). GODAN has published a guide to opening data for agriculture to help other countries to join the movement (Open Data Charter and GODAN 2018). This is clear evidence of GODAN progress against its logframe objective 'to empower the ecosystem by creating an overarching enabling environment for open data for agriculture and nutrition'.
Standard-setting has been identified as essential supply-side activity to improve interoperability and so increase open data availability and use. The many subfields of agriculture and other related fields each have their own data standards. However, these standards tend not to be interoperable with one another, making it difficult for users to combine data sets within and across value chains to garner insights (Schaap et al. 2019). GODAN Action mapped agricultural standards and found a low level of interoperability (Schaap et al. 2019 Farmers are at the centre of the ODAN ecosystem, as illustrated above in Figure 6.1. This includes global agribusinesses down to individual smallholders. GODAN has multinational agribusiness companies, including Syngenta, among its members. A power imbalance exists between actors with regard to data availability and affordability (Ferris and Rahman 2017). Organisations are able to interpret and apply data for agriculture and nutrition in proportion to their capital, connectivity and capacity. This has led to new ethical concerns about who owns farm data and who profits from its aggregation and sale, as well as concerns about the lack of transparency about how data is ultimately used by more powerful actors (ibid.). Given the profit motive of private companies, some of their data sets are seen as commercial assets and are thus not opened up, even when they hold the potential to improve sustainable development outcomes (Schaap et al. 2019). Although many agribusiness companies in the private sector lack transparency and have shown minimal interest in opening their data, some are beginning to see opening data as part of maintaining a 'licence to operate'. Syngenta is one company that has made open data a centrepiece of its transparency strategy (Schaap et al. 2019).
Farmers and farm machinery increasingly generate primary agricultural data that can potentially be used to improve yields and profits. Although this data may be open and freely available, an affordability barrier often remains for smallholder farmers who lack the means to collect or use such data. 'Financial cost [is] the most common challenge mentioned by GODAN partners when working towards open data' (Musker et al. 2018: 8). Private sector companies are less likely to face an affordability barrier than smallholders who may have to depend on intermediaries (Gurstein 2011). Intermediary organisations from government, the private sector, universities, or NGOs, have emerged to help farmers interpret and apply open data. Intermediaries 'develop portals, apps, and tools that allow farmers to benefit from data on a range of topics, such as weather, infestations, or soil quality, that would otherwise be unavailable to them' (Schaap et al. 2019: 40).
Intermediaries have also piloted accessibility solutions to overcome barriers of language and literacies. Communicating open data through Interactive Voice Response (IVR) has been proposed as a potential solution but has been found to be expensive for farmers and service providers (Ferris and Rahman 2017). Universities and academics are often intermediaries and play active roles in producing open agriculture and nutrition data; 16 or take part in data discovery networks like the RDA and the Interest Group on Agricultural Data (IGAD) (Schaap et al. 2019). Funders of agricultural research are increasingly requiring that data collected during projects they fund is made public, including USAID and the Gates Foundation (ibid.). Ferris and Rahman (2017) found that farmers often lack awareness of open data and its potential benefits, and that this limits its uptake once data is published openly. Lokers et al. (2019)  The open data for agriculture and nutrition research that does exist is focused primarily on supply-side potentials, problems, and policy alternatives rather than evaluation of outcomes. This means that there is no substantial evidence base to support claims of the efficacy of open data for agriculture and nutrition pro-poor outcomes.
GODAN does publish its own series of 'impact narratives', which are very effective in communicating potential, but which cannot be considered to be independent evidence of achievement against its intermediate high-level impacts. GODAN Action's nutrition impact narratives have shown how open data collated by intermediaries like the Global Nutrition Report could be interpreted and repurposed to fit regional and local needs through the African Nutrition Scorecard to support local awareness-raising and advocacy for data-based decision-making on policies, programmes, and targeted interventions that meet  ; the African Scorecard on Nutrition found that accessibility was improved through the provision of hardcopy reports.
GODAN's own research found that the programme struggles to ensure that open data benefits farmers (Musker et al. 2018). It is widely accepted that smallholder farmers tend to have relatively little capacity to exploit data and that the gap in abilities and agency is greater in low-income countries, in rural areas, and for women (Schaap et al. 2019). This is particularly important in the areas of ODAN because making effective use of open data for agriculture and nutrition often requires domain-specific expertise as well as digital and data literacies. For example, nutrition data includes scientific classifications and measurements that may not be in the general lexicon. Similarly, because weather data is technical and scientific in nature, using it in its raw form requires scientific meteorological expertise . We were unable to find independent research literature that provides evidence of measurable progress in demand-side initiatives to build the abilities and agency of marginalised groups to make effective use of open data for agricultural and nutrition. GODAN and GODAN Action have run many training workshops and its online e-learning course has had an impressive 4,448 participants (Janssen 2019). However, people who are able to participate in online training programmes have levels of connectivity, digital literacy and language skills that are not typical in marginalised communities.
Withholding agency from open data programmes can sometimes be strategic on the part of citizens. Even when data is open and available, some people or groups may have reason not to trust data coming from government or see no probable benefit from providing it . Citizens who mistrust powerful actors may withhold their agency and not engage in open data initiatives. For example, local land data-holders may choose to keep their data closed if they fear that sharing it with government might lead to their land title being usurped rather than legitimised (Mey et al. 2018). These fears are not unfounded. An early land registration digitisation project in Bangalore found that digitisation efforts led to increased corruption and facilitated the acquisition of settled but unregistered land by powerful private companies (Benjamin, Bhuvaneswari and Rajan 2007), amplifying existing patterns of (dis)advantage.
The need to address gender inequalities and under-representation in open data has only recently begun to receive sufficient attention (Feminist Open Government 2019). Berdou and Lokers (2019) found that nutritional data tends not to be sufficiently disaggregated by gender and other demographics, making it more difficult to use data to advocate for the needs of women or develop policies and solutions that meet their needs. GODAN explicitly recognised the problems of gender data bias and the under-representation of women in ODAN when it wrote its gender mainstreaming policy in 2016 (GODAN 2016). A number of commitments were made in the gender mainstreaming policy including to build a diverse organisation and to integrate gender issues into the theory of change and logframe. However, the theory of change and logframe have not been updated to reflect these commitments. Although the GODAN theory of change mentions gender equity as an objective, the logframe lists no gender activities and contains no pathway activities designed to redress the gender inequities that GODAN identified in its gender mainstreaming policy. Although GODAN has contributed significantly to the evidence base on ODAN, it has not focused research attention on the intersection between gender and open data for agriculture and nutrition.
GODAN is playing a significant role in producing an evidence base for open data in agriculture and nutrition. However, there is little independent empirical or peerreviewed literature measuring the impact of ODAN activities. Most of the evidence that does exist was produced by GODAN (Berdou et al. 2017

Conclusions and future issues in ODAN
Based on our review of the existing literature, GODAN appears to have made substantial progress in achieving its supply-side objectives of expanding the quantity of organisations signed up to open data principles, and has played a central role in securing the increase in availability of open data for agriculture and nutrition. GODAN has been successful in creating high-level awareness and policy commitments that have contributed to generating political will and an enabling environment for open data in agriculture and nutrition. GODAN has also produced or contributed to a range of highly valuable tools, policy engagements, and research papers that have further advanced its objectives. GODAN members have played a range of influential intermediary roles, raising awareness and assisting others to interpret and apply open data.
This success is inevitably uneven across countries as well as among stakeholders within countries. There is less evidence of farm-level awareness, especially among rural smallholders, or of demand-side progress more generally. We found no evidence of sustained or systematic work to address the data accessibility needs of speakers of indigenous languages, print-illiterate stakeholders, or people living with disabilities. The GODAN theory of change and logframe lack activities or a planned pathway of change to address gender diversity and the under-representation of women in open data.
The relative lack of attention to demand-side factors, including building the abilities and agency of marginalised groups, may mean that GODAN operational objectives are met but with the unintended consequence that the organisations most empowered to exploit open data could be multinational agribusiness corporations.
The examples of Ethiopia, Tanzania and the US illustrate the point that initial gains in securing high-level support for open data can also be reversed when more authoritarian governments close civic space. We were unable to access data on GODAN's membership in an open and accessible FAIR format. The GODAN website lists members' logos but this data is not readily downloadable or searchable.

Gaps in the existing research literature
The majority of the research to date has been focused on supply-side mechanisms.
There has been no independent evaluation of who benefits or who is being left behind in open agriculture and nutrition data.
There has been no independent evaluation of gender or diversity in ODAN.
There has been no independent evaluation of the extent of the international development impacts of open data for agriculture and nutrition.
The existing research on open data for agriculture and nutrition is over-reliant on GODAN actors and open data insiders. This creates bias in the data.

Recommendations
ODAN organisations should increase the emphasis paid to demand-side and intermediary initiatives in order to ensure that marginalised communities are not left behind in open data initiatives.
Funders should commission a strategic review of ODAN that encompasses the structural factors shaping the process of ODAN and which pays attention to intermediary and demand-side processes. The SCOTA Framework may provide a useful starting point for such a review.
Research commissioners should prioritise independent research that addresses the gaps identified in the existing literature to determine who benefits most from ODAN, and who is being left behind in open agriculture and nutrition data.