# 3. Setting up an R\&D Capability *Setting up an R\&D capability involves orchestrating many moving parts, and there are no formulas to guarantee success. In this chapter, we share our insights from establishing the Accelerator Labs as a global operation in 115 countries. We will discuss how to navigate the early stages – what to focus on? What does an R\&D unit look like – what roles and skills are essential? And what does it take to develop a distributed capability that drives learning across countries and contexts? While not exhaustive, with these insights, we aim to give you practical pointers for establishing your R\&D function.*Navigating the early stages ## Navigating the early stages There are thousands of ways to set up an R\&D function,[\[1\]](#footnote-1) but all involve orchestrating many moving parts simultaneously. Whether you call it a lab, unit or team, you'll find yourself juggling multiple competing priorities:[\[2\]](#footnote-2) * Identifying the needs or challenges you will address and articulating that as a compelling value proposition * Defining your strategic focus and deciding on your key methods and underlying logic * Recruiting the right talent, building the team, and establishing governance structures * Securing sustainable financial resources and gaining political buy-in from leadership, donors and other key stakeholders * Finding partners who can provide specific expertise, skills, or help you connect with key actors * Establishing systems and conditions to make outcomes, practices, and learning irreversible[\[3\]](#footnote-3) * Setting up infrastructure for documenting insights and outcomes, and establishing evaluation frameworks to demonstrate progress * Finding ambassadors who will promote your initiative while managing skeptics and resistance * Building an authentic brand and developing your communication strategy Each element matters, and entire guides could be written about any one of them. What's more, the specific challenges and priorities will vary greatly depending on your context. Rather than attempting to cover all these elements, we want to share a framework that helped us navigate the early stages of building the Accelerator Lab network – when we were building the aircraft while already airborne. This framework focuses on managing the dynamic balance between three dimensions (Figure 9) that we found ourselves constantly negotiating:

Figure 9: Three dimensions to balance: strategy, activities and building legitimacy when setting up an R&D function (Inspired by Centre for Public Impact, 2018)

* **Strategy:** *What is our north star? What is the* [*direction*](/undp-accelerator-labs/references/glossary.md#directionality) *of travel?* * **Activity:** *What do we do to get to that dot on the horizon?* * **Legitimacy:** *How do we create and maintain political buy-in?* These aren't sequential steps but interconnected areas of attention that depend on each other and co-evolve together. Strategy alone won't create legitimacy – you need to explain why the work matters and show tangible results. Activity without strategy won't achieve your ambitions – you'll look busy but will be wasting resources, time and energy on the wrong priorities. [Legitimacy](/undp-accelerator-labs/references/glossary.md#legitimacy) without activity will not last long – you need action to generate evidence, success stories, and insights.[\[4\]](#footnote-4) And activity without legitimacy won't build momentum – even good work struggles to get adopted without buy-in. The key point is, you need to balance all three areas and ensure they are aligned. ## Designing the R\&D team In the Accelerator Lab network, each team includes three core positions: Head of Experimentation, Head of Solution Mapping, and Head of Exploration. Your R\&D function will likely look different. Teams come in all sizes, and your context and strategic priorities will determine which capabilities matter most. With this in mind, when designing your team,[\[5\]](#footnote-5) we believe it's more helpful to think about the roles people can play and the skills they bring, rather than providing detailed job descriptions. Through our experience across our network, we've identified twelve such roles. They represent functions that team members can fulfil: where one person often plays multiple roles, and several people might share a single role, depending on personal capabilities and strengths. These roles are fluid; at various stages of an R\&D journey, certain roles become more prominent while others recede into the background.[\[6\]](#footnote-6) Their significance can change based on specific information needs, the momentum that is building or needs to be created, the stakeholders involved, or any unexpected opportunities that arise. To respond to these emerging needs and opportunities, a certain open-endedness is needed. Each role involves specific skills, approaches, and [mindsets](/undp-accelerator-labs/references/glossary.md#mindset) that enable teams to navigate [uncertainty](/undp-accelerator-labs/references/glossary.md#uncertainty), make sense of complex issues, and create positive change. But the roles also involve potential pitfalls when overdone or underperformed.

Explorer

Seeks possibilities

Investigates uncharted territories to discover new possibilities, emerging challenges and opportunities. Explores the adjacent possible and seeks out grassroots innovations to unearth existing solutions. Continuously scans for weak signals of change and emerging patterns across communities to map possible futures, helping communities prepare for what's coming while finding pathways to shape more desirable outcomes.

Key capabilities:

  • Scanning environments for weak signals and emerging patterns of change
  • Combining existing elements in novel ways to create new possibilities
  • Exploring alternative futures and scenarios through collective imagination

Pitfalls:

  • Staying in exploration mode without converting findings into actionable insights
  • Favoring exotic or trendy tech solutions over practical adjacent possibilities
  • Fixating on doom scenarios without exploring alternative pathways forward

Analyst

Crafts insights

Transforms complex data into meaningful insights that inform understanding and action. Collects, analyzes, and presents information in ways that reveal patterns, generate new perspectives, and enable evidence-based decision making, while empowering communities rather than extracting value from their data.

Key capabilities:

  • Systematic collection and analysis of traditional and non-traditional data
  • Translating complex data into accessible, actionable insights
  • Using data visualization and storytelling to help people make sense of complex and evolving situations

Pitfalls:

  • Extracting data from communities without ensuring they benefit from the analysis
  • Failing to connect quantitative findings with human experiences
  • Getting stuck in analysis paralysis and losing sight of what's actually needed

Ethnographer

Immerses into communities

Engages deeply with communities to understand their lived experiences, cultural contexts, and needs. Builds authentic relationships that enable empathetic insight into how people most affected by an issue experience challenges and develop their own coping strategies and grassroots innovations. Translates ethnographic insights into compelling stories and recommendations that inform decision-making and help develop more appropriate, effective solutions.

Key capabilities:

  • Empathetic observation and engagement with local knowledge holders
  • Building trust across different cultural contexts and with gatekeepers
  • Recognizing and valuing community assets and local knowledge systems

Pitfalls:

  • Imposing external frameworks rather than truly listening
  • Focusing on problems and deficits while overlooking community strengths and assets
  • Letting personal biases shape which voices are heard and amplified

Experimenter

Develops evidence

Designs and conducts structured tests to validate assumptions and generate reliable insights about what works, where, why, and for whom. Creates learning processes that balance rigor with speed, enabling intelligent failure that builds legitimacy for innovations through evidence. Designs coherent portfolios of experiments and interventions that collectively address complex challenges and create momentum for system change.

Key capabilities:

  • Designing experiments to test hypotheses and challenge assumptions
  • Balancing rigor with speed in testing approaches to accelerate learning
  • Co-designing experiments with communities and stakeholders to learn together and build ownership

Pitfalls:

  • Creating experimental designs misaligned with how the issue actually unfolds in practice[1]
  • Playing it too safe and avoiding the intelligent failures that generate real learning[2]
  • Treating communities as test subjects rather than active change-agents.[3]

Architect

Integrates elements

Synthesizes diverse elements into coherent structures and functioning solutions. Develops frameworks, processes and tools that help innovations work effectively across technical, social, and environmental domains. Combines data, technology, and people to create flows of information and knowledge that enable ecosystems to become smarter together.

Key capabilities:

  • Synthesizing diverse inputs into coherent frameworks that others can use, adapt and build upon
  • Facilitating interoperability between different technologies, capabilities and approaches
  • Balancing strategic, technical, legal, financial and human considerations in design

Pitfalls:

  • Designing overly rigid frameworks that don't respond to system dynamics and changing contexts
  • Overcomplicating designs with too many dependencies for solutions to work effectively
  • Losing sight of the big picture and people's needs and potential while focusing on technical integration

Visionary

Inspires change

Develops compelling shared visions that align stakeholders and inspire action. Helps groups imagine possibilities beyond current constraints, challenge outdated mental models, and create practical pathways toward transformative goals for sustainable development.

Key capabilities:

  • Finding and building shared purpose across diverse stakeholders and interests
  • Reframing situations to reveal new perspectives that mobilize collective action
  • Holding space for ambiguity while futures take shape

Pitfalls:

  • Creating visions too far ahead of where stakeholders are ready to go
  • Neglecting to build alliances and a movement to carry the vision forward
  • Overlooking quick wins that help people believe in the longer journey

Advocate

Creates legitimacy

Builds support and recognition for both the R&D function and the innovations it generates. Creates legitimacy by engaging decision-makers and influencers, articulating the added value with evidence, and identifying champions. Helps stakeholders understand why R&D matters for sustainable development and how experimental, learning-based approaches can address complex challenges.

Key capabilities:

  • Identifying and building relationships with influential stakeholders as champions
  • Translating R&D work into compelling narratives that resonate with different stakeholders and demonstrate the value add
  • Managing sceptics and converting critics into allies by sharing success stories, presenting evidence, and inviting them to participate in the process

Pitfalls:

  • Using technical jargon that alienates non-specialist audiences
  • Spending too much energy converting sceptics instead of empowering ambassadors
  • Over-stating the results or over-hyping the value of R&D, eroding credibility

Maker

Makes ideas real

Translates ideas into testable prototypes and proof of concepts that can be experienced, evaluated, and refined. Builds physical, experiential or digital manifestations of ideas that allow for rapid testing and iteration, helping to validate assumptions and surface tacit needs by demonstrating to stakeholders what's possible.

Key capabilities:

  • Bringing passion for making things, building quick and low-cost prototypes to test ideas early
  • Engaging stakeholders actively in design and testing to incorporate their perspectives and create buy-in
  • Designing structured experiments that maximize learning from each iteration

Pitfalls:

  • Overthinking instead of building, spending too much time imagining how something might work rather than making it real
  • Testing only in controlled settings instead of real-world conditions where solutions must work
  • Making prototypes without enough detail to yield meaningful feedback

Orchestrator

Forms collectives

Brings together diverse stakeholders and builds effective networks for collaboration and impact. Creates conditions for collective action by connecting the right people, establishing shared purposes, and developing "middleground" spaces where institutional and grassroots initiatives can work together effectively. Builds trust across different actors and sectors to enable authentic collaboration and knowledge sharing.

Key capabilities:

  • Mapping ecosystems to identify key actors, including gatekeepers and bridge-builders
  • Articulating compelling reasons for diverse actors to connect around shared purpose
  • Establishing relational infrastructures, in the middleground, that enable knowledge flows, collaboration and value creation

Pitfalls:

  • Ignoring existing networks or collectives and trying to build from scratch
  • Creating dependency where the collective only functions with orchestration
  • Over-planning the process instead of allowing the collective to evolve organically

Facilitator

Enables co-creation

Designs and guides collaborative processes that enable diverse groups to work productively together. Creates and provides safe spaces for collaboration, co-creation, and meaningful dialogue where diverse perspectives can interact. Manages power dynamics and helps groups navigate through uncertainty, tension, and perceived asymmetries of knowledge or power.

Key capabilities:

  • Designing and leading productive collective processes
  • Managing tensions and bridging different perspectives
  • Creating psychological safety for not knowing, authentic participation and learning

Pitfalls:

  • Over-facilitating by imposing structure or pushing forward when the group needs space, losing touch with participants' actual energy and interests
  • Giving too much airtime to extraverts or participants with power and seniority
  • Cramming in too many activities for the time available, not achieving depth in conversations or learning

Mapper

Navigates complexity

Makes visible the hidden structure of complex systems by mapping relationships, flows, and patterns that shape how challenges persist or change. Helps stakeholders see themselves within larger ecosystems and identify where small interventions might create a significant impact. Transforms overwhelming complexity into actionable insights by revealing connections between people, resources, problems, and possibilities.

Key capabilities:

  • Visualizing system dynamics in ways that help diverse stakeholders see their role and potential for influence
  • Understanding systems through different levels and perspectives, revealing how individual experiences connect to larger patterns
  • Building maps collaboratively with communities rather than extracting and representing their knowledge

Pitfalls:

  • Mistaking the map for the territory and missing what can't be captured visually
  • Focusing on aesthetics over function, creating visualizations that don't support shared understanding or decision-making
  • Working at the wrong level of abstraction, including too many details that obscure what matters for action and decision-making

Documenter

Captures knowledge

Records, organizes, and shares critical information and learnings throughout the R&D process. Makes insights accessible, usable and actionable for others by translating complex findings into formats that different audiences can understand and apply. Builds collective memory that enables ecosystems to learn faster and build on each other's discoveries.

Key capabilities:

  • Translating complex R&D findings into multiple formats (stories, data, frameworks) that different audiences can understand and use
  • Working out loud: sharing insights and learnings as they emerge rather than waiting for final reports
  • Building knowledge repositories and commons that enable others to find, adapt, and build upon innovations

Pitfalls:

  • Creating knowledge graveyards where information is stored but never accessed or used
  • Documenting only successes while missing the valuable lessons in failures and unexpected outcomes
  • Using formats and language that exclude the very communities who contributed the knowledge
*Table 3: The twelve roles within a R\&D team.* ## Building a globally distributed capability When we launched the Accelerator Labs, we set out to establish a learning network:[\[10\]](#footnote-10) a way to connect teams across the globe while keeping them grounded in their local realities. This network became the foundation for what would later evolve into a globally distributed R\&D capability. Through the network and its diversity we see patterns of emerging challenges and priorities, as well as new solutions and possibilities that might otherwise stay local. Let us first explain how we look at networks. A network consists of actors who are connected and interact with each other to achieve common goals. In everyday language, we might say we "built" a network, but that's misleading. Networks are not like skyscrapers that are usually built based on a predetermined plan or detailed blueprint. We consider networks as living systems.[\[11\]](#footnote-11) We can't predict and control how they'll evolve, what form they'll take, or whether they'll thrive. They grow through relationships, evolve through interactions, and strengthen through shared experiences. Therefore, instead of *building*, we prefer to talk about *curating*[\[12\]](#footnote-12) a network. Curating involves creating the conditions, initiating and coordinating activities, and providing the support systems and tools that help the network to: * Establish a relational infrastructure[\[13\]](#footnote-13) by connecting actors and strengthening their ties * Generate intelligence by sharing, analysing, and connecting insights * Develop a practice and logic that is in line with the network's purpose What do these conditions, activities, and support systems look like in practice? ### Creating conditions Networks are living systems: they emerge and thrive depending on the [conditions](/undp-accelerator-labs/references/glossary.md#condition) around them.[\[14\]](#footnote-14) While we can't control the network itself, we can design these conditions to a certain extent. With this in mind, we focused on creating conditions to help our Network self-organize, learn collectively, and generate value through emergence and serendipity. The following proved essential for establishing the Accelerator Lab Network: > #### **Preparing the landing strip** > > Before Labs arrived in their host offices, we invested in preparing the ground for a smooth landing. While recruitment was underway, we ran programs with the host offices to introduce the intent, logic, and key methods of the Accelerator Labs. Beyond the country offices, we delivered workshops with regional teams. These engagements helped us identify champions who could advocate for the approach, while also understanding skeptics' perspectives and what was needed to address their concerns and create legitimacy. Each interaction helped us refine our narrative, sharpen our value proposition, and build broader institutional support for a capability that was open-ended and ahead of demand. > > #### **Setting a mandate** > > A mandate provides both direction and autonomy, defining the “what” while leaving room for the “how.” We gave Labs a mandate for "directed improvisation"[\[15\]](#footnote-15) – setting a clear purpose to accelerate learning for the Sustainable Development Goals, while giving them freedom to determine how to best achieve it in their contexts. But we also set clear boundaries: no causing harm, no unethical behavior, no extractive data practices. Within these boundaries and purpose lies a vast space for Labs to experiment and find their own way. > > #### **Cultivating diversity** > > Our Network brings together people from different backgrounds, regions, and cultures, with varied career paths and educational foundations. They offer distinct perspectives on sustainable development and work in diverse ecosystems, addressing both shared challenges and context-specific problems. Despite this diversity, we're united by a shared mindset and a set of principles captured in our [fundamentals chapter](/undp-accelerator-labs/understanding-r-and-d/2.-the-fundamentals-of-collective-r-and-d.md). This balance – enough alignment to collaborate, enough difference to learn from each other – is essential for our work. By bringing different worldviews and knowledge systems together in our activities – Learning Circles, Tuesday calls, R\&D Raves – we enhance our collective intelligence.[\[16\]](#footnote-16) > > #### **Creating psychological safety** > > Learning in uncertainty requires vulnerability – admitting what we don't know,[\[17\]](#footnote-17) sharing what didn't work, and acknowledging we're navigating uncharted territory together.[\[18\]](#footnote-18) We foster this by modeling and rewarding curiosity, which propels us into the adjacent possible and often yields unexpected but valuable discoveries. We promote intelligent failures[\[19\]](#footnote-19) and treat almost everything as a prototype when we don't know how to do things. We create spaces where people can reflect on their actions and journeys, including how they experienced them emotionally. A simple way to start creating psychological safety is with "emotional check-ins"[\[20\]](#footnote-20) at the beginning of meetings and workshops, asking teams or participants: "How are you feeling?" “What's your presence level today and why?” or "What's on your mind?" Another way to create safety is to engage in “post-mortems”[\[21\]](#footnote-21) after large delivery efforts, creating a space to celebrate wins and learn from collective design processes. Innovation is emotional work;[\[22\]](#footnote-22) making space for discussing the emotional aspects of our work is critical for nurturing a learning culture. ### Coordinating activities

Figure 10: Overview of network activities that strengthen connections, enable collective learning, and advance our practice

A host of activities (Figure 10) keeps our Network vibrant and alive. These activities are initiated and coordinated both by the global team and by Labs themselves. Some activities recur frequently, establishing a rhythm or heartbeat for the Network. Others happen periodically, and some are one-off events designed for specific purposes. Together, these activities strengthen connections and create a sense of belonging, enable knowledge flows that drive collective learning, and provide spaces to reflect on and advance our practice.

Figures 11 & 12: Bootcamps for onboarding the first cohort of UNDP Accelerator Labs in Kigali (September 2019) and Quito (October 2019).

> #### **Convening weekly drop-in calls** > > Since the beginning, every Tuesday at 7 AM New York time the Network comes together on Zoom. These weekly calls began during recruitment as check-ins with country offices and evolved into "open mic" sessions once teams were onboarded. Over time, this weekly rhythm became like the Network's heartbeat. The content varies: Labs often share learnings from experiments and field research, the global team often facilitates discussions about future directions or presents back what the Network is doing and learning. Every now and then we invite UNDP policy teams to connect our R\&D work with ongoing development programming, or bring in external researchers working on topics like user-led innovation, serendipity, or learning systems. Beyond knowledge sharing, these calls also serve as a space to celebrate achievements and provide emotional support in times of crisis, or to commemorate the loss of a colleague. > > #### **Onboarding new recruits** > > When we onboarded the first cohort of 60 Labs and the second cohort of 31 Labs, we brought everyone together for intensive bootcamps to induct new Labbers into the Accelerator Labs. The first cohort gathered in person for these bootcamps (Figures 11 and 12), while the second cohort – due to the pandemic – connected entirely online. The bootcamps aimed to connect new recruits with the intent and mandate of the Labs, introduce our practice and key methods, provide guidance for navigating UNDP's bureaucracy, and – importantly – connect Labs with each other across regions. These cross-regional connections created the relational infrastructure for the Network to perform as a learning network. > > #### **Working out loud** > > One of the ideas we tested in this network experiment was for Labs to write frequent blogs, initially monthly, sharing openly with the rest of the world what they're working on, or planning to work on, and what they're learning.[\[23\]](#footnote-23) The hypothesis was that a lightweight approach to reporting could happen continuously, not just at the end (as reporting usually works).[\[24\]](#footnote-24) Working out loud through blogging enables us to engage with the ecosystem, signal what we're working on, and create opportunities for external partners to connect with Labs' work, which sometimes results in unusual partnerships. Having a continuous stream of blogs from the field also helps the global team and Network get an almost real-time sense of what is going on in the Network and spot patterns across countries. This became a core part of our intelligence system. To promote frequent blogging and spotlight well-crafted blogs, we instituted a blog award, handed out periodically during Tuesday calls when exceptional posts caught our attention. > > #### **Establishing learning rhythms through short cycles** > > Labs develop short learning plans outlining the challenges and learning questions they'll focus on, the methods they plan to use, and which ecosystem actors they'll partner with. Initially structured around 100-day learning cycles, the frequency has decreased over time as their work deepened. After each learning cycle, Labs report back on what they did, learned, and discovered. These plans and cycle reports have become another vital input for our knowledge base, helping us spot patterns across the Network, connect Labs working on similar challenges, and identify where specific expertise lives. This, in turn, allows us to match that expertise with both internal and external demands. > > #### **Designing for serendipity** > > Beyond our core activities, the Network organically initiated multiple gatherings that served specific purposes. The Heads of Experimentation hold bi-weekly meetings, as do Solution Mappers and Heads of Exploration, to share experiences and discuss technical aspects of their work and advance their practices. The global team regularly convenes leadership series with senior management to discuss future directions, address concerns, and maintain institutional support. Regional Retreats in 2022, held after the pandemic, strengthened connections, mapped regional priorities and explored future directions. Codification Fests in 2024 documented our practice, work that resulted in this guide. We also continuously run learning-focused events around specific questions, from Learning Circles to R\&D Raves[\[25\]](#footnote-25) discussing next practices,[\[26\]](#footnote-26) bringing together Labs, knowledge holders, and partners to accelerate learning on frontier issues.[\[27\]](#footnote-27) These various gatherings happen both online and in person, depending on purpose and context. At Lab gatherings (regional retreats, codification fests, and other convenings), we create structured opportunities for Labs to share what they're working on, what they're curious about, and what support they need. These exchanges often spark unexpected collaborations and yield surprising results.[\[28\]](#footnote-28) ### Establishing support systems We developed several support systems to maintain an almost real-time understanding of what the Network is doing and learning. These systems create the infrastructure for capturing, analyzing, documenting and sharing knowledge as it emerges across our globally distributed network. > #### **Going where the conversations are** > > Keeping a global network connected requires creating multiple channels for ongoing dialogue. With teams spanning all time zones, our conversations happen asynchronously around the clock. While Teams serves as our corporate platform, Labs organically adopted WhatsApp from the very first bootcamp; it's the go-to for quick questions and peer support. WhatsApp's accessibility on mobile phones and its global user base made it the natural choice. Most conversations – and certainly the most interesting ones – happen on WhatsApp. Rather than forcing Labs to use our corporate platform, we recognized that a network votes with its feet, we have to follow where it goes to be part of the daily conversation. > > #### **Using a little computational power** > > With 270+ people actively generating insights across 115 countries, we needed help making sense of the volume of information being created. We use data science to analyze multiple sources: blogs, learning plans, solution databases, experimentation records, and conversations on WhatsApp and Teams.[\[29\]](#footnote-29) This helps us understand what's happening in the Network and spot patterns. The informal chatter has proven particularly valuable: it captures how knowledge naturally flows between Labs. What people talk about reveals what they're working on, and what they're working on is what they're learning about. Developing a data governance framework and analyzing these conversations with AI tells us what people are discussing and shows us where knowledge lives in the Network: who to engage with when we quickly want to deepen our understanding of specific topics. > > #### **Creating a distributed knowledge base** > > Our knowledge base emerged by creating a mandate to share learning publicly, and through the digital exhaust of the activities of the Network. It helps us understand what Labs are doing and learning, while giving them access to their peers' work in other countries. Over time, this has become our collective memory. The knowledge base is decentralized: blogs live on our country websites. We also built a central database that captures learning plans, and co-designed with Labs databases to record experiments and grassroots solutions across the Network. We've tried to resist demands for more reporting and keep contribution processes lightweight, automating where possible (such as blog scraping) while remaining mindful of feature bloat that could make knowledge creation burdensome. This distributed knowledge base creates great opportunities for finding unexpected insights, knowledge nuggets or patterns, but that potential is only realized when processes are put in place to learn from serendipity. For example, we continuously scan blogs, analyze learning plans, and conduct 'pattern sweeps' to spot emerging insights and connections. > > #### **Developing an innovation commons** > > Innovation builds on the ideas, insights, and knowledge of others. Recognizing this, we've created the SDG Commons[\[30\]](#footnote-30) (Figure 13), an AI-powered digital public good that makes our work openly available to the global development community. It provides access to grassroots innovations, experiment results, blogs, and curated insights from across the Network. The platform enables pattern spotting from distributed intelligence across the Network. It includes thematic boards on our R\&D priorities, with consolidated knowledge from our Network on topics like digital financial inclusion, circular economy, and food systems. By sharing our data and knowledge openly, we enable other innovators to pick up where we left off and combine it with their own ideas, expertise, and solutions.[\[31\]](#footnote-31)

Figure 13: SDG Commons, an AI-powered platform providing open access to grassroots innovations and curated insights from the UNDP Accelerator Labs network.

## Evolving from a learning network to distributed R\&D We set an open-ended experimental mandate, allowing Labs in the Network to pursue the sustainable development problems and opportunities they saw as most relevant in their context. With labs spread across 115 countries each pursuing their own priorities, once we paid attention to the patterns, we gained access to a rich, real-time picture of what's emerging in sustainable development. A network of labs added unique value to UNDP by curating "liquid knowledge,"[\[32\]](#footnote-32) the knowledge that is usually difficult to embed in traditional bureaucratic structures. It includes the relational, experiential insights from our experiments alongside lived citizen experience, with a focus on those areas that drive new positions in digital and other system transformations. It created and captured programme signals that were local and context-specific, but then also shared across continents and thematic goals to confirm whether a particular phenomenon is an anomaly or an enduring new feature of sustainable development.

Figure 14: Distributed R&D cycle.

At one point, we needed to make a strategic shift. We were experimenting with new methods and technologies, innovating the “how.” But we realized we also needed to focus deliberately on the “what:” the new insights, positions, possibilities and value (Figure 14) that emerge from new ways of working. We called these "next practices": the actionable outcomes of R\&D.[\[33\]](#footnote-33) By synthesizing learning across the Network through a bottom-up innovation function, it was possible to identify where sustainable development had gotten stuck and where new opportunity spaces were emerging. ## Moving forward You may not run a global network of 91 Labs across 115 countries, and your R\&D function will likely look very different from ours. But whether you're setting up a small innovation team in a single organization or building a regional R\&D capability, the fundamentals remain the same. You'll need to balance strategy, activity, and legitimacy. You'll need people who can play different roles as circumstances demand. And you'll need to create conditions for learning, coordinate activities that build momentum, and establish reflection mechanisms that make sense of what you're discovering. What we've shared aren't formulas but patterns we've observed and approaches that have worked for us. Your context will demand its own interventions, activities and strategies. The key is to start, not to overplan, learn as you go, and remain open to the network or team showing you what it needs to thrive. Building R\&D capability is an ongoing experiment in creating the conditions for learning and adaptation. *** ## Notes 1. Setting up an R\&D function also means mistakes will be made, and you will be criticized for that. But only those who are, or have been, "in the arena" – as Theodore Roosevelt pointed out in his "Citizenship in a Republic," delivered at the Sorbonne in 1910 (Roosevelt, 1910) – understand the full complexity of balancing innovation with institutional realities, managing competing stakeholder expectations, and navigating the inevitable setbacks that come with pioneering new approaches. [↑](#footnote-ref-1) 2. For an in-depth exploration of these competing priorities, see Jesper Christiansen's (2019) reflections on the trade-offs inherent to running a lab. [↑](#footnote-ref-2) 3. Thinking about irreversibility – not to confuse with immutability – should perhaps start from day one, particularly in lab contexts where lifespans are often limited by funding cycles or changes in the institutional context or broader ecosystem. A useful tool for this is Kautsar Anggakara's Irreversibility Framework (2025). [↑](#footnote-ref-3) 4. See for example Kit Lykketoft’s (2016) reflections on creating legitimacy for Denmark’s MindLab. [↑](#footnote-ref-4) 5. It's beyond the scope of this guide to go deep into team design and development. But we do recommend the European Commission's research on collaborative policymaking (see Inchingolo et al., 2025) to enhance collaboration and become smarter together – not just as a collective, community or ecosystem – but also as an R\&D team. [↑](#footnote-ref-5) 6. Vaughn Tan's concept of "open-ended roles" is particularly relevant for R\&D teams navigating uncertainty. He suggests leaving 20% of each role explicitly undefined, allowing team members to shape this portion through "negotiated joining." See his blog "Unfrozen from the start" (Tan, 2023b) and his book "The Uncertainty Mindset" (Tan, 2019). Also see Tan (2015). [↑](#footnote-ref-6) 7. See Christensen, Leurs & Quaggiotti (2017) [↑](#footnote-ref-7) 8. As Dave Snowden (2012) suggests, we should design experiments with failure in mind – and a certain percentage should actually fail, otherwise we’re not pushing the boundaries. We usually learn more from failure than from confirming our hypotheses. [↑](#footnote-ref-8) 9. When designing experiments, we need to remember that experiments are never neutral: interventions in communities affect real lives. [↑](#footnote-ref-9) 10. For some introductory reading on learning networks see Fiona McKenzi’s (2021) scoping paper: “Building a culture of learning at scale” or this blog “ [Learning about learning networks](https://medium.com/centre-for-public-impact/learning-about-learning-networks-9c92553649ac),” (McKenzie, et al., 2023). Also worth reading is David Ehrlichman’s (2021) book: “ [Impact Networks](https://www.converge.net/book).” [↑](#footnote-ref-10) 11. In line with Ehrlichman (2021, p. 57). [↑](#footnote-ref-11) 12. While we use "curating," others have found different metaphors to avoid the language of "building" networks. David Ehrlichman (2021, p. 57) prefers "cultivating," drawing on the garden metaphor of nurturing growth, while June Holley (2013) speaks of "weaving" to emphasize how relationships form the fabric of networks. [↑](#footnote-ref-12) 13. Rye (2023) [↑](#footnote-ref-13) 14. Also see Ehrlichman (2021, p. 57) [↑](#footnote-ref-14) 15. We borrowed this term from Yuen Yuen Ang, who uses it to describe China's approach to development – combining central direction with local experimentation. See Ang (2016) “How China Escaped the Poverty Trap.” For how we applied it, see also Lucarelli (2019). [↑](#footnote-ref-15) 16. See Nesta (2019, p. 40) [↑](#footnote-ref-16) 17. See Adrian Brown (2019), in his blog, “The power of ignorance in policymaking.” [↑](#footnote-ref-17) 18. On the importance of psychological safety for learning and innovation, see Edmondson (1999 & 2018). [↑](#footnote-ref-18) 19. On intelligent failures and learning from failure, see Edmondson (2011). [↑](#footnote-ref-19) 20. For guidance see Razzetti (2019). [↑](#footnote-ref-20) 21. See for example the “Project Timeline” (tool 9a) of UNDP’s Hackers Toolkit (UNDP, 2017). [↑](#footnote-ref-21) 22. Emotions are an inherent part of innovation work, yet we rarely talk about this openly. Throughout our journey with the UNDP Accelerator Labs, we've experienced the full spectrum of emotions that come with navigating uncertainty and challenging the status quo. While emotions like optimism help us create momentum for change and build the trust necessary for collaboration, innovation work can also be emotionally taxing. Sustained negative emotions are a primary driver of burnout among practitioners. This is why we believe it's so important to acknowledge and honor the emotional dimension of our work. It creates psychological safety and nurtures our resilience for uncertainty work. [↑](#footnote-ref-22) 23. Visit our blogs at [↑](#footnote-ref-23) 24. The practice of working out loud turned out to be much harder than we initially thought, with several banana peels appearing along the way. It was a successful experiment as it created a mandate to communicate in a more humble way, and it led to a body of work that represents the diversity of the network. In practice, it was a communications shift more radical than initially expected, with many lessons for large public sector R\&D efforts. [↑](#footnote-ref-24) 25. R\&D Raves are learning spaces where we identify R\&D opportunities to connect system knowledge with experimentation, grassroots innovation and exploration. See Lucarelli (2025). [↑](#footnote-ref-25) 26. UNDP Accelerator Labs. (2025) [↑](#footnote-ref-26) 27. See UNDP Accelerator Labs (2025) [↑](#footnote-ref-27) 28. For example, our Codification Fests yielded unexpected results, including the collaborative creation of our catalogue of R\&D services lines (see UNDP Accelerator Labs, 2025a) that made visible the network's collective capabilities and offerings. [↑](#footnote-ref-28) 29. With a data governance framework consulted among network members. [↑](#footnote-ref-29) 30. [https://sdg-innovation-commons.org](https://sdg-innovation-commons.org/) [↑](#footnote-ref-30) 31. For a deeper exploration of innovation commons and how shared knowledge resources accelerate innovation, see Potts (2019). [↑](#footnote-ref-31) 32. See Van Eck, et al. (2025) [↑](#footnote-ref-32) 33. See Lucarelli (2025) [↑](#footnote-ref-33) --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://bas-leurss-personal-organization.gitbook.io/undp-accelerator-labs/doing-r-and-d/3.-setting-up-an-r-and-d-capability.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.