# Finding out what works Experimentation helps us test and improve [ideas](/undp-accelerator-labs/references/glossary.md#idea) and solutions, challenge our [assumptions](/undp-accelerator-labs/references/glossary.md#assumption) and create [evidence](/undp-accelerator-labs/references/glossary.md#evidence) for what works or where there is momentum for change. It creates a space for learning from failure – not only by showing what doesn't work, but by exposing important gaps in our knowledge and revealing new learning questions. This enables us to invest resources efficiently by identifying what works early in the R\&D process when it is still possible to change direction without big costs. Such early-stage proofs of concept often also build traction among ecosystem stakeholders, making R\&D more collaborative and generating momentum for system change. Experimentation involves multiple approaches best viewed as a continuum with three broad categories[\[1\]](#endnote-1): exploratory probes to understand system responses, trial-and-error to iterate and improve ideas, and formal tests to validate a solution or policy intervention. The choice of approach depends on whether outcomes are known or predefined, the need for quick insights versus rigor,[\[2\]](#endnote-2) available resources and skills, and what the unit of testing is: individual components or the complete solution. Experiments can have unforeseen effects: they can be good, or bad. We adhere to the principle to do no harm when we experiment, particularly when running behavioral trials. ## What we do to make big steps forward ### Creating evidence and legitimacy An experiment creates evidence that informs our next steps and gives an idea or solution [legitimacy](/undp-accelerator-labs/references/glossary.md#legitimacy), generating stakeholder interest to take it forward. For this, an experiment doesn't need to be large scale and expensive – we even prefer frugal experiments, using minimal resources to maximize learning. These often take the form of prototypes we can develop with a small investment. ### Prototyping to reveal needs However, we've discovered that [prototypes](/undp-accelerator-labs/references/glossary.md#portfolio) do more than just testing ideas quickly and creating evidence. Prototypes can catalyze R\&D in unexpected ways by making tacit needs explicit. When key stakeholders see a prototype, they often immediately recognize its value, even though they weren't previously aware of their needs or couldn't articulate them. They can only tell what they need, when they see it. When stakeholders recognize the value, they become eager to take the idea forward. If it's a key actor in an ecosystem, this can trigger ripple effects. ### Co-designing with stakeholders We never experiment alone – we are intentional about diverse and inclusive participation when we design and run experiments. We stimulate and enable other stakeholders to be involved, from grassroots innovators and entrepreneurs to government agencies. We often co-design experiments with these key stakeholders. This doesn't only help to spread the practice of experimentation, but also helps stakeholders take ownership of both the intervention and its results. Ideas and results carry further when people own them. ### Running multiple interconnected experiments This ownership becomes even more important when we run multiple interconnected experiments simultaneously as a [portfolio](/undp-accelerator-labs/references/glossary.md#portfolio). The logic of this approach is as follows: in complex systems, a single solution rarely creates systemic change; complexity needs to be addressed with complexity. We therefore probe a system at multiple points and levels through a set of connected interventions working together to create more desirable effects. These portfolios can be designed intentionally, or be more emergent and evolve into more intentional portfolios. ### Sharing our insights openly We practice [working out loud](/undp-accelerator-labs/references/glossary.md#working-out-loud) and openly share what we learn from our experiments – both what worked and what didn't. This transparency helps the ecosystem see what we're learning and creates opportunities for collaboration. For example, through our blogs, we frequently form unexpected partnerships with actors who learn about our work and reach out to get involved. {% hint style="info" %} #### Reflection questions These reflection questions guide us through designing, planning, and learning from experiments with ecosystem partners. * Which assumptions are we making about what might work, and how can we test these quickly and affordably? * Are we designing experiments to confirm what we already believe, or to discover what we don't know? * Which stakeholders need to be involved in designing and running this experiment for it to generate meaningful learning and ownership? * How do we strike a balance between speed and rigor? * What would "good enough" evidence look like for this stage of development? What constitutes credible evidence for stakeholders? * What could go wrong with this experiment, and how do we ensure we do no harm – especially with behavioral interventions? * How are we documenting not just what worked, but what didn't work and why? * How are we sharing our learning in ways that invite collaboration rather than just broadcasting results? * Which signals indicate there's momentum in the ecosystem to take this idea forward? {% endhint %} {% hint style="info" %} #### Methods and enabling technologies * [**Experimentation**](/undp-accelerator-labs/doing-r-and-d/6.-r-and-d-methods-and-enabling-technologies.md#experimentation) to test interventions and learn what works through trials ranging from quick probes to rigorous validation * [**Prototyping**](/undp-accelerator-labs/doing-r-and-d/6.-r-and-d-methods-and-enabling-technologies.md#prototyping) to test ideas and assumptions quickly and cheaply before committing more time and resources * [**Co-design**](/undp-accelerator-labs/doing-r-and-d/6.-r-and-d-methods-and-enabling-technologies.md#co-design) to design and run experiments together promoting ownership * [**Portfolio approach**](/undp-accelerator-labs/doing-r-and-d/6.-r-and-d-methods-and-enabling-technologies.md#portfolio-approach) to test interconnected experiments with ecosystem partners, enabling collective learning and triggering system transformation * [**Proof of concept**](/undp-accelerator-labs/doing-r-and-d/6.-r-and-d-methods-and-enabling-technologies.md#proof-of-concept) to test feasibility before committing more resources in further development * [**Behavioural insights**](/undp-accelerator-labs/doing-r-and-d/6.-r-and-d-methods-and-enabling-technologies.md#behavioral-insights) to provide evidence on human behavior and test nudges that influence positive choices and actions * [**Human-centred** **design**](/undp-accelerator-labs/doing-r-and-d/6.-r-and-d-methods-and-enabling-technologies.md#human-centered-design) to ensure people are closely involved in developing and testing ideas to fit their needs and context {% endhint %} *** ## Notes 1. See Christensen, Leurs & Quaggiotto (2017) [↑](#endnote-ref-1) 2. See Pop Ivanov et al. (2025) who describe the term “feasible rigor” as the art of generating robust evidence within the constraints of real-world development actions. [↑](#endnote-ref-2) --- # 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/5.-r-and-d-practices/finding-out-what-works.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.