From startup to scale: building evidence capabilities with a stage-based framework

• In Event: Roundtable Session 9
  In Refereed Roundtable Session: Digital Learning, Global Competencies, and Equity: Innovations, Evidence, and Policy Perspectives

Proposal

Technology- and AI-enabled education interventions are growing at a fast pace across diverse contexts, and hold vast potential to transform the learning lives of millions, particularly in the Global South. Despite data being cheap and readily available for ed-tech by comparison to “offline” education interventions, robust monitoring, evaluation and research (MER) practices remain scarce.
In IPA’s 2023 ed-tech evidence landscape review, we identified only a handful of use cases with a well-developed evidence base, and hypothesized several reasons for this: nascent technical capabilities; misaligned culture, and a lack of suitable tailored methodologies.
We also published a “stage based” evidence framework that argued for a simple approach of building MER capabilities aligned with the stage of scale of the intervention. We argue that evidence generation and use practices, rather than being a one-size-fits-all model where impact evaluations are the default, depend on the stage of an intervention: designing an intervention, refining it through pilots and tests, obtaining proof of concept, and then adapting and, eventually, scaling the intervention.
This presentation reports on applied research led by Innovations for Poverty Action’s Right-Fit Evidence Unit, through its Partnerships for Tech in Education (P4T-Ed) initiative, working with more than 10 EdTech providers across low- and middle-income countries. We have provided technical assistance using our stage-based framework, to build stronger cultures, capabilities, and methodologies for evidence generation and use in sector-leading organizations. In this presentation, we will share what we have learned about overcoming barriers to rigorous MER in ed-tech, how progress at each stage can help prepare for the next, and offer a more detailed picture of what “good looks like” for ed-tech MER at each stage of scale.
At the design stage, we have worked with startups and school networks to build clarity on their impact model by mapping user journeys and crafting a theory of change, and then articulating a MEL strategy around them. At the refine stage, we’ve helped partners upgrade technological infrastructure and analytical capabilities to enable rapid experimentation (such as A/B testing). For organizations preparing for rigorous impact evaluation in the prove stage, we’ve supported design and funding of rigorous studies. Finally, organizations that already had proof of concept and were aiming to adapt and scale to new markets or geographies received support to build data systems to ensure implementation fidelity.
By situating these insights within the CIES 2026 theme, we propose the session and subsequent discussion to be focused on five learning questions:
Why (or why not) should evidence generation and use practices depend on the stage of a tech- and AI-enabled education intervention?
What are the best practices for MER at each stage?
How can evidence intermediaries, technical assistance providers, and research organizations best support ed-tech providers to achieve those MER best practices?
The insights presented, while grounded in P4T-Ed engagements, address common challenges faced by governments, funders, and implementers worldwide. We aim to contribute to a more rigorous, transparent, and equitable MEL ecosystem, and ultimately help build stronger education systems that advance peace and prosperity.

Authors

• Benjamin J Tan, Innovations for Poverty Action

• Andres Parrado, Right-Fit Evidence Advisory Unit @Innovations for Poverty Action

• Stephen Gray

• Thomas Chupein, Innovations For Poverty Action

• Lucy Rimmington, Innovations for Poverty Action