From evidence gathering to translation into practice: Peer Learning with Dignitas, PAL Network and Rocket Learning 

In large-scale education work, organisations often face hard challenges as they seek to transform research and evidence into widespread adoption. After a year of rigorous research design and implementation, our seven Lab 3 FLN organisations (from India and Sub-Saharan Africa) were approaching this inflexion point. The gap between moving from “we know this works” to “the system has adopted this approach.” can be a stumbling block for promising interventions.  

Thus, we invited three organisations, represented by 

• Bilal Afroz, Head, Measurement and Evaluation (M&E) at Rocket Learning, India 

• Nancy Gikandi, Research and Development Manager at Dignitas, Kenya 

• Winny Cherotich, Action Manager at PAL Network, based in Kenya with members across 15 countries  

to share their hard-won lessons on navigating this transition during a peer learning session hosted by the Impact at Scale Labs’ Foundational Literacy and Numeracy cohort. The session was also attended by participants of GSF Early Years Lab and the Impact Accelerator. The group tackled three connected questions 

a. How do we build internal readiness for evidence-based shifts?
b. How do we enable governments to adopt evidence-informed strategies?
c. How do we bridge the rigour and relevance of research with the realities of education systems? 

Core insights that emerged across the three focus questions are highlighted below:  

A. How do we enable governments to adopt evidence-informed strategies?

1. Co-design with government from day one, through a champion and a working group. The most consistent principle across all three organisations was the importance of involving government partners from the very beginning as co-designers of the work. Dignitas starts with county governments, secures a single government champion, then convenes a technical working group spanning curriculum, assessment, teacher education and special needs. PAL Network’s member Uwezo Tanzania involved district and regional officials in inception, approvals, co-creation and training. This matters for scaling through government because shared ownership of the question produces shared ownership of the findings, which is what carries an approach into the system rather than leaving it as an external project.
    
2. Let evidence confront assumptions, and frame it as a solution to an existing problem. In Tanzania, officials who doubted PAL Network’s message that children were not learning despite high enrolment changed their position after going into the field themselves and watching a grade six child unable to read a grade two text. One sceptical quality assurance officer insisted that the worst-performing school he knew be included in a pilot. When 93 of 97 non-reading children had improved by the programme’s end, he became its strongest advocate, funding teacher transport and pushing for district-wide rollout.
    
3. Speaking the government’s language. Nancy Gikandi of Dignitas mentioned that organisations risk arriving as the presumed expert, only to have a government partner interrupt mid-presentation to ask whether they are aware a policy on the topic already exists. The lesson is to do the homework, acknowledge what is already in place, name the specific gap being addressed, and translate evidence into policy language. Decision-makers need meaning over methodology.  Rocket Learning made the same point that government buy-in improves when evidence is presented as a solution to an existing administrative problem, not as an external intervention being sold.
    
4. Relationships are the infrastructure. Nancy emphasizes that relationships are foundational infrastructure, shaped by power dynamics where organizations support, rather than override, existing system experts; when officials are trusted with real information and responsibility, they are more likely to buy in. Urmila from Peepul reinforces this, noting that effective government partnerships depend on partners feeling genuinely supported and aligned. She also highlights a shift in India, where the COVID-19 pandemic transformed government attitudes toward data from resistance to active demand. 
5. Make the data usable at every level. PAL Network simplifies evidence into village and district report cards, so a community member can see immediately how many children out of ten can read a text or do basic maths. Information must speak to each level of government, from the village elder to the national ministry.Dignitas learned that even a well-designed dashboard is not enough. When the team co-designed dashboards with Curriculum Support Officers (CSOs), Kenya’s middle-tier education leaders, officials asked for more than evidence. They wanted suggested solutions and analytics that told them what to do about the trends they were seeing. That feedback has driven roughly 20 iterations of the dashboard so far.
   
   
6. Presence is not the same as influence. Winny shared a reflection from PAL Network’s Nepal member centred on the lesson that dialogue invitations did not mean influence. As she put it: access may bring people into the room, but time is what turns presence into influence. A packed policy dialogue does not mean much when they don’t transform into sustained, layered engagement and real adoption. 
   

B. How do we bridge the rigour and relevance of research with the realities of education systems?

1. Rigour and scale can coexist. Rocket Learning’s experience with a J-PAL external evaluation illustrates how rigour and scale can coexist productively. A midline evaluation found that after nine months, parental engagement had improved significantly, but child learning outcomes had not. Rather than defend the programme, the team used the finding to strengthen it: adding more personalised nudges, more structured parent engagement, and stronger support for preschool workers. The revised “intensive model” improved child learning outcomes by approximately 0.2 standard deviations at an implementation cost of around US$1.17 per child per year, and was designed from the outset to be scalable. It now reaches approximately five million children annually. Alongside formal evaluation, the team runs continuous A/B tests on individual components, scaling only what proves itself. 
2. Pilot, look for signals, then open up. Dignitas applies a phased approach: every programme begins with a one-month pilot with a small group of teachers, gathering both qualitative and quantitative signals before opening to a wider cohort. This allows then detect early indicators of change, followed by a deliberate pause to reflect and adapt before scaling. 
3. Design for the system you are scaling into. PAL Network’s experience showed that adaptation was critical across each context. Its My Village programme, which combines learning camps, community libraries, SMS-based parental engagement and life skills sessions, took a different path to authorisation in Kenya, Nepal and Tanzania, shaped by each system’s norms. Existing structures matter can be used for scaling too. In Tanzania, monthly zonal teacher meetings became a channel for trained teachers to pass learning on to colleagues who had not yet joined. 

C. How do we build internal readiness for evidence-based shifts? 

1. Build a culture where evidence is everyone’s job. Nancy described how Dignitas has institutionalised an evidence-driven mindset, where every design or tool change must be backed by data, and teams regularly interrogate progress through quarterly and annual reflection meetings. A key cultural shift is moving away from blame toward learning, encouraging teams to surface challenges early and question assumptions even when things appear to be working. This is reinforced through “boot camps,” full-day sessions where teams analyze fresh field data and pair every insight with clear, actionable solutions for the educators they support.
    
2. Treat M&E as a learning engine. At Rocket Learning, monitoring and evaluation functions as a learning engine rather than a scorecard, actively helping teams identify what limits impact and where to adapt programming. Bilal illustrated this through several examples: over four years of child outcome assessments, children aged 3–6 consistently lagged in pre-literacy, particularly phonemic awareness, prompting an organization-wide shift in focus. The education and content team began testing targeted materials and instructional nudges while embedding pre-literacy more deliberately into twice-yearly Teacher Professional Development trainings, with operations teams reinforcing this priority through on-the-ground coaching. In another case, digital literacy emerged as a hidden mediator; because the programme operates via WhatsApp, parents and workers with limited platform familiarity engaged less, leading to weaker outcomes, which led the operations team to introduce digital literacy campaigns and stronger field support. Finally, a birth-to-three pilot in Haryana, conducted as a randomized controlled trial with Harvard University, revealed a clear dose-response relationship: while knowledge and attitudes improved even with low engagement, meaningful shifts in parental practices only occurred after crossing a certain engagement threshold, prompting a redesign toward more practice-oriented prompts and stronger support for caregiver action.

Conclusion  

The shift from proving “what works” to achieving system adoption is slow, non-linear, and often uncomfortable, but organizations that learn quickly, listen closely, and build deep partnerships can navigate it successfully.

To support this journey, GSF has developed a practitioner’s guide on measuring foundational literacy and numeracy, featuring case studies, assessment framework comparisons, and practical insights from seven organizations.