Evidence
Our approach to evidence
We believe that effective programs begin and end with evidence. We prioritize learning that is rapid, rigorous and immediately relevant—we want to get to work putting the learning to use. Great research informs our work, helps us measure our progress, and generates new, actionable insights that serve both our programs and the wider sector.
Over the past decade and over 60 rigorous, randomized experiments, we’ve built an approach to learning that sets us apart. We often begin with others’ research; too often evidence sits gathering dust, but we take evidence off the academic shelf and put it to use. All three programs in our portfolio are supported by randomized controlled trials (RCTs), and have been found to be some of the most effective, and cost-effective, programs studied.
​Beyond the initial evidence, our large, in-house team of world-class researchers continuously generates ongoing rigorous data on program fidelity and results. Many policy and practice efforts do not seek to systematically improve implementation: in fact, only 12% of education studies even measure implementation. To achieve impact at scale we focus on implementation—measuring it, accounting for it and improving it through systematic iteration. We’ve pioneered the use of iterative A/B testing in the social sector, and now run trials at a cadence of 10 per year to optimize for impact, cost-efficiency and scalability.
We view research not as an academic exercise, but as a practical tool for continuous learning and decision-making. Whether running replication trials in five countries, designing lightweight monitoring systems for scale, or supporting others to develop their own capacity for rapid, rigorous, repeated testing methods, we prioritize learning that is immediately useful. Many rigorous studies take ten years; we conduct ours termly. Our goal is to contribute to a broader movement of evidence-informed practice: one that builds on what works.​
Here we share some of the core evidence behind each of our programs. For more results and learnings, visit our insights page, and for more on A/B testing visit our A/B testing hub.
Choices
An RCT in Kenya, published in the American Economic Journal: Applied Economics, found the approach reduced teenage pregnancy—a proxy for unprotected sex and HIV risk—by 28 percent.
​
A separate RCT in Botswana involving 42,000 students and published in BMC Public Health, found a 40 percent reduction in teenage pregnancy when the program was delivered by near-peer facilitators.
Teaching at the Right Level
The TaRL approach consistently produces some of the largest effect sizes in the education literature. Developed by Pratham and J-PAL over 15 years, and rigorously evaluated across seven randomized studies, TaRL improves learning outcomes when implemented well.
​
In 2023, the Global Education Evidence Advisory Panel (GEEAP)—an interdisciplinary panel of leading global experts—named TaRL one of just three “best buy” interventions in its Cost-Effective Approaches to Improve Global Learning report, citing its strong evidence base and cost-effectiveness.
ConnectEd
An initial RCT, published in Nature Human Behaviour, found that ConnectEd delivered 0.89 standard deviations of learning per US$100—ranking it among the most cost-effective education interventions globally.
​
Further evidence, published in National Bureau of Economic Research (NBER) in collaboration with Columbia University, the University of Oxford, and J-PAL, showed consistent effectiveness across five countries, with average learning gains of 0.32 standard deviations. These effects translate to up to four years of high-quality instruction per $100 spent, placing ConnectEd in the top 1% most cost-effective education interventions globally.