A new study was just published in the Review of Educational Research: Identifying Effective Education Interventions in Sub-Saharan Africa: A Meta-Analysis of Impact Evaluations, by Katharine Conn from Columbia University. Here is a subset of findings (emphasis added):
I identify educational interventions with an impact on student learning in Sub-Saharan Africa. After a systematic literature search, I conducted a meta-analysis synthesizing 56 articles containing 66 separate experiments and quasi-experiments and 83 treatment arms…. A key finding is that programs that alter teacher pedagogy or classroom instructional techniques had an effect size approximately 0.30 standard deviations greater than all other types of programs combined. Limited evidence further suggests that pedagogical programs that employed adaptive instruction or teacher coaching were particularly effective.
In case you don’t have access, the earlier, open-access dissertation version has the same sample and the same findings reported in the abstract.
That version was one of six reviews that Anna Popova and I synthesized in our paper What Really Works to Improve Learning in Developing Countries? An Analysis of Divergent Findings in Systematic Reviews (open-access version).
David Evans' personal blog 
Limited evidence further suggests that pedagogical programs that employed adaptive instruction or teacher coaching were particularly effective.