Last Mile Education Technology: An Evaluation of Building Education Foundations through Innovation & Technology in Malawi
How to cite
Haroon, Maryiam; Gray-Lobe, Guthrie; Pugatch, Todd; Kadzamira, Esme; Kremer, Michael. 2026. "Last Mile Education Technology: An Evaluation of Building Education Foundations through Innovation & Technology in Malawi." AEA RCT Registry. AEARCTR-0018446. https://doi.org/10.1257/rct.18446-1.1
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
In Malawi, more than 80% of students at the end of grade 4 are unable to read a single familiar word (World Bank, 2018). High pupil-teacher ratios and limited resources constrain teachers’ ability to provide individualised instruction, contributing to the learning crisis in foundational literacy and numeracy. Digital solutions like personalised adaptive learning (PAL) can deliver high-quality, differentiated instruction at scale, but infrastructure and fixed costs have historically limited their reach in low-resource settings. The BEFIT program addresses infrastructure challenges by providing a comprehensive package that includes solar energy, tablets, and training, thereby lowering these fixed costs to support scalable implementation.
The Building Education Foundations through Innovation & Technology (BEFIT) Program will evaluate the large-scale implementation of onebillion’s onecourse software in Malawian primary schools. Implemented in partnership with Imagine Worldwide, the program aims to strengthen foundational literacy and numeracy skills. Evidence from nine small-scale trials in Malawi demonstrates the software’s effectiveness in improving early learning outcomes.
Using a randomised controlled trial with 348 schools in 11 districts of Malawi (surveyed at baseline) and then followed up with 250 schools at the first followup, the evaluation will assess the program’s effects on learning outcomes, cost-effectiveness, and long-term sustainability as management transitions from Imagine to the government. Specifically, we will test for the following:
Can BEFIT improve literacy and numeracy test scores for children in standards 1-4 at scale?
Can it improve student grade promotion, attendance, psychosocial indicators, and attention?
Can it enhance teacher practices, job satisfaction, and self-efficacy?
Additionally, if the study shows learning gains in numeracy and literacy from BEFIT, we will also examine the program’s cost-effectiveness compared to other experimentally evaluated education interventions in Malawi using the framework of Angrist et al. 2025.
Schools will be randomly assigned to one of three conditions: (i) control, (ii) standard dosage (150 minutes per week), or (iii) high dosage (300 minutes per week). This represents the first large-scale evaluation of a personalized adaptive learning (PAL) program in primary schools in Africa.
Intervention
The randomized evaluation assigned eligible schools to one of three conditions: (a) control, (b) standard BEFIT dosage, or (c) high BEFIT dosage. Two schools were allocated to receive a capitation grant and were not assigned to either of the groups.
Standard BEFIT dosage group: Schools are assigned to schedule student use of the onecourse software for up to five periods per week (about 150 minutes weekly, or 27 hours annually).
High BEFIT dosage group: Students are scheduled to use the software for approximately 240/245 minutes per week (about 43 hours annually).
Control group: Schools continue with existing practices.
Experimental design
Schools eligible for the study must meet three criteria: 1) no prior BEFIT implementation, 2) a suitable roof for solar panel installation and 3) deemed accessible by the implementing partner.
The research team shared a sampling frame with Imagine Worldwide to begin school eligibility assessments. Using the 2024 Education Management Information System (EMIS) data, we generated a randomly selected list of 574 schools across 12 districts (three rural and one urban per region). Single-gender schools were excluded to maximize precision in gender subgroup analysis. All 150 schools previously included in the National Numeracy Program (NNP) evaluation were retained. After Imagine Worldwide conducted eligibility checks, 441 schools were deemed eligible.
Due to budget constraints, the research team dropped 93 of the 441 eligible schools, including one district entirely. A baseline survey was completed with the remaining 348 schools across 11 districts. Information was collected from headteachers regarding school name, location, enrollment, number of teachers and classrooms, interest in technology, demand for the BEFIT program, and their experiences as school leaders. Additionally, a Student Tracking Sample was created by collecting information on randomly selected students, including their names and genders. These students were assessed using the Early Grade Reading Assessment (EGRA) and Early Grade Mathematics Assessment (EGMA) in Chichewa, along with an attention measure. A short household survey was administered to capture demographic and socioeconomic data. For the Teacher Tracking Sample, four teachers (one per grade, Standards 1–4) were identified from administrative data.
After baseline, schools were randomly assigned to one of three groups, stratified by tercile of the school’s predicted baseline achievement (capturing variation from geography, enrolment and pupil-teacher ratio), solar panel size and school level baseline achievement.
Control group (N =221): No BEFIT program
Standard-dosage group (N = 100): Students scheduled to use the one-course software for 150 minutes per week.
High-dosage group (N = 25): Students scheduled for 300 minutes per week.
Note: Two schools were excluded from the allocation to the treatment and control group as they were randomly selected for the capitation grant.
Follow-up surveys and assessments will be conducted with sampled schools (finalized at followup). In rounds after the program is fully transitioned to government management, additional students in Grades 1 and 2 will be added to the tracking sample. This will allow the evaluation to measure the program’s impact on new cohorts and compare learning outcomes with those observed in the original tracking sample.
Design details
Cited by
No citation data available for this registration yet.