Sabi is an initiative of Education for Equality
Evidence
Built on research. Judged on results.
Every design choice in Sabi traces back to a study. And from the first pilot call, Sabi is measured against the best phone-tutoring programmes in the world.
01 — The problem
Years in school, not years of learning.
7 in 10
Nigerian 10-year-olds cannot read a single sentence.
World Bank, 2022
98M
children are out of school in Sub-Saharan Africa.
UNESCO, 2024
16%
of planned learning actually happens in school.
Attendance × lesson completion × teacher presence
02 — The channel
Why a phone call, and not an app.
600 million people in Africa have no internet. A voice call over 2G reaches nearly everyone, on the phone they already own, with nothing to install.
03 — Design choices
Every choice has a study behind it.
Research from Nigeria, Ghana, Tanzania, Madagascar and global randomised trials.
9×Market maths beats abstract maths
Children solved comparable subtraction tasks nine times more often when framed as a market sale than in school format (48.5% vs 5.4%).
→ Why every lesson starts in the market.
Weber, Bogler & Vollmer 2024 · North-West Nigeria
1.5–2 yrsof learning in six weeks
A World Bank RCT in Nigeria found AI tutoring delivered one and a half to two years of learning gains in just six weeks.
→ Why the tutor is AI, not a recording.
World Bank 2024 · benefit–cost ratio 161–260
7% → 45%Reading, with no teacher in the room
In Tanzania, children given only tablets went from 7% to 45% reading ability in fifteen months.
→ Why a child can learn alone, at home.
Global Learning XPRIZE
$0.50–5per student, per year
Ten to a hundred times cheaper than the alternatives. At $2 a student, reaching 100,000 children costs $200,000 a year.
→ Why it can reach every family.
Self-hosted open-source stack
04 — The bar
What good looks like.
Published results from phone and chat tutoring. These are the numbers Sabi's pilot will be read against.
| Programme | Learning gain | Cost | Dose |
|---|---|---|---|
| ConnectEd (J-PAL) | 0.33 SD | ~$12 / child | ~3 hours over 8–9 weeks, by phone |
| Rori (Rising Academies) | 0.36 SD | ~$5 / child / year | ~32 hours over 8 months, by chat |
| Teaching at the Right Level | 19% → 79% | Varies | Readers before and after learning camps |
SD = standard deviations of learning gain. Teaching at the Right Level shows readers before → after.
05 — The pilot
Lagos, Q4 2026. Here's what we'll report.
100–500 students in high-burden local government areas. Every number below comes from consented children only, with no raw child audio in any export.
Level movement
The share of children who move up at least one reading or maths level from their first-call placement.
Learning gain
A short baseline and endline probe, so gains are reported as an effect size beside published programmes.
Dosage
Calls answered and completed, minutes of real lesson time, and how many children call back the next week.
Cost per child
Delivery cost per learner, and learning per dollar, so funders can compare like with like.
Go on, give Sabi a call.
Hear what a lesson sounds like, right here in your browser.