Learning Analytics
Performance prediction, engagement tracking, and early-warning systems that help teachers reach students before they fall away.
- Student performance prediction
- Engagement tracking
- Weak-area detection
- Dropout risk modeling
- Cohort comparison
- Teacher-facing dashboards
Why analytics is teacher-empowering
The image of “learning analytics” people often have — surveillance, automated scoring of students against opaque rubrics — is the wrong one to copy. The right framing is much simpler: what data would help a teacher reach the student who is quietly slipping?
Across Ethiopian secondary schools, dropout and failure are heavily concentrated. A few students per class are at risk; identifying them earlier means a teacher (or a parent, or a tutoring program) can act before the student disengages entirely. This is a high-leverage analytics problem because the intervention cost — a conversation, a study group, a follow-up call — is small relative to the lifelong cost of a student leaving school.
We treat analytics as a teacher tool first, an administrator tool second, and a researcher tool third — in that order.
What we’re researching
- Student performance prediction. Estimating exam outcomes from earlier-term work and practice scores, with honest uncertainty intervals. (A prediction with no error bar is just a guess.)
- Engagement tracking. Light-touch signals — practice frequency, problem-set completion time, return rates — that can flag disengagement without surveillance overhead.
- Weak-area detection. Pinpointing the specific topics within a subject that are dragging a student’s overall performance, so review time gets aimed precisely.
- Dropout risk modeling. Identifying students at elevated risk of dropping out while there is still time to intervene. Calibrated against real cohort outcomes, not toy benchmarks.
- Cohort comparison. Letting schools and districts see how their cohorts compare on apples-to-apples measures, without naming individual students.
- Teacher-facing dashboards. Designed for the constraints of an actual teacher’s day — fast to read, mobile-first, actionable suggestions rather than raw numbers.
Where we are
This pillar is exploring. We’re building the data foundation through other Atenu network deployments (scholarships engagement, practice platforms) and starting on dropout-risk modeling once we have enough longitudinal signal. No public analytics product yet.
If you are an education researcher with longitudinal Ethiopian secondary data, or a school willing to share anonymized engagement data for research, we’d love to talk.