Case · Education
EdTech — early dropout signal + pass-rate scenarios
PROBLEM
The Problem
240K learners' study logs, assessments, and environment data piled up, but the team could not answer "which learner is at high dropout risk in the next 4 weeks?" or "what study pattern would get them to pass?" — no model gave a justified, learner-specific answer.
APPROACH
XimTier Approach
Study activity, assessments, and environment are regressed to detect early dropout signals, with SHAP contributions for each at-risk learner. Reverse What-If solves the study-hours / problem-count / review-cycle mix needed to hit a target pass rate, and produces a rationale that is explainable to learners and parents.
OUTCOMES
Outcomes
Dropout detection 4 weeks early (84% accuracy)
Pre-simulated pass-rate scenarios
Learner / parent-facing explainable rationale
Tutor capacity +40% (learners per tutor)