Duration
My Role
Tools
Background
This project is part of a two-part case study:
• Part 1 — System Design, Governance & Data Architecture
• Part 2 — UX, Dashboards & Product Implementation
Problem Framing
Problem Space
Vocational schools serving low-income and rural communities often face a structural gap between their social mission and operational reality. While these institutions aim to protect disadvantaged students and prepare them for employment, they frequently operate without reliable data systems, consistent governance rules, or early-warning mechanisms. As a result, three recurring systemic issues emerge:
Late detection of disengagement — Declining attendance and academic performance are typically identified only at the end of the semester, when meaningful intervention is already limited.
Promotion without readiness — Students are advanced based on compassion rather than demonstrated competence, resulting in graduates who are not adequately prepared for the labor market.
High-stakes decisions without evidence — Admissions, promotion, and disciplinary decisions are often driven by intuition or informal discussions rather than structured, longitudinal data.
Together, these conditions create a silent failure mode: students remain formally enrolled and promoted while gradually disengaging, ultimately compromising their long-term outcomes.
Why I Started This Project
When disengagement is detected too late, the cost extends beyond academic failure to the erosion of long-term life opportunities. Students who graduate without meaningful skills face limited employability, reduced self-confidence, and an increased risk of long-term economic marginalization.
For educators, the absence of reliable data creates moral strain: teachers are forced to navigate trade-offs between compassion and fairness, and between institutional standards and individual hardship. At the institutional level, the consequences include reputational damage, declining trust from parents and industry partners, and a gradual deterioration in student intake quality.
Without structural change, these dynamics reinforce a self-perpetuating cycle:
underperformance → compassionate promotion → skill-poor graduates → weak alumni outcomes → declining trust → weaker future intakes.
Users & Stakeholders
Homeroom Teachers : responsible for student welfare, parent communication, and promotion recommendations.
Pain points:
Limited visibility into real-time attendance and performance
Late awareness of student disengagement
High emotional burden when making promotion decisions
Goals:
Detect problems earlier
Make fairer, evidence-based decisions
Intervene before disengagement becomes irreversible
Subject Teachers : responsible for academic assessment and classroom performance.
Pain points:
Manual grade tracking
Fragmented performance records
Limited coordination with homeroom teachers
Goals:
Record grades efficiently
Surface struggling students earlier
Maintain authority over academic assessment
Students: particularly those living alone in rented rooms or without close parental supervision.
Pain points:
Silent disengagement
Undetected absence
Lack of academic or emotional support
Goals:
Graduate with real skills
Feel supported rather than punished
Maintain dignity and autonomy
Pain points:
Late awareness of problems
Limited communication with schools
Lack of transparency
Goals:
Be informed early
Collaborate with teachers
Protect their children’s future
School Principal : final authority over promotion and disciplinary decisions.
Foundation : institutional owner; often risk-averse and resistant to operational change.
Education Authority : sets reporting requirements and curriculum standards.
System Logic
Design Principles
This system was not designed as a surveillance or enforcement tool. It was explicitly designed as an early-intervention and decision-support system — intended to surface risk patterns earlier, support humane responses, and improve the quality of institutional judgment.
Four principles guided all architectural decisions:
Intervention, Not Punishment
- The purpose of data collection is to enable earlier care, not stricter discipline.
- Alerts exist to trigger conversations and support — not sanctions.Evidence-Informed, Not Algorithm-Driven
- The system does not make promotion or disciplinary decisions.
- It only aggregates signals and presents structured evidence to human decision-makers.Governance-Aligned, Not Authority-Replacing
- Subject teachers retain academic authority.
- Homeroom teachers coordinate welfare responses.
- Principals retain final decision authority.Care-Oriented Data Use
- Student data is treated as a tool for protection and support.
- Not as a mechanism of control or monitoring.
Core Risk Signals & Scoring
Based on field observation and institutional practice, two variables were identified as the most reliable early indicators of disengagement:
Attendance Pattern
Academic Performance
• Daily quizzes
• Subject assignments
• Final exam results
Risk logic:
• Consistent low scores trigger performance risk
• Performance decay is tracked gradually over time
Rationale:
While teachers weight attendance and grades equally in promotion decisions, attendance collapse is a stronger predictor of future dropout.
The system uses a simple weighted risk index combining attendance and academic performance. However, this index is not treated as a truth machine. It exists only to prioritize teacher attention, surface silent deterioration, create a structured conversation starter.
Key design choices:
• Attendance and grades are weighted equally in promotion logic
• Attendance is weighted more heavily for dropout risk prediction
• Risk scores decay slowly over two weeks when behavior improves
This avoids punishing short-term setbacks while still surfacing sustained deterioration.
Decision & Communication Governance
The system was explicitly designed to support — not override — existing institutional authority structures.
Subject Teachers
• Input grades
• Flag academic concerns
• Retain authority over subject evaluation
• View only their subject-specific performance data
Homeroom Teachers
• View full student risk profiles
• Receive soft and critical alerts
• Initiate parent communication
• Log interventions
• Recommend welfare or academic actions
Principal
• Access aggregated school-level risk data
• Review promotion recommendations
• Make final high-stakes decisions
The system assumes that schools cannot intervene effectively without parental collaboration. For this reason, alerts are not sent directly to parents. All parent communication is mediated by the homeroom teacher.
Soft Warning (3 Days Absence)
• Homeroom teacher receives alert
• Teacher contacts parent through system
• Parent can reply directly inside the platform
• All communication is logged for auditabilityCritical Alert (7 Days Absence)
• Homeroom teacher receives urgent alert
• Teacher contacts parent
• Home visit is recommended
• Welfare context is documented
Cultural & Ethical Constraints
Many students in rural and low-income contexts live alone in rented rooms, lack parental supervision, or face unreported illness and economic hardship. In these cases, absence is not defiance — it is often vulnerability.
For this reason:
• Alerts are framed as care signals, not violations
• Intervention logs focus on context, not blame
• Risk scores are used to protect student rights, not to police behavior
The system is explicitly positioned as a welfare-first tool.
Conceptual Data Architecture
Data Architecture
This project required a minimal but coherent data architecture capable of supporting real-time monitoring, alerting, and intervention tracking without creating governance or privacy risks.
The system was intentionally designed around a small number of core entities and relationships to ensure clarity, auditability, and institutional trust.
Student
• Student ID
• Name
• Class
• Homeroom teacher ID
• Parent contact details
• Enrollment statusAttendance Record
• Student ID
• Date
• Status (Present / Absent / Excused)
• Recorded by (teacher / operator)Academic Record
• Student ID
• Subject ID
• Assessment type (quiz / assignment / exam)
• Score
• Date
• Entered by subject teacherRisk Profile
• Student ID
• Attendance risk score
• Performance risk score
• Combined risk index
• Last updated timestampAlert
• Alert ID
• Student ID
• Alert type (Soft / Critical)
• Trigger reason (attendance / grades)
• Created date
• Status (Open / Closed)Intervention Log
• Intervention ID
• Student ID
• Alert ID
• Action type (call / message / home visit)
• Notes
• Logged by
• TimestampParent Message
• Message ID
• Student ID
• Sender (homeroom teacher / parent)
• Message body
• Timestamp
• A Student has many Attendance Records
• A Student has many Academic Records
• A Student has one Risk Profile
• A Student can generate many Alerts
• An Alert can have many Intervention Logs
• A Student can have many Parent Messages
This architecture was chosen because the structure allows for:
• Real-time risk recalculation
• Transparent audit trails
• Clear ownership of actions
• Minimal data duplication
Most importantly, it keeps decision authority human while making deterioration patterns visible.