Project Overview
A UK Russell Group university with 28,000 students wanted to build a student success platform — identifying at...
Technology Stack
Compliance & Standards
The Challenge
A UK Russell Group university with 28,000 students wanted to build a student success platform — identifying at-risk students (likely to drop out or underperform) using learning management system (LMS) engagement data, library access data, and academic performance data. ICO guidance on higher education data processing, UK GDPR Article 9 (disability and mental health data), Equality Act 2010 (must not create discriminatory proxy variables), and UCAS data sharing requirements were the requirements. Budget: £90,000.
Our Approach
Moodle API
course engagement (login frequency, resource access, assignment submission timing, quiz scores).
Alma/Primo Library System API
- library book loans, database access, study room bookings.
- Student Records System (Banner/SITS): module marks, attendance, personal tutor contacts.
Timetabling
lecture attendance from smart card swipe data.
All data
anonymised pseudonymous ID in the analytics platform — student names not visible in risk dashboards.
Risk Identification Model
Random forest classifier (scikit-learn): trained on 4 years of historical student outcomes (degree completion, classification, average mark).
Features
engagement velocity (trend in LMS engagement over first 4 weeks), attendance trend, assignment submission timing (late submission pattern), and personal tutor contact frequency.
Deliberately excluded features
protected characteristics (disability, race, ethnicity, gender) — direct or proxy.
Fairness audit
model performance metrics disaggregated by protected characteristic annually.
ICO Guidance on Student Data
- students must be informed about analytics use of their data (privacy notice).
- Article 9 (disability, mental health): special category data never used in risk model.
Equality Act Fairness Audit
- model performance (false positive rate, false negative rate) must not be materially different across protected characteristics.
- Automated decision-making: risk score is advisory — personal tutor makes all intervention decisions (no automated contact).
Personal Tutor Intervention Workflow
At-risk dashboard: personal tutor sees their allocated students, risk score (high/medium/low — not a number), and contributing factors in plain English ("reduced library visits compared to start of term").
Intervention recording
tutor logs contact attempt, meeting outcome, action agreed.
Escalation
tutor flags student for student wellbeing team referral if tutor contact unsuccessful.
All data
FERPA-equivalent access controls — tutor sees only their allocated students.
The Results
Platform live at 16 weeks, £82,000 — under budget.
ICO audit: data processing confirmed lawful.
Equality Act fairness audit: model performance not materially different across protected characteristics.
At-risk identification accuracy: 76% of students who withdrew in pilot year were flagged by the model in week 6 (vs week 12 under previous approach).
Intervention success rate: 68% of contacted at-risk students improved engagement.
University-wide withdrawal rate: pilot cohort 12.4% vs non-pilot 15.1% (2.7 percentage point reduction).
“76% of students who withdrew were flagged in week 6 of term. We now have 6 weeks to intervene — previously we had no warning until withdrawal. The 2.7 percentage point reduction in withdrawal rate sounds small — at 28,000 students, that is 756 students per year who stayed in education. The Equality Act fairness audit gave the university Senate the confidence to deploy." — Pro-Vice-Chancellor Education, UK Russell Group University (name withheld)”
Project Details
Related Case Studies
View AllEdTech Assessment Platform — KCSIE Compliant, SIMS Integrated
A UK EdTech startup needed to build a formative assessment and pupil progress tracking platform for UK primary...
EdTech Learning Management System — UK Further Education College
A UK further education college with 12,000 students needed to replace their legacy VLE (Virtual Learning Envir...
EdTech Adaptive Learning Platform — UK GCSE Revision App
A UK EdTech startup wanted to build an adaptive learning platform for GCSE revision — serving Year 10 and Year...
Ready to Transform Your Business?
Let's discuss how our technical expertise can help you achieve remarkable results.