A regional public university with 22,000 students and a proud 85-year history was watching its mission erode one dropped student at a time. A 6-year graduation rate of 47%—14 points below the national average for comparable institutions—meant that more than half of the students who enrolled with aspirations never realized them. The advising team of 45 carried caseloads of 490 students each, making proactive outreach statistically impossible. The data to identify at-risk students existed across 8 disconnected systems, but no advisor had the time or tools to correlate it. The university needed a way to see every student clearly and act early enough to matter.
Education
AI Student Success Platform & Retention Transformation
Primary Outcome
Increased 6-year graduation rate from 47% to 58% and retained $28M in additional annual tuition revenue
58%
Graduation Rate
$28M
Tuition Retained
34%
Retention Improvement
18 Mo
Implementation
Project Overview
The Challenge
1. Impossible Advising Caseloads
With 45 advisors serving 22,000 students, each advisor carried a caseload of 490 students. Industry best practice is 250–300 for comprehensive advising and 150–175 for at-risk populations. At these caseloads, proactive outreach was mathematically impossible. Advisors spent 90% of their time responding to students who had already reached crisis—those who had already dropped a course, failed an exam, or stopped attending—rather than reaching students before problems escalated.
- 490:1 student-to-advisor ratio vs 250–300 best practice
- 90% of advising interactions reactive rather than proactive
- Students in crisis waiting average 8.3 days for advising appointment
2. Student Risk Data Fragmented Across 8 Systems
Attendance data lived in the LMS (Canvas), grades in Banner SIS, financial aid status in a separate system, library access records in the library management system, food pantry usage in a spreadsheet, campus housing records in yet another system, and student mental health contacts in a counseling system with strict access restrictions. No advisor could see a holistic picture of any student. A student missing class, struggling financially, and sleeping in their car was invisible until they withdrew.
- 8 separate systems with no real-time integration
- Financial aid processing delays causing enrollment interruptions mid-semester
- LMS engagement data not accessible to advisors in any workflow
3. Financial Aid Processing Delays Causing Enrollment Loss
Students whose financial aid packages were delayed, reduced, or miscalculated were frequently unable to register for classes until the error was resolved—a process averaging 3.2 weeks. Each week of delayed registration reduced the probability of that student completing the semester. 340 students annually dropped mid-semester or failed to register for the following semester due directly to financial aid processing errors or delays.
- 3.2-week average financial aid resolution time
- 340 students annually losing enrollment due to financial aid processing delays
- No early alert system for impending financial aid issues
4. No Early Intervention Capability
The university had a formal early alert process—faculty could submit a concern about a student—but it was paper-based and averaged 11 days from faculty concern to advisor contact. By the time an advisor reached a struggling student, the student had often already made the decision to withdraw. First-generation students, Pell Grant recipients, and students of color were disproportionately represented in withdrawal statistics and least likely to self-refer to advising.
The Solution
Unified Student Data Platform & Risk Scoring
Built a unified student data platform integrating all 8 systems—Canvas LMS, Banner SIS, financial aid, library, housing, food pantry, tutoring, and counseling (with appropriate privacy controls and FERPA-compliant access tiers). Machine learning models trained on 6 years of historical student outcomes generate daily risk scores for every enrolled student across four dimensions: academic engagement, financial stability, social connection, and life circumstance.
Data Integration
8 systems unified with FERPA-compliant access tiers; counseling data protected with additional consent controls
Risk Model
4-dimension risk scoring (academic, financial, social, circumstance) updated daily for all 22,000 students
AI-Powered Advisor Triage & Caseload Optimization
Rather than asking advisors to monitor all 490 students equally, the platform presents each advisor with a daily prioritized action queue—the 8–12 students most in need of contact right now, with specific risk factors and suggested conversation topics. Advisors document all contacts in the platform, enabling outcome tracking that feeds back into model improvement. Average advisor daily productive student contacts increased from 4 to 14.
- Daily prioritized action queue of 8–12 students per advisor
- Specific risk factors and suggested intervention approaches per student
- Advisor contact documentation with outcome tracking
- Same-day appointment availability for high-risk students flagged by AI
Financial Aid Early Warning & Automation
Integrated financial aid monitoring identifies students at risk of aid disruption 30 days before the issue becomes enrollment-threatening—time enough for financial aid counselors to resolve processing errors, connect students with emergency funding, or arrange payment plans. Automated processing rules resolved 68% of common financial aid errors without human intervention, reducing average resolution time from 3.2 weeks to 4 days.
- 30-day financial aid risk warning before enrollment impact
- Automated resolution of 68% of common financial aid processing errors
- Emergency fund connection automated for qualifying students
- Average resolution time reduced from 3.2 weeks to 4 days
Results & Outcomes
58%
6-Year Graduation Rate
The 6-year graduation rate increased from 47% to 58% over the 3-year cohort measured—the largest single-cohort improvement in the university's history. Improvement was most pronounced among first-generation students (+15 points) and Pell Grant recipients (+13 points)—the populations most underserved by the previous reactive advising model.
34%
Sophomore Retention Improvement
First-to-second year retention—the most predictive metric for eventual graduation—improved from 72% to 96% among students who received a proactive AI-triggered advising contact in their first semester. The platform identified and successfully intervened with 1,847 students who would have been lost without early action.
$28M
Additional Annual Tuition Revenue
Each retained student represents an average of $24,800 in annual net tuition revenue. The improvement in retention and graduation rates generates an estimated $28M in additional annual net tuition—approximately 18x the annual platform operating cost. The ROI calculation excludes reputational value of improved graduation rate rankings.
48 hrs
Time to Advisor Contact for At-Risk Students
Time from risk signal detection to advisor contact for flagged students fell from an average of 11 days (previous paper-based early alert) to 48 hours (AI-triggered priority queue). Students contacted within 48 hours of first risk signal showed 3x higher intervention success rates than students contacted later.
4 days
Financial Aid Resolution Time
Average financial aid processing resolution time fell from 3.2 weeks to 4 days, eliminating the mid-semester enrollment disruptions that were causing 340 students annually to lose their registration. Emergency fund connections were automated for qualifying students within 24 hours of need identification.
4.4/5
Student Experience Score
Student satisfaction with advising services increased from 3.1 to 4.4 out of 5. The most significant driver in qualitative feedback was the experience of being contacted proactively—students described advisors 'actually knowing who I am and what I'm dealing with' as a transformative change in their relationship with the institution.
Technologies Used
Data Integration
Ellucian Banner APICanvas LMS Data APIWorkday Student (Financial Aid)Azure Data Factory (ETL)
AI & Machine Learning
Python / scikit-learnAzure Machine LearningSnowflake Feature StoreSHAP Model Explainability
Advisor Platform
Custom Advisor Portal (React)Salesforce Education CloudTwilio (Student Outreach)Microsoft Power BI
Business Impact
1,847 Students Retained Who Would Have Left
Modeling based on pre-intervention risk scores and historical outcome data estimates that 1,847 students received timely intervention who would otherwise have withdrawn from the university. For each of those students, this represents not just a retained tuition dollar but a completed degree, an improved lifetime earnings trajectory, and a fulfilled commitment that the university made when it admitted them.
$28M Annual Revenue Impact
Beyond the human impact, the financial return is substantial and sustainable. The platform's annual operating cost of $1.6M generates an estimated $28M in retained tuition revenue—a 17.5x return that has made student success technology the highest-ROI investment in the university's IT portfolio. The Board of Trustees approved expansion to all 8 affiliated campuses based on this outcome.
Equity Gaps Narrowed
The graduation rate gap between first-generation and continuing-generation students narrowed from 19 points to 8 points. The gap between Pell Grant recipients and non-recipients narrowed from 22 points to 11 points. AI-driven proactive outreach proved most valuable for the students least likely to self-refer to advising—systematically counteracting the access inequities embedded in a purely reactive model.
Quick Project Info
Industry
Education
Services
AI Analytics, Student Success, SIS Integration
Duration
18 months
Client Overview
About the Client
A regional public university founded in 1939, serving 22,000 undergraduate and graduate students across 8 colleges and 140 degree programs. Classified as a Doctoral/Professional University by the Carnegie Classification, with 55% of students receiving Pell Grants and 38% identifying as first-generation college students.
Initial Situation
47% 6-year graduation rate (14 points below peer benchmark), 490:1 advisor-to-student ratio, student risk data fragmented across 8 systems, and a reactive advising model reaching students only after they had already decided to withdraw.
Start Your Transformation
Ready to achieve similar results? Let's discuss how we can help transform your business.
Schedule Consultation Explore Services