Review and feedback on Sankey diagrams and other visualizations and their use cases

Sankey Diagram Review & Feedback

Overview of Sankey Visualization

Jack presented a Sankey diagram built in Power BI to visualize individual student course-taking pathways across semesters and colleges.

Key Features

Displays:

• Semester → College → Course flow

• Student-level chronological course enrollment

Shows:

• Courses taken at the user’s college

• Courses taken at “Other College” (masked if outside district access level)

Supports:

• Filtering by student success momentum

• Cohort aggregation (multiple students selectable)

• Direct Query to Redshift for large fact tables (billions of rows)

Loads quickly via structured fact table (optimized distribution & sort keys)

Student Success Momentum Metric

• Value between 0 and 1

• Based on:

  • Recency of course-taking

  • Volume of courses

• Higher score = stronger momentum

• Used as filter to identify students potentially needing support

Example use case:

• Filter: Momentum < 0.85 and > 0.32

• Identify students who may have slowed enrollment activity.

Discussion emphasized more positive terminology than “intervention.”

Suggested terminology:

• “Engagement”

• Integration with Guided Pathways / intrusive advising frameworks

Key Use Cases Discussed

1. Student-Level Advising

• Counselor reviewing individual student progression

• Identify swirl (multi-college attendance)

• Spot off-ramping in structured programs (e.g., nursing)

2. Cohort-Level Analysis

• Analyze structured cohorts

• Evaluate pathway adherence

3. Auto-Award / Degree Audit Enhancement

• Integration with external college coursework

• Request to expose CB00 course identifiers

• Potential to:

  • Improve degree audit accuracy

  • Increase proactive transcript requests

  • Support auto-application for degree completion

Reported outcome:

• Colleges have doubled/tripled degree completion rates using proactive outreach.

4. Swirl Tracking

• District-level users can see named colleges within district

• Chancellor’s Office sees all colleges

• College-level users see “Other College” for outside institutions

Requested Enhancements

Visualization Feedback

Concern:

Sankey diagrams can be difficult to explain to non-technical audiences.

Suggested Alternatives:

• Aggregated bar charts

• Cohort-level summaries

• Traditional dashboard views layered over fact table

• Structured program sequence views

Conclusion:
Sankey remains useful for individual-level drilldown; alternative visuals recommended for executive or aggregated reporting.

Exploration of AI Applications to Enhance Data Warehouse Development and Utilization (continued)

AI Applications in the Data Warehouse (Continued Discussion)

AI Opportunity Areas Identified

1. AI-Assisted Data Navigation

Jack demonstrated an AI-powered MIS documentation assistant using:

• Large Language Model (LLM)

• Retrieval-Augmented Generation (RAG)

• Vector store of MIS documentation

Capabilities:

• Answer natural language questions

• Generate sample SQL queries

• Explain data quality and collection changes

• Clarify table locations

Example Use Cases:

• “Where is financial aid data stored?”

• “Where can I find demographic data for research?”

Guardrails:

• Limited to public documentation

• No direct student data access

• Queries provided for researcher validation

2. AI for Data Governance

Raised by Craig:

Potential applications:

• Automating repetitive governance tasks

• Metadata management

• Documentation assistance

• Data lineage validation

3. AI for MIS Error Detection & Anomaly Detection

Discussed by Christopher & Loris:

Use cases:

• Detect cohort inconsistencies

• Flag anomalous student status changes

• Identify single errors generating thousands of downstream errors

• Pre-submission MIS validation

Opportunity:

• AI-driven anomaly detection pre-submission

• Auto-feedback loops to departments

• Reduced manual MIS workload

Open question:

• Should validation occur pre- or post-submission?

• Integration with Common Cloud MIS tools?

4. AI in Student Support

Aligned with Vision 2030 priorities:

Potential:

• Force multiplier for student support staff

• Predictive engagement signals

• Enrollment fraud detection

• Natural language data interface

Craig noted:

• Risk tolerance varies across system

• AI governance guidance memo forthcoming

• AI microsite launching to centralize initiatives

5. AI Agents & Security Considerations

Discussion points:

• Not ready for AI agents directly executing warehouse queries

• Risk tolerance currently moderate-to-low

• Importance of guardrails

• Potential use of graph databases for join validation

Concept introduced:
“Vibe coding” — AI generating SQL dynamically (future possibility).