SLU Compliance Dashboard
Led a 5-member team to build a production-ready international student compliance dashboard for Saint Louis University, connecting 34,341 raw applicant records with 3,524 government SEVIS records across a PostgreSQL-to-Lo
Executive Summary
Led a 5-member team to build a production-ready international student compliance dashboard for Saint Louis University, connecting 34,341 raw applicant records with 3,524 government SEVIS records across a PostgreSQL-to-Looker-Studio pipeline. Engineered a bridge table to enable cross-system analysis when direct database joins were structurally impossible converting a data availability constraint into a validated integration. Delivered 14/14 SQL reconciliation with zero-tolerance validation, five strategic recommendation clusters, and an R1-R11 tiered risk monitoring framework for SL
2The Problem
SLU needed a unified compliance monitoring system for its F-1 international student pipeline. Three prerequisites deposit payment, I-20 issuance, and I-901 SEVIS fee determined enrollment eligibility and legal entry to the United States. Non-compliance risked institutional SEVIS certification, student visa denials, and revenue loss. SLU lacked visibility connecting applicant records with government compliance data.
Three datasets with no direct join capability: Connect (34,341 rows, 80% duplicate), SEVIS (3,524 records, 126 columns), and Applicant (unavailable). Sixteen critical document columns were 100% null in Connect. The SEVIS linkage field (FIN_ID) was 98.2% null. Financial data covered only 52.1% of students. The team had four weeks to build an auditable compliance dashboard despite incomplete and misaligned data.
Operating Constraints
Strict SQL validation requirement: all 14 queries had to match source data exactly. SEVIS data had compliance sensitivity requiring precise handling. Looker Studio was the only approved visualization platform. Fixed project timeline with deliverable gates.
3Objective
Build a live Looker Studio compliance dashboard for St. Louis University analyzing 3,524 SEVIS records, validate all SQL queries against source data (14/14 target), and deliver actionable compliance insights for international student program management.
4Approach & Method
Why This Approach
Validated all 14 SQL queries against source data before building visualizations to ensure dashboard accuracy skipping validation would have risked presenting incorrect compliance data to St. Louis University. Chose Looker Studio over Tableau because SLU was already in the Google ecosystem.
5Evidence & Artifacts
6Outcome & Results
- •Production dashboard deployed for SLU leadership operational use. Key metrics established as compliance baseline: 22.0% deposit rate, 34.8% I-20 issuance, 45.5% I-901 paid. 959 unpaid students identified across three actionable clusters. 61.5 percentage-point deposit-to-I-20 conversion gap quantified as strongest compliance lever. All metrics auditable and reproducible.
- •St. Louis University gained real-time visibility into international student compliance. The 14/14 validated SQL queries established a trusted data foundation for ongoing compliance monitoring.
Key Results
- •Production dashboard deployed for SLU leadership operational use. Key metrics established as compliance baseline: 22.0% deposit rate, 34.8% I-20 issuance, 45.5% I-901 paid. 959 unpaid students identified across three actionable clusters. 61.5 percentage-point deposit-to-I-20 conversion gap quantified as strongest compliance lever. All metrics auditable and reproducible.
Impact
- •St. Louis University gained real-time visibility into international student compliance. The 14/14 validated SQL queries established a trusted data foundation for ongoing compliance monitoring.
7What I Learned
- •The bridge table solution demonstrated that architectural problem-solving not just tool proficiency separates analytics projects that deliver institutional value from those that produce only charts. When direct joins are impossible, engineered workarounds backed by validation can unlock analysis that the original data structure was never designed to support.