Graduate Pathways System
The Graduate Pathways System modernized how Duke University collects and reports graduate career outcomes. By replacing decentralized spreadsheets with a structured survey, custom taxonomy, and Tableau-integrated reporting, the project enabled more accurate, scalable, and actionable data. It supported accreditation, grant funding, and program improvement while reducing administrative burden across the institution.
“If we want better and more accurate data and a higher rate of return, procedures and policies needed to be ironed out… We never want to pick up a spreadsheet again.”
Transforming Career Data into Institutional Strategy
Between 2015 and 2018, I led a full-scale transformation of how Duke University collects, analyzes, and reports graduate career outcomes. What began as a basic inquiry into outdated spreadsheet practices evolved into a system-wide redesign, rooted in stakeholder engagement, system integration, and strategic alignment with accreditation standards.
At the time, the process was entirely manual: individual programs tracked placements in their own formats, submitted Excel files by email, and relied on staff at the Graduate School to reconcile it all. Internal calculations revealed the scope of the problem—transforming and validating this data required the equivalent of 2.5 full-time employees each year. The process was expensive, error-prone, and unsustainable as demand for better outcomes data continued to grow.
The redesign replaced these institutional pain points with a scalable, survey-based system linked to university reporting tools, supported by a custom taxonomy, and eventually integrated into Tableau dashboards. We built new protocols, reconciled legacy data, and delivered program-level reports that improved compliance, informed decisions, and elevated the university’s data culture. By the end of the project, Duke was outperforming peer institutions in response rates and data quality, and was positioned to scale the system into a real-time, self-service portal accessible across the university.
🛠️ Updated coding instructions for departments
To replace a fragmented, labor-intensive process with a system that could scale across an entire university, I focused on five core pillars: survey design, data classification, legacy data reconciliation, stakeholder reporting, and infrastructure planning. Each component was designed not only to solve an immediate problem but to lay the groundwork for long-term institutional use.
At the heart of the system is a structured exit survey, which I designed and piloted for all graduating master’s and doctoral students. The survey asked graduates to share detailed information about their next steps—capturing everything from employment sector and job function to how their position relates to their degree. Survey logic was tailored to academic and non-academic pathways, and the instrument was built with future integration in mind, connecting to systems such as PeopleSoft and DADD. The embedded file shows the full range of structured inputs captured, including contact details, sector classification, research intensity, and degree relevance.
🏗️ The survey data fields
One of the most complex challenges in career outcomes work is agreeing on what counts as a research role, an academic appointment, or a discipline-related career. To bring consistency and structure to our data, I helped design and implement a custom classification framework that mapped open-text survey responses into three standardized dimensions: sector, career type, and job function.
We began by adapting elements of the CNGLS taxonomy and NAICS industry codes, then layered in our own logic based on common respondent patterns and institutional needs. This framework was designed to do more than just sort data—it enabled longitudinal comparisons, departmental benchmarking, and eligibility alignment for federal training grants. I also wrote guidance for coders and developed protocols to ensure accuracy and inter-rater reliability across staff.
Before the new system could move forward, I had to look back. Duke had more than a decade of career outcomes data stored in departmental spreadsheets—each with its own structure, shorthand, and assumptions. These legacy records were inconsistent, sometimes contradictory, and often unverifiable. But they also represented our only historical view of graduate placement.
📊 How data was reported historically
I led the effort to clean, verify, and reclassify this data using our new taxonomy. Every record was reviewed: I compared departmental edits, flagged inflated titles or unverifiable claims, and cross-referenced entries against LinkedIn and internal systems. I color-coded edits for transparency—what the department submitted, what I verified, and what remained uncertain—turning an opaque archive into an auditable, structured dataset.
Once the data was structured and validated, the next step was turning it into something useful—something departments, funders, and university leaders could actually act on. I developed custom reports for more than 90 graduate programs, providing each with targeted career outcomes data. For doctoral programs, we delivered both initial placements and 10-year trends; for master’s programs, we focused on high-quality initial outcomes, many of which had never been collected before. These reports were designed not just for compliance, but for impact. Programs could now use real data to support external reviews, improve advising, and make the case for training grants. At the same time, I began migrating this reporting framework to Tableau, enabling interactive dashboards with program-level filters, trend visualizations, and year-over-year comparisons.
🖼️ How data is reported today (PhD data is here | Master’s data is here)
To fully retire the spreadsheet era, I authored a proposal for a centralized, self-service portal that would allow departments to input, view, and update graduate career data in real time. This portal was designed to sit at the intersection of usability, governance, and data quality—reducing administrative burden while improving trust in the data.
The proposed system included pre-populated survey fields to minimize errors, user-level edit history for version control, and automatic timestamping to track the most recent verified data. I mapped out technical integrations with SISS (for student records), OIT (for authentication and access control), and Tableau (for dashboard visualization), ensuring the portal would work within Duke’s existing infrastructure.
🔧 Project Assets & Documentation
🚡 How to Update Data in REDCap
The collected information is utilized in the following core institutional endeavors: accreditation, the external review process, training grants, strategic planning, cross-institutional partnerships, program evaluation and improvement for the 90+ constituent graduate programs, and refinement of return on investment models. The data, policies and procedures generated from GPS were highly involved in a successful Innovations in Graduate Education grant from the National Science Foundation (#1806593). Yale University, Carnegie Mellon University, and MD Anderson UTHealth Graduate School adapted and implemented elements of GPS to meet their needs.
