The UDP Year in Review: Scale, Specificity, and Relativity

2025 was another landmark year for the Unizin Data Platform (UDP), which continued its evolutionary journey from foundational data infrastructure for Unizin members to a powerful accelerator for research, insights and application development. While each member has their own instance of the UDP, the common data model upon which it is built allows members to share insights, transfer knowledge, and adopt, apply, and adapt analytics tools developed across the consortium. This collaborative approach maximizes resources and enables everyone to “go far together” instead of going it alone. 

Insight at Scale 

In 2025, the sheer volume of data flowing through the UDP reached unprecedented levels. Unizin now manages 21 UDP tenants across its membership, including primary campuses, online learning organizations, and extension units.  Last year, Unizin members ran more than 122 million queries against these data stores to power insights around  student engagement analytics and enrollment management

During the year, Unizin’s UDP infrastructure processed more than 16.2 billion clickstream events from 24 different sources — an increase of 2.5 billion events over 2024. That’s over 500 clicks per second, on average, moving through Unizin’s infrastructure. The ability to capture behavioral data correlated with real-time student learning activities is accelerating the development of high-impact analytical tools designed to support student success. 

Specificity Beyond Risk Profiles

Across the consortium, the notion of humanizing learning data emerged as a common theme last year. Rather than using data to simply flag a student who may be struggling, members focused on ways to provide advisors, faculty and students themselves with helpful context concerning the specific factors that can contribute to student performance. 

The impact of this approach is evident in the Canvas Activity Score, developed and studied at Indiana University (IU). IU found that a simple three-variable model—assignments submitted, assignments due, and active minutes — could significantly predict student outcomes. The IU team translated data into contextual clues for student advisors to help inform their outreach to students. The IU team documented both measurable improvements in grades and an 80% increase in the odds of a student persisting to the next semester when advisors leveraged these insights. 

To help other members adopt and adapt the IU approach to their own campuses, Unizin packaged the logic from IU’s tool into the Student Activity Score data mart, making this representation of student activity available to every Unizin member, updated nightly.  In 2025 alone, 11 Unizin members used this data mart, accelerating the development of similar tools on other Unizin campuses. 

Inspired by the Canvas Activity Score model, the University of Iowa modified the Student Activity Score data mart to pilot its own advising tool. The Iowa version incorporates weighted gradebook data and allows advisors to identify students in the bottom 15% of engagement with a single click, enabling early outreach before students fall too far behind.

Bringing Relativity to Data

In 2025, Unizin members also focused on overcoming a significant and persistent technical hurdle: interpreting engagement data. Simple metrics, such as files viewed, can vary widely across disciplines and courses. To account for these variances, the UDP can now anchor performance data relative to other students in the same course section. 

This "relative performance" approach provides insight into individual performance within the context of the larger student cohort: to see if a student is in the bottom 20th percentile of their specific class, for example.  This context is crucial. Members of Unizin collaborative communities have noted that many metrics can be meaningless unless they are compared with peers in the same instructional environment. 

MyLearningAnalytics (MyLA) developed at the University of Michigan, delivers relative metrics to students. MyLA enables students to visualize their use of course materials and their grade performance in relation to other students in their class. MyLA dashboards can help students course-correct by letting them see how high-performing peers in their class engage with and use course materials and other resources. This can lead to better resource utilization and improved student performance. With Unizin’s ability to host MyLA for all members, 45,000 students across seven additional Unizin members have engaged with MyLA as part of their coursework.

Data Mart Expansion

Sharing knowledge and resources continues to drive the expansion of Unizin data marts and shorten the distance between inspiration and application. Data marts not only reduce the burden on members to develop complex code to query the UDP, they can be combined in interesting ways to do new things.  

Penn State’s Performance Outlook is an advisor-facing platform that provides a weekly assessment of students at risk of falling below a ‘C’ grade. Nine of the top 15 most predictive features in Performance Outlook are derived directly from Unizin data marts. One cornerstone metric of Performance Outlook is the "20-minute session", a Unizin-developed concept that clusters student clicks to approximate meaningful engagement time. As a relative performance metric, the 20-minute session allows advisors to see how a student is engaging with coursework as compared to the whole, while adhering to Penn State privacy policies that prevent the sharing of real-time grades. 

The University of California, Irvine, combined Penn State’s clickstream metrics and the data marts inspired by IU’s Canvas Activity Score to help power its Spark platform. UCI Spark is a student-facing tool designed to improve self-regulated learning and time management by providing personalized, curated suggestions (Sparks) that inspire action and student engagement. 

UCI Spark is just one of the latest examples of how the UDP, its common data model, and Unizin data marts continue to reduce the burden of building complex data integrations, pipelines and SQL code from scratch.  These tools not only save data scientists weeks or even months of development time, they also accelerate and advance the application of learning data to support student success.