1 of 26

2 of 26

Building Spark: A Campus-Agnostic Tool Supporting Student Success

Justin Camacho

Applications Programmer

Student Success Technologies

Office of Information Technology

Mike Walton

Manager

Applications Programmer

Student Success Technologies

Office of Information Technology

3 of 26

About UCI

Extremely diverse

#3 for Top Colleges in the West for Diversity

#4 in 30 best colleges for students with learning disabilities

�New Carnegie designation: Higher Access, Higher Earnings (1 of 3 in AAU)

HSI & AANAPISI

Hispanic and Asian American and Native American Pacific Islander Serving Institution

https://uci.edu/university-facts/

4 of 26

5 of 26

What is Spark?

Spark is a personalized, data-driven dashboard built on the Unizin Data Platform that links students with targeted resources to support well-being and success.

Learn more: http://tla.ovptl.uci.edu/spark

6 of 26

What is a spark?

7 of 26

Product Walkthrough

8 of 26

Dashboard

Application entry point

Tour on first visit

Displays up to 4 sparks (targeted actionable resource)
Refreshed weekly
Easy access to class-level analytics features
Mobile-friendly layout

9 of 26

Course Analytics

Visualizes course activity, assignment submission trends, and heatmap of activity over time
Goal is to increase awareness around LMS usage and activity
Displays related sparks

10 of 26

Spark Library

Browsable index of all sparks in the tool
Supports category filtering, favoriting, & history
As resource library expands, additional filter options may be added

11 of 26

Questionnaires

Supports variety of question types
Can differ between terms
Opportunity for goal setting, answering research questions, etc.
Results available in aggregate to researchers

12 of 26

Technical Details

13 of 26

Unizin

Unizin: member-based (13 R1 public universities) nonprofit that delivers solutions addressing the pressing and complex challenges of data, analytics, and digital content
Unizin Common Data Model (UCDM): integrates and normalizes data from the SIS, LMS, and learning tools
Unizin Data Platform (UDP): implementation of UCDM - data lake, warehouses, and marts that support use-case domains
Joined in 2022 / implemented in early 2023 / went live Sept. 2023

unizin.org

Questions? unizin@uci.edu

Unizin is a member-based nonprofit that delivers solutions addressing teh pressing and complex challenges of data, analytics, and digital content.

Of these solutions is the Unizin Common Data Model, or UCDM. The UCDM integrated and normalizes data from the SIS, LMS, and learning tools.

Unizin Data Platform (UDP) is the implementation of the UCDM - which consists of data lakes, warehouses, and marts that support use-case domains

In our case, we joined in 2022, implemented in early 2023 and went live in Sept. 2023. Not in Spark; it wasn’t launched yet. We have various use cases for the UDP, including another of our web applications and a dashboard for LMS admins. The reason we had a long implementation time was because of our homegrown SIS and our concurrent implementation of our Student Data Warehouse, but overall it’s saved us ~70K.

We are one of 13 R1 public universities who are members. Other members include NYU, Penn State, Indiana University. They also offer other affordable eTech services, not just the UDP, and you also get access to various committees they organize to foster collaboration and innovation between member campuses.

If you’re interested in learning more about Unizin, I’d encourage you to go to their website. It’s been really beneficial for us, which you’ll see specifically in Spark.

14 of 26

How it works

Relies on two key systems:

UCI Student Data Warehouse
UCI Canvas

Data flows into Unizin Data Platform (UDP)
Data further refined into data marts used by Spark
Available standalone or via Canvas LTI integration

15 of 26

Tech Stack

Development Frameworks:

Microsoft .Net Core, Microsoft Blazor

Databases:

Microsoft SQL Server hosted in Amazon RDS
Unizin Data Platform via Google BigQuery hosted in Google Cloud

Infrastructure:

Application hosting: Amazon AWS EC2
Database hosting: Amazon AWS RDS

Design Frameworks:

Tailwind CSS, DaisyUI

16 of 26

Modular Architecture

Designed for campus portability
Flexible for campus adoption via custom Authentication Services (ships with Shibboleth)
Exclusively leverages UDP for data to ease campus portability

17 of 26

Sharing with other campuses

Choose your Canvas Data source

Join Unizin & get access to UDP
Integrate with CD2

Authentication service flexibility

Ships with Shibboleth

Logging Service
Upload your spark library

vs.

18 of 26

Spark Upload

Upload custom spark fields - name, description, URL, etc.
Current: upload spreadsheet to database & run SQL script
Future: integrate into UI

19 of 26

Spark Criteria

20 of 26

Spark Criteria

21 of 26

Spark Criteria

5 criteria types

Field matches value range
Field matches exact value
Field matches value in set
Time (week #/term)
Performance/activity score range

Upload criteria spreadsheet to database & run SQL script

22 of 26

Learning Tools Interoperability (LTI)

Technical standard used to connect learning tools with an institution's learning environment without logging into each tool
Challenges

Browsers blocking third-party cookies
Performance
Duplicate code

Another piece is the LTI, Learning Tools Interoperability. It’s a technical standard used to connect learning tools with an institution’s learning environment (Canvas) without having to log into each tool. Think of it like an iFrame. For us, this is a button that opens up Spark within Canvas.

It may look simple enough but there were actually a lot of challenges to get this to work.

The LTI requires you to communicate back and forth from Canvas to our authentication endpoint. A number of languages and framework have a code library to do this for you. However, there was no such library for .NET, so there were a lot of security checks which we had to write mostly from scratch. Separately, browsers have also begun blocking third-party cookies, which are necessary to keep a user authenticated within the LTI. It also presented an issue with the Canvas masquerading feature. In Canvas you can masquerade as a different user to see what they are seeing. If you were masquerading in Canvas, the Spark LTI needed to know so that it could also masquerade you. In order do this, we needed third-party cookies. Without them, we had to follow additional security protocols to work around it.

All the back and forth negatively impacted performance. The page takes longer than we’d like to load.

We also ended up with some duplicate code. For instance, we have two dashboard endpoints, one for the LTI and one for the non-LTI. We consolidated code where we could, but we do still have two separate endpoints essentially doing very similar things, which is not ideal.

Overall however, we did see our engagement spike significantly after we launched the LTI, so it ended up being worth it.

23 of 26

Overall Challenges

Determining object IDs

Courses

Course space vs. SIS course
Different section types
Cross-listed courses

Users

Campus ID vs UDP Person ID vs Username

Integrating Blazor mid-development

We encountered a number of challenges along the way. We first has to determine the object IDs we were going to use. This may seem straightforward, but we allow instructors a lot of freedom in how they create their course spaces, which made things tricky for us. For UCI, Our canvas course and section structure doesn’t align with SIS data. For example, instructors can choose to combine all the LEC and DIS sections into one course space, or separate them all. They can combine cross listed courses into one space, or they can choose to not. With spark, we ultimately decided to map to SIS courses, so all sections appear separately. This is another advantage of the UDP since it normalizes SIS and Canvas Data, though this was still a decision we had to make. We went through a similar process deciding User IDs.

Another challenge was integrating Blazor mid-development. We built the tool on .NetCore. Mid-way through we wanted to experiment with front-end frameworks, so rather than rebuilding the tool on Blazor from the ground up, we injected Blazor in various places. This proved to cause some headaches because the .NetCore code didn’t always work seamlessly with the Blazor code. It would have been better if the whole tool was either one or the other.

24 of 26

Future

Comparative course analytics
Study planning tool
Performance enhancements
Incorporating non-LMS data

Career Center, campus orgs, etc.

Share with other campuses

Goal is to be easily portable for anyone using the UDP

25 of 26

Questions?

https://unizin.org/

https://tla.ovptl.uci.edu/spark/

Student Success Technologies

Office of Information Technology

Justin Camacho

jdcamac1@uci.edu

Applications Programmer

Mike Walton

mjwalton@uci.edu

Manager, Applications Programmer

Applications Programmer

Student Success Technologies

Office of Information Technology

26 of 26

Thank you!