- Adam Cebulski, Assistant Vice President and Chief of Staff, Southern Methodist University
- Sara Ousby, Director, Student Affairs Assessment, University of North Texas
Goals
- To discuss trends in big data and the implications for higher ed
- ID strategies for building data warehouses & analyzing data sets
- Share successes and challenges
- Story telling
- Strategies
Landscape of Big Data
3Vs: variety (lots of kinds of data); volume (more info than we know what to do with); velocity (collecting data at a higher rate than ever before).
There are tons of software packages that “do” big data, but buying software is not going to answer your problem! Big data translates into decision making through different processes, and that’s what we’re going to talk about.
Storytelling
Stories are far better at conveying what your data says than just the data itself. NASPA’s analytics study from 2017 identifies the following entry points for big data for predictive analytics: pre-enrollment > academics > motivation & self-efficacy > use of support services > student engagement
Stories are just data with soul. Stories cross the barriers of time, past, present and future, and allow us to experience the similarities between ourselves and through others, real and imagined.
Create a data story
Data + Narrative + Visuals
Case Study: SMU
We have no centralized data system, and we’re a Peoplesoft campus. We centralized OIT and brought on a new CIO from University of Illinois. We have a large Greek population and we experienced 315% increase in AOD offenses in one academic year. We introduced a number of programs and interventions to address this challenge.
- Why the large increase?
- Who is most at risk
- How and when to intervene?
- Campus partners: IR, OIT
- Data identification
We’re a Maxient campus, so we did a lot of ETL (extract, transform and load) processes to make this work from a technical perspective., Maxient offers no APIs.
We built a BEAM model: Business Event Analysis & Modeling
- Customer focused
- Flexible design
- Report neutral
- Prevents rework
- Saves time
Goal was to build a data warehouse to assist with our analysis and reporting. We started in 2017 and plan to launch in the next week with a dashboard as part of phase one. We needed to hire a data architect and data visualizer: these were university hires that “live” in OIT. At $125K each, these are not cheap resources (but they are an excellent investment).
A BEAM table consists of events and then we think about related events, i.e. sports game, finals, etc. that could be related. At the top we consider a range of other items associated with the charge/sanction, i.e. weather, did we win the game, what class level is the student, etc. We even pull in data about the students, such as if their parents are wealthy donors. This allows us to create a “star schema” which creates a comprehensive picture of the issue. Some of the criteria allow us to set a ranking for each of the events, which in turn allows us to prioritize items. One of the data points is which offices are responsible for addressing the issues. We started with 100, but grew to 279 unique variables that could be associated with a particular conduct case.
These variables allow us to build dashboards that rationalize the data for our staff (intervention or otherwise). The vast majority of people in the system were actually recruits. It’s mostly 1st and 2nd years that get caught up in our system. We were able to change policy immediately based on the insights our system provides.
Case Study: University of North Texas
We are 38,000 students in the DFW metroplex. We are minority majority, public tier one institution. 1st year residential requirement. Majority live in Denton County.
Our Questions
- What are the differences in retention for students who are engaged on campus?
- What are the differences in GPA for students who are engaged on campus?
- Campus partners: Data analytics & IR, IT shared serices
- Data Collection
We are going to pull card swipe data into our system soon! We’re going through the data dictionary of card swipes now, primarily using Excel and lots of pivot tables. We’re looking right now at correlation information with respect to retention.
We’ve had a lot of growth in card swipe usage. We have 220,000 card swipes into our student recreation center, and we plan to pull in the Career Center’s info next. There does appear to be a difference in retention of card swipers over non card swipers (81.18% vs. 64.02%).
Telling our story and making decisons
- Focus on completion
- 1st year students are those leaving at the greatest rates
- Most impact on FTIC
- Higher impact on men
Q: Are you planning an ROI analysis?
AC: We quantified every action with a dollar value. Our interventions have already saved over a million dollars so far. We swipe for EVERYTHING (we use CampusLabs).
Q: What does your data cleaning process look like?
AC: it’s awful! And, it’s ongoing. We’ve had to create many transformation tables, and we had a lot of silo’ed data that needed work.
SA: your data dictionary will go a long way in solving this challenge.
Q: are card swipes weighted equally?
SA: yes (for now). But we’re looking at this. Card swiping is now universal across the campus.
AC: we tie our NSSE and use ID Link to tie our data together.
