You are here

Sponsored Content

Predictive Analytics Strategies That Promote Student Success

Using data visualizations to identify struggling students and help them succeed
University Business, August 2016

Colleges and universities are under intense pressure to meet enrollment goals, improve retention rates, and shorten time to completion. Predictive analytics can play a crucial role in these efforts by providing insights that guide strategic decision making, improve enrollment management and promote student success.

In this web seminar, an associate vice president from Longwood University in Virginia discussed how the institution is using predictive analytics to help identify at-risk students, and seeing much greater impact than traditional student retention models, which often focus on struggling students at a point that is too late to intervene. By taking a different approach with their predictive models, Longwood has been able to put programs in place to help the university’s at-risk students succeed.

Senior Data Analyst
Rapid Insight Inc.

Rapid Insight was founded over a decade ago with a focus on making it easier for anyone to use data to make better decisions. We have customers in many different markets. In higher education, our customers include four-year institutions, as well as graduate schools and law schools. Our software is being used by over 200 institutions nationwide.

Our customers are using the software in a way that covers the entire student relationship lifecycle, from recruiting and enrollment to helping identify at-risk students and ensuring they achieve success. Then, later on after they have graduated and are out in the world working, it helps identify their potential for becoming a donor.

Our customers use our software to create analyses and reports, to build predictive models, and to create and distribute these dashboards and reports. And they do this without needing to be a database expert or having a degree in statistics.
Our tools help to integrate your data from any source, do the necessary cleanup, automate mining and predictive modeling, and then output it in any format needed.

When we discuss predictive modeling, we’re talking about looking at your past records, identifying some key metrics and significant indicators of success at your institution—particularly looking at whether or not a student was retained to the following year—and building upon all the information we have about the students.

You do this with multiple years of data—as many years as you think are relevant for your institution—and then feed that through our tools, through automated predictive modeling, to output a probability or a percent likelihood of retention.

When you have a class of new students where actual retention is unknown at the beginning of a semester, we can now apply this automated predictive modeling to calculate a student’s probability of retention. We have an individual probability score for each student within our database. And we can use that information to make data-driven decisions.

Of course, retention isn’t the goal of predictive modeling for every institution. When you get started with a predictive modeling effort, you’ll need to know what you are hoping to model, and the terms of success are influenced by your institution’s specific challenges and goals. One institution’s priorities might be slightly different than the next. And that was exactly the situation at Longwood University, where using more traditional or more common techniques for modeling success wasn’t working.

Associate Vice President for Enrollment Management
& Student Success
Longwood University (Va.)

Our retention rate was stuck in the mid- to high-70s for about six years. When I arrived at Longwood my position was brand new. We also had a new president join our school shortly after, and all of our leaders knew that we needed to take a much more strategic approach to retention.

That approach is focused on two things. One is that it’s the right thing to do for students—it’s the morally right thing to do. And it’s also a significant lever for funding and for revenue sources, which enables us to do more for our students.
When I arrived, the data analysis was not nearly as robust as it needed to be. We did not even have a lot of descriptive statistics around the details behind retention and graduation. And certainly we were not using predictive modeling.

We began looking at doing some models that, of course, predict the probability for a student to be retained. But from my perspective we were missing a critical predictor. We were taking information from a high school student and saying, “Are you going to stay for the entire year and come back the next fall?” Yet one particular piece of their performance—i.e., their first semester in college—is a much stronger, longer-term predictor.

So I wanted to be able to evaluate how we can use and be responsive to the new information we learn about this student, versus just being reliant on that initial information that we received during the application process. I also wanted to know how we could take that information and move it backwards to admissions, to inform our decisions about enrolling more students who we can then retain.

I had been approached by Rapid Insight several times, and I had just said, “Well, I have a Ph.D. I can do a regression model. I’ve got software that can do that for me.” But then I started exploring doing it myself, and then I became very interested in Rapid Insight, because creating the data sets to use for modeling in any software can be time-intensive. That opened my eyes to the solution they had for analysis, and I realized, “Why would I want to use any other statistical modeling tool even if I could? This just makes it that much easier.” I have much less legwork. And I can also take the models Rapid Insight develops and apply them much more quickly than I could with any other software.

That is how we came to be using Rapid Insight, because we needed to make some inroads fairly quickly. I was able to, in a matter of a few weeks, build a model to predict how a student would do in their first semester with their actual GPA, and then also predict whether they would be in good or bad academic standing.

It also allowed me to provide some key reporting and metrics for my vice president and the president. We were able to predict what our actual retention would be, and which students we needed to do outreach with over the summer, because the reality is not all students depart as a result of academics. So it’s been able to help us strategize how we interface with our students over the summer to continue their enrollment in the fall semester.

In the last four years we have moved our retention rate to an average of 80.5 percent, but have been able to reach as high as 83 percent. We are also very proud of some significant improvements we have been able to make in our graduation rate. Our four-year rate has increased 5 percent, our six-year rate has increased 6 percent, and we are projecting that we will increase both of those percentages slightly at the end of this year when we submit all of our official numbers with our State Council of Higher Education. We’re very excited to see that needle continuing to move in the right direction.

To watch this web seminar in its entirety, visit 

Register now for UBTech 2018

Register now for UBTech 2018, June 4-6 at the Mirage, Las Vegas. At UBTech 2018, you’ll network with a dynamic community of higher ed leaders who are shaping the future of campus technology and explore topics like cyber security, distance learning, campus learning space design, communications, personalized learning and more. Your UBTech registration also includes a free pass to the InfoComm exhibit hall.

Register now>>