Part II: How Predictive Analytics is Transforming Higher Education Retention
Welcome to the second post in a two-part series that will explore the benefits of using predictive models on enrollment and retention, and provide specific examples of how universities are accomplishing this work. This post is focused on retention. (Read Part I here to learn more about enrollment.) We’ll wrap up today’s post with tips on selecting an analytics software platform that can help you take the next step in your own data-driven journey.
Applying Predictive Analytics to Retention
Retention rates measure the percentage of first-time, full-time undergraduate students who return to the same institution the following fall. According to the most recent findings released by U.S. News & World Report in 2017, the average freshman retention rate for students entering in the fall between 2012 and 2015 was 78 percent—with the lowest rate reported at 41 percent. Boosting retention rates is critical to an institution’s financial health, the satisfaction of its current students, and the long-term perception of its value.
Like enrollment, retention rates can be improved by employing predictive analytics. Building a predictive model enables an institution to predict the probability of attrition, identify trends, and focus its efforts on the right students to increase retention rates. Furthermore, because the model requires retention data from a wide variety of campus sources, the outcome also provides a more complete, holistic view of each student.
Retention: Predictive Models at Work
Let’s look at two universities who have applied predictive modeling to improve their retention rates and student success programs.
Faced with a declining retention rate, Ball State University used predictive analytics to take action. They began by gathering student data from a vast number of sources, including academic preparedness from admissions, card swipe data from dining, other student life engagement data, recreation services, financial aid, and survey feedback. By implementing a best-in-class data preparation and analytics platform from Rapid Insight, they were able to make sense of the data and understand correlations. The question they were trying to answer: “Which students are most at risk to leave the university?” Using the model, they were able to go beyond their standard assumptions and integrate an additional 250+ variables in their analysis. The model produced a list of 400 students who had a 50% chance of not returning to Ball State. In the words of Dr. Kay Bales, the Vice President of Student Affairs and Dean of Students at Ball State, “No student outreach is a waste of time. But with the Rapid Insight tools, our team now has more confidence that we are accurately pinpointing the students who will benefit most.”
Dr. Bales team has been able to quickly test their hunches, such as finding a correlation between likelihood to persist and students who eat breakfast in the morning. Ball State has continued to refine their models with new information. Their efforts into becoming a more data-driven campus have paid off in big ways. In addition to increasing the retention rate by 6 points, they can now test theories before they begin new retention programs and be better stewards of their limited resources.
Let’s look at the University of Pittsburgh for another example.
In addition to using Rapid Insight for student deposit predictions (read Part I to learn more), Pitt applied the platform to predict student success rates and make more prudent financial aid judgements. The university needed to determine the extent to which student outreach and support was needed to ensure retention, and how to better forecast institutional spend.
Pitt began by identifying students from the past two years who performed lower than predicted. Michael Seidel, a Senior Data Analyst in the Office of Admissions and Financial Aid, leveraged software from Rapid Insight to sort the data into deciles, or bins of ten. By integrating the data with information from The College Board, he was able to detect the top two bins of students who looked most like the students who were identified as having underperformed during their first two or four semesters.
It was discovered that 406 students were 29% more likely to not succeed. This prediction was based on their performance before college and within the first two years at the university. This provided the Chief Enrollment Officer with a data-driven program to address at-risk freshman.
Within this data-driven management system, Seidel can pinpoint recurring trends and traits. For instance, he identified specific schools within the university that have a lower success factor. Results can often be ostensibly intuitive, but sometimes yield interesting and unforeseen surprises. For example, students with the highest probability of not succeeding are those with SAT scores between 1250-1290 and 1350-1390.“This is going to be really valuable moving forward. When you’re making admissions decisions, you want to look at this chart and ask, ‘Are they falling into one or more of these buckets?’ That might be a red flag,” Seidel said.
The University of Pittsburgh is also using predictive analytics to make more informed financial aid judgments. Seidel hopes to use predictive modeling to accurately and cost-effectively distribute financial aid funds. One of the most asked questions is: “How much institutional money are we going to spend if these students enroll at Pitt?” Rapid Insight’s data-driven management tools can answer that with Pitt’s existing data.
Using a Self-Service Platform for Data Prep and Predictive Analytics
The examples I shared with you today are all Rapid Insight customers who are leveraging the Veera analytics platform to boost retention rates. Regardless of the platform you choose, it’s critical that you own the decisions and knowledge that went into the predictive model. Look for a platform that provides complete transparency into the data preparation and predictive process. If you’re using the results to make key decisions within your institution, the results must be understandable and your work defendable.
Furthermore, look for a platform that enables you to change and update your models as needed without having to pay heavy fees. If you aren’t confident in your predictive modeling skillset, look for a vendor-partner who can teach you best practices and techniques. At the end of the day, you’ll need to be confident that you can explain the model scores to your team and how you got from raw data to the decisions you’ve made.
By democratizing data access across the institution and building a data-driven campus culture, universities like yours are empowered to make more effective decisions.
Start by determining how much tuition-revenue a simulated increase or decrease in your freshman retention rate can yield with our Retention Savings Calculator. Often, those who work in the field of student success or retention are hard-pressed to quantify their efforts. We’re making it easy for you!