Enrollment Management: Using Predictive Modeling
While there are no silver bullets in Enrollment Management our Founder has often said that if he could only utilize one tool out of the Enrollment Builders Tool Box it would be predictive modeling. This use of regression analysis is a great way for institutions to work smarter, not harder. This post is an overview of predictive modeling and some strategic applications in contemporary marketing, recruitment, and retention for higher education institutions.
What is predictive modeling in higher education?
Predictive modeling is a statistical technique used to determine the likelihood of a student performing some desired enrollment behavior. To do so, we study the behavior of previous students to identify variables that influenced their enrollment behavior. After we have identified the relevant predictor (independent) variables we construct a statistical model to predict future behavior.
Then, we apply this regression equation to current prospective or enrolled students to determine the likelihood that they will exhibit the desired outcome (dependent variable); that is to apply, enroll, persist, or graduate. Based on the student likelihood to reach the outcome variable, institutions can modify their recruitment or retention activity in order to maximize institutional human or financial resources.
Definitions within Analytics for Enrollment Management
- Logistic Regression – statistical method used to create a predictive model; gives a probabilistic result based on the value of the predictor variables
- Outcome or Dependent Variable – variable that is being predicted. Typical outcome variables include becoming an applicant, an admitted student, an enrolled student, or a retained student
- Predictive Modeling – creation of a statistical model of future behavior
- Predictor or Independent Variables – variables that explain the outcome or desired enrollment behavior
- Significant Predictor Variables – predictor variables that are important in determining the outcome/target
The Most Common Types of Predictive Models in Higher Education
- Search Model - search name to enroll
- Inquiry Model - inquiry to enroll (aka lead scoring model when used in performance-based marketing)
- Applicant Model - application to enroll
- Yield Model - admit to enroll
- Econometric Model - financial aid modeling, usually for admit to enroll
- Retention Model - enroll to first-year retention
3 Steps to Build a Predictive Model
- Build – Take 80% of the available historical records and create a predictive model where the outcome is known
- Validate – Take the remaining 20% of the available historical records and verify that the predictive model is doing a good job of predicting the known outcome
- Score – Once validated, use the predictive model with the current data set to actually predict the outcome
Turning Raw Scores Into Higher Education Enrollment Management Strategy
The output of a predictive model is a number representing the probability of a specific student achieving the desired outcome variable. It is a number between 0 and 1. For example, the probability that Suzy Student will persist one-year is 0.78.
Unfortunately, those raw number isn't very useful unless we know how they rank compared to other students. Consequently, we convert the raw score into a predictive model rank by ranking all of the records from highest to lowest, and by categorizing them into 10 deciles, or "buckets". Students in bucket 1 or more likely to exhibit the target enrollment behavior than students in bucket 10.
When placed on a gains chart, the data looks something like this:
These are data from a Graduate Program Inquiry Model from when our Founder was the Vice President of Enrollment Management at Tiffin University several years ago. Notice that students in Model Rank 1 (Bucket 1) are predicted to apply at a rate of 63.1% and enroll at a rate of 40.1%. Meanwhile students in Bucket 5 are expected to apply at a rate of 7.7% and enroll at a rate of 1.3%.
With knowledge like this institutions are able to modify recruitment and retention activity in order to utilize human and financial resources in a more strategic way. Here are a few sample strategies:
- Negotiate insertion orders with performance-based marketing vendors to only purchase leads of students in the top six buckets by using a lead scoring model.
- Save money on print, postage, and mailing services by only mailing printed material to students in the top four buckets by using an inquiry model. Send a pdf of a brochure to students in buckets 5-8, and don't sent anything to students in buckets 9 and 10.
- Prioritize follow-up telecounseling activity and increase the likelihood of productive calls by only reaching out to applicants in the top five buckets with an application model.
- Implement retention interventions in the first several weeks of enrollment with students who are at risk of attrition with a retention model. Interact with students in buckets 4-7, where human intervention could make the biggest difference.
Ready to Transform Your Enrollment Strategy with Predictive Analytics?
If you're ready to leverage data-driven insights for smarter decision-making, it's time to talk to the experts. Contact Enrollment Builders today to discover how our Higher Education Call Center Services and consulting can seamlessly integrate predictive analytics into your strategy, helping you achieve unparalleled results without spending more for leads that go nowhere. Don't leave your enrollment numbers to chance. Harness the power of predictive analytics with Enrollment Builders and elevate your institution to the next level.
