Predictive Modeling for Strategic Enrollment Management

One of the many uses for predictive analytics in higher education is to predict what will happen during various stages of the “enrollment funnel” (search names – inquiries – applied – accepted – enrolled). For those of you who might be considering how to incorporate predictive modeling to manage your enrollment funnel, we recently hosted a DataTalk panel discussion in which four experts shared stories and answered audience questions, which is available here.

In the meantime, I’ve collected a few examples of how predictive modeling can impact enrollment at the collegiate level. Here are four ways that colleges are utilizing predictive modeling to improve inquiry and applicant yields, increase enrollment rates, and boost student retention:

Feeding the Funnel – Ranking and prioritizing search names

What if, out of the thousands of search names that come in every year, you knew who was likely to apply and who you might be wasting time on? This is exactly the type of question that predictive modeling can answer. By building a model to predict which search names are most likely to apply or enroll with you, you can use the resulting probability scores to rank all names accordingly.

Similarly, you can also build models to predict which of your inquiries will become applicants. Ranking inquiries also allows you to develop cost-effective travel and communication strategies for reaching out to potential students. Small adjustments like changing communication type based on likelihood of application for each prospect can help you to allocate your resources more efficiently. Such savings can add up to tens of thousands or hundreds of thousands annually, depending on the size of your institution.

Narrowing the scope – Identifying the best prospects

By building a model to see which to see which students have been most likely to apply or enroll at your institution historically, you can get an idea of which characteristics correspond with eventual application or enrollment. Using these characteristics, you can change the way you approach your search lists. Say, for example, you find that a particular SAT band or geographic area has come into your model as being predictive; you can use this information to make more targeted requests for lists in the future, allowing you to buy fewer names and have a better application rate for the names you do buy.

Additionally, you can use these models to evaluate search lists themselves. You might find that some lists perform better than others, in which case you can adjust your future list purchases accordingly.

At the output – Estimating class size and financial aid outlays

Once applications come in, you can build a model to predict how likely each accepted applicant is to enroll. Then you can sort on these probabilities to focus on those students most likely to enroll with you. As an added bonus, by summing all these individual enrollment probabilities together, you’ll get an estimate of your incoming class size!

Similarly, predicting financial aid outlay is done by multiplying each enrollment probability by the amount of financial aid offered. By summing this new column, you’ll get the total expected financial aid expenditure.  You can also perform either of the above analyses before you send out acceptance packages in order to shape the class you’d like to see or the scholarships you’d like to give out.By running this type of simulation, you can ensure that your class meets set requirements for

Beyond enrollment – Anticipating student retention

Once you know which students are enrolling with you, you can use historical data to anticipate which of them might be at-risk. Flagging and focusing on at-risk students early in their academic careers – sometimes before they even reach campus – can improve retention rates, benefitting the institution as well as the individual students. The ROI on these types of models, especially when coupled with extra efforts to help retain the designated at-risk students, can reach million-dollar territory.

Building such models also gives you insight into what applicant characteristics might be correlated to at-risk behavior. This allows you to make informed decisions about your prospective students.

*

For more information about predictive modeling for strategic enrollment management, sign up for Tuesday’s panel here.

Leave a Reply

Your email address will not be published. Required fields are marked *