Predicting College Admission Enrollment

Big Data and College Admissions

Every year, colleges face the challenge of predicting which students will enroll at their institution.  Universities are using predictive modeling to provide an assessment of the probability of each student enrolling.  This allows the schools to use their limited resources efficiently.

Noel-Levitz provides a consulting service to universities that is aimed at improving communication with probable applicants.  They provide the ability to qualify a list of students before they purchase communications for them.  Normally, universities submit the data and the consulting firms have already been paid for the names, regardless if they rank high or not.  This new method allows the schools to provide a pool of applicants and then decide how to spend their money once the rankings come back.

Each student receives a score between 0.0 and 1.0.  For the fall of 2011, 90 percent of enrollments came from students with a .6 or higher.

Below: Fall 2011 results for campuses using predictive modeling to qualify student list purchases

 

This allows schools to purchase communication packages by score.  The author of the post has several recommendations of how to effectively utilize the service.’

1.       Purchase student names with high model scores. This allows you to buy names of those students who are more likely to enroll at your institution, eliminating potential waste by not buying students who will never enroll at your institution.

2.       Load the highest scoring names directly into your inquiry pool. Many of our campus partners will take a sub-set of their purchased list and treat them as inquiries. The value of the predictive modeling score is that you can accurately pinpoint which students are most likely to enroll.

3.       Segment your messaging by model score. Consider putting your search names into different buckets by score range. The highest model score students might receive the most robust communication flow (including written and electronic). Those with lower scores may receive a series of e-mail messages.

The company splits there categories over gates: Inquiry, Applicant, Admit, Net Deposit. The chart below shows the Fall 2011 results for campuses qualifying their inquiry pools with predictive modeling.  85 percent of the net deposits came from half of the inquiry pool.

This is a pretty interesting and tangible way of visualizing an applicant pool.  You don’t have to completely write off the .4 range and below.  Noel-Levitz allows you to still communicate with these low probability students at a substantially cheaper cost.  Below is an example of how schools could potentially use the model to prioritize communication strategies:

Allocating resources for applicant pool communication can seem like a daunting task.  With Big Data, schools are now able to effectively communicate with applicants that are most likely to enroll.  This saves them lots of money that they can uses to expand their high priority applicant pool communications.