Does Using Web Behavior Data Improve Enrollment Predictions?

Up until now, Capture Higher Ed did not use any of its marketing automation data from Capture Behavioral Engagement (CBE) in our Envision enrollment models. This is due to several statistical reasons, but the biggest is because our CBE data is only a few years old and we prefer to take advantage of three years of historical data to maximize the accuracy of our individual and group-level enrollment predictions.

This is different from our applicant models, where we can make predictions on fewer than three years of data, and where we already use CBE data for our predictions.

However, we know that admitted students who have shown interest in a school’s website are more likely to enroll. That’s the central thesis behind Capture’s Behavioral Intelligence platform. So, including our marketing automation data, in some form, should improve our enrollment predictions.

So, the big question.

Would including CBE data improve the accuracy of our enrollment models?

Now that CBE has been in operation for three full cycles, we have some schools with three years of data … so we ran tests to see if including CBE data would have improved our predictions. For all the schools, the individual-level predictions were more accurate compared to a model without CBE data (as measured by Area Under the Curve and Root Mean Squared Error). Also, in all cases, the yield prediction was more accurate.

Additionally, in all cases, there was a closer alignment between CAI scores and Envision predictions — students with higher Capture Affinity Index scores had higher predictions than those with lower scores. Now counselors don’t have to choose between targeting those students with high CAI scores or those with high Envision scores — in most cases, just target those with high Envision scores.

The takeaway: Using web behavior data on admitted students can improve our enrollment predictions both at the individual and group levels, for schools with three years of CBE data. And Capture’s Envision enrollment models are now the only predictive models in higher education to use this sort of behavioral data.

By John Foster, Senior Data Analyst, Capture Higher Ed