Enrollment Forecasts are Incomplete

“Vibes” do not an enrollment goal make.

Yet far too often, when looking to validate our enrollment forecasts, we find ourselves relying on anecdotes and vibes reported to us by admissions counselors.

It’s a problematic approach that has led many of us on the wrong path at least once in our careers.

As admissions directors, we are tasked with not just guiding the team but predicting the future enrollment decisions of incoming students.  Of course, we’ve learned to look at the indicators of enrollment: applications, FAFSAs, visits, admits, and deposits.  By measuring historic trends on those numbers, we can get reasonably close. 

However we get there; an important step is to bounce the raw data against the admissions counselors’ experience interacting with the class.  It’s a necessary step but one that is fraught with potential inaccuracies.

Here’s the problem: we’re bouncing the forecasts off of the admissions counselors too early.

The metrics above – that’s what we need to manage up.  They should control expectations, be early indicators of a successful or struggling plan, and help align resources appropriately.

Those metrics are not how we should be managing down.  We are asking admissions counselors to respond to the wrong data set.

Admissions counselors can vibe-check the outputs of a forecasting model that is measuring and monitoring engagement throughout the cycle.  Imagine if in addition to the metrics we measure for the President and the Board, we presented counselors with hard data on year to year email clicks in their territory and projected yield rates from that?  How would the conversation be enhanced if website engagement was up or down 18% in a counselor’s territory year over year, and we used that as the basis for a territory forecast?

Admissions directors can improve their forecasts and minimize the data disadvantage by building a monitoring tool that is meant to measure student engagement, and then, and only then, ask the counselors to provide qualitative data on top of it.

I can tell you this – I’ve watched the data science team at enroll ml spin through hundreds and thousands of variables, evaluating tens of millions of rows of data.

“Vibes” has yet to prove to be a good predictor of enrollment.