How Predictive Do You Want to Be? Enrollment Analytics and Easy Answers

Imagine sending a letter to a prospective student that says, “We regret to inform that we are unable to consider you as a prospective student. Other students who have graduated from your high school have a poor academic history at our college.” Harsh, huh?

This is not unlike a letter that American Express card member, Kevin Johnson received last year.  “Other customers who have used their card at establishments where you recently shopped have a poor repayment history with American Express,” the letter said.  In an effort to manage default risk, credit card companies are studying shopping habits, which mortgage lender a customer uses and whether the customer owns a home in an area where housing prices are declining.

After contacting the company, Mr. Johnson said, “I understand the need for and the power of predictive analytics … but I think they have crossed the line.”

Today, many in higher education – institutions are vendors-alike – are pushing and pursuing the concept of predictive analytics to optimize enrollment and reduce ‘institutional enrollment default risk’.

The question is how far is your institution willing to go?

Colleges and universities are using simple predictive scoring or modeling techniques to better streamline their student marketing data purchases. What is it really? Look at your previous year’s inquiry pool, determine the characteristics of those that enroll, and then buy names of students that display those characteristics. These models are routinely based on simple geo-demographics.

But even for this level of analytics, many institutions struggle with the data to support these models. Last year, I interviewed a former consultant of a well-known enrollment management services company. He commented that it often took a full-year for their customers to be able to provide a clean, actionable data set to even apply the model.

But, what if data or the technology, for that matter, wasn’t an issue? Would you be willing to go beyond simple geo-demographics and look at patterns of behavior?

When it comes to the issue of student-retention, the easy answer is to put the onus on the front-end of the institution. “If we only recruited and marketed to better students, our retention or student success rates would improve” is a common refrain heard across academia. In other words, if only the admissions department could find students with the perfect DNA to fit our unique institutional culture, retention issues would disappear.

The problem with this easy answer approach is: 1) it requires intensive analytics into academic, geo-demographic, and behavioral-pattern data (see American Express), 2) the political will to act upon it, and 3) it diminishes the impact your college or university has upon students.  Taking the proverbial easy way out, in essence, says that NOTHING transformative happens when a student enrolls on your campus. Rather, the reason that students succeed (or not) is simply a matter of DNA.

UPDATED THOUGHTS ***********************************************************

Certainly, predictive modeling can be effective in focusing limited resources on recruiting students that make sense for your college or university. An advanced model requires an intensive data set of not only geo-demographics, but student behavior as well. Secondly, it requires the political will to act upon it.

Will/should you stop enrolling students that demonstrate behaviors of previous unsuccessful students? Tough question.

The larger problem with the 'retention is a recruitment issue alone' view is that it diminishes what can happen on college and university campuses. In other words, following this view to the limit, it's logical conclusion is that nothing transformative happens when students enroll. Which, of course is non-sense.

When it comes to predictive analytics, how far is your college or university willing to go? I invite your comments.

Comments

[Shawn M. Brown]

Tim,

Good post. My answer to your bottom line question is, "pretty far." With the assistance of one of those well-known enrollment management services companies, we have implemented predictive models for student recruitment at my institution. The measurable results include an increase in size of the incoming classes by 40-50% while at the same time we held steady the academic profile. So far, our retention numbers have been higher, too.

We addressed the first two problems you cited. First, we started with good data. (We continue to identify ways to collect better data.) Second, we had the will to act upon it. (Being under-enrolled with the trend line heading down helped gain support for the use of predictive models.)

I don't fully understand your third problem with enrollment analytics. I agree with your proverbial "it's just DNA" comment as the mission and character of our institution is unchanged. At the same time, as participants of the NSSE survey, we can show that we do transform lives. Our predictive models have simply helped us to identify the students on whom we can have the greatest impact.

Shawn

[Tim Copeland]

Hi Shawn, thanks for your comments and congratulations on your institution's success.

I realize I didn't frame that last point as I intended. The real easy answer approach is to simply put the retention issue back on the front end of the institution. So if the problem is simply a matter of 'DNA', taking that belief to the limit it's logical end is that nothing transformative happens within the institution to change students. Which of course, is not true.

An edit to the blog post is forthcoming!

Tim