Education analytics project helps Marist, students make the grade

Analytics paints picture of at-risk students

The process works by aggregating basic student data, such as GPAs, SAT scores, where students are from and other demographic data, and then combining that information with course-specific data, such as how often students submit assignments and engage with instructors through online forums. By analyzing all that information through predictive modeling and data mining, Marist can build an accurate picture of who is likely to drop a particular class, Thirsk said.

But the work doesn't stop there. Thirsk said that knowing how to help students who are likely to drop classes is equally as important as identifying them in the first place. When the education analytics initiative began as a pilot project in 2013, his team would notify students identified as likely to drop a class that they were falling behind. The analysts thought doing so would give the students a nudge in the right direction -- but ultimately, more students went ahead and dropped classes after getting notifications than the analytics team would have expected if it had done nothing.

Now the team takes a more subtle approach. Initially, the identity of students who are found to be falling behind is kept confidential. The students aren't notified of their status, and instructors aren't told who in their classes is seen as likely to drop out. Instead, they're notified that they have students who are falling behind and that they should do more to engage the class as a whole. If that doesn't work, instructors then can unlock the identities of at-risk students to do more personalized interventions.

More than just statistics

Marist's experience shows that this kind of initiative is much more than just a stats problem. Thirsk said he brought in experts from throughout the college, including the education department, to make sure that it gets the follow-up right, not just the initial analysis. "The intervention and what we do to process the intervention is just as important as figuring out who's in trouble," he said.

So far the updated approach is working. Thirsk said that when Marist started the initiative, about 60% of students identified as at-risk dropped classes. Today, that number is down to 40%.

Now that the effectiveness of the program has been proven internally, Marist is also focused on encouraging other schools to follow its lead. The analytics team is starting to promote its methods through webinars, and since the entire project is based on open source technology, other colleges and universities should be able to implement similar education analytics approaches.

In addition to putting the required analytics tools and capabilities in place, Thirsk said it's important to tap into existing knowledge within a campus, whether that means statistics professors or education instructors. Or both: He sees a cross-disciplinary team as a key to success. Other schools "can do this if they have the skill," he said.

Ed Burns is site editor of SearchBusinessAnalytics. Email him at eburns@techtarget.com and follow him on Twitter: @EdBurnsTT.