We live in the age of big data. Whether it’s the way we’re sold to, how a favourite sports team performs, or the way healthcare is delivered, big data is already influencing great swathes of our personal and public lives.
Universities are no exception, and as is often the case it is US institutions that are leading the way.
What’s less well publicised is what exactly they’re doing, so in our cover story this week we delve into the detail, focusing in particular on the impact that data is having on student retention and success.
On a recent trip to the US, I was talked through a case study of one institution’s work in this area by Gregory Fenves, president of the University of Texas at Austin.
Five years ago, the proportion of its students who graduated within four years stood at 50 per cent.
This prestigious public university, the flagship of the UT system, has a particular set of challenges as a result of its unusual automatic admissions process, which guarantees a place to a percentage of the best-performing students in every high school in the district.
For a school in a privileged area, this may result in a quota of students who are well prepared for study at a world top-50 university; but for others, coming top of the class at high school secures them a place, but does not give them the foundation they need to succeed.
The challenge is how to close what Fenves calls the “persistence gap” between the well prepared and those who stand to gain the most but are also most likely to drop out.
Fenves says that in “five short years”, a combination of learning analytics and a sensitively designed framework for intervention has enabled the university to make “remarkable” improvements.
The four-year graduation rate is now over 60 per cent, with a target of 70 per cent.
As in the examples explored in our feature, UT Austin tracks dozens of “markers” to keep tabs on students’ academic chances. Much of this is sensitive personal data, which is one of the reasons that the university (unlike some others we speak to in our cover story) has established its own data systems internally, rather than partnering with a private sector firm.
As well as flagging up characteristics known to make a student more academically vulnerable, these systems can spot when an isolated low mark in a particular course is a cause for major concern, even if the student is performing well in other parts of the curriculum.
The second part of the university’s approach is how it processes and responds to the data.
Fenves says that this should not “just be about getting students through faster, it’s also about improving the quality of the education”.
Among the most effective interventions at UT Austin is what it calls the University Leadership Network, which uses data analytics to identify 500 first-year students who could be at risk.
“Instead of just identifying those students and saying: ‘We don’t think you’re going to graduate, you’ve got to go into this programme so we can try to keep you in school’, the genius was to have every one of our freshmen go into an interest group of some sort – we have over 400,” Fenves explains. “The students who our analytics show are least likely to graduate go into groups that are tailored to support them. Most are the first in their family to go to college so don’t have the family network [to provide experience and support], most are Pell Grant eligible [a federal government subsidy for those in financial need], most are students of colour.”
Alongside these steps to create support networks and a sense of belonging, the university also ties financial aid to what Fenves calls “work-study” placements, which could be working in the accounting department or some similarly professional internship, ensuring that financial aid is “earned” while gaining experience and the employability skills that will be needed later.
This case study is one example of how big data, and the increasingly sophisticated way it is martialled, can contribute to a step change in tackling issues that universities have been chipping away at for years. But it also demonstrates that while the data helps diagnose the symptoms, it is human intervention that provides the remedy. And the results can be transformational.