23 October 2017 Register to receive our free newsletter by email each week
Advanced Search
View Printable Version
Mining the power of data to boost student success

Data analytics is becoming increasingly important to improving the effectiveness of almost every profession and academia is no different, but knowing what data is important and how to use it is critical.

This was among the key messages of the third annual conference of Siyaphumelela (‘We succeed’), an initiative to improve the capacity of South African universities to collect and analyse student data to boost student success.

Held at the end of June in Johannesburg, the conference saw leaders from the five universities selected to the Siyaphumelela programme profile their efforts and challenges in using data to foster student success. The five universities include the Durban University of Technology, Nelson Mandela University, the University of Pretoria, University of the Free State and the University of the Witwatersrand.

Dr Mark David Milliron, co-founder and chief learning officer at Civitas Learning, told the conference that one of the main challenges facing universities when it came to data collection and use was that universities often only measure data that is convenient to collect.

The ‘right’ conversations

Therefore, they consistently draw correlations from information that is convenient to collect and which they use to determine causes. This, he said, has led to a host of demographically and financially oriented conversations as opposed to some of the broader discussions that are needed about what the key obstacles are to student success.

“That is why we, as a community of practice, need to challenge ourselves,” said Milliron. “Finding the right data and understanding which of these data elements really work – that's statistician work. But, to get it to the right people in the right way, that's artistry. That means testing data to find the right way to incorporate it to get the best out of faculties and advisors.”

As an example, he referred to early-warning systems at the University of Arizona in the United States, where at-risk students would be alerted by automated messages to improve their academic performance. This was initially thought to be a grand solution to the problem of poor student performance but, after testing, it was revealed that students felt suffocated and demoralised by the way the initiative had been executed.

Demoralising messages

Said Milliron: “What happened was: one, they were using imprecise academically fuelled triggers that weren't identifying model students reliably; secondly, they started sending messages to at-risk students that basically felt like a slap in the face.

"The message was along the lines of, 'Our systems have shown that you have missed this many lectures or that your marks are this low' and students basically felt that the university was sending them a message to say 'we have noticed that you're failing'. And it ended up causing a spiral in students.”

After testing the data, the university realised that the way data is brought to students matters significantly and the message was changed to one of encouragement and support.

“One of the messages we sent to students said: 'We are proud of you, you're succeeding at this institution, but we know university is hard. These are the challenges that students are facing: childcare, financial issues, transportation issues... if you are currently dealing with any of these challenges, please reach out to us and we will help you… Congratulations on your continued success.’”

He said that the message was the product of three iterations. It was also discovered that messages with generalised subject lines were not opened as much and that mails that came from a team didn't get opened as much as those that came from an individual.

Poor data literacy

Professor Wendy Kilfoil, director of the Department for Education Innovation at the University of Pretoria, said data literacy was generally inadequate in the country’s universities. She said educators tended to rely on excel spreadsheets, which they were often unable to interpret.

“We don't really have the capacity for data analytics in this country. We're not producing many data scientists. The other thing is that we collect certain data for the government but we don't look beyond that to think about what other data we need. It is normally only academic data,” said Kilfoil, adding that universities are not able to get access to a variety of data sources.

Thus, while demographic data and information on success and failure rates were relatively easy to find, formative data from a variety of systems and sources was difficult to find.

Referring to her own institution, she said there still exists a ‘silo’ approach whereby departments and lecturers keep their own records of academic data on spreadsheets, making it impossible to predict trends and try to address potential crises in students’ academic journeys.

This meant that when students sought advice on what combination of courses they should take next, Kilfoil and her colleagues were hindered by the lack of access to wider data. “One cannot access a student's marks so you can't identify, outside of a particular module, which courses a student struggles with and which subjects they are strong performers in,” she said.

Finding a solution

Professor Alan Amory from the South African Institute for Distance Education or SAIDE, which organised the conference, said there were many datasets in the South African education system, ranging from school data to matric data, higher education data and there are also data available for employment.

The challenge, therefore, was to find some way to harness all the data in some kind of system that is not only available to some, but to all institutions and perhaps even to private institutions or NGOs that operate within the sector to help students. He said SAIDE had been working with the five universities to collate data and make it more accessible.

“In our first meeting, we had a wide-ranging discussion about how we need to go about finding a solution to this problem. It was decided to create a proposal to think about how we could use this data. We had conversations about how we put data in a warehouse, who are the audiences for this data, how to use the data for research … these discussions will result in a concept document that we will put forward at the next convening meeting,” he said.
Receive UWN's free weekly e-newsletters

Email address *
First name *
Last name *
Post code / Zip code *
Country *
Organisation / institution *
Job title *
Please send me UWN’s Global Edition      Africa Edition     Both
I receive my email on my mobile phone
I have read the Terms & Conditions *