This article is more than 5 years old

How analytics can support student success

Charles Prince, of the University of East London, explains how predictive analytics can use data to improve the student experience.
This article is more than 5 years old

Charles B. W. Prince is Chief Higher Education Officer at the Global Education Executives, LLC.

It’s time for higher education and further education institutions to invest more in predictive analytics.

Such approaches aren’t just about improving retention and graduation rates, nor are they for those institutions that suffer from low rates, but they represent an opportunity to improve the overall student experience.

Students deserve more personalised learning in higher education. Predictive analytics can help make learning and support services more targeted, identifying students according to their characteristics, and understanding the risk of them disengaging or dropping out. Some institutions might struggle to provide this kind of approach due to budget constraints. Therefore, to explore personalised learning from a “low-impact” budget perspective, I’d like to share how we use predictive analytics to improve the experience of students at a predominantly black, Asian, and minority ethnic (BAME) institution – the University of East London (UEL).

Trying predictive analytics

There has been much debate about the use of predictive analytics in higher education. Some see it as a way to improve the student experience with more efficiently-targeted resources, others worry it is invasive and ineffective. Done right, I think it’s the former.

Since the 2016/17 academic year UEL has invested in a predictive analytics system and platform with Civitas Learning, to better understand students – in a way that allows us to shape our educational offerings, and academic and career support – to better suit their needs. Civitas is a US-based company that provides various products built using institutional data to provide clients with predictive analytics about students who are disengaging or dropping out.

Predictive analytics systems process student data from a secured, central repository (SITS/DELTA) with the aim of supporting students with professional services. The use of the data helps target students with specific interventions and allows us to ensure we are widening access and participation to all students in more meaningful ways. All students are informed about how UEL processes data at the time of entry. As at many other universities, we continue to work collaboratively with IT to ensure all of our processes are transparent to students, faculty, and staff.

How the data can help

In our predictive analytics system, UEL is only expected to have a 75% retention rate. This is based on the number of students enrolled at the institution over the past five years that have dropped out. In the simplest terms, this predicted score is calculated by the algorithm within the Civitas system, which evaluates all student records and identifies the characteristics of those students who continue and complete their studies, and those that do not. The system then applies those outcomes to currently enrolled students, controlling for their characteristics.

Figure 1: Snapshot of a predictive analytics dashboard

Source: Civitas Learning and UEL

This is one of the reasons why we appreciate predictive analytics because it supports targeted approaches from those university services focused on improving engagement, retention, and graduation rates. Predictive analytics helps us to identify which students to focus our limited resources on, in order to better understand their needs and respond to their actual live behaviour, rather than just their feedback after it’s too late.

The data can also have a powerful predictive ability, depending on the subject area and cohort. For example, students in one programme and one year might have a very different set of predictors than students in another. Students who might be more at-risk in our modelling could include those without GCSEs, and/or who are re-enrolling into the university but have not passed one class.

These data points are important for us in the Centre for Student Success (CfSS) team, as targeting students based upon their levels of risk allows for limited resources to be focused in ways that can show the most direct impact. Evidence of such practices should also be important to the Office for Students through direct monitoring and TEF submissions. They will receive this information through our initial registration but also through our ongoing data submissions.

Focusing on students that need support

One area that has seen improvements at UEL is the expanded use of diagnostic testing with students. This has become a standard method for supporting students when they arrive at the institution. But, instead of assessing every student, we are able to target students who might need the diagnostic, and then provide them with plans that help support them in their academic achievement. So far, the students that have been engaged with the diagnostic, have seen a 3.5 percentage points increase in their assessment grades. Once we have expanded the diagnostic to more students, then we should be able to tailor our student support services even better.

Since CfSS also provides career support, we can target students with on-campus jobs and internships. We take into consideration predicted risk scores as well as their level of bursary qualification, and other data. This ensures that the career support we provide increases their attendance and engagement in the classroom this year, but ultimately, also encourages them to return to the institution next year. The programme has been in effect this first year and once September 2018 comes, we will be able to measure its impact on student retention.  

Finally, predictive analytics will also transform academic advising and professional advising, supporting students who are at-risk or are taking a break from studies. We will be able to assess whether such advice is effective and what components of the advising are directly related to retention goals. Again, advising isn’t for every student, but it can be incredibly effective for those students who are having the most difficult time at university. Predictive analytics will allow us to determine the correct pattern of these meetings, and also which type of advising has better results with different students. It allows us to ensure that students are being assigned to appropriate advisors and advising approaches, to help them to achieve and also to get the biggest return on our investment.

Our future investment in predictive analytics

Future investments in predictive analytics will also help assess the impact of teaching and learning. It will allow us to better predict course completion rates and results, and to see which types of students find which courses harder. As we get a stronger grasp on improving the student experience in this area, we can use research-based strategies, our strategic teaching and learning objectives, and inclusive pedagogy – all to improve the student experience, which, should result in improving retention and graduation rates.

Due to several regulatory changes, predictive analytics will also allow us to build collaborative university infrastructure that allows professional services and academics to work more closely together, in line with the institution’s strategic goals and objectives. This will require more investment in the expansion of the platform, so that other areas of our predictive analytics system includes course data, advising data, and module-level development that assesses the impact of teaching and learning.

Translating to policy

The Office for Students should aim to build a strong predictive analytics framework for all the providers it will be regulating. It isn’t just about data though, but actually having providers and institutions develop plans that use authentic data, from the time of a student’s entry through to graduation, and beyond. I am supportive of OfS’s aims and think that there is an opportunity to move the conversation about data and analytics to a new place for higher education in the UK.

This approach can also be translated into policy, as OfS will be able to shape the way institutions are developed and rewarded for their actions in this area. For example, instead of using LEO data, funding for courses could be based on predictive analytics versus what actually happens with student behaviour. What a student does or doesn’t do over five years cannot be a realistic judgement of the university. However, predictive analytics can help show what interventions have had a direct impact on student success. That data could inform university standards, rankings, and even funding.

UCAS could also look at how they might use predictive analytics, to inform post-18 choices. Institutions are going to be hard to change. From experience, just talking about predictive analytics and translating the data into practice is a major challenge in itself. However, it makes it easier to inform students, based on their characteristics, of which universities they will be more successful at than others.

If you want to inform students, give them the information as a powerful way for them to make the most appropriate decisions about the right university for them. By giving them this level of power, then maybe institutions will be more willing to change quicker and more effectively. That is not as commonly done now as it would be if a true predictive analytics approach was implemented. This could change the face of higher education by mandating institutions to reveal to students who they are really built to educate.


2 responses to “How analytics can support student success

  1. Excellent article. As institutions are increasingly faced with thinning fiscal resources, platforms and tools like this will assist them in ensuring that any student vulnerability points have been prepared for.

  2. Those who have knowledge, don’t predict. Those who predict, don’t have knowledge.

    Lao Tzu

Leave a Reply