Data against drop-outs: Using big data to manage student attrition

Dr Ellen Wagner, Chief Research and Strategy Officer for the Predictive Analytics Reporting (PAR) Framework

Dr Ellen Wagner, Chief Research and Strategy Officer for the Predictive Analytics Reporting (PAR) Framework

Most experts agree that recent shifts in educational technology are driving change in teacher-pupil relationships and pupil expectations. With the attendant rise of the concept of the higher education student as consumer, student retention is becoming an increasingly important issue for educational institutions. In the US, decreasing graduation rates are causing concern and damaging President Obama’s drive to make sure that by 2020 the US boasts the world’s highest proportion of college graduates (50% of the population).



By George Bodie


Students failing to complete their education represent a problem globally for already financially challenged higher education institutions. Every student ‘lost’ is seen as a financial loss, as universities miss out on fees, government funding and potential future alumni contributions. And of course for students, dropping out can mean losing initial investments and valuable time, as well as the ignominy of being labelled a ‘drop out’.


As higher education provision continues to grow, so too does concern regarding student ‘attrition’ (the reduction in numbers of students attending courses as time goes by). According to UNESCO, the percentage of adults worldwide who have received tertiary education rose from 19% to 29% between 2000 and 2010. This growth has continued in the first half of the decade, albeit at a slower pace. Such growth is subject to what Philip Altbach, Director of the Centre for International Higher Education at Boston College, has called the ‘law of expansion’. Apart from a small number of elite institutions, Altbach claims that any large expansion in a country’s tertiary education sector will be matched by a decline in quality of both education and students – as students of a wider range of ability are being taught often by less qualified staff under conditions of stretched public funding.


Does the growth of higher education mean then, by necessity, that student attrition rates will grow in turn? Not necessarily. Using the analytical power of big data, the Predictive Analytics Reporting (PAR) Framework is offering an innovative way of understanding student loss, allowing institutions to identify causes and thus halt the flow of students dropping out. With more than 351 unique member campuses, over 2.6 million anonymous student records and 24 million institutionally de-identified course level records, PAR seeks to provide a holistic perspective in order to improve student outcomes on a broad scale across the US.


Dr Ellen Wagner, Chief Research and Strategy Officer for the framework explains: “PAR uses predictive analytics to find students likely to be at risk of dropping out, finds the variables /reasons they are likely to drop out, which then helps target interventions most likely to provide support for that student at the point of need.”


Using predictive techniques usually found in business intelligence settings to aid educational decision-making, PAR does not only seek to predict who is at risk, but also provides analytical data which measures the effectiveness of interventions. According to Dr Wagner, “We spend as much of our energy on intervention measurement as we do on the predictive models and descriptive benchmarks we have developed.”


PAR thus seeks to turn the problem of massification on its head, using the mass data that is produced by increasing numbers of students to limit the problems that usually come with this growth in numbers. In keeping with the zeitgeist, PAR functions on a collaborative basis, with each institution contributing data and experience, leading to an exchange of assets and best practice. Comparing the data, PAR then seeks to identify what is known as ‘common data definitions’ – common variables which apply to the entire dataset. Key variables are centred on student demographics, course level records, student level academic records and institutional records. These variables produce a ‘student success matrix’, the end result of which is to predict which students are at risk, and identifying the correct tools needed to help them succeed.


So far, Dr Wagner notes that there have been a number of stand-out lessons arising out of the PAR’s implementation. The first is that “change is good”, but must be collaborative: “The fact is, higher education is a follow-the-leader industry. We all know we need to navigate change, but most of us really want others to go first. The best way to deal with this is to establish a trusted group that can collaborate to move the needle on improving student success as a collaborative venture involving multiple schools.”


Furthermore, Wagner states that when working collaboratively, data definitions must be shared: “Common data definitions are a game changer for scalable, generalisable, repeatable learner analytics. If we all know precisely what we are talking about and have common measures for those conversations that makes it possible for others to join in.”


Dr Wagner points out that lastly, clarity is key: “Change happens when fuelled by transparency and trust.”


This trust is helped by the fact that the PAR framework has been designed by educators, for educators. Dr Ellen Wagner is a household name in the e-learning world, having worked as a tenured university professor and administrator for a number of years, before serving as Executive Director of WICHE Cooperative for Educational Technologies and Senior Director of Worldwide eLearning Solutions for Adobe. At this year’s OEB, Dr Wagner will host a session entitled ‘Leveraging Analytics to Improve Student Success within a Community of Engaged Institutions’, with Dr Karen Vignare, who currently serves as Vice Provost at the Centre for Innovation in Learning at the University of Maryland University College, an institutional member of the PAR framework.


Dr Wagner and Dr Vignare, who is responsible for analysing and implementing innovation at UMUC, will expand upon the successes achieved and challenges encountered in the development of the PAR framework in order to open up a debate (audience participation will be encouraged) regarding how data analytics can help us understand the rapidly shifting world of higher education today.


Participants will discover fresh insights into collecting, using and leveraging data analytics. For those interested in how the use of big data can quell the problems usually associated with massification, the session will provide invaluable insight from two of the leading figures in data analytics and e-learning.


For more information visit:

Leave a Reply

Your email address will not be published.