The ‘so what’ rule of data science and student retention
As a child, I loved science. I loved it because science always had an answer; a maths test was never subjective. As I moved through the education system to degree level however, I realised just how wrong I was. Science is constantly in flux - evolving, changing, nuanced. To be a great scientist, you must challenge everything.
Data science is no exception. People revere words like ‘analytics’, ‘big data’ and ‘machine learning’, treating data science like a mythical sage who works with algorithms instead of tarot cards. Analytics solutions are often expected to be fix-all, fast-acting, one-time cures; black boxes that can be put in place to tell you all the answers. This is a formula for failure.
"Analytics solutions are not one-time plasters to stick over deeper issues..."
One of the big areas of data science growth in higher education is student retention; why do students drop out? How can we catch at-risk individuals? Harnessing the data about your students is extremely powerful, but many solutions currently on the market are failing to go beyond the ‘black box’ analytics model. There is no ‘one-size-fits-all’ answer to why students drop out of university - if there was, student retention would quickly cease to be a problem for institutions around the country.
Take the University of Manchester and Manchester Metropolitan University. Two universities of similar sizes (1st and 3rd in the country by size, respectively), in the same city. UoM has a first year continuation rate of 94%, with MMU at 88%. For UoM, 2.3% of students transfer out to other higher education institutions, and the figure for MMU is similar at 2.6%. The difference, lies in those withdrawing from higher education altogether. 3.4% of UoM students drop out of higher education altogether, whereas for MMU the figure is over double at 8.6%.
These reasons for non-continuation cannot be equated. A student dropping out of higher education entirely is in a very different place from one who transfers. It is not as simple as flagging those with low attendance or fluctuating grades - institutions need to embrace a data culture which allows them to deep-dive into their own personal microcosm of student experience.
"Instead, ask “So what?”"
Flagging at-risk students is a great first step to improving student retention, but it is ultimately a reactive measure, not a proactive one. A great retention solution goes beyond risk factors, delving into each finding to ask the most important question in data science - “So what?”.
Turning a statistic into an actionable insight is not easy, but the return on investment is always worth it. Take, for example, a particular course with a low continuation rate. Do you flag those students as ‘at risk’ and simply add in extra support for those individuals? It’s not the wrong answer, but it’s an expensive and not particularly nuanced one. Instead, ask “So what?”. Ask why. Focus on breaking down the data you have and challenging it. You may find the issue points to a misdescribed course in a prospectus - a mistake that actually puts you in violation of consumer rights laws. This is an easy fix once you get down to the bare bones of the problem, but the positive impact on the institution in terms of avoiding fines and retaining future students is potentially massive.
Of course, all this relies on a strong data infrastructure and a culture shift towards a data driven mindset. Data science is just that - a science. Analytics solutions are not one-time plasters to stick over deeper issues; if used right, they can transform your understanding of not just your students, but your entire student experience and, most importantly, how to make it better.