The power of data


By Jon Kirk

There’s plenty of competition and change in the Higher Education sector. Which means that Universities, inevitably, are looking to use smarter, more cost-effective tactics when recruiting students. For an agency like SMRS, this means providing the tools and advice to its clients, helping them to make smarter and better-informed decisions.

Using data in the right way, all of us (universities and agencies) will benefit with more sophisticated online targeting, and improvements in online tracking. It’ll enable us to identify an even clearer link between advertising and student recruitment.

This idea formed a main part of a 2015 project by Birkbeck, University of London, and SMRS. Together, we used anonymised student data to create strategies – strategies that’d improve both recruitment and retention. Essentially, Higher Education Institutes (HEIs) collect more data about potential ‘customers’ than most sectors – excluding credit and insurance companies.

The idea for the project came from an exercise in media consumption research, where we identified the dominant Experian Mosaic Profiles of Birkbeck’s unique student population. This led us to adjust our media activity.

This insight was easy to find. We used student postcode data – and both Birkbeck and SMRS realised the potential of using applicant and student data in this process. It would unearth further, unique findings, based on the sheer amount of data variables available to us.

We wanted insights on which demographic, or behaviours, would affect a student’s likelihood to enrol and stay. We used a process similar to what a financial service would use; almost like assessing someone’s credit score. applicant’s credit worthiness. But first and foremost, we wanted to use this information to help applicants and students enrol, and stay with Birkbeck. This information wasn’t going to be prioritised in any way.

The Process

1. Get the raw data and

2. Organise the data

This involved collecting 5 years’ of student data and 38,000 cells from 120 variables on our students’ behaviour. We noticed that London travel zones were important to Birkbeck students, and we cleansed this data by removing variables that weren’t universal to all students. After all, not every student completes every part of an application form.

3. Analyse the data and

4. Generate ‘Decision Trees’

At this point, the number of variables and the number of data units was too great to analyse on our own. So we used data analytics software. From here, our Data Analyst used data analytics software to develop ‘Decision’ trees, and identified the significant variables that would affect an applicant’s likelihood to apply. Then, we could work out how likely a student would stay with the University.

5. Analyse the results

Due to the granularity (and detail) of data from the decision trees, and the fact that we identified different factors and variables for different Schools and Departments, we spent a significant amount of desk-based analysis to make sense of the decision tree models, and make sense of it.

Outputs and actions from here

It’s clear that we’ve been able to identify behavioural and demographical characteristics – ones that identify students as more or less likely to enrol. Then, we’ve been able to adjust marketing tactics accordingly.

Behavioural and demographical characteristics have also been identified as factors that encourage students to stay – which has led to an internal project at Birkbeck, University of London. They’ll use this to put measures in place, and improve the retention rate of groups apparently less likely to be retained.