Using cold, hard data science to provide warm and meaningful student experiences
Undergraduate recruitment had reduced significantly for two consecutive years at Anglia Ruskin University (ARU). A new approach was needed. One that would more accurately define and better engage their target audiences. Pooling our decades of experience, our solution resulted in attracting 34% more applications – with significant increases across every faculty. We also increased geographical diversity, grew Clearing applications by 36% and reduced the cost-per-conversion by more than 40%. We even won the HEIST award for Best Use of Data and Insight 2021.
How did we do it?
ARU have 20,000 undergraduates, multiple faculties and several campuses. We also knew they attracted applications from a broad geographic footprint but there was too heavy a reliance on the East of England. Our Head of Consultancy Ed Layt, insight specialist Golnaz Sadoughi, and Lead Client Partners Jon Kirk and Vicky Green focused on understanding the dizzying makeup of their prospective student audience to make more informed targeting decisions, develop stronger engagement strategies and ultimately, increase demand.
Better defining ARU’s audience
Our first step was to complete detailed machine learning analysis of application and enrolment data from the last three years. This comprised of joining up data from different systems, undertaking a complex data cleansing exercise and enhancing the data with third-party markers such as Indices of Multiple Deprivation.
We then used various statistical modelling techniques to establish propensity to enrol. This started with decision trees to pinpoint the most important variables from our past applicant data that impact the likelihood of someone enrolling on a course at ARU. We then looked at the correlation of data variables on reaching enrolment and tested the model with one covariate at a time. The significant variables identified were included in the concluding stage of logistic regression modelling. Here, the relationship between variables were finalised and accurate statistically significant propensity segments were created.
Then we segmented this analysis by main scheme and clearing applicants, faculty of interest and campus of application. This produced 56 individual segments. These were matched against market data so we could accurately understand the size of each target market, ARU’s market share, the potential opportunity and who they needed to win market share from.
How the data and analysis was used
This enabled us to make informed decisions around where to target, how to weight budgets and the messages that were likely to resonate with the audience profiles and be distinct from the competition. Different engagement strategies were also created for different objectives. And we were even able to implement specific conversion strategies for different audience segments by lead scoring applicants.
While the campaign was live, we also constantly monitored performance to make sure everything stayed on track.
Building data for the future
As we have outputs for each faculty, we’ve used our industry-leading tools to explore the data down to subject and course level to inform specific media campaigns, now and in the future. Depending on what’s needed, this can help to increase the quality and/or quantity of applications – as well as reduce costs. For example, for ARU, their Google Ads cost-per-conversion from October 2020 to January 2021 was £14.43 compared to £20.52 for the same period 12 months prior. This is particularly relevant due to the granularity of the paid search strategy that was informed by the segmentation research.
The potential of data science is staggering. And something we’re experts in. To hear more about how we could help you through propensity modelling, take a look at our short video here.