It’s been a while since I finished university but it’s always a pleasure to visit one. The last time, I was there not as a student or a speaker, but as a technology vendor and I was surprised by how much had changed. UK and European universities are transforming into American-style academic institutions which use market research to help set their curriculum, identify student segmentation for their marketing campaigns and fight over prestige charts in a “University league”.
Each industry is focused on a few Key Performance Indicators. For banking it might be “days past due” (the number of days people are late with their payments), for mobile operators “churn” (how many people left their subscription base). What is it for higher education? Drop out - the percentage of students who drop out.
The economics behind this number are pretty simple. Imagine a university tuition fee of £8,000 annually with an average drop-out rate of 20% in the first year. A student who finishes a 3 year bachelor-degree programme is worth 3 x £8,000 in tuition, or £24,000, plus income from accommodation, food and books s/he buys on campus, to the university. If a university has let's say 20,000 students, with a 20% drop-out rate in the first year, the university loses 2years x 8,000tuition x 4,000 drop-out students, or £64 million. If they could get the drop-out rate lower by 2% to 18% - 400 less student would drop out and university would have an extra 2years x8,000tuition x400, or £6.4 million income over two years.
Is it worth implementing an early warning system that would flag potential drop-outs using Business intelligence like mobile operators do? The numbers say definitely YES. The question is how, especially taking into account the academic nature of the university.
Frankly, I can't imagine a university making a huge investment into a BI mega-vendor platform (SAP, IBM, Oracle), building an enterprise data-warehouse with reporting, and investing millions of pounds into it over a period of years like banks and telcos do. One needs a specific culture and previous experience for such an endeavour to take place, let alone succeed. I would propose a looser approach. Don't start with implementing new technologies for data integration, but put a thin semantic layer over the existing data sources (by which I mean, spreadsheets scattered across the faculties ;-) and do a human-driven “soft integration”. How would that work?
Step 1) Register all existing relevant spreadsheets into one catalogue of reports and extract the key terms (column names) used in them.
Step 2) Let users who produce these spreadsheets define the key terms they use (student, applicant, conversion rate …)
Step 3) Identify which existing spreadsheets can be combined to answer questions leading to the identification of possible drop outs.
A few sample questions that come to mind might be:
Who's stopped visiting the library in the last two months? (Is there a table of last library entry times and student ID's?)
Who's not attended tutorials or seminars within the last two months? (Is there a table where faculty staff record attendance?)
Anything else you can come up with?
And finally combine these lists and send a study advisor to talk to the students who frequent them and find out what's wrong.
This approach doesn't require a huge budget, resources or significant cultural change. How about a small data try-out? A university should be the right place for experiments ;)