Hi all. I’m writing jointly today with colleagues Andrew Drinkwater and Pat Lougheed from Plaid Analytics, a company with whom HESA is teaming up to offer services related to improving the state of data collection, analysis, and use on campuses across the country. We’re not going to spend time giving you an outline of what we’re offering (although do click here for more if this interests you), but we do want to talk about how we see data environments evolving in higher education institutions, and how they could be improved. 

Having good data matters particularly for institutional efficiency and transparency, both of which matter even more in tough times, such as the ones we are currently enduring. Put simply, the waste and inefficiency that could be sloughed off as “the price of doing business” in good times are no longer affordable. To avoid enshittification, our institutions need to make harder and harder decisions and do so at a substantially accelerated pace.

The main thing to understand about institutional data is that it’s not usually the case that institutions “lack” data. Some may be missing some data here and there, particularly for student and employer preferences, but for the most part, the bigger issue is that they have more data than they can use efficiently, for several reasons.

One big reason is data timeliness. Often, what data is available internally is “as of” some date when a data snapshot was created—term census dates for enrolments, fiscal year/quarter end for finance data, etc. These dates exist because of government reporting deadlines, but it’s hard to build a data-informed culture when everyone waits for the next reporting period only to use data which is immediately out of date.

Another important reason is what you might call data literacy mismatches. Broadly, the people collecting the data and the people turning it into displays and the people using it to make managerial decisions don’t always have a particularly good of understanding of each other’s perspectives, and the people at the top aren’t always good at seeing how data can be used to explore alternative futures (i.e. exploratory analytics) as opposed to describing a current or recent situation. And so what you get is a lot of data that doesn’t quite answer the right questions, presented in a way that doesn’t quite enable decision-making, sort of like trying to read a small-print novel through two stacked sets of reading glasses.

But the biggest reason is that institutions are simply larger and more complex than they were 30 years ago, and efficient decision-making means there is a need for more actionable data in more people’s hands. To understand its situation, an institution doesn’t just need to understand the data it is collecting about its many different facets (prospective and current students, alumni, academic and graduate outcomes, finances, facilities, etc.) it also needs to understand how all the different buckets of data they have interact with one another. 

The problem fundamentally is one of data silos. In our experience, many institutions are very, very tight with data, not just in the sense of being afraid to put data in the public sphere, but even in allowing data to be made free to circulate within institutions. Too few people have access to needed data, even when it is mission-critical (you would not believe the number of conversations we have with Deans and Department chairs who simply can’t get access to data on enrolments and applications from their own institutional research/ registrarial data shops). In such an environment, adding data can help (sometimes), but it is the liberation of data—allowing datasets to be not just in the open for people to see, but to allow people within an institution to get the different data sets to talk to each other. 

 And then, there is the issue of AI and LLMs, which is forcing many institutions to up their game (in the US, at least; it’s not clear that many people have given this much thought north of the border, yet, but trust us, it’s coming). How many institutions can say that their key financial, personnel, and recruitment teams know how to leverage their existing data using AI, or are in a position to adopt new tools that could be an improvement? This isn’t simply a question of whether they are enthusiastic about AI—it’s a question of whether institutional data governance is in sufficiently good shape that AI tools could generate accurate analysis. Where data definitions are unclear, data lineage isn’t understood, and ultimately, data field meaning isn’t obvious to stakeholders, using AI to interpret your data will just get you to a wrong answer faster.

 That means having common understandings across the institution of what problems the institution is actually trying to solve, then working backwards from that to design data capture and display systems, data silos be damned.

• This isn’t an easy problem to solve, but there are some steps everyone could probably take if they wished:
• Understand what questions various institutional stakeholders want answered, with a bias towards making data as open as possible (while respecting important safeguards on personal information).
• Map data needs and current data resources. Understand how many of the above questions can be answered with existing data, provided only data silos could be dismantled.
• Understand how different audiences within an institution can best interact with the data (not everyone needs the same displays).
• Only once you have done all this, start thinking about what extra data you need in various areas, and design a data collection strategy.
• Designate a lead for data governance, someone who can bring various players to the table and make decisions regarding how to liberate data and make it more available and more useful. We see collaborative approaches to data (and data governance) being the most effective—usually a partnership between institutional research, information technology, and the provost’s office. 

The second, potentially much harder, problem is how to create space for new types of analysis. How can the culture of the institution be changed to one where units are encouraged to play with data on their own, coming up with their own analyses using linked data sets, potentially using new methods such as AI? We suspect this is probably easier for institutions to tackle after they’ve dealt with their more basic data problems, but it couldn’t hurt to designate a team right now to explore how existing data can be more fully exploited for analytical purposes using AI and what kinds of investments need to be made to get there.

In any event, universities and colleges are unlikely to flourish in future unless they have a flourishing data culture. Today, we’ve laid out some ideas about how to make sure managers get the data they need; tomorrow, we’ll lay out how to make sure this data gets used to improve governance as well.

And of course, if you’re intrigued by these ideas and want to know how to put them into practice at your institution, get in touch with us.