When the chancellor first asked for a map of CU Boulder’s international activity, he said he imagined, and I’m paraphrasing, a globe with sticky notes on it. For the folks in the Reporting & Analytics team the chance to build a view of the campus’s international activity was a prime example for why this team was formed. The project presents two big problems. First, how do you combine data from (very) disparate sources all around campus into one space? Second, how do you represent all of that information in a meaningful way that is both informative and engaging? The third could be, how do you build all of this in the span of a couple of months with limited resources, but we won’t dive into that one here .
The only real solution to the first problem, of gathering and combining data from different systems, was leg work. We worked with the project requestors to hammer out a series of meaningful metrics, and each metric was assigned the name of a “Data Gatekeeper”. Our AVC, Steve Vassallo, got in touch with each of these gatekeepers and asked them for simple csv data dumps from their databases. Each csv was aggregated to a country level, and then combined via vlookups into a single Excel sheet. It’s the tried and true use of Excel, quick, dirty, unsustainable and highly effective. It took a lot of reading through rows of data, wrangling different perspectives and compromising to get all of this into a single file.
Once the data was created, it was passed to Reporting & Analytics to turn into visualizations. At heart, most of the folks here are tried and true data geeks, so there wasn’t much chance of putting sticky notes on a globe and walking away. We started out by considering what was being asked of us: a dashboard at least, right? Dashboards are interesting tools. When you consider what a dashboard was ten years ago, you might first think about your car’s dash. What did it tell you? How fast are you going; what’s the oil temperature; how many rpm’s is the engine running at. All useful points of data that you glance at every once in a while and glance away, back to the road where you need to apply that information and maybe change the way you’re driving. Nowadays, someone says dashboard and you may think of a dozen different apps on your phone that tell you how much you walked that day, if you’re over your grocery budget, even how your phone’s battery is doing and what apps are eating up most of the power. It’s cliché, but I’ll say it, data is everywhere, and as a result, so are dashboards.
I’ve built my share of dashboards in the past, but something about this project just didn’t seem to lend itself to one. Would someone open up this tool, ask “How many students studied abroad in Spain in Fall of 2015?” grab the number, close the tool, and go off to use the data to perhaps change some policy? Maybe. But the chancellor’s request was open enough to interpretation that we could exercise a little creativity, of course understanding the risk that this could impact our ability to deliver a beta version in a timely manner.
Being the data geeks that we are, and coming from a wide variety of backgrounds, my own being a BA and MA in English Literature, we decided to merge “data visualization” with “dashboard” and see if some hybrid would work. Why we approached it this way is because, to us, a dashboard is informative, it quickly tells you what you want to know. A data visualization, when done right, is engaging. To be clearer, when I write “data visualization” I don’t mean bar charts, bubble graphs, tree diagrams, or, hopefully never, a pie chart. I really don’t mean any kind of chart that a software package would spit out when you feed it data and click a button. In my mind, to be a data visualization requires some level of customization where what you’re displaying highlights some specific attribute of the data that you’re working with. Sine we’re working with data from around the world, my very boring first thought was, let’s make a map.
The dashboard component would be through Tableau, representing the data that we have on multiple levels, a world overview, a continent overview and country-level detail view, giving you access to all of the specific data points that you may want to view. The idea was simple: put the two together, some sort of visualization of a map with a detailed dashboard that makes use of all of the “best practices” that we could think of and reasonably apply in a short time period, to make a tool that is both informative and engaging. Easier said than done, and I won’t say that we succeeded, but the effort is our experiment towards making this happen.
We thought about what we learned a long time ago about cubism (I told you we’re geeks), and about how a traditional dashboard is like linear perspectives developed in the Renaissance. A dashboard is a linear interaction, ask a question and get an answer, then repeat at multiple grains of detail, i.e., aggregate your information up to a summary then drill down to the components that make up that summary. If we could combine that with some visualization that incorporated not an informative, linear interaction, but enticed the user to engage with the dashboard in a way that they would click around, generally play with the tool, and discover information as often as seek it out to answer a specific question, we have some semblance of simultaneity. It’s cliché to say the least, using ideas from over a hundred years ago to inform something that could easily be accomplished with the equivalent of sticky notes on a globe. But in using the final product, it does have an element of play, of engagement, that offsets the linear functionality that it’s meant to provide, which is to tell us specific information about CU Boulder’s international activity. By animating the globe and passing the country that was clicked on over to the dashboard as a filter, we’re trying to invite the user to click around and be curious, which hopefully leads to some discovery. The animation is purposeful (and wholly stolen from Mike Bostocks amazing blog posts on how to use his equally amazing library, D3 https://bost.ocks.org/mike/), it starts slow and speeds up, which, if you’re like me, makes you want to keep clicking on the globe and watch it spin. Every time you get that animation, though, information about the country that you clicked on comes up. Wikipedia is probably the best at this type of engaging, time-sucking, and informative form of interaction, and while the BI world is still beholden to the idea of giving people the information they need as quickly as possible, to be informative, efficient, and fast, I do believe that it’s helpful to add in an element that makes your user engage with more than just the data, but with the interface. It’s an idea that we’re still kicking the tires on, trying to figure out how to implement it.
The future for this project is a longer road with a lot of improvements. We’re building a data infrastructure using a nosql database as our warehouse that will automate the collection, cleaning and access of all of this data, so that we’re dealing with live and robust datasets. We’re also hiring more skilled programmers to take the mess that is this beta product and turn it into a slicker environment, where you don’t just click on the globe but can load it on your tablet and spin it with a swipe of your finger, and where the data in the dashboard feeds back to the globe so that the user can get a spatial representation of the numbers in the dashboard. It’s pretty exciting to try and figure out how we’re going to balance creating an engaging environment with a highly informative one, and in the process, build a suite of tools that are helpful for the campus as a whole.
Director, Reporting & Analytics
Office of Data Analytics