Speaking the Language of Data Visualization
Understanding data visualization technologies is critical to recognizing and responding to enrollment trends and patterns — particularly for community colleges facing precipitous decline.
As a 21st-century consumer, my home is full of sophisticated technology that is deceptively simple. On any given day I may use my smart speakers to get directions, find a recipe or adjust my thermostat, all without considering the complexity behind what drives this technology. Similarly, those of us in higher ed who use data visualization technologies to reveal underlying institutional priorities and pain points may also feel a pull between sophistication and simplicity.
In both instances, learning something new or seeking to achieve a desired result requires the first step: asking questions. However, the real challenge lies in the art "asking" technologies the right questions. An ask that is made in a way that is outside of how the technologies understand will quickly lead to unexpected complications or undesired results. This is particularly true when it comes to data visualization technologies, which graphically represent important data points and offer a helpful way to reveal trends and patterns.
Making Community College Enrollment Data Actionable
This year has provided community colleges with many difficult questions. Recent data released by the National Student Clearinghouse Research Center has shown a sharp decline in community college student enrollment — particularly among first-time students, where we see a drop of nearly 23 percent from last year's numbers. Community colleges across the United States are asking: How do we best serve our communities as they face unprecedented challenges to health and economic security? How do we effectively deliver upon our mission in such circumstances? How do we ensure that students who do enroll are retained and supported as they persist through to degree completion?
As higher education professionals, we bring human distinctions and nuances into our interactions with data visualization software. All of the questions above are fair, but are asked in ways that make visualizing data difficult. Just as my home smart speaker cannot possibly know which of the lights I am asking it to turn on without clarification, a data visualization program needs similar elaboration to run a successful query. For example, how are we defining "serve"? What metrics do we use as evidence of our mission? What data captures our retention, support and persistence initiatives?
Five Steps for Clarifying Data Visualization Outcomes
To help community colleges analyze their data for actionable decision making through data visualization technology, let's walk through five steps needed to clearly establish analytic priorities to visualize.
1) Identify the problem or phenomenon. In theory, this is the easy part. Most institutions abound with common narratives about what is going well on their campus and what needs improvement. We want this to be a general starting point, which we then filter to be more specific.
For example, the common narrative may be that "retention needs improvement." Successful analytics require that we better articulate the problem/phenomenon. It is a clearer starting point to state, "While COVID-19 has recently complicated matters, enrollment trends at our institution indicate that year-to-year retention has actually been declining since 2016."
2) Choose a specific focus area. Remember my earlier comments about "asking" technology questions correctly? I can ask a smart speaker to play "music" — which will work, but without a focus the algorithm will data-dredge and take random guesses at what I will find interesting. Asking for a given genre, artist, album or song will produce more relevant results.
Similarly, with higher ed analytics, one can run a query about student retention, but it will likely yield information that isn't relevant or specific enough to make actionable. Instead, focus on something more specific, such as the efficacy of specific student success interventions in correlation to student retention.
3) Identify items to analyze. With a problem or phenomenon stated and an area of focus identified, one can consider what items from this area generate data to visualize and analyze. It may seem obvious, but worth stating: It is impossible to visualize and analyze data that does not exist. Arriving at this moment will either be an affirmation of the data collection efforts your campus has invested in or a call to action to shift data initiatives to ensure the data you need for your focus area can be collected moving forward.
Following the retention example, an institution may identify items such as early course registration or targeted engagement outreach by focusing on student success interventions.
4) Generate questions, one item at a time. If a campus has been diligent in data collection efforts, it is tempting at this point to throw all the data together and see what meaning comes out. Patience must be exercised, though: An abundance of data does not inherently produce an abundance of meaning.
At this stage, be methodical in producing a series of analytic questions for each of the items your institution has generated data on related to your focus area. Following our retention example, one may ask a whole series of questions:
- Are students who participate in early registration retained at a higher rate than their peers who register later in the term?
- Are students who participate in early registration enrolling in high "DFW" rated courses at the same frequency as their peers who register later in the term?
- Does participation in early registration show an impact on performance of course learning outcomes?
5) Refine a question you can act upon. Each of the questions above challenges us to view an item from a slightly different angle and consider what actionable steps we might take from what we learn. Further refinement of our language tasks us to go beyond common language and state explicitly what we wish to know of our data.
For example, consider the first unrefined question above: Are students who participate in early registration retained at a higher rate than their peers who register later in the term? In this question, how are "students" defined? Are we concerned about full-time, part-time or both? How is "early registration" defined? Precision in our definitions allows for greater accuracy in our query.
A refined version of the question might be: Excluding students who registered for less than 12 credits, are students who participate in early registration (within the first two weeks of registration for the next term) retained at a higher rate than their peers who register later in the term? While this question would be a mouthful for one person to ask another, an analytic inquiry will respond with precisely the desired information, and your institution can respond to accordingly.
It can be easy to forget that our engagement with data visualization technology truly does require fluency in speaking the language of data. Our daily consumer technologies generally do a good job of hiding this reality from us — they act as a translator because they have a fiscal incentive in anticipating our asks, inquiries and interests. When faced with analyzing our own institutional data, no such translator exists, and it is up to us as agents of change at our institution to learn how to "ask correctly."