The idea of an advisor telling a student who just received an associate's degree to next enroll in a Ph.D. program seems farcical. But, leaders in higher education act similarly when it comes to analytics by embarking to create predictive models without first building a baseline level of understanding around data and analytics.
As a result, we hear all too often, “my institution needs predictive analytics.” This is not where you should start. An institution must build its analytics muscle to reach the stage where predictive analytics can be powerful and transformative.
The desire to use predictive analytics is rooted in the best of intentions. Institutions including University of Maryland University College, Georgia State, University of Kentucky and others have seen success in increasing retention and improving student success through implementing this kind of technology. But diving into analytics blindly underscores a misguided approach to higher education data: Relying on predictive analytics alone will inevitably fall short in delivering value to an institution or its students.
To understand why that is, let’s take a step back. Predictive analytics only recently became part of higher education vernacular, but the technique has been in use for far longer. For example, meteorologists have been able to use atmospheric data to predict the weather since 1950 and in 1973 market researchers developed the Black-Scholes model to predict the optimal price for stock options over time. Today, predictive analytics is widespread in healthcare, marketing, retail, and other industries.
In the higher education landscape, predictive analytics use statistical techniques to find patterns in historical and transactional data to identify risks and opportunities. For example, an institution may want to build a predictive model to understand which students enrolled in the fall are likely to persist to the spring semester. However, in higher education the term has already lost meaning, headed in the direction of “disruption” or “personalization”—and not focused enough on delivering intentional impact at colleges and universities.
Long term success can only be accomplished by looking across the institution and not assuming that predictive analytics is a starting point. Ignoring the foundational elements of analytics practice—descriptive and diagnostic data—will leave an institution ill-equipped to take action on predictions regarding areas such as student recruitment, retention, managing costs and fundraising.
The great irony is that historical data, trends, and business processes directly impact the effectiveness of predictive models. By ignoring historical trends, institutions allow inadvertent blind spots to persist that in turn spur policies, practices, and processes that could end up magnifying problems that an analytics initiative originally aimed to solve.
So, while predictive analytics is often pursued as a panacea, it should instead be treated as a final step in an institution’s analytics journey.
To get there, analytics initiatives should be about leveraging the right metric and the right model to answer a college or university’s most pressing questions. For example, pinpointing necessary descriptive statistics, including admissions funnel, retention, and graduation rates is an essential first step for an analytics initiative to deliver significant value.
Still, in many cases, the question an institution is asking can be answered sufficiently without building a predictive model. Instead of jumping straight to predictive analytics, institutions should work through the four stages of analytics maturity: data integration; descriptive statistics; visualization, exploration, and analysis. Navigating these stages requires building subject matter expertise and cross-functional teams who can validate the models and the conclusions drawn from the data. And, only after building those teams will an institution be ready to finally take on predictive modeling.
One can never be too prepared, though. Additionally, data professionals and leaders in higher education may also want to consider the following tips when pursuing predictive analytics:
- Start with the most important question. Questions will vary by institution but could include “are we discounting tuition for the right students” or “are there significant opportunities to improve student persistence?” Analytics are more than numbers in a spreadsheet or pretty graphs. Asking questions that lead to action and answers is what matters most.
- Technology isn’t the silver bullet; technology is an enabler. Like moths to a flame, data professionals are sometimes attracted and distracted by tools and technology, but the newest or flashiest dashboard isn’t always the right fit for a particular institution’s needs and goals.
- Value data storytelling as much as the data itself. Often leaders at institutions focus only on pulling reports. Data storytelling is the art of leveraging insights from analytics, providing context and commentary, as well as making data easily understandable and actionable. Investments in building this expertise in a team will pay off.
- Define your drivers. Analytics can provide a strategy for transformation, but if leaders are unwilling to make that change, the effort is in vain.
- Be patient but persistent. Building a team and campus that understands how to best collect and act upon student analytics takes time.
By supporting and training staff and faculty in analytics before big data initiatives take off, higher education leaders gain a full picture of what is happening across the student lifecycle and will see a more significant return on investment of the institution’s dollars, time and resources. More importantly, higher education will stop seeking predictive analytics as a starting point and instead ask “are we ready?”
