Predictive Analytics and Higher Education

Predictive analytics is the practice of extracting information from existing data sets in order to determine patterns and predict future outcomes and trends. In essence, predictive analytics makes predictions about the future, for example, enrollment targets. Organizations are turning to predictive analytics to increase revenues, mitigate risk, and meet goals.

There is a need for predictive analytics in higher education because of increasing pressure to enroll and retain students with limited recruitment resources. According to a 2017 survey of college and university admissions directors, 66 percent missed their enrollment targets,1 and 70-80 percent of four-year institutions have either reduced or kept flat their recruitment and admissions budgets2.

These pressures have lead to growing interest among higher education institutions for predictive analytics tools: 41 percent of colleges and universities are using data for forecasting and predictive analytics3. With heightened interest among schools, there are an unprecedented number of vendors labeling portions of their products or services as “predictive analytics” or “predictive modeling.”

There is not an industry consensus for what constitutes “predictive analytics” or what level of accuracy an institution may expect to glean from a solution labeled “predictive modeling.” Some provide data visualizations that are just that, data that looks pretty, while others provide insights that can inform decision making. The range of products can leave buyers unsure how to readily distinguish between offerings.

There are some essential features that you should look for when considering a predictive analytics solution. One of the most important features is the data science process: Cross Industry Standard Process for Data Mining (CRISP-DM). CRISP-DM consists of several steps that all have to be completed successfully for the whole project to be successful. Think of it like protecting your home against intruders, one mistake and it does not work.

The steps include:

  • Business Understanding
  • Data Understanding
  • Data Preparation
  • Modeling
  • Evaluation
  • Deployment

At Othot, we follow CRISP-DM to deliver actionable insights to our customers that can impact the future. Starting with step one, we identify the right questions, what we call our High Impact Questions (HIQs), that matter most to your institution and build a solution around that specific question.

Over the next few weeks, I’ll talk more specifically about CRISP-DM, as well as some other some features to consider if you are evaluating predictive analytics tools. In the meantime, if you have any questions or comments, please contact us at

3 KPMG 2015-201 higher education industry outlook survey -

By Mark Voortman | October 11, 2017