Last week on the Othot blog, I talked about why higher education institutions should use predictive analytics. This week I’ll discuss one area where predictive analytics can have a tremendous impact: enrollment management.
A predictive analytics tool for enrollment should answer the following question, “What is the likelihood of a student to enroll?” The answer should be a percentage, like “This student is 5 percent likely to enroll.”
Data science bases the “Likelihood to Enroll” score on which students in your prospect pool have characteristics in common with students who enrolled at your institution in prior years, or alternatively, possess the characteristics of students who chose not to enroll.
Prediction can change depending on where you are in the enrollment lifecycle. An ideal solution would give you real-time total enrollment updated throughout the 18-24 month enrollment process.
Predictive analytics can also help an institution shape a class with characteristics that differ from its historic profile, like greater ethnic diversity, higher academic profile, gender (e.g. more women in their computer science program), geography (more out-of-state), etc.
Knowing which students are most likely to enroll and who to target makes you smarter about where to spend your institution’s admissions resources and marketing and financial aid dollars, which is arguably the highest purpose of any predictive analytics enrollment tool.
There are many predictive analytics tools on the market that offer enrollment management solutions and it can be difficult to distinguish among them. One way is to check the data science ‘engine’ that is driving the solution. Predictive analytics tools vary greatly in the strength of the data science they’re built upon, and a solution that provides inaccurate predictions would be more of a liability than an asset.
Key characteristics that make one solution more accurate than another include:
- Data Science – Is it based on linear regression, or is it machine learning?
- Model – If the tool uses machine learning, does the solution use one model for all institutions, or does it choose from multiple algorithms and parameters to find the best fit for your own institution’s data and population?
- Lifecycle Considerations – Does the solution use the same model for each stage of the enrollment process (prospect, inquiry, applicant, admittee)? Or does it build different models for each stage?
We’ll start to dig into the first characteristic, data science, in the next post. If you have any questions, please contact us at firstname.lastname@example.org.