Eight Questions To Ask Your Predictive Modeler

Eight Questions To Ask Your Predictive Modeler

A lack of competition in the market tends to favor long-standing, entrenched products and services. Just a few years ago, taxis were the only way to get rides if you didn’t own a car or access to mass transit. And while few people would champion the experience of riding in a cab, there was simply no other comparison until other, better services entered the market.

In the enrollment management space, there have been few options to challenge orthodoxy in the predictive modeling space. This has led to prediction products that use outdated technology and institutional leaders without the option to compare products or demand results based on what really matters: out-of-sample model accuracy.

In an environment like this — where technology and know-how are so vital — it’s more important than ever to ask the right questions. Here are eight good ones to ask your current predictive modeling provider.

Is your current provider:

  1. Giving you reports on model performance and accuracy and explaining what they mean?
  2. Utilizing methods beyond logistic regression for predicting yes/no outcomes?
  3. Using multiple types of data: institutional, contextual (publicly available demographic), and behavioral data (such as those provided by Capture Behavioral Engagement)?
  4. Using automatic feature selection to optimize model training?
  5. Dynamically scoring each prospect, every time new data becomes available?
  6. Re-training your model within the enrollment cycle?
  7. Publishing their accuracy statistics on their website?

If every answer is yes, then there is one last question:

8. Is your current provider willing to compete?

In the world of data mining and predictive analytics, accuracy is the only currency that matters and competition is the best way to find the most accurate, highest performing model.

Capture Higher Ed is willing to analyze your historical enrollment data and evaluate the accuracy of an Envision predictive model at no cost to you. Put us up against your current model and see how we fare. We are confident that you will see enrollment predictions in a whole new way.

By Thom Golden, Ph.D., Vice President of Data Science; Brad Weiner, Ph.D., Director of Data Science; and Pete Barwis, Ph.D., Senior Data Scientist, Capture Higher Ed