The achievement gap that plagues students from lower socio-economic strata is well documented, and most acutely affects first generation college students from historically underrepresented minority populations. In recent years, many companies have developed “early warning” alert systems to identify students in crisis early enough for campus professionals to execute a corrective intervention in the hops of rectifying the situation and retaining the student.

In this presentation, we will look at the accuracy of one such predictive model. We will follow a freshman cohort through their junior year, and measure the degree to which the initial predictive score created for the student upon admission was accurate, and look at changes in those scores as the available predictive data pool grew. In particular we will answer the following questions:

• Has the student been retained?
• If so, how accurate was the predictive score and has it changed?
• If not, how accurate was the predictive score, and what was the score at the end of the semester prior to their departure?
• If the student has left the university, are we able to identify their current academic status elsewhere?
• Are there differences in retention by race & gender?
• Is the cost of the subscription to the service offset by gains in retention?