This article, and most of the discussion here, appears to be conflating two things.
1. The University of Arizona tries to predict which students are at risk of dropping out using a bunch of data about them, and then offers some extra support to these students. They have been doing this for four years and have improved their retention rate about 3% in the last year.
2. A researcher at the University of Arizona has been experimenting with using location tracking to predict which students will drop out. She has found that using this data allows her to make more accurate predictions than <i>the ones that the university currently uses</i>. She hopes that they will one day start using her data/methods.