The Intelligent Tutor software is based on predicting which skills are most in need of improvement, and when practiced, which skills need to be taught in what sequence, in other words, Artificial Intelligence applied to language learning. 

Performance metrics are based on gross skills (the B sound) and the sub-skills (the B sound at the beginning of a word, in the middle of a word, at the end of a word, the B next to an "r" in a blend, etc are all different sub-skills).  

As a student improves by successfully completing exercises, their current performance skill level increases.  However when a student is unable to successfully complete exercises, the student's skill scores will not increase and may decrease.   

See also https://en.wikipedia.org/wiki/Bayesian_inference