A report from the U.S. National Academy of Sciences (Freeman et al., 2014) has shown that the use of active learning methods led to significant increases in student exam scores. This presentation demonstrates lessons learned from collecting student behavior data during class from several courses at the University of Michigan. These data show that attendance, by itself, is not related to student grades. Grades are, however, highly related to active participation in the form of number of slides viewed, video captures reviewed after class, volume of notes taken and number of times indicating confusion. Analytics also revealed unexpected relationships between grades and students' emotional state. It is argued that the growth of student behavior data will provide new opportunities to provide students with detailed feedback and provide instructors and advisors new insight into why students succeed or fail.