Students learn scientific concepts and mathematical calculations relating to scientific principles by repetition and reinforcement. Teachers and instructors cannot practically spend the long time required during tutorials to patiently teach students the calculations. Usually, teachers assign homework to provide practice to students, hoping that spending time on the task and providing appropriate feedback will make students stronger in the concepts. However, these assignments do not provide insights into how the students approached a problem, or which part of the problem stymied them. To provide fine-grained details about how students interacted with physics problems, a tutor was designed with state of the art software tools provided by Carnegie Mellon University's CTAT (Cognitive Tutor Authoring Tools). In this article, the results of deploying the computer based tutor and the lessons learned from about 2000 fine-grained instances of student interactions with the tutoring software are presented.