Online "smart" courseware, eTextbooks, MOOCs, and similar recent innovations are leading to major changes in education. One aspect in common for all of these courseware innovations is their ability to capture large amounts of information about the users of these systems, often referred to as "learner analytics". The resulting massive streams of interaction data can potentially provide valuable, actionable information to students, instructors, and systems developers. But this can only happen if the data are transformed appropriately from the level of button presses and mouse movements to interpretations of user behavior. In this talk, I will illustrate these issues through two case studies: (1) an eTextbook system named OpenDSA, and (2) a system to collect and analyze the progress of intermediate-level student programmers doing class projects.