Learning analytics (LA) are summaries, visualizations, and analyses of student data that could improve learning in multiple ways, for example by supporting teachers. However, not much research is available yet concerning how LA may support teachers to diagnose student progress and to intervene during student learning activities. There is evidence for two mechanisms, namely that LA tools can (1) aggregate information to a manageable level and thereby lower information load, and (2) increase the teacher's confidence and specificity concerning the diagnosis of the situation. By means of micro-analytic study of teachers' real-time use of LA tools, these mechanisms are further unpacked to answer the question how the mechanisms occur and how these results influence the broader field of LA.