Purpose: Digital learning has become a global trend. Partly or fully automatic learning systems are integrated into education in schools and universities on a previously unseen scale. Learning systems have a lot of potential for re-education, life-long learning and for increasing access to educational resources. Learning systems create massive amounts of data about learning behaviour. Analysing that data for educational decision making has become an important track of research. The purpose of this paper is to analyse data from an intermediate-level computer science course, which was taught to 141 students in spring 2018 at University of Turku, Department of Future Technologies, Finland. Design/methodology/approach: The available variables included number of submissions, submission times, variables of groupwork and final grades. Associations between these variables were looked at to reveal patterns in students' learning behaviour. Findings: It was found that time usage differs per different grades so that students with grade 4 out of 5 used most time. Also, it was found that studying at night is connected to weaker learning outcomes than studying during daytime. Several issues in relation to groupwork were revealed. For example, associations were found between prior skills, preference for individual vs groupwork, and course learning outcomes. Research limitations/implications: The research was limited by the domain of available variables, which is a common limitation in learning analytics research. Practical implications: The practical implications include important ideas for future research and interventions in digital learning. Social implications: The importance of research on soft skills, social skills and collaboration is highlighted. Originality/value: The paper points a number of important ideas for future research. One important observation is that some research questions in learning analytics need qualitative approaches, which need to be added to the toolbox of learning analytics research.