The focus of this study was on identifying patterns of student behavior to support data-informed decision-making which would then improve the learning experience and learning outcomes of online English language courses. Learning analytics approach (or more specifically cluster analysis) was used to identify engagement patterns in online learning. Relevant information was obtained from learning behavior data, and student behavior was modeled based on the seven indicators of engagement. Cluster analysis results were presented in the form of three scenarios: (1) grammar-focused semester, (2) vocabulary-focused semester and, (3) overall one-year course activity (grammar-focused and vocabulary-focused semester together). Each scenario helped identify four clusters and revealed different behavioral patterns (Scenario 1: "Restrained", "Focused", "Cautious", and "Social"; Scenario 2: "Result-oriented", "Restrained", "Cautious", and "Active"; Scenario 3: "Cautious", "Restrained", "Result-oriented", and "Focused"). Some patterns are stable and present in all three scenarios ("Restrained" and "Cautious"), while others seem to be specific to a particular learning context. The subsequent expert analysis phase offers an interpretation of the identified patterns, guidelines for the adjustment of the learning process, and a consideration of pedagogical, ethical and technical implications of this approach. Depending on the level of automation, targeted adjustment might be done by human or nonhuman operator. Taking this into consideration, this paper proposes an approach to managing behavioral engagement patterns of students in a course-based virtual learning environment. The approach encompasses four steps which need to be implemented periodically depending on the course material dynamics, so the future delivery methods in the e-learning course could be adapted accordingly.