Universities collect a great amount of data on students that is not used to get a full potential benefit for all stakeholders. Nowadays, educational data mining can be used for this purpose. The paper analyses the existing state of the art in available solutions to analyze the data that universities have to enable more informed decisions from all stakeholders, namely, study administration, faculty members, and students themselves. The paper is the first conceptual step in implementing an educational data mining ecosystem based on the existing data on student activity. It describes the available data sources from one side and gathers the needs of corresponding stakeholders from the other side. It enables formulating the task for machine learning based data analysis for early identification of the learner’s characteristics and enabling individualized tutoring as well as informed and early communication about the actual situation of each learner, so enabling various assistance mechanisms to improve student’s learning experience, wellbeing during the studies and ultimately decrease dropout. The paper presents a conceptual framework to create an ecosystem of educational data mining to deliver the necessary data to a corresponding stakeholder as early as possible.