Digital technologies are becoming more present and essential in all sectors of our life. In education, the intensive usage of digital learning devices contributes to generating a large amount of trace data from digital learning activities. Intelligent exploitation of these traces represents a valuable asset for both device producers (to improve the design of the devices) and consumers (learners and teachers). In this paper, we first share our vision for better exploitation by teachers, of traces from middle schoolers' digital activities generated by their use of tools and digital learning services during different classes. This vision is a part of the AT41 project funded by the French Ministry of Education. This exploitation has to meet the requirements of the different teachers. Conducting such a project is not an easy task, because it has to consider the following issues: (i) the lack of comprehensive and clear methodology to design and exploit these traces, (ii) heterogeneity of teacher requirements that complicates their elicitation and analysis, and (iii) the diversity of trace sources. Secondly, we propose a requirement-driven architecture for Learning Analytics composed of a well-identified life cycle. This architecture is augmented by learner traces. It offers a repository storing both teacher requirements and traces to facilitate the Learning Analytics in generating relevant and valuable indicators.