Activity-centric data gather feedback on students' learning to enhance learning effectiveness. The heterogeneity and multigranularity of such data require existing data models to perform complex on-the-fly computation when responding to queries of specific granularity. This, in turn, results in latency. In addition, existing data models are inefficient in storing computed results, which are often required for follow-up analysis. These follow-up analyses depend largely on stakeholder objectives, which often impose constraints to the analysis process. In this article, we propose a context-based data model that addresses two challenges associated with learning analytics: the increased processing time due to multiple data granularities required from various stakeholder objectives, and the lack of support for archiving and updating of new information that exist during iterative analysis. We demonstrate how the proposed model can help support analysis and visualization via the XuetangX and ASSISTments datasets. Results show that although our proposed data model requires preprocessing, it requires less time to process queries associated with learning-activity attributes of various granularities compared to existing data models.