This paper sets out to construct and present the evolving conceptual framework of the SpeakApps projects to consider the application of learning analytics to facilitate synchronous and asynchronous oral language skills within this CALL context. Drawing from both the CALL and wider theoretical and empirical literature of learner analytics, the framework sets to theorise and problematise one construct, such as the authenticity of language production and interaction during a CALL mediated session. It is considered within this paper that a CALL learning analytics approach is usefully conceptualised from a student-centred perspective as opposed to an interventionist perspective based on action-oriented approaches to language learning. The paper considers that structured data--usually represented by statistics from Virtual Learning Environments (VLEs) or Learning Management System (LMS)--and unstructured data (user generated content) can be harnessed to support the oral language learning process. The paper considers the implications of using this framework with the SpeakApps platform. [For full proceedings, see ED565087.]