Speaking is an essential part of human communication, making it a major topic in second language teaching. Effective speaking skills comprise a number of components (e.g. pronunciation, the use of grammar and vocabulary). Developing the ability to speak fluently as well as in a socially and pragmatically appropriate manner is thus a complex task. Despite the attention given to speaking in language classes and pedagogical resources such as textbooks, speaking is an area that learners notoriously struggle with. Data-driven learning (DDL) is an effective method for incorporating a number of speaking skills in the teaching process as it can address some of the key issues in teaching speaking that have not been given enough attention in mainstream classes and resources. In particular, corpus-based resources and activities offer a unique opportunity for learners to experience the ‘messiness’ of spoken production and to develop their awareness of linguistic features that are associated with real-life communication. In addition, exposure to a realistic language use can increase learners’ confidence in their speaking abilities, thus helping them to take part in spoken interaction outside the classroom.