This chapter discusses the rationales for promoting the data-driven learning (DDL) approach in Cantonese learning and teaching. Instead of relying on prescribed grammar rules stated in textbooks, it is more important for learners, through authentic and natural language data, to explore, internalize and generalize the patterns and usage of the target language. Using the Corpus of Mid-20th Century Hong Kong Cantonese developed at the Education University of Hong Kong, we illustrate the types of linguistic information that can be provided by corpora but not by traditional learning materials such as textbooks and dictionaries: Usage frequency, collocation information, authentic language examples. The DDL approach also concerns that Cantonese learning materials should be more practical and communicative. With the increasing amount of Cantonese language data available in the cyber world, it is anticipated that more Cantonese learning materials will be compiled, which can make the learning process of Cantonese more flexible, exploratory and motivating. Copyright © 2024 selection and editorial matter, Siu-lun Lee; individual chapters, the contributors.