This paper reports on findings from the MoCoDa2 project which is creating a corpus of private CMC interactions from smartphone apps based on donations by their users. Different from other projects in the field, the project involves users not only as donators but also as editors of their data: In a web-based editing environment which provides users with access to their raw data, they are supported in pseudonymising their data and enhancing them with rich metadata on the interactional context, meta-data on the interlocutors and their relations, and on embedded media files. The resulting corpus will be a useful resource not only for quantitative but also for qualitative CMC research. For representation and annotation of the data the project builds on best practices from previous projects in the field and cooperates with a language technology partner.