This paper presents an analysis of how people construct playlists and mixes. Interviews with practitioners and postings made to a web site are analyzed using a grounded theory approach to extract themes and categorizations. The information sought is often encapsulated as music information retrieval tasks, albeit not as the traditional "known item search" paradigm. The collated data is analyzed and trends identified and discussed in relation to
music information retrieval algorithms that could help support such activity.
