Social media is a fundamental source of information on human behaviour; it has been used to study everything from health information behaviour to the functions of democracy. Using this data for research, though, is challenging, particularly given requirements to retain research data post publication. Social media posts are often ephemeral, and notoriously difficult to re-find; a problem exacerbated by shifting social media company policies and algorithms. Nonetheless, research standards require retention of data. While at-scale digital libraries and archives of social media material have been studied, day to day collection practices of digital humanists are less well understood. We present a qualitative examination of the collection practices of 14 digital humanists, with a view to understanding how to support them in creating these important personal digital archives.