Understanding how narratives spread and influence target audiences in dynamic information environments like social media has a broad range of applications, e.g., to anticipate public reaction to policy changes or disinformation campaigns. In this paper we describe our ongoing efforts to define, collect, detect, and annotate narratives online in order to analyze and model the spread of these narratives. We also share lessons learned and our vision for how to improve these sub-tasks. We define narrative as a collection of statements expressing a point of view on some topic. These statements may explain events, interpret the motives of actors, or emphasize specific perspectives. We believe this minimally-restricted definition helps our findings apply across disciplines.