Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating similarity groupings of software projects. One of the methods includes computing respective values for a plurality of analysis metrics associated with each software development project of a plurality of software development projects, wherein the analysis metrics include snapshot metrics that represent respective properties of the commit history of snapshots in the software development project, functionality metrics that represent respective properties of software elements in the software development project, or both. A similarity grouping is computed for the primary software development project based on the respective computed values for the plurality of analysis metrics for the plurality of software development projects, wherein the similarity grouping for the primary software development project comprises fewer than all of the plurality of software development projects.