Engineering team scaling research
Throughput, coordination overhead, and topology data from 412 codebases.
A study of how engineering output scales with team size, drawn from public commit logs across 412 open-source codebases.
Throughput
Across 412 codebases, teams beyond 8 engineers showed a 23% drop in median PR throughput per engineer, controlling for tenure and language. The drop continues monotonically through team sizes up to ~25, where the trend flattens — likely because larger teams have already absorbed the coordination cost into their structure.
Team topologies
Stream-aligned teams (those organized around a product surface rather than a function) showed 40% less cross-team review traffic and faster cycle times than functionally-split teams of the same size. The effect held across both monolithic and service-oriented architectures.