What Hidden GitHub PR Patterns Reveal About Your Team’s Efficiency?

Researchers from the Federal University of Pará (Brazil) and the University of British Columbia (Canada) examined 56 GitHub projects, recording every issue and pull request (PR) and representing them as nodes and edges in a Neo4j graph.

Research Methodology

Issues and PRs were captured with their timestamps, authors, and linkage information. The resulting graph visualizes how work items flow from issue creation to PR submission, review, and merge.

Identified Workflow Patterns

The analysis revealed eight distinct workflow types, accounting for over 1,000 development actions. Key patterns include:

Competitive PRs – multiple contributors submit overlapping PRs for the same feature, often resulting in only one being accepted.

Duplicate issues – several independent issues describe the same problem, typically caused by major changes or poor communication.

Split (decomposed) PRs – a large feature is broken into linked PRs, each handling a portion of the work.

Multi‑issue PRs – a single PR attempts to resolve several issues, slowing review due to increased complexity.

Approximately 35.7% of relationships involve simple issue resolution. Duplicate‑issue nodes were rare, appearing only 15 times among 90,000 nodes.

Implications for Project Management

These patterns reflect a project’s maturity and its need for structured, highly organized collaboration. Competitive PRs can force stricter code‑review standards, while split PRs often improve code quality and review efficiency in larger projects.

Visualization Tool

The team built a visualizer called Workflows Explorer to display the graph results.

Future Publication

The full paper, titled “Using PR‑Issue Graph Topology to Reveal Software Development Workflows,” will be presented at the ACM International Conference on Software Engineering Foundations in Brazil on July 18. Authors include Cleidson de Souza, Jesse Wong, Dongwook Yoon, and Ivan Beschastnikh.