Continuous Deployment at Facebook and OANDA: Practices, Quantitative Analysis, and Lessons Learned

Continuous deployment is a software engineering practice that enables frequent, incremental releases to production, adopted by companies such as Facebook and OANDA.

The paper describes the core elements of continuous deployment, provides background on agile and DevOps, and outlines the deployment process, testing, release engineering, and configuration management.

Two case studies are examined: Facebook, with thousands of engineers and a codebase serving billions of users, and OANDA, a smaller forex trading platform. Data from Git repositories, deployment logs, and fault databases were collected for the period 2008‑2014.

Quantitative analysis shows that despite a 20‑fold increase in team size and a 50‑fold growth in code, productivity (measured by lines of code deployed per developer per week) remained stable, and the rate of severe production failures did not increase proportionally.

Key findings include: (1) continuous deployment scales with organization size; (2) management support and automated tooling are critical; (3) developers prefer faster release cycles; and (4) quality can be maintained without a separate test team.

The authors discuss cultural, managerial, and technical lessons, limitations of the study, related work, and conclude that continuous deployment is feasible at large scale but requires sustained investment and careful resource monitoring.