Continuous Delivery Practices and Evolution at Meituan Delivery

Meituan Delivery, launched in 2013, has grown to handle over 30 million daily orders, requiring a delivery process that can quickly bring new features to users across multiple product lines such as food, flash‑sale, and advertising.

What is continuous delivery? It is defined as a framework that delivers valuable software early and frequently, encompassing continuous integration, testing, and deployment. In Meituan’s context, it links product, design, development, QA, and market release in a tight feedback loop.

Delivery model evolution: The early “serial” model involved eight stages per version and a six‑week cycle. To meet growing demand, a bi‑weekly model was introduced, parallelizing review, development, and testing, reducing the cycle to 5.5 weeks and increasing releases to 22‑24 per year. Further refinement added a weekly dynamic‑release window, allowing pure dynamic features to ship within a week and agile changes to follow the bi‑weekly cadence.

Key practices: Standardized review windows (W1‑W7), clear role responsibilities, AB grouping for balanced workload, and dynamic‑page frameworks (Mach, MRN) to reduce native release friction.

Automation: An automated version‑management system (Tide) integrates Gulf scheduling, CI jobs, and monitoring platforms (EVA, CAT, Sniffer) to handle branch creation, packaging, testing reminders, and gray‑to‑full rollout without manual intervention.

CI pipeline: Five stages—pre‑prep, PR check, development, testing, and release—include specialized checks such as independent compilation configuration, PR merge verification, and automatic shell‑engine updates.

Deployment & release: Distinguishes deployment (testing environment) from release (production), with defined approval steps, automated gray‑scale rollout, and continuous monitoring to ensure stability.

Principles for sustainable delivery: Continuous automation, early problem solving, versioning of code, data, and configuration, rapid feedback, and ongoing incremental improvement.

Future outlook: Focus on fine‑grained CI checks, differentiated PR rules, and shifting more testing to automation to further accelerate delivery while maintaining quality.