Evolution of Payment Channel Automation Management at Meituan-Dianping

With the rapid growth of payment transaction volume, Meituan-Dianping's payment team expanded and integrated many payment channels. The instability of banks and third‑party systems caused frequent channel failures, making manual maintenance insufficient. Therefore, a comprehensive payment‑channel automation management system became essential.

Initial Stage (June 2014 – September 2015) The monitoring system could only detect anomalies and alert engineers. Fault handling required manual steps: disabling the channel in the routing configuration, diverting traffic, contacting the bank or third‑party, and re‑enabling the channel after confirmation. Problems included high miss‑alert rates, heavy manual workload, long recovery times, and risk of secondary failures.

Semi‑Automatic Stage (October 2015 – October 2016) Monitoring algorithms were improved (≈95% accuracy) and an automatic “set‑channel‑unavailable” API was added. The routing system supported real‑time status changes and one‑click manual overrides. Fault handling combined automatic traffic diversion with manual confirmation, reducing human effort but still relying on external parties for recovery.

Full‑Automatic Stage (November 2016 – present) The system can automatically detect failures, disable channels, and perform controlled “ramp‑up” traffic tests to verify recovery. A state‑machine governs the ramp‑up process, adjusting volume based on success rates to avoid secondary failures. The routing layer was refactored to support per‑bank handling within cross‑card channels, local caching for low latency, and robust disaster‑recovery mechanisms (fallback data, degradation paths).

Key Improvements Across Stages

Optimized monitoring algorithms and added one‑click status view.

Implemented both automatic and manual channel switching.

Reduced manpower cost dramatically.

Integrated routing with monitoring to enable real‑time channel state changes.

Performance Comparison Data shows significant reductions in fault‑response time, average manpower cost, and average recovery delay after each stage.

Summary and Outlook The automation journey has greatly enhanced fault handling capability, system availability, and operational efficiency. Future work includes further improving monitoring accuracy (>99%), fully automating fault notifications to banks/third parties, handling network anomalies automatically, and prioritizing low‑complaint user groups during ramp‑up.