TS Operations System and Practices at Ctrip's Public Technical Service Center

In August 2021, Ctrip's Technical Assurance Center restructured its release‑system support team into a Public Technical Service (TS) team to continuously improve operational efficiency and service quality.

The article defines key terms such as generic R&D tools, one‑click fault reporting, service accounts, IM+, TS, and TS Butler (the management platform).

When user numbers grow, dedicated full‑time TS becomes necessary because documentation alone cannot meet diverse user needs and rapid feature iteration.

The TS organization evolved through three models (illustrated with diagrams), with the current Model 3 offering the best balance of coverage and efficiency.

Since adopting Model 3 in August 2021, the team now supports nine generic R&D tools, achieving a 53% overall self‑service rate, an 81% first‑line resolution rate, and a 50% reduction in TS onboarding time.

The TS operation system is visualized with a context diagram that shows relationships among users, the service account, and various internal services.

Five standout components of the system are highlighted: (1) AI‑enabled service account to boost self‑service, (2) automated wiki and “five‑question” push, (3) a crawler tool to ensure wiki link availability, (4) streamlined tagging to reduce effort, and (5) a data‑driven monitoring framework.

To enable AI chat, the team created a dedicated service account on TripPal, configured an AI bot, populated it with over 700 wiki entries, and achieved a 79.5% top‑4 accuracy after three training rounds.

Wiki push strategies (active push, one‑click fault push, and five‑question push) are described with response categories and illustrated with performance charts.

The crawler tool checks external and internal links in wiki pages, validates content against expected topics, and sends email reports; its architecture is shown in a diagram.

Tagging is optimized by combining product, team, and module information into a single tag, enabling quick retrieval and reducing labeling effort, as demonstrated with the “Captain” product example.

A comprehensive monitoring dashboard tracks total and per‑product self‑service rates, first‑line resolution, incident handling time, user satisfaction, PV/UV, incident volume, wiki usage, and individual engineer metrics.

Operational insights reveal higher quality wiki, reduced manual effort, cross‑product support, and significant time savings for both developers and users.

Future plans include proactive training, systematic feedback to developers, further tool enhancements, and leveraging AI advancements to lower learning barriers for both TS engineers and users.