End-to-End Dual-Mode Agile Transformation of the C3 Credit System: Practices, Pipeline, and Outcomes

The C3 system supports a large offline and online credit business for the bank, acting as a central platform that connects multiple channels to core accounting systems, with complex technology, massive codebase, and high reliability requirements.

Key problems identified were fragmented configuration management across development, testing, and production branches; manual version delivery processes; and poor work‑management visibility leading to inefficient collaboration.

To address these, the team built an end‑to‑end, fully online R&D platform using a customized TFS+Git setup, establishing a “requirement‑task‑delivery” management framework, a dual‑mode (waterfall + agile) branch model, and an automated continuous‑delivery pipeline that automates building, deployment, code checks, smoke testing, and notifications, cutting 90% of release waiting time.

Agile practices were introduced via an electronic Scrum board, implementing a PDCA cycle with daily stand‑ups, iteration reviews, and OKR‑based quantitative assessment, completing 22 iterations and 6,800 tasks.

The dual‑mode delivery architecture defines separate branch strategies for waterfall and fast‑iteration projects, supports both scheduled (waterfall) and on‑demand (agile) builds, and provides dedicated environments for each mode.

A reliable continuous‑delivery pipeline automates version packaging, testing, and release, enabling developers to self‑service publish test and pre‑production versions, while Jenkins‑driven AIX builds ensure consistent C‑code compilation across front‑ and back‑end teams.

Code quality is enforced through mandatory multi‑level reviews, continuous incremental and full builds, a custom static‑analysis plugin (TFS API + JTest), and branch‑alignment checks to keep feature branches synchronized with test and production branches.

Automated regression testing, built on Ant, JUnit, and Selenium, runs daily core loan process cases, reducing validation time and improving release reliability.

The full traceability chain links requirements, tasks, branches, commits, builds, deployments, and production releases, providing bi‑directional visibility from demand to delivery.

Organizationally, a cross‑department C3 configuration‑management workgroup led the transition, conducting pilot trials, A/B parallel runs, and phased rollouts, culminating in a complete switch on October 25, 2018.

Post‑transition, the team observed significantly shortened delivery cycles, higher product quality, and reduced labor waste, establishing a benchmark for large‑scale IT system R&D transformation.

Future work focuses on further pipeline automation, distributed application delivery, dynamic gray releases, and expanded automated functional and load testing.

Finally, the article includes a recruitment notice seeking DevOps engineers with experience in C#, HTML/CSS, JavaScript, PowerShell, T‑SQL, IDEs, VSTS/TFS, and Scrum/Kanban.