Digital Quality Measurement System at Qunar: Building, Implementing, and Operating a Comprehensive R&D Metrics Framework

Qunar built a digital quality measurement system to enable precise monitoring and management of system stability, supporting decision‑making for global stability managers.

The article explains how more than 100 initial metrics were identified, refined to about 60 key indicators, and categorized by measurement object (product, project, application, people), service target (executives, team leads, staff), and measurement method (layered, leveled, classified).

A five‑step implementation path—metric definition, data collection, processing, evaluation, and continuous improvement—guides the rollout, with iterative refinement based on feedback.

The platform architecture consists of a data source layer, a model layer, and a presentation layer, standardizing data formats, aggregating metrics into a data lake, scoring indicators, and providing multi‑level dashboards (global, project, application) for trend, comparison, clustering, and drill‑down analysis.

Governance combines team collaboration and PDCA‑style processes: a digital governance committee sets quarterly goals, tracks metric changes, rewards improvements, and reports to senior management, while monthly data reviews drive corrective actions.

Beyond quality, the system addresses software decay by defining a complexity model that measures static attributes (code size, cyclomatic complexity) and call attributes (internal/external dependencies), normalizing values to a 0‑10 scale and weighting them appropriately.

Practical outcomes include reduced fault rates (from 0.57% to ~0.2%), 100% quality gate activation, full coverage of core system stress testing, and a long‑term governance mechanism that caps annual complexity growth at 10% and triggers alerts when thresholds are exceeded.

The article concludes with future plans to integrate AI for predictive analysis, enhance training, and expand automated fault‑drill exercises.