Unlocking Software R&D Efficiency: How Metrics Drive Better Development

Overview

In a live round‑table, authors of “The Authoritative Guide to Software R&D Efficiency” discuss how to apply measurement (度量) in enterprises, sharing insights from WeBank, Tencent, and Kuaishou.

Key Concepts – “Dao, Fa, Shu, Qi”

Dao – the direction: clarify what problems measurement should solve, such as showing results and processes.

Fa – the methods: adopt proven frameworks like GQM, value‑flow management, DORA, and avoid reinventing the wheel.

Shu – the practice: concrete scenarios (demand throughput, bug‑rate, code quality, production incident review) and how metrics guide behavior.

Qi – the tools: building or adopting platforms (e.g., DevLake) that integrate data from requirement, code, CI/CD, and deployment to provide self‑service dashboards.

Insights from Speakers

Yuwei emphasizes that metrics should guide correct R&D behavior, focusing on result and process visibility, and warns against using metrics for personnel evaluation.

Ren stresses “management precedes data”: define goals first, then collect data to support decisions; GQM helps structure this.

Xiong highlights that metrics can solve technical problems but not people problems, and advocates for a “divide‑and‑conquer” approach.

Zhang summarizes three dimensions: “see”, “analyze”, “guide & empower”, and warns against over‑precision.

Practical Recommendations

Start with clear goals (quality, efficiency, value, cost, capability), use GQM to derive questions and metrics, collect data automatically, keep the measurement framework simple, and iterate the platform based on feedback.

Adopt open‑source solutions like DevLake when possible, and complement them with custom extensions for specific toolchains.

Conclusion

Effective R&D measurement combines clear purpose, proven methods, concrete practices, and robust tooling to improve software delivery while avoiding common pitfalls such as metric misuse or one‑size‑fits‑all approaches.