Efficiency Measurement and Engineering Culture: From Farming to Industrialization

In a 2021 Tencent CI Day technical management salon, the speaker argues that efficiency measurement is indispensable and that a strong engineering culture can only develop after engineers are in place.

The presentation frames software development as a modern craft where many skilled workers produce components for a continuously running machine, highlighting the need for systematic, scientific, and quantifiable engineering practices.

It contrasts the old "farming" R&D model with an industrialized approach, defining industrialization as letting machines do what they excel at and humans focus on their strengths. High automation and clear responsibility lead to higher personal efficiency and a sense of engineering superiority.

However, industrialization requires a high degree of continuous delivery capability; otherwise, measurement efforts become costly and may hinder other work.

The speaker outlines the steps an engineer typically follows: discussing and analyzing requirements, writing code (including automated tests), submitting CLs, handling review comments, fixing pipeline failures, and analyzing feature performance data.

Benefits of this approach include higher automation, focus on high‑value tasks, clear accountability, and improved personal efficiency, but the cost of measurement can be high due to the labor required to collect, analyze, and disseminate data.

Before measuring, three questions must be answered: what to measure, why and what results are expected, and what actions will follow the results.

Measurement should be avoided when the team cannot act on the results, when results will become obsolete quickly, when metrics serve vanity purposes, or when metrics are insufficiently precise.

Key quotes from Peter Drucker are presented: efficiency is "doing things right" while effectiveness is "doing the right things," and when the two conflict, focus on effectiveness first.

The talk concludes that improvement should prioritize external, outcome‑oriented metrics (e.g., CLCT as a north‑star metric) and maintain a global view of process metric interrelations, while recognizing that cultural change follows the establishment of engineering practices.