Improving Engineering Capability in Silicon Valley Software Teams: Four Key Practices

Two years ago, the author wrote an unpublished piece describing a common software product development model used by well‑known Silicon Valley internet companies, focusing on the notion of "engineering capability" as a concrete expression of a team's technical competence.

Definition of engineering capability : It is the ability of a group with diverse expertise to quickly produce high‑quality products through systematic collaboration aided by auxiliary tools. This capability can be categorized as short‑term (often achieved by extending work hours or borrowing staff) or sustainable (gained by improving tools, skills, and processes while keeping work time and headcount constant).

The author emphasizes that engineering productivity is the sum of all engineers' individual contributions, likened to a chemical reaction where larger teams increase entropy and energy loss.

Four key practices to boost engineering capability :

1. Small batch (lean "single‑piece flow"): Even large, month‑long requirements are broken into many small tasks and merged into the main codebase frequently.

2. Standardization : Standardize behaviors and processes so engineers naturally follow correct procedures, minimizing exceptional workflows that disrupt production rhythm.

3. Simplification : Build on standardization by further reducing management complexity and unnecessary communication, making collaboration more direct.

4. Automation : With simplification in place, develop advanced auxiliary tools that further raise the team's overall engineering capability.

To achieve small batch , teams adopt continuous experimentation, quickly forming hypotheses for high‑risk features and validating them in short cycles, as illustrated by the Etsy example of testing search result quantity within a day.

For standardization , the article cites Google's code‑submission workflow, which features a highly uniform process with few exceptions, reducing special‑case coordination.

Regarding simplification , the author notes that after standardizing, processes can be further optimized; for instance, Google's code‑review tool shows test coverage directly, eliminating the need to switch contexts.

Automation is exemplified by Google's open‑source C++ build tool Bazel, which enforces a unified build file format and single‑dependency rule, dramatically lowering software complexity, improving compile speed, and enabling selective automated testing.

All three practices lead to simpler process rules, making auxiliary production tools more effective; when every engineer follows the same workflow, overall productivity rises, even if each tool saves only ten minutes per engineer per day.

In conclusion, as team size grows, problem complexity increases exponentially, so adopting engineering‑centric and systematic thinking for every "exceptional case" is essential.