Improving Efficiency through “No‑Test” Operations in a Trading Platform

In the third part of the trading platform quality operation series, the focus shifts to "no‑test" operations, the only discussion on efficiency improvement within the series.

Background: In 2019 the platform launched nearly 2,000 demands, each QA handling over five parallel projects, with effort ranging from fractions of a person‑day to dozens, highlighting the fragmented and numerous nature of the work.

Existing process includes demand review, technical review, case review, CR, and ShowCase before testing, which compresses testing schedules and leads to efficiency and manpower issues, as well as a limited quality awareness among developers.

Goals: (1) Improve product delivery efficiency; (2) Enhance developers’ quality awareness.

"No‑test" Specification: Two levels – Level 1 requires no QA involvement; Level 2 requires QA review. Scope is small, non‑core, simple‑logic development. Process: developers submit a no‑test request, automated regression runs, QA reviews, then release.

Operation – White‑list: Initially few users, so a dynamic white‑list of high‑quality developers was created, refreshed based on project count and issue count, encouraging proactive no‑test requests and leading by example.

Operation – Data‑driven: As white‑list usage declined, the approach shifted to data‑driven operation, automating data collection, visualizing metrics such as no‑test project ratio and issues, and using automated cleaning and configurable dashboards.

Results: No‑test project ratio stayed above 40%; 2020 saved 170 QA person‑days; developers’ quality awareness improved, making them more cautious about no‑test submissions.

Iterative "Small‑Urgent" Policy: Defines small, temporary, urgent demands that qualify for no‑test, addressing rapid delivery needs and reducing QA bottlenecks. Introduces responsibility zones and on‑call QA to handle urgent cases without lengthy approvals.

Effect: Weekly delivery rate of no‑test projects rose above 90%.

Conclusion: No‑test is not merely removing QA but requires periodic review and continuous improvement; while it boosts efficiency, maintaining quality through release checks and developer awareness remains essential.