Data Point Governance and Quality Management in ZhiZhu QA Process

1. Background

At ZhiZhu, every recycled device goes through a strict official inspection line, and each operation, detection content, device information, and execution time are reported as data‑point events to a digital asset system. These data points are crucial for analyzing and managing the inspection process, directly affecting timeliness and accuracy.

2. Problem

Before 2024, data‑point testing ended after client development self‑testing, lacking QA involvement and product acceptance, leading to inaccurate, missed, or false reports that hindered business analysis and decision‑making. The team decided to establish a digital data‑point quality management mechanism driven by QA.

3. Data‑Point Governance

3.1 Phase One – Manual Execution

Initially, quality was ensured manually. In 2024, QA took ownership of data‑point testing, defining a digital quality management process that aligns product, development, and QA roles, achieving 100% QA coverage of data‑point tests.

3.2 Phase Two – System Monitoring

Manual methods could not scale with the growing number of data points. A digital governance platform was built with capabilities such as real‑time validation, user‑behavior trees, and dashboards.

3.2.1 Real‑Time Validation

Testing many data points across multiple clients is resource‑intensive. Many parameters follow predictable rules (e.g., start time < end time, start time < current time, end time ≤ current time). The team automated these checks by consuming binlog‑derived Kafka messages and applying rule‑based validation.

......

The validation runs continuously in both offline and online environments, automatically flagging violations and notifying responsible parties, improving test efficiency and accuracy while also catching historical issues.

3.2.2 User‑Behavior Tree

Raw JSON data points are cumbersome. The team transformed them into a visual “user‑behavior tree” that displays engineers' actions and key information, aiding rapid issue diagnosis and serving as a clue for product, development, and operations teams.

3.2.3 Data‑Point Dashboard

Two dashboards were created using QuickBi:

Data‑point reporting dashboard: monitors overall reporting trends and dimensions such as category, client, and version, enabling timely alerts and root‑cause analysis.

Real‑time interception dashboard: highlights abnormal reports, allowing quick analysis of client‑ or version‑specific quality and detailed breakdowns.

3.2.4 Bi‑Weekly Quality Analysis Report

Every two weeks, a detailed quality report is produced, summarizing data‑point metrics and correlating them with inspection business data, driving continuous feedback and remediation.

4. Summary

Over six months, the combined manual and automated governance reduced new abnormal data points by 83% (to only one case) and increased testing efficiency by ~40%, providing a sustainable quality assurance framework for future data‑point initiatives.

5. Future Outlook

Make real‑time validation rules configurable via dynamic services like Apollo.

Enrich validation rules and integrate business data for more accurate anomaly detection.

Build a mapping between client features and data points to assess impact of code changes and guide testing.