Data Governance: Challenges, Framework, and Implementation Practices

The article begins by listing six major issues that data governance aims to solve: the disconnect between data development and governance, siloed "chimney" data development, lack of unified control across platforms, unquantified monitoring of governance processes, insufficient fine‑grained cost‑value management of data, and the absence of a closed‑loop governance lifecycle.

It then presents a three‑part outline: (1) what data governance solves, (2) the governance system, and (3) a brief discussion of implementation.

The governance system is described as a full‑link framework built on actual enterprise and industry scenarios, leveraging data‑governance products, processes, policies, and management to create an end‑to‑end workflow.

Implementation details include:

Overall tool solution: integrating governance into the entire data‑development lifecycle, establishing a unified platform for data assets, metadata, lineage, and visual monitoring.

Development‑governance integration: embedding governance standards, design, and modeling into data development, following the principle “design first, develop second, standardize, then govern.”

Standardized modeling: defining data models, metrics, and dictionaries based on national, industry, and enterprise standards to ensure consistent definitions and high‑quality assets.

Metadata asset governance: managing business, technical, and management metadata, publishing assets, and using ROI‑based health diagnostics for fine‑grained asset management.

Lake‑out governance: registering external data sources (e.g., MySQL, Oracle), collecting metadata, assessing governance readiness, and publishing assets for business use.

Lake‑in governance: similar registration, enrichment, approval, and publishing workflow for data stored within the data lake, with mechanisms for issue escalation and asset deprecation.

Governance policies: covering development standards (naming, modeling, scheduling), metric management (definition, calculation, lineage), data‑quality management (pre‑rules, monitoring, analysis, traceability), organizational structure (governance teams, administrators, specialists), and ongoing operation & knowledge‑capture.

The article concludes with a visual summary of the data‑asset closed‑loop, emphasizing problem discovery, targeted tooling, continuous monitoring, and operational activities such as governance contests and specialized projects.

Speaker: Fu Zheng, NetEase DataFun Big Data Product Expert, former senior trainer at Huawei, with extensive experience in data development, governance, and cross‑industry projects.