OneData Methodology: Building a Unified Data Warehouse Architecture and Governance Framework

Background : Rapid business growth and frequent cross‑departmental iterations have led to serious data‑quality problems in the existing data warehouse. Lack of unified standards, insufficient data‑quality monitoring, scattered business knowledge, and unclear data‑layer responsibilities hinder data governance.

Goal : Based on the current big‑data platform, adopt the mature OneData methodology to construct a reasonable data‑system architecture, data standards, model standards, and development patterns, thereby ensuring fast data support for evolving business and driving business development.

OneData Exploration

Industry Experience : Alibaba’s OneData standard (see Figure‑1) provides a reference for data‑norms, model design, ETL standards, and supporting tools.

Our Thinking : We analyze Alibaba’s OneData and identify three characteristics (uniformity, uniqueness, standardization) and three effects (high scalability, strong reusability, low cost). Based on these, we define our own OneData theory and practice.

Core Ideas : Avoid duplicate construction and metric redundancy across design, development, deployment, and usage, ensuring consistent data definitions and enabling a full‑link data asset chain, standardized data output, and a unified public data layer.

Core Characteristics :

Three characteristics: uniformity, uniqueness, standardization.

Three effects: high scalability, strong reusability, low cost.

Strategy – Unified Intake

We propose two unified methods: unified business intake and unified design intake. This includes building a global knowledge base, consolidating business knowledge, and aligning application documentation to a single source.

Strategy – Unified Output

To guarantee data quality and promote data usage, we define a unified data‑output strategy, covering delivery standardization and data‑asset management.

Delivery Standardization : Five key properties (accuracy, consistency, uniqueness, timeliness, stability) are illustrated in Figure‑12.

Data‑Asset Management : Using the internal “Origin Data Platform”, we achieve unified metric management, unified dimension management, and a single export point for dimension and metric metadata (see Figure‑13).

Model Design and Review Responsibilities are shown in Figure‑10, outlining the duties of model designers and auditors.

Implementation Details

Unified Business Intake : Build a global knowledge base to ensure consistent business understanding.

Unified Design Intake : Define model layers (ODS → DWD → DWT → DWA → APP) and enforce clear layer responsibilities (see Figure‑4 and Figure‑5). Model layers are divided into four levels and data flow is standardized.

Topic Division : Two types of topics – business‑oriented and analysis‑oriented – are introduced to reduce abstraction and development cost.

Naming Conventions :

Table naming rule:

表名称 = 类型 + 业务主题 + 子主题 + 表含义 + 存储格式 + 更新频率 +结尾

(see Figure‑6).

Metric naming follows a structured pattern with base roots, business modifiers, date modifiers, and aggregation modifiers (Figures‑7 to‑9).

Cleaning Rules : A set of 24 data‑cleaning standards is provided (Figure‑11).

Unified Application Intake : Refactor the application support process to avoid siloed development and ensure a one‑to‑one mapping between applications and documentation (Figure‑11).

Unified Data Export : Define delivery standards and data‑asset management to ensure high‑quality, reusable data outputs.

Results

Process improvements, a panoramic data‑warehouse view, and an asset‑management list are presented (Figures‑14 to‑16). Comparative analysis shows significant value gains after OneData adoption (Figures‑17).

Conclusion and Outlook : The OneData framework provides a stable, reliable data‑warehouse foundation, ensuring data quality and supporting business decisions. Future work includes real‑time data‑warehouse integration, expansion to new business domains, and the development of an enterprise‑level One Entity data platform (Data‑as‑a‑Service).