Mastering Data Asset Management: From Inventory to Value Realization

01 Data Asset Inventory

Data asset management consists of three parts: data inventory, data governance, and data value realization. First, identify what data exists (inventory), transform it into valuable data assets (governance), and finally use it to drive business outcomes.

Data sources include internal raw data (business, management, IT/OA, monitoring) and external data (third‑party, web‑crawled). After data exchange, the data passes through an ODS (operational data store) layer that aligns with source systems and may perform standardization/cleaning. The warehouse layer builds thematic subjects (e.g., product, customer) and performs processing such as denormalization and standardization. Typical warehouse layers are source, standard, summary, metric, and data‑mart layers. Data marts export processed data for high‑concurrency queries in specific applications. Structured data, together with semi‑structured data (e.g., logs) and unstructured data (e.g., audio, video, email), form a big‑data platform.

2. Data Classification and Grading

After data mapping, clarify the distribution of confidential and sensitive data. Data classification groups data by source, content, and usage. Data grading assigns sensitivity levels (public, internal, sensitive) based on business value, sensitivity, and impact.

3. Data Responsibility

Once data is classified and graded, confirm data ownership to support permission design, governance, and operations.

4. Data Permission Management

Permission design focuses on access control and protection of sensitive information (e.g., ID number masking), linking data grading and responsibility.

02 Data Governance

1. Metadata

Metadata describes data itself. Business metadata defines meaning, rules, and indicators; technical metadata records storage location, model, field types, ETL scripts, etc.; management metadata captures responsible departments, owners, processes, and versions.

Business metadata: definitions, rules, indicators. Technical metadata: storage, model, field details, ETL/SQL scripts. Management metadata: organizational ownership and processes.

With metadata, we can trace data lineage, analyze relationships, and assess data hot‑cold patterns.

2. Data Standards

Standards define formats, naming conventions, and value rules to ensure consistent data exchange and avoid ambiguity.

Metadata must align with standards—for example, a growth‑rate metric should specify whether it is year‑over‑year or month‑over‑month, and field values should follow a unified coding scheme.

3. Data Quality

Data standards support quality. Common quality issues stem from entry errors, faulty code logic, or data loss during transfer; they are mitigated by validation at each processing stage.

4. Master Data

Master data are high‑value, highly shared, and relatively stable entities such as customers, products, and personnel. Challenges include multiple entry points, duplication, inconsistency, and data silos.

① High value: core business entities. ② High sharing: used across departments and systems. ③ Relative stability: low change frequency compared with transactional data.

Management approaches: (1) dedicated systems (e.g., CRM for customers), (2) centralized MDM platforms synchronizing data from various sources, (3) warehouse‑level themes, though the latter is less effective.

5. Data Security

Security spans the entire data lifecycle. Storage security relies on physical safeguards and secure hardware; transmission security uses encryption and network controls; usage security enforces access, download, sharing, and disposal policies.

03 Data Value Realization

With clean, governed data, value can be delivered through three service types:

Basic data services : query, multi‑dimensional analysis, funnel and path analysis via SQL.

Tag profiling services : customer profiling, precise marketing, using tags and models such as RFM.

Algorithmic model services : recommendation, risk control, and other domain‑specific AI models exposed as online APIs.