Data Model Governance Practices at Taobao (Alibaba)

Background and Problems Taobao (now Taobao Group) faced rapid data growth, increasing scale, stability issues, and declining consumption efficiency, prompting the need for model governance. Key challenges included data volume explosion, invalid data, high proportion of unassigned tables, and non‑standard usage by non‑data‑research roles.

Solution Overview The solution is built on four "1" capabilities: a standardized methodology, an evaluation system, incremental control, and stock governance. Core strategies focus on improving high‑quality data supply, enhancing consumption efficiency, and controlling model complexity through scale reduction and consumption relationship simplification.

Model Governance Practices Specific practices include:

Automated invalid table decommissioning with >95% identification accuracy and safe rollback mechanisms, reducing invalid tables to under 5%.

Source‑to‑target (ODS) duplicate import governance with >99% detection accuracy, business ownership attribution, and rule‑based import control.

Data handover automation that triggers evaluation and governance steps during role changes, ensuring both data and business documentation are transferred.

Public layer specialized operations to increase reuse and coverage, employing data albums for discovery and targeted outreach, achieving significant coverage improvements.

Incremental control by embedding governance rules into development tools, providing real‑time recommendations and checks.

Productization through intelligent modeling, data maps, and a governance center that supports one‑click migration, automated code generation, and health assessment.

Future Planning Planned initiatives aim to further enhance supply‑side efficiency, enforce architectural standards, integrate evaluation and governance, and leverage emerging technologies such as proactive metadata, large language models, automatic code generation, and automated lineage switching.

Q&A Highlights The Q&A covers topics such as consumption relationships, elegant decommissioning, duplicate import detection, balancing agile data usage with governance, visual modeling capabilities beyond ER diagrams, large‑model assistance for similarity detection, one‑click migration mechanics, coordination of overlapping data products, technical challenges and highlights of model governance, and DataWorks trial availability.

Overall, the presentation shares practical insights, metrics, and tool integrations that illustrate how systematic data model governance can improve data quality, reduce costs, and accelerate business value in a large‑scale e‑commerce environment.