Systematic Data Governance Framework and Practices at Meituan Accommodation

Introduction

The Meituan Accommodation data governance team reflects on years of data‑warehouse construction and governance experience, describing a transition from passive, single‑point governance to proactive, systematic, and automated governance.

Background

Since the product launch in 2014, the accommodation business has moved from rapid expansion to a stable, refined‑operation phase, raising requirements for data cost, efficiency, security, and value. Multiple business lines (accommodation, tickets, vacation) have differing data‑center lifecycles, making a unified, standardized, and automated governance approach essential.

Governance System Thinking

To address fragmented practices—​inconsistent cognition, duplicated work, unclear boundaries, non‑MECE governance, and low efficiency—the team defined a comprehensive governance system that integrates management, standards, capability, execution, evaluation, and vision layers.

3.1 What Is a Systematic Data‑Governance Framework?

It combines a management system, method system, evaluation system, standard system, and tool system to continuously support data‑governance implementation, analogous to a retailer’s sales, product, supply‑chain, and logistics systems.

3.2 How Systematic Governance Solves Existing Problems

Methodology: Top‑down framework design defines scope, roles, goals, methods, and tools before tackling specific issues.

Technical Means: A unified metadata and metric system evaluates data warehouses and applications, backed by governance tools.

Operational Strategy: Issues are prioritized by impact and benefit from both manager and owner perspectives.

3.3 Framework Construction

The framework is built around the team’s governance goals and iteratively refined through feedback.

Management Layer: Legislation, organizational processes, responsibilities, and incentives.

Standard Layer: Development standards, SOPs, and technical specifications.

Capability Layer: Metadata‑driven measurement, tool‑based detection, and systematic remediation.

Execution Layer: Action plans covering pre‑constraints, mid‑monitoring, and post‑governance across seven governance domains.

Evaluation Layer: Health‑score metrics, specialized reports, and ROI assessment.

Vision: Long‑term governance objectives.

Practical Implementation

4.1 Standardization

Standardization turns policies into concrete standards and SOPs, addressing three pain points: missing process norms, poor implementation conditions, and ad‑hoc construction.

Process norms are codified for data‑development, testing, and health‑assessment.

Documentation is centralized in a knowledge‑center with unified access control.

Tools such as the internal "八卦炉" ETL testing platform enforce testing SOPs.

4.2 Digitization

Digitization builds a metadata warehouse and metric system covering lifecycle, team‑goal, and problem‑object perspectives. The team created over 112 indicators (57 technical, 36 demand, 19 fault) to monitor quality, safety, efficiency, cost, and value.

Asset statistics: >3000 Hive tables, >2000 ETL tasks, ~100 tasks per engineer.

Asset‑level scoring uses the Analytic Hierarchy Process (AHP) to weight factors such as downstream type, downstream count, usage heat, and link depth, producing L1‑L5 grades.

4.3 Systematization

The "Data Hundred Products – Governance Center" platform integrates asset management, problem analysis, automated governance, process tracking, and result evaluation in a one‑stop portal.

Asset Panorama: Dashboard, catalog, and personal asset views for managers and data engineers.

Management Center: Core‑metric dashboards, demand management, fault management, and team‑operation modules.

Governance Center: Issue overview, quantitative analysis, SOP‑driven remediation, and progress monitoring.

Automation is realized through the SOP Automation Tool, which converts SOP documents into configurable UI components (rich text, tables, IFrames, single/multi‑choice controls) and exposes them via URLs for one‑click governance.

Governance Process

Discover & Set Goals: Identify problems from the data‑development perspective and define achievable targets.

Decompose & Metricize: Break problems into measurable indicators using metadata collection.

Design SOP & Tooling: Create SOPs, verify standards, and build tool‑supported solutions for pre‑, mid‑, and post‑governance.

Promote & Operate: Deploy solutions with role‑specific strategies, monitor impact, and avoid conflicts with user interests.

Summarize & Iterate: Capture lessons, refine the methodology, and explore smarter, more automated solutions.

Results & Outlook

The systematic framework has delivered noticeable improvements: unified standards across teams, higher data quality through automated testing, and consistent governance outcomes via SOP automation. Cost savings are estimated at several million RMB annually. Future work will focus on intelligent metadata services, AI‑assisted standard creation, and further automation of governance scenarios.