Meituan Hotel & Travel Data Governance: Journey, Practices, and Future Directions

Background

Data governance has become a hot topic in recent years, especially for large internet companies. As Meituan's hotel‑travel business grew rapidly from 2014 to 2018, data volume and complexity exploded, exposing issues in data quality, cost, security, standardization and operational efficiency.

Key Problems Identified

Quality problems such as timeliness, accuracy, consistency, and logical alignment of metrics.

Rapidly rising storage and compute costs.

Security risks for user‑level data.

Lack of unified data standards across business lines.

Low efficiency in data development and management.

Current Data Situation

From 2017‑2018, production tasks doubled annually, and data growth exceeded two‑fold each year. Without governance, future task complexity and cost would become unsustainable.

Governance Practice

1. Standardization and Organizational Assurance

A full‑link data standard method was established covering data collection, warehouse development, metric management and lifecycle, supported by a cross‑functional Data Management Committee that unites technical and business teams.

2. Technical System

Automation is emphasized to reduce manual effort. Four major quality issues are tackled:

Warehouse comprehensiveness – standardized modeling and configuration tools.

Metric consistency – unified metric definitions and management.

Application consistency – unified data service interfaces.

Product entry consistency – a single entry for all data products.

Key tools include model development utilities, naming‑standard tools, and release‑rule monitoring, all integrated into a continuous validation pipeline.

3. Data Warehouse Modeling Standards

Modeling standards are divided into pre‑, mid‑ and post‑development phases, with automated configuration tools and rule‑based validation to ensure compliance.

4. Unified Metric Management System

Metrics are managed in three layers (physical tables, models, metrics) within a metadata system, providing consistent metric logic for all downstream applications.

5. Unified Data Service Platform

The platform offers pull‑based (high‑frequency) and push‑based (daily) data access, organized into import, storage, service, control, and interface layers to guarantee accuracy, monitoring and permission control.

6. Unified Product Entry

Data products are grouped into three categories—analysis/decision tools for managers, sales/operation products for business, and asset‑management products—ensuring users access consistent data through a single portal.

Overall Architecture

The three‑layer architecture (standardized modeling → unified metrics → unified services → unified entry) secures data quality while supporting organizational processes and monitoring.

Data Operation Efficiency

Three user problems are addressed: discoverability, understandability, and usability of data. Solutions include a data guide system, a Q&A bot, and knowledge‑base integration, reducing manual assistance to less than 20% of requests.

Data Cost Management

Compute cost reduction (invalid task governance, long‑task optimization, resource utilization, unified resource management).

Storage cost reduction (cold‑data governance, duplicate data handling, lifecycle management, compression).

Log collection cost reduction (downstream monitoring, reporting optimization, invalid point‑cut optimization).

Data Security

Security is handled through pre‑emptive encryption, mid‑process de‑identification, and post‑process auditing, following five principles: ciphertext transmission, latest‑decryption, minimal‑scope extraction, minimal‑authorization, and end‑to‑end audit.

Measurement Indicators

A five‑category metric system (quality, cost, security, usability, value) with daily and periodic monitoring tracks the health of data governance.

Governance Stages

Three stages are defined: passive (ad‑hoc fixes), proactive (systematic, long‑term planning), and automated (policy‑driven, AI‑assisted remediation). Meituan's hotel‑travel data governance is currently transitioning from proactive to automated, requiring further investment to achieve full automation.

Conclusion

The sharing concludes with thanks to the audience.