How Leading Logistics Companies Master Data Governance for Cost and Stability

Introduction

In August 2022, Wang Haihua, a big‑data expert and infrastructure lead at Huolala, organized the "Data Governance Practice" forum of the DataFun Summit 2022, inviting data‑governance specialists from Zhongtong, SF Express, and Huolala to share their experiences on governance systems, platform construction, and practical implementation.

Zhongtong Technology

Senior architect Xue Shimin presented "Zhongtong Data Governance Practice", discussing the drivers behind Zhongtong’s data governance and their implementation experience.

Drivers of Data Governance

Xue identified business pain points and national regulations as the main motivations for Zhongtong’s data‑governance initiatives.

Zhongtong aims to improve data governance, achieve data aggregation and connectivity, understand data‑asset status, ensure data security and compliance, and unlock business value.

Practical Experience

The practice covers three dimensions: data‑quality governance, data‑model governance, and metadata governance.

Data‑quality governance : Zhongtong uses a comprehensive monitoring process and platform to detect quality issues. Rules are applied at each processing layer (ODS/DW/DM/Application); detection points trigger checks, output results, and generate alerts for anomalies.

Business‑topic analysis drives the resolution of data‑quality problems.

Data‑model governance : Due to rapid business growth, fast project iteration, and inconsistent data construction, Zhongtong faced low model reuse, unstable timeliness, and rising resource costs. They addressed these by establishing standards, process controls, and model tiering to improve reuse.

Metadata governance : Challenges such as low reuse, unstable timeliness, and high resource costs were tackled through standardized definitions, controlled processes, and hierarchical model management.

SF Express Technology

SF Express data‑governance expert Wang Minjun delivered "Enterprise Data Governance Practice at SF Express", covering the evolution of the governance framework, the overall architecture, and practical experiences.

Evolution of Data Governance Framework

Before 2020: Business‑driven, fragmented capabilities across domains.

2020‑2021: Support business analysis and decision‑making, gradually improve governance capabilities.

2022 onward: Build an end‑to‑end, efficient governance system to support decisions and operational efficiency.

Data Governance Framework

The framework rests on platform tools and governance organization, supporting data production management, consumption management, and architecture management. Core tool capabilities include master‑data management, data‑quality management, and a data marketplace.

Policies such as master‑data standards, metric‑definition standards, and data‑security policies guide capability building.

SF Express Data Governance Practices

Four key success factors are identified: senior leadership support, operational organization (business + technology) for continuous governance, combined assessment and incentives, and phased implementation. Two critical elements for advancing governance are leadership mechanisms and the approach to governance entry points.

Master‑data governance follows four steps:

Identify master data and define owners.

Establish data standards.

Determine trusted source systems.

Monitor and improve data quality.

The identification of master data and owners is detailed through three perspectives—business, control, and technical—to exhaustively list company master data, followed by classification according to SF Express’s master‑data framework, and finally assigning owners and responsibilities.

Huolala

Huolala shared how platform construction and project practice achieve cost governance and stability assurance.

Data Governance Platform Construction

Data engineers Chen Yuan and Zhang Fang presented "Huolala Data Governance Platform Construction" covering data‑quality and cost‑governance platforms, including design ideas, technical architecture, and outcomes.

With growing data volume and deeper business demand for data value, Huolala faced rising storage costs and recurring quality issues. Productized capabilities were adopted to address these challenges.

The governance capability consists of three platforms: metadata management, data‑quality management, and data‑security management. The metadata platform provides model management, asset management, lineage analysis, and cost control, supporting both pre‑emptive constraints and post‑incident troubleshooting. The quality platform offers end‑to‑end detection, monitoring, and alerting; issues trigger alerts or task circuit‑breakers to prevent downstream impact. The backend uses stateless micro‑services with multi‑instance schedulers, ensuring stable detection. A hybrid engine routes over 80% of SQL to Presto, enabling P80 quality‑check tasks to finish within 5 seconds.

Currently, the quality platform covers 100 % of core tables, detecting over 300 quality issues per month and safeguarding data‑link stability.

Huolala’s metadata platform follows a “standard‑govern‑capability‑operate” principle, building a global data map, full‑link lineage, cost‑governance, model standardization, and asset management.

Cost governance leverages resource budgeting and asset‑measurement, establishing health scores and blacklists to continuously operate toward cost targets.

For storage governance, Huolala applied hot‑warm‑cold‑frozen tiering and lifecycle management. Frozen data (≈33 % of tables, unused for 90 days) was moved to cheaper storage, reducing storage growth to zero over eight months and cutting total storage cost by 54 %.

Stability Assurance in Data Governance

Data‑asset lead Li Renquan presented "Stability Assurance in Huolala’s Big‑Data Governance". He described the problem‑identification‑solution‑implementation‑validation workflow.

Facing continuous business growth and high IT costs, Huolala encountered rising data‑quality incidents and latency in core pipelines. He analyzed the full data lifecycle—generation, ingestion, processing, and service—to pinpoint stability‑impacting root causes, proposing pre‑emptive, monitoring, and post‑mortem measures.

The solution spans platform support, organizational guarantees, policy construction, and project execution, illustrated by Huolala’s overall data‑asset architecture.

Implementation details include proactive standards and processes, real‑time monitoring with on‑call alerts, and post‑incident reviews to shift from reactive to proactive problem solving.

Since Q4 2021, these measures have yielded measurable improvements, with future directions focusing on intelligent prediction algorithms for metric accuracy, smart alert strategies for on‑call satisfaction, and one‑click automated governance.

Author: Chen Yuan, Senior Big‑Data Engineer at Huolala, responsible for data‑governance platform and unified engine services.