Systematic Modeling for Delivery Data Governance

The rapid growth of the digital economy brings new opportunities and challenges for enterprises. Effective data governance—breaking data silos, unlocking business value, and ensuring data security—has become a hot topic. This article shares Meituan Delivery’s experience in building a unified delivery data "foundation" through systematic modeling.

Preface : The article introduces the construction and practice of a delivery data foundation that bridges data definition, model design, and data production, aiming to provide references for data‑governance practitioners.

What is Systematic Modeling? It is based on dimensional modeling theory and a pre‑governance mindset, driving metadata throughout the process. It unifies three stages—data demand, model design, and data production—so that business definitions, logical models, and physical implementations are consistent.

Figure 1: Overview of Systematic Modeling

The approach emphasizes two unifications: (1) aligning data demand with model design, and (2) aligning model design with physical implementation, preventing “silo‑style” development.

Why Systematic Modeling? Previously, delivery data construction suffered from fragmented demand management, model design, and development, leading to inconsistent data architecture, duplicated effort, and low data trust. Systematic modeling addresses these issues by standardizing data construction and eliminating siloed development.

3.1 It enables effective management of data architecture from the source, removing siloed development.

3.2 Standardized metadata reduces business confusion when retrieving and understanding data.

How to Perform Systematic Modeling The process starts at the source, linking data definition, model design, and ETL development to achieve “design‑as‑development”. High‑level modeling defines business indicators and dimensions, which then drive detailed physical modeling.

Figure 2: Systematic Modeling Workflow

High‑Level Model Design involves extracting atomic and calculated indicators, assigning them to business processes, and designing consistent dimensions across processes.

Detailed Model Design translates high‑level models into physical tables, respecting data‑warehouse layers (B3, B2, B1). It uses metadata to generate DDL/DML automatically, ensuring consistency and reducing manual effort.

Figure 3: Detailed Model Design Process

Pre‑Release Checkpoints include consistency verification of identical indicators from different sources, alignment of business definitions with implementation, compliance checks (e.g., primary key uniqueness), and impact analysis of downstream tasks.

Summary Systematic modeling has improved data quality and efficiency for Meituan Delivery. It unifies data definition, model design, and production, supports self‑service analytics, and transforms data governance from a “post‑hoc” activity to a proactive, metadata‑driven practice.

Authors: Wang Peng, Xin Xing, Xiao Fei – Meituan Delivery Data Team.