Meituan's Supply Chain System: Architecture, Challenges, and Automation

1. What is the Supply Chain

In the fast‑changing O2O battlefield, Meituan's success depends on its behind‑the‑scenes technology team – the Supply Chain (SCP). The system converts merchant agreements into electronic contracts, refines proposals into detailed, searchable product descriptions, and ensures legal and financial compliance before users can purchase them on the C‑end.

The production process involves multiple roles: business development (BD) initiates the deal, contracts are signed, details are entered into the business portal, headquarters audits the proposal, and the CMS packages the content for various C‑end channels, creating a complete project.

Beyond traditional group‑buying, Meituan expands into hotels, travel, food delivery, and breakfast, each with varying degrees of standardization, posing additional challenges for the supply‑chain system.

2. Challenges of the Supply Chain System

2.1 Complex, Fine‑Grained Structured Data

Different O2O categories require vastly different attribute sets (e.g., restaurant location, hotel room type, Wi‑Fi availability). To support both B‑end efficiency and C‑end searchability, Meituan abstracts reusable attributes into modules such as “purchase notice” and structures data for multiple channels.

2.2 Flexible and Varied Sales Modes

Hotels, for example, have numerous combinations of room type, breakfast inclusion, and holiday pricing. Recording each combination separately would explode the workload; therefore the system must support multiple sales modes from a single data entry.

2.3 Dynamic Category and Attribute Adjustments

New sub‑categories (e.g., Wi‑Fi in self‑service restaurants) require additional fields, and splitting existing categories (e.g., Chengdu vs. Chongqing hotpot) demands UI and storage changes. The supply‑chain must allow zero‑code configuration for such adjustments.

2.4 Configurable Audit Workflow

Different business channels have distinct audit requirements. The system must allow dynamic configuration of audit nodes, participants, and routing to adapt quickly to changing policies.

3. Addressing the Challenges

3.1 Building an O2O Service Model

Meituan abstracted a “product center” that maps merchant‑provided service units (e.g., big‑bed room, breakfast, Wi‑Fi) to sales units and then to C‑end products via SKU attributes (price, consumption rules, etc.). This model decouples business logic from implementation.

3.2 Event‑Driven Decoupling

A workflow engine was introduced to make audit processes configurable, allowing dynamic definition of nodes, participants, and routing. This also decouples data collection from business logic, enabling independent analysis of process metrics.

3.3 Automation Everywhere

3.3.1 Project 908

By segmenting the supply‑chain process and applying targeted automation (e.g., auto‑approval, auto‑writing), Meituan reduced per‑order cost by 90% and increased efficiency eightfold, a project internally named “908”.

3.3.2 Ongoing Automation

The automation mindset continues: workflow handles repeated audits, an attribute center enables rapid category/attribute changes without code, and configuration pages let business users adjust rules in minutes.

4. Achievements

4.1 Faster Delivery

For a typical discount purchase, the end‑to‑end supply‑chain development effort dropped from 30 person‑days to 5 person‑days.

4.2 Cost Reduction and Efficiency

After eliminating manual review and writing, the dedicated review team (nearly 100 people) was dissolved, and with a 1000% order volume increase, per‑order cost fell by over 90%.

5. Summary

From a technical perspective, the supply‑chain workflow starts with BD initiating a request, the backend renders a dynamic form (DF) based on channel and category, the Attribute Center (AC) validates the input, and the data is transformed for the Product Center. Gravity schedules audit tasks, changes are recorded in the Change Center, and after approval, the CMS assembles the final content using dynamic templates for various C‑end channels. The system follows an MVC pattern with Model (Product/Change Center), View (Dynamic Form & Template), and Controller (Attribute Center & Workflow), achieving high availability and automation.