JD Mu Media Data Platform

Introduction

JD Mu Media Data Platform is the data marketplace of the Digital Marketing Ecosystem department, integrating the group's internal and external offline unique data capabilities. The goal is to drive offline marketing business development with data intelligence at its core and empower advertisers.

Data Platform Capabilities

• Precise Audience Targeting
• Panoramic Area Insights
• Smart Media Resource Selection
• Industry Analysis Report
• Federated Modeling Capability

Supports Upper-layer Applications

• Business Site Selection
• Customer Insights
• Outlet Recommendation
• Precise Customer Acquisition
• Success Case - Star Map
• Success Case - M+
• Moly Media Index

Job Content

Data Access

• External sources include the purchase of POI and AOI data, deep industry data crawling for POI, purchasing data such as offline trajectories and profiles from telecom operators, and accessing data collected from IoT devices.
• Internal sources include accessing group labels, traffic behavior data, ID integration, etc.
• Challenge: Develop a comprehensive and reasonable data acquisition plan based on different data characteristics, compliance, and cost constraints.

Data Governance

• Merge and de-duplicate multiple POI sources, define standard address levels, and extract and integrate deep POI information.
• Online behavior data model: Depict user online behaviors towards products and stores, such as browsing, following, adding to cart, and placing orders.
• Offline behavior data model: Describe users' offline visit locations, trajectories, time periods, and frequency.
  Media data model: Standardize offline media, link media outlet POI, scene classification, etc.
• Challenge: Integration and refinement of multi-source heterogeneous data.

Data Mining

• Traffic estimation model: Use operator data and IoT data as positive samples, and use POI data, reporting data, user profiles, statistical yearbook data, etc., as feature factors. Estimate POI traffic based on DNN. The estimation accuracy reaches the community level and covers the whole country.
• POI engine: Extract and complete address elements from unstructured addresses using named entity recognition and address trees, and use various innovative methods to mine AOI.
• Offline behavior tags: Based on proprietary reporting points, mine 17 user high-frequency offline behavior tags.

Data Services

• POI-related: Address-POI matching, POI query, multi-dimensional POI statistical analysis, POI traffic, TA concentration.
• Crowd-related: ID mapping, crowd selection and construction, profiling, crossover analysis, TA concentration.
  Media-related: Media query, media statistical analysis.
• IoT data access: Define interfaces for store creation, device binding, and data reporting.
• Highlight: The flexibility and versatility of the interfaces.

Work Achievements

• Completed the construction of the data platform, becoming one of the core capabilities of the department.
  Covered 70 million+ POI, 38 million+ effective traffic POI, a total of 3 billion+ IDs, covering 1 billion+ users, 62,000+ labels, and 10 million+ screens; covering 4000+ categories, 60+ industries, and 5 major offline scenes.
• Provided a one-stop data interface for advertising marketing, becoming the data capability cornerstone for upper-level applications like Star Map (DMP), Mu Media Intelligent Investment (DSP), and M+ (advertising monitoring).
• Became the main data source for industry analysis reports and external federated modeling collaborations.