Design and Implementation of an International Business Data Platform for JD.com's 618 Promotion

As JD International makes its first comprehensive appearance in the group's 618 promotion preparation, it operates across Southeast Asian e‑commerce sites (Thailand, Indonesia), newly added European and domestic businesses (MCA, B2B), as well as third‑party and main‑site sales channels. To provide real‑time operational guidance during the promotion, it needs to aggregate traffic, user, and sales data from all lines into a single management and planning system, confronting challenges related to international business traits, data distribution, personnel, current data construction, and technical architecture.

International Business Characteristics – The business spans multiple formats, countries, regions, languages, time zones, and currencies, each subject to distinct legal regulations, which creates significant data‑building challenges.

Data Distribution and Personnel – Six business lines are deployed in different data centers (e.g., Indonesia in Jakarta, Thailand in Hong Kong moving to Singapore, MCA in Amsterdam, B2B/main‑site/third‑party in domestic centers). Data processing and analysis responsibilities vary: some are handled by business analysts, others by data‑development teams, and some by staff who have quickly learned big‑data technologies.

Building the international data platform therefore requires a professional data team, a standardized data‑center, uniform metric definitions, and consistent calculation frequencies.

Data Status and Team Building – To meet the demand for a promotion‑specific dashboard, a "618 Promotion Dashboard Project" was created, identifying three dashboard scenarios (traffic, users, transactions) and eight common metrics. A "Virtual Data Development" group was formed, comprising product, data, and development experts from each line.

Dashboard System Architecture and Idea – The goal is to integrate all international business data, provide real‑time unified metrics, and display cumulative and daily trends for eight common indicators from June 1 to June 18 on a JDV‑based large‑screen dashboard.

Data Sources – The data originates from heterogeneous systems using ClickHouse, Hadoop, MySQL, Elasticsearch, with latency ranging from seconds to ten minutes and currencies including Thai Baht, Indonesian Rupiah, USD, and CNY.

Cross‑Region Integration – Two integration approaches are proposed: (1) each line offers an HTTP data service that the domestic aggregation service consumes, minimizing impact on existing pipelines; (2) define common data tables (e.g., product, merchant, order, marketing) for standardized processing and reporting.

Capability Reuse – To ensure cross‑device compatibility and efficient rendering, the front‑end leverages JDV's existing large‑screen resources and data orchestration capabilities, collaborating with the retail middle‑platform and data‑intelligence teams.

Architecture Plan

Short‑term:

Utilize the big‑data platform (Flink, Spark, JMQ, MySQL, ClickHouse) for computation, storage, analysis, and data pushing.

Build an international data aggregation layer accessed via internal HTTP, handling accumulation, currency conversion, and formatting, with caching for real‑time performance.

Expose aggregation services to JDV dashboards, which cache data locally in the browser.

Long‑term:

Establish a unified global data hub routing user, transaction, marketing, browsing, product, and inventory data to regional data centers.

Define international data standards for naming, storage size, retention, analysis frequency, and ensure proper data‑tracking for new services.

Construct an international data platform that integrates internal and external data, delivering insights for operational guidance and experience improvement.

Summary – The practice of multi‑source international data application demonstrates a flexible, component‑driven large‑screen dashboard system with built‑in permissions, reusable components, and efficient data interaction, significantly reducing duplicate development and enhancing performance. Ongoing enhancements aim to create a universal international data platform driven by business and customer needs, continuously improving usability and extensibility.