Intelligent Supply‑Demand Matching, Assistants, and Content Operations on Alibaba's 1688 B2B Platform

Background

In the Taobao/ Tmall shopping experience, users mainly see front‑end recommendations, but the massive behind‑the‑scenes work is rarely understood. This talk introduces the intelligent B2B (ToB) operational scenario of 1688, focusing on efficiency, functionality, and process automation.

Three Main Topics

Intelligent Supply‑Demand Matching

Intelligent Assistant

Intelligent Content Operations

Intelligent Supply‑Demand Matching

The matching problem is split into supply side and marketing side. The supply side determines which goods to sell and which factories or merchants to source from. Market definition follows four layers: discovery, analysis, evaluation, and operation, involving indicator prediction, sales forecasting, pricing, and emerging market detection.

Example: for beverages, categories such as carbonated, water, and tea are defined; water further splits into packaged and pure water, both showing growth, indicating a strong market.

Market data maps product graphs and user queries onto CPV (Category‑Product‑Value) attributes, enabling feature clustering and label aggregation to build a market model that can predict the market for unseen products.

Market analysis uses metrics such as buyer behavior, transaction volume, traffic, and supply. Models like GBDT and W&D predict outcomes, and the Boston matrix classifies markets into mature, premium, risky, and emerging, guiding tailored operational strategies.

The overall framework shows how the matching system outputs tasks to the operation platform, such as supplement orders for merchants and marketing plans for operators, revealing a blue‑ocean market.

Intelligent Assistant

The assistant workflow includes three steps: insight & customer deepening, linking market demand with product/merchant graphs, and connecting product and merchant growth paths.

Insight & customer deepening – mining multi‑source data.

Link market demand – integrating market definition, analysis data, and product/merchant graphs.

Bridge product‑merchant growth – delivering strategies to merchants.

Strategy examples: revenue‑driving product operations, supply‑side strategies based on downstream market insights, and membership or traffic‑analysis benefits for merchants.

Delivery channels include merchant workbench, DingTalk, Qianji Assistant, and Alibaba Communication.

Use case: the Qianji Assistant distributes supply strategies and demand orders to operators, who generate SOPs with explainable models.

Process chain: problem discovery → big‑data analysis of merchant onboarding capability → solution attribution → strategy distribution → SOP generation with interpretable factors.

Intelligent Content Operations

Human‑machine collaboration reduces manual copywriting costs. Three sub‑tasks:

AI‑generated product descriptions and recommendation reasons.

AI‑generated short titles for personalized displays.

AI‑generated tags that reconcile buyer‑focused and seller‑focused attributes.

After deployment, conversion improved by 10%.

Research example: "Automatic Generation of Pattern‑controlled Product Description in E‑commerce" (WWW'19) uses a coordinate encoder for diversity, an attention‑based decoder for de‑duplication, and incorporates style classification into the loss function.

Summary

The talk covered three pillars—intelligent supply‑demand matching, intelligent assistants, and intelligent content operations—along with related research on product, venue, and factory knowledge graphs, and explainable human‑machine collaboration.