Big Data 16 min read

NetEase Yanxuan Tag and Profile Platform Construction Practice

This article presents NetEase Yanxuan's comprehensive approach to building a tag and profile middle‑platform, detailing the concepts of tags and profiles, their business value, the platform's architecture and functional modules, and the methodologies for constructing tag systems and applying profiles to drive data‑driven operations.

DataFunTalk
DataFunTalk
DataFunTalk
NetEase Yanxuan Tag and Profile Platform Construction Practice

Introduction

Amid the wave of digital transformation, enterprises increasingly focus on accumulating and applying their data assets. Profiles, as an important form of data asset, have attracted growing attention. NetEase Yanxuan, a self‑operated e‑commerce platform with long business chains and many scenarios, involves core business entities such as users, products, suppliers, and channels. Using profiles to gain insights into these entities is crucial for refined operations. Based on this background, Yanxuan built an industry‑specific tag and profile middle‑platform. This article introduces the tools and methodology of Yanxuan's tag and profile construction.

What Are Tags and Profiles

Tag: A description of a specific dimension of a business entity, representing a business‑oriented data organization form. For example, a restaurant labeled as "must‑eat" on a review site or a movie's rating on Douban.

Profile: A collection of multiple tags that together describe an entity across several dimensions, such as a game character’s strength, intelligence, and agility attributes.

Value of Tags and Profiles

Tags: Provide information, enabling more scientific and accurate decision‑making; they are business‑oriented, turning data into actionable insights.

Profiles: As a set of tags, they deliver multi‑dimensional orthogonal information, helping to more accurately and vividly understand entities.

Why Build a Tag and Profile Platform

Yanxuan builds the platform to achieve two goals: (1) address common user‑value needs, and (2) accelerate data assetization and value realization (commercial value).

1. Solving Common Needs: While many industries focus on user tags and user profiles for precise marketing and consumer insight, Yanxuan’s scenarios are broader—consumer insight, supplier sourcing, hot‑product creation—covering entities like users, products, suppliers, and channels.

2. Accelerating Data Assetization: Tags, as business‑oriented data structures, enable rapid creation, comprehensive management, and convenient usage, thereby speeding up the transformation of data into value.

Tag and Profile Platform Tools

1. Product Overview: The platform provides end‑to‑end data‑driven capabilities from data management, tag extraction, to insight analysis.

2. Functional Features:

Data Management: Unified management of business entities and their global data, supporting Hive, Kudu, ES, and HBase storage engines.

Tag Extraction: Supports basic tags (any data type, custom rules) and composite tags (boolean, predefined rules). Extraction methods include direct field association and SQL‑based custom calculations.

Insight Analysis: Enables entity selection via tags, manual upload, or group processing, and provides profile analysis such as single‑tag distribution, multi‑tag cross analysis, group comparison, TGI analysis, and time‑dimension comparison.

Tag System Construction Methodology

Bottom‑Up Approach: Initially driven by product‑technology teams to bootstrap core scenarios, then gradually handed over to business teams for organic growth. Example: building a supplier tag system by first researching core supplier‑sourcing scenarios.

Top‑Down Approach: Derive tags from business processes and entity lifecycles aligned with commercial goals. Example: reducing supply‑chain costs leads to inventory‑related tags such as seasonal product tags and historical sales tags.

Profile Application

Profiles become actionable by comparing groups to discover significant features and translating them into concrete operational strategies. Common analysis methods include:

In‑group feature distribution (e.g., gender, city, preference).

Between‑group feature comparison using TGI to highlight distinctive traits.

Cross‑time dimension comparison to evaluate the impact of marketing tools.

Case studies: product profiling revealed that top‑selling items are priced in the 0‑40 range and favor home‑style designs; user profiling identified super‑member traits (female, Shanghai resident, car owner, overseas fashion preference), guiding acquisition, activation, and retention strategies.

Conclusion

The article introduced the concepts of tags and profiles, their value, Yanxuan’s motivations for building a middle‑platform, the platform’s functional modules, and the methodologies for tag system construction (bottom‑up and top‑down) and profile application (various comparative analyses), supplemented with practical Yanxuan case studies. It also noted challenges such as comprehensive data collection, tag accuracy, and system stability.

big databusiness intelligencedata platformUser Profilingtaggingdata asset
DataFunTalk
Written by

DataFunTalk

Dedicated to sharing and discussing big data and AI technology applications, aiming to empower a million data scientists. Regularly hosts live tech talks and curates articles on big data, recommendation/search algorithms, advertising algorithms, NLP, intelligent risk control, autonomous driving, and machine learning/deep learning.

0 followers
Reader feedback

How this landed with the community

login Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.