Building a Complete Retail Industry Tagging System for Data‑Driven Operations

Guide: This article provides a step‑by‑step tutorial on constructing a complete retail‑industry tag taxonomy.

Background D‑E‑Commerce is a retail e‑commerce company that integrates online transactions, offline logistics, finance, and community into a unified ecosystem, offering a leading all‑category one‑stop platform in its segment.

In recent years, the company’s core business has hit a bottleneck; the fixed product display model no longer meets market demand, leading to issues such as labor‑intensive ad placement and marketing maintenance, lack of precise profiling and data support for marketing, high marketing costs with low ROI, and abundant business data that cannot be transformed into valuable assets.

After the CEO studied data‑mid‑platform concepts, the company quickly formulated a strategic plan to build a retail data middle platform: collecting user behavior via event tracking, standardizing data entry across channels, and establishing a complete retail industry tag taxonomy to support data‑driven operations and personalized recommendation scenarios, thereby improving decision‑making, user stickiness, and commercial conversion efficiency.

The project team’s data middle platform follows a generic architecture but incorporates industry‑specific concepts such as “person‑goods‑scene” and D‑E‑Commerce’s unique requirements, resulting in a retail data asset system .

The architecture consists of three layers—platform foundation, core assets, and upper‑level applications—mirroring a typical data‑mid‑platform structure, with industry‑specific implementations in the asset and service layers (see Figure 2).

In the backend, the tag taxonomy covers core objects like Consumers (person) , Merchants (person) , Shops (entity) , Products (entity) , and Marketing Activities (relationship) . Each object includes multiple first‑level categories (e.g., basic attributes, interests, behavior, location, credit for consumers) totaling hundreds of tags.

These backend tags can be exposed as data‑service interfaces or interactive applications for business users, forming the frontend tag structure (see Figure 3).

For the frontend, D‑E‑Commerce focuses on insight analysis and “one‑to‑many” personalization: insight analysis includes merchant profiling, consumer profiling, product perspective, and activity effectiveness; the “one‑to‑many” scenario leverages consumer and product tags to deliver precise personalized recommendations.

- END -