Private Domain Tag System Construction

The Specific Problems to Solve

In private domain operations, tag systems often face several key pain points. Ambiguous definitions: The same tag, like "high activity," has different meanings across teams (e.g., community active vs. order active), leading to unified usage chaos and poor decisions. Tag explosion: Blindly creating hundreds of tags without management results in campaigns still relying on bulk sends, making tags ineffective. Fragmented data sources: Tags from WeChat Work, mini-programs, and order systems misalign (e.g., inconsistent "membership level" across three systems), creating data silos. Lack of governance: Without clear owner or online/offline rules, tags degrade over time, becoming "rotten" assets that waste operational resources. These problems prevent tag systems from effectively supporting goals like segmented outreach, personalized recommendations, or churn prevention.

How the Skill Works: Core Capabilities and Key Steps

This skill builds a sustainable, governable, and non-rotting private domain user tag system from scratch. The core output is a four-part deliverable:

• Tag hierarchy architecture: Defines the layered structure of tags (e.g., first-level categories as fact, behavior, prediction) for clear classification, avoiding tag clutter.
• Tag dictionary and metadata: Includes 18 fields with detailed definitions, such as tag name, definition, owner (business + data dual ownership), and quality score, ensuring tags are well-documented and traceable.
• Governance SOP: Establishes standard operating procedures, including the owner system, definition arbitration rules (prioritizing target use cases), online/offline thresholds (e.g., quality score below 0.6 triggers offline), and quantifiable quality score formulas (weighted calculations).
• Implementation roadmap: Plans 6 phases—inventory existing tags, design architecture, execute tagging, validate quality, launch monitoring, and iterate optimization—for step-by-step execution.

The skill takes inputs like business model, private domain carrier (e.g., WeChat Work/community), existing tag list, and data sources (tracking/orders), then generates a structured blueprint based on methodology. For instance, if data sources are missing, it degrades to fact and order tags, marking others as "awaiting data sources." If tag definitions conflict, it applies arbitration rules to map or offline tags. After delivery, an Excel template (generated via openpyxl) is provided for easy filling, with pre-formatted headers and sample rows for operations/data teams.

Applicable Boundaries and Considerations

This skill is a pure methodology output, free to use without login or API keys, with the agent directly producing the four-part deliverable. It focuses on "system construction and governance" (enterprise-level blueprint), not single-audience profile cards—if users need to populate specific audience profiles, they should use the companion skill. The tag system requires iteration, with all criteria quantifiable (avoiding vague terms like "good" or "moderate"), and prediction tags must include score, confidence, and timestamp. Missing inputs follow exception scenarios: e.g., business model or private domain carrier must be queried, not assumed. Reports can be submitted to the server for logging, but failure does not block delivery. Final outputs need to be adapted to the enterprise's actual data sources, tools, and organizational divisions; major decisions recommend manual review.