Categorization and Tagging
Proper categorization improves both AI retrieval accuracy and customer self-service browsing.
Category vs. Tag
Categories are hierarchical and exclusive. An article belongs to one category. Tags are flat and non-exclusive. An article can have multiple tags.
Example:
- Category: Billing > Invoices
- Tags: payment, receipt, PDF, download, tax
Building Your Taxonomy
Start with a customer-centric taxonomy based on what customers are trying to do:
Tier 1 (Top-Level Categories) Keep to 5-8 maximum. More creates decision fatigue.
Tier 2 (Subcategories) 2-6 per top-level category. Only create subcategories when a top-level category exceeds 15 articles.
Example taxonomy:
Account
├── Login & Security
├── Profile Settings
└── Team Management
Billing
├── Plans & Pricing
├── Invoices
└── Refunds
Product
├── Dashboard
├── Integrations
└── API
Tagging Strategy
Use these tag types:
- Feature tags: dashboard, API, widget, reports
- Action tags: setup, configure, troubleshoot, migrate
- Audience tags: admin, agent, developer, end-user
- Status tags: new-feature, deprecated, beta
AI Retrieval Impact
Good categorization improves AI chatbot accuracy because:
- The AI can narrow search scope to relevant categories
- Related articles surface as follow-up suggestions
- Confidence scores increase when matches are category-specific
Maintenance Rules
- Review taxonomy quarterly as your product evolves
- Merge categories with fewer than 3 articles
- Split categories with more than 20 articles
- Audit tags monthly — remove unused tags and consolidate synonyms
- Assign a category owner responsible for accuracy
Mapping Customer Language
Customers rarely use your internal terminology. Create a synonym map:
| Your Term | Customer Says |
|---|---|
| Workspace | Account, organization, company |
| Widget | Chat bubble, chat window, popup |
| Knowledge base | Help center, FAQ, docs |
Add these synonyms as tags to improve AI matching.
Next up: How to measure whether your knowledge base is actually working.