Match classification architecture to domain
Design multi-class classification systems with mutually exclusive categories when items can only be one type, and multi-label systems when items can legitimately belong to multiple categories simultaneously.
When This Fires
- Designing any classification system (knowledge base, project management, content organization)
- When items frequently belong to multiple categories and the system forces single placement
- When duplicating content to make it appear in multiple folders
- During any information architecture design
Common Failure Mode
Defaulting to folders (multi-class) for everything because they're familiar. Folders enforce exclusive placement — each item goes in exactly one place. For genuinely exclusive domains, this works. For overlapping domains (notes about topics, articles about themes, tasks affecting multiple projects), folders force arbitrary placement and destroy cross-category retrievability.
The Protocol
When designing a classification system: (1) Ask: "Can items in this domain legitimately belong to multiple categories?" (2) If no (exclusive) → multi-class: folders, single-select, status fields. Each item has exactly one home. (3) If yes (overlapping) → multi-label: tags, multi-select, polyhierarchy. Each item can have multiple homes. (4) If mixed → use multi-class for the exclusive dimension (status) and multi-label for the overlapping dimension (topics). Most real systems need both.