Eleanor Roach's two principles of categorization explain why classification feels intuitive when it's right and jarring when it's wrong: cognitive economy means categories should give people maximum information for minimum mental effort, and perceived world structure means people expect digital categories to map onto the same kind of structure they perceive in the physical world. Morville and Rosenfeld identify four concrete challenges that make classification genuinely difficult in practice: ambiguity, since language itself is imprecise and category boundaries blur; heterogeneity, when content varies too much in type or granularity for one scheme to fit cleanly; differences in perspective, since people bring different mental models shaped by their own knowledge and experience; and internal politics, where departments disagree about categorization for reasons that have nothing to do with users.
Metadata is the tool that makes classification operational. Three types matter most: descriptive metadata, title, topic, keywords, describes what a content object actually is, and has the greatest impact on navigation and search. Structural metadata describes a content object's internal structure, enabling the same content to be assembled and displayed differently in different contexts. Administrative metadata, owner, dates, status, supports the business processes of managing that content, like workflow and publishing.
Every "which category does this belong in" argument on a real project is usually one of these four challenges wearing a disguise, recognizing which one you're facing tells you what kind of resolution is even possible.
1: Ambiguity. Two teams using the same word, "Litigation," to mean different things is a direct case of language's inherent imprecision creating unclear category boundaries.
2: Heterogeneity. Case notes, video depositions, scanned documents, and structured billing records are too dissimilar in type and structure for one classification approach to fit all of them cleanly.
3: Differences in perspective. Junior associates and senior partners bring different mental models shaped by their different roles, task-focused versus client-focused, to the same content.
4: Internal politics. The resistance here isn't really about which scheme is better: it's about control, which is exactly what the internal politics challenge describes.
For problem 3, administrative metadata is the most direct fix, since fields like client name and matter number are exactly the kind of business-context information administrative metadata is built to capture. Rather than forcing associates and partners to agree on a single categorization scheme, both client/matter administrative metadata and task-based descriptive metadata could be applied to the same document simultaneously, letting each group filter and browse by whichever attribute matches how they actually think, without requiring either side to give up their own mental model.This is a useful general lesson: differences in perspective don't always need to be resolved by picking a winner. Sometimes the right fix is applying multiple metadata fields to the same content so different audiences can each find it their own way.