Production-Oriented Taxonomy Design

A framework for building classification systems based on the design principles embedded in NAICS, applicable beyond industry classification to any domain requiring consistent, analyzable categorization.

The Framework

NAICS demonstrates that robust taxonomy design requires explicit decisions across five dimensions:

1. Choose Your Lens

Decide whether to classify by supply-side characteristics (how things are produced/created) or demand-side characteristics (what markets they serve/what outputs they produce). Neither is neutral. Document the choice and its implications.

Supply-side enables: productivity analysis, input-output relationships, process comparison. Demand-side enables: market sizing, customer segmentation, competitive analysis.

2. Commit to a Single Principle

Mixed-principle taxonomies create categories that look comparable but aren’t. Pick one organizing concept and apply it consistently, even when it produces counterintuitive groupings. The payoff is analytical coherence (every comparison in the system is meaningful.

3. Define Your Unit

What entity gets classified? The choice cascades through everything. NAICS chose establishments (locations) over enterprises (companies) because their analytical goals required site-level granularity. Your unit of classification determines what questions you can answer cleanly.

4. Design for Governance

Build revision processes into the system from the start:

• Define review cycles (NAICS uses five years)
• Specify criteria for when changes are warranted
• Accept that adaptation breaks time-series continuity
• Plan for historical restatement where possible

5. Encode Collaboration Boundaries

When multiple authorities share a classification, use hierarchical codes to separate shared agreement from local extension. NAICS reserves the sixth digit for national detail while harmonizing digits 1-5 across countries. The structure itself enforces the governance model.

When to Use This Framework

This framework is most valuable when:

• The classification will support statistical analysis over time
• Multiple parties need to agree on shared categories
• Consistency matters more than perfect fit for any single use case
• The domain is dynamic and will require periodic revision

Limitations

• Requires discipline to maintain single-principle purity
• May produce unintuitive groupings that frustrate domain experts
• Time-series breaks are inevitable as the domain evolves
• Not optimized for discovery or browsing (built for analysis

The Deeper Pattern

What makes NAICS instructive isn’t the specific choices it made, it’s that every choice is explicit. The manual documents why production-oriented over market-oriented, why establishments over enterprises, why five-year cycles.

Most taxonomies are implicit compromises. Making the design decisions visible allows them to be evaluated, debated, and, when necessary, changed deliberately rather than accidentally.

Related:, 02-molecule—faceted-classification