Building a knowledge graph that classifies 5.3 million queries instantly required rethinking how we structure search data from the ground up.
The Challenge
Raw keyword data is chaos. Millions of queries with no structure, no hierarchy, no actionable grouping. Manual classification? 1,000 keywords takes 2 hours. We had 5.3 million.
5-Level Taxonomy
We built a hierarchical classification system: Categories → Subcategories → Intents → Topics → Keywords. Each level adds specificity while maintaining the ability to roll up for strategic analysis.
Architecture
- —913 L1 Categories covering the entire industry landscape
- —Vector embeddings for semantic similarity matching
- —<1ms classification speed per query
- —5.3M keywords processed instantly vs. 2 hours for 1,000 manually
Impact on Content Strategy
The knowledge graph powers our content gap analysis, internal linking recommendations, and programmatic SEO strategy. Every piece of content maps to a node in the graph, creating a complete picture of topical coverage and authority.
Market Intelligence
Beyond SEO, the knowledge graph enables market share analysis with nationwide and geo-focused insights (state, city, zip code), customizable topic filters, and L1-L6 hierarchical reporting.