Query Types
Definition
Query types are taxonomies that classify search queries by their structural characteristics, vocabulary, and user behavior patterns — distinct from Search Intent (which focuses on goal) but related.
Baymard’s 8 E-commerce Query Types
Baymard Institute’s taxonomy for e-commerce search, derived from large-scale usability testing. The percentage is the share of benchmarked sites that fail to support that type — see Ecommerce Search UX - 8 Query Types.
| Type | Example | Sites with issues | Challenge |
|---|---|---|---|
| Exact | ”Nikon D800” | 12% | Match a specific model or SKU exactly |
| Product Type | ”chairs” | 20% | Broad type; behaves like a category page request |
| Feature | ”wireless noise cancelling headphones” | 39% | Attribute matching against structured data |
| Use Case | ”running shoes for marathons” | 43% | Activity implies attributes never stated |
| Abbreviation and Symbol | ”13in laptop”, “tv w/ hdmi” | 54% | Tokenisation and normalisation of symbols and shorthand |
| Compatibility | ”case for MacBook Air M2” | 44% | Structured relation between two products |
| Symptom | ”my back hurts pillow” | 37% | Infer product category from a described problem |
| Non-Product | ”return policy” | 66% | Service content, not products |
The ranking is the finding: the types sites handle worst are the ones furthest from the catalog’s own vocabulary. Only “chairs” is an example drawn from the source; the rest are illustrative.
Other Common Type Labels
Widely used in practice but not part of Baymard’s taxonomy — worth keeping distinct when citing sources:
| Type | Example | Note |
|---|---|---|
| Category | ”cameras” | Overlaps Baymard’s “Product Type” |
| Branded | ”Nike shoes” | Filter + rank within brand |
| Thematic | ”birthday gift for photographer” | Discovery; benefits from semantic retrieval |
| Comparative | ”mirrorless vs DSLR” | Educational content, not a product listing |
Academic Query Type Taxonomies
Factoid vs. Non-Factoid
- Factoid: has a short, definitive answer (“When was Napoleon born?“)
- Non-factoid: requires explanation, opinion, or multiple facts
Keyword vs. Natural Language
- Keyword: “iphone 15 review 2024”
- Natural language: “What do people think of the iPhone 15?”
Head vs. Torso vs. Tail
Based on query frequency:
- Head (>1000 daily): “shoes” — simple, broad, easy to optimize
- Torso (10–1000): “running shoes women” — specific, addressable
- Tail (<10): “waterproof trail running shoes wide toe box women” — rare, hard to optimize individually
Query Length and Complexity
| Length | Characteristics | Best Handled By |
|---|---|---|
| 1 word | Ambiguous, broad | Diversification + personalization |
| 2–3 words | Most common | Standard hybrid retrieval |
| 4–6 words | Specific | Semantic search excels |
| 7+ words | Conversational/complex | Asymmetric Semantic Search, RAG |
Query Type in Search Sampling
When building Judgment Lists, stratified sampling by query type ensures evaluation covers all types proportionally rather than being dominated by head queries.
Related Concepts
-
Query Classification — assigning queries to these types in production
-
Search Intent — goal-based classification (overlaps)
-
Query Understanding — broader processing framework
-
Judgment Lists — query type used in sampling strategy
-
Diversity Metrics — exploratory/thematic queries need diversity
-
Asymmetric Semantic Search — long queries benefit from asymmetric models
-
Query Specificity — the degree of constraint in a query; head ↔ tail correlates with low ↔ high specificity
People
- Daniel Tunkelang — query type taxonomy; query understanding framework