MOC — Search UX and Discovery
Map of content for the Search UX & Discovery family: how users express a need, scan and refine results, recover from failure, and discover things they weren’t explicitly looking for. Covers query input, results presentation, faceted navigation and filtering, search scopes, zero-results recovery, and discovery/inspiration patterns.
Start here
Topic hub: Search UX — the full narrative (interaction loop, presentation, facets, zero results, discovery, measurement). This MOC is the link-out index into the notes behind it.
Query Input UX
- Concept: Autocomplete — prefix trie, neural suggestions, UX patterns
- Topic: Autocomplete and Autosuggest — query-input UX in full
- Bootstrapping Autosuggest — Giovanni Fernandez-Kincade: goals and metrics first
- 13 Design Patterns for Autocomplete Suggestions — Baymard: 27% get it wrong; 9 UX patterns
- How to Really Scale Autocomplete — bonsai.io; search_as_you_type + significant_terms; 416 RPS at 45ms p99
- LLM-Powered Query Extraction for Autocomplete — David Albrecht; GPT-4.1-nano solves cold-start
- Affordances for Conversational Search — Daniel Tunkelang’s affordance gap: showing users what the system can do
Spelling Correction
- Concept: Spelling Correction — Norvig’s algorithm, SymSpell, context-aware
- How to Write a Spelling Corrector — Norvig’s classic probabilistic algorithm; 21 lines of Python
- 1000x Faster Spelling Correction Algorithm - SymSpell — Wolf Garbe: O(1) via delete-only preprocessing
- Modeling Spelling Correction for Search at Etsy — industry application at scale
Synonyms
- Concept: Synonyms — lexical vs. semantic synonyms; index-time vs. query-time expansion
- Real Talk About Synonyms and Search (Tunkelang)
- Boosting the Power of Elasticsearch with Synonyms
Results Presentation
- Concept: Presentation Bias — what’s shown shapes what’s clicked; only-shown results can generate feedback
- Concept: Position Bias — top results get clicks regardless of quality
- Query Understanding - Search Results Presentation — layout, snippets, components matched to intent
- Concept: Results Merchandising — curating/pinning results to tell a product story
- Concept: Results Boosting — score-level promotion/demotion behind the presentation layer
- The Pinball Pattern - Complex Search-Results Pages Change Search Behavior — NN/g: nonlinear SERP scanning
Faceted Navigation & Filtering
- Concept: Faceted Search — dynamic filters, constraints vs. preferences, counts, dynamic facets
- Facets of Faceted Search (Tunkelang)
- Facets - Constraints or Preferences (Tunkelang) — the key facet-UX distinction
- Facets, But Which Ones (Tunkelang) — choosing which facets to expose
- 7 Filtering Implementations That Make Macy’s Best-in-Class (Baymard)
Search Scopes
- Concept: Search Scopes — pre-query narrowing (department/type/“search within”); the sticky-scope failure mode
- Concept: Federated Search — querying multiple scopes/sources at once
- Query Understanding - Query Scoping — automatic scope detection from the query
- Scoped Search - Dangerous but Sometimes Useful — NN/g: the sticky-scope failure mode
Zero Results & Recovery
- Concept: Zero Results — causes and recovery strategies; findability floor
- Concept: Query Relaxation — drop/loosen constraints to refill an empty page
- Search UX: 6 Essential Elements for ‘No Results’ Pages (Baymard)
- Strategies for Using Alternative Queries to Mitigate Zero Results
Discovery & Inspiration
- Topic: Search Result Diversity — the discovery angle of result presentation
- Three Pillars of Search Quality - Discovery and Inspiration — Andreas Wagner: findability vs. discovery vs. relevance
- Broad and Ambiguous Search Queries — Tunkelang: recognizing when results need diversification
- Thoughts on Search Result Diversity — Tunkelang on diversity as a UX goal
- Concept: MMR — Maximal Marginal Relevance; the workhorse diversification method
Query Types & UX
- Ecommerce Search UX - 8 Query Types — Baymard’s e-commerce query taxonomy
- Query Types — how intent type sets presentation expectations
- Targeting Broad Queries in Search — Etsy approach
Measuring & Experimenting on UX
- Concept: A-B Testing for Search
- A-B Testing for Search is Different — Tunkelang: session-level analysis
- Concept: Clicks Residual · Zero Results — engagement and failure signals
- Good Abandonment on Search Results Pages — NN/g: no-click sessions aren’t always failures
- Search-Log Analysis - The Most Overlooked Opportunity in UX Research — NN/g: query logs as UX research
- Common Pitfalls of Search Experimentation