pattern playground

Searching

Added

A transactional method for retrieving specific information by matching actor intent (query) with system content. Unlike Navigating (where actors move through structure) or Browsing (where actors scan for patterns), Search relies on the system to do the heavy lifting of evaluation and discrimination.

Anatomy

The search experience spans three distinct phases, often distributed across different UI components:

1. Entry & query formulation

The point where the actor expresses intent.

  • Keyword search: Matches exact strings or tags. Best for specific items (names, IDs).
  • Natural language prompt: Describes the desired outcome or characteristics. Best for complex queries that would otherwise require heavy filtering.
  • Scope:
    • Global: Searches the entire system.
    • Local: Searches only within the current context.

2. Processing & translation

  • Lexical and statistical retrieval: explicit algorithms relying on patterns in the text.
  • AI intent translation: The system bridges the gap between the actor’s language and the system’s taxonomy.

3. Results & refinement

  • Instant results: Popover or dropdown.
  • Search results view: A dedicated data view.

Stability & undo

Search shouldn’t destroy context. It should be easily reversible, returning the actor to their previous state (e.g., scroll position) without friction.

  • Combobox — the entry-phase mechanism for instant-results

Consequences

  • a result set now exists — ordered by relevance rather than by the collection's own structure — and presenting it is a design problem of its own
  • a search can fail, leaving the actor with intent but no result — browsing the structure ("browse by category instead") is the fallback seeking move

Resources & references

Appears in

Structuring the space

  • Growing past the model, step 1 — within Add sideways paths before changing shape: the escape hatch for the actor who knows the item rather than the place

Composing views

Related patterns

Precedes

  • Data view — a result set now exists — ordered by relevance rather than by the collection's own structure — and presenting it is a design problem of its own

Complements

  • Autocomplete — real-time narrowing of query suggestions as the actor types
  • Bounded choice — where the move escalates when the set outgrows narrowing
  • Fully connected — alternative navigation path
  • Hub and spoke — For actors who know their destination, a command menu and searching reach a spoke without the hub, and deep linking reaches it from outside the structure altogether.
  • Multilevel tree — escape hatch when hierarchy fails
  • Pan and zoom — enables jumping to locations
  • Navigation — the fallback seeking move when a search fails; browse and search co-deploy over the same structure
  • Filtering — often used after a search to refine results; in AI-search, filters are the output of the search query
  • Prompt — modern search inputs are evolving into prompt interfaces, accepting instructions alongside keywords
  • Suggestion — "did you mean?" and related patterns that guide query formulation

Related to

  • Intent & Interaction — search read as a navigation behaviour — how the seeking move fits the actor's broader movement through the system
  • Reference material — retrieval does implicitly what attachment does by hand — choosing what the system works from
  • Selection — Search shrinks the population from which selections are drawn

Preceded by

  • Information architecture — the underlying structure and labelling system that search indexes — without good IA, search is brittle
  • Tag — an applied label is a facet a search can narrow by

Instantiated by

  • Command menu — searching is the broader seeking move, applied to a command vocabulary