Search behaviour pattern analysis

Description

Classify search sessions into quit, narrow, expand and pearl-growing patterns, and detect the antipatterns pogosticking and thrashing, to diagnose search design problems.

A shared vocabulary turns search logs and observed sessions into diagnoses. Quit: every search ends in an exit, and analytics must show whether users succeeded or gave up, especially after no-results pages. Narrow: refining after a broad first query, the second most common pattern. Expand: broadening by dropping terms, relaxing facets or browsing a thesaurus. Pearl growing: finding one good document and mining its terms and metadata for more, supported by 'find similar' features. Two antipatterns are caused by design: pogosticking, bouncing repeatedly between the results page and individual results because snippets lack information scent, and thrashing, repeating small variations of a flawed first query, an anchoring effect. Each pattern points to a design response.

Attributes

Loss type
Siloed
Data
Mixed
Effort
Days

Steps

  1. Collect search analytics with session identifiers and, if possible, observe or record a sample of search sessions.
  2. Identify exits, especially after no-results pages, and estimate success versus abandonment.
  3. Code narrowing and expanding sequences of queries, facets and filters.
  4. Look for pearl growing: use of similar-item features or terms mined from a good result.
  5. Flag pogosticking (repeated rapid returns from results to the results page) and thrashing (chains of near-identical reformulations).
  6. Respond with design changes such as clearer no-results pages with spelling suggestions, partial matching, facets, richer snippets, and autocomplete built from reformulation data.

Sources

  • Search Patterns (Morville & Callender, 2010)

Also known as

  • Search session pattern coding
  • Query reformulation analysis
  • Pogosticking and thrashing analysis

Categories

  • Content
  • Information Architecture & Content