Ecommerce Search UX — 8 Query Types

Source: https://baymard.com/blog/ecommerce-search-query-types Author: Baymard Institute

Baymard’s taxonomy of the query types users enter into e-commerce search, derived from large-scale usability testing across desktop sites, mobile sites and apps. Each type is reported with the share of benchmarked sites that fail to support it — the house style described in Baymard Institute. The reference taxonomy for Query Types in e-commerce.


The 8 Query Types

Percentages are the share of benchmarked sites with issues supporting that type.

#Query typeSites with issues
1”Exact” searches12%
2”Product Type” searches20%
3”Feature” searches39%
4”Use Case” searches43%
5”Abbreviation and Symbol” searches54%
6”Compatibility” searches44%
7”Symptom” searches37%
8”Non-Product” searches66%

The ordering is informative on its own: the types sites handle worst are the ones furthest from a catalog’s own vocabulary. Exact and product-type searches map onto a category tree directly and are largely solved. Non-product, abbreviation/symbol and use-case searches have no natural node to map to, and two thirds of sites fail the worst of them.

One worked example from the article: a “Product Type” search for “chairs” at CB2 returns a search results page rather than a category page, but includes “Related Searches” that narrow 522 items down to a specific chair category, with no irrelevant items in the list.

Benchmark Context

The article summarises 10,000+ usability scores across the 5 search topics that make up the e-commerce search experience, plotted over 170+ benchmarked sites and apps.

  • Only 44% of desktop and mobile sites and apps achieve a “decent” or “good” overall Search UX performance.
  • 46% of desktop, 58% of mobile and 64% of app experiences are “mediocre or worse”.
  • A handful of exceptions exist — 4 desktop sites and 1 mobile site scored “perfect” — so the ceiling is reachable.

Baymard’s stated framing is that abandonment is rarely caused by one severe issue; it accumulates from medium-level issues encountered across several searches.

Why This Matters for Classification

A query-type taxonomy is a label set, and this one is defined by observed user behaviour rather than by catalog structure. That is the useful property: it names the demand a catalog-derived taxonomy cannot express. Query types that no category node covers are exactly the ones that end up in a classifier’s residual bucket — see Query Classification.

Companies