Search experience playbooks
Help people find the right result through useful queries, filters, facets, and relevance. Choose a scenario, then the decision you need to make.
Product search intent drift
A product query may quietly shift from shopping intent to support, account, or navigation intent. Separate the job behind the words before changing ranking or scope.
- Triage product search intent drift before changing behavior
- Prioritize product search intent drift with a clear tradeoff
- Investigate product search intent drift with a reproducible brief
- Verify product search intent boundaries across representative paths
- Learn from product search intent drift for the next review cycle
Product search synonym gap
People use alternate names, abbreviations, and regional vocabulary for the same product concept. Find the evidence for a synonym before adding a rule that could broaden unrelated queries.
Search boost conflict
A merchandising boost, pin, or freshness rule can push a result above a stronger textual match. Make the competing signals visible before changing relevance behavior.
Facet value coverage
A filter can be technically present while omitting values that matter for the current catalog. Compare attribute coverage with returned facets before calling the control complete.
Facet count trust
Facet counts can disagree with visible results when filters, pagination, indexing, or cache boundaries differ. Reconcile the numbers users rely on before redesigning the filter panel.
Filter state persistence
A search view may lose filters when someone paginates, sorts, returns from a result, or opens the same query on a narrow screen. Define the state that must survive each transition.
Filter combination dead ends
Two individually useful filters can create a confusing dead end together. Distinguish a valid narrow result from an impossible combination and give the empty state an honest recovery path.
Autocomplete noisy suggestions
Suggestions can save typing or distract people with popularity, stale terms, and irrelevant matches. Review what each suggestion helps someone do and whether its ranking follows the typed prefix.
Autocomplete keyboard selection
A suggestion list can look correct while focus, Escape, Enter, or highlighted selection behaves differently for keyboard and assistive technology users. Test the interaction contract, not only the screenshot.
- Triage autocomplete keyboard selection before changing interaction
- Prioritize autocomplete keyboard selection with a clear contract
- Investigate autocomplete keyboard selection with an event trace
- Verify autocomplete keyboard selection across input methods
- Learn from autocomplete keyboard selection failures
Query spelling variants
Typos, casing, hyphenation, and language variants can create false zero-result signals or silent intent changes. Separate helpful correction from a result set that should remain literal.
Zero-results recovery
No results can mean missing inventory, a narrow filter, a vocabulary mismatch, or a broken scope. Make the empty state distinguish those cases and offer a next useful move.
Zero-query discovery
An empty search field can be a useful starting point or an unhelpful blank screen. Decide what discovery evidence belongs there without pretending recent or popular terms are universal.
Result snippet context
A result title can match a query while its excerpt omits the context needed to judge relevance. Review snippet selection, highlighting, and truncation across the layouts people actually use.
- Triage search result snippets that hide useful context
- Prioritize search snippet context with a useful result choice
- Investigate search result snippets with query-to-context traces
- Verify search result snippets across content and screen widths
- Learn from search snippet context and result-choice mistakes
Result sort expectation
People read “relevance,” “newest,” “price,” and category position as different promises. Compare the visible sort control with the order produced under each query and view.
Availability filter confidence
An availability filter may reflect an index snapshot, a region, a parent product, or a sellable variant. Establish what “available” means before promising that every result can be obtained.
Category search boundary
Category browsing and global search can share controls while using different scope rules. Make the route, breadcrumb, phrase, and category filter tell the same story.
Duplicate product results
Variants, bundles, regional records, and duplicate index documents can crowd a result page. Determine which rows are useful choices and which repeat the same product identity.
Multi-index search scope
A unified search may combine products, documentation, projects, or account records with different relevance and access rules. Establish which indexes belong in each context.
Relevance explanation gap
Reviewers can argue about ranking without seeing the fields, boosts, freshness, or pins that influenced a result. Keep a compact explanation trail for tuning decisions.
Query learning loop
Search improvement needs a record of reformulation, refinement, selection, and abandonment patterns. Define a careful learning loop that distinguishes evidence from an anecdotal query.