AI Search
AI Search combines language understanding and product signals to return useful results even when customers use everyday wording.
Explanation →From AI and ranking to conversion, evaluation, indexing and performance.
AI Search combines language understanding and product signals to return useful results even when customers use everyday wording.
Explanation →Semantic search looks beyond exact words to understand related meanings, contexts and product needs.
Explanation →Lexical search matches the words and character sequences in a query, making it dependable for precise product terms.
Explanation →Vector search represents queries and products as numerical meaning signals to find conceptually related items.
Explanation →Hybrid search combines lexical precision with semantic understanding, so exact identifiers and broader intent both count.
Explanation →Query understanding identifies the likely product, attribute, brand or task behind a customer search.
Explanation →Query expansion adds carefully selected related terms, helping customers find products without needing to know the catalogue wording.
Explanation →Search intent describes what a customer is trying to accomplish, such as comparing, replacing or buying a product.
Explanation →Typo tolerance recognizes common spelling mistakes and input variations without turning unrelated products into matches.
Explanation →Synonym management connects customer vocabulary to catalogue language while keeping each relationship reviewable and controlled.
Explanation →Search ranking determines the order of matching products using relevance, availability, business and behavioral signals.
Explanation →A ranking rule makes a specific ordering decision explicit, so teams can review its purpose, scope and impact.
Explanation →Boost and bury temporarily raise or lower selected products when there is a clear commercial or customer-service reason.
Explanation →A relevance score expresses how well a product matches a query, based on the signals your search engine can explain.
Explanation →A relevance judgment records whether a result is good, acceptable or poor for a query, creating evidence for improvement.
Explanation →A regression test checks known queries after a change, preventing earlier search quality gains from being lost.
Explanation →CTR measures how often a search result is clicked after it is shown, revealing whether results attract attention.
Explanation →A search reformulation happens when a customer changes their query, often signalling that the first result set was not helpful.
Explanation →Zero-result rate shows the share of searches that return no products, helping teams prioritize missing data or vocabulary.
Explanation →Search conversion links a search journey to a later order, showing whether discovery contributes to revenue.
Explanation →An attribution window defines how long a product click or search can receive credit for a later conversion.
Explanation →Conversion rate shows the percentage of search journeys or product clicks that lead to an order.
Explanation →Index freshness indicates how recently product, price, stock and availability changes became searchable.
Explanation →Query latency measures the time needed to return search results, helping teams protect a fast customer experience.
Explanation →Fallback search provides a safe alternative when the primary search route cannot produce a suitable response.
Explanation →Faceted search lets customers refine results with structured filters such as brand, size, price or availability.
Explanation →Autocomplete suggests useful queries, categories or products while customers type, reducing effort and guiding discovery.
Explanation →Product discovery covers the ways customers find relevant products beyond a single exact search query.
Explanation →Personalization adapts search or recommendations using relevant context, while respecting privacy and keeping the experience understandable.
Explanation →NLP, or natural language processing, helps systems interpret human language, including wording, grammar and intent.
Explanation →Embeddings convert text or products into numerical representations that help systems compare meaning and similarity.
Explanation →