Failed Searches
summary
An e-commerce team analyzes the top 1,000 failed queries from three months, classifies causes, and adjusts stemming, decompounding, synonyms, indexing and redirects to reduce misses. When misses still occur they launch an alternative search-phrase recommender that uses customers’ subsequent searches and purchases to suggest related brands and products. Certain search phrases see about a 30% uplift in findable alternatives, examples include mapping missing brands like Daily Paper to colorful kidswear or rain cape to rain poncho, and overall session conversion increases.
AI-written summary · about
chapters (5)
related talks
49mReplatforming Online Search for a Major High Street Retailer
39mAI-Powered Search in Modern E-Commerce
37mSimilarity Search for Product Matching @ Semantics3
13mUnconference Session: A Neat Trick for Fuzzy Search
43mSearch Logs + Machine Learning = Auto-Tagged Inventory
47mWhy, What, When and How of Running Diff Types of Sear...
14mSearch on Steroids
44mFinding relevancy: Improving your application search bar!