Smart Search Refinement to Fix Irrelevant Results on Meesho
Why a query like 'black formal shoes under 600' fails on Meesho, and three fixes ranked by RICE
✳The problem they went after
Shoppers type natural language into Meesho search and get back repetitive, irrelevant or badly ranked results. They scroll past duplicates, run the query again, and some of them leave for a competitor app. On a catalogue-driven commerce platform that hits search-to-product-click conversion directly, which is the core engagement metric.
✳What they actually did
Three solutions, scored with RICE. Smart Refine Chips generate filter tags automatically from the user's original query and rank highest at 18, cheap to build and immediately useful. Result De-duplication clusters identical and near-identical listings into expandable cards, scoring 13.5 and cutting scroll fatigue at the source. A Search Rescue feature surfaces alternative query suggestions after twenty items pass without a click, so a failing search recovers instead of restarting. Wireframes trace each flow from query to product click.
✳What you can take from it
Worth stealing
- Smart Refine Chips generated from the raw query
- De-duplication into expandable clusters, RICE score 13.5
- Search Rescue triggers after twenty items with no click
- Risk named: de-duplication reading as a smaller catalogue
Read this if
Anyone working on marketplace search or query understanding
The transferable bitChips can extract the wrong attribute, de-duplication can make the catalogue look smaller than it is, and ranking logic can misclassify attributes. Naming those failure modes before build is what separates a search idea from a search plan, and it is the part of this submission most worth reading twice.
✳The submission itself
Open the PDF →Reading on a phone? The embedded viewer is cramped — open the PDF in a new tab instead.
