Suyoung Mel Kwon
Selected Works
CommerceSearch UXVector SearchAI

“Just type it out. We'll find what you need.”

AI Search: Recovering Lost Revenue

Enhancing search experience & driving sales

Period
May – Aug 2024
Role
PMA/B Test DesignCEO CommunicationQA
Team
1 ML engineer1 backend engineer
Org
Kurly
10s Summary

What?

Cut "no results" searches from 6.8% to 0.22%, unlocking ₩5B (~$3.6M) in new monthly revenue.

Why?

The search engine couldn't handle typos, spacing, or synonyms, so 7% of searches on Kurly's biggest revenue channel hit a dead end.

How?

When the old search came up empty, we fell back to semantic search with Vertex AI, so "cabage" still finds cabbage.

AI Search: Recovering Lost Revenue, case study page 1 of 4AI Search: Recovering Lost Revenue, case study page 2 of 4AI Search: Recovering Lost Revenue, case study page 3 of 4AI Search: Recovering Lost Revenue, case study page 4 of 4

Tap a page to open it at full resolution.

Related Links
← Previous

AI Curation: Scaling Themed Campaigns