Title: Semantic / Aesthetic-Based Search
Summary:
Enable users to search the marketplace by style, aesthetic, or vibe (e.g., "steampunk", "cyberpunk", "minimalist") rather than relying solely on exact text matches against item names and descriptions.
Epic:
- E3: Marketing – Primary Sales
Pillars:
- P8: Browse – Explore the catalog
Impacted Metrics:
- M23 (# of unique sessions/week on marketplace)
- M24 (Conversion rate — sessions → purchase)
Problem
Buyers browsing the marketplace can only search by literal keywords that happen to appear in item names or descriptions. A user looking for "steampunk" gear will miss items tagged as "Victorian goggles" or "brass armor" — there is no semantic or aesthetic layer connecting items by visual style. This limits discoverability, especially for new users who don't know specific item names and browse by taste rather than title.
To validate:
- Check the current metrics about marketplace search functionality usage. What do users search for? Are they converting after searching? How much are they scrolling?
Proposal
Introduce a semantic search layer that maps items to aesthetic/style tags (either via manual creator-applied tags, AI-based visual classification, or a combination). When a user searches "steampunk," the system returns items whose visual style matches — not just those with "steampunk" in the title. This could be implemented through:
- A curated taxonomy of aesthetic tags (e.g., Cyberpunk, Fantasy, Streetwear, Minimalist, Retro) that creators select during submission.
- An AI/ML-based visual similarity engine that clusters items by appearance.
- A hybrid approach where curated tags bootstrap the system and ML refines over time.
The goal is to make the marketplace browsable by "vibe" — the way users actually think about fashion — rather than requiring exact keyword recall.
Title: Semantic / Aesthetic-Based Search
Summary:
Enable users to search the marketplace by style, aesthetic, or vibe (e.g., "steampunk", "cyberpunk", "minimalist") rather than relying solely on exact text matches against item names and descriptions.
Epic:
Pillars:
Impacted Metrics:
Problem
Buyers browsing the marketplace can only search by literal keywords that happen to appear in item names or descriptions. A user looking for "steampunk" gear will miss items tagged as "Victorian goggles" or "brass armor" — there is no semantic or aesthetic layer connecting items by visual style. This limits discoverability, especially for new users who don't know specific item names and browse by taste rather than title.
To validate:
Proposal
Introduce a semantic search layer that maps items to aesthetic/style tags (either via manual creator-applied tags, AI-based visual classification, or a combination). When a user searches "steampunk," the system returns items whose visual style matches — not just those with "steampunk" in the title. This could be implemented through:
The goal is to make the marketplace browsable by "vibe" — the way users actually think about fashion — rather than requiring exact keyword recall.