The keyword problem
Traditional marketplace search matches words. If your listing says "motorcycle" and the buyer types "moto", they get zero results. If someone searches "quiet workspace near city center" and your listing says "studio office downtown" — no match.
This isn't a configuration problem. It's a fundamental limitation of keyword matching.
What semantic search actually does
Semantic search converts text into mathematical representations (vectors) that capture meaning, not just characters. Two phrases that mean similar things end up close together in vector space — even if they share zero words.
In practice, this means:
- "moto" → finds "motorcycles", "motorbikes", "scooters"
- "cozy place" → finds "studio apartment", "small flat", "compact unit"
- "something for my kid's birthday" → finds "children's party services", "kids entertainment"
The search understands intent, not just keywords.
Impact on marketplace metrics
When buyers find what they're looking for:
- Bounce rate drops — they don't leave after one failed search
- Conversion increases — relevant results lead to transactions
- Return visits increase — "this marketplace actually finds what I need"
For operators, it also reveals demand signals. When you see what buyers tried to search for, you know exactly what supply to add.
How it works with Sharetribe
Empasa's semantic search runs alongside your existing Sharetribe search. Your template can call the semantic endpoint instead of (or in addition to) the default keyword search. No database changes, no reindexing your marketplace — it reads your listing data and builds the vector index automatically.
Want to see semantic search in action? Start a free trial and connect your marketplace.