How Etsy Uses Thermodynamics to Help You Search for “Geeky”
Source: https://codeascraft.com/2015/08/31/how-etsy-uses-thermodynamics-to-help-you-search-for-geeky/ Author: Etsy Engineering
Summary
Etsy applies concepts from statistical thermodynamics — specifically entropy — to measure and optimize result diversity for broad, ambiguous queries. Companion piece to Targeting Broad Queries in Search.
The Thermodynamics Analogy
In thermodynamics, entropy measures disorder/spread of a system. Applied to search results:
- Low entropy = results all look similar (converged, narrow)
- High entropy = results span many categories and styles (diverse, exploratory)
For a query like “geeky”, optimal results have high entropy — they should show geeky necklaces, t-shirts, mugs, art prints, etc., not five variations of the same item.
Key Technical Idea
- Measure result entropy using Shannon entropy over category/attribute distributions
- Set entropy targets per query based on query type (broad = high target, specific = low)
- Re-rank or inject results to meet entropy target while maintaining base relevance
Why This Matters
Pure relevance ranking on broad queries produces “winner takes all” results — the most popular category dominates. Entropy-based diversification corrects this by ensuring the result set represents the full intent space.