Cosy score
A transparent, evolving heuristic to estimate how cosy a place feels.
What goes into the score
- Overall rating (out of 10) – normalized and weighted.
- Amenities warmth – amenities associated with cosiness (e.g., fireplace, bathtub, spa/sauna, garden).
- Language – descriptive text mentioning cosy/cozy, warm, intimate, quiet, romantic, character, etc.
- Scale – small/boutique places tend to feel cosier than very large hotels.
How we blend it
We combine the signals into a single 0–10 score. Ratings provide the base, amenities and language nudge cosiness up, and very large scale can apply a small penalty. The goal is to surface warm, characterful stays while staying simple and transparent.
Limitations
This is a starting point. As we ingest more data (reviews, images), we’ll refine the model with NLP and image signals. Feedback is welcome.