Luxury Intel is my attempt to fix two problems when booking luxury hotels:

1. Google ratings are useless in this segment. Everything sits between 4.7 and 4.9 stars. There is no signal in average ratings.

2. Prices are deliberately opaque. If you want to obtain a year of nightly rates for a handful of hotels you're stuck clicking through calendar widgets for hours.

Luxury Intel this by:

1. Calculating ratings that differentiate using >400k reviews and >20k pairwise comparisons across specific axes (service, food, rooms, location, value, ambiance), and overall.

2. Scraping property websites for a year of forward nightly rates for each property. Then displaying the data in a price curve and weekday/month heat map.

A few notable findings:

- Aman is mid-pack overall, with bi-modal scores. The brand is very weak on Food and Value while very strong on Ambiance.

- Price only weakly predicts quality. Correlation between median nightly rate and overall rating is 0.26. The median rating is essentially flat across the $200, $500, $1000, and $2000 price tiers. The premium only starts to show up above $2000.

- Service is the dominant driver of overall ratings (r=0.86). Location is almost uncorrelated with it (r=0.15).

- Rooms and Location are negatively correlated (r=-0.20). The hotels with the best locations tend to have weaker rooms, and the properties with the best rooms tend to be in more remote locations.

Stack: Python, Astro, Node, solo project built in evenings.

I find it useful and I hope you do too.