Hi Product Hunt, I'm the solo maker of CatchFlaw.

Star averages bury bad stays: two thirds of Fukuoka hotels sit between 3.9 and 4.4 on Google, yet among the 24 hotels rated exactly 4.1, the share of "severe problem" or "never again" reviews ranges from 2% to 13%.

So I built something that reads the reviews for you.

What it does

• Collects 66k public Google and Trip.com reviews for 203 Fukuoka hotels

• An LLM reads the 34k text reviews and tags each one with 21 sub-topics (cleanliness, smell, noise, room, staff, location, safety) and a severity, plus a per-review "would not return" judgment

• Each hotel gets one number, disappointment probability: the recency-weighted share of reviews that report a severe problem or refuse to return, shrunk toward the city average when the sample is small

• City average is 6%; hotels range from 2% to 23%

• Rare but critical issues (bugs, mold, room security) are shown as counts, not probabilities

• Every number links to the original review, its date and language

• Recomputed weekly, static site, no login, free

What it is not

• Not a booking site (some hotel pages carry a clearly marked Agoda affiliate link)

• Not a ranking: hotels with fewer than 100 reviews in the last year are excluded from ordering, and position is shown as a band ("top 25%"), not a rank

Fukuoka only for now (a short flight from Korea).

The data is public (CC BY 4.0): https://huggingface.co/datasets/...

Full list of 203 hotels: https://catchflaw.com/hotels?utm_source=producthunt&utm_medium=referral&utm_campaign=backlink2611

How the number is computed: https://catchflaw.com/about?utm_source=producthunt&utm_medium=referral&utm_campaign=backlink2611

Happy to answer anything about LLM classification consistency, the scoring, or which city should be next.