Global Hotel Price Index

Methodology

• What we measure: the median of live nightly hotel rates (Agoda availability, 2 adults) sampled per city over a rolling 30-day window. The median resists one-off spikes better than an average.
• 30-day trend: the current window’s median vs the prior 30 days. Shown only when both windows have a sufficient sample.
• What has to be true to be listed: a city needs at least 4 readings in the current window, and each reading has to be a median over at least 5 sampled properties. Thin samples — two hotels that happen to be expensive — describe those two hotels, not the city, so they are dropped rather than ranked. 242 cities clear that bar today.
• What we refuse to publish: a nightly rate at or above $1,000 is a mis-tagged package, not a room. And for reference markets we hold a plausible range: a city whose computed median lands grossly outside it means the source feed broke, not that the market moved, so we withhold the city instead of printing a number we can’t defend. Excluded cities are queued for a targeted re-pull, so they return on their own once the sampling recovers.
• Excluded from this run: none — every sampled city passed the checks above. (This is read from the live index, not a fixed claim.)
• Directional, not a census: rates reflect sampled availability, not every hotel in a city. Treat it as a market-level index.