16.3% of guests render an explicit price verdict in their review text. Among those guests, "overpriced" predicts non-return at a rate that no other signal in the dataset matches: not friction, not sentiment, not the published score.
Churn = unlikely_return + will_not_return, NLP-inferred from review text. Denominator: reviews with a price verdict (16.3% of all reviews). Newcombe difference: 54.3pp (CI 52.5–56.0pp). Replicates in 2025-only data: overpriced 55.8%, worth-it 2.0%.
Note: stronger NLP signals (explicit detractor language, low ratings) predict churn at higher rates; this finding ranks high on predictiveness among non-extreme signals, as it surfaces even when rating and sentiment appear neutral.
Among guests rating exactly 8/10 with neutral sentiment: overpriced churn 17.3% vs worth-it 2.7%. Among all guests rating 8 or higher: 17.1% overpriced churn vs 2.6% for the rest. The score is identical. The verdict is not.
This guest gave a score that would look fine on any dashboard. The sentiment is neutral: no flag, no crisis. But 17.3% of guests in this exact combination signal non-return: a rate 6.4x higher than the guest one cell over in the same table.
denominator: reviews rated 8/10, neutral sentiment, with overpriced price verdict · cohort-117Same rating. Same sentiment. The worth-it guest churns at 2.7%. The rating gives no signal of the gap: the 17.3% and the 2.7% sit behind identical published scores on the same booking page.
denominator: reviews rated 8/10, neutral sentiment, with worth-it price verdict · cohort-117Among all guests rating 8 or higher (the population most investors and operators consider "satisfied"), the overpriced verdict still predicts 17.1% churn vs 2.6% for the non-overpriced population at the same score range.
denominator: reviews rated 8–10 with price verdict; among overpriced vs non-overpriced (residual price signal control) · cohort-117In the same cohort, Luxury underperformers reach 38 to 42%. The overpriced rate separates Luxury properties more cleanly than any friction metric in the dataset. (Anonymised: tier + country only, no hotel or location named.)
denominator: price-mentioning reviews per hotel, luxury segment; hotel-level comparison · private to named engagementsThe natural fix is to lower the rate. The data disagrees. 95.1% of overpriced reviews carry an identified friction event. Worth-it reviews: 34.2%. The difference is not the ADR; it is what happened during the stay.
Among guests who rated 8 or higher, signalled churn over price, and named specific items: 60.2% point to breakfast, parking, minibar, or other add-ons as the friction making the stay feel overpriced. The room rate is rarely the cited item.
denominator: guests rated 8+, price verdict = overpriced, named friction items, with churn signal · 259,647 reviews · 117 hotels
This matters for the decision shape. "Lower the rate" is a revenue action. "Remove the breakfast surcharge" or "fix the parking policy" is an operational action. The read names which specific ancillary is doing the damage, and the decision is different from cutting ADR.
STR covers the market. Reputation platforms cover the operator's rating. This finding lives in the third layer: the why under the rating, read from outside, for the seat deciding on the asset.
The finding is public. The named version, on a specific asset, with its specific ancillary friction identified, is a paid engagement. The next one is yours.
Volunteered, before you ask.
Everything above re-runs from the same public base. We'll show you the working before you ever pay for a number.
This page shows the pattern across the cohort. The named engagement shows you the overpriced rate for a specific property, which specific ancillaries are driving it, and the Wilson-CI-bounded decision case for changing the policy or the capex. One hotel you name, delivered in 14 days.
You just read a finding that no reputation platform will send you. Your pipeline is full of the same scores with the same verdict hiding inside.
STR reads the market. Reputation platforms read the operator's rating. What this page demonstrates is the third layer: the why under the rating, read from outside, for the seat deciding the asset.
You pay only if it surfaces a finding worth acting on; you set the threshold; if it doesn't clear it, there's no invoice. One hotel you name, delivered in 14 days.
Know someone underwriting a hotel on its rating? Forward this page. The number that breaks them out of it is the first one up.