Across 117 European hotels and 259,113 guest reviews, 1 in 13 of your repeat guests is writing it — and 81% of them tell us, in the same review, that they won't be back. Your dashboard reports +76 top-line loyalty. Underneath sits a cohort at −60.
Of 59,090 repeat-guest reviews across the cohort, 7.6% (4,468) explicitly write that the property got worse than their last stay. 81% indicate, in the same review, that they will not return. The dashboard reports the top-line average. It cannot surface what guests wrote in the comparison field.
Top-line loyalty score
The aggregate of all repeat-guest reviews, averaged by rating. Visible on every dashboard.
Decliner-cohort score
The 7.6% writing "got worse." 81% won't return. The dashboard reports the average; the cohort underneath sits here.
For a 100-key property at €250 ADR with a 23% repeat-share, the retention envelope around this cohort is roughly €180,000 per year. At properties carrying 15-22% decline rate (the worst five in the cohort), the envelope reaches €350-540,000.
Top change-detail tokens across the 4,468 decliner reviews:
Each finding survives within-hotel paired comparisons and sample-size gates — n ≥ 39 per claim (Wilson), Wilson 95% CIs on proportions, Newcombe intervals on paired differences.
7.6% of repeat guests write that the property got worse. 81% won't return — in the same review.
+76 top-line. −60 cohort underneath. The dashboard cannot surface what guests wrote in the comparison field. ~€180K/year retention envelope at a 100-key property; €350-540K at the worst five.
When check-in fails AND staff service fails in the same stay, 82% of those guests detract.
Independent probability would predict 48%. The +33.9 pp excess is compounding. Recovery, once both have failed, doesn't work — 89.5% detractor with recovery attempted vs 85.3% without. The window closes before staff intervene.
On wedding, anniversary, honeymoon stays — one friction event lifts the detractor rate 5–8× baseline. Standard recovery makes it worse.
Wedding + friction = 7.7× baseline. Honeymoon 5.3×. Anniversary 4.3×. Recovery attempted = 22.2% detract vs 18.6% without. Generic scripts read transactional to guests who said "this stay matters."
Top-quartile hotels mention 4.2× more distinct named staff per 100 reviews than bottom — controlled for review length.
Top quartile averages 10.2 named individuals per 100 reviews. Bottom averages 2.4. Proactive recovery narration is 3.3× richer. Friction prevalence is 45% lower. None of this is capex.
ADR and verified guest loyalty are uncorrelated. Top-decile-ADR luxury properties sit in our bottom loyalty quartile. Mid-ADR Porto hotels lead.
Capex, ADR, brand flag, location anchor — none separate the top quartile from the bottom. The five-star price tag does not, on its own, deliver the five-star loyalty.
When friction occurs and staff narrate the recovery, advocacy more than doubles. Only 14% of friction reviews carry the narration.
Narrated-recovery reviews advocate at 13.95% vs 5.60% for un-narrated. Narrated recovery outperforms clean (friction-free) promoter reviews on advocacy. Fixing is the floor. Narrating is the ceiling.
"For the asset side, the question is no longer what the rating is. It is what the cohort underneath is writing — and whether they are leaving."
A natural-language analysis layer reads every review across Booking, TripAdvisor, Expedia and Google. Forty-plus fields per review. Run unattended on the full review window.
Luxury, upper-upscale, upscale, midscale — the full price spectrum European hotels operate at.
Portugal, Spain, Italy, Switzerland, France, the Netherlands, Brazil, Tanzania.
Booking.com, TripAdvisor, Expedia, Google — read natively across languages, no translation step.
2003-06-19 → 2026-05-04. Recent-12-month sub-tables anchor all current-state claims.
Sample-size floor n ≥ 39 per claim (Wilson 95%, ≤15pp half-width at p=0.5). Wilson CI on proportions. Newcombe intervals on within-hotel paired differences.
Every numeric claim re-runnable from a single export bundle. Within-hotel paired comparisons defend against tier-mix and cohort-selection confounds.
You've read the cohort. The next step is a scoped read on a property or portfolio of yours — anchored on its actual peer set.
Tell us who you are, the asset (or portfolio), and the question the deck would answer. We reply within two business days with a scoped engagement note and timing.
No demos · No free property analyses · Project-priced engagements only