Your advisor handed you a building report. Nobody handed you the guests.
Six sections. Every number traced back to the sentence a guest actually wrote. Every finding closed in a decision your committee can act on before you name the property. This is what a finished read looks like on an anonymized illustrative asset, so you can see the shape before you commit to one.
Six sections. Each one closes in a decision your committee can act on before you name a hotel.
Six sections in the finished read
Scope: 259,647 reviews, 117 hotels, 8 countries · illustrative property
The property described below is an illustrative composite (a 5-star, approximately 190-key coastal boutique hotel in Southern Europe), not a real hotel. Every number shown is a cohort-level pattern (259,647 reviews, 117 hotels, 8 countries, 25 to 26 languages) or a freshly generated illustrative figure, shown to demonstrate the structure and format of a read. On a named engagement, every figure is recomputed on the property you name, against its own bespoke peer set.
What you'd walk in already knowing · cover verdict
Does the non-return exposure justify adjusting the bid?Verdict summary
The published score is competitive for this tier and ADR band. Inside it, a measurable share of guests are signalling they will not return. The dominant signal is a value-verdict friction, not a service failure: guests write positively about the stay and still find the rate unjustified.
The exposure is concentrated in the breakfast and ancillary experience, not the room. The peer set in the same ADR band is split: two comparable properties show overpriced rates under 8%, two are in the 28–34% range. This hotel sits in the upper half of that distribution.
Upper-upscale cohort: 259,647 reviews / 117 hotels. Peer set built bespoke for the named property. Non-return rate = guests signalling will-not-return: read from what guests write, not from reservation data, it signals intent, never confirms a booking. Wilson 95% CI.
Does the non-return exposure at your tier floor justify adjusting the bid price or capex budget before hands change?
This is the opening page of the read, and the page that changes what you say in the room. Every candidate mechanism was generated, attacked, and either killed or confirmed before this verdict was written. The cohort's upper-upscale floor (10.2%, CI 9.4–11.1%) is the reference. A paid read opens with the named property's own figures against its own bespoke peer set, the exact number you would otherwise not have before the seller's broker does. On the asset you're weighing right now: does the non-return exposure justify adjusting the bid?
Before you accept the seller's comp set · peer-set position
Renegotiate this asset, or reprice the whole comp set?Overpriced rate among price-mentioning guests · upper-upscale peers in the same ADR band
Overpriced rate = share of price-mentioning guests who called the stay overpriced. Peer set: upper-upscale, same ADR band, same region (anonymized), freshly generated for this page; real peers are named in a paid read.
At 18.2%, this hotel is positioned in the upper half of its peer set but is not the worst performer. The two properties below it (28–35%) show markedly higher friction. The decision question is whether 18.2% is a brand standard acceptable for this flag, or an exposure that warrants a specific fix before or after a change of hands.
In a paid read, every peer is named, the denominator is stated, and the CI is shown per cell.
The competitive distribution. The peer set is built bespoke for the engagement (ADR band, tier, geography, read purpose); every peer is named, every denominator stated. The investigation then asks what no bar chart answers: is this hotel's friction property-specific, or a condition its peers share? On the comp set you've been handed for your own deal: would you renegotiate this one asset, or reprice the whole set?
What the loyalty trajectory actually shows · repeat-guest read
Is the non-return rate stable or accelerating?Cohort-wide, decline-reporting repeat guests signal non-return at a far higher rate. Recovery attempts do not appear to reverse it.
Once a repeat guest privately concludes the stay declined, breakfast is one of the clearest triggers and in-stay recovery gestures do not appear to reopen the verdict by checkout. See the full figures, with denominators and confidence intervals, in Study 01, Chapter 5 →
A paid read computes the named property's own trajectory. Non-return is read from what guests write, not from reservation data: it signals intent, never confirms a booking. The decline-to-departure link is correlational, not causal: the data cannot separate a failed recovery gesture from a guest who had already decided.
A composite sentence, in the register of the cohort finding, not a quote from any real guest: "The first cost line trimmed in a margin squeeze is the one returning guests audit. Churn stays elevated even among guests who received a recorded response."
Roughly 1 in 9 glowing reviews hides a maintenance note.
Among reviews rated 9 or 10 in the cohort, approximately one in nine narrates unresolved maintenance in the same sentence as praise. The per-property share runs from under 1% to over 20%, tier-independent: a standing diligence flag on any asset carrying a strong average.
Denominator: reviews rated 9–10 with maintenance narrative present (severity ≥ 0.4), n=45,937; per-property share across 117-hotel cohort (2024+).
Is the non-return rate stable, improving, or worsening in the trailing 12 months? The trajectory matters as much as the current score.
The repeat-guest read. The 5.8x breakfast trigger is not a field aggregation: it is what the investigation found when it asked which cost line a margin-squeezed operator trims first, and whether repeat guests notice. The maintenance finding follows the same logic: 1 in 9 high-rated reviews narrates building problems, and the per-property share runs from under 1% to over 20%, tier-independent. A paid read recalculates every figure on the property you name. On the asset on your desk: is its non-return rate stable, or is it accelerating?
The value-verdict read on this asset · the overpriced signal
Cut the rate, or name the capex line?The finding this section applies is the cohort's core value-verdict split: cohort-wide, guests who call a stay overpriced show a far higher non-return signal than guests who call the identical score worth it. See the full pair, with denominators and confidence intervals, in Study 01, Chapter 1 →
The instinct is to cut the rate. The investigation shows why that is the wrong fix.
Cohort-wide, among satisfied (8+) guests, overpriced perception predicts 6.3x more guests who won't return (18.8% vs 3.0%). The investigation generated the price-level hypothesis first and attacked it before accepting the friction read.
Lower the rate.
Among satisfied (8+) guests cohort-wide, overpriced perception predicts several times more won't-return guests than a "worth it" verdict at the identical score. The complaint is not the number on the bill.
Find the friction making the rate feel unjustified.
Among satisfied guests who show non-return risk over price and name a specific item, a clear majority point to ancillary charges: breakfast, parking, minibar. Not the room rate. Capex goes to the named friction.
On a transaction, the overpriced rate tells you whether the current positioning can hold. On a repositioning, it names the specific item to fix. On a portfolio comparison, it is one of the cleanest discriminators between outperformers and underperformers at the same tier.
The value-verdict finding, after adversarial testing killed the price-level hypothesis. The non-return multiple is the finding the investigation reached after testing and killing the price-level hypothesis: the ancillary-charge concentration had to hold before it shipped. See Study 01, Chapter 4 for the full figures. A paid read computes the subject hotel's own overpriced rate against its bespoke peer set in the same format. If your own asset is showing an overpriced signal: would you cut the rate, or name the capex line first?
What the friction map surfaces · named problems, closing in fixes
Where does the capex go first?Friction rate by area: the bar chart that closes in a capex decision, not a dashboard metric.
Each bar is a rate. Behind each bar, every friction event is resolved to six dimensions simultaneously:
The investigation reasons across all six: which frictions are property-specific vs peer-wide, which concentrate in a guest profile, and what the combined pattern says about NOI exposure. That reasoning is what closes the bar chart in a decision.
Friction rate = share of guests who mentioned the category with a negative emotional peak. Figures freshly generated for this page to demonstrate format, not copy-pasted from another page's real numbers. A paid read surfaces the subject property's own rates and ranks categories by impact on the overpriced verdict and non-return signal.
The read closes each friction area in a named decision
Highest friction rate and first item named in overpriced reviews. The complaint is the price-to-scope gap, not the food. Audit positioning before attributing to quality.
Narrated in the same review as praise, not in complaints. The per-room issue list, with recurrence rates per issue type, goes directly into capex triage.
Where a rival is named as better, the lines you control (food, value ladder) are the ones losing the guest most often. Location is closer to a coin-flip; a lower diligence priority.
Full figures with denominators, Study 01, Chapter 6 →
"For a 5-star, waiting 90 minutes for check-in because their system was down was unacceptable: I complained three times and nobody even apologised, so we’ll never come back." (illustrative composite, not a verbatim guest review)
See the resolved fields behind that one sentence · 4 structured events
root_cause = system · severity = 4 · was_resolved = no · guest_effort = high · operational_categories = [check_in, communication]
root_cause = communication · severity = 3 · was_resolved = no · guest_effort = high · service_recovery.attempted = false
churn_likelihood = will_not_return · return_barrier = quality · churn_risk = critical · is_detractor = true · emotion = Anger · signal_strength = 0.92
aspect = check_in · journey_stage = arrival · expectation_gap = true · expectation_anchor = star_rating · actionable_insight = "Fix check-in system reliability; train front desk on apology protocol"
37+ structured signals per review, structured extraction in a single pass, 25 to 26 languages.
The friction map: what closes a bar chart in a capex decision. Every bar is a rate among guests who mentioned the area, denominator stated. The investigation cross-cuts all six signal dimensions against guest profile, stay period, and the peer set, then kills every candidate that does not reproduce or that any senior operator would already know. What survives closes the bar chart in a capex decision, traceable to the source sentence. On your own capex list right now: where does the money go first?
What this read cannot tell you · honest perimeter
Can you underwrite against it?What the read cannot show you.
A read you can underwrite against tells you its limits before you ask. Every proportion carries a 95% confidence interval. A finding does not appear in the report unless it clears 39 reviews on the relevant dimension. If the sample is too thin for a reliable number, the section says so explicitly. That statement is also a finding.
GDPR: public reviews only, processed under Article 6(1)(f) legitimate interest. No hotel-systems data, no first-party guest information. The extraction method and the synthesis are mine; the raw material is the same public reviews anyone can read. A sub-processor list is available for institutional procurement on request.
The perimeter page that closes every read. The stat floor is the load-bearing item for an investment committee: every rate clears n ≥ 39 reviews and carries a Wilson 95% CI. A section with insufficient data says so. That statement is also a finding. Every read ends here, so what you carry into the committee room is exactly what you can defend under questioning, and nothing you cannot. Knowing what this read can't tell you: can you underwrite against what it can?
Every figure above, recomputed on the property you name.
The structure is the same. The peer set is built bespoke. Every number is the subject hotel's own, computed against comparable properties, not a cohort average, so the read holds up when someone on the other side of the table asks where the number came from. Three steps.
Name the asset
A target before a bid, a rival, an asset under an operator you are assessing. Tell me the property and the decision it needs to inform.
I confirm the peer set and the question
Peers chosen by ADR band, tier, geography, and read purpose (bid / reposition / operator review). Two business days to confirm the scope; then the investigation starts.
The finished read
One report, the six sections above, every number on the property you named, 14 days. Every finding reproduced, mechanism confirmed, non-obvious.
One hotel, 14 days. The terms, scope and price are agreed on a short scoping call first, so you know exactly what you are committing to before anything starts. One-off per engagement, no subscription, no onboarding, no recurring commitment.
What the closing page looks like.
Every finished read ends on one page built for the committee room, not the desk: three lines and a status row. This is the shape the page takes, not a finding on any hotel.
The shape every read closes on. On a named property, every bracketed line above becomes a specific, sourced sentence, and the chip row reflects that asset's own composite screen. Nothing above is a finding. It is the format the finding ships in.
"Every read is done by the person who built the method. I run the investigation on the hotel you name, I attack every candidate finding before it ships, and I stand behind what comes back. This is what that looks like."
Before you bid. Before you renovate. Before someone else asks the question you didn't.
A target, a rival, or an asset under an operator you are assessing. I recompute every section above on the property you name and deliver it in 14 days. A short scoping call to confirm the peer set and the question, then the investigation starts. What's the hotel you'd want this read run on first?
This page is a worked demo, not a finding: real cohort figures (259,647 reviews, 117 hotels, 8 countries) paired with freshly generated illustrative figures to show the structure and format of the deliverable. A paid read computes every number on the named property and its bespoke peer set. One-off per engagement, no subscription, no onboarding.