Every percentage carries a Wilson 95% CI and a row count; every quote is row-cited on the engagement deliverable.
The guest reviews — read for the owner.
We read every public review a property has and hand you a finished analysis — the problems quietly draining the asset's economics, and the decision each one forces. Any property, from public reviews alone.
A rating average flattens the reviews to one number.
The asset turns on what it leaves out.
Guests talked about breakfast, and the score dipped. An average can't tell you which of three things that is — and the difference is what moves the asset.
Each review gets read end to end — 40+ structured fields, in its own language — then read again as a set. Everything below comes out of that, anchored to the exact sentence behind it.
One review, turned into a decision.
One real public review. A score flattens it to a flag and an average. Hover any highlighted phrase and watch the field it becomes light up — then watch the fields add up to a decision.
What the rating returns
A low score and three tags. They tell you it went wrong — not what it costs, or what to fix.
What the read returns
Eight Porto hotels. One rating, five operator grades — and the rate doesn't track the grade.
All eight sit at roughly nine out of ten on guest rating — a search filter lumps them together as peers. The grade runs A–F on the guest-experience signal in each operator's reviews, cohort-relative, beside the rate it charges. Anonymised A–H, and every grade traces back to the guests' own words.
| Hotel | ADR | NLS | Operator culture | Service recovery | Repeat guests | Overall |
|---|---|---|---|---|---|---|
| Hotel B | €679 | +50 | A | A | B | A |
| Hotel A | €520 | +45 | A | A | A | A |
| Hotel G | €249 | +33 | B | C | C | B |
| Hotel C | €268 | +37 | C | C | C | C |
| Hotel D | €248 | +25 | C | D | D | D |
| Hotel F | €207 | +35 | D | C | C | D |
| Hotel E | €592 | +41 | C | D | D ↓ | D |
| Hotel H | €708 | +18 | F | F | F | F |
8 Porto premium hotels (upper-upscale to luxury) · 18,550 reviews · all clustered at ~9/10 guest rating · grades are cohort-relative. ADR = average offered nightly rate. NLS = net loyalty score (−100 to +100), a strict-language read of repeat intent. D ↓ marks a property declining since an ownership change. The set's two highest rates sit at opposite ends of the grade.
Three reads a rating average can't produce.
Each finding traces to the verbatim behind it, carries its window and sample size, and closes in a decision the right seat can act on.
Repeat guests identified by comparator language; window and n labelled on every shift claim.
Association, not a causal claim — verified across all eight hotels, each above the sample floor (Newcombe 95% CIs).
Every finding closes in a number you can defend.
We don't hand you a sentiment score. We hand you a decision — and the number behind it, worked on your own ADR, with the math shown so your analyst can check it.
Finding7.6% of repeat-guest reviews say "worse than last time" — and 81% of them won't return, yet many still rate 8/10, so a rating average misses it entirely.
BridgeLost repeat room-nights × your ADR × repeat-guest lifetime value. Repeat revenue carries near-zero acquisition cost — the most expensive kind to lose.
≈ a low-six-figure annual envelope · illustrative on a 100-key asset · the figure on your asset is computed in the paid readDecisionSize the retention envelope and act before the rating moves.
FindingReviews naming a staff member are 44–60pp more likely to carry explicit advocacy. The goodwill concentrates around a handful of people.
BridgeWhen a named advocacy-driver leaves, a measurable share of return intent is exposed — a cost no rating average can see.
DecisionProtect and price the people guests come back for.
The read shows the decision and the real rate. Your asset-specific € is computed on the paid deliverable, on your actual ADR, occupancy and repeat-rate. We under-claim on purpose — the number you can put in a memo beats the number that looks good on a slide.
Their tools answer to the front desk. We answer to the owner.
STR covers the market; the operator's reputation tool covers the rating. We sit on the other side of the table — the why under the rating, for the seat deciding on the asset.
The reputation & ops layer.
Buyer: GM, front office, brand reputation manager. Job: respond to reviews, raise the score, train the team.
- SAASPer-property monthly subscription; the hotel is the customer.
- DATAAggregate scores and category sentiment; single-number indices.
- FORMAn always-on dashboard the team logs into.
- NEEDSYour own property, as a paying client.
What lands in the underwriting file.
Buyer: asset manager, transaction advisor, operator-selection advisor, owner. Job: decide on the asset — buy, hold, swap, intervene.
- MODELProject-priced engagement, designed to drop into an investment memo.
- DATAPer-review structured fields — which cohort leaves in silence, which staff carry advocacy, what moves before the rating.
- FORMOne designed PDF per engagement — the product is the read, not access.
- NEEDSNothing from the hotel. Any property, from public reviews — yours, a rival's, a target.
A PDF your committee will actually read.
The cohort above, exactly as it ships: the grade, the quotes behind it, the decision each one points to. Asset-side or operator-side, built for your seat.
What it is — and what it isn't.
Where the read works
No operator cooperation. No PMS feed, no NDA, no site visit — public reviews only. That's what lets us read a rival or a target you have no relationship with.
Periodic, on purpose. Watching your own book day to day is the operator platform's job. We're the outside read on the properties it can't reach — targets, rivals, a hotel that just changed hands.
What reviews don't capture — and what we won't claim
We name the lever and the decision; we don't cost the build. That's your cost engineer. And whether your team executes is your call.
A thin finding stays directional. Anything resting on a handful of reviews is flagged, never sold as fact — every shift claim carries its window and its n.
From brief to deck in 10–14 days.
Four steps, each with a date attached. No sales call to sit through.
The asset, the question, the peer set.
You send the property (or portfolio), the question the deck should answer, and a peer set if you have one. We propose one if you don't.
Engagement note in two business days.
One page: scope, peer-set confirmation, deliverable shape, fixed price, dates. You confirm or adjust.
One designed PDF deck.
Anchored on the agreed peer set, calibrated to the question. No raw extracts — the deck is the deliverable.
Recurring cadence (optional).
For portfolio mandates: the same deck refreshed quarterly, or anchored on diligence windows. Same engine, same definitions.
Two founders. One read.
Rui Andrade
MSc in Informatics Engineering, University of Porto. Built the extraction pipeline and the read end to end — and has personally read more hotel reviews into structured data than he'd care to count.
José Andrade
PhD candidate in hotel intelligence at the University of Vigo (Spain), after 26 years in industry. He keeps the read honest against how hotel markets actually behave, and owns the methodology behind each case.
Name an asset. Get the read.
Name an asset — yours, a rival's, or one you're weighing — and we'll read it against its real peer set. That's the whole first step.
Tell us who you are and what you're working on — a specific asset, or just the question on your mind. We reply within two business days.
No demos · project-priced · two assets, €4,000, and you're the judge of whether it earned its keep
No asset in mind yet?
Book a 20-minute call — no brief required. We'll talk through what you're weighing and whether a read would actually earn its place.
Book a call