How the HRI Works

Every hotel rating you have ever used is broken in the same few ways. A 9.2 from 23 reviews is displayed exactly like a 9.2 from 12,000. Google does not know what TripAdvisor says about the same property, so a hotel that looks great on one platform and mediocre on another shows you only the flattering half. A renovation from 2024 drowns in averaged-in complaints from 2018. Expert recognition, the Forbes inspections, the World's 50 Best lists, the awards that actually involve someone checking, sits in a press release instead of the score. And one number is supposed to mean the same thing to a honeymooning couple and a family of five.

Rating agencies solved problems like these for bonds a century ago. Search engines solved them for web pages twenty-five years ago. Nobody has solved them for hotels, mostly because the companies big enough to do it make their money from the hotels being rated.

We built the HRI to fix that. This page shows you exactly how it works. No rating platform publishes its formula. We think that is precisely the problem, so here is ours.

The HRI at a Glance

The Hotelsrating Index is a single score from 0 to 100, computed nightly for every hotel in our database, from four components:

The weighted sum is then adjusted for data confidence and calibrated against the full corpus of hotels, so that a score of 62 always means "the median hotel in the world" and a score of 90 always means "top 5%".

Every hotel page includes a "Why this score?" breakdown showing each component's value. If you disagree with a score, you can see exactly which part of it you disagree with.

Component 1: Consensus (CRS), 40%

We aggregate ratings from Google, TripAdvisor, Booking.com, Expedia and professional databases. But we do not average them, because averaging is where other ratings go wrong. Four corrections are applied.

1. Small samples are pulled toward the platform's norm. A 9.6 from 8 reviews is not a 9.6. We apply Bayesian shrinkage: a hotel's platform score is blended with that platform's average, weighted by review count. A hotel with 8 reviews moves most of the way toward the platform norm; a hotel with 5,000 reviews keeps its score almost untouched. The crossover point is 35 reviews. This is the statistically honest replacement for the raw average, and it means new or tiny properties cannot rocket to the top on three enthusiastic reviews.

2. Each platform is normalized against itself. Booking.com scores cluster around 8.1. Google clusters around 4.3 out of 5. Comparing the raw numbers is meaningless, so we convert every platform's scores to standard deviations from that platform's own mean before combining them. A hotel is "good on Booking" only relative to other hotels on Booking.

3. Platforms are merged by trust and evidence, not raw volume. Sources are weighted by a published trust factor and by the logarithm of review count, so 12,000 Google reviews inform the score more than 200 TripAdvisor reviews, but cannot completely silence them.

4. Disagreement between platforms is a warning, and we treat it as one. When three or more platforms cover a hotel and their normalized scores diverge, the consensus score is pulled toward the middle of the scale, by up to 35% of its distance from it. A hotel that Google loves and TripAdvisor does not is a hotel we are less sure about, and the score says so. No other rating system looks at cross-platform disagreement at all.

Component 2: Expert Anchor (EXP), 35%

This is the part of the HRI no one else has, and the reason small exceptional hotels can outrank big adequate ones.

Awards become score. We maintain a verified registry of expert recognition: Forbes Travel Guide star ratings, the World's 50 Best Hotels, World Travel Awards, Condé Nast Traveler lists, Travel + Leisure World's Best, TripAdvisor Travelers' Choice, and Michelin Keys. Each carries a published prestige weight:

Weights reflect our assessment of each program's rigor and independence, including how much an award depends on actual inspection versus marketing participation. Awards decay with a half-life of three years: a 2021 award counts for roughly a quarter of a 2026 one. Excellence has to be maintained.

Trust propagates through a graph. Here we borrowed the core insight of PageRank, the algorithm that made Google work: authority is not a property of a page in isolation, it flows through a network and settles into a stable ranking. We build a graph where hotels are nodes, connected by shared recognition (awarded by the same body in the same year), shared ownership or brand, and geographic peer groups. Award-holding hotels seed the graph with their prestige, and trust then propagates iteratively until it converges, exactly the way PageRank converges on the web.

The practical effect: a boutique hotel that keeps company with the world's best, through its group, its repeated co-recognition, its peer set, inherits a measurable fraction of that standing. A hotel chain's reputation flows, diluted by the size of the chain, to its individual properties. Connections are normalized so that no single award cohort or mega-chain can flood the graph. And a hotel with no awards and no notable affiliations simply gets nothing from this channel, which is as it should be.

Component 3: Delivery (FAC), 15%

Star classifications promise things. This component checks whether the property delivers them.

For every combination of star class and location type (beach, city, mountain, countryside), we compute what the peer group actually offers across a curated set of 18 meaningful facilities, pools, spas, gyms, accessibility, kids' facilities, business infrastructure and the like. Self-reported decorative amenities are excluded; we only count facilities that matter and that we can increasingly verify.

A hotel that delivers more than its class and setting promise earns a bonus, capped so it cannot be gamed by checkbox inflation. A five-star property missing what five-star peers consider standard takes a penalty twice the size of any bonus. Underdelivery is worse than overdelivery, and the formula says so.

Component 4: Trajectory (TRJ), 10%

Hotels change. A great GM arrives, a renovation completes, an owner starts cutting corners. Trajectory compares recent consensus against the prior period and rewards properties that are demonstrably improving.

Current status: this component requires dated individual reviews, which we are in the process of ingesting through official APIs. Until then, trajectory is held neutral for all hotels and the score is driven by the other three components. We say this here because pretending otherwise would violate the next section of this page.

Confidence, Calibration and Medals

Confidence. Every hotel gets a data-confidence assessment based on how many platforms cover it, total review volume, verified information, recognition history and brand affiliation. Crucially, low confidence does not destroy a score, it makes the score conservative: thin-data hotels are blended toward the median of their star class rather than punished toward zero. An excellent hotel we know little about reads as "probably around average for its class until proven otherwise", not as "bad".

Calibration. The final score is mapped onto a fixed scale against the entire corpus: the median hotel in the world sits at 62, the top 5% begins at 90, the top 0.5% at 97. These anchors are recomputed nightly and published, so the scale cannot silently drift and a 90 means the same thing this year and next.

Recalibrated nightly · last update 2026-06-10 15:16 UTC

Medals. Gold, Silver and Bronze are percentile-based: Gold is the top 3% of all hotels, Silver the next 7%, Bronze the next 15%. Medals are earned against the whole world, not bought, and not negotiable. There is no process by which a hotel can apply for one.

Personalization: One Formula, Your Axis

The HRI is the same for everyone, it is the hotel's intrinsic quality. But fit is personal. If you choose to tell us how you travel (family, honeymoon, remote work, nightlife and so on), we compute a Personalized Fit Index on top of the HRI.

Two rules keep this honest. First, fit can only re-order, never inflate: a hotel's personalized score can be reduced by up to 25% if it clearly does not match how you travel, but it can never exceed the hotel's HRI. A bad hotel cannot become good by matching your vibe. Second, the hotel's underlying HRI is always displayed alongside any personalized view.

Verified Inspections

Hotels may commission an independent on-site inspection: a standardized 40-60 point protocol with mandatory geotagged, time-stamped video evidence, scored by both our AI pipeline and human review. Verified properties carry the HRI Verified badge, and verified facts replace self-reported ones in the Delivery component and raise data confidence.

Three things about inspections, stated plainly:

• The hotel pays for the inspection taking place. It does not pay for the result, and the price is identical whatever the result turns out to be.
• The hotel cannot choose its inspector. We assign inspectors, we pay them a fixed fee regardless of findings, and their long-term standing with us depends on their accuracy, not on hotel satisfaction.
• The results publish either way. If the inspection finds that the advertised spa is a massage chair in a converted storage room, that is what the data will now say. There is no option to suppress an unfavorable outcome, and commissioning an inspection constitutes acceptance of this.

An inspection can lower a hotel's score. It has, and it will again.

Integrity Policy

This is the part of the page our business model answers to.

1. The score is not for sale. No payment of any kind, inspection fees, subscriptions, advertising, marketplace commissions, partnership revenue, affects any hotel's HRI, medal, or position in any ranking. There is no premium placement, no sponsored ranking slot, and no rate card for visibility within rated results. There never will be.

2. Commercial products and the rating are separated by a firewall. We earn money from analytics subscriptions for hoteliers, from independent inspections, from a clearly labeled media marketplace, from booking affiliate links, and from data licensing. None of these flows into the formula. Specifically: affiliate commission rates play no role in ranking; sponsored creator content is labeled, lives in a separate section of the hotel page, and is never an input to the HRI; analytics subscribers see their score explained, not improved.

3. Inspections follow the assessor model. Like Forbes Travel Guide inspections and unlike pay-for-badge schemes, payment buys the assessment, never the outcome. Inspector assignment, fixed inspector compensation, and unconditional publication are described above and are non-negotiable terms of the service.

4. Attempts to buy the score are logged and refused. Any request to alter, suppress, or guarantee a score or medal in exchange for payment, business, or content spend is declined and recorded in an internal refusal log. Patterns of attempted manipulation (review flooding, facility misreporting, coordinated pressure) can themselves reduce a hotel's data confidence.

5. Hotels have the right to reply. Any hotel may claim its page free of charge, respond publicly to its score breakdown, and submit factual corrections (a closed pool, a new spa, a completed renovation) with evidence. Factual corrections are reviewed and, where verified, change the underlying data, which is the only legitimate way a score changes.

6. The methodology is public and versioned. The formula, the weights, the prestige table and the calibration anchors are published on this page. Changes go through the changelog below with effective dates. We will not adjust the formula quietly, and we will not adjust it for anyone in particular.

7. We rate ourselves out of conflicts we cannot firewall. Hotels owned by, operated by, or commercially entangled with our company or its principals beyond arm's-length services are flagged as such on their pages, and we will disclose the nature of the relationship.

What the HRI Is Not

Honesty about limits is part of methodology.

• It is not a guarantee. The HRI is a statistical synthesis of evidence. Hotels have bad nights, and data lags reality by days to months depending on the signal.
• It is not complete. Coverage is thinnest for small independent properties in less-traveled regions, which is exactly where confidence adjustment matters most. We state confidence rather than hide it.
• It is not yet measuring trajectory. As noted above, the trajectory component is neutral until dated review ingestion completes. The changelog will record the day it goes live.
• It is not opinion-free. The prestige weights and component weights encode judgments, ours, about what rigor and quality mean. We believe publishing those judgments, with the reasoning, is the most any rating can honestly do. The alternative is encoding judgments and hiding them, which is what everyone else does.

Versioning and Changelog

Material changes to weights, prestige values, or calibration anchors will appear here with effective dates and a brief rationale. Scores recompute nightly; methodology changes do not.

Questions, Corrections, Challenges

If you believe a score is wrong, the breakdown on the hotel's page will show you which component to challenge, and the claim process above is how to challenge it. If you believe the methodology is wrong, we want to hear that too: [email protected].

We will not always agree with you. But you will always be able to see exactly what you are disagreeing with.