How the Reality Index is calculated
Two numbers for every hotel, not one: a Reality Index built from signals that are hard to fake, and a separate Trust Index — whether these reviews can be trusted at all. Platforms show one pretty figure; we deliberately show two, and neither has to match the platform's own score.
Recent impressions count more than old ones: a review from last month matters more than one from two years ago. Few recent reviews — we say so honestly.
If one category (service, say) sharply dips below the overall score, the index shows it — instead of hiding it behind a pretty average.
Spot a fake-looking review burst or a serious complaint (bedbugs, theft, etc.) — we don't bury it inside a number, we flag it.
Trust Index — a separate 0–100 number next to the index. It's about how reliable the reviews are, not how good the hotel is, so it deliberately doesn't move in lockstep with the platform score.
How the index is assembled
This is not a simple weighted average. First we take a freshness-weighted average of real reviews — the fewer reviews there are, the more cautiously the score is pulled toward the mean (so a hotel with 5 reviews can't shoot to the top on one lucky five-star). Then two honest correctors kick in: a category gap and a trend. That's why there's no high index without real confirmation — even for a hotel with a 9+ platform score.
Review freshness
baseA review from the last 3 months weighs 1.0, from 6 months ago — 0.7, a year ago — 0.45, and very old ones barely matter (hotels change, and old opinions go stale). The weighted average is cautiously pulled toward the mean when there's little data — the hotel gets honestly flagged "thin data".
Category gap
corrector · up to −15%If the overall score is high but one category (say, "Service") is noticeably lower than the rest, the index honestly drops. This is the difference between "9.4 on the platform" and "38 in reality" — see the journal article.
Trend
corrector · up to −10%We compare recent reviews (≤3 mo) with older ones (6 months to a year ago). If recent ones are noticeably worse, the hotel has likely gotten worse, and old good reviews shouldn't mask that.
Trust index
separate numberNot a corrector on the index above — a second, fully independent 0–100 number: review volume, agreement across sources, language diversity, and the absence of bursts or manipulation. It's deliberately not tied to the score itself, so it doesn't have to match the platform's rating the way the index on the left does.
Toggle these — the index recalculates with the real engine right here.
What we show separately, instead of hiding it in the number
Some signals are too important to dissolve into a single figure. Instead of a silent penalty — a visible warning next to the index, and the decision stays with you.
Critical complaints
Bedbugs, theft, unsafe — if even one review mentions it, we won't let it get lost among hundreds of good ones.
Suspicious review pattern
A burst of dates in one week, duplicated text, a complaint in the text paired with a 10/10 score — looks like manipulation, and we flag it as such.
Guest-type breakdown
Not a penalty — honest information: "great for couples, weaker for families" tells you more than one averaged number.
Trust index
A 0–100 number, separate from the index itself. Built from review volume, consistency, source agreement and language diversity — and deliberately independent of the score's magnitude.
What we deliberately don't count
These things make other platforms' ratings pretty but dishonest. Here they don't affect the index — at all.
The hotel's promo photos
Polished shots from the press kit don't add a single point. The only proof is live guest photos.
Paid promotion
A hotel can't buy a higher spot. Affiliate commission has no effect on the index.
Stars and category
"Five stars" is about the service class, not about whether guests actually liked it. We look at the guests.
The hotel's self-description
"A cosy boutique hotel in the heart of the city" is marketing. What goes into the index are the facts from reviews.
Few reviews — too early to judge. We honestly flag such hotels as "thin data" and show the index as indicative rather than hiding the emptiness behind a pretty number.
Where the data comes from
Signals are collected from public reviews and guest photos via LiteAPI, and our engine calculates the index from them. No manual tweaks in a hotel's favour.
Collecting signals
Reviews, category scores and guest photos for a specific destination.
Checking
We look for date bursts, duplicated text, and a clear mismatch between score and complaint — and flag it, not hide it.
Calculating the index
A freshness-weighted average, adjusted by category gaps and trend, gives a 0–100 score and a verdict.
Updating
The index is recalculated regularly — fresh reviews change the picture.