What a Hotel Review Score of 8.5 Actually Tells You

You are choosing between two hotels. One scores 8.5, the other 8.9. The second is better, obviously — except that it very often is not, and the number is close to the least informative thing on either page. Review scores are genuinely useful, but almost none of that usefulness is in the headline figure. It is in the sample size, the shape of the distribution, the dates and the subscores, and reading all four takes about two minutes once you know what you are looking at.

First, an honest note on sourcing

We could not read the booking platforms' own explanations of how their scores are calculated: Booking.com's partner help page for guest review scores returned 403 to an automated request on 6 September 2026, and its public reviews page returned 404. So nothing in this article describes any platform's weighting, its review window, its minimum-review threshold or how it maps a number onto a word like "fabulous". We would rather tell you that than paraphrase a third-party summary and present it as the platform's own rule. What follows is about how to read a score — which you can verify on the property page in front of you without trusting either us or the platform.

Why every score looks high

Because the people who fill in a review are not a random sample of guests, and because a rating scale that most people use only at the top end has very little room left to express a difference. The result is compression: a large number of properties sit close together near the ceiling, and the gaps that matter are smaller than they look. Once you notice this, the whole exercise changes shape. You are not reading a mark out of ten. You are reading a position inside a narrow band near the top of the scale, where a few tenths carry more meaning than they appear to and where anything genuinely low is a loud signal.

“You are not reading a mark out of ten. You are reading a position inside a narrow band near the ceiling — which is why the distribution tells you more than the average.”

Is 8.5 good?

It is a fine score and it tells you almost nothing on its own. Review scores cluster near the top of their scale, so the useful range is compressed into the last two points and a single number cannot separate a consistently decent hotel from an inconsistent one. Read the sample size, the spread of individual ratings and the subscores before you read the average. Two hotels can share an identical average and be completely different propositions: one where nearly every guest rated it around the average, and one where half the guests loved it and a quarter had a bad night. The first is a reliable choice. The second is a gamble, and the average will never tell you which you are looking at.

The four things to read, in order

Sample size: how many is enough

Enough that one memorable stay cannot move it. A handful of reviews is an anecdote — a single furious guest or a single delighted one swings the average visibly. Prefer a slightly lower score built on a large number of reviews over a near-perfect score built on very few, and be especially wary of a perfect score with a tiny sample. There is a simple test that needs no arithmetic: imagine one more review at the bottom of the scale and ask whether it would visibly move the number. If the answer is yes, the number is not yet describing the hotel; it is describing the last few people who stayed there.

Distribution: the part almost nobody opens

Most platforms will show you the breakdown of scores rather than only the average, and it is the single most useful click on the page. What you want is a distribution with a heavy top and a thin tail. What should worry you is a bimodal shape — a big group near the top and a distinct second group near the bottom — because that is the signature of inconsistency: a hotel with some excellent rooms and some poor ones, a service level that depends on which shift you meet, or a building with a good wing and a bad one. Averages are designed to hide exactly this, and you are the one who will be allocated a room.

Recency: a score is a historical document

A review score is an average over a period, and hotels change inside that period. Refurbishments, a change of owner, a new management company or the loss of a long-serving general manager all move the actual experience long before they move the score, because the older reviews are still in the average holding it up or dragging it down. Sort by most recent and read the last stretch on its own. If the recent reviews are noticeably better or worse than the average, believe the recent ones. If several recent reviews mention building work, believe them too, and ring the hotel.

Subscores: the actionable part

The subscores, and specifically the gaps between them. A property that rates highly for staff and cleanliness but poorly for location is telling you something precise and actionable, which the single average hides completely. The subscore that matters is the one attached to the thing you are actually buying. If you are there for two nights of sightseeing, location outweighs breakfast. If you are there to work, wi-fi and quiet outweigh the pool. If you are travelling with a baby, cleanliness and quiet outweigh everything. A hotel with a lower average and the right subscore profile is very often the better booking, and this is the one place where the review data genuinely beats your own judgement.

• Cleanliness low, everything else high: usually a real and consistent problem, and the hardest thing on the list for a hotel to fix quickly.
• Location low: frequently not a problem at all, because "location" often means far from a station rather than unpleasant. Check a map before you discount it.
• Staff low: the loudest warning sign in the set, because it is what you will meet when something goes wrong.
• Value low with a high overall: the hotel is good and you are being asked to pay for more than you need.
• Facilities high, comfort low: often a big resort with tired rooms — read the recent reviews for which building or wing.

Who wrote it, and what they wanted

Most platforms let you filter reviews by traveller type — solo, couple, family, business — and the filter is worth more than another ten minutes of reading. A business hotel rated by families will look worse than it is; a family resort rated by couples will too. Read the reviews from people who came for the same reason you are coming, and ignore the rest. Read the negative reviews from that group in particular, because what you are looking for is not whether people complained but whether they complained about something that would bother you. Somebody furious about the absence of a kettle is telling you nothing if you do not want a kettle.

Are the reviews real?

The big booking platforms restrict reviewing to guests who completed a stay booked through them, which makes fabricated reviews much harder than on open review sites. It does not make the score neutral: who chooses to write, what they were expecting and when they stayed all shape it. Treat a verified-stay score as real but biased, not as fake. The bias that matters most is who bothers. People write after an unusually good stay and after an unusually bad one, and the enormous middle — the stays that were simply fine — is under-represented on every platform there has ever been.

The honest counter-argument

All of this can be overdone. Beyond a certain point, review analysis is a way of feeling in control rather than a way of booking better, and the marginal value of the fortieth review is close to zero. A sensible ceiling: check the sample size, glance at the distribution, read the last handful of reviews and the two subscores that matter for your trip, then decide. If two hotels survive that, book the one with free cancellation and stop — you can re-shop it later, which is the method in how to get the best hotel rate, and the flexibility is worth more than the fifth decimal place of a score. The reasons to prefer a flexible rate in the first place are in free cancellation versus non-refundable.

One last thing a score will never tell you: what the stay will actually cost. That is a separate exercise, and it starts at your hotel bill is always bigger than the price you booked, with resort fees and city and tourist taxes as the two most common surprises. And if you are choosing between platforms as well as hotels, Expedia versus Booking.com covers how the same property can total differently on each.

Frequently asked questions

Is 8.5 a good hotel review score?

It is a fine score and it tells you almost nothing on its own. Review scores cluster near the top of their scale, so the useful range is compressed into the last two points and a single number cannot separate a consistently decent hotel from an inconsistent one. Read the sample size, the spread of individual ratings and the subscores before you read the average.

Why do hotel review scores all look so high?

Because the people who fill in a review are not a random sample of guests, and because a rating scale that most people use only at the top end has very little room left to express a difference. The result is compression: a large number of properties sit close together near the ceiling, and the gaps that matter are smaller than they look.

How many reviews does a hotel need before the score means anything?

Enough that one memorable stay cannot move it. A handful of reviews is an anecdote — a single furious guest or a single delighted one swings the average visibly. Prefer a slightly lower score built on a large number of reviews over a near-perfect score built on very few, and be especially wary of a perfect score with a tiny sample.

Which part of a review score should I actually read?

The subscores, and specifically the gaps between them. A property that rates highly for staff and cleanliness but poorly for location is telling you something precise and actionable, which the single average hides completely. The subscore that matters is the one attached to the thing you are actually buying.

Are hotel reviews on booking platforms fake?

The big booking platforms restrict reviewing to guests who completed a stay booked through them, which makes fabricated reviews much harder than on open review sites. It does not make the score neutral: who chooses to write, what they were expecting and when they stayed all shape it. Treat a verified-stay score as real but biased, not as fake.

No platform's scoring formula, review window, minimum-review threshold or score-to-label mapping is described anywhere above, because the platforms' own help pages refused automated requests on 6 September 2026 and we will not paraphrase a third-party summary as if it were the source. Everything here is about reading what is on the page in front of you. Reviewed by the NorwegianSpark SA editorial team.