We Rated 100 Hotel Stays. Our Favourites Weren’t Always Worth Repeating.

Some of the best hotels we’ve ever stayed in are places I wouldn’t book again. They were worth the price once. Not twice.

That sounds like a contradiction, so let me put numbers on it. Mel and I went back through 100 of our hotel stays, and for every one I answered a single question: knowing what I know now, would I hand over the same amount again?

Fifty answers were Yes. Twenty-four were “Only if it were cheaper.” Twenty-six were an outright No. So, for half our stays, I wouldn’t pay the same again.

We expected the No pile to be full of bad hotels, and a few were. What we didn’t expect was to find some of our favourite places in the world sitting in the middle pile, looking slightly sheepish.This is the story of how we worked out why, and how it reshaped the TripWorthIt™ scorer. 

We thought the problem was the star ratings

Every hotel we’ve ever booked looked brilliant on paper. A 9.2. “Exceptional.” Hundreds of reviews from people who loved the breakfast.

So, our starting assumption was simple. Crowd ratings are inflated, and we could do better.

What we could actually show was narrower than that. For every stay we gathered the platform scores from Booking.com, Agoda, Google, Tripadvisor and Expedia, backed by roughly 667,000 reviews between them. On Booking.com our hotels ranged from 7.1 to 10.0, and 83 of the 95 it rated were Fabulous, Superb or Exceptional. When almost everything lands in the top three bands, the bands stop separating much.

That’s bunching, though, not proof of inflation. These were hotels we chose, often because they scored well, so a high average is partly our own doing.

Then the sharper number. Booking.com rated 54 of our hotels Superb or Exceptional. I’d happily pay the same again for 31 of them. Fifteen I’d only rebook if they were cheaper, and eight got an outright No. One of those eight had a perfect 10.0.

Are Booking.com ratings reliable?

Yes, for what they measure. They’re honest averages, and they’re good at warning you when something is wrong. They’re much weaker at telling you whether a hotel is right for you, at your price.

Three things stood out in our data:

• Fewer reviews, less certainty. That perfect 10.0 came from seven reviews. In our sample, hotels with under 300 Booking.com reviews averaged 9.2, against 8.8 for those with over 1,000, and their scores showed no link to whether we’d go back. Above 1,000 reviews, the score lined up with my answers moderately well. That’s around 30 hotels in each group, so read it as a pattern in our stays rather than a rule for yours.
• The sub-scores move together. Staff, Facilities, Cleanliness, Comfort and Value rise and fall closely together, on Booking.com and on Tripadvisor, and far more closely than our own pillars do. In our sample, they added relatively little beyond the headline score.
• Location is the exception. It was the only sub-score that measured something different, and the only one that agreed with our own Location scores.

None of this makes reviewers dishonest. A crowd score blends the couple who stayed a week, the businessman who stayed a night and the guest who happened to get the lovely waiter. Booking.com does ask about value for money, but it averages that across everyone who paid whatever they paid.

So that was the story we set out to tell. The crowd versus us.

It turned out to be the wrong story.

Instead, “worth the price” turned out to be three questions

The surprise wasn’t in the crowd’s scores. It was in ours.

When we rescored all 100 stays properly (more on how below), the hotels I’d rebook “only if cheaper” scored slightly higher on average than the ones I’d happily rebook: 3.63 out of 5 against 3.58.

They weren’t worse hotels. Several are among the best we’ve ever stayed in.

Four of our ten highest-quality stays sit in that middle pile: Emerald Faarufushi and JA Manafaru in the Maldives, Golfo Dulce Retreat in Costa Rica, and Taru Villas Rampart Street in Galle. All four also beat our price benchmark: each delivered more than stays at a similar price in our own collection usually did.

And I still wouldn’t pay it again.

That’s when it clicked.

First: “Was it worth it?” is really three questions wearing one coat:

Second: “Was it good?” How well did the hotel deliver, pillar by pillar?

Third: “Was it good for the price? Did it deliver more than stays at that price usually did, in our own records?

Would I buy it again? Knowing what I know now, would I hand over the same amount?

The answers don’t have to match. Some stays are worth it exactly once.

What does “worth the price” actually mean for a hotel?

For us it now means two separate things. The first is a comparison: what did this hotel deliver against what similar money bought us elsewhere? The second is a decision: is this a price I want to pay again, for this experience?

A villa at £800 a night can beat every comparable stay we’ve had and still be a once-in-a-decade purchase. An £80 villa can deliver so far above its price that we’d rebook tomorrow without a second thought.

The comparison you can calculate. The decision is personal, and no rating can make it for you.

Before we trusted any of that, we had to fix our own scores

Here’s the uncomfortable bit. When we first scored our hotels, we weren’t nearly as objective as we thought.

We wore pink-tinted glasses

We adore Emerald Faarufushi. Twice, we scored its Location higher than Velassaru’s.Getting to Emerald means a seaplane, and Maldives seaplanes only fly in daylight. Land at Malé after dark and your first night is somewhere else. Velassaru is a short speedboat ride from the airport, at any hour, with the spray coming over the bow and the island already in sight. 

We loved the hotel, and the love leaked into a score that should have been a plain matter of access.

Whole numbers left no room for “a bit better than fine”

Our early scores were almost all 3s and 4s. A hotel that was slightly better than fine had nowhere to go. On those whole-number scores, our own model matched my rebook answers less than half the time.

“Good” meant something different to each of us

Mel and I scored separately, and Location is where we disagreed most. One of us was scoring how easy it was to get around. The other was scoring whether we’d wanted to leave at all.

Both were real impressions. They weren’t the same assessment.

How do you score a hotel fairly?

You describe every score before anyone gives one. I borrowed that from the appraisal systems I used to sit through in my risk-modelling days: agree what a 3 looks like, and what a 5 looks like, before anyone starts.

TripWorthIt™ now scores five pillars, in this order:

• Location & Access: how easy it is to arrive, get out, and reach what people come to the area for
• Service & Ease: how smoothly the stay runs: check-in, housekeeping, the Wi-Fi
• Hotel Offer: the food, the drink and the extras
• Rest & Recharge: how well you sleep and switch off
• Property & Setting: the grounds, the pool, the place itself

Every pillar has a plain question, three things to check, and a description for each point on the scale:

1. Seriously deficient
2. Below standard
3. Does the job: nothing wrong, nothing special
4. Consistently strong
5. Exceptional and difficult to improve, and meant to be rare

Scores are now decimals too, so “a bit better than fine” can finally be a 3.4.

We tested the new wording on 12 hotels first, scoring separately, then talked through the three where we differed. After that conversation, 90% of our scores landed within 0.2 of each other. Then we rescored all 100. Across the full set our agreement was looser but held up: every pillar correlated at 0.81 or better, and our average gap was about a quarter of a point.

Emerald’s Location now sits below Velassaru’s, where it belongs. Our rebook model now predicts my answer about two times in three on stays it wasn’t fitted to, where always guessing Yes gets you half. (We changed the prompts and the decimals at the same time, so the decimals can’t take all the credit.)

As for the rankings, nine of our top ten for quality survived the rescore. Further down, about 20 hotels moved twenty places or more. An £80 villa, Agandau House, climbed from 31st to 7th for value. Rescoring didn’t change who was best. It changed how sure we could be about everyone else.

The same hotel can be two different purchases

Even honest scores don’t settle it, because a hotel’s worth depends on what you need it to do.

So TripWorthIt™ asks one thing before it adds anything up: what job did the hotel have?

• 🗺️ The Explorer uses the hotel as a base. Location & Access counts most, and the grounds matter less because you’re out all day.
• 🏖️ The Relaxer treats the hotel as the destination. Rest, Property, Service and Offer carry the weight, and Location barely registers once you’ve arrived.

The pillar scores don’t change between the two. Only the weighting does. For a hotel that’s good at everything, the gap is small: Hotel ICON in Hong Kong scores 4.31 for us as a base and 4.24 as a destination. That tiny gap happens to straddle our Superb line, which tells you how much to read into any single line. For a lopsided hotel, the gap can be more than half a point.

Try it below. We’ve loaded an imaginary island resort: wonderful to stay in, awkward to reach. Switch its job and watch the badge. Then drag the sliders and score a hotel of your own.

Does the type of trip change a hotel’s score?

It changes the overall score, never the pillar scores. And it matters more often than you’d think.

My background is catastrophe risk modelling, where you never trust a single number without asking how much it could move. So we did what I’d have done at work. We ran 5,000 simulated versions of our recorded scores, nudging each pillar by roughly the amount Mel and I actually differ, and checked how often our rebook model gave each hotel the same verdict. About six in ten of those verdicts held in at least four runs out of five. The rest are closer calls than any star rating admits. If anything, that’s optimistic, because we measured our disagreement after we’d talked our scores through.

Then we tested the weights. We reweighted the five pillars 5,000 random ways, everything between “only the bed matters” and “only the location matters”, and checked each hotel against our price benchmark:

• 31 hotels came out ahead of their price in at least nine of every ten weightings
• 35 came out behind it in at least nine of every ten
• 34 flipped between ahead and behind, depending on the weighting

That last group is why TripWorthIt™ asks about the hotel’s job before it adds anything up. A star rating can’t tell you whether a remote villa with no bar is a dream retreat or an expensive mistake. How you weight the pillars can.

What 100 stays taught us about booking the next one

This is how we now read crowd scores, and our own:

• Use crowd scores as a veto, not a verdict. A low score usually means something real is wrong. A high one tells you far less.
• Read the Location sub-score. In our data, it was the one crowd number that measured something of its own.
• Check the review count. A 9.5 from 40 reviews is a much less certain number than an 8.9 from 4,000.
• Decide the hotel’s job before you compare. A great base and a great destination are different purchases.
• Ask the rebook question before you book. Would I pay this twice? If the honest answer is “once is enough”, that’s fine. Just know it going in.

Ten of our 100 stays cleared our Superb line of 4.25 out of 5. They’re the ones we still talk about over dinner, and they’re the heart of the film, so we’ll save the full list for that.

What TripWorthIt™ can’t tell you

We’d rather you trusted it for the right reasons.

Our price benchmark comes from our own 100 stays. It isn’t a survey of the hotel market, and our tastes aren’t yours.

We also scored with hindsight. We knew how we felt about each stay when we scored it, while Booking.com’s reviewers were strangers scoring on the day. That flatters our model, and we know it. The fairer test is to score new stays at checkout and decide later whether we’d go back, so that’s what we’re doing from now on.

And no calculator answers the third question. It can show you what a stay delivered and how that compares with its price. Whether you’d pay it again is yours to decide.

Try it on your last hotel

The full TripWorthIt™ scorer puts the sliders first. Pick the hotel’s job, score the five pillars and add what you paid per night. Then, before we show you the answer, guess: did it beat its price?

You’ll get two results: how good the stay was for its job, and how it compares with what stays at that price delivered in our records. The third question, whether you’d pay it again, stays with you.

Would you hand over the same amount again?

Score your last stay with TripWorthIt™ →

Frequently asked questions

Are Booking.com ratings accurate?

They’re accurate averages of what guests reported, and that’s the catch. Of the 54 hotels Booking.com rated Superb or Exceptional in our sample, I’d pay the same again for only 31. We now use crowd scores to rule hotels out, then decide for ourselves which of the survivors is right for the trip.

Why are so many hotels rated “Superb” or “Exceptional”?

Because scores bunch at the top. On Booking.com, 83 of the 95 hotels it rated in our sample were Fabulous or better. Agoda labelled 59 of the 85 it rates as “Exceptional”. Part of that is us: we chose these hotels, often because they scored well. But when most hotels share the top bands, the labels stop separating them.

What is TripWorthIt™?

It’s our free hotel scorer. You rate a stay on five pillars, tell it whether the hotel was a base or the destination, and add the nightly price. It gives you an overall score for that job, and tells you how the stay compares with what stays at that price delivered across our own 100.

What’s the difference between a good hotel and a good-value hotel?

A good hotel delivers well. A good-value hotel delivers more than stays at its price usually do; we measure that against our own 100 stays. Some of our highest-scoring stays were both, and I still wouldn’t pay for them again, because “good value” and “worth repeating” are separate questions.

What does choosing Explorer or Relaxer change?

Only the weighting. Your five pillar scores stay exactly as you entered them. As an Explorer, Location & Access counts most; as a Relaxer, Rest, Property, Service and Offer take over. A well-rounded hotel barely moves. A beautiful but remote one can jump a whole badge.

How many of your 100 hotels scored Superb?

Ten, using our line of 4.25 out of 5. Move the line to 4.0 and it’s 16; raise it to 4.5 and it’s four. Even among those ten, I’d pay the same again for only six. The other four are “only if cheaper”, which is this whole post in one line.