Guest satisfaction metrics explained: review scores, NPS and complaint rates

Guest satisfaction metrics are several different measures, not one: review scores on each booking site's own scale, survey averages and top-answer shares, Net Promoter Score, complaint rates and response times. Each comes from a different question, scale and group of guests, so they are read side by side and never averaged into a single satisfaction number.

For example, the monthly report arrives with four numbers that all carry the word satisfaction: 8.9 on one booking site, 4.3 on another, a survey average of 4.1 out of 5, and a Net Promoter Score of 42. Someone at the table asks which one is right. All of them are, and none of them measures the same thing. After reading this article you will be able to name what each number is built from, refuse to average them, and cut your report to the few measures your team can actually act on.

Why does one satisfaction score hide several different measures?#

Because each number comes from a different question, asked of a different group of guests, at a different moment, and placed on a different scale. Four things vary between measures: who is asked (every guest, only guests who chose to write, only guests who opened the email), when they are asked (at the desk, the next morning, weeks later), the question itself (rate your stay, recommend us, rate cleanliness), and the scale (1 to 5, 1 to 10, 0 to 10, a count of complaints).

A score is a summary of answers, not a measurement of the stay. Two hotels with the same score can carry completely different mixes of problems underneath it. Every source is also self-selected in its own way: review writers choose to write, survey respondents choose to answer, and a complaint log only holds guests who spoke to a person at the desk.

Averaging across sources mixes scales and populations and produces a number nobody can act on: no team can fix an average of a 1 to 10 score and a 1 to 5 score. The practical rule is labelling. The measures also move at different speeds: a review score can shift over weeks, a complaint rate over days, a survey average only when the next survey closes.

What does a hotel review score actually measure?#

It measures the opinions of guests who chose to write a review, expressed on that site's own scale and combined by that site's own published method. Booking.com's "How we work" page (opens in a new tab) (updated 31 May 2025) says each review score is between 1 and 10, that a property's score is assigned by customers rather than by Booking.com, and that it is calculated with a weighted system in which more recent reviews have a bigger impact.

The same page lists separate subscores for location, cleanliness, staff, comfort, facilities, value for money and free Wi-Fi, submitted independently of the overall score. A subscore is therefore not a slice of the total, and the subscores do not add up to it.

Three different things sit next to each other on a listing: the 1 to 5 yellow stars are a star rating that the site does not assign, the 1 to 5 yellow squares are a quality rating that the site does assign, and the 1 to 10 number in the blue square is the review score. Reports that call all three "the rating" are unreadable.

Because recent reviews carry more weight, a score can drift as older reviews fade, even in a month with few new reviews. Before explaining a move, check how many new reviews actually arrived. Other sites use their own scales and methods; read each site's help or partner page, write down the scale and the date you read it, and keep that note with the report, because methods change without announcement.

What do survey measures tell you that a review score cannot?#

A survey lets the hotel choose the question, the timing and the guest group, so it can measure one step of the stay and reach guests who never write a review. Three survey figures, defined once: the average score (the sum of answers divided by the number of answers), the top-answer share or top-two share (the percentage choosing the best, or the two best, options), and the response rate (answers divided by invitations sent).

The average hides the shape of the answers. A 4.1 out of 5 can mean everyone was mildly positive, or half the guests were delighted and a fifth were angry. Print the share of the lowest answers next to the average.

The response rate is part of the result, not a footnote. Twelve answers from 400 invitations is a different claim from 120 answers from 400. Fix the moment of asking and keep it fixed: moving the survey from the morning after departure to the moment of checkout changes what guests are reporting about, so the trend becomes a change of timing rather than a change of satisfaction. Keep the question wording fixed too, because a reworded question breaks the series, and the break is usually discovered months later by someone comparing two charts.

A survey is the only one of these measures where the hotel controls the question, which is why it is the right place to ask about a specific step: check-in, the room, breakfast, the bill. If you are building one, start with the questions, the timing and the follow-up of a guest satisfaction survey.

What is Net Promoter Score, and what does it not tell a hotel?#

Net Promoter Score is one number built from one question: how likely a guest is to recommend the hotel, answered on a zero-to-ten scale. Bain & Company's Net Promoter System site (opens in a new tab) defines the groups: guests who answer 9 or 10 are promoters, 7 or 8 are passives, and 0 to 6 are detractors. The score is the percentage of promoters minus the percentage of detractors (Bain & Company, "Measuring Your Net Promoter Score", 2026).

The result runs from minus 100 to plus 100 by construction, so a score of 42 is not "42 percent satisfied". It is a gap between two groups of guests.

For example, with 200 answers:

NPS = 60 minus 15 = 45.

It is one question about intention, not a record of what happened during the stay. It tells you whether guests would recommend the hotel; it does not tell you what to fix, and it does not name a department. It is sensitive to sample size: with 30 answers, one guest moves the score by about three points, so a monthly NPS from a small sample is mostly noise. Report the number of answers beside the score.

It is also sensitive to the exact wording, the language it is asked in and the moment it is asked. Comparing the NPS of two properties that serve different source markets is weak evidence of anything. Passives are the group most reports ignore, and often the group a hotel can actually move: they were satisfied enough to return but not enough to recommend. Do not import a target from another hotel or from a conference slide; there is no public benchmark that matches your mix of guests, market and season, so the only honest comparison is your own history.

Which operational measures belong next to the scores?#

Scores say how guests felt; operational measures say what the hotel did, and they can be counted from the hotel's own records without waiting for a guest to write anything. The complaint rate is complaints logged in the period divided by guests or stays in the period. Write down what counts as a complaint (a logged issue, not every remark at the desk) and keep the definition fixed.

Request closure time is the median hours from a logged request to its closure, not the average, because one forgotten request drags an average far away from the normal case. Review reply time is the hours from a review appearing to the hotel's reply, plus the share of reviews that received a reply at all; the share matters more than the speed, since an unanswered review is a public unanswered review. If your team is working on this, see how to reply to hotel reviews by complaint type.

Repeat-guest share is guests with a previous stay divided by guests in the period. It needs identity matching, and a returning guest who books through a different channel may not be recognised, so treat it as a floor rather than a precise figure.

Each of these has a definition problem before it has a data problem. Write the definition once, in the report header, and change it only with a note on the date. These measures respond within days and point at a team, which is why they belong beside the scores rather than in a separate operational report nobody reads. For the mechanics of tracking them, see logging, assigning and closing each guest request.

Which three numbers should a small hotel watch each month?#

One review score per site you sell on, one survey top-answer share with its response rate, and one complaint rate with its closure time. Three, not twelve, because each number needs an owner and a follow-up action. A page of twelve indicators is read once and then ignored, while three with names beside them get discussed.

Read direction, not level. Compare each measure with the same measure three months earlier and with the same month last year, on the same definition and the same scale. Flag method changes on the chart: a site changes its weighting, the hotel changes its survey wording, the complaint log changes hands. Each of these breaks the series, and an unmarked break reads as a real move.

Do not attach a staff bonus to a review score alone. It rewards chasing scores and quietly discourages staff from logging complaints, which hides exactly the problem the score was meant to reveal. Keep revenue measures such as occupancy, ADR and RevPAR in a separate report: they answer a different question, and mixing them with satisfaction numbers invites the false reading that one causes the other.

Give the three numbers a monthly owner meeting of twenty minutes: what moved, what the operational measures say about why, and one action with a date.

Glossary: what each satisfaction measure is built from, and what it cannot tell#

Each measure below answers one question, and the last column is the reason not to read any of them alone. Definitions marked as the site's own come from that site's published method; definitions marked as the hotel's own are internal conventions and must be written down to stay comparable. Record the date you read a review site's method page, because methods change and an old note is better than a wrong memory.

Keep this glossary in the front of the monthly report so that a new manager reads the same definitions as the person who built the series. Where a measure depends on guests choosing to act, the count of guests who acted belongs in the same line as the result.

What to do next#

Take your next monthly report and write the four labels beside every satisfaction number in it: source, question, scale and denominator. Then cut the page to three measures, give each one an owner, and compare each with its own history rather than with another hotel. The measures in this article are only useful as a set: the scores say how guests felt, the operational numbers say what the hotel did, and the trend says whether either is changing.