Put four published hotel cancellation rates next to each other and they will not agree. They rarely do, and the reason is almost never that the markets differ.
It is that "cancellation rate" is four different measurements sharing one name.
The four denominators
Cancelled ÷ decided. Cancelled over confirmed plus cancelled. Bookings still pending are excluded from the denominator entirely. This is what I publish, and on more than 34,500 decided bookings from one Gulf desk it comes to a quarter.
Cancelled ÷ all rows. Everything in the file, including bookings that never reached a decision. Same data, and it reads close to 24%. Neither figure is wrong. They answer different questions: the first asks how often a decision goes against you, the second asks what share of your total workload evaporates.
Cancelled ÷ confirmed. A ratio dressed as a rate. It always looks better and it cannot be compared to either of the above.
Cancelled room-nights ÷ total room-nights. A completely different unit. A cancelled fourteen-night booking and a cancelled one-night booking are the same event in the first three measures and are not remotely the same in this one.
Before you compare two numbers, find out which of these four each one is. Most of the time that single question dissolves the disagreement.
The three other things that move the number
What counts as a cancellation. Does a no-show count? An amendment that drops one room from a booking of four? A booking cancelled and immediately rebooked at a different rate? Three defensible answers, three different rates.
Whether the sample is a market or a channel. A rate drawn from direct consumer bookings on a refundable rate and a rate drawn from B2B wholesale bookings are measuring different behaviour under different terms. Neither generalises to the other, and both get quoted as "the industry".
Whether destinations were pooled. This is the big one. On my own file the rate ranges from under 14% to close to 40% by destination alone. A single headline rate is therefore a statement about a portfolio mix, not about hotels. Change the destination mix and the headline moves without any behaviour changing at all.
That last point is worth sitting with. Two desks in the same city with the same customers can publish cancellation rates twenty points apart purely because one sells more long-haul.
How to read one properly
Four questions, in order:
- What is the denominator? If it is not stated, the number is not usable.
- What counts as a cancellation? No-shows and amendments in or out.
- What is the sample? Which channel, which market, what size, over what period.
- Is it pooled across destinations? If yes, the number describes a mix, not a risk.
If a published benchmark cannot answer all four, it is not a benchmark. It is a headline.
Why I state my definitions before my findings
Because the alternative is to be uncitable.
A number without a definition cannot be checked, cannot be compared, and cannot be used by anybody whose reputation depends on being right. It circulates for a while and then dies, because the first person to interrogate it finds nothing underneath.
So the methodology page states the denominator, the exclusion rules, the minimum observation counts required to enter a ranking, and the exact row counts before and after cleaning. The FAQ states what the sample is and what it is not. The dataset is versioned, licensed CC BY 4.0, and has a permanent identifier, so a number quoted from it today can still be traced years from now.
None of that makes the findings more impressive. It makes them checkable, which is a different and more durable thing.
If you take one habit from this: when two travel numbers disagree, check the denominator before you check the market. Nine times out of ten, that is the entire disagreement.
Source: the Gulf Outbound Hotel Index, version 1.0 — more than 36,100 B2B wholesale hotel bookings from one Gulf desk. Findings and full tables · Methodology · CC BY 4.0 · 10.5281/zenodo.21796038