A cancelled night was still priced.

In the SiteMinder 2025 set, the US cancellation rate was 18.78 percent. The year before, it was 19.22 percent. The global rate in that set was 19.15 percent. The drop is moderate. The level is not. Close to one booking in five never became a stay.

That does not mean four bookings in five are being miscounted, or that every booking will be cancelled. It means a meaningful share of what is booked is not occupied. Each of those bookings had a night attached, and the rate for that night was already set before the cancellation. Someone decided what it was worth, sold it, and the guest cancelled before arriving. The price existed. The stay did not.

The problem shows up when pace is read. Pace often adds up bookings and treats them as nights that will be slept. Done that way, it is also counting nights that will not be occupied. On a high-demand date, the hotel can look fuller than it will be. The person reading that figure thinks there is less room to move than there is, because part of what is on the books will drop off before arrival.

Cancellations are not, for that reason, a management failure. They are part of how people book. 18.78 percent is a level, not an accusation. What is an error is treating the booking as if it were already a stay. A booking is an intention with a price. A stay is someone who arrives.

The horizon has to show the cancellation separately, not add it into occupancy as if the hotel were full. The night already had a price in both cases, whether it is occupied or not. The question the figure has to answer is a different one: whether anyone is going to sleep in it.

So do not read booking pace as if it were already occupancy. In that set, close to one booking in five never became a stay, and that night already had a price. Separate cancellations from expected arrivals. If the hotel looks full only because the bookings are on the books, the room you have left is larger than that number shows.