Twice this week I sat down to do something ordinary. I tried to buy one night at the top of the alpine season, at properties I have been tracking for two months, on the dates every underwriting model treats as the strongest of the year. Nothing exotic. One room, one night, booked the way any guest would book it.
It could not be done. Not at a high price, not at an absurd price. The transaction the model assumes is taking place was not on offer.
At St. Moritz, on 11 August, none of the five properties I follow would sell 31 December as a single night. Four returned no bookable availability at all. The fifth would take the date only inside a seven-night stay beginning the evening before. At Courchevel the following day, across seven properties and twenty-one attempts, not one produced a bookable single night. A week earlier the same test on the same panel had produced exactly one. The direction of travel over seven days is not the point. The floor is.
Two details matter before anyone reaches for the obvious explanation. The booking channels are working: at every property in the test, other dates in the same calendar return a rate and a confirmation path, so this is not a website that has stopped functioning or a hotel that sells only by telephone. And these dates sit roughly one hundred and forty days out. This is not the last-minute market closing up. It is August, and the last week of December is already not for sale in the form the model expects.
What the engine offers instead is a stay. Minimum lengths of three to seven nights, sometimes a fixed window with a mandated arrival day, occasionally a package with a price attached to the whole rather than to any night inside it. The building is available. The room-night is not the thing being sold.
That distinction has a consequence that survives the trip from the booking screen to the valuation.
Almost every method of putting a number on a seasonal hotel runs through a rate multiplied by an occupancy. Both terms are expressed per night, because that is the unit hotels have always used. If the peak of the year is not transacted by the night, then the per-night figure covering those dates is not an observation. It is a division. Somebody has taken a seven-night price and cut it into seven, or taken an aggregate and spread it across the days it might plausibly have covered. The arithmetic is fine. The status of the resulting number is different from the status of a rate somebody actually paid for one night, and the two arrive on the page looking identical.
The same applies to occupancy at the peak, which stops behaving like a measure of demand and starts behaving like a count of blocks. A property selling its strongest fortnight in seven-night units is closer, commercially, to a seasonal rental with service attached than to a hotel responding continuously to demand. Its peak revenue depends on a small number of decisions taken months in advance by a small number of parties. That is a different risk shape from the one implied by a smooth occupancy curve, and it does not surface anywhere in an occupancy figure.
I want to be exact about what I do not know, because the limit is real and it changes what can honestly be claimed. From outside, three situations produce an identical screen. The rooms may be genuinely sold. The property may have closed single-night sales deliberately, which is a routine commercial decision at peak. Or the dates may be held off public channels and released through other routes. These mean very different things about demand, and I cannot distinguish between them from where I sit. What I can establish, and what is dated and checkable, is narrower: on those days, at those properties, the night was not purchasable through the channel a buyer would use.
That narrower fact is enough to matter, because the reconstruction happens whether or not anyone knows which of the three applies. A per-night figure for the alpine peak gets produced, it enters a model, and it travels onward with the same confidence as every other line. The person receiving it three steps later has no way of knowing that the unit underneath it was never sold that way.
There is a quieter version of the problem that I find harder to shake. A market where the peak is only available in long blocks has, in effect, chosen its guest. Someone who can commit to seven nights in the last week of December, months ahead, is a specific kind of buyer with specific alternatives. The property has not just priced its peak. It has selected who is allowed to purchase it, and that selection is invisible in any figure that reports what the peak achieved.
On the record — Alpine peak, August 2026: Confidential Markets holds that at the top of the alpine winter season the room-night is not the transacted unit, and that per-night figures published or modelled for those dates are therefore derived rather than observed. Our position is that this is structural to how these markets sell their peak and not an artefact of how early we looked. We expect that as the dates approach, single nights will appear at some properties, and that where they appear it will be at the edges, through late releases in lower categories or gaps left between blocks, rather than as a general return to nightly sale. We are clear about what would falsify us: if by the end of October the same panel shows single nights broadly bookable at the majority of properties across those dates, then what we recorded in August is a booking-window effect and our reading of it as structural is wrong. The panel and the dates are fixed, the readings are weekly, and the answer will exist either way.
None of this requires anyone to distrust the numbers they are given for these markets. It requires one question, asked early enough to be useful, about the peak season figures in any alpine file: was this rate transacted for a night, or assembled from something that was sold as a week. In most of the files I have seen described, nobody asks, and the number does not carry the answer with it.
—
The next Field Note continues the series.
Confidential Markets is independent research on the structural and cyclical risk of Europe's scarce ultra-luxury hospitality micro-markets. The market, before the asset.